me lu ju'i lobypli li'u 18 moi

From Lojban
Jump to navigation Jump to search

For a full list of issues, see zo'ei la'e "lu ju'i lobypli li'u".
Previous issue: me lu ju'i lobypli li'u 17 moi.

Number 18 - May-June 1993
Copyright 1993, The Logical Language Group, Inc.
2904 Beau Lane, Fairfax VA 22031 USA (703)385-0273
Permission granted to copy, without charge to recipient, when for purpose of promotion of Loglan/Lojban.

Logfest 93 - July 9-12

rafsi List Revised and Baselined

ju'i lobypli (JL) is the quarterly journal of The Logical Language Group, Inc., known in these pages as la lojbangirz. la lojbangirz. is a non-profit organization formed for the purpose of completing and spreading the logical human language "Lojban - A Realization of Loglan" (commonly called "Lojban"), and informing the community about logical languages in general.

la lojbangirz. is a non-profit organization under Section 501(c)(3) of the U.S. Internal Revenue Code. Your donations (not contributions to your voluntary balance) are tax-deductible on U.S. and most state income taxes. Donors are notified at the end of each year of their total deductible donations.

For purposes of terminology, "Lojban" refers to a specific version of a logical human language, the generic language and associated research project having been called "Loglan" since its invention by Dr. James Cooke Brown in 1954. Statements referring to "Loglan/Lojban" refer to both the generic language and to Lojban as a specific instance of that language. The Lojban version of Loglan was created as an alternative because Dr. Brown and his organization claims copyright on everything in his version, including each individual word of the vocabulary. The Lojban vocabulary and grammar and all language definition materials, by contrast, are public domain. Anyone may freely use Lojban for any purpose without permission or royalty. la lojbangirz. believes that such free usage is a necessary condition for an engineered language like Loglan/Lojban to become a true human language, and to succeed in the various goals that have been proposed for its use.

Press run for this issue of ju'i lobypli: 130. We now have about 720 people receiving our publications, and 250 more awaiting textbook publication.

Important Notices

Important: Your mailing label indicates the last issue of your subscription. If that issue is JL18, we need to hear from you, preferably with money for another year's subscription (US$28 North America, US$35 elsewhere).

Note the new network address on page 2 for the Planned Languages Server if you wish to obtain electronic copies of our materials. The address published last issue turned out to be incorrect.

Your Mailing Label

Your mailing label reports your current mailing status, and your current voluntary balance including this issue. Please notify us of changes in your activity/interest level. Balances reflect contributions received thru 15 June 1993. Mailing codes (and approximate balance needs) are:

Activity/Interest Level:     Highest Package Received (Price Each)     Other codes:
B - Observer                 0 - Introductory Materials ($5)           JL JL Subscription ($28-$35/yr)
C - Active Supporter         1 - Word Lists and Language Description ($15) (followed by expiration issue #)
D - Lojban Student           2 - Language Design Information ($10)     * indicates subscription prepaid
E - Lojban Practitioner      3 - Draft Teaching Materials ($30)        LK LK Subscription ($5-$6/- yr)
R Review Copy (no charge)                                              UP Automatic Updates (>$20 balance)

Please keep us informed of changes in your mailing address, and US subscribers are asked to provide ZIP+4 codes whenever you know them.

Contents of This Issue

The biggest news this issue is the baselining of the rafsi list, the last major piece of the language to be frozen before dictionary publication. Two articles this issue deal with the Lojban rafsi, and the latest change, and the revised list is included with the issue.

As soon as this issue goes to the printers, I will be starting to work intensively on dictionary publication, with the intent to have something to show off at LogFest, our annual gathering here in July. See the news section for more on the dictionary work, and on LogFest 93. Because JL issues are taking 1-2 months to prepare, I am not going to be able to get JL on the hoped for quarterly schedule and also get the dictionary and textbook published this year. As such, I will not start work on JL19 until September, to enable me to work all summer on getting the dictionary out. I also had to cut off work on this issue rather abruptly, though hopefully without too much loss in quality. Details in the news section.

This issue summarizes all grammar changes proposed for the dictionary rebaselining, and the revised E-BNF form of the grammar. Articles detail the rationale behind several of the changes, with a focus on the most significant change relating to relative clauses. A selection of articles deal with usage issues that have come up on Lojban List, and we have a couple of more philosophical discussions on the goals of the language. As is usual, material derived from the Lojban List computer, as well as from the 'conlang' mailing list, is edited, revised, and corrected from the original.

There are 3 longer Lojban texts in this issue, one related to the ckafybarja project discussed in JL17. The discussions of grammar and usage issues., though have a lot of Lojban text in them, perhaps as much as in the longer pieces. I made an effort to update all lujvo in this issue to the new rafsi baseline, so that you can use the lists accompanying this issue to interpret them. However, since I did this manually, don't be surprised if I missed one or two.

                          Table of Contents
Brief Glossary of Lojban Terms                                    ---3
News - JL Status, Subscriptions, Finances, LogFest 93             ---3
 Other News:  DC Weekly Group, Bradford Group, UK LogFest, Book
 Status                                                           ---4
Language Development Status:  gismu, rafsi, Grammar               ---5
Lojban Proto-Reference Book (Dictionary) - Preliminary Outline    ---6
Sample English-to Lojban dictionary (intermediate step)           ---8
On Lojban rafsi; Revised rafsi Assignments                        ---9
On lujvo, by Greg Higley; "General Purpose lujvo", by Greg Higley;
 Greg Proposes and Explains some lujvo                           ---20
le lojbo ce ciska                                    ---21, 47, 50, 66
Grammar Changes                                                  ---22
A Change to the Relative Clause Grammar: Quantification and noi, by
 Greg Higley; sumti and Relative Clauses, by Colin Fine; Lojbab's
 Solution; Excerpts from the Original Changes 20 and 21, with
 Discussion                                                      ---30
Articles Related to Usage/Grammar/Word Change Proposals: New JOI,
 by Greg Higley; kau, by Greg Higley                             ---43
Empathy in Attitudinals, by John Cowan; Summary of cmavo Changes in
 selma'o UI; Punctuation Proposals, by Nick Nicholas             ---48
More Usage Questions: Dean Gahlon asks a simple question; SVO Order
 in Lojban; And Rosta on "se", "te", & lujvo;  On the Grammar and
 Range of Free Modifiers; Comments on the Tense System, by Greg
 Higley; ko'a stizu; Questions on Logical Connection; Causality in
 Lojban; On le and lo and Existence; A Heated Exchange?          ---53

Language Goals:  Lojban and Metaphydsical Bias; Sapir-Whorfian
 Thoughts; Metacognition-friendly Languages                      ---61
A Lojban-to-Prolog semantic analyzer, by Nick Nicholas           ---67
Enclosures - Lojban Machine Grammar: E-BNF Version, dated 12 June
 1993; 06/01/93 Lojban baseline rafsi list

Computer Net Information

Via Usenet/UUCP/Internet, you can send messages and text files (including things for JL publication) to la lojbangirz./Bob at:

(This supersedes the prior "snark" address.)

You can also join the Lojban List mailing list (currently around 70 subscribers). Send a single line message (automatically processed) containing only:

"subscribe lojban yourfirstname yourlastname" to:

If you have problems needing human intervention, send to:

Send traffic for the mailing list to:

Please keep us informed if your network mailing address changes.

Compuserve subscribers can also participate. Precede any of the above addresses with INTERNET: and use your normal Compuserve mail facility. If you want to participate on Lojban List, you should be prepared to read your mail at least every couple of days; otherwise your mailbox fills up and you are dropped from the mailing-list. FIDOnet subscribers can also participate, although the connection is not especially robust. Write to us for details if you don't know how to access the Internet network.

A good portion of our materials are available on-line from the Planned Languages Server (PLS). See JL16, or send the messages "help" and "send lojban readme" to the server address: 

This is a new address since JL17 was published.

The following explicitly identifies people who are referred to by initials in JL. 'Athelstan' is that person's real name, used in his public life, and is not a pseudonym.

'pc' - Dr. John Parks-Clifford, Professor of Logic and Philosophy at the University of Missouri - St. Louis and Vice-President of la lojbangirz.; he is usually addressed as 'pc' by the community.

'Bob', 'Lojbab' - Bob LeChevalier - President of la lojbangirz., and editor of ju'i lobypli and le lojbo karni.

'Nora' - Nora LeChevalier - Secretary/Treasurer of la lojbangirz., Bob's wife, author of LogFlash.

'JCB', 'Dr. Brown' - Dr. James Cooke Brown, inventor of the language, and founder of the Loglan project.

'The Institute', 'TLI' - The Loglan Institute, Inc., JCB's organization for spreading his version of Loglan, which we call 'Institute Loglan'.

'Loglan' - refers to the generic language or language project, of which 'Lojban' is the most successful version, and 'Institute Loglan' another. 'Loglan/Lojban' is used in discussions about Lojban to make it particularly clear that the statement applies to the generic language as well.

'PLS' - The Planned Languages Server, a no-charge computer-network-accessed distribution center for materials on Lojban (and other artificial languages). See pg. 2 for email address.

Brief Glossary of Lojban Terms

Following are definitions of frequently used Lojban terms. Longer explanations are in the Overview of Lojban. cmavo - Lojban structure words

gismu - Lojban root words; currently 1342;

rafsi - short combining-forms for the gismu;

lujvo - compound words built from rafsi;

le'avla - words borrowed from other languages (there are people who would like to see another term, with a better metaphor, for this concept, but "le'avla" will remain a valid term for the indefinite future; suggestions are welcome);

brivla - Lojban predicate words, consisting of gismu, lujvo, and le'avla; (a few cmavo have the grammar of a brivla);

tanru - Lojban 'binary' metaphors, the most productive and creative expression form of the language, unambiguous in syntax/grammar, but ambiguous in semantics/meaning; tanru generally have a modifying portion (generally on the left) that serves the function of an English adjective or adverb, and a modified portion (on the right).

sumti - the arguments of a logical predicate;

selbri - Lojban predicates which indicate a relation among one or more sumti. A selbri is most often a brivla or tanru; the concept was formerly called "kunbri" in error in some of our early publications;

bridi - Lojban predications, the basic grammatical structure of the language; a bridi expresses a complete relationship: the selbri expresses the relation and the sumti express the various things being related;

selma'o - grammatical categories of Lojban words; the basis of the unambiguous formal grammar of the language. Traditionally and erroneously called "lexeme" in the Loglan community. These categories typically have a name derived from one word in that grammatical category; the name is all capitals, except that an apostrophe is replaced by a small letter 'h' (this is an artifact of the computer language "C" in which the formal Lojban grammar is defined for the YACC processor; C forbids apostrophes in 'tokens' representing single words.


JL Status

I remain short of my goal of publishing every three months, at least partially because getting all of the mailings out the door last issue took more than a month in the first place. But hopefully 4 months is better than the delays we had been having.

I delayed a little in hopes of seeing some more submissions for the ckafybarja (coffeehouse) writing project, especially from those of you who first became aware of the project with the publication of last issue. Nick Nicholas revised one piece that was in progress when JL17 was published. Then, at the last minute, he submitted a character description on behalf of a friend. But otherwise, alas, only silence. As a result, the period for submission of characters and/or setting ideas has been extended indefinitely, until the various people who have contributed feel that enough has been submitted to either vote, or to at least turn fully to the Lojban writing endeavor that is intended.

Unfortunately, this issue of JL has taken even longer to produce, almost 2 months from the day I started. And I had thought that the issue was partially done when I started. Family life, supporting the computer network discussion, and administrative tasks have kept me from working efficiently, and the types of materials we are publishing are taking longer to edit than older issues, because of the need to ensure clarity and accuracy of technical content.

The books have been too long delayed while I tried to get JL on a more frequent schedule. We've improved the JL frequency, though not to the quarterly level I want, or need in order to get 2nd class mailing from the Postal Service. As such, I have decided to cut off work on this JL and go directly to work on the books for the whole summer. I will not be starting JL19 until September (which means publication probably in October, or perhaps even November). Hopefully the dictionary will be done by then, and maybe (but not likely) the textbook. I'm sure that the decision to put book publication higher priority than regular JL publication is one which the community will find acceptable, provided that we maintain some minimum publication frequency; 3 issues this year, while not the desired 4, is considerably better than we did the last two years.

This delay will also serve to give more time for people to submit writings for the ckafybarja project, per the above discussion, before the next decision point. Let's see some more participation this time.

As partial recompense for the delay, this issue is larger than intended. Our prices were set on an assumed average of 60-70 pages per issue, but both of the last two issues have been longer than that. I will wait till next issue to decide, but if issues continue to run long, I may have to increase the subscription price by about $1 per issue ($4 per 4 issues) as of next issue. Orders and renewals until then (up to a maximum 8 issues prepaid) will be at the rate of US$28 for 4 issues (US) and US$35 overseas.

Because the rafsi change baseline took place at a date just before publication, and because the issue was so long already, I've put a minimum of Lojban text in this issue. Next issue will probably have quite a bit more text, since Nora is working on a program that will convert lujvo based on pre-baseline rafsi to the new baseline.


We are now fully on the subscription system, and for the most part, people who have not sent a request for JL are no longer receiving it. We have a slightly smaller subscriber list than last issue, but we know that everyone getting the issue really wants it.

I now have to get publication solidly onto the quarterly schedule, in order to get 2nd class status, which means it probably won't happen this year while book publication takes precedence. Until then, people will get JL a little quicker, via first class mail, and of course we are going to still be losing some money as a result.

As of the publication date, we have around 120 JL subscribers. For about 25 of these, JL18 is listed as their last issue, but I expect at least half of these to renew their subscription based on the experience of the last 4 months. Thus, the number of (all paid) subscribers will drop to around 110 for JL19, and seems likely to stabilize at around that level until books are published (when it hopefully will increase). US recipients will continue to get their issue by first class mail.


We continue to expend money faster than we are taking it in, but the rate of hemorrhage has slowed (at least until this issue goes out). We already had a deficit for 1993 of a couple of thousand dollars by April, which has been remedied by the delay in publication, and significant donations from Jeff Prothero (totalling $1500 so far this year, or almost 1/2 of our income). We will still need a fund raising drive in order to make it through the year. I intend to ask for donations in the letter that announces publication of the first book. Substantial donations and/or massive orders will also be necessary to keep the price for the Lojban books reasonable, since small print-runs alone will add several dollars to the price of each book, and we cannot afford a larger print run. (Expected publication costs will run around $10,000. Donors welcome!)

LogFest 93

The dates for LogFest 93, and the annual meeting of la lojbangirz. has been set. The gathering will take place at Lojbab's house in Fairfax VA (per the la lojbangirz. address and phone number) the weekend of 9-12 July 1993 (we traditionally open up on Friday, but schedule few organized activities for that day; people can feel free to arrive on Saturday the 10th, to come for only one day, etc.). As in previous years, families are welcome, although we are requesting that attendees bring sleeping bags, etc. if possible. One or more tents will be set up in the yard as applicable to ensure plenty of sleeping space.

The invitation to families is a bit more meaningful this year, since we now have two kids. Child care duties will presumably be shared among the relevant adults to maximize people's abilities to participate in activities.

Interest in participating in LogFest seems a bit higher than in previous years, perhaps because more people believe that they can do something with the language, and that books to help learn and use the language will shortly be coming out. Preliminary positive responses from around 20 people suggest that we will set a new turnout record this year.

There is no required admission fee for LogFest. Our costs for putting on LogFest have averaged $20-$30 per attendee in previous years, and we ask attendees to donate at least enough to cover their share if possible. But we don't want money to stand in the way of your attending if you are interested in coming.

As is typical for LogFests, we expect that this year will consist of mostly English-language activities, with an emphasis on Lojban-teaching and learning activities for those new or less experienced in the language. There will probably be significant discussion of the ckafybarja Project, which was significantly developed at last year's gathering.

Several Lojbanists have expressed serious interest in having a major emphasis on Lojban conversation at this gathering, and we believe that there are enough people skilled enough in the language that we can do this, while providing mentoring/tutoring to those who are unable to understand what is being said without help.

We are also trying to arrange international Lojban conversation during LogFest, most likely by live 'interaction' on the computer networks with Colin Fine and other British Lojbanists, and Nick Nicholas in Australia, using the "IRC" function (see 'Other News' below). Those not able to attend LogFest, but who have Internet access may want to contact us at prior to LogFest, and we will try to set some definite times, so that you can also participate in these sessions.

While the books will not be published before LogFest, I will be making a major effort to have copies of some or all of the books-in-progress available for people to look at, and possibly to use during Lojban sessions.

The annual meeting will take place on 11 July 1993 at 10:30 AM. At this point, there is much less on the agenda than in previous years, and we are hoping that this means that the meeting will be shorter than usual. (People planning to attend who would like to see a policy topic discussed at the meeting are welcome to suggest agenda items.)

Other News

DC Weekly Group - The DC weekly group, consisting of 4 Lojbanists (with a 5th planning to start regular particpation this month), continues to meet, and do a little conversation each week in Lojban. We seem to have plateaued in skill level, since only a couple of us are spending much time on Lojban on other days of the week, and my activities are not the type that enhance my Lojban skills.

Bradford Group - Colin Fine's group in Bradford, UK, continues to grow and to meet regularly, and from postings on the net, is probably achieving a sophistication in Lojban use at least comparable to us in DC. There are 3 participants at this writing.

UK LogFest - Colin Fine and Iain Alexander have been actively recruiting Lojbanists in the United Kingdom, and the numbers are growing significantly, now approximately 40. In addition, a higher percentage of British Lojbanists are active students of the language, whereas many American Lojbanists seem to be holding back on learning the language.

As a result of the increased numbers, Colin and Iain proposed that a LogFest gathering be held in the UK this year, and this idea met with ready agreement from other Lojbanists. At publication, it appears that the UK LogFest will be held in September, probably at Colin's house in Bradford. Lojbanists throughout the UK, and indeed all of Europe, are encouraged to attend. Independent of JL publication, when a date for this LogFest is firmly set, we will try to send notice to all European Lojbanists of the details for this gathering.

Colin is also planning a gathering the weekend of the American LogFest, as a 'dry run' for the bigger event, and Lojbanists are welcome to visit that weekend as well.

For further details, please contact Colin Fine at (44) 274 733680 (home) or 274 733466 x3915 (work), or by mail at 33 Pemberton Drive, Bradford, West Yorkshire BD7 1RA, UK

CIX - A possible bolster to Colin's efforts to build a UK Lojban group was the formation within the last couple of months of a Lojban discussion group on the UK computer network 'CIX'. This group has grown rapidly, and is reported to have some 25 participants. Lojban List traffic is echoed to this group, and Colin plans to obtain CIX access later this year to assist those interested in studying Lojban in furthering their progress.

IRC - Colin Fine, Nick Nicholas, and Mark Shoulson started a pattern of using the computer network system called "Internet Relay Chat" or IRC, in order to enable 'live' Lojban conversation between Lojbanists otherwise isolated. A group of Lojbanists is thus now meeting irregularly on the computer networks to converse in Lojban, recently including David Young and Sylvia Rutiser from the DC Lojban group. If you are on the Internet with access to the IRC function, and want to participate, contact us by e-mail per page 2.

As described above, we are hoping to use the IRC facility in conjunction with LogFest, to bring more people into the activities here.

Legal - The trademark on 'Loglan' has now been officially cancelled, in accordance with the court order following our legal victory on this issue. TLI did not include the trademark claim in the first publication after the cancellation.

We have now paid off the legal debt, with money contributed by Lojbab and Jeff Prothero.

The Loglan Institute - There is little to report about the Loglan Institute these days; not much seems to be going on. The organization continues to exist, and may be gaining supporters, although at considerable expense. TLI had an advertisement in the April 1993 Scientific American, although they reported in Lognet that they spent an amount for the ad that would take an enormous response in order to break even. TLI has apparently set up a computer network mailing list, but people who have subscribed to it report no activity.

TLI may be nearing completion of their own dictionary revision, which will be issued in electronic form (a price of $50 has been mentioned). They are also reporting work on a substantial revision on the rules of their language version, in order to make it, like Lojban, truly 'self-segregating' at the word level (i.e., unambiguity demands that you always be able to break a stream of Loglan/Lojban sounds down into individual words uniquely; the TLI language version has been seriously defective in this area).

This will be the last issue containing a regular report on TLI; we will, of course, continue to report any real news about the organization that I receive either through official or unofficial channels. But with the end of the legal battle, there seems to be little interest among the Lojban community in hearing about TLI, so long as they seem to be avoiding resolution of our differences.

Book Status

Work continues on the books, but we cannot report any completion dates yet. Highest priority remains the dictionary/reference, and that occupies most of Lojbab's time in between JL issues, along with the administrative tasks involved in keeping the organization running (including responding to orders and questions from the community by mail). Unfortunately, these latter tasks continue to take too much time, with the inevitable continued delays. There is some significant progress though. In this issue, however, are two reports on the dictionary/reference: an outline, and a sample discussing our approach to doing the English-order portion of the dictionary.

As the outline shows, the contents of the reference book have swollen to the point that we are strongly considering issuing the reference as two books - one more of a reference per se, while the other is a pure dictionary of English-Lojban and Lojban-English, emphasizing content words. A major reason for this has been Nick Nicholas's excellent and extensive work on lujvo, which promises to give us several thousand entries in each direction in the dictionary if it is completed. Nick is also writing a paper describing his treatment of place structures in lujvo-making, which will also be included in the reference book.

John Cowan has completed a revision of the entire content of the draft textbook lessons, reorganizing the materials and updating them to the current language. The results will be merged with the new work that Lojbab has done towards a textbook, and will then result in the draft textbook.

John also has continued writing his survey papers covering the entirety of the language from the standpoint of the grammar, which will be assembled into the Lojban Reference Grammar. This still will be the last of the scheduled books to be completed, since John has several papers left to write, and all of the papers must yet be reviewed by several people before they are finalized.

Language Development Status


Last issue we noted adding of 4 new gismu to support the new international metric prefixes, but did not list the words. They are (with the international prefix in parentheses):

gocti 10-24 (yocto-)
gotro 1024 (yotta-)
zepti 10-21 (zepto-)
zetro 1021 (zetta-)

The major work on the gismu list continues to be the resolution of a few open issues on place structures. These issues will be decided as we prepare the dictionary reference. As soon as these issues are decided, the gismu list will be split into two forms, the current form that is intended for use with LogFlash, and a version oriented towards dictionary formatting. Once we have two lists, keeping them matching with each other will be a substantial requirement. In case of conflict, the dictionary format listing will be presumed to have precedence.


We are baselining the rafsi list, as changed and published in this issue, effective June 1, 1993. We had intended to have the baseline effective with the book publication, but the books aren't out, and the pending change has had a noticeable effect on people's willingness to make and use lujvo, as well as to write in Lojban in general. Since we expect no changes in the few months before the book comes out, it seems logical to make the change effective now. We are issuing a new list of rafsi as an attachment to this issue, in all of the various orders typically used by Lojbanists, and including the lujvo-making algorithm now excluding le'avla lujvo, which are handled by inserting "zei" between components, with no rafsi used. The place structures are not included in the rafsi list (a full gismu list in both Lojban and keyword order, would be larger than this issue).

Included in this issue is a discussion of why the Lojban rafsi system works the way it does, and a report indicating why the changes were made and how we went about making the changes. Greg Higley also discusses his ideas on lujvo-making, and gives some samples of the words he has invented. (Other Lojbanists are invited to submit lujvo that you have coined, along with commentary/explanations of how you came to choose those words).

Nora is integrating ad hoc software programs into a software capability to correct and revise older texts written with the earlier rafsi list. The current procedure is sufficiently complicated, and the baseline so close to publication, that I had to conevrt all lujvo manually this time. Luckily, this issue has less text than last issue.


This issue contains a complete summary of the changes to the Lojban grammar that are pending, and an attachment includes the revised E-BNF notation form of the Lojban grammar incorporating those changes. The grammar is effectively being rebaselined with this publication, as we are using a parser incorporating the changes to evaluate Lojban text, and do not otherwise intend to continue using the previous grammar baseline in any way. On the other hand, there is still the possibility of minor corrections before the official rebaselining in conjunction with book publication. If you have any disagreements with any of the proposed changes, we need to hear from you as soon as possible, but we will consider any comments.

The previous version of the E-BNF had typographical errors, making it difficult for some to use. Enough Lojbanists are actively using the E-BNF as a tool of studying the language that we felt that this should not wait any longer for published revision. Special thanks to John Cowan for devising and maintaining the E-BNF.

We are not yet publishing a new version of the formal grammar definition (the 'YACC' grammar), which will appear in the published reference book. Note that the E-BNF, while computer-ish in style, is not the formal definition that has been verified as unambiguous. It was prepared manually from the formal definition, and has been checked many times, but the YACC grammar takes precedence in case of disagreement between the two versions.

The summary of proposed changes, which may be written rather technically for some readers, shows that there continue to be minor changes proposed in the Lojban grammar, nearly all of which are extensions to the expressive power of the language. As John Cowan continues writing the papers that will eventually comprise the Lojban 'reference grammar', minor problems may be discovered that require further changes. We are hoping that all of these will be found before the first book is published, when the official rebaselining will take effect.

On the other hand, these changes are so minor that almost none of them affect any text written thus far. Some changes enable new usages where it was found that existing forms were leading to unacceptable semantic situations (see the discussions below of relative clauses - change 20, and JOI - changes 30 and 31 for examples of such changes). As a result of these changes, the changed semantics of some of the older forms may render some older texts as inaccurate, even while still being grammatical.

This issue also contains edited discussions that led to some of the more significant proposals being adopted. These proposals often started as discussions of Lojban stylistics, and understanding these discussions will help you gain a better understanding of how you must think about what you are trying to say in order to properly phrase the Lojban. Note that many of the participants in these discussions are not especially advanced, or skilled, Lojbanists. It is worthwhile to plow through the occasional jargon-ridden passages (there is a limit to how much this editor feels he can change what people write, even for the sake of clarity) to follow the thought processes of these new and more advanced Lojban students. You'll learn a lot about the language and how it works, and maybe a little bit about how people at different levels of skill approach problems of expression in the language.

Lojban Proto-Reference Book

Preliminary Outline with estimated page counts by section

The following is the outline for the proto-reference book which Lojbab is using as of publication time. It includes a description of each section contemplated for inclusion, and an estimated page count. Major tables, forming the bulk of the book, are the most unpredictable portions in length; these are marked with asterisks (*). The estimated page counts in the following are in most cases just that - estimates (a bar indicates a page count for several related sections). The text is not in general written in any final form, although almost all of the materials exist in some preliminary form that mostly requires editing, rather than new writing.

Due to space and publication cost, some of the materials listed in the outline may be left out. For example, many people would not be that interested in the gismu list etymologies, especially since they are in a rather preliminary form that may make them less easy to use than they eventually will be. On the other hand, the features documented in the outline are those that define Lojban officially, and all may be helpful to both language learners and to people looking over our shoulder to examine the quality of the Lojban design.

A study of the outline shows that, with the exception of the dictionary proper, no section of the book is particularly long, such that omitting it would substantially reduce the size of the books. The only real tradeoff that might make a major difference would be to avoid the practice of listing most data twice - once in the full dictionary, and once in a list specific to the type of information being presented.

However, the nature of the language is such that people will want and need those separate lists fully as much as any combined dictionary list. When you are making new words, you need a handy list of the gismu and their rafsi, and other data, especially existing lujvo, would be a distraction. Similarly, people tend to use lists of cmavo in selma'o order as often, if not more often, than they use alphabetical lists.

The reference will include three attempts that have been made to devise a thesaurus-style semantic index for Lojban. None of the efforts really can be considered authoritative, and indeed, Lojbab believes that there is a significant problem with the standard thesaurus technique, which tends to be more noun/adjective-oriented than verboriented. In dealing with a predicate language, which is probably more like a verborientation - most of the words have been categorized on the basis of the meaning of their x1 place, which is often not the only place that is important to classify.

However, semantic indexing of the gismu list seems to be something that most people have some use for, given the number of people who have reported doing something of that type on their own. Since we cannot produce a definitive and verified thesaurus solution, it seems better to present all three efforts, and let the user of the book decide which best suits his purpose and his understanding of the Lojban vocabulary system. Of course, this takes more pages, but we cannot honestly say, without a lot more research than we are likely to have time for in the next year, which effort is most accurate and/or useful, and what entries in each list are correct. Take all groupings therefore, with a large grain of salt, recognizing that at least one person, the compiler of the particular list, saw a semantic similarity between the various gismu that are grouped together.

Comments on the outline, are of course welcomed.

 Pages Section Description
  4    Table of Contents

  4     About Lojban
  3     About this book

     Lojban Orthography
  1     Letters and symbols
  3  |  optional conventions
  |     Cyrillic Lojban
  |     Dates
  1  |  compounds
  |    text layout

     Lojban Phonology
  2     consonants
  1     permissible initials
  1     permissible medials
  2     vowels, diphthongs, divowels
  2  |  syllables
  |    hyphen
  |    buffering
  1     stress
  1  |  rhythm, phrasing
  |    intonation

     Lojban Morphology
  1     Summary of types and how to tell them apart
  1  |   cmene (names)
  |     cmavo
  |      V
  |      VV
  |      CV
  |      CVV
  1  |   brivla
  |      gismu
  1  |     lujvo
  |        rafsi
  4         lujvo-making algorithm /tosmabru
  2         scoring/choice of form
  1  |     le'avla
  |      le'avla lujvo
  3     Resolver algorithm

  2      About the E-BNF
  3      *E-BNF
  1      *selma'o/E-BNF terminal index
       YACC Grammar
  8      About the YACC Grammar
  1      Parser algorithm
  20     *YACC Grammar
  8      *selma'o/YACC grammar
     terminal index
  1      *selma'o list
  20     *short alphabetical definition,
           subcategories with cmavo in each subcategory
  20    *YACC terminal list, definition, examples of each type?

       The formation of gismu
  3      Lojbanizing rules used
  45    *composite gismu etymologies (may be omitted for space)
  1      *cultural gismu
  1  |   *metric gismu
  |     *internal gismu
       Place structures of gismu
  30    *Lojban gismu (rafsi, definition) Lojban order
  35    *gismu keywords; keywords/phrases for each place by gismu
  35    *Lojban and English order (no place structures)
  10    * cmavo in Lojban order
  10    * cmavo in selma'o/subtype/alphabetical order
  2      * cmavo compounds typically written as one word
  8      * non-Lojban alphabet and symbol set conventions
  1      * unassigned cmavo
  2      * experimental cmavo
  1      Categories within pro-sumti (KOhA)
  3      Categories within UI
  2  |  Use of BAI to add places/cases
  |     *list of BAIs typically used to add cases
  |     *list of BAIs typically used as sumti modifiers
  1      Assignment of rafsi
  8      *rafsi, by type,
  8      *rafsi, pure alphabetical
  20   How to determine place structures of lujvo
       lujvo lists
  45    *lujvo actually in use - estimated ~1800
  45    *proposed lujvo (possibly intermingled with preceding) systematically created (using "se", "te", "ve", "xe", "nu", "ka", "ni", "ri'a", "gau", etc.  estimated ~3000
  22    *pre Eaton/TLI lists (heavily weeded and edited) - estimated ~1500
  15    *collected old proposals ~1000
  1     Lojbanizing of names
  4      *some personal names
  4      *some country/language names
  3      types of le'avla
  1      the culture word issue
  3      *cultural le'avla
  3      *some food items
  3      *some plants/animals
  3      *element words
  198  *Lojban order dictionary ???  (composed of all preceding lists) [gismu (25), cmavo (20), rafsi (8), cmene (names) (6), le'avla (12), lujvo(127)]
  310  *English-order dictionary [page counts dependent on Lojban order counts:  gismu (est. pg. x 5), cmavo (x 2), names(x 1), le'avla(x 1), lujvo(x 1)]

       systems of categorization
  4      *Roget's/Athelstan/Lojbab
  4      *Carter
  4      *Cowan
  40   *gismu to category for each type
  30   *category to gismu for each type
  10   *English-order cross-index of categories

  30 Appendix - *Glossary of Lojban/Linguistic Terminology
     Appendix - Correspondences with historical TLI Loglan
  2     Alternate Orthography for Lojban
       Lojban gismu correspondence to historical TLI Loglan gismu and lujvo
  12    *Lojban gismu order
  8      *historical Loglan gismu order
       Lojban selma'o
        correspondence to historical TLI Loglan selma'o
  3      *Lojban selma'o order
  3      *historical Loglan selma'o order
       Lojban cmavo correspondence to historical TLI Loglan cmavo
  10    *Lojban cmavo order
  6      *historical Loglan cmavo order

  8  Index
  502 pages reference +
  508 pages dictionary +
   92 pages thesaurus +
   82 pages appendices =

Sample English-to-Lojban dictionary (intermediate step)

The following is a sample of the output from a KWIC (Key Word In Context) tool that John Cowan wrote specifically to help automate creating the English-to-Lojban dictionary. This is a trial effort, which will almost certainly play a part in the creation of the English portion of the dictionary. There may be some differences in style or format. Comments are welcome as to how usable you find this style of presentation of the vocabulary.

This format is that used by the Oxford Dictionary of English Proverbs, which has the problem of deciding how to alphabetize a list of proverbs. Just using the first word (or even the first content word) is not enough; what if you remember only the word "devil" from "Needs must when the devil drives"? Each proverb is listed, therefore, under all its content words. The word is rotated to the front, followed by a comma; the place from which it was removed is marked by a "|" character (omitted at the beginning or end).

John took a similar approach here. The entire place structure definition is processed, and the corresponding gismu is attached to the end, set off by a "¯" sign. The rafsi, if any, are appended in parentheses. This version of the program omits all words appearing more than 20 times in the input; there is no point in listing words under "x4" or "event" or "the". An exception is made when the word is also the LogFlash keyword: thus "zvati" appears under "at", but no other word does because "at" is too frequent. Two different fonts and three sizes are shown. We will probably use trhe smallest that we think can be clearly read in reproduction. Comments welcome, especially from those with vision problems.

abdomen: x1 is a / the | / belly / lower trunk of x2; ¯betfu (bef be'u)

able: x1 is | to do / be / capable of doing / being x2 under conditions x3; ¯kakne (kak ka'e)

above: x1 is directly | / upwards-from x2 in gravity / frame of reference x4; ¯gapru (gar)

abrupt: x1 is sudden / | / sharply changes at stage / point x2 in process / property / function x3; ¯suksa (suk)

absolute: x1 is a fact / reality / truth, in the | ; ¯fatci (fac)

absorbs: x1 soaks up / | / sucks up x2 from x3 into x4; ¯cokcu (cok cko co'u)

abstracted: x1 is | / generalized / idealized from x2 by rules x3; ¯sucta (suc)

academy: x1 is a school / institute / | at x2 teaching subject x3 to audience / community x4 operated by x5; ¯ckule (cu'e)

accessing: x1 is a street / avenue / lane / drive / cul-de-sac / way / alley / at x2 | x3; ¯klaji (laj)

accident: x1 is an | / unintentional on the part of x2; x1 is an accident; ¯snuti (nut nu'i)

accommodates: x1 contains / holds / encloses / | / includes contents x2 within; x1 is a vessel containing x2; ¯vasru (vas vau)

accompanies: x1 is with / | / is a companion of x2, in state / condition / enterprise x3; ¯kansa (kas)

accompaniment: x1 dances to | x2; ¯dansu

accomplishes: x1 succeeds in / achieves / completes / | x2; ¯snada

according: x1 is a dimension of space / object x2 | to rules / model x3; ¯cimde

according: x1 is a family / clan / tribe with members x2 bonded / tied / joined | to standard x3; ¯lanzu (laz)

according: x1 is a history of x2 | to x3 / from point-of-view x3; ¯citri (cir)

according: x1 is an heir to / is to inherit x2 from x3 | to rule x4; ¯cerda (ced)

according: x1 is polite / courteous in matter x2 | to standard / custom x3; ¯clite (lit)

according: x1 is to the east / eastern side of x2 | to frame of reference x3; ¯stuna

according: x1 is to the north / northern side of x2 | to frame-of-reference x3; ¯berti (ber)

according: x1 is to the south / southern side of x2 | to frame of reference x3; ¯snanu

according: x1 is to the west / western side of x2 | to frame of reference x3; ¯stici

according: x1 is / reflects a pattern of forms / events x2 arranged | to structure x3; ¯morna (mor mo'a)

account: x1 is an | / bill / invoice for goods / services x2, billed to x3, billed by x4; ¯janta (jat ja'a)

accountable: x1 is responsible / | for x2 to judge / authority x3; ¯fuzme (fuz fu'e)

accruing: x1 is a profit / gain / benefit / advantage to x2 | / resulting from activity / process x3; ¯prali (pal)

accuracy: x1 measures / evaluates x2 as x3 units on scale x4, with | x5; ¯merli (mel mei)

achieve: x1 helps / assists / aids object / person x2 do / | / maintain event / activity x3; ¯sidju (sid dju)

achieves: x1 succeeds in / | / completes / accomplishes x2; ¯snada

acid: x1 is a quantity of / contains / is made of | of composition x2; x1 is acidic; ¯slami

acidic: x1 is a quantity of / contains / is made of acid of composition x2; x1 is; ¯slami

acids: x1 is a quantity of protein / albumin of type x2 composed of amino; ¯lanbi

acquires: x1 gets / | / obtains x2 from source x3; ¯cpacu (cpa)

acrid: x1 is bitter / | / sharply disagreeable to x2; ¯kurki

across: x1 is a bridge over / | x2 between x3 and x4; ¯cripu (rip)

across: x1 is located | x2 from x3; x1 is opposite x3; ¯ragve (rav)

across: x1 ranges / extends / spans / reaches | / over interval / gap / area x2; ¯kuspe (kup ku'e)

act: x1 is an event / state / | of violence; ¯vlile (vil)

actions: x1 is kind to x2 in | / behavior x3; ¯xendo (xed xe'o)

actions: x1 tries / attempts to do / attain x2 by | / method x3; ¯troci (roc ro'i)

An alternative being considered,and shown as a second example, isto repeat the English words intheir context, marked by format tomake them easy to spot. Creatingsuch an alternative format issignificantly more cumbersome, andobviously takes a bit more spacesince the words are spelled out,but many would find it easier toread. In a dictionary, even smallpercentage changes in definitionlength can make a difference ofseveral pages in the result.

Since the Lojban dictionary is going to be expensive to produce, brevity could make a difference it what we have to charge for the result.

If you have a strong preference in this utility vs. cost tradeoff, make it known to us as soon as possible.

abdomen: x1 is a / the abdomen / belly / lower trunk of x2; ¯betfu (bef be'u)

able: x1 is able to do / be / capable of doing / being x2 under conditions x3; ¯kakne (kak ka'e)

above: x1 is directly above / upwards-from x2 in gravity / frame of reference x4; ¯gapru (gar)

abrupt: x1 is sudden / abrupt / sharply changes at stage / point x2 in process / property / function x3; ¯suksa (suk)

absolute: x1 is a fact / reality / truth, in the absolute; ¯fatci (fac)

absorbs: x1 soaks up / absorbs / sucks up x2 from x3 into x4; ¯cokcu (cok cko co'u)

abstracted: x1 is abstracted / generalized / idealized from x2 by rules x3; ¯sucta (suc)

academy: x1 is a school / institute / academy at x2 teaching subject x3 to audience / community x4 operated by x5; ¯ckule (cu'e)

accessing: x1 is a street / avenue / lane / drive / cul-de-sac / way / alley / at x2 accessing x3; ¯klaji (laj)

accident: x1 is an accident / unintentional on the part of x2; x1 is an accident; ¯snuti (nut nu'i)

accommodates: x1 contains / holds / encloses / accommodates / includes contents x2 within; x1 is a vessel containing x2; ¯vasru (vas vau)

accompanies: x1 is with / accompanies / is a companion of x2, in state / condition / enterprise x3; ¯kansa (kas) accompaniment: x1 dances to accompaniment x2; ¯dansu

accomplishes: x1 succeeds in / achieves / completes / accomplishes x2; ¯snada

according: x1 is a dimension of space / object x2 according to rules / model x3; ¯cimde

according: x1 is a family / clan / tribe with members x2 bonded / tied / joined according to standard x3; ¯lanzu (laz)

according: x1 is a history of x2 according to x3 / from point-of-view x3; ¯citri (cir)

according: x1 is an heir to / is to inherit x2 from x3 according to rule x4; ¯cerda (ced)

according: x1 is polite / courteous in matter x2 according to standard / custom x3; ¯clite (lit)

according: x1 is to the east / eastern side of x2 according to frame of reference x3; ¯stuna

according: x1 is to the north / northern side of x2 according to frame-of-reference x3; ¯berti (ber)

according: x1 is to the south / southern side of x2 according to frame of reference x3; ¯snanu

according: x1 is to the west / western side of x2 according to frame of reference x3; ¯stici

according: x1 is / reflects a pattern of forms / events x2 arranged according to structure x3; ¯morna (mor mo'a)

account: x1 is an account / bill / invoice for goods / services x2, billed to x3, billed by x4; ¯janta (jat ja'a)

accountable: x1 is responsible / accountable for x2 to judge / authority x3; ¯fuzme (fuz fu'e)

accruing: x1 is a profit / gain / benefit / advantage to x2 accruing / resulting from activity / process x3; ¯prali (pal)

accuracy: x1 measures / evaluates x2 as x3 units on scale x4, with accuracy x5; ¯merli (mel mei)

achieve: x1 helps / assists / aids object / person x2 do / achieve / maintain event / activity x3; ¯sidju (sid dju)

achieves: x1 succeeds in / achieves / completes / accomplishes x2; ¯snada

acid: x1 is a quantity of / contains / is made of acid ofcomposition x2; x1 is acidic; ¯slami

acidic: x1 is a quantity of / contains / is made of acid of composition x2; x1 is; ¯slami

acids: x1 is a quantity of protein / albumin of type x2 composed of amino; ¯lanbi

acquires: x1 gets / acidic / obtains x2 from source x3; ¯cpacu (cpa)

acrid: x1 is bitter / acrid / sharply disagreeable to x2; ¯kurki

across: x1 is a bridge over / across x2 between x3 and x4; ¯cripu (rip)

across: x1 is located across x2 from x3; x1 is opposite x3; ¯ragve (rav)

across: x1 ranges / extends / spans / reaches across / over interval / gap / area x2; ¯kuspe (kup ku'e)

act: x1 is an event / state / act of violence; ¯vlile (vil)

actions: x1 is kind to x2 in actions / behavior x3; ¯xendo (xed xe'o)

actions: x1 tries / attempts to do / attain x2 by actions / method x3; ¯troci (roc ro'i)

On Lojban rafsi

by Lojbab

Occasionally people new to the project have criticized Lojban's rafsi system, generally claiming that the system is overly complex or hard to learn. I contend otherwise, based on personal experience and on observation of those who have already learned the system. What may appear extremely complex and rule-bound, in practice turns out to be quite easy. The system also has the advantage that you need not learn everything at once - you can use the system while knowing only a fraction of the rules and the rafsi.

As a sample of the criticism, here is Rick Harrison, commenting on the "conlang" computer mailing list:

The vast majority of constructed language enthusiasts agree that a planned language should have no allomorphs, i.e. each root-word should have only one form which should not change due to conjugation, declension, compounding, or other grammatical processes. Allomorphs increase the difficulty of memorizing a vocabulary and give no benefit in return. It appears that Loglan and Lojban suffer from rampant allomorphy. Any given 5-letter predicate might have 0, 1, 2, or 3 triliteral allomorphs to be used in compound words. Unless I am mistaken, there's no way to predict whether a given predicate has allomorphs, and if so, what those allomorphs might be; each predicate's allomorphs must be memorized.

Lojban rafsi are the word-forms used to make compound words, and are the 'allomorphs' that Rick is talking about. I, of course disagree with Rick's statements and his conclusions. In particular, I believe that:

  • 'allomorphy', like many other aspects of the design of a constructed language is a design feature that may be used as a trade-off to prevent other problems or to provide other advantages. In the discussion that follows, I will present our rational, showing that Lojban's system does both;
  • the need to clearly distinguish between a multi-word metaphor and a single word compound derived from that metaphor means that some sort of allomorphy is necessary. The only other alternative is to add an extraneous particle as glue between the components of one of these two types of concept combination (which we do in the case of le'avla lujvo, but only because there is no other general solution for an arbitrary word-form that maintains unambiguity). In general such particle addition violates Zipf's Law when the compound is to be used frequently. Zipf's Law predicts that words which are frequently used will be shorter than less frequent ones. I have considerably more faith in this principle as a basis for constructed language design than I do in the purported difficulties arising from allomorphy, especially with a system like Lojban's that is carefully designed.
    (One oft-recurring suggestion for change, generally by critics of the language such as Rick, has been to let the short forms serve as the roots themselves. Not only are there far too few such possible roots, but such a usage would detract from the words available for use as cmavo, the normal interpretation of a CVV form that is a separate word. In addition, short rafsi are far more densely-packed among the set of possible forms than the gismu - nearly all such short forms are used. This results in a significant loss of redundancy that would make the language harder to resolve with such condensed forms. Indeed, Lojban allows the long-form for any compound built of 5-letter rafsi, to alternate for any compound built with the shorter rafsi forms to be used equivalently with identical meaning, to reduce noisy environment redundancy problems. Finally, of course, if the short forms were the roots, there would be no capability for further shortening in conformance with Zipf's Law, and indeed either compounds or non-compound metaphors would have to be longer than the separate words that compose them.)
  • all words in a language have to memorized eventually, if you are to achieve fluency. 'allomorphy', at least as used in Lojban, makes learning that vocabulary easier in general, and there are significant benefits in addition to vocabulary learning, in that you can create new words on an ad hoc basis, even when you are still a language novice, and you can usefully analyze words you don't know. The added memorization implied by the rafsi, even if you memorize every single one of them (which no one has), is but a very small percentage of the total vocabulary needed for fluent adult conversation, but provides immediate benefit for even small amounts of learning.
    The first time you see a compound, you will probably take it apart. Perhaps even the first few times. But you cannot become even moderately fluent in any language if you need to analyze the etymology of every word you want to read, speak, or understand. Words that occur at all frequently must be internalized as a unit of meaning. If there are 50,000 concepts that are needed for adult conversation (a reasonable guess), then you will need to memorize 50,000 words, at one word per concept. This number cannot be reduced, except by polysemy (one word representing multiple concepts), and I cannot see Rick or anyone arguing that polysemy makes learning a language easier.
  • there is indeed a way to predict whether a Lojban root has rafsi, and there are constraints that greatly limit what those rafsi might be. In addition, because the assignment of rafsi is maximized, almost every possible rafsi has some meaning. This has the result that every rafsi that you learn to associate with its gismu reduces the possibilities for other words. This makes learning the others easier, and by the time you've learned even 1/2 the rafsi (or maybe less if they're the right ones), you can generally guess the rest as you need them.

Let me discuss the rationale, first. Lojban lujvo, or compound words, represent the myriad of predicate relations that are not reflected in the gismu roots. As predicate words in Lojban (as opposed to tanru, the phrases from which lujvo are often derived), they each have a unique meaning (and associated place structure). This meaning need not be memorized by the Lojban learner - the rafsi system allows you to unambiguously take the word apart to see the tanru components that went into building the compound. You may then assume that the compound represents the most common and/or most plausible interpretation of that phrase, and you will rarely be incorrect.

Thus, as you come to know more and more of the rafsi through using the language, you become less and less dependent on a dictionary or word list to help you understand new words as you come across them. The ability to dispense with a dictionary in everyday Lojban use is the major goal and benefit of the rafsi system - it is virtually impossible to achieve fluency in a language until you are willing and able to try to use it spontaneously without looking words up that you don't know.

The ability to do without a dictionary offers a major advantage in the growth of the Lojban vocabulary, a critical aspect of the language's first years. Lojbanists, whether new or experienced, can create new words on an ad hoc basis while speaking and writing, using the rafsi system to do so quickly and easily. Doing so, you know that for a given concept represented by a tanru, there is only one lujvo structure that will represent that concept. You won't be inventing a word only to find out later that someone else expressed the same tanru concept in a different form, and that their version is right and your version is wrong.

The system of rafsi replaced an earlier Loglan system (changed in 1982) wherein compounds were formed by mashing parts of each component together without a system, with the result that you could only guess what components went into making a lujvo. The only requirement was that the resulting compound had to be 2 mod 3 characters long.

The learning problem proved severe when people actually tried to both learn the existing compounds and to make new compounds, after the first printed dictionary came out in 1975. The specific solution embedded in Lojban took 5 years to develop (1978-82), with experimentation at several steps along the way (involving many people, though unfortunately almost all native English speakers). The design you see today was not adopted lightly. Several other changes in the phonology and the morphology were also made at the same time, with all designed to mutually consistent with each other and with the goals of the language. Thus the system of rafsi was not a patchwork ad-hoc solution that doesn't fit the rest of the language - it is an integral part of the system.

In the old system, when composing a new lujvo, there were a large number of possible forms for combining the gismu components, and you would have had to look each of them up to make sure that the word had not been already created. Even if it had not been already made (and since dictionaries are inherent outdated in this respect by the time they are published, you would not be certain), you would then look up your proposed compound, to make sure that it had not been already used to represent a different, unrelated tanru. As such, mastering these early versions of the language effectively required you to memorize words in order to learn and use them, with relatively minor and undependable clues in the word-form to aid in your recognition.

With the current Lojban system, the situation is reversed. You only memorize those lujvo which you find yourself using often (in which case you memorize them simply by using them often enough that they come to mind without thinking about it). You invent new words on an ad hoc basis, knowing that someone else independently inventing a word for the same concept will likely end up with the same word, but that in any case, the word you invent will almost certainly be correct, in that it will not represent any concept other than the one you have in mind.

Briefly reviewing the Lojban rafsi system, each Lojban gismu has between 2 and 5 combining forms. Two of these are trivially and uniquely determined. The gismu itself may be used as its own combining form when it is in the final position of the lujvo. In addition, there is a related 4letter form, obtained by dropping the final vowel from the gismu, which may be used in any non-final position, by gluing it on to the following component with a "y" (pronounced as a schwa, the final sound in the English word "sofa"). Since no two gismu concepts differ only in the final vowel, this means that each concept has two combining forms, which can always be used in forming compounds that can be uniquely broken down to recognize the components.

Using only these two 'long' rafsi forms, the 4-letter and the full 5-letter gismu form, the beginning Lojbanist can use the full expressive power of the language, while memorizing no rafsi. There are no exceptions to these rules, and no complications, and the resulting word, (called the 'unreduced form') is always correct and acceptable.

The complications arise only when you become a more advanced student of the language. When you can speak and write in a language quickly, you don't want really long words for relatively simple concepts. It is fairly common to devise lujvo made up of 4 (or more) components, sometime for concepts that are used every day. Most people would be unsatisfied with a language that required them to use a 20-letter word with 8 syllables for a very common concept.

Indeed, an analysis of natural languages called Zipf's Law indicates that the length of words in actual use is inversely related to their frequency of use - the most frequently used words in a language are the shortest ones, and long words are rarely used. In languages such as English, when a commonly expressed concept is represented by a long word or phrase, common usage turns it into a contraction (like "didn't", or into an acronym or abbreviation. Examples include "TV" for "television", "TB" for "tuberculosis", "ASAP" for "as soon as possible", and "CIA" for "Central Intelligence Agency", reducing 9 syllables to only 3). It is believed by many linguists that the multitude of declensions and conjugations found in languages today are the remnants of earlier contractions.

Note that such acronyms as "TV" lose significant information about word meaning available in longer forms. "Television", for those who know the Latin roots that formed the word, reveals some aspects of the word's meaning; "TV" does not. "CIA" can stand for a variety of longer expressions, and there is no clue except context to indicate that a government organization is the intended meaning. A common English word that is apparently a short form, "OK", has completely lost its origin (leaving only unconfirmable speculations). When that happens, these compounds become like roots in themselves that must be memorized separately. This increases the difficulty of language learning, unacceptable in a constructed language like Loglan/ Lojban.

To relieve this pressure for short forms for common words, those Lojban gismu which have been found most useful in compounds have been assigned additional 3-letter short rafsi. A Lojban word may have up to one of each of the following forms: a CVC-form, a CVV-form, and/or a CCV-form, where C and V stand for consonants and vowels that are found in the source word. These short-forms may be preferred because they combine to form shorter words, sometimes with fewer syllables, than the 4-letter and 5-letter rafsi.

As a result, therefore, more than one rafsi may be used to represent a gismu/concept in making a compound, since the 4- and 5- letter forms still exist. In addition, because these shorter forms are found in other words, or even standing alone as words (cmavo) in themselves in the case of CVV forms, you need to have rules that prevent the compounds from breaking up incorrectly. Language design decisions force tradeoffs between the need to maximize the number of words that can be contracted and the requirement to retain the integrity of the compounds that are formed and the ability to break them down into recognizable meaning components.

The nature of the sounds that make up words, and the imperfections in human speech and hearing give rise to further complication in a system of word compression. Certain sounds, when adjacent to each other may provoke mispronunciation or may be misheard by a listener. Linguists also know that certain sound combinations tend to be unstable and to change with time. In designing Lojban, we had to plan ahead to avoid combinations that would likely lead to the Lojban of 2100 being significantly different from the Lojban of the first dictionary.

All of these tradeoffs have been dealt with in the current Lojban design; yet the rules for lujvo-making remain relatively simple. Some rafsi are forbidden in some word positions. Depending on word-position and adjacent rafsi, you may have to add a "hyphen" letter to make a word pronounceable, or to keep the sounds from breaking up into two words when heard by a listener.

If the rules are too difficult for your level of proficiency, you always can fall back to the long form rafsi mentioned above. You can do so because a firm rule of the Lojban design is that, if there is more than one possible rafsi combining form, the choice of form does not affect the resulting meaning. The shortest form of a word means the same as the long form. An English example where this is true is "television", which can be seen as a short form of the two components "tele" and "vision". "TV", a further shortening of the same components is taken as identical in meaning to "television". This invariance is true for all Lojban compounds, even when dozens of possible shortened forms are possible.

Dozens of forms can be possible when more than one short rafsi is assigned to a gismu. We want to assign multiple short forms, because the effects of sound interactions and the Lojban word-formation rules may prevent one particular rafsi from being used in some situations. Thus an additional short rafsi increases the likelihood that some short form is possible in a particular difficult combination; it also may mean that in other combinations where there are no sound restrictions, you will have a multitude of choices.

Of course, the rule that all of these choices will have a single common meaning means most of them will never be used. Probably only the longest form (which will be used by language beginners) and the shortest form will be used. If there is more than one 'shortest form', different people may choose different ones are preferable for a while, but usage will relatively quickly tend to settle on one of the choices. We have defined a formal scoring rules to help people pick the form that is most likely to be settled on, but it is not necessary to use it - choose the form that sounds best to you and others may agree.

Let me now turn to a Lojban example. Following is a long compound that has appeared in Lojban text:

nolraitruti'u (5 syllables)
(princess - specifically the daughter of a king/queen, as opposed to Princess Di of the UK)

If there were no short forms, this word would have to be:

noblytrajyturnytixnu (8 syllables)

Given that it is desired that you expect to memorize the Lojban word, learning it as a unitary word rather than by puzzling it together every time from its components, it should be obvious that the shorter word "nolraitruti'u" is better than the longer one. If you lived in a country with royalty such as the UK that had such a princess (as Elizabeth was before she became queen) and were prone to reading, writing, and talking about such a princess a lot, which word would you prefer to say or write?

I argue that "princess" is not that infrequent a concept, certainly deserving of a single word. The British, so I understand, do make distinctions between the various types of princess, at least in terms of how they are titled, so that the distinction is socially and linguistically important. Lojban must have separate words if there are clearly two separate concepts, as there are in this case (the 'Di' variety of princess might be 5 terms: noble-superlative-governor-son-spouse).

The longer 8-syllable form is permitted as an alternative to the short form, and might be used either in noisy environments where the longer word has all those extra sounds as redundancy checks, or by beginners who have not yet memorized the short rafsi or the compound, and are creating the compound on the fly (as this word has been created every time it has been used thus far since we have no dictionary nor people who have memorized such words). The long forms are of course needed when the words are not compounded, or you would not be able to tell a compound from a root from a structure word.

Loglan/Lojban has reached what I believe is an optimal tradeoff between redundancy and brevity, ease of learning and unambiguity of the morphology. If other solutions exist, they are unlikely to meet all the goals for the language.

Let me now turn to two hidden assumptions that Rick and others make when criticizing Lojban, assumptions I believe are incorrect:

  1. that there is a way of reducing the amount of memorization needed to gain fluency in a conlang below some arbitrary minimum, and
  2. that memorizing allomorphs is difficult.

Assuming that the set of thoughts that might be expressed linguistically should be about the same, regardless of the language, there are only so many options available for expressing those thoughts. If there is 'one word per concept', then a speaker must have memorized a separate word for each concept in order to achieve fluency. If polysemy exists, then speaker has an added burden: to memorize a somewhat smaller set of words, but to also memorize the multiple meanings of those words (including meanings he may rarely use) and some means of pragmatically distinguishing which meaning is intended.

There's no way around this. Fluent speakers don't often invent words or even derive new prefix/suffix formations when conversing. Productive language formation (i.e. inventing new words) takes time to think, and taking that time in the middle of a conversation breaks up fluency. There is some minimum amount that must be learned, even in the most regular of conlangs; no design trick can reduce this.

For a given language, for each concept you expect to talk or hear about in fluent speech, you must learn 1) at least one word for the concept, 2) the association of that word with that specific concept, and not to other concepts (including false friends from the native language), 3) any other meanings or usages associated with that word, including both polysemy and pragmatic considerations (what phrases may be appended to sentences using that word, etc. For example, if you stick an object on an intransitive verb "*I sit the store", or attach certain prepositional phrases to a word that doesn't expect them "*I give from Mary across the store" you get nonsense in any language, ungrammatical garbage in most of them.) It takes memorization to turn words into sense.

Thus, for people who are really going to use a language, the only thing you can do is ease the memorization process to make it easier to do that required memorization, to get from novice to fluency.

One way - the most frequent among conlang inventors - is to build lots of memory hooks to some natural language(s). In doing so, you risk semantics transfer that might make your conlang not truly an independent language. An example of this problem is the oft-heard debate about the Esperanto prefix "mal-" which in that language means "opposite of", but in many European languages means "bad". People native to those languages seem to often complain about 'derogative' implications of words containing "mal-", when such implications are not part of Esperanto in any way. You can't avoid this kind of problem - all languages will have 'false friends' that mislead you in learning similar-appearing new words in a new language. You can minimize it through other methods of aiding the learning process.

One way, occurring in Esperanto, is the use of affixes (such as "mal-") that modify meanings of words in certain semi-regular ways. Thus, by learning a few words and these few productive affixes, you multiply the vocabulary that results from memorization. New people then learn from seeing words that they can easily decompose - after seeing these words over and over, they suddenly find that they know both the word-formation rules, the affixes, and the compounds.

Lojban in effect carries the Esperanto technique to the ultimate extreme. Rather than a couple dozen short affixes, we allow every root to have an affix, and then make those affixes resemble the roots in very regular ways. For all Loj-ban lujvo, you automatically know that any resemblances to words of other languages are accidental, since those lujvo are always composite of simpler words in Lojban and are not derived from any other language.

As for the second assumption, I assert that Rick is wrong, and that

A very regular conlang can have allomorphs that are easy to memorize and Lojban has such a system that actually makes compound words more learnable than they might otherwise be.

There are three parts to my argument on this point:

  1. the nature of 'memorizing' of a word is non-trivial in the first place;
  2. Lojban's system is designed to provide differing aids to the novice, the experienced learner, and the expert Lojbanist, allowing the different levels of skill to concentrate on those aspects of word 'memorizing' that are easiest for their skill level and most productive for them;
  3. the Lojban allomorphs, being made in predictable ways from the gismu are relatively easy to memorize.

There are two phases to memorizing a word. In the Lojban literature, we call these phases "recognition" and "recall". In recognition, the goal is to look at a word, and be able to recognize its conceptual meaning. In recall, you must be able to go from a concept in-mind, and determine the word that represents that concept.

Recognition is by far the easier of the two skills to master, and it is the most important for the new Lojban learner. Such a new learner will probably be reading far more Lojban text than he/she will write (or if learning verbally, will hear far more than he/she speaks). When learning to recognize words in a foreign language, you can rely on aspects of the word that you are trying to learn that in some way remind you of a corresponding word in the other language.

As evidence for the difference in difficulty, people using our software tool 'LogFlash' will practice 'recognition' of a Lojban gismu, and must get it correct 3 times correctly before they attempt 'recall'. Depending on individual skill at learning, and the amount of time spent studying in advance of a first test, a Lojbanist will range from 20% to perhaps 70% correct. However, having gotten a word correct once, the minimum score for the 2nd attempt ranges from 60% to 90% correct, and the 3rd time after two correct recognitions in a row, results in over 90% correct (most errors are typos). However, the first recall attempt, which follows the 3 successful recognitions, tends to range from only 30% to 70% again, almost as if learning to recognize the word gave absolutely no advantage to learning to recall it. (Words successfully recalled once are recalled 90% correctly on the next recall attempt. However, recall skill decays relatively quickly without practice, dropping to the original 30-70% level within a couple of weeks if there have been only two test sessions. Recognition skill drops off much more slowly.

As applied to the rafsi components of lujvo, given no clues to meaning from context, the early Lojban student will still quickly gain the ability to recognize and identify the meaning of the rafsi after having to look it up a couple of times. In reality, of course, context clues may tell you what a word must mean, allowing you to recognize the components, which contribute to that meaning, even more easily. Since the early Lojban student must recognize far more words (and hence rafsi) than he must recall or generate, this is the key skill at this early stage.

At this stage, a Lojbanist generally knows few gismu or rafsi, so he/she will tend to learn them in tandem. Since the rafsi closely resemble their corresponding gismu (as I'll explain below), learning gismu helps in learning the corresponding rafsi and vice versa. Simpler Lojban texts will probably have a higher percentage of gismu than more advanced texts, and thus more words can be simply looked up in the word lists. (When the dictionary is available, I suspect that simpler texts will tend to rely more on words in the published vocabulary than on coining of new words.)

From the recognition standpoint, the lujvo-making algorithm is incredibly simple. Break a lujvo at every 'y', dropping the 'y's, then break all remaining chunks of more than 5 letters by removing 3 letter chunks from the front. You will be left with 3 letter pieces, which of course are short rafsi, at most one 5 letter piece at the end of the word, which is a well-formed gismu, and 4-letter pieces which are gismu missing their final vowel, which can be trivially identified in a gismu list. (While le'avla borrowings are rare, especially in beginning texts, they can be most readily identified either by a 3-or-more letter consonant cluster with a syllabic 'r' or 'n' after the first 3 letters - the classifier rafsi - or more simply by the fact that they fail to break down into 3 and 5 letter chunks that are all valid rafsi, as described above. le'avla never contain a 'y', so 4-letter rafsi will not occur.)

As you start to write in the language, you will already know a few gismu from reading, and maybe a few rafsi. You then have to learn to make lujvo. Initially, this can be done using long-form rafsi, with no complications. Learning long-form rafsi is equivalent to learning gismu, so no memorization is being wasted on this stage. Ideally you will memorize all of the gismu, or at least most of them. Your continued reading will teach you some shorter rafsi, because you've looked them up enough times that you no longer need to do so. These are probably going to be the most common rafsi, the ones that you will most likely need earliest in your own efforts to coin lujvo. You will also acquire a fairly instinctive feel for the conditions under which 'y' is inserted to break up impermissible consonant clusters in lujvo, but the written rules are clearly and formally stated for cases that aren't obvious. As a learner, if you insert an extra 'y' in error, you will be understood; the occasions where extra 'y's cause word breakup problems are extremely rare, and only affect fluent speech streams of spoken Lojban.

By the time you know most of the gismu, through LogFlash or by some other learning technique, you will already have recognition control on many rafsi, and perhaps even recall of a few of them. Only then is it worthwhile to start memorizing rafsi directly, and at that point it becomes quite easy to do so.

Look first at recognition. When you know almost all of the gismu, then for any given rafsi, you probably can identify all of the gismu it could represent (about 1/4 of the rafsi can only stand for one possible gismu, and many of the rest have only 2 or 3 possibilities). But since no gismu has more than one of each of the different forms of 3-letter rafsi, you will be able to eliminate some of the possibilities because you know another rafsi for that word.

Recall of rafsi is made easier by the fact that, for any given gismu, there are only a few possible rafsi, and no more than one of each of the forms. A CVCCV gismu (form C1V1C2C3V2) must have rafsi from among the 5 forms CVC {C1V1C2 or C1V1C3}, CVV {C1V1V2, with or without the apostrophe between the vowels}, and CCV {C2C3V2 or rarely C1C2V1 and the consonant cluster must be a permissible initial}. (By the time it becomes a factor, you will have learned which letter combinations are not permissible initials, since there no Lojban words start with them). A CCVCV gismu (form C1C2V1C3V2) must choose rafsi from among CVC {C1V1C3 or C2V2C3}, CVV {C1V1V2 or C2V1V2, with or without the apostrophe between the vowels}, and CCV {C1C2V1 and the consonant cluster must be a permissible initial}. In other words, up to 3 from among 5 possibilities, and you can eliminate any possibilities that you know are assigned to other words. You don't need to know all of the rafsi for a given word at first, since you can always use the long forms till you are sure of the short forms. Thus, you use what you know, and acquire new rafsi as you need them. Of course, every rafsi you can recall, you can almost certainly also recognize.

As an example, take the gismu "bangu" The possible rafsi are "ban", "bag", "bau", "ba'u" (the 2 CCV forms bna and ngu are ruled out because of impermissible initials). There can be only 1 CVC and only one CVV rafsi, so "bangu" has at most 2 rafsi. It turns out that they are:

bangu ban C1V1C2 (CVC) language
      bau C1V1V2 (CVV)

and readers of this article have probably already learned the "ban" rafsi, since it occurs in the name of the language, Lojban.

It should be easily seen in this example that the more rafsi you actually do know, the easy it becomes to learn the rest. You have a closed set of three-letter forms, nearly all of which has a meaning. By the time you know a third of the rafsi, a 1/4 guess becomes a 1/2 guess. By the time you know 2/3 of the rafsi, you probably can deduce 90% of them without a word list, because you can determine so many by elimination of alternatives.

Of course, learning the rafsi helps you cement in your knowledge of the gismu themselves. If you know 'bau' is a rafsi for the word for "language" (bangu), you know that C1 is b, V1 is a, and V2 is u. This rather reduces the burden of learning the other two letters. If you know the other rafsi is "ban", then you know that either C2 or C3 is 'n', and you can almost certainly guess the word at that point. (In speech you can probably get away with slurring over the other consonant and the listener will guess what word you wanted from context.)

Revised rafsi Assignments

The Lojban rafsi list, the set of affixes associated with the various gismu and a few cmavo, has explicitly not been baselined along with the gismu list during the last few years. This is because the initial assignment of rafsi was based on merely educated guesses on what was needed, with some highly suspect data as the basis for those guesses. The intent has been to wait as long as feasible to build a data base of actual lujvo-making usage before making the assignments permanent. The rafsi assignment list has been exceptionally stable over the intervening years partly to encourage lujvo-making, and partly because there was no bona fide basis to make judgements about rafsi needs without usage data.

Now, with the impending dictionary publication, we want to have rafsi assignments with a greater confidence of adequacy and stability. Indeed, the publication of a dictionary that we hope to be able to sell in book form for a few years requires that we baseline the list. The tradition in Lojban design has been to have a thorough review immediately prior to any baseline decision. This report describes the results of such a review.

In July and August of 1992, the complete set of rafsi was reanalyzed based on the 4 years of actual usage since the original analysis. Because of new data, the report proposed many changes to the set of rafsi. These changes were reviewed by a committee from the community, and almost half the changes were thrown out at least partially in the interest of language conservatism.

With this rafsi retuning and recent re-examinations of all Lojban gismu place structures, all aspects of the Lojban design will have had two or more separate thorough reviews, separated significantly in time, to ensure that the design can stand the test of time. While the proposed changes are a fairly high percentage of the total set of rafsi assignments, the set of assignments seems to me (who knows the set of rafsi best, to be much the same as it was before.

For both efforts at assigning Lojban rafsi, they have been assigned using a method developed by JCB for old Loglan during the 1979-82 timeframe, and described in TLI publication "Notebook 2", believed to be out-of-print; the document was not all that useful, mainly being a 200-page catalog of supporting data for what I describe much more briefly here without such complete data. JCB called his process 'tuning' the rafsi list, or 'optimizing' it for 'coverage'.

'Coverage' refers to the extent to which words are used in lujvo compounds, which is of course the major use of rafsi (they are also used to a more limited extent in names and le'avla borrowings, the latter of which has been taken into account in my latest review, as noted below). The goal is to ensure that a maximal percentage of Lojban lujvo compounds can be composed from 'short' (CVC, CCV, or CVV/CV'V form) rafsi.

This goal is based on the paradigm known as Zipf's Law, which has been fully embraced by the Loglan design for at least the last two decades. The Loglan/Lojban paradigm actually goes beyond the 'law' as inferred by Zipf, which merely observed a tendency in language and other phenomena to inversely relate length of a phenomenon to frequency. As the original law is descriptive rather than prescriptive, it has been questioned on occasion as a design principle for Loglan. I do not intend to defend this design principle, merely to state that it is a central tenet of the Lojban design philosophy in accordance with our policy of following JCB's central design tenets for Loglan.

Applying Zipf's Law to Loglan design, we have assumed that the law will, whether we allow for it or not, govern the evolution of the language as it becomes used widely in less-controlled circumstances as we expect in the future. We want to try to see where the language will end up (presumably in a state consistent with Zipf's Law), and design features into the language that will allow for that evolution to take place smoothly, without actually needing to change the language design when it occurs. To the extent that we can foresee the future of the language, we want to make the changes now, and not later, when people have already learned the vocabulary.

One result suggested by Zipf's Law is that words of greater frequency in usage tend to be shorter. If a word comes into greater use, it is observed that it becomes shortened, either by natural word compression. Such compression might include the compression of sounds as in "cannot" to "can't", or the tying words together in compounds like lujvo rather than leaving them as longer tanru (e.g. the English lujvo "grandfather" - interesting in that many pronounce it with a silent 'd' as Zipf appears to continue to shorten the word after its written form has been frozen in spelling). Similar processes include the use of acronyms, a phenomenon which Lojban supports but tries to discourage.

Now there are other reasons for making lujvo other than merely frequency of usage. One obvious reason is to get a more useful place structure, whereas a tanru has the place structure of the final term. But the inherent unpredictability of lujvo place structures (notwithstanding various proposals for regularizing them) means that most lujvo will be made because someone sees that the word/concept in question will be used multiple times in multiple contexts, and hence justifies being thought of as a 'word', rather than a phrase.

At this stage there is not a lot of a priori decision making going on regarding lujvo-making. People usually make lujvo when the concept is expressed by a single word in the language they are translating from. But this is a valid practice, and indeed is most common when compounds are 'borrowed' from other languages, a process called 'loan translation'. Of course, not all Lojban lujvo that have been proposed correspond to single words in other languages, so even at this point, Lojban is evidencing its own trends in concept/ word formation independent of other languages.

It is presumed that under Zipf's Law most people will make lujvo to cover concepts of higher frequency, leaving as phrases those concepts that occur once, or in specific, isolated, context-dependent situations. Thus JCB put a priority on making gismu and lujvo to represent concepts found in the one generally recognized cross-language study of the use of concepts in languages (as opposed to words), Helen Eaton's study from the 1920s and 1930s. Unfortunately that study is outdated, and its association with 4 European languages makes this data questionable as the sole basis for a modern language design. Now that we have actual Lojban usage to include in the design evaluation, for the first time we can downgrade the importance of Eaton's study.

History of the Loglan/Lojban rafsi system - The use of rafsi in languages, including conlangs, is not particularly controversial. Esperanto, for example, has a wide variety of prefixes and suffixes which operate roughly as Loglan's rafsi do. The extent to which Loglan/Lojban uses and indeed depends on rafsi may be more controversial.

Pre-1982 Loglan had haphazard compound formation, with the effect that compressed compounds had a structure such that etymology and hence implied meaning could not be elicited from the word. As a result, the 'correct form' of a compound had to be memorized, and to a great extent, a given compound could be looked at with relatively little possibility of recognition of its compound nature or of its implied meaning.

The GMR (Great Morphological Revolution) redesign in 1978-1982 incorporated the concept of 'resolvable affixes' (rafsi) such that the fact that a word is a compound could be recognized on sight, and the nature of its etymology and hence significant clues as to its meaning could be recognized by identifying the rafsi of which the word was composed. In the spirit of Loglan's design, resolvable affixes were to be unambiguously assigned to specific gismu roots, so that recognizing the rafsi identified a unique etymology, and rules that allowed a compound to be unambiguously recognized as being composed of these, and only these, rafsi.

The Loglan/Lojban design now allows for both 'long' and 'short' rafsi. Long rafsi are identical to the basic gismu (all of CVCCV or CCVCV form) for final position in a compound only or have the final vowel replaced by a 'y' (pronounced as a schwa) in non-final positions. Thus the long form of a compound for "broda brode" will be "brodybrode". (The 'example' gismu "brodV" are the only gismu in the language that share the same final vowel and hence have ambiguous lujvo compounds - but then they are used most often for making examples. The current reanalysis has given a limited alternative to this ambiguity for those rare usages of these that are non-exemplary).

It must be clearly understood that there is no guarantee that a lujvo compound means exactly what one would infer from the source metaphor. Language use is rather too chaotic to assume that. Indeed, Lojban policy is to assume that the source metaphor is ambiguous and context-dependent, whereas upon adopting a shorter compound form, that form becomes a single word in its own right with a unique meaning and place structure like all other Lojban content words (brivla).

Zipf's Law, plus this distinction between metaphor and compound, require that the compounds be both shorter than and distinguishable from the source metaphor. All Lojban gismu can form long-form compounds of this sort; the use of 'y' replacement in non-final rafsi assures that there is unique resolution, while also ensuring that the words do not fall apart. In accordance with Zipf's Law, all such compounds are at least trivially shorter than the uncompressed 'metaphor' (tanru) from which they are formed. If short rafsi exist, the compound can be shorter still.

Since all Loglan rafsi occur only in bound forms (inside compounds), it was recognized that some shorter forms than the five-letter rafsi could be used. Unambiguous word-resolution limited this set of shorter rafsi to CVC, CCV, and CVV forms, where in Lojban a VV pair might be one of the four primary diphthongs or a disyllable vowel pair (which is marked with an apostrophe ' to indicate a devoiced, non-glottal-stop glide, which English speakers usually approximate with an 'h' sound.) Older Loglan forms do not have the distinction between a diphthong (such as "oi") and its corresponding divowel form ("o'i", pronounced as in "toe heel"), hence have fewer possible CVV rafsi. (Note that the CVV rafsi are totally unrelated to the CVV-form cmavo. The rafsi occur only in bound form, and the rules for lujvo-making mean that the rafsi can never be heard as separate words. In some cases, a rafsi may have a meaning related to that of the cmavo spelled the same way (and this is recognized as a good memory hook to aid in learning the words), but such matches occur only because the cmavo assignments were also chosen where possible to be associated with gismu which would suggest the cmavo's meaning.

Since all gismu in the language are considered one part of speech and syntactically identical, it is a language requirement that all gismu be allowed to serve in all positions within compounds; we cannot have a limited set that is more 'worthy' of use as prefixes or suffixes in compounds. We can use Zipf's Law to assign short rafsi based on other factors, the minimum requirement that all gismu have combining forms for all positions sets the dictum justifying the universal availability of 4-letter + 'y' and 5-letter, 'long-form rafsi' that can be used for any gismu.

Given the current rules for Lojban sounds and word-making forms, There are 1445 possible Lojban CVC rafsi, 493 CVV rafsi, and 240 CCV rafsi. The rules for combining these compounds:

  • forbid a CVC rafsi in final position;
  • require a 'y' inserted between rafsi:
    • when they are conjoined so as to result in certain 'proscribed medial consonant clusters';
    • to prevent 'assimilation' that would make it hard to distinguish that combination from some other combination;
    • as glue in two other special circumstances where a compound might break up into smaller pieces;
  • require a syllabic 'r' or 'n' (rules determine which is used) to glue on a CVV rafsi in first position where it might 'fall off' in spoken contexts and be mistaken for a separate unrelated structure word (cmavo) of the same CVV form. (CVV rafsi do not need to be glued on the front only in a two-part lujvo where the final term is a CCV rafsi, because the Lojban's penultimate stress rules hold the pieces together).

Including current new word proposals, there are 1342 Lojban root words, and 93 cmavo that are useful in delineating meanings of compounds that are also given short rafsi (where possible the rafsi is the same as the cmavo, but this isn't always possible.) Since there are only 733 rafsi that can be used in final position (CVV and CCV forms), it is not possible to assign such a short rafsi to each root, in spite of the theory that permits any of them to appear in final position. Because Loglan/Lojban words were created based on recognition scores in source natural languages, they are not uniformly spread around the alphabet. We wanted to make the rafsi set easily learnable, so we limited the set of possible rafsi for a given gismu to specific permutations built from certain letters of the word. Thus for "broda", possible rafsi include only -bod-, -rod-, -bro-, -bo'a-, and ro'a-.

In some cases, there's no trouble assigning a rafsi to a gismu - there is only one gismu with the letters permitting use of the rafsi given the rules for deriving possible rafsi. This determines perhaps 550 rafsi in the first pass (1 in 3.5 of the CVC rafsi, 1 in 5 of the CVV rafsi, and 1 in 6 of the CCV rafsi). But given that no gismu could have more than one of a given type of rafsi, and some simplifying assumptions (such as noting that a gismu having a CCV did not need a CVC or a CVV rafsi, especially if it would prevent another from using that rafsi), another 500 rafsi are trivially decided, perhaps 1/2 of the total.

On the other hand there were some rafsi that are extremely difficult to assign. In the recent retuning, for example, there were 33 competitor-words that could use -ci'a-, and 33 for the two possibilities -sai- and - sa'i-, while as many as 500 rafsi (mostly CVC, but nearly 100 of the more valuable final position rafsi) could not be used by any gismu. Only reinventing significant numbers of gismu, choosing lower recognition score word-forms could significantly improve this maldistribution, and such a change would not be considered under our baseline policy. (Only one gismu has previously been reinvented to get a usable final position rafsi, mleca, meaning "less than". As part of this retuning, the gismu for "daytime" is being changed to "donri" to allow it a good rafsi. This second change was considered only because the word was added to the set of gismu so recently, that it is not on the published gismu list, and hence is little known.)

Because of the limited set of rafsi, we want to make the rafsi assignments optimal for our word set, so as to minimize the length of compounds formed in accordance with Zipf's Law (presumed to be most of them). This means that we have to 'tune' the set of assignments based on some type of usage statistics.

When we first assigned rafsi in 1987-8 after constructing the gismu roots, there were no usage statistics. Older versions of Loglan had been used in only very scattered bits of text, and were based on a set of only around 900 gismu roots, including a bunch that were judged inappropriate as 'basic roots' like words for 'billiards' and 'football', and were hence not retained into Lojban. Most of these words had been used in a set of predefined compounds JCB's 1974-5 dictionary chosen because they represented the most common concepts in 4 European languages (based on Helen Eaton's study). This data is suspect of being both European-biased and outdated, though no better study is known.

The metaphors underlying the 1974-5 compounds were often culturally biased, and relied on English-language based conventions unrelated to the Loglan words they were built on. Classic bad examples of underlying metaphors in that dictionary include "man-do" for "to man a ship" (which can easily be done by a woman, and has no functional association with manhood), and the word for "kill" (now a Lojban root), based on "dead-make" where the word for "make" means "x constructs y from components/materials z" (meanwhile ignoring the 4 completely different Loglan words for indicating causality). Indeed "- make" was used in some 500 compounds, and non-specific "-do" and "-cause" (associated with only one of the 4 causality words) in several hundred more each, making a substantial part of the old Loglan vocabulary rather restricted in semantic variation. The Lojban vocabulary is intended to be far more analytical in terms of the Lojban meanings of the words, and current actual usage ranges over a much wider variety of roots. But the older Loglan data necessarily dominated our initial rafsi assignments.

Our other source of data besides JCB's dictionary were words invented by Loglanists, either in efforts to cover the rest of Eaton's word lists, or to cover concepts not in the dictionary that were needed by people in the few texts in Loglan that were attempted. These were generally either patterned on the already poor examples in the 1974-45 dictionary, or, even worse, were built on haphazard ad-hoc methodologies generally in ignorance of the rules for compound-making that had been set down. These included the much lambasted (for obvious reasons) "dog-woman" for the pejorative equivalent of English "bitch", "one-future-one" for "in sequence" ('one' is a cmavo and had no final position rafsi, so the word-inventor just used the CV-form cmavo, resulting in an illegal word), and "water-pass_ through-skin" for "sweat" (the latter uses the worst possible term order; Loglan grouping would lead one to expect the metaphor to refer to a kind of skin, whereas the English verb "to sweat" might be a kind of 'passing-through', and the English noun 'sweat' might be a kind of 'water'). There was of course no frequency data for any of these words, other than the frequency inferred from Eaton's list for that subset, which basically implied that all such 'Eaton words' would be among the most frequent words in Loglan and hence should wherever possible have short forms.

In 1979-82, JCB did a statistical analysis of the words in his dictionary, choosing a set of resolvable affixes to minimize the percentage of words that could not be written with short forms. In 1987, Lojbab repeated that analysis, using that data, along with a hundred pages of notes on words proposed for Loglan in the intervening years, most of the low quality exemplified above. Only some of the additional Eaton data was incorporated; we didn't have the software tools to handle such a large data volume, and didn't want the language design overwhelmed by the poor quality of most of the metaphors. Because of a lack of software tools, we compiled statistics manually (probably making errors, and including some entries multiple times when they were invented independently by different sources. But the result was still a significantly broader semantic field of words - approximately 97% of the lujvo in Loglan's compounds were reducible to short forms in JCB's 1982 tuning; the 1987 tuning based on a much larger set of words only achieved 94.6% reduction.

It was recognized from the start that these initial assignments would have to be re-evaluated based on actual usage, of which there could not be any until we had a stable gismu list. This requirement leads to a 'Catch-22' situation where you have to have people learn the rafsi well enough to use them naturally, while preserving the flexibility to change them. Change will naturally be resisted by people who have taken the trouble to learn something, and the Lojban project has been strongly committed to recognizing and respecting the amount of effort that goes into learning a language, and not demand unnecessary relearning through constant change.

Since I was the likely person to do the eventual retuning, I (Lojbab) made it a point to be the first to learn the set of rafsi (using the old version of LogFlash 2 developed especially for this purpose), and made it a point to try use them heavily when writing in the language. Thus the re-learning penalty if there are changes falls at least as hard on me as on anyone. We also recognized that we could probably only do this reanalysis once - uncontrolled change in the language is debilitating to morale, so we've waited till the 'last minute' before dictionary publication.

Unfortunately, the minimal amount of change in the rafsi list over the last couple of years misled some into thinking that the rafsi were baselined with the gismu list, so we often repeated the statement of its not being baselined. Still, we avoided changes, because people won't use something that is constantly shifting underfoot like sand. Even when new gismu were added, we shied away from changing any rafsi to accommodate them (though we assigned them rafsi from the unassigned set when they were available).

Luckily, what has happened fit our needs quite well. Few people actually learned the rafsi in any systematic manner like I did (I know of no one besides me who completed even one run-through of the rafsi list with LogFlash 2, and only a few have reported even trying to use the program.

Some people, like Nick Nicholas, have used lujvo heavily in writing, though he clearly hasn't memorized most of the rafsi (one of the few problems with Nick's texts has been trying to figure out what his words were supposed to be when he fails to look up a rafsi and guesses wrong - that many people are able to do so shows that the language doesn't require people to memorize every rafsi in order to communicate effectively). Nick also makes good Lojban lujvo, since he supports the idea of conventions in lujvo-making to a great extent. Though I disagree with making conventional standards for lujvo at this point in the language development, conventions generally lead to choosing appropriate components and getting them into a plausibly acceptable order, a result clearly better than the strange efforts by some of the old Loglanists.

Because of Nick's and others' heavy usage we considered certain rafsi assignments to be 'sacred' as part of the reanalysis. For example, we could not seriously consider changing -loj- for logji/logic and -ban- for bangu/language, since that would change the name of the language. Likewise, other commonly used words were considered inviolate, like "selbri", "le'avla", "brivla" (though some of these assignments did vary before the gismu list was baselined: bridi used to have the rafsi -rid-, and 'brivla' was at one time 'ridvla' (but this lujvo would now indicate a source metaphor of 'fairy-word'). The current word "selbri" in our early documentation is "kunbri", -kun- having been reassigned from kunti/empty to kunra/ mineral). But our documentation is now too extensive for us to lightly change such words, and indeed my threshold against change was to protect a few dozen rafsi absolutely against change, and only reluctantly consider changes to another large group. Thus "blari'o"/bluish-green had some claim for 'sacredness' (but not absolute), even though it has only appeared to my knowledge in one set of examples - the recently published Diagrammed Summary.

Still, if rafsi were to reflect frequency of usage, that means that some of the most frequently used words had to change rafsi, so as to get one more useful given its typical position in a word. Since the possible-rafsi-space is densely packed with the existing assignments, though, retuning by assigning a rafsi to word A generally means freeing that rafsi from word B, which then needs a rafsi currently used by word C, hopefully moving down a list until you get to a word used seldom enough that people won't so much mind it not having a rafsi.

In July, 1992 I used software tools to process some 3 Megabytes of Lojban text and English commentary on Lojban text, identifying some 2700 lujvo created and their frequency of usage. (Because the processor could not distinguish English from Lojban, a few English words crept in because they looked like Lojban lujvo; e.g., the English word "simple" might be a lujvo based on the unlikely metaphor "mutual-paper" - this mis-classification happened relatively rarely.) The frequency data was used logarithmically to weight usage data - a word used twice got a score of 2, used four times got a score of 3, eight times getting 4, etc., up to words like selbri and brivla used several hundred times and getting weights of at least 10. This weighting supports both the Zipf's Law basis of the language, and pretty effectively made sure to protect rafsi assignments that are 'sacred'.

I also used different tools to process the Eaton proposals into the statistics. As noted, these metaphors aren't too good, but the words in question cover a broader semantic spectrum than actual Lojban usage. Also many of the meta-phors are bad mostly in being phrased in a weird-for-Lojban order, as in the above example "skin-pass through-water". Thus even these poorly-made words give suggestions as to gismu that need rafsi coverage, though should be ignored in deciding whether a word gets a final-position or initial position oriented rafsi assignment. Words in the Eaton file were only given a weight of '1', and multiple-occurring usages in Lojban text thus far outweighed these terms. Eaton proposals thus probably only served primarily to break ties in the 'competition', and to ensure that the broadest possible range of words was represented.

The new statistics obviously tracked more closely with actual usage. However, the 'coverage percentage' of the current rafsi assignments dropped to only 92.6. This sounds pretty good, but is almost 3 times as bad as JCB's original tuning, and 50% worse than the rafsi assignments had been under the original statistics. The actual Lojban usage data was less than 1/2 of the total weighted data, and was even more poorly covered, around 89.7%.

In addition, since 'coverage percentage' does not reflect hyphenation, the quality of the coverage was even more mediocre. For example, the cmavo, 'ka', much used in lujvo in recent times, was originally assigned the rafsi 'kaz'. 'kaz' is hyphenated before c/f/k/p/s/t/x/j because of the compounding rules. These letters form cover more than 60% of the actual rafsi in non-initial positions weighted for actual usage. Nick Nicholas and 2 others thus asked that 'ka' be given a less-hyphenated rafsi.

In a couple of cases, I overruled a statistical quirk after verifying that, for example, that it was based on some particularly bad metaphors in the Eaton data. But for the most part, statistics led the decisions. The resulting proposal improved coverage only a small amount, from 92.6% to 93.8%, but coverage of the actual Lojban usage portion of the data improved more significantly, from 89.7% to 92.8%. Given the constraints to minimize changes to 'sacred rafsi', this was about as good as could be hoped.

Review by the community led to elimination of many of these changes, since people considered a few more rafsi assignments to be 'sacred' than I did in my analysis. But the disapproved changes had only minor effect on coverage statistics (no percentage has actually been calculated based on the final assignments appearing in this issue).

Methodology - This section deals with details of the methodology I used, and may be skipped by people not interested in such details.

As stated above, I gathered statistics on usage of gismu in various positions in lujvo. These positions were 1st/3-or-more term lujvo (allowing any rafsi, but CVV rafsi must always be hyphenated), 1st of 2-term lujvo (any rafsi is permitted, but CVV are only sometimes hyphenated), middle of 3-or more terms (any rafsi is permitted, but CVV/CCV never need a hyphen afterwards), and final term (CVC rafsi forbidden, CVV/CCV about equally useful, but CCV is one syllable shorter than a CVV with an apostrophe, and is thus preferable for the highest usage words).

Given these rules, it is clear that CCV rafsi are the most flexible. A word with a CCV rafsi never needs a hyphen afterwards, and needs a hyphen before it only part of the time when preceded by a CVC (an unvoiced-initial CCV is hyphenated about 25% of the time, a voiced-non-liquid CCV about 40% of the time, and mlV/mrV rafsi are hyphenated less than 10% of the time).

CVV rafsi can be used in any position but almost always require a hyphen in initial position. Since there are more than enough words that need CCV and CVV words for final positions alone, I emphasized using CVV rafsi for concepts concentrated in final positions in the data words but relatively little usage in initial positions in metaphors, CCV rafsi for words with significant final position concentration, but also having high usage in other positions - in other words with high overall position scores. CVCs are reserved primarily for words concentrated in the first positions. (CVC assignments were also favored for gismu often used as le'avla classifiers, because CVC rafsi are the easiest to use as classifiers.)

I presumed to 'tune' at first assuming only that a few 'sacred' rafsi would remain untouched, but otherwise assuming all assignments were freely determinable without reference to the past. With this assumption, 30-50% of the rafsi could be assigned either as 'sacred', or as having little or no competition for the rafsi best suited for them.

For the most part, I proceeded as if I were starting to assign words from scratch, using 'sacredness' only to dictate choices when they came up. The alternative would be to identify specific words that needed new rafsi as a change from the current set (such as 'ka'), and 'force them' into a new assignment cascading along a chain of rafsi assignments until a rafsi was found that wasn't already assigned to a word. This paradigm is very useful for understanding the actual effects of a series of changes in retuning. As a methodology, however, it is highly suspect. There is no obvious test for when a word 'needs' a rafsi other than direct comparison of the statistics. People's instincts can be woefully inaccurate on this score. Thus, while 'ka' indeed turned out to justify a rafsi, Nick Nicholas also proposed giving 'drata' -dra- , taking it from drani. It turned out that both statistics and actual lujvo data show that drata is almost never used in final position, while drani often is.

I made a few other assumptions that explicitly deviated from the original rafsi assignments, based on understanding the word-making implications of the lujvo-making algorithm better. Words with CCV rafsi are hyphenated so seldom that it rarely improves coverage to give the gismu another rafsi in addition. Thus, once I assigned a CCV rafsi to a word, I ruled out adding a CVC or CVV rafsi for that word as unneeded, unless no other word could benefit from the rafsi. Only 'sacredness' was allowed to interfere with this principle, hence zmadu, with no competition for -zma, was assigned that rafsi, and did not need -zad- or -mau-. 'mau' was deemed moderately 'sacred', though, and was kept with zmadu anyway. Unusual for a word with a CCV, this extra rafsi may be occasional useful since it starts with a different letter than the -zma-, hence is useful to avoid hyphenation in about 25% of lujvo where it is preceded by a CVC. However zad- was freed and is no longer assigned to "zmadu" or to any other Lojban word.

A much larger variety of gismu have now been used in lujvo; in a couple areas of the alphabet, something had to give. For example, to assign one of 'kal/kam/kan/kar' to 'ka', 1 of the existing 4 words using those rafsi had to give up its CVC assignment. Each of these CVC rafsi was the only assignment for its corresponding gismu, so this decision was going to deprive a word of having any rafsi at all (there was no possibility of a chain of changes displacing a CVV or a CCV rafsi).

In actual lujvo usage data, CVV rafsi have been avoided in initial positions in favor of CVC rafsi, especially when they are di-syllable (with an apostrophe between the vowels). Indeed, even when a CVC also requires a hyphen afterwards, it has been preferred to a CVVr in the same position.

This actually contradicts the experience of JCB when he did 'taste-tests' to determine the lujvo-making choices of the old Loglan community - his conclusion then was that people tended to like vowel-rich compounds as more melodious and easier to pronounce than words with many consonant clusters. (A possible counter-explanation is that consonants provide better aids to word recognition, and are thus preferred by people who want to easily recognize the components in a written lujvo; such a tendency was not measured in the 'Taste Tests' conducted by JCB.) Because of this tendency, I de-emphasized CVV scores in the initial positions, assigning them almost solely on the basis of final position usage. The following data shows one example.

gismu le'avla 1st/3 1st/2 mid end
sanga  0       1     16    2   27
stagi  3       0     0     0   4
New assignment gismu old assignment
sag   sa'a     sanga sag
-              stagi -

sanga gained the rafsi -sa'a- based on extensive new use in final position, a score of '27' in that position guaranteed it such a selection. Having the rafsi "sa'a", it is arguable that the word no longer needs the rafsi 'sag', and it should have been used for 'stagi'/ vegetable, which has 3 usages (all in le'avla); though all other usages of "stagi" thus far are in final position where a CVC rafsi would do no good. I overruled this change, recognizing that with the substantial score for sanga in 1st of 3+ terms (1) and 1st of 2 terms (16), there would be a lot of instances of sa'ar- that lujvo-makers have dispreferred given a choice.

Generally I let a word with a CVV rafsi keep a CVC in addition only if the score for initial position usages exceeded all competitors by at least 5/1 ratio, as it did in this case. Else we would end up with a few words having almost all the rafsi. I gave slightly better favor to words to keep a CVC rafsi assignment that they had had previously, as sanga had previously had -sag-, and indeed that was the determining factor in this example, consistent with the goal to minimize unnecessary change.

Another assumption was more subjective. For the original rafsi assignments, a requirement was that all culture words be given a rafsi. Since each such culture word associated with a country automatically had at least 8 identifiable lujvo (e.g., merkyjecta merkybangu merkyrupnu merkyfepni merkykulnu merkyturni merkygugde merkynatmi, etc.) this policy was justified, and indeed 8 usages was generally enough statistically to warrant such a lujvo in the original tuning. But since then, the culture words have come under a lot of attack, and some Lojbanists have said they will avoid using them. At least one person specifically recommended freeing their rafsi assignments for use by other words (though 'sacredness' would preserve the heavily used 'gic'/glico, 'lob'+'jbo'/lojbo and 'mer'/ merko. Similarly, a variety of words associated with chemical elements have been attacked - most of their usages are figurative ones dating from the JCB era, and figurative tanru metaphors are now dispreferred in Lojban usage. Finally, all metric units were presumed to have a defined lujvo for each metric prefix (about 16).

I downgraded all statistics for these words by at least a factor of two, even when doing so meant that the calculated coverage would decrease. For example, because Nick translated some texts from Ancient Greek, there were some usages of 'xelso' in final position. This warranted giving xelso the assignment of 'xle', currently held by 'naxle' (canal) which has no actual usages indicating that 'xle' would be useful in addition to its noncompetitive CVC assignment of 'nax'. Nick specifically recommended against "xelso", and I took his recommendation more broadly to apply to all such cultural compounds. Some gismu in this set lost all of their rafsi assignments because of the down-weighting, many of these being culture words which were borderline to even have gotten a gismu in the first place.

Measurement word scores were down-weighted by a similar argument. "snidu" had its CVV - si'u- removed in favor of the slightly lower scoring simxu, a change that would not have been considered based on pure statistics. Nora argued that, while all metric prefixes were theoretical compounds for snidu, in natural languages of metric countries which also permit such compounds only a few metric prefixes are actually used with each measurement. Thus we may talk of milliseconds, but seldom deciseconds, dekaseconds, or exaseconds. On the other hand, Nora favored retaining -gra- for grake/gram because its most frequent use is in the compound kilogram which in Lojban would require a CCV rafsi to avoid hyphenation. In this case, I did not downgrade the scores, and grake kept 'gra'. Thus, some amount of subjective judgement was used in deciding assignments for culture/metric/element words.

I painstakingly assigned rafsi to each gismu, working approximately 12 hours a day for 3 weeks. This was a largely manual job involving cross-checking among several dozen pages of statistics. It is hopefully a one-time job and hence was not worth the effort to develop programs to do the analysis automatically. Perhaps a good spreadsheet might have saved some time, but I don't have a spreadsheet that could handle this much data, and designing and testing a standalone program would have probably taken more time than I spent.

Four metric gismu proposed by John Cowan were included as effectively equivalent to all other metric words; the exact form of these words was selected to minimize rafsi assignment problems, since we had to modify the actual prefixes to fit Lojban gismu anyway.

When I was done with this exercise, I looked at unassigned rafsi and tried to find cmavo that could reasonably have a use for them, in some cases proposing a CCV for a CV cmavo by inserting a consonant. Since the cmavo assignments have proven to be most unpredictable and unsatisfying based on statistics, this seemed like a wise course. For cmavo, I felt that it is better to assign a rafsi and drop it if it isn't used after the 5 year baseline than to not assign one and have the cmavo be difficult or unable to be used in lujvo (in which case we might never know they were needed). The community overruled me on this, choosing to leave rafsi unassigned in borderline situations, thus minimizing the memorization of possibly useless data, and noting that any cmavo can be incorporated into a lujvo-equivalent using "zei", though this is not Zipfean.

Lest people worry, I expect that after the 5-year baseline, while usage might provide data warranting significant retuning of the rafsi list, the assumed philosophy will be to oppose revising rafsi assignments. At this point we are concluding a design phase; after 5 years of usage, we can only justify fixing what has demonstrably been found unreasonable or void by actual usage.

I put the results into the computer, and made lists of chains of changes as described above, to make them easier to understand. A couple of chains proved to offer questionable improvement and were backed out. Where changes seemed to affect 'sacred' rafsi disproportionately, I created alternate changes for the community to select from.

The resulting set of change proposals was posted to Lojban List. Several Lojbanists commented on the draft version of this report included with that proposal, and several people indicated a desire to vote on individual changes. As a result, a large number of the changes I proposed were rejected (some involving changing of rafsi that others considered 'sacred', but mostly involving assignments of rafsi to cmavo that were not certain to need them). The community also asked for a couple of other guidelines to be factored into the analysis, such as minimizing the number of gismu with multiple rafsi assignments (especially 3 rafsi assigned to a single gismu), unless there was really a good reason for them. This led to some new changes.

Summary of results

In the baseline version approved after review, there are a total of 457 changes in rafsi assignments, about 30% of the total, affecting 372 total gismu and cmavo. This overstates the actual change rate, since in most cases, giving a rafsi to one word means taking it away from another, giving 2 actual changes. The adopted total is significantly lower than the original proposal, which would have changed 590 rafsi. The community rejected about 1/3 of the proposed changes, though it requested a small number that I did not have in my original report.

Following is a more detailed breakdown.

CVC rafsi
   1445 possible rafsi
   915 assigned. (64%)
   257 assignments changed (28%)

Of the changes, 97 words lost CVC assignments where they once had them (many of these also had a CCV or CVV and didn't need both). 100 words gained CVC rafsi where they did not have one before. 60 words actually changed from one CVC to a different one, generally as part of a cascading chain.

CVV rafsi
   493 possible rafsi 
   421 assigned (85%) 
   149 assignments changed (35%)

Of the changes, 66 words lost a rafsi without replacement and 65 gained a rafsi they didn't have before. Some of the 18 remaining assignments involve switches between a CVV and its corresponding CV'V to give a word with a lot of initial position usages the monosyllable rafsi. Monosyllable CVVs seem not to be as rejected by Lojbanist lujvo-makers in that position as disyllable ones, perhaps because the resulting word seems shorter.

CCV rafsi
   240 possible rafsi 
   209 assigned (87%) 
   51 assignments changed (25%) 

Many CCV changes were switches with CVV assignments, sometimes freeing up a CVC rafsi (since a CVV word may need a CVC rafsi while a CCV rarely does), thus cutting off a long chain of changes that might have affected several more words. 24 words lost a CCV rafsi, while 27 gained one that they did not have before. No words changed CCV assignments, an option that was rarely possible.

The numbers and percentages of changes may seem large, given the small benefit in coverage (that benefit is actually even lower than the benefit mentioned above; The numbers above were calculated based on the original proposal fore changes, some of which did not occur. However, percentage coverage seems now to be a less significant figure than the degree of failure to cover words that have a great deal of usage in lujvo.

While the overall coverage percentage changed by only a small amount, most 'problem words' were given useful rafsi. In the 1987 rafsi assignments, a word was considered a 'problem word' if it had more than an uncovered score of '4'; i.e. more than 4 lujvo where no reduced form could be used. No problem word in the original data had an uncovered score more than 8.

By comparison, no less than 111 words had 'uncovered' scores more than 10 when I started tuning and 52 had scores exceeding 15. The worst words had uncovered scores exceeding 30. This means that there were an awful lot of lujvo using these words in ways for which they did not have short rafsi. These numbers, though large, do not affect the coverage percentage much; the latter percentage includes some fully-covered words with weights of several hundreds.

As a result of tuning, 37 words with scores over 15 were reduced to a score less than 15, while 7 others were forced above that level to make room for them. Thus there are now only 15 such really severe problem words, with the highest scores being two words with uncovered scores of 19 and 3 words with 18 (one of the latter being "snidu", for which we decided to discount the numerous metric lujvo).

There remain 51 words with scores above 10, so the total number of problem words was cut by more than 1/2.

The enclosures give complete lists, in several orders, of the new rafsi assignments.

[This issue contains several essays written by Greg Higley. Greg, who is not on the computer networks, has only contributed irregularly on Lojban topics, but has still been able to affect the language design with his insightful comments. (Note that his examples and translations are not necessarily sanctioned, but are sometimes of the nature of discussion or proposal. See the comments after each essay, which sometimes indicate that a given example was either ungrammatical or means something other than what Greg intended.)

The essays are generally located with other essays on similar topics, so that this issue forms a cohesive flow.

On lujvo

by Greg Higley

I'd like to make a few comments on nu jvozba. As I've read, the current policy of la lojbangirz. is "Let a thousand flowers bloom." While at first I was opposed to this, I now see the wisdom of it: How could it be otherwise? I've decided after much thought to disband the lujvo pulji and let the prisoners go. (nu jvozba "lujvo-making"; lujvo pulji "lujvo-police")

But this doesn't mean that I don't have anything to say on the topic of nu jvozba! Au contraire, mon frŠre! I have actually come 180ø from my old viewpoint: I'd like to suggest - since 'suggest' is really all I can do - that a different view of lujvo be adopted.

As I understand it, a lujvo, as currently defined, is a tanru that has been "compressed" into a single word, and that has been assigned a fixed meaning. (And I guess a new place structure, as well.) Thus the essential difference between the tanru "remna sovda" and the lujvo "remso'a" is that the former does not have a fixed meaning, it might mean "the human's egg", i.e., the one he had for breakfast, or it could mean the same thing as (what I'm suggesting for) remso'a, namely "human ovum", i.e. the female human reproductive cell.

I see lujvo more as "abbreviations" than "fixed tanru": I don't think a lujvo has to be so exact that its meaning is crystal clear. Then we'd have huge lujvo. I see the parts of a lujvo as forming a "memory hook" which can be used to remember its meaning, and which, knowing the concept, can be used to remember the lujvo. I don't think that, seeing a lujvo on a page, you should instantly be able to know what it means. Rather, finding out what it means, you should then be able to more easily remember it. Case in point is "le'avla". This is a word well-known to Lojbanists, but let us assume that we've never seen it before. Would you know what it meant, just by looking at it? You could rely on the context in which it occurs, but what if there were no context, or what if the context wasn't informative enough? You could probably make some educated guesses, but let's face it, "le'avla" is not a very clear lujvo as lujvo go. Expanding it into a tanru is just as unhelpful: "lebna valsi" is just as nebulous. And yet I'd like to argue that this is just exactly how lujvo should be made! Once you discover the meaning of "le'avla", you aren't likely to forget it: You can now see why it means what it does. This is similar to the process that goes on with an abbreviation, although thankfully lujvo have clearer parts than abbreviations. You can't necessarily figure out the meaning from the abbreviation, but you can figure out the abbreviation from the meaning. With lujvo, it might be more accurate to say that, given a list of lujvo, you could pick out the one that corresponds to the concept in question.

"General Purpose Lujvo"

by Greg Higley

One of the reasons why I don't do much translating from English to Lojban, or from Welsh to Lojban, is that in order to do this with any reasonable degree of accuracy, you have to make lujvo. Well, I do make them, but I usually don't start out with an English or Welsh word or concept that I'd like to translate into English. I start out with the gismu list and just start combining, trying to see which combinations suggest meaningful concepts. This is how I arrived at the idea of "General Purpose Lujvo".

While making lujvo in this way, I'd often come across a word which had no exact equivalent in English, but which seemed to be useful nevertheless. A good example is "zaltapla". This is anything ground up and made into a patty. It doesn't have to be meat, doesn't even have to be food. If you're eating a hamburger, and you call it "le zaltapla", you aren't likely to be misunderstood, and you can always get more specific if you want. I find that this makes Lojban much more interesting, because it divides the semantic space in a different, perhaps "Lojbanic" way, and it helps me to think "Lojbanically". If you wanted to say "That hamburger looks good" in Lojban, you're likely to try to make the word for "hamburger" very specific. While there's nothing wrong with this - clarity is a good thing. I think doing this makes Lojban no more than a code into which we translate the pre-existing concepts of other languages. With GPL, or even lujvo that are unique, but with specific meanings (SPL "Specific Purpose Lujvo"?), we can build a language that is not just a code, but a living language of its own, that divides the semantic space in its own way.

Mark Shoulson:

Higley makes a good point, and it touches a little on something that I've been thinking about a lot myself. I feel that a lot of the Lojban text written suffers from overuse of lujvo owing to a tendency to try to reproduce the specificity afforded by natural-language terms. I try to use more tanru than lujvo, and to be as non-specific as I can, while still saying what I want to say (with a few exceptions; e.g. I don't use prenu as "person" in the English meaning of "human being" - that's a "remna". "prenu" is more of "thinking being" or even "soul" (minus the religious and non-bodily connotations)). So I avoided Nick's "beipre" for "waiter": what did the "prenu" rafsi add? The waiter is just "that which carried the coffee": "le bevri be lei ckafi". Sometimes you may need to be more specific, that's okay. But I think you'll find that you don't need to be specific as often as you might think at first. That the "bevri" was also a "prenu" gets cleared up later, when conversation is initiated.

Higley's view of lujvo as "abbreviations" rather than "fixed tanru" is very cogent and, I suspect, very close to the official view of what lujvo should be. His example, le'avla is a good example. After all, "le'avla" expands to "lebna valsi" which is "take word" or better, "taker word" - a word which is somehow associated with a taker, perhaps. A more pedantic "jvozba" would have made it "selyle'avla", for "se lebna valsi": "taken-thing word", much closer to the meaning: a word which is taken. Note, though, that that's not what we use, nor should it be: you can't trust an expanded lujvo 100%, you can only assume that it's close to what the lujvo means. lujvo are intended to be dictionary words, having their own definitions not precisely derived from their associated tanru (the "selpinxe"/"se pinxe" problem I had before is another good example. "selpinxe" is a good lujvo for "na'o se pinxe", i.e. "a beverage", as opposed to just plain "se pinxe" which could mean "ca'a se pinxe", "liquid-thing-sliding-down-someone's-throat".)

Colin Fine:

I agree somewhat with Greg, and wholeheartedly with Mark, especially about inappropriate specificity. (I recall once inviting people to join me in a campaign against precision!)

I also like to play around with possible lujvo - and go beyond the obvious when trying to coin them.

One thing I do in text is that I will sometimes use a more precise lujvo the first time I introduce a concept, and then omit a term or two from it thereafter. Thus having once said "samymrilu" I will thereafter quite happily use "mrilu" later in the passage.

Greg proposes and explains some lujvo

lujvo velcki
bromalsi "synagogue"
musymalsi "mosque"
xisymalsi "church" etc.
[And so on. It's no new discovery that the names of the major religious edifice(s) can be made with "malsi".]
jelspo "destroy by burning"
[This is the basic meaning. More colloquial translations might be "put to the torch, burn down, burn up" and many others. "-spo" can be added to many words to create interesting lujvo of this type:]
po'aspo "destroy by (causing to) explode"
[It is the x2 place of "po'aspo" that does the exploding. "lenu ta spoja cu po'aspo ti" covers any x1 explosions nicely.]
zdabartu x1 is exterior to/outside of the nest/dwelling of x2
[As in "Mom, I'm going outside.": "doi mamta .i mi zdabartu klama".]
zdane'i x1 is inside of/interior to the nest/dwelling of x2
[As in "I'm staying in": "mi zdanei stali". It could also roughly mean, "at home" - as long as x2 is the same as x1.]
zdasta x1 stays at home x2
zaltapla x1 is a tile/patty/etc. made of ground-up material x2
[This is one of the "General Purpose lujvo" I talked about in my comments on lujvo.]
rartapla x1 is a naturally occurring tile-shape of composition x2
taktapla x1 is a ceramic tile of specific ceramic x2
drutapla x1 is a ceiling/roof tile of composition x2
[A GPL. It isn't specific as to whether ceiling or roof tiles are needed. But if you're tiling your roof, and you say, "Joe, hand me that drutapla", you aren't likely to be misunderstood. It's the same thing in English. When tiling a roof, you don't keep repeating "roof tile" over and over. You eventually just say "tile".]
zdabartu drutapla "roof tile"
[There may well be an easier way to say this. "bartu drutapla" might not clearly mean "roof tile". I don't know anything about carpentry or the like, but "bartu drutapla" could be some kind of "exterior ceiling tile" as opposed to an interior one.]
po'ertutra x1 is territory (property) owned by x2
ni'ablo x1 is a submarine
[I experimented with a number of different terms for "submarine", but I think this sums it up nicely. I had "sfeni'ablo", but "sfe-" turned out to be rather redundant: What else would it be under but the surface?]
zalre'u x1 is ground meat from source x2 [A good GPL.]
remso'a x1 is a human ovum from woman x2
remtsi x1 is human sperm from human x2
cticinza ["cinza" used for eating]
benmro brain-dead
[Lojbab: A little unclear what you mean by this - the most common colloquial usage of the English, of course, is merely a form of "mabla". If you are referring to the medical state, this seems fine.]
jiksre x1 errs socially in x2 ["social faux-pas".]
menmikce [A general purpose lujvo: "psychiatrist, psychologist, counselor" etc.]

le lojbo se ciska

New ckafybarja Submissions

It appears that theres been little work done on the ckafybarja project since JL17, and I am beginning to think that the schedule for the planning phase was much too ambitious. The only new material received was one English-language personality sketch, giving us 3 to choose from. Since there are planned to be 7 such common characters, we don't have enough to make much of a choice yet.

Also included later this issue is Nick Nicholas's second ckafybarja piece, written last year, which he was revising at the time JL17 was being prepared.

Character Sketch

by Zoe Velonis

She had the kind of body that clothes couldn't contain. It wasn't that she was so fat that she burst out of whatever she wore, that her flesh strained against the warp and the weft, but that she had the kind of body that clothes just shouldn't confine. Her bra straps were forever falling down: she'd go about the kitchen tugging at one absent-mindedly as she stirred a concoction. Buttons would fly off at a moment's notice, turning up later in a bowl of soup. The zipper of her jeans had to be anchored with a safety pin else it would slowly creep down, leaving her blushing.

Her naked body was voluptuous, resplendent, Rubenesque. Never of the personality to subscribe to the feminine beauty myth, she exuded both femininity and beauty, from her thighs to her belly to her gloriously round and pendulous breasts.

He would come to her at night, creep into her bed and bury himself in her warm, soft flesh; nestling his face between her thighs and reaching up for huge handfuls of her breasts, marvelling at her bounty as she tossed her head and moaned with pleasure. She surrounded him, took him in, made him feel complete.

In the daytime she never gave any sign that she knew of his nightly visits. She was the cook, he a busboy, and there was no hint of affection or shared pleasure, much less gratitude, in her voice as she thrust dishes at him, giving him instructions in a firm, clipped voice that bore no contradictions.

He'd worshipped her beauty for weeks, in the beginning, longing for her, his flesh aching for her, his mind consumed by the demands of his loins. He'd sneak outside the cafe' at night, stare up at the window he knew was her room as she turned on the light. He'd watch, hypnotized, as she languorously disrobed, brushed her hair, leaned out of the window to breathe deeply of the night air. Her breasts shone like twin moons as she drank in the night, erasing the scents of garlic and rosemary, butter and tomatoes from her nostrils. Once, as he watched, she laughed, a low, quiet chuckle, and opened her arms in an embrace. "Come up then, why don't you," she said, her voice rich with a melodiousness off nuance that it never had during the day. His breath caught in his ribs, clung there until he remembered and opened his lungs again. "Me?" he asked, desperately grateful that his voice didn't display that annoying habit it had lately, of cracking when he particularly wanted it not to. "No, the other people who are out there watching me every night," she said, the laughter still in her voice.

So he went back into the cafe', past the night janitor who whistled as he wiped down tables and mopped the floor, who gave him a knowing wink that made him all the more nervous. He went through the kitchen and paused at the foot of the stairs, put, finally, one foot on the first protesting step.

Thirteen stairs, he counted, and crossed himself. He turned down the hall, past the head waiter's room, the manager's to her room. As he stood outside, breathing heavily, his pants distended with his desire, she opened the door.

Her nakedness was more than he'd dreamed of. Not perfect: he could see the silvery stretch marks on her breasts and thighs, the moles and freckles, the pits and scars of age. But her imperfection only made her more achingly real, more desirable, and his genitals throbbed against his jeans. Breath came in short gasps.

"Have you ever been with a woman before?" she asked.

Mute, he shook his head. It was the truth: his absentminded penetration of his sister's best friend when they were all playing doctor behind the abandoned barn didn't count. She took his hand and led him into the room, whose walls were covered with tapestry bedspreads that exuded odors of frankincense and patchouli. She guided him to the bed and undressed him carefully, opened herself to him and then, when he had spent his first desire in her, taught him how to pleasure a woman as well as himself.

He realized, at one point, that he didn't know her name, that she didn't know his. Somehow it seemed desperately urgent that she whisper his name at her climax, but when he told her, she only laughed.

And now she was just another part of the day to him, the thing that he escaped to when his shift was over each night, threading his way through the tables and up the stairs to her soft, endless flesh. She was always the same, never cried or wept or showed that anything touched her emotion.

Her laughter, though rich, was only amusement, never joy or happiness; and he wondered if the walls would echo with her moans of pleasure without him, if she even needed him. So one night he stayed away.

She looked the same the next day, but the one after, her face seemed drawn. He watched her carefully, but she never said anything to him or to anyone, and although for a month she grew paler and thinner, stopped tugging at her bra straps, and although her cooking grew bland and tasteless, the decline finally ended. Her color came back and her voluptuousness was even more irresistible. He thought that she had found a new lover and, jealous more than he had thought himself capable of being, he mounted the stairs one night to see.

There were no sounds from her room and he had almost turned away when he heard her low rumble of a laugh. He opened the door quietly and peered into the darkness.

The window was open, making the tapestried bedspreads billow in the air, sending out whiffs of their scent like tendrilled ivy. And she...her bed faced the window and on the ceiling was a mirror. She lay, legs spread wide to the night, looking up at herself, and laughed a laugh of joy and happiness. As he watched, she moaned and tossed her head in that way he knew so well, and then she cried out, syllables that formed what he knew must be her name, and wept, tears of release and happiness as well as pain and emotion.

He crept out, closing the door softly behind him, and tried to blank out the emptiness inside him with alcohol, tried to forget that the night and the mirror and her own hand had done what he never could.

It was then that the cafe' began to become very popular, then that its cook began to acquire her reputation for food with the indefinable passion, mer'aki, for being a chef unparalleled by any before.

Grammar Changes

The next section of this issue is the largest, and deals primarily with changes to the grammar. We first present the proposed changes to the Lojban grammar baseline, which will become official with book publication. Detailed discussions of a few of these, recorded at the time they were proposed, will reveal a bit about how the decisions to change the grammar are made, and perhaps show that such decisions are never made lightly.

The largest portion of this discussion is devoted to the change in Lojban relative clauses, which is centered on Change Proposal number 20, but also led indirectly to several other changes.

In most cases, proposals discussed in this section have been adopted in some form, although not always in the form originally proposed in the discussion. Sometimes, for example, we were able to resolve a problem just by explaining things a little better, or possibly by making a change to the cmavo list (adding or deleting a word, or changing the selma'o or detailed definition).

Proposed Changes 1-32 to the 2nd Baseline Lojban Grammar

[Terminology note: Ek, JEk, GIhEk, ZIhEk, GUhEk, JOIk, etc., have traditionally been used to refer to the sets of logical/non-logical connectives of the appropriate type, and their compounds that involve negation of either the preceding or following term (or scalar negation of the connective in the case of JOI). This is a useful shorthand when talking about these families of compounds that are function identically in the grammar.]

Executive Summary:

  1. Change Ek+KE and GIhEk+KE to lowest precedence
  2. Add JEk+BO construction
  3. Add various new free modifier locations
  4. Add ZEI compounds
  5. Allow observative after GI in forethought connected sentences
  6. Regularize BOI with free modifiers
  7. Simplify relative-clause connection to "zi'e" only
  8. Allow I+BO at the beginning of text
  9. Allow bare NAI at the beginning of text
  10. Allow any kind of JOI in forethought
  11. Remove POhO
  12. Allow full selbri after NIhE
  13. Disallow NAhE in forethought termsets
  14. Allow multiple I or I+BO at the beginning of text
  15. Allow conversion of abstract and negated selbri
  16. Allow ZAhO+NAI for contradictory negation of event contours
  17. Merge LUhI into LAhE; make NAhE+BO equivalent to LAhE
  18. Merge BRODA and LEhAVLA into BRIVLA
  19. Regularize rule names in YACC and E-BNF versions and update comments
  20. Revise grammar of relative clause incorporation in sumti
  22. Change description of Step 5 in preparsing to match reality
  23. Allow CUhE to be logically connected to other tenses; forbid NAhE+KI
  24. Allow KI after CAhA (and including it) rather than before
  25. Disallow NA [tag] after CO in inverted tanru
  26. Allow only selbri rather than bridi-tail after NAhU
  27. Allow I, I+BO, NIhO after TUhE
  28. Create NAU+tag as a non-logical connective (probably ANNULLED)
  29. Change MAhO from lerfu-to-operator conversion to mekso-to-operator
  30. Allow afterthought JOI in termsets
  31. Allow JOI+BO and JOI+KE parallel to E+BO, JE+BO, and JE+KE
  32. Allow JAI without following tag, as unclefter

Currently, the logical connective constructs Ek+KE (and GIhEk+KE) have higher precedence (bind more tightly) than either Ek+BO (GIhEk+BO) or Ek(GIhEk) constructs.
Give Ek+KE (GIhEk+KE) the lowest precedence among Eks (GIhEks).
In 1987 (NB3 = Notebook 3 TLI) Loglan, the equivalent of Ek+KE and GIhEk+KE had low precedence. In the first Lojban baseline, Ek+KEs had been changed to high precedence, and in the second baseline, GIhEk+KEs were changed to follow. In writing the logical connective paper, considering constructs like

A .e B .ake C .e D

suggested that the most reasonable interpretation is:

(A .e B) .ake (C .e D)

Therefore, this change restores the original Loglan situation, which supports that grouping.

Currently, there is no way to group tanru components logically in pure afterthought. The only alternatives are:

X je Y ja Z

which groups left to right

(X je Y) ja Z 


X je ke Y ja Z [ke'e]

which groups right to left

X je (ke Y ja Z [ke'e]) 

but is a hybrid of forethought and afterthought.

X je (Y ja bo Z) 

analogously to

A .e (B .abo C) 

in sumti.
Uniformity and flexibility.

Allow free modifiers (such as subscripts, vocatives, and metalinguistic comments) in the following new places:

after LOhO when not elided
after LAhE for both sumti and MEX operands
after CO
after CEI
after NU[NAI]
after NA preceding a selbri or a GEk-bridi-tail
after NAhE BO
after NAhE, except in tenses and within NAhE+BO (which are lexer compounds)
after TUhE
after TEhU when not elided

Increased flexibility.

There is no way to construct lujvo that involve le'avla or cmavo, unless the cmavo have been assigned rafsi.
Add the metalinguistic cmavo "zei" (selma'o ZEI) which will join the word before it and the word after it into a construct treated by the parser as of selma'o BRIVLA. More than two words can be joined by using multiple "zei"s. The words "zo", "zoi", "la'o", "lo'u", "le'u", and "fa'o" cannot participate, since they are delimiters of quoted text, which will be resolved by the lexer before compounding with "zei".
Other methods of incorporating le'avla into lujvo are extremely error-prone and subject to a multitude of special-case tests. No method of incorporating cmavo into lujvo has ever existed, encouraging speculative assignment of rafsi to cmavo that might be used in lujvo. (TLI Loglan allows incorporating lerfu into compounds using a 'magic' compounding method.)

It is not currently grammatical to say:

ge mi klama le zarci gi klama fa mi le zdani

Allow logically connected sentences wherein the first sentence has terms before the selbri but the second one does not. (The reverse situation is still forbidden, because it looks like bridi-tail connection to a LALR(1) parser.)
The previous restriction was arbitrary and unnecessary.

"boi" gets special treatment unlike that of all other elidable terminators. In all other cases, free modifiers may optionally appear after the elidable terminator (in which case it can't be elided). Free modifiers must be placed before "boi", however, because "boi" is used to terminate subscripts, and subscripts are a species of free modifier.
Regularize the rules for "boi" so that it takes free modifiers after it, except that no free modifiers at all are permitted on a "boi" that terminates a subscript. ("ve'o" already has this split personality: no free modifiers if it is terminating a subscript, but allowed otherwise.)
Simplicity and regularity. A new convention is needed for subscripts on subscripts, however; so we simply declare that consecutive subscripts are taken to be nested.

Multiple relative clauses can only be placed on a single sumti by connecting them with logical connectives, namely ZIhEks.
Eliminate ZIhEks except for a single cmavo, "zi'e" of selma'o ZIhE, which places two relative clauses on the same sumti but does not count as a logical connection.
There is some doubt whether any of the ZIhEks make sense other than "zi'e", which puts both relative clauses into effect. Unlike other logical connectives, ZIhEks cannot be split up into multiple sentences. The existing implementation of ZIhEks was incomplete, and did not allow the full functionality of other logical connectives, and there is no easy way to make them work. Analysis shows that the most likely combinations of relative clauses can be easily expressed with other types of logical connectives within a single relative clause. The only restriction this places on the language is the as-yet-unused situation of a non-AND connection between two relatives of different types (restrictive and nonrestrictive).

Mark Shoulson comments: This one I have some trouble with. I'll concede that in most cases, GIhEks and the like within the relative clause will suffice for logical connection, but there are some things that we lose by dropping ZIhEks. For one thing, how could we do logical connections (other than "AND", of course) between restrictive and non-restrictive clauses? Granted, I can't think of much of an application for such an animal, but it may be a needed construct.
Also, we lose logical connections between NOI phrases and GOI descriptions. This one actually has applications. For example, a system of locking things on many MUDs (Multi-User Dimensions: text-based, multi-user, user-extensible thingies that are sort of adventure games or chat programs, (or something in-between) depending on how people choose to use them) often works with methods like "A person who is carrying the key, or who is Herman, can pass through this door." In the old method, this is neatly done with "lo prenu poi ponse le ckiku zi'a po'u la xerman. cuka'e pagre levi vorme". No muss, no fuss. In the new method, we'd have to expand out the "po'u" to get "lo prenu poi ponse le ckiku gi'a du la xerman. li'u", which granted is okay, but loses the whole point of having "po'u" in the first place (it can always be expanded). (actually, an even more Lojbanic translation, in the old grammar, would be "lo prenu pe le ckiku zi'a po'u la xerman.", taking advantage of the symmetrical nature of "pe").
John Cowan responds: Mark has presented the first useful rationale for "zi'a" that I have ever seen: "poi broda zi'a po'u la xerman." Nonetheless, I still think that the logical problems of "poi broda zi'V noi brode" are overwhelming; if we were going to split up NOI and POI (and GOI and PO) into separate selma'o, there might be a rationale, but we aren't.

Currently, a text can begin with a bare ".i" or an I+JEk, but not with an ".ibabo".
Allow I+BO, I+JEk+BO, I+tense+BO, and I+JEk+tense+BO at the beginning of text.
Allows people to complete each other's expressions by adding causals, presuppositions, and the like.

Theoretically a text may begin with "nai", and this bare "nai" is taken as attitudinal. However, the parser does not currently handle bare initial "nai" in embedded texts within quotes or parentheses.
Allow bare initial "nai" explicitly within the grammar rather than as a preparser hack.
Uniformity and consistency.

Forethought JOIks (also known as JOIGIks) are restricted in their syntax. In particular, GAhO brackets are not permitted in forethought.
Permit any sort of JOIk, so that JOIGIks are any JOIk + "gi".
Simplicity and uniformity, plus the ability to specify GAhO brackets on forethought intervals.

Three kinds of fragmentary utterances (bare I with or without JEk or modal, bare number, bare NA) currently have a special terminator "po'o" (of selma'o POhO). This terminator is always elidable.
Remove POhO.
Earlier versions of the grammar required POhO, possibly due to an implementation weakness in the YACC version used in developing that grammar. It is never necessary because it can always be elided, so it serves no purpose except to clutter the grammar.

Only a restricted form of selbri (simple selbri plus optional linked sumti) are currently allowed after NIhE.
Allow any kind of selbri.
The former restriction was meant to remove ambiguity, but now that the TEhU delimiter has been introduced, it does the necessary job, and so a full selbri is permissible. This grammar is also parallel to that of MOhE, which allows a full sumti.

In forethought termsets, a NAhE is allowed just after the NUhI.
Disallow this NAhE.
Nobody can figure out what it might mean to have a scalar negation of a termset, a construct which currently exists solely to implement a certain kind of logical connective. What does it mean to scalar-negate not a term but the logical connection of two or more terms?
Change 30 makes explicit the use of non-logical connectives in termsets, and scalar negation of such non-logical termsets makes some sense, possibly enough to justify the status quo, even though no usage has yet been found to support it.
This change has been incorporated in the current draft of the new baseline, but will be reconsidered at least once before final baseline for book publication. If any Lojbanists can propose an authentic use for the construct, this will be considered in the final decision.

Only a single instance of I or I+BO (and their related compounds) is allowed at the beginning of text (per change 8 above).
Allow multiple Is or I+BOs consecutively.
Symmetry and simplicity. With the elimination of POhO, multiple Is are now allowed at the end of texts and between sentences.

It is not possible to convert an abstract selbri [NU + bridi] or one that has been (scalar) negated [NAhE + selbri].
Allow these forms. The place structure of [NAhE + selbri] is that of the original selbri.
Simplicity and uniformity.

PU and FAhA allow -NAI for contradictory negation. This is not very useful on tenses (punai = na pu), but very useful for sumti tcita to deny that the relationship holds. ZAhO cannot take -NAI, although it is also useful as a sumti tcita.
Allow ZAhO+NAI.
Consistency and general usefulness:

mi morsi ca'onai le nu mi jmive

I am dead, but it is not the case that this is so during my life.

There are three kinds of qualifiers which can be prefixed to a sumti, giving another sumti:

  • LAhE provides indirect reference, indirect discourse, and sumti raising;
  • LUhI changes sumti between individuals, sets, and masses;
  • [NAhE+BO] provides sumti scalar negation.

LUhI has terminator LUhU; the others have no terminators. LAhE is also allowed on mekso operands.
Merge LAhE and LUhI into a single selma'o, with the current grammar of LUhI but named LAhE (for compatibility with the past). Allow the same grammar for sumti and for MEX operands. Change NAhE+BO grammar to be the same as LAhE, thus allowing it on operands as well.
Proposed changes to the sumti grammar (including Change 20 below) make LAhE and NAhE+BO messy without terminators. Merging them with LUhI allows greater generality (expanding the expressiveness of the language) and simplicity, without needing to add a new terminator. NAhE+BO is a compound and cannot be merged directly, but can be made grammatically equivalent.

Technically, brivla fall into three selma'o: LEhAVLA (for le'avla), BRODA (for broda/brode/brodi/brodo/ brodu), and BRIVLA (for everything else).
Merge LEhAVLA and BRODA into BRIVLA.
The grammar is identical and the machine parser has never bothered to make the distinction anyway. It is a relic of long-ago pre-baseline versions.

Various rule names:

bri_string -> selbri
bri_unit -> tanru_unit
header_terms -> prenex
utt_string -> paragraph
cmene_A_404 -> cmene_404
ekroot -> ek_root
no_FIhO_PU_mod -> simple_tag
sentenceA -> sentence_A
indicators_412 -> indicators_A_412
bridi_valsi_408 -> bridi_valsi_A_408
JOIk_JEk_957 -> simple_JOIk_JEk_957
PA_812 -> number_812
PA_root_961 -> number_root_961
BY_string_817 -> lerfu_string_817
BY_string_A_986 -> lerfu_string_root_986
modal_972 -> simple_tense_aspect_972
modal_A_973 -> simple_tense_aspect_A_973
modal_B_974 -> modal_974
modal_C_975 -> modal_A_975
BY_987 -> lerfu_word_987
space_time_* -> space_* (where "*" stands for each of several letters)
interval_mod_1050 -> interval_modifier_1050
interval_prop_1051 -> interval_property_1051.

Consistency between the YACC grammar and the E-BNF version and other documents. Also, this results in no two rules differing only in number. (Some rules have the same names as selma'o, though.)

(See JL18 text article!)
Relative clauses on descriptions are grouped by the parser so as to attach to sumti before outside quantifiers are put on. The actual semantics of what is being attached has been pragmatically determined, and analysis has now shown that this can theoretically be vague/ambiguous or even limiting to expression in the language, though work-arounds probably exist for all problems raised.
Allow the distinction between a relative clause attaching to the "inside set", excluding external quantifiers, of a description. A relative clause outside the KU will refer to the entire sumti. A relative clause inside the KU will generally be preposed so as to parallel the historical pseudo-possessive which is recognized as a transformation of an inside-set relative clause. However, postposed relative clauses will be inside by default, matching the way in which the parser inserts elidable terminators (i.e. only if needed).
Comparable expansion of the relative clause possibilities inside vocatives is incorporated in this proposal.
The current grammar appears to group relative clauses with the "inside set" of a description sumti, that portion of a sumti including from the LE to the KU which includes the inside quantifier and not the outside quantifier. In the case of non-restrictive "lo" descriptions, and possibly some others, this is not what is normally intended.
Example: "pa lo sipna noi melbi" groups as "pa <lo sipna noi melbi>" apparently adding the incidental claim that "all sleepers are beautiful".
The problem manifests itself in various forms more completely documented in a long paper by Colin Fine, but the bottom line is that the existing grammar is vague as to what a relative clause attaches to, and there are definable cases where this vagueness can lead to unacceptable ambiguity.
The proposed solution has the secondary virtues of:

  1. making pseudo-possessives visibly match the parallel inside-set relative clauses, but without overt relative clause marking;
  2. making it obvious how to express a pseudo-possessive with a quantifier ("le ci mi broda" is a complete sentence and not a sumti, since "le ci mi" is a complete sumti. With preposed inside-set relative clauses, "le pecimi broda" is unambiguously a sumti.); and
  3. the problematical "[quantifier] [quantifier] [description]" is eliminated from the language (analysis can give a meaning for this expression of "[quantifier] lo [quantifier] lo [description]", and it has even been used once or twice, but experience has shown that the analysis is counter-intuitive to many people, who see also "[quantifier1] lo [description] [quantifier2]-mei" as plausible).

Postposed inside relatives are allowed in all descriptions, so the preposed/postposed distinction becomes a forethought/afterthought distinction, which can be valuable. Existing texts retain their currently official inside-relative interpretation (unless the KU is explicitly present, a rarity), which is arguably desirable as the default (though it must be recognized that there are text examples where the speaker obviously wanted to apply the relative clause to the externally quantified sumti.) The negative tradeoff of this is that KU becomes always required when you want an external relative clause. (Other options were considered and rejected by the net-based Lojban community.)
Preposed relative clauses (but not relative phrases) will almost always require a terminator, though monosyllabic "vau" is usually as applicable as "ku'o".
The following analyzes all definite and indefinite relative clause cases.

Descriptor External quantifier present internal quantifier present noi/poi
le no no poi
le sipna poi melbi
[ro (le su'o sipna poi melbi ku)]
The sleepers who are beautiful...
le no no noi
le sipna noi melbi
[ro (le su'o sipna noi melbi ku)]
The sleepers, who are beautiful...
le no yes poi
le ci sipna poi melbi
ro (le ci sipna poi melbi ku)
The 3 sleepers who are beautiful...
le no yes noi
le ci sipna noi melbi
ro (le ci sipna noi melbi ku)
The 3 sleepers, who are beautiful...
le yes no poi
ci le sipna poi melbi
[ci (le su'oci sipna poi melbi ku)]
3 of the sleepers who are beautiful...
le yes no noi
ci le sipna noi melbi
[ci (le su'oci sipna noi melbi ku)]
3 of the sleepers, who are beautiful...
le yes yes poi
re le ci sipna poi melbi
re (le ci sipna poi melbi ku) re le ci sipna ku poi melbi
[re (le ci sipna ku)] poi melbi
[The] two of the 3 sleepers who are beautiful...

The Lojban in this case makes the distinction based on presence of the "ku", forcing the speaker to think about the distinction when important.

le yes yes noi
re le ci sipna noi melbi
re (le ci sipna noi melbi ku) re le ci sipna ku noi melbi
[re (le ci sipna ku)] noi melbi
Two of the 3 sleepers, who are beautiful...

The Lojban in this case makes the distinction based on presence of the "ku", forcing the speaker to think about the distinction when important.

lo no no poi
lo sipna poi melbi
[su'o (lo ro sipna poi melbi ku)]
Sleepers who are beautiful...
lo no no noi
lo sipna noi melbi
[su'o (lo ro sipna noi melbi ku)]
Sleepers, who are beautiful...
lo no yes poi
lo ci sipna poi melbi
su'o (lo ci sipna poi melbi ku)
At least one of the 3 in the universe that sleep who are beautiful...

(the following is a more likely example:)

lomi ci cukta poi melbi
su'o (lomi ci cukta poi melbi ku)
At least one of my 3 books that are beautiful...

(Quantifying the inside set emphasizes it so that the restriction applying to it seems natural - natural enough that English requires forcing an indefinite description if there is an inside quantifier.)

lo no yes noi
lo ci sipna noi melbi
su'o (lo ci sipna noi melbi ku)
At least one of the 3 in the universe that sleep, who are beautiful...
lo yes no poi
ci lo sipna poi melbi
[ci (lo rosu'oci sipna poi melbi ku)]
3 sleepers who are beautiful...

With no inside quantifier, the English becomes an indefinite, and there is no suggestion that there is an inside-set, much less that the relative clause relates to it. Likewise in the current Lojban which is equivalent to the indefinite

ci sipna poi melbi

(which under this change will have the ku after the melbi to separate from other sumti). The restrictive clause unambiguously talks only about the 3 sleepers, since in an indefinite there is no internal quantifier to put secondary focus on the inside set - the set of all sleepers. If the inside quantifier "ro" was present, under this change the restrictive clause would attach to the inside set unless explicitly closed off with "ku".

ci lo ro sipna poi melbi
ci (lo ro sipna poi melbi)
Three out of all sleepers who are beautiful.
ci lo ro sipna ku poi melbi
ci (lo ro sipna ku) poi melbi
[The only] three of all sleepers who are beautiful.
lo yes no noi
ci lo sipna noi melbi
[ci (lo [rosu'oci] sipna noi melbi [ku])]
3 sleepers, who are beautiful...

(The English again becomes an indefinite and the incidental clause goes outside. Note that this time, the English remains ambiguous and odd-sounding no matter how you phrase it:

?3 of sleepers, who are beautiful...
?3 of those sleepers, who are beautiful...

unless you go to

3 who sleep, who are beautiful...

which is better reflected in Lojban as

ci da poi sipna zi'e noi melbi

which accurately puts the relative clause outside.

lo yes yes poi
re lo ci sipna poi melbi
re (lo ci sipna poi melbi ku)
re lo ci sipna ku poi melbi
[re (lo ci sipna ku)] poi melbi
Two of 3 sleepers who are beautiful...

(The English is totally ambiguous as to which sleepers are beautiful, and the Lojban in this case makes the distinction based on presence of the "ku", forcing the speaker to think about the distinction when important.)

lo yes yes noi
re lo ci sipna noi melbi
re (lo ci sipna noi melbi ku)
re lo ci sipna ku noi melbi
[re (lo ci sipna ku)] noi melbi
Two of 3 sleepers, who are beautiful...

(The unlikely English is totally ambiguous as to which sleepers are beautiful, and the Lojban in this case makes the distinction based on presence of the "ku", forcing the speaker to think about the distinction when important.)

IMPORTANT NOTE: Change 20 affects nearly all of the sumti grammar rules. There may be unforeseen side effects, although analysis so far has shown that the only reduction in expression is the confusing "[quantifier] [quantifier] [description]" which has a much clearer equivalent.
However, the introduction of such a major change at this late stage of the project makes it highly controversial, as any problems may show up too late to be easily fixed (i.e. after books are published).


Bring the description of lexer compounding (Step 5 of the preparser) in the comments at the beginning of the grammar into conformance with the way the current implementation (as well as all its predecessors) actually do things.
The comments in question were written presuming that the parser would use method 5b, i.e. insertion of lexer tokens. All actual practice has employed method 5a, i.e. replacement of lexer compounds by single tokens. It seemed to be more useful to document actual practice: 5a and 5b have different ordering implications.

The current rules for connecting "cu'e", the tense/modal question, with other tenses using JEks or JOIks are erroneous and hopelessly irrational. "cu'e je bai" is legal but "bai je cu'e" is not. Also, "na'e ki" is legal but meaningless.
Put "cu'e" on a level with space/time tenses and with modals. No modifiers such as scalar negation are allowed to affect it. This is what Imaginary Journeys (John Cowan's paper on Lojban tenses published with JL16) says. Put bare "ki" on the same level; this does not affect "ki" following modals or tenses.
The YACC grammar said one thing, the E-BNF another, and Imaginary Jourmeys a third. The Imaginary Journeys version is clearly what makes sense. NAhE+KI was the unintended result of a previous fix intended to get bare KI working.

In complex tenses, the optional CAhA (for potentiality) comes after KI, and therefore cannot be made sticky.
Place the optional CAhA before the optional KI.
Sticky CAhA is not unreasonable.

It is currently legal, though pointless, to insert NA (contradictory bridi negation) after the CO of an inverted tanru, rather than in its usual place at the beginning of the selbri. Furthermore, it is possible to follow such a NA with a tag or another NA or various combinations.
Disallow them by splitting up current rule 131, which conflates CO handling with NA handling.
The disallowed constructs have never been used by anybody, have no advantages over the normal use of tenses/negation at the beginning of the selbri, and may tend to confuse people if used - they look like a negation/tense that applies only to the second half of the selbri, a meaningless notion.

NAhU is used to construct a mekso operator out of a regular Lojban predicate. The current grammar allows a bridi-tail to be used after NAhU.
Allow a selbri only, with no following sumti.
In a context like

li by. na'u broda te'u cy.
the number B # C 

where "#" represents the nonce operator, the elidable terminator "te'u" turns out to be always required. If it is omitted, the "cy." is interpreted as part of the bridi-tail. Reducing the generality of what is permitted makes elidability much more likely.
The original reason for allowing the bridi-tail was that some of the places of the general predicate may be non-numerical, and allowing sumti permits those places to be "plugged up" and not used in the operator. However, the same effect can be achieved by binding any such sumti into the selbri with "'o".

Normally, I, I+BO, and NIhO are allowed only between sentences; for special effects, however, they may also be used at the beginning of text. This initial use is not permitted, however, in portions of text grouped by "tu'e...tu'u". (See change 8, 9, and 14 for related beginning-of-text changes.)
Allow I, I+BO, and NIhO after TUhE.
Increased flexibility. In particular, leading "ni'oni'oni'o..." may be required to set the maximum level of "ni'o" nesting that will be used in the text enclosed by "tu'e...tu'u".

CHANGE 28: (Probably ANNULLED)
The draft textbook had a cmavo "moi" used to attach a relative phrase to a sumti 'modally'. i.e. neither restrictively or non-restrictively. As part of an early cmavo change, "moi" was combine into the non-restrictive "ne" because at the time there was not seen to be any logical distinction between the two. This was an error.
The relative-phrase introducer "ne" is used before a tagged sumti in two different ways: to add incidental information (the non-restrictive equivalent of "pe"), and to attach a new sumti to the bridi, modally associating it with some already existing sumti. Paradigm cases are:

mi nelci la .apasionatas ne fi'e la betoven.
I like the Appassionata, created by Beethoven.


la djan. nelci la betis. ne semau la meris.
John likes Betty more than (he likes) Mary.

respectively. In the former sentence, "ne fi'e la betoven." means no more than "noi la betoven. finti"; in the latter sentence, however, "ne semau la meris." does not mean "noi la meris. se zmadu", since the information is essential to the bridi, not merely incidental. That is, John may like Betty more than Mary, but not really 'like' Betty or Mary at all. In fact, the second example generally means:

le ni la djan. nelci la betis. cu zmadu le ni la djan. nelci la meris.
The amount-of John's liking Betty is-more-than the amount-of John's liking Mary.

The confusion between the two types of "ne" is unacceptably ambiguous. The second type is especially valuable with "semau" and "seme'a", and has seen considerable use, but this use is contrary to the nominal definition of `ne'. [See Greg Higley's article on JOI, elsewhere in this issue, for a discussion that was closely inter-related with this change.]
Assign the cmavo "nau" to the latter use. Since "sumti NAU tag sumti" is really a kind of non-logical connection between sumti, it no longer makes sense to treat it as a relative phrase; this grammar change makes "NAU tag" a kind of non-logical connective, usable between sumti, tanru units, operators, and operands only.
This mechanism only works correctly if a second place is implicitly given the modal or tense tag. For tenses, the second place is the space/time origin; for the comparatives, it is what is being compared; for the causals, it is the effect (and vice versa). But for a tag such as "bau", using the x2 place of "bangu" simply isn't useful.
For most uses of this construction, the right thing to do is to use the actual underlying gismu, which has all the necessary places: recast pure comparisons using "zmadu", "mleca", or "dunli". If you want to simultaneously make positive and comparative claims, use ".esemaubo". To apply tags separately to the two parts of a non-logical connective ("I in Lojban, with you in English, discuss"), use Change 30's non-logical termset connection.
It has been argued that the standard use of "semau" in relative phrases is logically misleading. If we are saying that "John likes Betty more than (he likes) Mary", the essential claim is not "likes"/"nelci" but "zmadu" as stated above, and the main bridi should therefore be "zmadu". This essential logical structure is hidden by the status quo, and to some extent by the proposed change. The counter-argument to this, that natural language usage of comparison warrants an abbreviated form, is logically unsound.
Change 28 will probably not be accepted, and is not incorporated into the published E-BNF, but is being retained here until all interested parties have seen the arguments on all sides.
Clarify that "ne semau" is non-restrictive, not simply comparative. This means that the example Lojban sentence above requires that John like both Betty and Mary, in order for the non-restrictive "ne semau" phrase to be true. By comparison, the English can be used if John likes Betty, but doesn't like Mary.
This clarification requires no grammar change, but substantial reworking of draft textbook lesson 6.

The flag "ma'o" (of selma'o MAhO) is used to convert a letteral string to a mekso operator. It serves to disambiguate uses of "f" or "g" as names of functions from the identical-looking uses of "x" or "y" as names of variables.
Allow any mekso to follow "ma'o". This involves changing the terminator to "te'u", the general mekso terminator.
Some flavors of mathematics (lambda calculus, algebra of functions) blur the distinction between operators and operands. Currently, an operator can be changed into an operand with "ni'ena'u", which transforms the operator into a matching selbri and then the selbri into an operand. The reverse transaction is not readily possible.
There is a potential semantic ambiguity in "ma'o fy. [te'u]" if "fy." is already in use as a variable: it comes to mean "the function whose value is always 'f'". However, mathematicians do not normally use "f" as a normal variable, so this case should not arise in practice.

Termsets are defined with logical connectives only. Forethought non-logical connectives (JOIGIks) are allowed also, but only as a by-product of their grammatical equivalence with GEks.
Explicitly allow afterthought non-logical connectives (JOIks) in termsets.
Sentences like:

nu'i mi bau la lojban nu'u joi do bau la gliban. cu casnu
I in-language Lojban joined-with you in-language English discuss. 

are not possible without termsets. The effect of a non-logically connected termset is to non-logically connect each of the corresponding terms in an inseparably cross-linked way.

Logical connections can be grouped closely (with BO) or loosely (with KE), but non-logical connectives cannot, except in forethought. This is a hangover from Loglan days, when there was only one non-logical connective and grouping was irrelevant.
Allow JOIk+BO between sumti, tanru units, and operands; and JOIk+KE between sumti and operands. We already allow JOIk+KE in tanru and operators, because no cmavo compounding is required.
Completeness: "the set of red-joi-blue and green-joi-black things" can now be done with "cebo" as the middle "and".

Currently, "jai" (selma'o JAI) is used only with a following tag (tense or modal), and causes a modal conversion analogous to the regular conversions expressed with SE. The sumti normally tagged by the modal is shifted into the x1 place, and the regular x1 place is moved to an auxiliary place tagged with "fai" (selma'o FA).
Allow "jai" with no following tag. The semantics is to extract a place from the subordinate bridi within the abstract description normally appearing in the x1 place, and raise it to the x1 level. The abstract description goes to the "fai" place. For example:

le nu mi catra la djim. cu jenca la djein.
the event-of my killing Jim shocks Jane.


mi jai jenca la djein. fai le nu [mi] catra la djim.
I shock Jane by the event-of [my] killing Jim.

Exactly which place is extracted from the subordinate bridi is left vague.
This construction is a sort of sumti-raising; it differs from the "tu'a" type because it marks the selbri rather than the sumti. The whole abstraction is preserved in the "fai" place if it is wanted, and "le jai jenca" can be used to mean "the one who shocks" (where "le jenca" would be "the event which is shocking"). In this case, "jai" is equivalent to "jai gau".
Note that this type of sumti-raising is semantically ambiguous, as is "tu'a" sumti-raising. The natural raised sumti may not always be the actor. In the above example, the bracketed "mi" is implied to be the agent because it is omitted from the abstraction in the "fai" place. If Jim were also omitted from the abstraction:

mi jai jenca la djein. fai le nu catra.
I shock Jane by the event-of killing.

it is not clear whether it is my doing the killing or being the one killed is the event that shocks Jane (ignoring the pragmatics of whether someone who was killed could/would be making such a statement; well-known American essays such as the hypothetical statements by people who have died in traffic accidents after drinking alcohol come to mind). What is known is that the speaker wants to emphasize the role of "mi", whichever role he played in the killing.
If it is necessary to raise from an abstraction which is not in x1, a regular SE conversion following (and therefore inside) the "jai" can be used to get the abstraction to x1:

lo nazbi jai te frica do mi fai leka [lo nazbi ...]
A nose is the difference between you and me.

(exactly what about the nose that is different is quite vague.

A Change to Relative Clause Grammar (Change 20)

[The following is an extract of the discussions that led to the most significant grammar change in the language since mid-1989, long before we baselined the Lojban grammar (that change was the one that incorporated the structures in the Negation paper). Although the relative clause change discussed below is fundamental to a major structure in the language, it is almost invisible to the average Lojbanists: few texts that have been written require changes. It was also taught in passing in less than an hour to beginning students, with no real difficulty.

The extensive discussion, and the serious resistance to even what turned out to be a very low-impact change should stress for Lojbanists the commitment that the design team has to language stability. On the other hand, the outstanding and detailed technical analysis that Colin Fine and others put into this change is both informative of the 'nitty gritty' of this change and its philosophical underpinnings, and of several broader aspects of the Lojban design philosophy, which are mentioned in passing during the discussion. I believe that the result, while technically detailed, should be fairly understandable to relatively novice Lojban students using only the Diagrammed Summary of Lojban Grammar due to the detailed translations that accompany the examples. I also note that Iain and Veijo, when participating in this discussion, had started studying the language only a couple of months before, and hence considered themselves to be beginners at the time they wrote (though their analyses were generally quite correct).

The following is presented in several parts. First comes Greg Higley's paper, actually submitted after the decisions had been made on this issue, but developed independently of Colin's work. Then follows excerpts from Colin's original analytical paper, which we have footnoted with some of the discussion that resulted on each point (edited to make the interaction more evident), Then follow comments from Iain Alexander, Veijo Vilva's (showing his perspective as a non-Indo-European language native speaker), and a few others, which did not fit well as annotation in the two original papers. After all this discussion, Colin responded with a rebuttal. This rebuttal was convincing to Lojbab, who hit upon a satisfactory solution through a rather serendipitous consideration of a lesser change proposed by John Cowan. That proposal constituted Change Proposals 20 and 21. Change 20 as adopted is found earlier in this issue in the summary of grammar changes, while Change 21 was rejected by the community. The optional portions of Change 20 and the whole of Change 21 are included here, as well as some the commentary that led to the final decision.

Quantification and noi

by Greg Higley

A potential problem has come to my attention regarding the quantification of sumti modified by relative bridi. Since this "problem" almost invariably pops up when "noi" is involved, I will discuss it as it relates to "noi" only, and its occurrence with other relative clause cmavo can be inferred. This problem does not seem to occur with "poi".

All sumti that are not explicitly are implicitly quantified. In the following discussion I will deal only with those that are made by the addition of a gadri (article) to a selbri. With all such sumti, whether the quantification is implicit or explicit, there are two "points" of quantification, one (the selected subset) before the gadri and one (the "inner" set - so called because of its position) after it. (I shall henceforth refer to the "inner" set as I and the selected subset as S.)

Put simply, the question/problem is this: In a non-restrictive relative clause, does the cmavo "ke'a" refer to I or to S?[1] If we take the analogy of "poi", it refers directly to I, and thus to S as a subset of I. In the sentence "mi pu viska ci le vo prenu poi ca vave'a litru", "four people were moving around in a medium-sized area a medium distance away, but I saw only three of them". Thus "ke'a" refers to I. If we replace "poi" with "noi" in this example, we get "mi pu viska ci le vo prenu noi ca vave'a litru". For this a colloquial English translation will be helpful: "I saw three of the four people, who were (at the same time) traveling (i.e. moving on/across/via some unspecified surface) a medium distance away in a medium-sized area." Based on the English translation, it is quite impossible to tell, in the absence of context, whether three or four people were "traveling", although it is certainly clear that only three were visible to me. Since of course we cannot take the analogy of English ^ we would be rightly guilty of malglico - we must conclude that "noi" is analogous with "poi" in this respect[2], and that "ke'a" always refers to I in a non-restrictive relative clause.

But here's where we run into a problem. If "noi" and "poi" are analogous in this respect, many Lojbanists, myself included, are making the mistake of assuming that "ke'a" can sometimes refer to S, particularly if S is quantified explicitly and I is not. The examples below will show what I mean:


mi viska ci le vo ninmu noi melbi
I see three of the four women, who are beautiful.


mi viska ci le ninmu noi melbi
I see three of the women, who are beautiful.


mi viska le ninmu noi melbi
I see the woman, who is beautiful.
I see the women, who are beautiful.


mi viska ci ninmu noi melbi
I see three (of the set of all?) women, who are beautiful.

Look carefully at these examples and their colloquial English translations. If "ke'a" always refers to I, then we may run into occasional problems, particularly if we definitely do not want it to refer to I. As for example 4, I would venture to guess that most Lojbanists would not take "ke'a" as referring to all women! But this is the interpretation we must accept if "ke'a" always refers to I. If, on the other hand, "ke'a" always refers to S in noi clauses, we run into the problem from the other end. For this, look at example 1. What if we want to say that "all of the women are, incidentally, beautiful, while I only see three of them"?

One solution to this is to divide "ke'a" into two cmavo. One that refers to I, and another that refers to S. For the following examples, I have assigned the experimental cmavo "xai" the meaning of S-referring relative sumti, and "ke'a" refers to I:


mi viska ci le vo ninmu noi xai melbi
Three women are beautiful (out of the set of four that I happen to have in mind) and the same three are seen.


mi viska ci le vo ninmu noi ke'a melbi
Three are seen and four are beautiful.


mi viska ci le ninmu noi xai melbi
Three women are seen (as always) and the same three are beautiful (out of the set of all that I have in mind).


mi viska ci le ninmu noi ke'a melbi
All of the women are beautiful, and three of the same are seen.

3. (skipped)


mi viska ci ninmu noi xai melbi
Three are seen, and three are beautiful, and we avoid the problem of having to call the whole lot beautiful!


mi viska ci ninmu noi ke'a melbi
Three are seen, and the members of the set of all women are beautiful.

Another possibility has come to my mind, and the grammar may very well specify exactly this, but I'll call it to your attention anyway. What it involves is the quantification of "ke'a" itself. If we allow "ke'a" to refer to all of I, then we can echo the quantification of I or S to show the one to which we are referring, and thus we won't need two cmavo. If this seems rather hazy, the following examples should clear it up:

1. mi viska ci le vo ninmu noi ci ke'a melbi

Here we know that "three of the women are beautiful", because the S quantification is echoed with "ke'a". (Remember that "ke'a" is always quantified as "all of I", so "ci ke'a" means "three of the four", and the rule would state that these three must be S.)

2. mi viska ci le vo ninmu noi ro ke'a melbi

Here "four women are beautiful".

3. mi viska ci le vo ninmu noi ke'a melbi

Here we don't know whether three or four are beautiful, and only context will help us.

4. mi viska ci le vo ninmu noi paboi ci ke'a melbi
"I see three of the four women, of which one of the three (of all four) is beautiful." And this woman is a member of S.
5. mi viska ci le vo ninmu noi paboi ro ke'a melbi
"I see three of the four women, of which one of the four is beautiful." And not necessarily any of S.
6. mi viska ci le vo ninmu noi su'oboi ci ke'a melbi
"I see three of the four women, of whom at least one of the three is beautiful." Etc.

I frankly don't know which one of these systems (two cmavo or one with special quantification rules) will work best, but I am partial to the latter method. Our intuition will still be of great help to us when deciphering relative clauses - as shown by the fact that, so far as I know, no one has noticed this problem before - so it will still often be possible to omit the relative pronoun. One last possibility would be that "noi" clauses always refer to S and "poi" clauses always to I, but that will run into some problems, as you may already see.

What does the baselined grammar say about all this? I'd love to know.


  1. In referring to I, "ke'a" always refers to S as a subset of I. But the question here is whether "ke'a" might ever refer directly to S, thus excluding some members of I.
  2. Since we have no reason to think otherwise. I have never seen a rule of the grammar that specifically states whether "ke'a" refers to I or to S.

Sumti and Relative Clauses

by Colin Fine

I believe there are some hidden problems with the semantics and syntax of relative clauses and quantifiers. In this paper I discuss the problems, and suggest some solutions.

1. Relative clauses

The syntax of relative-clauses is:


: relative_clause_A_111
| relative_clause_110

i.e., a constituent consisting of a left-associative list of individual relative clauses.

I believe this is a faulty analysis. To see where the problem lies, consider a relative clause as a semantic operator: it takes as its argument (the referent of) a sumti - some more or less specified set of entities - and delivers another set (or a sumti which refers to this set - it doesn't matter very much whether we take the operator as acting on sumti or their referents).

In the case of an incidental relative (ne, noi, goi), the membership of the result set is identical to that of the argument set - all we have done is made a subsidiary claim about its members. e.g.

lo sipna
[some of] all sleepers
lo sipna noi melbi
[Some of] all sleepers, by the way, they are beautiful

The problem is in determining which sleepers are beautiful, 'all of them', or just the 'some' that we are talking about in this sentence.

My argument is that if you follow the parse, it means 'all of them', because it parses as (su'o) [lo sipna [noi melbi]] with the (implied) quantifier unequivocally outside the scope of the relative.]

The set of all sleepers is selected by "lo sipna", and unchanged by the incidental relative.

A restrictive relative clause, on the other hand, in general delivers a different set from its argument. e.g.

lo sipna
[some of] all sleepers
lo sipna poi melbi
[some of] all those sleepers who are beautiful.

Clearly each successive restrictive will deliver a further altered set:

lo sipna poi melbi zi'e poi mi prami ke'a 
[some of] {{all those sleepers who are beautiful} whom I love}

and logically we have a left-associative structure in which the relative-clauses is not an independent constituent.

Thus far, I have established that the grouping in the Lojban syntax is logically erroneous; but this might not be very important. The next sections show how it does matter.[1]

2. Mixed relatives

First, note that incidental relatives certainly associate (in fact, commute):

lo sipna noi melbi zi'e noi vasxu
"sleepers, who are beautiful, and who breathe"

does not depend on any grouping, and is even the same (except maybe for some pragmatics) as

lo sipna noi vasxu zi'e noi melbi 
"sleepers who breathe and who are beautiful"

Probably, the same is true for restrictives:

lo sipna poi melbi zi'e poi mi vasxu ke'a
"sleepers who are beautiful and who breathe"

probably always delivers the same set as

lo sipna poi mi vasxu ke'a zi'e poi melbi.  
"sleepers, who breathe, and who are beautiful"

(I am not convinced this is always true).

The first problems appear when we mix the two. Does

lo sipna poi mi vasxu ke'a zi'e noi melbi

mean the same as

lo sipna noi melbi zi'e poi me vasxu ke'a?

As far as I know, the answer is not currently defined in Lojban.

I believe that the first is (or should be) saying "(incidentally) that all the sleepers that I love are beautiful", whereas the second says that "all sleepers are beautiful", even though it is then going on to talk about only "those whom I love".[2]

Though this is a problem, I don't think it is a big one, mainly because the only common occasion for mixing the two has been with "goi":

le prenu goi ko'a zi'e poi mi viska ke'a
le prenu poi mi viska ke'a zi'e goi ko'a
"The people whom I saw, (henceforward x1)"

and even there, the technical difference (whether x1 refers to all people or just the one(s) I see) is often vitiated by the intensionality of "le" as opposed to "lo".

If this were all, we could probably get by with the existing syntax, and adding one of two interpretative rules to the (pu'o) semantics. Either:

"Take the relative clauses in order; each restrictive clause selects some subset from the current set of designated entities and makes that the current set; each incidental clause makes that subsidiary remark about the current set"

or, more simply:

"Take all the restrictive clauses together and apply them to get the final set; then interpret each incidental clause as commenting on that final set"

which is certainly simpler, though very grubby.

3. External quantifiers

Where the problem starts to become bigger is with quantifiers. There are actually two semantically different occurrences of these, which I shall call "external" and "internal". Internal quantifiers are within descriptions, considered below in section 4. External quantifiers occur in rule

    : sumti_E_96 
    | quantifier_300 sumti_E_96

(and also in indefinite sumti, which I will come to below), and I suggest that they are semantically similar to a restrictive clause.

That is to say,

ci lo cukta "three books"

is roughly equivalent to something like

lo cukta poi lu'i roke'a cu cimei
"books such that the set of all of them is a threesome"

(I am not claiming that this is a precise paraphrase, or a transformation; my point is that, like a restrictive clause, the quantifier performs a substantive selection operation on the set of referents).

In fact, external quantifiers do not bind as tightly as restrictive clauses, so a phrase like

ci lo sipna poi melbi


three of (those sleepers who are beautiful),

and the current parse

ci [[lo sipna] [poi melbi]]

corresponds with this interpretation.[3]

But if we then introduce incidental relatives, the current syntax does not give the right answer.


ci lo sipna noi melbi

currently parses as

ci [lo sipna noi melbi]
three out of [all sleepers, who incidentally are all beautiful]

but I believe that almost all seljbo would interpret it as

[ci [lo sipna]] [noi melbi] 
[three out of all sleepers], who are beautiful.

Similarly with quantifiers and both types of relative:

ci lo sipna goi ko'a zi'e poi mi nelci ke'a

The current syntax makes this

ci [lo sipna [goi ko'a zi'e poi mi nelci ke'a]]

i.e. ko'a is either "all sleepers", or "all the sleepers I like", but in no way just three of them.

In summary, incidental relatives belong outside the external quantifier, but restrictive ones belong inside.

4. Internal quantifiers

When we look inside a description we get a different kind of quantifier, with different properties:

le ci sipna
the three sleepers

It seems to me that this is semantically an incidental rather than a restrictive construction.[4]

As I understand it

lo vo prenu

makes the subsidiary claim that there are only four persons, which is an incidental claim to the description, and not a restriction.

This does not give any problem with explicit incidental clauses:

lo mo'a temci noi sutra simlu
the too-few time intervals (that seem fast)

but the interaction with explicit restrictives is wrong:

lo ci sipna poi mi nelci ke'a

is at present unequivocally

[lo ci sipna] [poi mi nelci ke'a]
those among [the three sleepers] whom I like

whereas what it should mean is

lo ci [sipna poi mi nelci ke'a]

i.e. the sleepers that I like, of whom there are in fact three.[5]

So as with external quantifiers, incidental relatives belong outside, but restrictive ones belong inside.

5. Indefinite sumti

(pe'i these are an annoying mistake, complicating the syntax just so that we can omit a word here there and thereby muddy the logical structure. However, we have them and we can cope.)[6]

Transformationally, as I understand it

<quantifier> <selbri>
e.g. ze prenu

is exactly equivalent to

<quantifier> lo <selbri> 
ze lo prenu

and we have precisely the same difficulties as with any other external quantifier, except that the <quantifier> and the optional <relative clauses> are introduced at the same point in the syntax (indefinite_sumti_94), so for example

ze prenu poi gleki

parses as

[ze prenu [poi gleki]]

with three constituents, and not explicitly as

[ze [prenu [poi gleki]]]

in the way

ze lo prenu poi gleki

does. i.e. the syntax is equivocal here.

6. Preposed possessives

The other anomaly in the current grammar is the preposed possessive (the optional sumti_E_96 in sumti_tail_113):

le mi cukta

This is precisely equivalent to

le cukta pe mi

This does not interact problematically with relative clauses, of either type:

lo mi cukta poi xunre
= lo cukta pe mi zi'e poi xunre

restricts the set of books to those which are both mine and red.

lo mi cukta noi xunre 
= lo cukta pe mi zi'e noi xunre

restricts the set to "books which are mine", and comments that they (my books) "are-red".

But it does not work at all with internal quantifiers.

lomi ci cukta

which is always used to mean

'my three books', i.e.
'all books, restricted to those belong to me, there are three of these' 
(= lo ci [cukta pe mi])

is actually defined to be

lo mi [ci cukta] = [lo ci cukta] pe mi 
'my books, out of the three' , 
'all books (there are three), restricted to those which belong to me'[7]


*lo ci mi cukta

which has some hope of meaning what we want, is not even valid![8]

(It is true that these forms with 'lo' are relatively unusual, and it is more common to use 'le', which once again gets round the logical problems by pragmatics; but I think the problems are there nonetheless.)

7. Summary of the problems

There are two basic problems, one of them in two parts.

1a. restrictive relatives belong inside external quantifiers, incidental relatives belong outside.

1b. restrictive relatives belong inside internal quantifiers, incidental relatives belong outside.[9]

2. preposed possessives belong inside internal quantifiers.[10]

8. Suggestions for problem 1

There are a number of possibilities I can think of.

a) Nothing.

Thus far, we have found this area to be workable. However, wait until you try to teach the semantics to a computer. This will require rules something like the following:
Quantified sumti: Store the quantifier, then go ahead and interpret the sumti including any relative clauses. Then select the specified number from the set of denoted items. If there are any incidental clauses stored, now apply them.
Internal quantifier: Store the quantifier, and go ahead and interpret the selbri, and carry the set of denoted items forward.
Relative clauses: Interpret each clause in turn. If it is a restrictive, select appropriately from the current set of denoted items. If it is an incidental, remember it.
At the end of the relative clauses, if there is an internal quantifier stored, use it to select an appropriate number from the set. Then carry the set forward.

Possible, but hideous, and not worthy of something described as a logical language. (And preposed possessives will give a further complication).

b) The minimal change I can see is to require all restrictives to precede all incidentals, and modify the grammar as follows to reinterpret almost what we have ... [detailed proposal omitted, since it was rejected]

I believe this will produce just the same surface strings as we have at present, except that all incidentals will have to follow all restrictives.[11]
I and GOI have to be split, and that ZIhE performs some very strange functions).

The only thing in favour of this suggestion is that it does the minimum damage to existing texts. It complicates the syntax remarkably and - in the name of compatibility - confusingly.

c) My preference is to introduce three specific locations for relatives, thus so'a lo panono cukta poi mi nelci ku poi dopa'a nelci ku'o noi cfika would parse as

{[so'a {<lo panono {cukta poi mi nelci} ku> poi dopa'a nelci ku'o]}] [noi cfika]}
[almost all of those of <the hundred {books I like}> that you also like] which incidentally are fiction...

... [Colin's detailed proposal eliminated][12]

d) I considered a solution with arbitrarily nested scopes, each of which was limited by a quantifier and/or restrictives, and each of which could have an incidental attached to it, thus:

 { <  [so'i
       { <lo tarci
       > poi se viska tu'a le
terdi ku'o
      } noi melbi ku'o
   > poi mi di'i catlu ke'a ku'o
 } no'u la ze mensi

but this requires a much more complicated grammar, and I think it can be managed by structures already existing at a higher level (KE or LUhI). At any rate, I did not investigate its syntax carefully.

I think (c) is the best solution. It does not do a lot of injury to existing texts: as long as they don't mix restrictive and incidental clauses, they will still parse; if they do, the two sets need to be sorted out, and the first (restrictive) set ended by a KUhO/GEhU (or by a KU if there is a description). And the scoping will make sense.

Note that something like

le ci cukta poi mi nelci

will parse as

le [ci [cukta [poi mi nelci]]]

but you can force the restriction outside by

[le [ci cukta ku] [poi mi nelci]]

which I claim is selecting "those I like" from among "the three books".[13]

9. Suggestions for problem 2

[Omitted - there was no consensus that Colin's #2 was a problem.]

10. Conclusion

I have presented at length some logical problems in our current sumti grammar, and made some suggestions:

  1. Withdraw the "<quantifier><quantifier><selbri>" form of indefinite sumti[14]
  2. Distinguish restrictive from incidental clauses, and define three distinct places where they may occur: incidental ones only outside quantified arguments, restrictive ones both inside external quantifiers, and inside internal quantifiers in descriptions.
  3. Reverse the order of the internal quantifier and the preposed possessive in descriptions. The three suggestions are all independent of one another.

I have not looked at vocatives: since they do not include quantifiers, they do not really have a problem, though for consistency they should be changed consistently with any changes to solve problem 1.[15]

Commentary from: Iain Alexander

First of all, let me point out that the latest Diagrammed Grammar Summary appears to support one of your proposals. At the bottom of page 19, it describes a "description sumti" as

[number] le [number] [sumti] [modal] selbri [ku]

which is your solution (c) to problem 2.[16]

In general, however: there is no rule that says that the deep(er) structure of a language (natural, artificial, computer, whatever) has to correspond to the surface structure. (This is obvious, isn't it.)

On the other hand, it is kind of nice if it does, particularly if it's easy. This is particularly true when you've got people like myself who have access to the grammar definition, which gives the syntax, but tells you essentially nothing about the semantics of any given construction.[17]

Some of it we intuit from the corresponding English language construction - we are after all still a predominantly English-language group - but this is in itself dangerous. Many of the discussions I've seen or been involved in recently (and some I've never started, because I saw what was going on in time) have been a result of confusing an English gloss for a Lojban definition - mainly of gismu rather than grammar rules, but then there are more of the former.

There's a lot of stuff in the language which needs careful definition, which is a lot of work, and it's not even obvious how you can best present some of it.

In any case, I think I'm saying that although your concerns are theoretically unimportant, in practical terms they are extremely reasonable, and I am in favour of any such rationalisation which makes it easier to get to grips with the grammar - I would need to read it all through again before committing myself to any of your particular solutions. But this is coupled with a warning that much of the grammar, possibly even including this part after your improvements, needs semantic clarification, and we as a group need to find some way of handling this.[18]

Commentary from: Veijo Vilva

My initial reaction to Colin's paper was to agree with him but Iain's cautionary note about Anglocentrism sent me thinking (as the only non-Indo-European in this group).

I thought of Colin's example sentences and their close relatives in view of the current Finnish pragmatics and after a while I wasn't too happy anymore. The original parses also seemed quite necessary and changing the parsing would have necessitated the introduction of new alternate ways of similar simplicity to express the original 'grammatical' meaning.

... [much of Veijo's commentary is deleted as it supports options that were rejected. Among these was further elaboration of the concept that "zi'e"-joined relative clauses were nested, which was an erroneous assumption on the part of both Colin and Veijo. The apparent demand for nested relative clauses led to change proposal 21, but support for nested clauses did not persist, and change 21 was annulled.]

In general I find that properly combining le/lo, internal/external quantifiers and restrictive/incidental relatives gives about all the semantic variants I might want. It may take some juggling at the natural language level to get just the wording you are accustomed with - but often finding the proper wording to express just the shade of meaning you are after in a natural language expression in general may be more difficult and even beyond the capabilities of most people.

I think we ought to get away from translating and to start taking Lojban as is. It has it's own ways of expressing ideas and it is very important to avoid imposing an alien strait jacket upon it.

My approach was based firstly on the fact that I am, as a newcomer, still struggling to express ideas and to understand ideas expressed and secondly, after all, this interplay of expression and understanding is what a language is all about.

Lojban is an emerging language which still is in a state of flux in many respects. We have a relatively limited corpus of existing text which is at least partly outdated. Some of this text has been created by people at a relatively early stage in their development as Lojbanists and may contain usages which necessarily haven't been so thoroughly analyzed at the time of writing but may have been 'instinctive' choices reflecting more the linguistic background of the writer than the grammar of Lojban. When I spoke of translating above I didn't mean that to be taken quite literally. What I mean is that when we are dealing with a completely different language like Lojban we mustn't always expect to see things expressed in an 'instinctive' way. We have a grammar which defines the framework within which we are trying to express ideas and before we modify it I think we must see whether it is possible to express the ideas we might want to express - even in an 'alien' way. After that we must make a choice: do we accept the 'alien' way or do we modify the grammar. I think that at this stage we still have the option of specifying the way various things are expressed.

4. Internal quantifiers

On this issue, I would use the following structure allowed by the present grammar:

le ci [le sipna poi mi nelci ke'a]
The three of [the sleepers that I like].
i.e. the sleepers that I like, of whom there are in fact three.  
(Produced from: "LE_562 [quantifier_300 sumti_90] gap_450) [sumti_tail_113]"

The meaning is quite obvious - in fact it matches exactly the first English gloss.

This produces a kind of intermediate quantification - it is internal in the total structure but external to the restrictive relative clause. The only blemish I can see is that it is occasionally necessary to use a double KU terminator.

Colin's example for solution (c):

*so'a lo panono cukta poi mi nelci ku poi dopa'a nelci ku'o noi cfika

would become:

so'a lo panono le cukta poi mi nelci ku poi dopa'a nelci ku'o noi cfika

Neither is a candidate for casual conversation but I prefer the latter one (conforming with the present syntax).

And the ones in the discussion:

Colin's proposal               present grammar
le ci cukta poi mi nelci    => le ci le cukta [ku] poi mi nelci 
le ci cukta ku poi mi nelci => le ci cukta [ku] poi mi nelci

To me the present way is in this case more obvious.[19]

6. Preposed possessives

By the way, the last production in the present definition allows constructs like:

le paboi ciboi ze cukta so the indefinite sumti cause trouble also here. Perhaps we ought to prune them off totally as the easiest solution?[20]

Response from John Cowan:

I believe that Colin's main error lies in ignoring the uses of relative clauses with non-description sumti. If anything, the use of relative clauses with da-series variables is even more important. Colin's proposal to separate incidental and restrictive clauses, and to place the latter within the scope of "le...ku", does nothing for "da poi" constructions.

Colin rebuts: It is true that I did not specifically discuss relative clauses with non-descriptive sumti; however I did not ignore them:
My contention is that as a matter of current fact we interpret relative clauses with non-descriptions as (necessarily) outside the sumti (but inside the (external) quantifier), whereas we interpret relative clauses with descriptions as inside the sumti and the internal quantifier. (I am ignoring incidentals here, which are currently a problem, as I explained).
ci da poi sipna
ci [da poi sipna]
three (out of) (those x who are sleepers)
lo ci prenu poi sipna
lo [ci [prenu poi sipna]]
some ((persons who are sleepers) (incidentally there are three))
but our existing parse matches in the first case, but not the second. My suggestion 1(c) is to change the syntax so that these two currently valid sumti will still both be valid, but will parse reflecting the semantics I have given.[21]
Thus my proposal is not 'to separate incidental and restrictive clauses, and to place the latter within the scope of "le ... ku"'. It is to separate incidental and restrictive clauses, and to define two different places of attachment for the latter: one within descriptions and the other outside the sumti-4.[22]
All existing strings that do not involve incidentals will remain valid, but they will parse differently according as there is a description or not. As an extra, it will be possible to place the restrictive string outside the description explicitly (and therefore outside the internal quantifier) by using "KU".[23]

[Cowan continues:] I also believe that the notion of "restrictive relative clause" is far more semantically deep than can be reasonably addressed by mere syntactic manipulations, requiring its own semantic processing module.

First, it seems clear (and Colin implicitly recognizes) that all talk of relative clauses and phrases can be reduced to "noi" and "poi" only. The alternatives are "voi" clauses (which Colin ignores) and relative phrases with "ne", "pe", '"ne" + BAI', '"pe" + BAI', "po", "po'e", "no'u", and "po'u". All of these may be reduced as follows: voi -> poi mi skicu fo da poi

ne -> noi srana
pe -> poi srana 
ne + BAI -> poi BRIVLA [where BRIVLA is the source of BAI] 
pe + BAI -> noi BRIVLA [ditto] 
po -> poi steci 
po'e -> poi se ponse [with additional connotation of inalienability] 
no'u -> noi du 
po'u -> poi du

These transformations are not necessarily claimed to be exact or to work in all cases, but they indicate the basic mechanism involved.

I suspect, that the current attachment point of "relative-clauses" is too far down in the sumti hierarchy: the fact that it appears twice is ipso facto suspicious. I will make an attempt to do the necessary YACCing to determine if the connection point can be moved up closer to, or within, "sumti-3<93>". External quantifiers should be processed either before or on the same level as relative clauses.[24]

[Colin rebuts]: Obviously, I don't agree that "relative-clauses" is too far down in the hierarchy - it is both too far down and not low enough.
Incidentally, the fact that it appears twice is purely a requirement of mabla indefinite descriptions.[25]

... Further, while I would be keen to have a transformational description of the language, I would vastly prefer one limited to transformations within the syntactic structure, not just of surface strings; i.e. that did not allow shifts into or out of constituents, as this would require.


  1. Lojbab and John Cowan note:
    "zi'e" is a degenerate logical connective (reduced from a large set of connectives in Change 7, decribed above), a sumti with two relative clauses, restrictive or non-restrictive, or both, is applying both relatives simultaneously. By the principles of Lojban logical connectives, Colin's example must be interpreted as
    lo sipna poi ge ke'a melbi gi mi prami ke'a
    [some of] {all those sleepers who both are beautiful and whom I love}

    Thus multiple restrictions are not 'successive' restrictions, but in effect tantamount to a logical AND on the restrictions.

    Whether there should be a successive restriction capability, is arguable.

    A key point about Lojban grammar, especially where 'grouping' is concerned, is that the groupings produced by the parser go beyond what is needed to resolve the grammar, and impose a structure that is not necessarily there. Thus the 'left-grouping'- ness of relative modifiers is an artifact of LALR1 grammar that exists because you cannot have multiple relative clauses without some grouping - the grouping is not intended to have implication for semantics.

    Here is where reasoning from "da poi ..." comes into play. Restrictive clauses have a deep effect on "da"; they do not simply say that in addition to fitting into its existing bridi "da" must also fit into another bridi; instead, the meaning of "da" is changed from "some object" to "some object chosen from the universe specified by the 'poi'". This is shown by the fact that "da" thereafter has a meaning incorporating the restriction: it is not local to the current sumti, but is pervasive until another "da poi" appears.

    By similar reasoning, "lo mi ci sipna", which means "lo ci sipna [ku] pe mi" exactly, and is roughly equivalent to "lo ci sipna [ku] poi [ke'a] srana mi", asserts that "the number of sleepers is three" within the domain "things associated with me", as opposed to "lo ci sipna" by itself, which claims that "there are three sleepers within the general (unrestricted) domain". (In either case, the quantification claim is incidental.)

    Once this domain restriction has been done, the meaning of the sumti can be evaluated. At this time, the incidental clause can be understood as applying to the sumti in its entirety, and making a subordinate bridi (possibly compound) which is incidentally asserted. Note that this analysis implies that "ke'a" means different things within restrictive and incidental clauses: in a restrictive clause, it refers to the meaning the sumti would have if no restriction were in effect; in an incidental clause, it refers to the sumti as-is with any restriction in effect. Therefore,

    ro da poi mlatu cu mabru
    all things which-are cats are-mammals

    has an utterly different meaning from

    ro da noi mlatu cu mabru 
    all things (which incidentally are cats) are mammals

    which says that "everything is a mammal", and what's more, "everything is a cat, too".

    Colin rebuts: Your explanation of the effect of "da poi" is very clear, and more succinct than my own. We are in complete agreement. Further, your discussion of "ke'a" exactly demonstrates my point: that logically restrictive and incidental clauses belong at different places in the parse.

    Lojbab: It appears that Colin is arguing that because a word has different semantics in the two different constructs, the two constructs must have a different syntax. There are numerous cases to the contrary in the language, as for example the fact that "da" has completely different semantics than most any other member of KOhA, while all members of KOhA are considered syntactically equivalent (indeed, this consideration has led to useful and serendipitous realizations, as for the use of prenex non-definite-sumti for topic construction, and the use of prenex bu'a-series, which is especially anomolous in semantics, for 2nd order predicate logic with no special grammar needing to be defined.

  2. Lojbab: The two are defined to mean the same, though I'll agree that it isn't written in any of our published materials.
    Order in Lojban does not necessarily imply succession. The obvious example being NA negation, which does not affect quantifiers in this left-to-right succession fashion in the way that English negation does. Similarly, stated order of sumti does not imply any particular importance.
  3. Lojbab: As I said above, the parse within the sumti may be implying more structure than is semantically significant. In a restrictive relative clause it does not matter whether "lo sipna" or "ci lo sipna" is restricted by "poi melbi"; you still get the same result. Thus it remains arguable.
    I think that the problem almost goes away, by stating that you can interpret all relative clauses to be 'outer'; then, if you want them to apply to an 'inner set' you do so by sticking another descriptor outside: "le <ci lo sipna poi melbi> ku". But you cannot do this when you leave the quantifiers implicit: "*le [su'o] lo [ro] sipna poi melbi ku". This has convinced me there is a problem to fix.
    Note that the classic Loglan descriptor is "le", and not "lo". Colin tacitly agrees that this intensional descriptor doesn't really suffer from these problems (a "le" description means precisely what I want it to mean). The only reason this issue surfaces at all is for "lo" with its default outside quantifier that is "su'opa" and not "ro". This was a change from old Loglan, which set the default of the equivalent to "lo" (lea) at the equivalent of "ro", making it only useful for universal claims. Nora's first reaction to this whole problem was thus - if you have problems, just use "le".
  4. Lojbab: Incidental in the case of "lo", identifying in the case of "le".
  5. Lojbab: This grouping is bogus, since a restriction cannot apply until after there is a description - a sumti - whereas you have marked it to apply to a selbri. The "ke'a" can stand for nothing until you have identified that a single place of sipna is being used as the description, and this takes the descriptor.
  6. Lojbab: JCB spent 25 years waxing wishy-washy on indefinite constructions, agreeing to eliminate them because they caused problems in the grammar, but finally deciding that they were too natural for him and other Loglanists who actually used the language. So he said "make the grammar fit it", and they did, and it remains so. This is of course what used to be called the "se sorme" ("seven sisters" in older versions of Loglan) question.
    "lo" inherently muddies the logical waters, and logical purists would prefer either that you use "da poi" or "le", and skip "lo". Indefinite sumti are no muddier than the rest of "lo".
  7. Lojbab: Fallacious. The "lo" and the "mi" cannot be semantically separated from the "ci" by an artifactual bracket. Especially since you have identified the "ci" as also being equivalent to a relative clause, you should make all transforms of a type at once if you wish consistency:
    "lo mi ci cukta" = "lo cukta pe mi zi'e noi cimei
  8. Lojbab: If it were valid, the "ci" would quantify "mi", which is why it is forbidden. There is no way to make the grammar work with quantifier before the pseudo-possessive, unless we choose to eliminate the established [LE + quantifier + sumti] construct which has existed historically in the language and is more important.
    "lo mi ci cukta" is defined to transforms into "[lo ci cukta] pe mi". It cannot transform as Colin wishes. Since relativization is inherently a function of a sumti and not a bridi (or a selbri), "lo ci [cukta pe mi]" makes no sense, since "cukta pe mi" makes no sense, is not grammatical, and shouldn't be. The structure of a sumti does not group that way. In this sense, the E-BNF grammar makes more sense than the YACC grammar. The essentials are the "lo" and the "cukta" in the description - the quantifiers, pseudo-posessives, and relative clauses are all optional. But they are all at the same level, not grouped more tightly just because there are brackets present; the brackets are an artifact of the way it is easiest to write YACC grammars and should NOT be assumed to have semantic import.
    "pe mi" must be a restriction on "lo cukta", and the only consideration is the relevance of the inside quantifier "ci" (and any outside quantifiers too perhaps). My initial guess is that the inside quantifier might indeed transform to another relative clause, which is incidental:
    lo ci cukta
    = lo cukta noi cimei 
    lo ci cukta pe mi 
    = lo cukta pe mi zi'e noi cimei or lo cukta noi cimei zi'e pe mi

    In any case, I think it is historically clear that the outside quantifier on "lo" exists as a selection from the well-defined sumti that exists without the quantifier present: "ci lo cukta pe mi" is "ci [lo cukta pe mi]; however the bracketed text is parsed internally - it selects 3 out of that inner-sumti.

    Since this answer is different for restrictive and non-restrictive clauses (which traditionally have been interpreted to apply to the set after the quantifier is attached) convinces me that Colin is right.

    Where there is some grounds for argument is that the quantifiers on a descriptor should be bound into the descriptor since the descriptor expands to some default quantifier if they aren't present. I'm not sure what I think of

    ci lo mi vo cukta pe broda 

    parsing as

    <([ci] lo [mi] [vo]) cukta [KU]) pe broda [GEhU]>

    which loses the semantics of selection implicit in the outer quantifier, but perhaps answers Colin's objection on other grounds. I think that this would reflect the semantic expansion of "lo" better than the current grammar, but lose the selection implication. (I couldn't make a YACC grammar work with this approach to sumti descriptions, so it is a non-solution).

  9. Lojbab: I would instead label the problem as being that the grammar is vague as to whether relative clauses apply to inside sets or outside sets, and that this is probably logically unacceptable.
    Given that the grammar needs to be changed to permit inside and outside sets to have distinctive relative clauses, I present the following example, using Colin's syntax, where both types (restrictive and incidental) are used constructively both inside and outside. A bit contrived but plausible.
    mi cuxna ci lo xa cukta poi mi nelrai zi'e goi ko'a ku poi cfika zi'e ne semau leko'a ci drata
    I choose (the) 3 of the 6 books that I most-like (the 6 being ko'a) which are fiction [over] their (the 6's) three others.

    I'm sure it is clear that restrictive clauses can apply to either set; the point of the above example is that, at least for "goi" assignment, non-restrictive clauses can be used on the inside set. That pragmatics can lead to either interpretation of either type of relative that makes me see this as a true problem worthy of the degree of change needed even at this late date. Otherwise the late date would cause me to consider this merely a semantic interpretation problem, rather than an ambiguity problem; logical ambiguities must be fixed in a logical language, while semantic questions can be left for pragmatic usage to decide.

  10. Lojbab: I don't agree, since I consider quantifiers and possessives to be at the same level - they both relate to the inside set, and there is only one such inside set that has meaning for a simple description.
  11. Lojbab: I do not see that you have made any case for requiring any particular ordering of relative clauses, since ZIhEks imply no ordering that can be interpreted as erroneous. I've also devised examples wherein both restrictive and non-restrictive clauses could be applied to either inner or outer sets. Restricting what can be said in order to the more common usages seems too extreme a solution.
    The ordering is also not pragmatically acceptable because people want the incidental "goi ko'a" (which might be intended to apply to either inside or outside quantified sets) to be as close as possible to the description that it marks. A restrictive relative clause can be quite lengthy, and if it also has complications and relative clauses within it, the incidental information becomes worthless. Thus
    le ci broda poi brode da de di ku'o goi ko'a

    is dispreferred, in favor of

    le ci broda goi ko'a poi brode da de di [ku'o]
  12. Lojbab: I do not see the essential difference between the 1st and 2nd of the three relative clause positions in your example, and believe that the image of their difference is due to the fallacious "*cukta poi mi nelci" which is ungrammatical and meaningless. You want "[books (that I like AND that you also like)] of which there are 100", with the incidental clause applying to the outer set.
    However, even reducing this to 2 clause positions, one inside the KU and one outside, would at first glance mean that the KU is no longer elidable when you wish to put an outside relative clause. This may be ameliorated by your distinction between incidentals and restrictives, but I think that distinction is pragmatic - what is most often wanted - not what is plausible in language use. For example, in your just previous example, what if you wanted to merely restrict the "100 books" to "those I like", but note incidentally that "you also like them", before noting that "most of them are fiction". Or perhaps you want to restrict the set to "most of them that are fiction", associating with the outside quantifier.
  13. Lojbab: Since restrictive clauses are often outside, this has the effect of requiring the KU terminator much more often than it has been. We've worked quite hard to make Lojban not be a "ku-ku" language, as older versions of Loglan tend to be when expressed at the natural level of sentence complexity that we have found typical of Lojban usage. [Note: the proposal actually adopted does have this characteristic of requiring more KU terminators.
  14. Colin's argument on this was deleted from the article text, since it really is a separate issue. This construct is an artifact of the older versions of the grammar - something that was permitted, ended up being used rarely (mostly by people here in DC), and therefore preserved. Since there was little justification for its existence and it was difficult to preserve under the changes proposed to resolve the other problems Colin identified, it is passing from the language without much fuss.
  15. Lojbab: Almost. Vocatives can have quantifiers, but only in the context of the sumti_90 internal grammar, and hence are taken care of by whatever we do for the latter. However, the solution proposed requires some changes to the vocative grammar, though consistent with the other changes being made.
  16. Lojbab: This is a typo. Correct is
    [number] le [sumti] [number] [modal] selbri [ku]

    This basic structure becomes incomplete because it doesn't include the preposed relative clause. However at the point in the Grammar Summary in which it is presented, relative clauses have not yet been covered. At some point this will need to be corrected.

  17. Lojbab: JCB had a strong policy on the machine grammar matching the 'human grammar' as closely as possible, to the point of starting fights about it and putting in kludges in the grammar to make it work rather than accepting even small changes in what he saw as the human grammar. This is similar, if not identical to 'deep structure' matching 'surface structure', and our policy has been to preserve this as much as possible.
  18. Lojbab: Such clarification is desirable where possible - there is little likelihood of overdefining the language, but it shouldn't be necessary. Lojban is already by far the most thoroughly defined language there is. I don't expect that there will need to be that much more depth provided at this point than what will be in the final set of papers being written by John after the spirit of Imaginary Journeys [the JL16 tense supplement.
    Colin responds: Iain and Bob rightly point out that it is not essential that the deep structure/semantic parse follow the surface structure; but it is highly desirable. I also believe that getting this sort of disparity straightened out is a valuable step in the process of understanding what we mean and what we are skating over in learning and talking Lojban - for me probably the foremost attraction of the language.
    Lojbab: Agreed, and most of the little changes that have been accepted by John Cowan and me since the last baseline have been little cleanups that arose from the writing of his papers like Imaginary Journeys, which exercise I see as essentially dealing with this step in defining the language. The main question, apparently now resolved in favor of change, is whether the degree of change necessary is warranted by the level of confusion possible. This is by far the biggest change in the grammar since the MEX change before the first baseline, which affected a then-unused part of the grammar, or the even earlier negation-paper and abstraction clause lenu[ke] changes while the first Lojban class was being taught, which did affect usage. It thus takes enormous justification at this late stage for a change of such magnitude.
  19. Lojbab: Veijo's approach was also Nora's initially-proposed response. A limitation is that it doesn't allow you to leave the outer quantifier unspecified, if it is default. However, it does seem clear that, if we left the language unchanged, it is still possible using this construct, if not always convenient, to express anything you need to in the language.
    However, it seems silly to require the extra "le" and the explicit quantifier to force a relative clause inside; also, the current grammar does suggest to some like Colin that the default clause is already inside.
  20. Lojbab: It may be an easy solution but not an acceptable one, since it removes a significant expressive form of the language, and indeed one of historical import with a clear JCB pronouncement. In this time of baselined grammars, that is three strikes against a deletion supported primarily by the argument that the grammar generates messy, probably useless, strings. As long as the strings are not syntactically ambiguous, we can tolerate them, though logical ambiguity also warrants consideration, as indicated by our current discussion. But no one has clearly claimed indefinites are themselves logically ambiguous, only that they make certain aspects of the grammar messy.
  21. Lojbab: Current interpretation is [ci da] [poi sipna] - the sumti has an implicit claim that there are exactly ci who are sleepers. Except that in a second usage after such a definition but within its scope, say "re da poi prenu", the implication is [re le <ci da (still poi sipna)>] [poi prenu].
  22. Lojbab: The second half of this and is what this restatement caused me to recognize was the fundamental problem. Hence this clarification was well-timed, Colin. The latter, though actually more problematical for the rare incidental that can go in either scope, was convincing to me. I remain unconvinced of the need to grammatically distinguish incidental and restrictive clauses.
  23. Lojbab: But since this is often the desired expression, it makes KU less elidable.
  24. Lojbab: Moving relative clauses up to a higher level affects the rules for interaction with "la'e". Both "la'e" and relative clauses are things we'd like to move up, but which can cause problems because of being open-ended.
    That is "la'e lo prenu poi nanmu" is ambiguous without a terminator for the "la'e" construct, unless "la'e" grabs constructs above the rule for relative clause attachment but is itself also above that rule. It is possible that grammar flexibility could be increased with an elidable terminator for "la'e". ... [such a terminator was indeed added as part of the solution].
  25. Lojbab: The reason for repeated occurance of relative clauses in the rules is indeed the perversity of indefinites, which interact badly with virtually everything. They are a blotch on the grammar, and this has been recognized for ages. JCB even agreed to remove them for years because of this, but they kept creeping back into his and others' usages, and he finally said that they were obviously intuitively a part of the grammar, hence the grammar must be made to fit them, regardless of how ugly it was. But in order to put them at a proper place, they in effect need a parallel set of rule structures from the standard sumti structures. Thus, if you look at the baseline grammar sumti rules as a forked tree just above the indefinite rule, you will probably see that the relative clause rule appears just once in each fork, at about the same level. Only by moving the indefinite fork further down (which restricts what you can indefinitize), can you eliminate the problem of which rule goes highest.
    In the adopted solution, we did so, by putting an elidable terminator on LAhE and NAhE-BO constructs, the LUhU of grouped sumti, and putting them all in the same grammar rule. Thus, except its use in MEX (modified to be consistent with the sumti grammar usage), in effect LAhE is the same selma'o as LUhI and the lexer-constructed selma'o NAhE_BO, though its actual usage is very distinctive.

Lojbab's solution

The relative clause change I proposed in response to Colin's paper and ensuing discussion will be found as Changes 20 and 21 in the proposed baseline changes. (Change 21 was annulled, and is not included in the list above, but may be found below, along with parts of Change 20 that were deleted as a result of discussion.)

A change of this magnitude is very controversial. Cowan and I were originally opposed to any change, primarily on the basis that the language design is too firmly baselined to permit such a degree of fiddling as was necessary, and the possible unforeseen side effects of this change are enormous. We were for the most part unconvinced until a late stage that the logical problem Colin was talking about was indeed serious enough to warrant the type of change we believed was needed to solve it, one that might render much existing Lojban texts incorrect.

An earlier major proposal like this had Colin and a few others basically arguing that if the language has an irregularity, it is still permissible to change it because not all that many have learned the language to a point where it would hurt them to relearn. In that case, Nick sided against the proposal. Nora did also, seeing herself as guardian of language stability, since she knows how many people were driven off by similar attempts to stick one more necessary improvement after another in old Loglan in the 1970s and early 1980s. The cost of continued change is not only relearning, but a reluctance of new people to try to learn a language that they might have to relearn.

On the other side of the fence is someone like John Hodges, who, while opposed to unnecessarily fiddling with the language in general, sees that Lojban's main hope as a language in the future depends on its logical integrity, and flaws in that integrity must be resolved even if it costs significant relearning for those of us already studying the language. (A limitation on this position is that, for most of the logical issues that have faced Lojban in the last few years, formal logic gives no clear and single answer. Different schools of thought on logic solve the problems differently. Thus, Lojban research has had to forge its own school of logical thought based on what is necessary to make the language self-consistent.)

John Cowan, who has frequently proposed minor changes in the last three years, almost all of which were adopted, has come to understand the third aspect of the problem: if the language is ever to be documented, it must stop changing. The mere existence and serious consideration of this proposal stopped his work on the sumti paper, and its adoption forced a totally redesign of that paper, not to mention changes to a lot of documentation already completed. Similar changes in the future pose equally drastic threats to already completed and in progress documentation efforts. If the language isn't documented; no one can learn it.

A proposal of this magnitude serious affects on-going learning and teaching efforts. At the time of this proposal, it affected the then ongoing DC class - I had to decide which version of relative clauses to teach within a week, since relative clauses was indeed the topic of the week.

These comments are thus set forth as a warning - that while we want to make the language right and it is worthwhile finding such problems, proposals alike this are stressful to the project, the design team that is trying to finish the project, the language and the community, and thus are decidedly unwelcome. This doesn't mean that questions should not be raised - I hope people will do so, but the expectation must be that most such problems as are identified from here on out will be merely documented as problems, with no change to the language.

[Colin and others reassured me immediately in response to the above that there were no other pending major issues, and indeed, none have been raised in the 10 months since this discussion].

Colin's last rebuttal on the issue finally convinced both me and Nora that the problem required fixing. Cowan remained less than convinced that a change of this magnitude at this late date was tolerable even if the problem is real, but went along with the consensus. Nora's priority in this issue is to minimize the effect on existing text and documentation and this led to a complication in the proposal. All three of us were fairly certain that Colin's solution, which is to separate the grammars of restrictive and incidental clauses, is not the right solution, and also results in too much complication to grammar, documentation, and teaching.

My solution instead attempted to see the problem as a restriction in what can be said in the language, specifically in where a relative clause may/must be attached. Indeed, my solution is mostly an unexpected side benefit of trying to add preposed inside relative clauses as a way around the oft-occurring problem of the invalid sumti form "*<le [ci mi] broda ku>" that mucked up my attempts to understand what Colin was arguing during the above discussion (that text parses as a complete sentence: "<le ci mi> [cu] broda" and the "ku" is therefore invalid).

My solution to that problem was to allow the preposed relative "le pe ci mi broda ku".

In proposing this to John Cowan, I did not realize that the real argument was centering on the distinction between inside and outside quantified sets, since I had not yet read Colin's paper. Cowan had put the issue to me in terms of an attempt to attach relative clauses to explicitly include the outside quantifier without mentioning that there was a reason why someone might want to also relatively modify the inside set as well. Thus I saw the solution as merely explicitly moving the relative clauses indisputably outside. (A major side effect of this turned out to be the need to put a terminator on LAhE clauses, which in turn has resulted in the simplification of the language indicated by Change 18. That change is numbered first because we agreed on it before the full proposal reached its full glory. Changes 17 and 19 are also side effects related to Changes 20 and 21.)

I also attempted to pretty up the grammar by combining indefinites with relative clauses in one place. The rule I proposed basically saw the use of an inside quantifier as "le <indefinite>" or "le <quantifier> lo <description>", a plausible but arguable proposition. Cowan talked me out of this to minimize change - it would require a "ku" after relative clauses for all indefinites (in addition to the relative clause terminator), though real speakers don't need it because indefinites have no explicit 'inside' set to be modified). The grammatical rule stayed in without mentioning indefinites, because by then the change was evolving to the current proposal. The remnant of this side exercise became option 3 under the change, and was the assumed default in the discussion.

Upon seeing Colin's writings, my first inclination was to say that inside clauses could be solved under this plan in a way that Veijo proposed: "le ci le <description> ku poi broda ku", and indeed it is a tribute to Veijo that this almost works. However, when there is no explicit outside quantifier, a problem that only manifests with "lo" and family, since "le" has a "ro" outside quantifier as default. For

lo sipna noi melbi"

I raised the following question with Nora: "Since the default quantification expands to

su'o lo ro sipna noi melbi

is the unexpanded form claiming that the 'indefinite sleeper' is beautiful, or are all of the sleepers?"

The answer appeared to depend on whether you expanded the quantifiers or not - the unexpanded form appears to be outside because we haven't explicitly quantified the inside; the expanded form seems more ambiguous. The problem is even worse when repeated with "poi", and Nora declared that something was indeed 'broken'. You cannot use Veijo's solution to fix this since "*le [su'o] lo sipna poi melbi ku" isn't grammatical with the "su'o" left implicit and unstated.

Thus we needed some kind of inside relative clause, and I looked at my working proposal and said, voila - it is already there. The preposed relative clause is indisputably 'inside', and I even had a postposed inside relative available when the inside set is quantified, based on the internal-indefinite rule.

Indefinites were separated back out per Cowan's argument, as mentioned above, but the result is highlighted in the following extracts from the E-BNF. Note that I consider the question of nesting of relative clauses and a couple of other things that came up, as side issues, but they also appear in the rules quoted.

[The following E-BNF is not the proposal as finally adopted, which deleted Change Proposal 21. However, it is fairly hard to understand the three original options of Change 20, along with Change 21, without this version of the E-BNF.]

sumti-3<93> = sumti-4 | gek sumti gik sumti-3
sumti-4<94> = sumti-5 | quantifier selbri /KU#/ | sumti4 relative-clauses
sumti-5<95> = sumti-6 | quantifier sumti-5
sumti-6<96> = (LAhE # | NAhE BO #) sumti /LUhU#/ | gek sumti gik sumti-4 | KOhA # | letteral-string /BOI#/ | LA CMENE ... # | (LA | LE) sumti-tail /KU#/ | LI mex /LOhO#/ | ZO any-word # | LU text /LIhU/ # | LOhU any-word ... LEhU # | ZOI any-word anything any-word #
sumti-tail<111> = relative-clauses sumti-tail | [sumti-6 [nested-relative-clauses]] sumti-tail-1
sumti-tail-1<112> = selbri | sumti-tail-2 | quantifier sumti
sumti-tail-2<113> = quantifier selbri | sumti-tail-2 relative-clauses
nested-relative-clauses<120> = relative-clauses ...
relative-clauses<121> = relative-clause [ZIhE relative-clause] ...
relative-clause<122> = GOI term /GEhU#/ | NOI sentence /KUhO#/
free<32> = SEI # [term ... [CU #]] selbri /SEhU/ | SOI sumti [sumti] /SEhU/ | vocative selbri [nested-relative-clauses] /DOhU/ | vocative relative-clauses sumti-tail /DOhU/ | vocative CMENE ... # [nested-relative-clauses] /DOhU/ | vocative [sumti] /DOhU/ | (number | letteral-string) MAI | TO text /TOI/ | XI number /BOI/ | XI letteral-string /BOI/ | XI VEI mex /VEhO/

The rule that proposed for 113 is the remnant of the attempt to merge indefinites and inside quantifiers. It allows inside postposed relative quantifiers before the "ku" if-and-only-if there is also an inside quantifier. My argument for this was that it allows the most natural meshing with the defaults assumed in the past language, which perhaps have been excessively English-based, but in any event are indeed historical and at least plausible interpretations.

John Cowan did not like this idea, because it makes "lo sipna poi melbi" and "su'o lo ro sipna poi melbi" group differently even though one is the defined transformation of the other. I argued that the transformation must include the "ku" explicitly before expanding, and thus there is no inconsistency. "lo sipna ku poi melbi" expands to "su'o lo ro sipna ku poi melbi". However, the inside restriction requires that the relative clause be preposed in order to contract it "su'o lo ro sipna poi melbi ku" -> "lo poi melbi vau/ku'o sipna ku"

Note that you need a terminator on the preposed relative clauses most of the time. I would use "vau", though "ku'o" is more exact, because "vau" is monosyllabic and the idea of preposing is to contract.

Excerpts from Change 20 and 21 as originally proposed but not in the final proposal

Options relating to allowing postposed relative clauses inside the KU (referring to inside-sets, and thus paralleling the preposed equivalent) lead to a complicated tradeoff, which is left for the community to resolve. Option 3) is believed closest to the current grammar and semantics, and is the default selection described by the E-BNF above.

  1. If postposed inside relatives are allowed in all descriptions, then the preposed/postposed distinction becomes a forethought/afterthought distinction, which can be valuable. It also makes existing texts retain their currently official inside-relative interpretation (unless the KU is explicitly present, a rarity), which is arguably desirable as the default (though it must be recognized that there are text examples where the speaker obviously wanted to apply the relative clause to the externally quantified sumti.) The negative tradeoff of this is that KU becomes always required when you want an external relative clause.
  2. If postposed inside relatives are never allowed, then all existing usages will become parsed as external relatives whether or not a KU is present. This is probably equally valid as 1) as a default, and makes a simpler, easier-to-teach grammar, since one learns the rule: prepose inside, postpose outside. The negative tradeoffs are that this eliminates the forethought/afterthought distinction, forcing the speaker to form all inside restrictions before starting the description. Somewhat more of older texts will be misinterpreted under the new parse, since they use postposed relative clauses, but are often intended to refer to the inside set.
  3. A third option is to allow postposed inside relatives only when there is an inside quantifier. Though it seems counter-intuitive that this would handle almost all problems with existing texts, in fact it appears to do so. Another negative aspect is that "lo broda noi/poi brode" (external relative) would have a different parse than "su'o lo ro broda noi/poi broda" (internal relative), which is merely the same sumti with implicit quantifiers made explicit. This could make it more difficult to teach, though it might make natural expression easier if relative clauses end up grouping correctly most often without the KU.

A note applicable to all options is that preposed relative clauses (but not relative phrases) will almost always require a terminator, though monosyllabic "vau" is usually as applicable as "ku'o". Since preposed relative clauses require a terminator, 1) or 3) may be advantageous in that they always allow the afterthought construction which does not require a terminator (but may require explicit KU too often, especially in option 1).

Allow nesting of relative clauses, distinct from ZIhEk grouping which retains relative clauses at the same level (commutative and associative, with all restrictions taking place before non-restrictive uses).
This change is mostly made moot by the addition of both inside and outside relative clauses, which probably renders the need for nesting to be negligible.
It is argued that natural language speakers will process relative clauses as they come to them, making "zi'e" grouping unnatural if in keeping with the logical aspects of the language. (Actual Lojban usage suggests that people will prefer to put "goi" assignments, which are nonrestrictive, closer to the sumti than restrictive ones, even when the wish the assignment to include the restriction.)
The advantages are that nesting allows variable assignment to intermediate restrictions:

lo sipna goi ko'a poi melbi goi ko'e poi mi nelci [ke'a] goi ko'i
("ke'a" in this case would seem to be the same as "ko'e", requiring

"ke'axire" to get the equivalent of "ko'a" if it was useful for some reason. Another argument is that "voi" restrictive clauses, which are intensional, would be implicitly nested. As yet there has been no example of a multiple "voi" relative clause to support this since "voi" is new in the language and remains seldom-used. Thus the bottom line is that some would like this option, and it is an expansion of the language that dovetails well with Change 20.

Commentary on the Proposed Change that led to the version that was adopted

Iain Alexander:

sumti-6<96>: How do we attach relative clauses unambiguously to a whole "<GEk ... GIk>" or "sumti ek sumti"... I think the only way to do that is some kind of terminator or grouping mechanism. Similarly we do need to say things like "Three of the people who voted", or "Three of the men who voted". But you can either use some sort of inside quantifier or use "ci lu'a ... lu'u", so we're covered.

If LUhI is the answer [yes, it is], then I'll accept that.

sumti-5<95>: I notice that in getting rid of multiple quantifiers on an indefinite description, you've ended up with multiple quantifiers on a sumti-6 :-)

Change 20. I've tossed this around various ways, and I've more or less convinced myself that, if it comes down to it, I can probably live with all the options, including no change. The argument revolves round the ability to force the required grouping, either by using one of the (LAhE that used to be LUhI), to force an inside quantifier, or an explicit "ku" to force an outside one.

The existing grammar has some potential ambiguities, such as

<quantifier selbri /KU/ relatives> 


<quantifier (LA | LE) sumti-tail /KU/ relatives>

(which latter is an expanded instance of sumti-3) - with the "ku" elided and no explicit grouping, it could be interpreted either way. You can regard this as a bug or a feature, depending on your point of view. The way the grammar is actually laid out suggests an outside relative for the former, and an inside one for the latter (but that's with or without the "ku").

In fact all versions seem to imply an outside relative for the former implicit indefinite, which is reasonable enough. However on balance, I suspect the ambiguities are too confusing.

On balance, I prefer an occasional extra "ku" to an occasional extra LAhE. The "ku" is shorter, and the LAhE carries an extra unwanted semantic implication. In the "poi" case, the distinction between some cases with and without the "ku" is vanishingly small, e.g. "lo sipna", "le ci sipna". In the "noi" case, I think if anything the "ku" helps to make the point, echoing the pause resulting from the comma in the English - but that may be excessively parochial.

I like the preposed relatives for variety, but I'm too fond of postposed relatives not to use them even at the expense of a little awkwardness.

I'm not so keen on option 2, since it means you will always need a LUhA to force an inside postposed relative.

The decision between options 1 and 3 is closer. If I were to work out all the cases, it might turn out that extra LAhE in option 3 were sufficiently fewer than extra "ku" in option 1 to tip the scales to option 3, but at the moment, I lean towards option 1.

21. The only problem with this appears to be cases like

"le prenu goi ko'a poi mi nelci ko'a goi le prenu poi mi nelci".
"le prenu <goi [ko'a poi mi nelci {ko'a goi <le prenu poi mi nelci>}]>".

... [Some complicated analysis by Iain showed that use of multiply nested and variable-assigned relative clauses, one of the few benefits of Change 21, are very non-intuitive. They often group differently than you would expect unless you put a lot of terminator in.]

In the current grammar, we could have said

lu'a lu'a lo sipna [vau] [ku] goi ko'a [ge'u] lu'u
poi [ke'a] melbi [vau] [ku'o] zi'e goi ko'e [ge'u] lu'u 
poi mi nelci [ke'a] [vau] [ku'o] zi'e goi ko'i [ge'u]

This is obviously cumbersome, but then the whole idea of three nested relatives with intermediate variable assignments is cumbersome. We already appear to have relatives coming out our ears in descriptions (preposed, nested, inside postposed and outside postposed). I'm generally in favour of flexibility, but perhaps enough is enough. Put me down as a NO, although not a very loud one.

Veijo Vilva

Lojbab's analysis of the (de)merits of the various options seems reasonable. My ranking of the options is, however, 2 3 1.

At this stage option 2 seems to be clearly the best choice and the difference between the other two is minimal. All the options are, however, acceptable to me.

1. option 2 seems to be the basic option, the other two are just elaborations of it : 2 < 3 < 1

2. Basically 3 and 1 just add ways to express the same things. I am not very concerned about the lack of the forethought/afterthought distinction in option 2. Most afterthoughts are, after all, incidental in nature and can be considered external.

3. Option 2 will cause perhaps the greatest amount of changes in the existing texts but the corpus is not too large at the moment. In five years time the situation will be different, I hope. It is always easier to expand the language later on, if the need arises, because it doesn't necessarily mean changes to the existing texts. I think it is wiser to adopt option 2 now and check the need for and syntactical consequences of options 3 and 1 very carefully during the five-year waiting period.

4. Not much goes to waste if the use of relative clauses is taught according to option 2 as it is the core option (besides being the easiest to teach).

5. I have tried to estimate the consequences of using only the preposed form of internal relative clauses based on the knowledge I have about different languages. I do read reasonably well Finnish, English, German and Swedish. In addition I know the basic grammar of Japanese quite well (my reading isn't too good). There are great differences between these languages in the use of preposed clauses. English is quite limited in this respect, Finnish coming as a good second. Swedish and German are reasonable and in Japanese it seems to be the only possibility and is quite well developed.

I have often been quite frustrated writing Finnish because of the inherent limitations of the so-called pro-sentences which can be preposed. It takes extreme care in the formulation of the postposed relative clauses to make exactly the point I am after as it is all too easy to write ambivalent sentences. The possibility to use the preposed restrictives would usually solve the problems but the limitations in the Finnish system are too severe. In Japanese the problem is reversed. You can prepose complete sentences but differentiating between restrictive and incidental clauses may be difficult. I have never had, however, difficulties in understanding and using the preposed clauses of Japanese in general.

My general feeling is that the use of preposed relatives shouldn't cause unsurmountable difficulties.

The beauty of the preposed restrictive clauses is in that you define beforehand what you will be talking about - it's kind of having a local prenex. The incidental information is clearly separated and there is less chance for confusion. I feel this is so important that I'd be willing to give up in exchange the nested relative clauses I have been advocating. (NB. Even though the nested relatives do offer some theoretical advantages, we may be asking for trouble in the form of lots of incomprehensible exercises of cleverness if we adopt them.)

It is also noteworthy that a sumti with preposed relatives is a very clearly demarcated entity and in x1 position there won't be the separation caused by a postposed relative between the main sumti and the selbri.

6. One of the weaknesses of option 3 is that the legality of the postposed internals - which many feel are more natural - is dependent on the existence of the internal quantifier. In the heat of a conversation it's all too easy to forget the rule and use the postposed form even when not appropriate.

In option 1 the need to juggle the KU's is a real drawback and a possible source of confusion. The flexibility of opt1 may be more illusory than real. It might well turn out that in practice this extra flexibility would be more of a burden. It is also more difficult to check the consequences of the adoption of option 1 to the whole sumti grammar.

Option 2 has no apparent weaknesses and is in a way a quite balanced choice between two worlds as the restrictives will mostly be preposed and the incidentals postposed - so everybody ought to be happy :-).

I think we ought to use the design of the language as a tool to enforce clearer ways of expression - as long as the adopted design doesn't hinder expression. How many of us do really customarily strive for exact expression? Most of the scientific articles I have read during the last 25 years have been full of ambiguous sentences - irrespective of the language they have been written in. Quite few authors seem to have the ambition, the talent and/or the time to hone their expressions to clarity. It would be a real bonus if a language were designed so as to gently push the users in the right direction. Maximum flexibility in a language sets also the greatest demands on the user to avoid ambiguous ways of expression. We are in a unique position and we ought to do our best to find the correct balance between regulation and flexibility. I feel that the expressive power of Lojban at its present stage of development is such that even if we adopt the most restrictive one of the options, it is quite impossible to prevent a really determined individual from presenting his thoughts in a muddled way - so I think we needn't worry.

Colin Fine:

What particularly delights me is that your proposal in effect matches both much of my recent suggestion, and also the call I made the other month for pre-posed relatives. I did not expect this bounty.

I understand le do'o reluctance to make a change of this size this late, but I believe it is a noticeable improvement to the grammar, so I certainly support it. I definitely favour option 1) (which is the closest to my suggestion), but would accept 3). I am least happy with 2).

A few more specific comments: "le pe ci mi broda" was exactly what I argued for the other month.

I have one or two queries about the grammar you exhibited: 1) the E-BNF has "gek" in both sumti-3 and sumti-6, which surprised me, and indeed it seems to be only in sumti-3 in the YACC. This prevents you from saying

*ci ge le broda gi ko'a and
*[ge le broda gi ko'a ] poi melbi 

which seem fine to me - they're not very intuitive, and if you really want them you can nest explicitly though sumti-6 with LAhE or else LE <quantifier> <sumti>. I take it that this is actually just a bug in the proposed E-BNF. [Yes]

2) I found it a bit odd that both sumti-4 and sumti-5 can start with quantifier, but I take it LALR-1 can handle this.

3) I also found it odd that multiple "zi'e zei claxu" [without-zi'e] relative clauses are sometimes left-branching sisters of a constituent (sumti-tail), sometimes right-branching ditto (sumti-4) and sometimes a constituent in their own right (nested-relative-clauses). I accept that this is an artifact of writing grammar for YACC, but I think it is unfortunate for a "nu'o" syntactic-semantic description of the language, not to mention any transformational account.

The three options: I favour option 1) because it is the most orthogonal - I don't like the way that forethought/afterthought either have different meanings (2) or depend on other structures, whose relevance may not be immediately obvious (3). Note that the part of my argument which you have rejected is my claim that the unmarked position for incidentals should be external, while that for restrictives is internal; option 1 reflects that belief in the (more important pe'i) case of restrictives.

Preposed relatives: I didn't say that "postposed relatives are abnormal to all but English speakers in an AN (adjective-noun)-ordered language"! That's a much stronger claim than I ever intended to make. I said that some languages have only pre-posed relatives, and I don't see why Lojban should not extend its flexibility to allow those.

I note that we will have the option of teaching pseudo-possessives as a special case of preposed-relatives, thus

le mi zdani

as elliptic for

le pe mi zdani 

just as

ze mensi

is elliptic for

ze lo mensi.

I don't say we have to do this, but it is an option.

Mark Shoulson:

I prefer options (1) and (3) greatly over option (2), perhaps with slight preference to (1). I don't like the restrictiveness of (2); I want to be able to put my relatives as afterthought even if they're inside, thank you. (3) seems kludgy, and I don't much mind the odd "ku" thrown in here and there to make (1) work. For one thing, it's usually close to right even without the "ku", and for another, "ku" is a short, quick syllable, and we've already gotten used to using it with the very common conjunction "joi" ("lo nanmu ku joi lo ninmu", etc.) And don't screw around with reversing "ku'o" and "vau"; much work for little gain.

John Cowan

Infinite quantifiers on a sumti: I agree that this is a useless wart and that it should go. One quantifier is enough; if you want more, use "lo I lo J lo K broda". [It went.]

Relative clauses vs. logical connectives: I don't agree that it makes sense to attach a relative clause to logically connected sumti. Remember that logical connection expands to separate sentences. If this really needs to be done, use LAhE.

[Mark Shoulson responds: Oh, no. It is very sensible. I ran into it when I started playing with the Tower of Babel story. If you check your text, God descended to see "the city and the tower which the sons of Man had built." I think we'd all agree that that's a very natural construction, and that "which the sons of Man had built" obviously applies to both the city and the tower. Logically (and non-logically, for that matter) conjoined sumti are as natural to language as simple ones, and are as likely to be relativized as a unit. I used a LUhI/LUhU set to handle this case, as "lu'a le tcadu .e le kamju lu'u poi loi remna cu zbasu" (I thought the logical ".e" worked here, but maybe not...). It could be that termsets are the best answer to this type of problem, but it is not true that this type of construction is nonsensical or uncommon.

[Colin Fine replies: But John specifically referred to "logical connectives" and your example is better translated with a non-logical.

[Mark: Well, allowing one entails allowing the other, so it amounts to the same thing. And I did consider using a non-logical (perhaps "ce"), though I figured that the observation could be independent, simply "seeing one" and "seeing the other", as if in two sentences, and thus using the logical ".e". Stylistic point of contention, of course, and I'm open to correction.]

[Colin continues: Nonetheless, I agree with you [Mark] - a logical ".e" is possible there, though I don't think it is a good translation; and in any case, there are plenty of examples with ".a" or ".onai"

mu'ulu<< mi darno viska le xirma .onai le xasli .i le sego'i cu lacpu le karce >>li'u
e.g. " I see far off a horse or donkey(. It's) pulling a cart"

This is one way to say it, and there is another with a connection inside the description, "le xirma jonai xasli noi lacpu le karce", but I don't know how to get it with connected sumti and a "noi", which is what I want to use. (The Lojban above does not express whether the second sentence is restrictive or incidental).

[Lojbab: Non-connected sentences are inherently incidental.]

[Mark replies: The only way, currently, to do it is using LUhI/LUhU. Pick the one that makes the most sense. I'd go with "lu'a". Thus:

mi darno viska lu'a le xirma .onai le xasli lu'u poi/noi ke'a lacpu le karce

Simple enough, but I suspect common enough to warrant finding a way to do it without the "lu'a" and unelidable "lu'u". Can our tired, overworked "bo" help? No, I think it's already in use in that place...]

[As a result of the above discussion, option 1 was selected, and the proposal was modified to account for the comments.

Usage Questions and Grammar/Word Proposals Related to Usage


by Greg Higley

Has it ever been considered that some of the members of selma'o BAI might be better construed as members of a conjunctive selma'o such as JOI? In particular we have "mau" and "me'a". To borrow a natural language analogy, aren't these much more like conjunctions than like prepositions, much more like non-logical connectives than like sumti tcita?

Take a look at a sentence with a JOI connective:

(1) mi djica lo vanju ku ce lo djacu
"I want the wine and the water."

Here both wine and water are se djica. This sentence can be expanded to:

(2) mi djica lo vanju .ice mi djica lo djacu.
"I want the wine and also I want the water."

The "force" of the x2 place of djica is distributed to both sumti linked by ce. Now look at a sentence containing semau "more than":

(3) mi djica lo vanju ne semau lo djacu
"I want the wine more than (I want) the water."

Here the sumti "lo vanju" is the x2 place of djica, and "semau lo djacu" is simply linked to it as a modifier. Awkward!

It is clear semantically, though it is not true grammatically in this case, that lo djacu is a kind of "spiritual" x2 place of djica. Why not make it one explicitly? Think how much clearer and easier it would be to say:

(4a) mi djica lo djacu ku mau lo vanju
I want the water, exceeded by the wine.


(4b) mi djica lo vanju ku semau lo djacu
I want the water, more than the wine.

- regarding these as JOI.

In this way they could even be used in tanru, just as the members of JOI are. We could say:

(5) le karce cu xunre semau narju
"The car was more red than orange."

With the current definition of the grammar, I can't even imagine how to say something like this. You can see how much easier it is to do if we change the grammar of mau and me'a.

Sentences too could be linked much more easily this way. We could say:

(6) le karce cu xunre .isemau ri narju.
The car is red. More than it is orange.

I think the main reason why "mau" and "me'a" were included in BAI in the first place is that when the list of gismu were sorted to look for candidates for inclusion in the BAI set, "zmadu" and "mleca" seemed obvious choices. But I think it's fairly clear that they are conjunctive and not modificatory in nature, as evidenced by the current awkwardness of their usage. Please consider changing their status. (I am currently looking through BAI to see if any others of its members need to be put into a new conjunctive selma'o.) Actually, zo me'a du lu semau li'u .ije zo mau du lu seme'a li'u. This is a little redundant. I suggest me'a for "less than" and mau for "more than". This is opposite to the current definition, but seems more intuitively correct. Their conversions, seme'a and semau would be unnecessary. Keeping their place structure integrity would be irrelevant, since they would no longer be BAI.

Try "playing around" with these as conjunctive cmavo, and see if they aren't much easier to use.

Below are a few sentences designed to show the potential range of use of my suggested definition of me'a and mau:

(7) mi mau la djan djica lenu klama ta
I more than John want to go there.

(8) mi djica lenu klama ta .imau la djan. go'i
I want to go there more than John does.

(9) mi djica lenu klama ta me'a la rom.
I want to go there less than to Rome.

(10) mi pumauca nelci lo vanju 
I was more than I am fond of wine.

(11) mi dzukla mau bajykla 
I am more a walker than a runner.

Perhaps you can think of some more structures in which mau and me'a might be useful.

Mark Shoulson:

Oh, my. "mau" and "me'a" as JOIs. The scary part is that it makes a lot of sense. I don't feel strongly enough to join Higley in calling for their re-classification, mostly because it's a major change in concept and in syntax, and it would invalidate a lot of text. But if by some bizarre set of circumstances reclassifying them gains support, I wouldn't be opposed, much. Gotta think about this more.

Colin Fine:

I accept the point you are making in [Example 1], but the example is flawed.

"jo'u", "joi", "ce" are non-logical connectives delivering the three basic types of sumti: individuals, masses, sets. (This is one of Lojban's few obligatory grammatical categories, and, interestingly, it is not shared by any other language that I know of).


mi djica lo vanju ku ce lo djacu
I want the set containing wine and water

does not say anything about wanting wine or water. Use 'jo'u' or else use 'lu'i'.

The same applies to the '.ice' construction - except that it is very unclear what on earth it means. I think it is constructing a set of sentences, but I'm not sure. In any case, it has been well established that you cannot in general expand non-logical connectives [into multiple sentences] in this way.

All of which does not affect your point ...

The effect you want in 4a/4b can be achieved with the current grammar, admittedly less elegantly:

mi djica lo vanju .esemaubo lo djacu

asserts that both are wanted and that there is a "semau" between them.

(Note that this gives a possibility of variation lacking in your method:

mi djica lo vanju .anaisemaubo lo djacu
I want wine only if, but more than, water.)

Your version of (6) is the form most closely approached by the current grammar:

le karce cu xunre .isemaubo ri narju

What your suggestion does ignore is the possibility that there are uses of "mau" which are genuinely sumti tcita (attached to a selbri). I agree these are not frequent, but there are some:

mi gleki semau tu'a le prujeftu
"I am happier than last week"

Probably you can always find a paraphrase (often using "zmadu"), but the fact is that there are current uses of "mau" which your proposal does not meet (note that you can almost always paraphrase a sumti tcita with the corresponding gismu, but this does not make them useless).

If they were changed to JOI, [using "mau" and "me'a" instead of "semau" and "seme'a"] would make some sense: place structures for most BAI are counter-intuitive until you understand the principle. However, note that JOIk in the grammar has an optional 'SE' anyway - at present the only asymmetric JOI is 'ce'o', but conversion is permitted for all of them.

[On Greg's (7), (8), (9):] These are all good, but can be expressed with "[j]esemaubo".

[On Greg's (10):] This is exciting. I can't see an easy way of doing it at present. The best I can think of is:

mipepu .esemaubo mipeca cu nelci lo vanju
I of the past, more than I of the present, am fond of wine.

[On Greg's (11):] Poor example - I took that as "I walk more than I run", which is different in English, but the principle stands.

mi dzukla gi'esemaubo bajykla 

- but that has a different structure, because yours is one tanru, mine is not.

This example also shows the general problem with "mau" - the scale is not expressed. This is a problem with the existing "mau" too, but it is possible to add a "ci'u" or "ji'u" phrase. I'm not sure that would work with "mau" in JOI.

I agree [with Mark] that it makes a lot of sense, and is quite attractive. I don't agree that "it's a major change in concept and in syntax" - on the contrary, it is shifting two words from one selma'o to another (existing) one. It would invalidate a lot of text.

However, I think that unless Greg can convince me that he can cope with existing structures, I will not support the change.

Result: Change 28 was proposed in response to this issue, but currently there is no support to implement it. Changes 30 and 31 indirectly derive from this change. The ensuing discussions on the topic have led to significant rewriting of material in the draft textbook, and a couple of the minor grammar changes above, which enhance the expression of joined sumti in the 'termset' construct.


by Greg Higley

As I understand it, the cmavo "kau" indicates that the value of that which it "modifies" is known, presumably to the speaker, but there are instances where this is apparently not the case. Thus if I say

mi djuno le du'u pakau le prenu pu dzuli'u le loldi
I know that one of the people walked on the floor, and I know which one.  
I know which one of the people walked on the floor.

I am indicating that the referent of "pakau le prenu" is known (to me). Thus "kau" means something like "referent known". And if I just say "pakau le prenu pu dzuli'u le loldi" apparently the meaning is the same as when "djuno" was the main selbri. And here's where we run into a problem. How do we know to whom the referent is known? Is "kau" somehow connected to the x1 sumti of "djuno" and any other related gismu? For if I say

la djos. djuno le du'u pakau le prenu pu dzuli'u le loldi

apparently it is to Joe (and not to me?) that the referent of "pakau le prenu" is known. If "kau" does not always indicate that it is the speaker who knows the referent, what is the standard for determining this? For

la djos djuno le du'u pakau le prenu pu dzuli'u le loldi

could mean

Joe knows that one of the people walked on the floor, and I know which one.

But this seems contrary to intuition. What is the standard? Is there one?

In the examples that came with the article on "kau", it was used with words which might be classed as "indefinites" and "interrogatives", and apparently these were used interchangeably. For our purposes, an indefinite is a word like "zo'e", while an interrogative is a word such as "ma" (which, as I'll show, is a close relative of "zo'e"). I think it would be useful and advantageous to split the use of "kau" as it is used with indefinites and interrogatives. With interrogatives, "kau" could be used to ask a question, while indicating that the speaker already knows the answer. Thus a teacher could ask her students

mi makau zukte makau

What am I doing and to what end? and her students would realize that she wasn't just asking this for her (mental) health.

With indefinites on the other hand (and I class such things as "pa le prenu" among them), "kau" would perform its simple duty of letting us know that the referent is known.

mi zo'ekau zukte zo'ekau

means something like

I'm doing something-known-to-me for some purpose-known-to-me.

And thus

 mi djuno le du'u do du zo'ekau
I know that you are someone-known-to-me.  
I know who you are.

becomes easy.

Has anyone yet noted the strong relationship between "kau" and "ki'a"? The former indicates that the referent is known, and the latter asks for clarification. Both can be used to express "which one of the people" but in semantically different situations. Still, the relationship between them is clear, and perhaps worth exploring further.

Also note that "zo'eki'a" is virtually identical - if not completely identical - to "ma" in meaning. In fact, it is probably possible to form the whole range of interrogatives by affixing "ki'a" to their corresponding indefinites. (Japanese, I believe, does something similar.) I am not suggesting that this be done. It would be unnecessarily verbose. But it is worth noting the relationship.

Nick Nicholas:

[Who does "kau" refer to?] An outstanding question. I have held that the knower of "kau" is the knower of the bridi it is in, implicit or not. "John knows which one." I also wished that extended to observative attitudinals such as "za'a", which gave rise to reaction from Lojbab. This issue is unresolved, but I agree with you on the above solution being counter-intuitive. "se'i"/"se'inai" exist as (kludgy) patchwork disambiguators at the moment. But no consensus on default interpretation was reached.

I hope this distinction [between interrogatives and indefinites], which is pretty elegant and clear, wasn't passed over in the specification of "kau" (although I remember at the time that I felt I understood "kau" better than Lojban Central :). But yes, that's correct.

By the way, as John Cowan will no doubt point out, "kau" is not restricted to knowing/"djuno", but can extend to all sorts of analogous concepts like believing, opining etc.

Colin Fine:

I don't believe that "se'i" works like that at all. As things stand at present, all discursives, like all attitudinals (other than "pei") strictly refer to the speaker's intentions/quality of knowledge/attitude. I have on occasion wanted a way to indicate somebody else's attitude etc., but I'm not convinced that it is desirable. ("se'i" is about whether the speaker's attitude relates to "vo'a", not about whose attitude it is).[1]

On reflection, I think [Greg's] is a good distinction. However, if this is the case, then "kau" does not, as I thought, remove the 'performative' quality of question-words ("ma" etc) - then various texts of mine, and I think others, are wrong.

mi djuno le du'u le cukta cu zvati makau

is still asking a question of the hearer, which was not my previous understanding of it.

By the way, "kau" is not restricted to "knowing/djuno, but can extend to all sorts of analogous concepts like believing, opining etc." Asking, too!

Iain Alexander:

"kau" was the subject of the first comment I posted on the list. My interpretation of John Cowan's response is that "kau" isn't about "knowledge", it's about abstraction, in particular, the identity of the concept it's attached to. So "lekau prenu" is "the identity of the person".

Since it's a UI, it can be attached to almost anything, to denote the identity of, e.g. a logical connective. The current official position is that exactly which member of the selma'o (or presumably, which gismu) is used is not important, although it might indicate something about the type of value expected.

With this interpretation, "le pakau prenu" means "the number of people", i.e. essentially the same as "leni prenu".

In practice, it frequently occurs inside a "du'u" abstraction, with the side-effect of 'inverting' the whole construct to refer to the identity of whatever is tagged, within the given context. To my mind, this means it changes the meaning of "du'u". Further complications arise if the "du'u" is nested, in which case subscripts need to be used to indicate that the "kau" is relative to an outer "du'u". Things might be simpler if a separate cmavo, say "xau", in selma'o NU, was allocated for this usage, meaning "x1 is the identity of whatever is tagged with "kau" in [bridi]".

Nora LeChevalier: My understanding of "kau" is that it flags the 'key item' for any bridi. Thus,

mi djica lenu pakau le prenu pu dzuli'u le loldi

doesn't say that I know the one who walked on the floor, but rather that I desire that particular one. It can be used to say "John is the one I want to walk on the floor":

mi djica lenu pakau le prenu ku po'u la djan. pu dzuli'u le loldi

"zo'eki'a" can appear after usage of "zo'e" as more of a metalinguistic comment (What do you mean "zo'e" - "zo'e" can't be the right word here!) and is thus similar to "na'i". "ma" has no such function. Using "zo'eki'a" for "ma" would deny the important usage that prompted invention of "ki'a". "ki'a" is a request - for clarification - and would be inappropriate except in response to someone else using the words that you are questioning.

  1. Iain comments: Regarding Colin's comment on "se'i",] as I understand it, the way to indicate someone else's attitude etc. is to use something like "sei [vo'a] jinvi".
    Colin responds: or "fi'o jinvi ko'a"...
    You can do this with most UI, but it sometimes needs some thought to find a suitable brivla. Anybody got any ideas about the selbri corresponding to ".ai"?
    Iain replies: The closest I've come up with is "terzu'e". As mentioned in my comments on Nick's mekso translation, "ca'e" isn't very easy either.
    Lojbab: When we first created the attitudinal list, we had a gismu or brivla equivalent for each attitudinal - this was part of the criteria in choosing the original gismu list: a primitive emotion word should have a primitive root. The redesign of the attitudinal space, and the major expansion that result therefrom kinda messed this up. The distinctions that are permitted now using attitudinals are more diverse than there are yet defined gismu and brivla in that semantic space.
    By recollection, the old meaning of ".ai" could simply be handled by "balvi". The sense that JCB had for ".ai" was like unto the American sailor's response "Aye, Aye! Sir!", hence the cognate. But we certainly now have the capacity to distinguish between "intent", "prediction", and "expectation" using the attitudinals, and "balvi" no longer satisfies me for ".ai". My choice of the top of my head would be "platu" using the new place structure that puts a planner in x1, instead of a plan.
    As for "ca'e", I can see a lot of these questions coming. Someone want to tackle a list of gismu/brivla for the entire attitudinal list? Editted and enhanced, it will probably be added to the dictionary-in-progress. "ca'e" doesn't seem that hard: "smuni xusra" or "smuni cuxna" or "smuni jdice" seem like tanru on which to base a lujvo for "define". Hmm. Add in "sruma" in combination with the above to add to the possibilities.

le lojbo se ciska (cont)

Speaking of "kau", the following Lojban text makes use of the word. See Nick's footnoted comment for his further views on "kau".

Following is Colin Fine's translation into Lojban of a familiar children's fairy tale. It is the first text to be vetted under the 'editor de jour' concept described in JL17. Nick Nicholas served as the reviewing editor. In this case Nick recommended publication, making some comments. Colin declined to make Nick's suggested changes, which therefore appear as footnoted comments. All lujvo have been updated to the new rafsi list enclosed with this issue (manually by Lojbab, so please forgive any errors).

The translation immediately follows, unlike our normal practice, due to the length of this issue.

®lu le nolraixline ga'u le dembi li'u¯

cmene di'e noi se finti la xans. krIstian. Andrsn.

=.itu'e tu'e

lisri le nolrainanla[1] goi ko'a

=.i ko'a djica lo nolraixli =.i ri mulno be loka nolraixli be'o gi'o se zanru ko'a =.isemu'ibo ko'a fe'eroroi litru gi'e sisku pa go'i =.iku'i roroi nabmi =.i sa'e ge lo nolraixli cu raumei ju'o gi lo ni ri nolraixli ku ko'a na se birti .!uu =.i roroiku le no'e drani vau[2] =.i ko'a ki'u se'irzdakla gi'e badri lenu na'epu'i cpacu lo nolraixli mulno

ni'o pa vanci cu ki jaica ke selte'a vilti'a =.i lindi joi savru joi carvi joi camcilce =.i zo'e darxi le tcavro =.i le sorna'a nolraitru ki'u minde lenu le vorme cu karbi'o =.i le bartu cu nolraixli =.i ri selkecmlu .!uuse'inai ri'a tu'a lo carvi .ebo lo xlali vilti'a =.i mo'ini'a flecu lo djacu vi le kerfa .e le taxfu =.i flecu ji'a pa'o le cutci file cucti'e le cucyzbi [tosa'a pamoi pinka toi] =.i cusku fa ra ledu'u ra nolraixli mulno

ni'o ®lu .!ue =.i cipra li'u¯ se seisku le sorna'a truspe goi fo'e =.ije ri bacru noda ku'i gi'e klama le sipku'a gi'e vimcu ro le ckabu'u gi'e punji le pa dembi le ckazbe =.ijebabo fo'e cpacu reno vresraki'e gi'e cpana punji ri le dembi =.i pa'aku reno datkypi'u gairki'e co'a cpana le sraki'e =.i ro go'i cu se vreta le nolraixli goi fo'a ca'o le nicte

ni'o co'i le cerni cu preti fo fo'a fe leli'i fo'a capu[3] sipna ge'ekau[4]

=.i ®lu .!oicairo'o [seisa'a selsku be fo'a] =.i mi su'eso'uroi .!uu ga'orga'i le kanla ca'o piro le nicte =.i ?ma za'anai ?pausai nenri le ckana[5] =.i mi puca'o vreta le raktu jdari =.i piro lemi xadni ri'a bunre joi blanu =.i to'e zdile .!oisai li'u¯

=.i seni'ibo co'i djuno ledu'u fo'a nolraixli je'a mulno ki'u lenu fo'a[6] fi le reno sraki'e ku jo'u le reno gairki'e cu ganse fe le dembi =.i lo ckaji be loka ganse du'i la'edi'u cu nolraixli mulno ju'o

ni'o le nolrainanla goi ko'a co'a speni fo'a =.i ko'a seki'u djuno ledu'u vo'a kansa le mulno be loka nolraixli =.i le dembi ba se punji fi la larku'a [tosa'a remoi pinka toi] =.i caji'a go'i[7] =.ijo noda capu vimcu .!iacu'i tu'u ni'o di'u jetnu lisri .!uo.ui


ni'oni'o di'e pinka

=.i pamai le lujvo po'u zo cucyzbi cu satci te fanva fe ®zoi.dy. Naesen paa Skoen .dy.¯ =.i mi nelci le di'u bangrdanska tanru

=.i remai [tu'e la larku'a po'u ®la'o .dy. Kunstkammeret .dy.¯ cu ga'orbi'o ca le nanca be li pabirepa gi'eseri'abo ca'a teke carmi morji caze'u le lisri =.i le'i ca'a jmaji noi selzda le tolci'o ke nolraitru ckusro dinju cu selcmi so'i vrici ne mu'u lo prucedra lisri ku ce lo naiske lisri ku ce lo rarske cizra tu'u] =.i di'u se krasi le pinka ne bau la dansk. fo la xans. briks. jo'u la .anker. iensn.

Colin's translation:

The Princess on the Pea

There was once a prince, who wanted a princess for himself, but she had to be a real princess. So he went all round the world trying to find one, but there was always some hindrance: there were plenty of princesses, but whether they were real princesses, he could never be sure - there was always something that wasn't quite right. So he went home and was sad, because he so much wanted a genuine princess.

One evening there was a frightful storm. There was lightning and thunder, the rain poured down, it was dreadful! There was a knocking on the town gate, and the old king ordered it opened.

It was a princess standing outside. But God how she looked in the rain and the storm! The water ran down her hair and her clothes, and went in at the toes of her shoes and out at the heels. And she said she was a real princess.

"We'll see about that!" thought the old queen, but she said nothing. She went to the bedroom, took off all the bedclothes, and put a pea on the base of the bed. Then she took twenty mattresses and put them on top of the pea, and then twenty eiderdowns on top of the mattresses.

And that's where the princess was to lie that night.

In the morning, they asked her how she had slept.

"Oh, terribly!", said the princess. "I hardly closed my eyes the whole night! God knows what there was in the bed! I was lying on something hard, and I'm black and blue everywhere! It's quite horrible!"

So they could see that she was a real princess, since she had felt the pea through twenty mattresses and twenty quilts. Nobody but a real princess could be that sensitive.

The prince took her for his wife, for now he knew that he had a real princess, and the pea was put into the Kunstkammer, where it is still to be seen, if nobody has taken it away.

You see, it's a true story!

Note (from Blix & Jensen): The Kunstkammer ("art chamber") closed in 1821 and was therefore fresh in memory at the time of the tale. The collection was housed in the old Royal Library, and contained many different things: old sagas, ethnographic tales, curiosities of natural history, and so on.

Sylvia Rutiser, of the DC-area Lojban group, attempted her own independent translation, though she did not complete it. Since Sylvia is a moderately skilled Lojbanist, her effort is a reasonable standard for a learning Lojbanist to strive for. Significant differences between the following and Colin's version of what he intended, are areas where either Colin wasn't clear, or used a construct that even Sylvia could not figure out (Sylvia admitted having some unanswered questions when she completed the translation; in some cases, the wording may be strange due to these questions).

"The princess and the bean" names this that was invented by Hans Christian Anderson.

This is a story of the prince. He desires a princess. She is complete in the quality of "princessness" if and only if she is approved by him (I question this). Therefore, he travels everywhere and seeks such a princess. However, there are always problems. To be precise, there were enough princesses, and he was not certain if they were all princesses. Always something was not correct. Therefore, he went home and was sad about not being able to get a complete princess.

(Set time) One evening it was stormy. Lightning and rain and intense wildness. Something hits the city gate. The old king therefore commands that the door be opened. The outside thing is a princess. She was pitiful seeming because of the rain and storm. Water flowed off her hair and clothing.

"Surprise! Test. Intent" is said to herself by the old queen. And she said nothing and goes to the sleeproom and removes all the bed-cloth and puts one bean on the bed-frame. She then takes twenty mattresses and sets them on the bean. Each respectively twenty duck-feather cover cushions upon the mattresses. All of this is reclined on by the princess through the night

In the morning she is questioned about the experience of her sleeping (emotion unspecified)

"Ouch! she said I my eyes all night. Why? I observe ( question follows) in the bed. I continuously reclined on the troubling hard thing. All of my body (therefore) is brown mixed with blue. Not funny. Complaint!"

Therefore it is known that she is a princess truly complete, because (reason) the event that she (through 20 mattresses and 20 coverlets) felt the bean.



  1. Lojbab: Colin chose to base his words for "princess" and "prince" on "nanla" and "nixli", which explicitly denote immaturity, even though it seems from the story context that the prince, at least, is an adult (he is taking the princess as a wife, and it appears to be his volition rather than an arranged marriage in the royal youth. Better choices are "nanmu" and "ninmu", which explicitly do not imply maturity. "nakni" and "fetsi" might also do, though they do not necessarily imply 'human'; however, "person-ness" is implied by the "royal-" status - the story could easily be told about a non-human but vaguely humanoid intelligent species.
  2. Nick: Hm. Because I equate the referent of "lo nolraixli" with the earlier one in the tale (He seeks a princess), this sounds like "the princess is enough". But of course, the Lojban doesn't say that at all; "nolraixli" is quantified afresh here. Still, might it not make more sense to say "raumei lo nolraixli" or "loi nolraixli cu raumei"? I don't recall the place structure of "mei" right now. And I'd have said "roroiku da no'e drani" (note that, for quantification, the "roroiku" has to go before the "da", else we assert that there is one thing always awry, rather than one thing each time. (We do need a quantification paper badly).
  3. Lojbab: John Cowan has expressed the opinion that, under the rules as interpreted by his tense paper, cmavo compounds based on "ca" no longer have perfective intent.
  4. Nick: This remains a clever use of "kau", and should get mentioned in any write-up about it. By the way, from my reading, it does seem that lambda calculus is the best way to explain "kau". For those unfamiliar with it: lambda calculus explains math at a deep level. 'LAMBDA(x.x+x)' is the function taking x as an argument and returning x+x. Lambda(x.x+x) 1 is a function application to 1, and evaluates to 2. The lambda expression itself is a function waiting for an argument. Lojban selbri aren't lambda function; their arguments are filled with "zo'e", or explicit values. In "I know who did it", though, the predicate "did it" is crying out for an argument to fill in x1: (LAMBDA "zo'ekau"."zo'ekau gasnu ri"). For that matter, a lot of the elliptical places, as John Cowan has mentioned, get explained by it: Being a parent is difficult - not being a parent of John, or of Mary, but (LAMBDA "zo'ekau"."mi rirni zo'ekau").
  5. Nick: I don't like "ga'orga'i", but that's a matter of taste. I rather like the "za'anai ?pausai".
  6. Nick: I think you need a "kei" before "ki'u": her feeling the pea does not cause her to be a princess, but causes them to know it.
  7. Nick: I don't know about "go'i" - what is true now is that the pea remains there, not that it is still being placed there.

Empathy in Attitudinals - A Proposal by John Cowan

[This proposal deals with an issue discussed in footnotes from the last technical article on "kau", though the proposal arose separately.]

As part of reviewing the cmavo list for inclusion in the dictionary, I have been thinking about the current uses of attitudinals. As originally specified, the attitudinal indicators of selma'o UI were solely to specify the speaker's attitudes. Thus ".ui" expresses the speaker's happiness.

However, there has been an increasing pull toward allowing attitudinals, suitably marked, to express other people's feelings as well. In particular, "se'inai" has been employed as an attitudinal modifier for this purpose.

I find this use objectionable for two reasons: 1) It conflicts with the original purpose of "se'i"/"se'inai" as described in the attitudinal paper; 2) support for emotional empathy should not be done with a negated cmavo.

The original purpose of "se'i" was to indicate that the object (not the subject) of the feeling was oneself rather than another. Thus, where ".au" means "desire", ".ause'i" means "I want it" whereas ".ause'inai" means "I want you to have it". This function obviously conflicts with using ".ause'inai" to mean "You want it".

There exists a general mechanism for expressing complex attitudes: "sei" followed by a bridi with limited syntax. With this machinery, "You want it" becomes "sei do djica". However, it is often hard to decide exactly which selbri should be used to express a particular attitude, and for the case of attributing feelings to another, some additional support may be useful.

Some natural languages support this feature to a limited degree. I am told that in Swedish the word "uffda" signifies ".oiro'o in empathy" - you say it not when you stub your toe but when you observe someone else do so.

[The proposal was formulated as: we propose "dai" as an attitudinal indicating "speaker empathy", secondarily allowing someone to attribute attitudinals to others in speech or text. The former meaning of that cmavo (in selma'o KOhA), which has seen no actual use, has been assigned to "do'i".]

Nick Nicholas:

The empathy attitudinal is something whose time has come: do it, John, do it!

Jim Carter:

I have found it useful for attitudinals to describe the attitude of the subject of the bridi which the attitudinal is in. In the most common usages this will be the speaker, and a fair number of other-person usages are also subsumed automatically.

Of course this was all worked out for Old Loglan. Some of the new UI's in Lojban may be more speaker-tropic than the old ones - and in fact I was very tempted to make a blanket exception that .ua- .ue-.ui-.uo-.uu always referred to the speaker, not the subject. Also there was a strong distinction between "discursives" and "attitudinals", and the item related by the discursives was usually or always "the previous discourse" rather than "the speaker". (Example: le bi'u cribe = the bear which is absent from the previous discourse, not the bear which the speaker is not familiar with.) The point of these weaselwords is that we should specify with each UI a default argument selected from speaker, subject or previous discourse.


I accept the idea, most especially for narration, such as Ivan's translation (in JL17), where the attitudinals expressed are those of the characters, and not of the author. I suggest that a combination of "sei" metalinguistics and the proposed "dai" could be used to indicate whose point of view is indicated in freely inserted attitudinals. Or a long scope attitudinal attached to "dai" at the beginning of a story like Ivan's, merely leads to the obvious interpretation that all attitudes expressed in a story are those attributed empathically by the speaker to the characters.

On the other hand, I will strongly encourage the emphasis on empathy, and not that you are in any way claiming an attitude on the part of another person. We never really know what another person is thinking, or feeling; we can only empathically identify with them. hence an empathic attitude is still the speaker's attitude, and the Lojban attitudinal system remains consistent. Note that there are cultures where it is taboo, or even impossible in the language, to express the thoughts/feelings of another person, on the grounds that this is either impossible or an invasion of personal space.

Summary of cmavo Changes in selma'o UI

Here is a list of changes to "selma'o" UI since the attitudinal paper, for those who track such things:

"lu'a" (loosely speaking) was based on "kluza", a malglico metaphor; it has been replaced by "sa'e" (based on "satci") with meanings reversed.

"jo'a" was introduced as the opposite of "na'i": it specifies that the text is correct as written, like English "[sic]". "na'inai" would mean the same thing, but seemed too confusing as an affirmation.

"pau" is an optional signal at the beginning of a question, and was omitted from the attitudinal paper in error. "paunai" signals a rhetorical question.

"kau" is attached to the focus of an indirect question: it does not connote knowledge particularly.

"e'e" was changed to "competence - incompetence".

"re'e" was added as a new category modifier, parallel to the "ro'V" series; it means "spiritual" and takes the place of old "e'e".

"vu'i" (virtue - sin) was changed to "vu'e" to match the new gismu "vrude".

"se'a" is a new attitudinal modifier meaning "self-sufficiency - dependency", based on demonstrated need in Japanese and other cultures.

"be'u" is a new attitudinal modifier meaning "lack - satisfaction - satiation".

"ta'u" and "ta'unai" were switched in meaning.

The former term "observational" has been replaced with "evidential", to agree with linguistics norms, and to avoid confusion with "observative".

"se'o" is a new evidential meaning "I know by internal experience (dream, vision, or personal revelation)".

"ka'u" is a new evidential meaning "I know by cultural means".

"su'a" is now both an evidential and a discursive, displacing the old discursive for "in general - in particular".

"ju'a" is a new vague evidential: "I state"; particularly useful in "ju'apei" = "How do you know?"

"bi'u" signals new information: "lebi'u cribe" is a newly mentioned bear, as distinct from "lebi'unai cribe" which is a bear we've heard about before.

"dai" newly assigned to indicate empathic identification of another's feelings.

"po'o" has been proposed as a discursive for the sense of "only" meaning exclusively, or uniquely, within a context. There is some debate about this addition, since there is no way to specify the context using the UI grammar.

Punctuation proposals from Nick Nicholas

To the current list of optional punctuation symbols, used to highlight sentence structure, I consider worthy of attention:

"!" for UI words. Given the presence of "." before VV UI-words, maybe limit his to CVV UI-words. "!ca'e", ".ui" or ".!ui" "{","}","[","]" to highlight structure of tanru and of various grammar constructs like POI-clauses. "le cmima {bele [{vofli bo minji} jeva'i vinji] jenmi be'o} {poi vitke loi xendo} cu bebna"

John Cowan has reemphasized the need for a symbol to indicate the start of a sentence, given that ".i" is not distinctive enough. The most appropriate such mark would be a section-symbol or a paragraph-symbol (respectively, the two interlocking S's on top of each other, and the reversed filled-in P ). Neither of these is ASCII. I don't see why we don't revive John Hodges's proposal, in JL10, that we revive the "=" for that purpose. If we need something chunkier, perhaps a "@" or a "#".

John Cowan comments:

These [use of "{","}","[","]"] are OK, but anyone using them must be warned that they never affect the official interpretations [should there be contradiction].

Chris Handley:

I tend to agree with John, but more strongly. In any situation where there are two ways of specifying something (structure, relationships, dates, whatever) one of them will be wrong sometime. How many times have you seen a notice of a meeting that said something like "Tuesday, 1 March 1993" and then missed the meeting because it was on the Monday?

Nora LeChevalier:

I am opposed to structure markings, because these break audio-visual isomorphism. All other optional punctuation marks in the language appear with a specific words that correspond, and hence are 'read off' by reading the associated bracket word.

If brackets are needed in writing these cases, what correspondingly distinguishes the grouping in speech?

le lojbo se ciska

Nick's Second ckafybarja Text

I have to admit that Nick Nicholas's proposals to use bracketing to make it easier for a reader to figure out a complex text structure might be useful, or even necessary, for Nick's writings. The following is Nick's submission for the ckafybarja project, an elaborate and stylistically complex character study.

I said that I would print all ckafybarja submissions so they can be evaluated by the community. Unfortunately, I have to admit that I could not read the Lojban, even with the bracketing that Nick inserted 'to make it easier'.

Unfortunately for Nick, I agree with Nora that Lojban's audiovisual isomorphism requires that the grammar be understandable based on what is supplied in the words themselves. Lojban's design presumes that all 'punctuation' is spoken. As such, punctuation that is inserted to make a text easier to read must be algorithmically derivable from the text structure itself. The bracketing that Nick included in the following text occasionally violated the grammatical structures of the language. For many of his markings, I saw no obvious explanation that allowing me to predict what bracketing he felt to be useful, and what he felt it was unimportant to include.

In addition to bracketing explicitly, Nick tends to write many cmavo as compounds when there is neither a grammatical link between the words, nor a common English word as translation. He and I clearly have different ideas as to what should constitute a Lojban 'word'. In one case he wrote "na'igo'i", which in a side comment he says is patterned after "nago'i". But "na'igo'i" is in error if he wishes the "na'i" to apply to go'i, so this kind of compounding must not be allowed to creep into the language.

[My own policy, rather utilitarian, is that a compound is a single word if it forms a gestalt image in the mind that is more than the components. To the extent that the gestalt differs from the components, the word needs to be put into a dictionary. If the compound is not made of words linked grammatically, a dictionary cannot define the word as having a single meaning - violating the Lojban design - and you may need to break the compound down into components in order to figure out what the role of the individual words is. A secondary factor is that automated processing of Lojban text, including the spelling checker I use in preparing JL and the Lojban glosser Nora is writing has trouble dealing with irregular compounds. It seems likely that learning Lojbanists will have the same problem - if a word is not found in standard word lists, many Lojbanists do not know what to do with it.]

This issue, I had to check to make sure any lujvo were properly formed, and update them to the new rafsi list. Irregular cmavo compounding made this work more difficult. When I see a compound like "na'igo'i" that counters grammatical sense, I have to rule out the possibility that it might be a mismade lujvo or a typo, omitting the hyphen 'r' that would make it valid. Since many rafsi represent gismu that are related in meaning to the cmavo of the same form, it is plausible that an irregular compound will be seem semantically plausible as a erroneous form for a lujvo. I am thus coming to believe that Lojban does not have the redundancy to support significant cmavo compounding. Even fluent speakers of a language make typos when writing, and learning Lojbanists (which all of us are) make even more typos, but also grammatical and lujvo-making errors, that irregular word forms can hide.

Thus, I removed all of Nick's markings, and expanded most of his compounds, prior to inserting my own efforts to structure the text. I then inserted those markers that I could come up with simple algorithmic rules for (I did this manually, so there may be some inconsistencies). New sentences are marked with an equals sign (=), per Nick's suggestion, and I also left 3 spaces before the mark. I added quotation marks, parentheses (and brackets for parentheses marked as editorial), and question marks for question words and exclamation points for attitudinals. I figured any more marks would make the text simply too punctuated, and indeed in places it seems to have exceeded reason already.

Nick's text unfortunately gave few clues for paragraphing. The unfortunate result was a block of Lojban that was extremely hard to read, even with (or especially with) the forest of punctuation marks.

Nick's style of quotation made it impossible to try to follow English-like practices of starting new quotations in a new paragraph. He has quotations in the beginning of sentences, in the middle of sentences, at the end of sentences. In one place he has a series of alternating quotes and names in a single sentence with no clue for the Lojbanist as to how to link the two (we have metalinguistic structures specifically designed to communicate the 'he said'/'she said' of conversation, but Nick did not use these.

Nick's parentheses are especially confusing - a parenthetical note attaches grammatically as a free modifier to the previous word, and Nick's placements often made no sense by this rule. If a parenthetical needs to be broken into a separate thought, as in Nick's long digression near the end of his story, it must be separated from the previous word by an ".i" (using Lojban metalinguistic markers to refer to the outside text as needed).

I decided to double indent paragraphs, and to single indent new sentences that were immediately followed by a start of quotation mark or which immediately followed after a quotation ended. This seems something an automatic algorithm can do, and it helps a little in making the text easier on the eyes, if not on the brain.

Nick's compounds are expanded unless they are compounds that would be joined by the lexer component of the Lojban parser (and sometimes I expand those, since lexer compounds can be arbitrarily long), or unless they are of patterns that have traditionally been written as compounds in Loglan/Lojban writings like "lenu" and "lemi", and "leca" (which usually means that they have a simple English word or phrase in translation that makes it easy to think of the compound as a unit). I generally separated indicators from the words they follow, whereas Nick generally writes them as compounds.

Until someone convinces me differently, I am going to take a hard-nosed attitude towards text structure. I need people to keep their style simple enough that the rules of the language convey what they are supposed to. I hope Nick and everyone else forgives what I did to his text. I hope this effort, if nothing else, leads to some agreements for the future on standards for text submission and for editing.

Nick's character sketches are certainly interesting, even if you need to read the English text. Good luck and encouragement to those who try the Lojban!

All footnotes are by Lojbab, except where marked otherwise.

kafybarja #2


®lu go'e =.ibaboke'u ko'u bacru ®lu ko seljde loi mabru li'u¯ li'u¯

=.i lei puze'a tirna cu milxe ke se cfipu cmila =.i la paul. bacru ®lu mabru tcini .!u'iru'e li'u¯ gi'e cevni melbi co dasni lo xekri birtu'ucau .!i'ero'u nercreka =.i ge lerci tcika vi le barja gi carmi melbi co xekri fa le tsani za'a loi selca'o nenri prenu =.i so'o ve barja mo'u cliva =.i la lizbet. na'e go'i cadykei be le xekri tedykre[1] be la paul. kalsa be'o se mlifanza cisma no'e zanru le xajmi =.ivu.!u'esaibo ti'e xekri kalsa tu'a loi juntytri .!ii poi vlipa joi vlile joi ke daspo joi finti vau .!u'e =.i ki vive'i kamjikca simsa go'i .!i'unai

no'i la liz. dasni lo grusi notcreka (to le no'a cu se kanla loi danmo blanu za'a toi) be ®lu lenu prami cu ca'e nu nelci carmi se trina lo prenu ju nakni ju fetsi vau !pa'ero'a li'u¯ ne loi lerfu co xekri =.i mi (to lego'i cu se kerfa loi na'e kalsa za'a toi) cairmau me leli'i grusi =.i grusi fa lemi plokarlycreka .e le palku .e le kosycreka noi jgena se dasni ru'u le xadmidju =.i su'o prenu cu ba'anaika'uta'o sanga bacru ®lu =.i RUSpre ce RUSta'u ce rusxirXEMkla li'u¯

=.i na'i go'i sa'e =.i {lu'e ry. ce'o .ubu ce'o sy.}[2] cu ka'u drani se basti {lu'e xy. ce'o. ebu ce'o ky. ce'o ry. ce'o .ybu} =.ita'ocu'i su'u xekri kei vi le kafybarja

ni'oremo'o su'o bevri cu .!a'acu'i masno kasydzu zo'i loi ve barja =.i la paul. cu tavla (to le no'a mebri cu jurja'o =.i le laurxampre pu'i vlipa .!i'e.i'onai cei bu'a[3] toi) fi leli'i gletro

=.i ®lu =.i mi du'eroi .!u'anaizo'o se gletro =.i mi purlamcte[4] seku'i go'i la liz. .!oinai .!u'i li'u¯ ®lu =.i ?xu purpla[5] go'i zo'o li'u¯ ®lu =.ipe'i .!ianai snuti li'u¯ ®lu =.i la paul. jikfazgau ?.iepei doi liz. .!u'iru'e li'u¯ ®lu =.i carmi jikfazgau ju'o .!iu =.i ko co'u xlapre .!u'i li'u¯ ®lu =.izo'o tu'a ko bapli .!e'inai li'u¯ (to bu'a .!o'e fi leka smaji ke lamji prami joi pendo noi su'anai se mupli na'ebo lecaca'a seltra .!i'o toi) ®lu =.i mi le'o go'i li'u¯

=.i la paul. ce la liz. co'a cisma simtipyda'a ni'a le jubme =.i la liz. (to gasta bo demxa'e ce margu bo jamfu ce xamsi bo kanla vau .!io toi) certu lezu'o ca'arcau damba =.i mi se mliburna ctacarna co na'eke ca'arcau damba certu gi'e zgana le barja ni'o le paltylu'i[6] ku jo'u le jukpa puza cliva =.ija'ebo le barja cu tatpi smaji =.iji'a le trixe be le barja be'o noi di'i krasi leka to'e cando gi'e kurfa kei ki'u lepu'u re ru vi ri zdidabysnu (to ®lu =.i do te sluji le birka lo mleca be la'e mi .!o'a li'u¯ ®lu =.i .!e'u mi'o cipra .!a'e birvrajvi =.i .!ai le pritu =.i do djuno .!o'ocu'i ledu'u le zunle pe mi tsame'a le pritu birka doi paul. li'u¯ ®lu =.i ?xu purpla go'i zo'o li'u¯ toi) tigni fi loi ve barja cu ca malmliselgu'i ke dukri'a kunti =.i la paul. jinga fi la liz. fe lenu birvrajvi =.i la liz. go'i fi mi fu'i (to ba'e dukri'a kunti toi) =.i lerci tcika vi le barja =.i mi'a pu'o jbuboikei

ni'ocimo'o le jatna ®lu =.i .!a'o do joi le pendo be do cu xaufri ca leca vanci li'u¯ mi ®lu =.i go'i .!io =.i ca pamoi zu'o mi vitke le barja ca lo relmoicte[7] =.iza'a .!u'eru'e lei ve barja cu clira cliva ca le cabdei li'u¯

=.i le re jibni be mi depcni catlu le jatna =.i le jatna ®lu =.i go'i ki'u leka lei cibdei na'o cabdei lenu mutce gunka kei vi levi tcadu =.ita'o do noi ta'e klama le barja ca lei xavycte cu punai pe'i penmi la xiron. noi vi sidju li'u¯

=.i la paul. ce la liz. smaji casnu lenu ri jo'u ra ba litru la'e le merko =.i le jatna cu degji jarco le clani ke blabi creka xadyti'e be le cnino be mi gi'e cisma bacru ®lu =.i .!ai mi bazi benji ri do ge'e li'u¯ gi'e cliva =.i la liz. bacru ®lu =.i do li'a selxagmau[8] mi'a tu'a le bangu .!o'o li'u¯

=.i le re se cimei na lojbo =.i mi ®lu =.i nu vlipa jivna zo'o =.i mi jitro joi seltro li'u¯ la paul. ®lu =.i ca ro nu za'u prenu cu simfra cu nu vlipa jivna ru'a =.i go'i cu'u la djen. vecu'u le samsnuci'e =.iseni'ibozo'o. .!iecu'i go'eje'u li'u¯ mi ®lu =.i ?xu purpla go'i zo'o li'u¯ la paul. ®lu le xaupre za'ota'e bacru ®lu ?xu purpla li'u¯ li'u¯ mi ®lu =.i ?xu purpla go'i zo'o li'u¯ la liz. ®lu =.i co'a lerci =.i doi paul. do pu nupre lenu mi'o clira sipna =.i mi cu'urzu'e co bavlamdei li'u¯ la paul. ®lu =.i .!u'i ?xu purpla go'i li'u¯ mi ®lu =.i .!ua mi se sitna li'u¯

[tosa'a lemu'e sitna na dunli lemu'e xusra =.i la paul. cu nalri'i bacru do'i[9] pe zo ®simfra¯ gi'u xusra =.i loi cmavo be zo ®zo'o¯ na'o banzu lenu lo te sitna lo se xusra cu frica =.i lemu'e mi se sitna cu te ciste lo pemci joi kelci jenai xusra plitadji be la paul. bei le bangu bei lenu jikca pluja =.i na nibli fa le nunsitna lenu morna sinma =.i na nibli na'ebo le sego'i .!u'i =.i mi mutce mezo®to¯ tavla =.i ?xu !se'izo'o purpla go'i toi]

la liz. ®lu malxlu zo'o li'u¯

=.i lerci tcika vi le barja =.i mi'a puba'o jbuboikei =.i lei bevri cu .!a'acu'i masno bo kalsydzu fa'u sutra bo kalsydzu fa'u cando =.i casnu loi sancrfrikative .e loi relcinpampre girvlici'e .e loi nalzva pendo ca'o le nicte noi sruri be lo ba'a vu trene co pelxu gusni nenri pamei ke sirji darno xemkla zmitra ke snura grusi nalkalsa kunti be'o .!uo xekri

Translation of Nick's Coffeehouse Text


"That's right. And then he says, 'Beware of the mammals.'"

Those who have been listening smile with mild confusion. Paul says "So, it's a mammal kind of situation!", and is godlike-beautiful in his black tank-top (mmm...). It's late in the cafe, and the night is pitch-beautiful dark to those inside, on the other side of the window. A few cafe patrons have already left. Lizbet, who hasn't, toys with the chaos of Paul's hair, smiling slightly annoyed in disapproval of the joke. Far, far away, I hear, there are black chaoses of gravity, that strongly and violently both destroy and create! Right here and now, social-wise, something similar is happening...

Now, Liz is wearing a grey T-shirt (her eyes are smoky blue, I see), saying "Love is an intense fondness and attraction to a person whether male or female!" in black letters. I (her hair is not a chaos) am more into greyness. Grey are my shirt and my pants and my sweater tied around my waist. It has been sung, I recall, in my culture: "For grey he was, and grey he wore, and grey too was his steed." Actually, not precisely so. The string "G.R.E.Y." should be replaced with the string "B.L.A.C.K.".

To sum up (or to expand!), there's a blackness going on in the cafe.


Waiters, I suppose, are ambling slowly past the patrons. Paul is talking (his brow looks serious. The loud joker has been known to show strength - how I envy!) on topping. "I get topped too often, I'm afraid. But I did top Liz last night! Hehehe!" "On purpose? :) " "Oh, I think it was an accident!" "Paul is being a pest, don't you think so, Liz?" "Quite a pest! Stop being a bastard, love!" "Oh yeah? Make me!" (... he has been known to show strength in a quiet, close love/friendship - which is not exemplified by this behavior in particular!) "I will!"

Paul and Liz start smilingly kicking each other under the table. Liz (fists of steel, legs of mercury, eyes of the sea...) is an expert in self-defence. I, not being an expert in self-defence, turn around in slight embarrassment and observe the cafe.

The dish-washer and the cook have left. As a result the cafe is tired-quiet. Also, the back of the cafe, normally the source of bustling and comfort because of the two of them debated there for our -

("Your biceps are smaller than mine! Ha!"

"Yeah? Let's test them! Arm-wrestle. The right! You know my left is weaker than my right, Paul!" "On purpose? :) " )

- amusement, is now ill-lit, and anguishingly empty. Paul beats Liz at arm-wrestling. Liz beats me, surprise surprise. (Anguishingly empty.) It's late at the cafe. We're about to play pool.


The Manager: "I hope you and your friends are enjoying the evening?" Me: "Indeed, sir. This is the first time I've been at the cafe on a Tuesday. I see the patrons are leaving early today!" My two neighbors patiently look at the manager. The Manager: "That's because Wednesdays get quite busy in this town. By the way, since you usually come into the bar on Saturday nights, you will not have met Xiron[10], who has been helping out here."

Paul and Liz are quietly talking about their trip to the States. The Manager points out to me the long, white-shirted back of someone new to me and smiling says: "I'll (hm...) send him to you later", and leaves.

Liz says "You... clearly have the advantage of language over us." Two of the threesome do not speak Lojban.

Me: "It's power conflict! I top and am topped."

Paul: "At any time more than one persons interact, there is a power conflict. Jen says so on the electronic news, so it must be true!"

Me: "On purpose?"

Paul: "Our good man here has been saying 'On purpose' a bit too long."

Me: "... On purpose?"

Liz: "It's getting late. Paul, you promised we'd get to bed early. I'm busy tomorrow."

Paul: "On purpose?"

Me: "Aha! I've been quoted!" [Editorial digression. Quotation is not equivalent to assertion. Paul informally utters the "Interacts" sentence, independent of whether or not he is asserting it. My being quoted is part of the poetic, or playful, rather than assertional usage of language by Paul to make his social interactions complex. The quotation does not imply emulation. Nor does it imply non-emulation! I use parentheses a lot. On purpose? :) ]

Liz: "You're a bad influence. smile"

It's late at the cafe. We have been playing pool. The waiters, I suppose, are ambling slow and ambling fast and idling. We're talking fricatives and bisexual politics and absent friends during a night that, surrounding a distant putative train, lonesome yellow lit interior / direct distant vehicle automaton / secure grey unchaos empty, is (THE END) black.


  1. Nick translates this as "chaos", for which he used the gismu "kalsa" elsewhere in the piece; I get nothing from the metaphor "earth-hair".
  2. These strings could have been done more clearly using the Mex grammar, which allows you to talk about strings of letters and numbers as strings. "me'o ry.ubusy." and "me'o xy.ebukyry.ybu" would be the corresponding string expressions. Since lerfu used as sumti (as is the case in this text) are presumed to be anaphoric abbreviations, rather than literal text, this version really isn't correct, though it can be figured out.
  3. This usage is wrong. "bu'a" is one of the existential predicate variables, equivalent to "da" for sumti. Acting like "goi" does for sumti, "cei" is the selbri assignment marker used to assign values to the unbound selbri variables of the brodV-series. The latter series corresponds to "ko'a" series for sumti, and not for "da" series, and is clearly what Nick intends in this usage, since he anaphorically repeats the bridi of this sentence in the next parenthesis by back reference to "*bu'a".
    On the other hand, the mechanisms available for defining or restricting bu'a series variables are relatively undefined.
  4. The lujvo-scoring algorithm given with the rafsi lists this issue would give a slight preference to "prulamcte" over "purlamcte".
  5. The lujvo-scoring algorithm given with the rafsi lists this issue would give a slight preference to "prupla" over "purpla".
  6. I would have used the more general "ctitcilu'i" for "dish-washer".
  7. This one lost me for a little bit, since the names of the days of the week do not include the rafsi for "moi", and Nick did not use "moi" elsewhere in the story for "Saturday night". (Actually, the English translation doesn't mention it being night, but the previous sentence mentions evening. Since we worked hard to give Lojban culturally neutral definitions for the parts of the day, word choice here could be significant to some.)
  8. The lujvo-scoring algorithm given with the rafsi lists this issue would give a slight preference to "selxaumau" over "selxagmau".
  9. This is "dai" on older cmavo lists; see "dai" in the list of new members of UI elsewhere in this issue.
  10. The Lojban is obviously a reference to the character proposed by Veijo, and described in JL17. Apparently Nick votes in favor of Xiron (though he inexplicably spelled it 'Chiron' in his version of this English translation). Nick appears to add the stipulation that Saturday is Xiron's regular day off.

More Usage Questions

Following are essays on usage questions that are perhaps less technical, and have not led to significant proposals for change. In most cases, they are further explanations of usage issues discussed in earlier publications.

Dean Gahlon asks a simple question

This is a very basic question; hopefully, the answer will also be simpler. The canonical form of a Lojban sentence seems to be something like this:

le nanmu cu citka le cripu
The man eats the bridge. 

My question is: are the following two forms equivalent to this (as I think they should be, given my understanding of place structures), or have I missed something?

citka le nanmu le cripu
le cripu se citka le nanmu

Also, if these are correct, are there any other variants on the sentence that are grammatical? (And yes, I am aware that the event described in these sentences is rather unlikely, but I wanted to keep this simple, and there appears to be no gismu for "bicycle".)

Lojbab responds:

In each case: almost, but not quite, equivalent.

Starting with "le nanmu cu citka le cripu". This is identical to "le nanmu le cripu cu citka", and both have the same x1 and same x2.

To put both sumti after "citka", you must mark the first, because Lojban assumes that if there is no sumti before the selbri "citka" that you have omitted the x1. You must thus mark the x1 place with "fa" which says that the following is the x1 place:

citka fa le nanmu le cripu

Using "fe", the marker for the x2 place, you can derive even more forms basically mixing "fa le nanmu" "fe le cripu" and "citka" in all combinatoric orders, inserting a "cu" if either of the sumti is before the "citka".

All of these are equivalent in a broad sense, the difference being one of emphasis: the thing at the front of the sentence is typical the thing of highest emphasis, and the thing at the end of secondary emphasis. The rules for emphasis are pragmatic mostly, and are based on our experiences rather than a formal prescription.

If you insert "se", the result is a 'conversion' and 'equivalent' becomes a trickier proposition.

le cripu cu se citka le nanmu 

(note that "cu" is needed) expresses the same relationship as the above sentences, but there is a minor difference in that the labels 'x1' and 'x2' are reversed, and you have to use "fa" and "fe" appropriate to the new numbering to rearrange the terms, but all of the options listed above are still possible, with "se citka" as the central selbri.

There is some question whether a conversion 'means the same thing', though, because the other things you can do to a converted predicate have different meanings: "le citka" (the eater) is different from "le se citka" (the thing eaten) in a later back reference to the above sentence relationship.

There is some question whether "le nu citka" and "le nu se citka" have the same meaning, with or without the x1 and x2 filled in. Again, they abstract the same relationship, and the resulting 'event' being described is the same event. But pragmatically, we would often construe different meaning to the use of one over the other.

Mark Shoulson gives his answer:

In Lojban, the order of sumti with respect to selbri is fairly free. The usual way of doing things is, as here, in "SVO" form (scare quotes because it's not really applicable in Lojban): x1 place, then selbri, then remaining sumti. The other common form is "SOV" form: "le nanmu le cripu cu citka". This is also fine. Presumably, with many sumti, there's nothing wrong with putting the selbri anywhere among them (but see below). So, "mi le briju cu klama le zdani" ("I to the-office go from the-nest") is OK, too.

Using "VSO" form, "citka le nanmu le cripu", is quite grammatical, but poses a different problem. By current usage, since VSO is not a common word-order in many languages, the "selbri-first" word-order is reserved for "observative" sentences - ones with the x1 place ellipsized. Thus, the above sentence would probably be understood to mean "(something) eats the man ??? the bridge" - since "citka" only has 2 places, it would be unclear how the bridge related to it all.

As to using "le cripu cu se citka le nanmu" (the "cu" is necessary here, otherwise we get "the bridgish eaten-thing"); that's another bit of hairy semantics. I like to consider it quite the same as "le nanmu cu citka le cripu", but even I, like most others, often consider a SE-converted selbri somehow to have a different semantic loading than an unconverted one. So, when I hear "se citka" I think "is-eaten", and thus would get a different meaning for "le cripu cu se citka [zo'e]" as opposed to "[zo'e] citka le cripu", even though both have the same brivla (citka), and the same sumti ("zo'e" [elliptical "it"] in the eater position (so to speak), and "le cripu" in the eaten position).

'Course, you may not have gotten up to this yet, but there are other ways to mangle the word-order in a Lojban predication. There's selma'o FA, which allows totally free reorganization (basically, the chief words in FA are "fa", "fe", "fi", "fo", & "fu", which mark the next following sumti as belonging in the x1, x2, x3, x4, & x5 places of the current bridi, respectively. Following a FA-marked sumti, subsequent unmarked sumti are considered to continue sequentially from the point specified by the FA.) Needless to say, this allows you to construct truly confusing sentences, put more than one sumti into the same place with no conjunction, etc.

SVO Order in Lojban

JCB's Rationale, with commentary by John Cowan, Colin Fine, Lojbab

During a computer network discussion of word order in constructed languages, the rationale for the predominant SVO (subject-verb-object) order used in Loglan/Lojban came to be discussed. This article summarizes that discussion. For those unfamiliar with the grammatical word-order terminology, with regard to Loglan/Lojban it is generally presumed that the selbri is the "verb", the first sumti is the "subject", and all other sumti are the "object". The reasons for this will come out in the article. Note that Colin's 'proposal' is one that is proposed quite often, and the commentary may thus help people better understand the rationale for the current design.

Any such discussion must start with the original rationale for Loglan, which was that of James Cooke Brown (JCB). The following text was originally written by JCB in 1967-68, published as part of Chapter 6 of his book Loglan 2: Methods of Construction, and reprinted in The Loglanist 1:2, p. 54ff. These publications are long out of print and hard to find.

John Cowan(JC):

It provides an interesting insight into the mind of a language designer at work.


[JCB begins by defending SVO as the order of choice because of its prevalence in Chinese, English, Russian, Spanish, French, and German, 6 of his 8 source languages.]
"There was a time, however, when [VSO] order was seriously, if briefly, considered for Loglan. This order has a certain traditional charm for logicians - witness the standard schematic notation 'Fxy' for a two-place predicate, for example - and for certain purposes of manipulation it has undeniable advantages. But for a spoken and, at the same time, uninflected language the VSO order turns out to be quite unsuitable. The argument which discloses that result may bear repeating here.
We note first that, on the most fundamental grounds, arguments are not to be distinguished except by word order in Loglan. Thus we entertain no "case endings", or other marking devices, by which "Subjects" can be intrinsically distinguished from "Objects".[1]

One form of the argument then hinges on the management of imperatives.[2]

John Cowan: Both Loglan and Lojban have to some extent withdrawn from the original rejection of case marking, and have created a set of optional case tags. However, neither form of the language uses them much. In Lojban, the argument about "imperatives" which follows must be replaced by an exactly parallel argument about "observatives", since Lojban interprets a V-first sentence as an elliptical subject without imperative coloring. I have added bracketed comments to the next paragraph giving the Lojban, as distinct from the Loglan, viewpoint.

JCB (cont.)

Now imperatives [Lojban: observatives] are almost invariably short forms; there is apparently little scope for long-windedness in giving warnings or commands [Lojban: drawing the hearer's attention to things in the environment]. Moreover, the first argument of an imperatively [Lojban: observatively] used predicate is almost always the hearer [Lojban: understood from context], and as the omission of any constant feature of a message cannot reduce its information content, first arguments are nearly always [Lojban: always] omitted in the imperative [Lojban: observative] mode (e.g. as in English 'Go!' - [Lojban: 'Delicious!']). But if we omit the first argument from the form PAA (Predicate-Argument-Argument) - for arguments, note, are to be taken as indistinguishable - we obtain a result that does not differ from the result of omitting a second argument, or a third. Therefore the adoption of the PAA schema as the standard order for the Loglan sentence deprives us of a good way of defining imperatives [Lojban: observatives]. In fact, it deprives us of the only way of defining imperatives that is consistent with the other patterns of an uninflected language.

John Cowan: Lojban resolved this by making use of a special "imperative 2nd person pronoun" which may appear as any argument, thus permitting more complex imperative forms while remaining "uninflected". This enabled us to use a missing argument to indicate an observative.

JCB (cont.):

Similar difficulties arise with specified descriptions. Thus if 'He gave the horse to John' is to become something like 'Gave he the horse John', how do you say 'the giver of the horse to John'? A form like 'the give the horse John' will not do, since it is the designation of the giver, not the gift, which normally follows the predicate. Only by introducing some sort of dummy argument into the 'Fxyz' form, e.g. 'F-yz', can we keep the meaning clear. But this is awkward. These seemed good reasons not to use the VSO form, especially as the SVO form does not suffer this disaster. Thus, the schema APA yields an unmistakable PA in the imperative [Lojban: observative] mood.

Incidentally, the SOV order ('He the horse John gave') collapses into the same kind of ambiguity under the pressure of abbreviation. (Is 'The horse John give' an imperative, or an incomplete declaration?) Thus, curiously enough, and independent of any facts about the distribution of these arrangements among languages, we would have been forced to abandon the logicians' notational convention anyway. For once incomplete or abbreviated forms are considered - and in a spoken language they are far more frequent than unabbreviated forms - the predicate can no longer be treated as a prefix or a suffix of its uninflected arguments ('Fxy' or 'xyF') but must be treated as an infix ('xFy'). It is only of such initially infixed arrangements that the fragments left by the removal of uninflected arguments (e.g. 'xF' and 'Fy') remain reconstructible and, hence, grammatically clear.[3]

Colin Fine then commented on JCB's rationale:

It is remarkable how weak these arguments are, from the perspective of 25 years later.

Consider the following.

- 1. The major justification was in terms of imperatives.
This was a strong argument as long as "the only way of defining imperatives that is consistent with the other patterns of an uninflected language" was to omit the leading argument. But as John points out, we have an elegant and flexible alternative method. (JCB's original argument about imperatives stressed the importance of minimal morphological material in them, and gave examples from natural languages; but in fact there are plenty of contrary examples with more morphology in them, such as polite imperatives in German "gehen sie!".)
- 2. Given that the omitted first place now signals an observative rather than an imperative, the argument becomes feeble. Even if observatives had continued to be used as apparently intended, statements such as "there is apparently little scope for long-windedness in ... drawing the hearer's attention to things in the environment" are highly dubious. It is true that there are short observatives ("Delicious!") but equally there are long and tortuous ones ("A man on a unicycle eating cream cakes!"). Furthermore, I observe that 'observatives' are not in practice limited to this use in current Lojban writing and speaking, but that lojbo feel free to omit the x1 in just the same way as they do any other argument. Indeed, constructions like
"cumki falenu ..." (it is possible that ...)
where the x1 is postposed by an explicit x1 marker ('fa'), are syntactically equivalent to observatives, and not unusual with words like 'cumki'.

I would analyze the current situation in Lojban thus:

  1. A bridi consists of a string consisting of zero or more terms (optionally tagged sumti) and one selbri. The selbri may occur first, last, or between any two terms.
  2. The case where the selbri comes first has some special properties of interpretation (below), and is therefore treated as a special construction, called, for historical reasons, 'observative'.
  3. An untagged sumti S is interpreted as follows (ignore all terms tagged with BAI, tense or FIhO in this):
    1. if the preceding sumti is tagged with an explicit positional marker (FA) indicating the Xn place of the selbri, or is interpreted by recursive application of these rules as filling the Xn place, S fills the X(n+1) place. b) if no sumti precedes, S fills the x1 place except in the case of an observative when it fills the x2 place.
  4. (a stylistic or discourse observation) a syntactic observative (with x1 unstated) is often appropriate for uses that might be referred to stylistically as observatives, such as "kukte" ("Delicious!").

But it is equally useful where the x1 is omitted because pragmatically reconstructible (for example in narrative: "la maik. mu'o klama .i rinsa mi'a" ("Mike arrived. [He] greeted us.") ) or for structural reasons to do with clause weight ("cumki falenu loi xarju cu vofli da'i" = [it is] possible that pigs might fly).

Lojbab: There are stylistic and pragmatic uses for the "observative" word-order/x1- omission other than spontaneous, brief observation ("Delicious!"). But the latter was the justification for providing the short form.


Thus, while observatives currently exist as a distinct grammatical structure in Lojban, they are distinguished only by a special rule of default interpretation. The argument originally advanced in respect of imperatives really does not seem to have any weight once transposed.

The second argument advanced was in respect of selgadri (specified descriptions) [ed. note: sumti of the form "le {selbri} be {x2 sumti} bei {x3 sumti ...} be'o"]. Remarkably, this argument is actually stronger in respect of Lojban than it was for Loglan (at least when I knew it, in the late 70's) because Loglan then had a series of words that meant "befe, befi, befo, befu" i.e. the links indicated the place of the following argument. (There was no 'bei' equivalent). Given this, his argument that "the give the horse John" could not be interpreted as "The giver of the horse to John" because there was an omitted argument, is simply false.

Lojbab: Actually, Lojban "be" is the exact equivalent of the first of these, and "bei" the second of these, provided that there is no use of the fa/fe/fi-series of tags. Loglan eliminated the higher-numbered places in the early 1980s, combining them into "bei", as part of the development of the unambiguous machine grammar, as part of the recognition that sumti numbering need not be a function of the syntax (i.e., that the grammar should not be counting the number of sumti in a bridi - in other words, that you did not need separate grammar structures for 2-place bridi, 3 place bridi, etc.). This was still in the 70s, I think, but it might ave been 80- 81. Older versions of Loglan rarely made use of omitted sumti (at least partially because so little text of any complexity or naturalness was written), so it was never analyzed in the 70s version, how, for example, you would skip the x2 sumti in a specified description. You could not merely leave out the "befe" equivalent term and jump to the "befi" term. So older Loglan is really the same as Lojban with regard to the argument that follows.

Colin (cont.):

In current Lojban, the argument does have some weight, since "be"/"bei" are merely syntactic glue, and do not specify the role of the following term. However, it is not convincing, for the following reason:

At present, as sketched above, there is a rule of interpretation which says that if the first unmarked sumti in a bridi follows the selbri, it is to be taken as the x2, not the x1. There is no a priori reason not to apply the same rule to linked sumti - except that it would be simpler, because there are only following sumti.

In short, a VSO version of Lojban could be created by making two changes to interpretation, and no changes to syntax, viz.:

  1. In a bridi, the first untagged sumti is always the x1, whether it precedes or follows the selbri;
  2. In a selbri with linked sumti, the first untagged sumti is the x2, and the meaning of the selbri as a taurpau or selgadypau (tanru or description component) is the x1. To specify the x1 (meaningless in a selgadri), FA must be used.

The first removes a complexity from the current rule, the second inserts it back in elsewhere.

The effects on usage would be:

Current                   VSO

1. Normal bridi with leading sumti would not be affected:

mi viska ta or mi viska   ta viska mi ta

2. True observatives with no positional sumti would not be affected:

kukte carvi vi lei bartu  kukte carvi vi lei bartu
or                        or
vi lei bartu ku carvi     vi lei bartu ku carvi

3. True observatives with following arguments would require a FA:

batke le gerku            batke fe le gerku
or                        or
ta batke le gerku         batke ta le gerku

4. bridi with omitted x1 would require a FA:

.i suksa bacru di'e       .i suksa bacru fe di'e
or                        or
.i suksa bacru ri di'e    .i ri suksa bacru di'e

5. selgadri with linked sumti would not be affected:

le batke be le gerku      le batke be le gerku

Of the two patterns which would require change, I believe 3. is very rare. 4. is undoubtedly common in current writing; but it is also very common to omit the x2, even when there is an x3 - we are used to using FA a great deal.

I am not actually advocating this change. But I think it would be perfectly workable, as well as slightly more elegant. But the arguments against it are very weak indeed.

John Cowan on this proposal:

Not really enough. Consider the very common form:

le prenu poi klama le zarci cu blanu 
the man who goes to-the store is-blue

Under a VSO interpretation, "le zarci" would be the x1 place of "klama", not the x2 place. By the way, this goes literally into 4th-edition Loglan as:

le pernu jao godzi le marte ga blanu

and so we see that JCB isn't even consistent: within a relative clause, omitting x1 has no imperative sense.

Colin (cont.):

Some further observations on current Lojban:

1. I assume that a bridi which has tagged terms (but not FA) preceding the selbri, and untagged ones after, is still technically an observative, and interpreted according to the observative rule, i.e.:

ne'i le purdi ga'a mi mu'i leza'i birti kei cu preti ta mi


In the garden, watched by me, in order to be certain, (something was) a question about that to me.

rather than

... that was a question about me. Thus my account above is not complete.

2. Thus the observative rule applies when there are no untagged or FA-tagged sumti preceding the selbri. This is a different rule from that for "cu": "cu" is permitted if and only if there is at least one preceding term, of any kind. I have more than once tripped over this rule - I don't see why you should not be permitted to use "cu" with an initial selbri if you wish - but as it stands there is a rule, and these two rules which you might have expected to coincide in their application do not. (On the other hand, one is purely syntactic, and the other interpretive, so there is no a priori reason why they should agree).

Lojbab: The reason is that CU is grammatically a separator - it comes between leading terms and the selbri. If there are no leading terms, there is nothing to 'separate'.

There really is no relation between the two rules. The interpretation of how sumti are to be counted in complex sentences is a semantics convention - one which probably could have gone either way. It is not covered by the formal grammar, whereas the locations where CU is permitted is specified by the grammar. We chose to make the language totally transparent to tagged sumti in counting regardless of where, or what, they are, as an aid to teaching. (Note that a sumti marked with a case tag is automatically not a numbered sumti, even if the case is merely marking a semantic role normal carried by a sumti in the bridi.

Thus, in:

bau la lojban. mi [cu] tavla
In language Lojban, I talk. 

untagged "mi" is still x1, even though the language (which is x4 of "tavla" when unmarked) is specified first, because the case tag is present. By comparison, if the x4 sumti is tagged with the x4 marker "fo", you need to use "fa" (marking x1 on "mi", or the latter would be understood as (the undefined-for-"tavla") x5:

fo la lojban. fa mi [cu] tavla
In language Lojban, I talk.


  1. I leave the argument behind this remark, however, to the reader.
  2. It could as well be based on specified descriptions; see below.
  3. In these analyses, by the way, we may have isolated the ambiguity-avoidance mechanism behind one of Greenberg's most interesting universals, namely that all SOV languages have case systems (his Universal 41). I am surprised that the principle does not hold for VSO languages as well. If it did, we should then have strong evidence for the even more interesting converse principle that only SVO languages can be analytic: a fact we suspect anyway, but we would then know why.

And Rosta on "se", "te", & lujvo

jerna x1 earns x2 for work x3

"le se jerna" can mean anything that is earned. Suppose one wanted a lujvo specifically meaning "wages": could "seljerna" be such a lujvo? (i.e. does it have to be synonymous with "se jerna"?)

If "seljerna" needn't be synonymous with "se jerna", then, I wonder, is there a way of forming a lujvo that yields the x1 place of the source gismu, but isn't synonymous with the source gismu? Put another way: if "le se jerna" doesn't equal "le seljerna" then "le jerna" doesn't equal ________? (what corresponding lujvo)

If "le se jerna" = "le seljerna", then why is Colin always using seljerna-type lujvo?

Lojbab replies:

"le se jerna" need not be identical semantically to "le seljerna", but it will probably be close and nearly always interchangeable, probably an idealized value. A good example is Mark Shoulson's "selpinxe" (beverage) vs. "se pinxe" (something drunk). For "seljerna", I would presume that if one wanted to be specific that it was money that was earned, you could add "jdini" (money) or "pegji" (pay) to the compound, but given the stylistic bent people have these days for omitting such info where it is obvious from context, I can see why people would not. Therefore it is safe to say that at this point it is not yet clear whether "seljerna" is limited only to monetary wages, but that Colin probably does not want the value to be as broadly construed as "se jerna" might allow.

In this case, I tend to rely on my English instincts: if what I am translating is a single word in English, I am more likely to use a "seljerna" lujvo, whereas if it takes a phrase to say it in English, and the Lojban isn't exactly a paragon of trailblazing eloquence, I am more likely to leave the "se" separate. The fact that not all concepts that might be thought of as "se jerna" will also be "seljerna" is a natural consequence of the fact that a new word has been created. "seljerna" of course exists because there are times when you want to make a lujvo in which it is important to make it clear what aspect of a selbri modifying another. For train travel,

"selkla stana" (destination station) is clearly different from "terkla stana" (origin station). For such longer metaphors, you don't want to be stuck with the length of a tanru expression for everyday usage, and you don't want to be stuck with the place structure of the final lujvo term. So you need to be able to make "se"-based lujvo.

Any restriction from "se jerna" to "seljerna" is vague. And's question seems to be whether we have a similar short lujvo form that makes a vague restriction on the gismu itself. The answer is that we do not, since no one has suggested why such would be useful. Most often, when you make a lujvo, you have a specific concept in mind, and are going to choose a word that conveys that concept clearly, but briefly. "seljerna" does convey some information, that it is making a relationship involving something that is also the x2 of a similar relationship involving "jerna". If you want to make a word that in some way restricts the concept of "jerna", you will naturally make a lujvo that suggests something about what kind of restriction you have in mind.

The only other reason for And's suggestion, which came up in ensuing discussion of this topic, seems to have been that a symmetry is lacking without such a form. We really don't want to use up cmavo and rafsi space for the sake of idealized symmetries.

On the Grammar and Range of Free Modifiers

['Free modifiers' (the rule labelled "free" in the E-BNF) are the grammar structure which includes discursives, vocatives, subscripts, metalinguistic comments. Put briefly, free modifiers work like attitudinals, and modify the previous word of the sentence, or modify the whole sentence if found at the beginning of the sentence. Free modifiers would be as entirely free as attitudinals as to where they lie in a sentence, except that they have internal grammar (sometimes quite complex), and that grammar can interact in complicated ways with the grammar in the surrounding sentence. Thus, in the Lojban design, we had to limit the places where free modifiers could occur to specifically enumerated places (The list gets occasionally extended because someone thinks of a new place they want to use a free modifier, and John Cowan is able to successfully get YACC to accept free modifiers in that situation. This type of change has been a substantial fraction of the grammar changes approved in the last few years.) At one point, free modifiers were much simpler and more restrictive, and included the set of attitudinals, which now can be located anywhere in a sentence. Jim Carter proposed making free modifiers the grammatical equivalent of sumti. We chose the attitudinal model instead, and this essay discusses that decision.]

Free modifiers (and attitudinals) were never considered the equivalent of sumti for Lojban because they inherently modify the previous structure (except at the beginning of the sentence). They are thus more like the attitudinals, which we keep distinct and grammar free - more on this in a moment. Free modifiers include subscripts, and there are innumerable reasons in Lojban to use subscripts metalinguistically in ways that a sumti attachment would simply not support. The grammatical free modifiers are those with sufficiently complex grammar to require parsing, and hence cannot be totally free, but we remove as many constraints as possible (Loglan IS about removing unneeded constraints).

On the whole, though, Jim will find these grammatical free modifiers to be not all that unlike sumti in their grammatical location - but they group differently in the sentence than sumti.

Attitudinals are intended to be grammar free expressions because for the most part they are intended to be at the subliminal level. Like the hesitation noise, .y. and the English "you know" (Lojban pei?), these are to be stuck in where they fit, where you feel the intuitive need to express them. Unlike Carter, we do not feel these are abbreviations for claims; they are expressions. They are the equivalent of tone of voice, which in English and most other languages is controlled down to the word level or even more refined. (The Joy of Yiddish starts off with a sentence with contrastive stress applied in something like a dozen different places in the sentence to get different semantic interpretations of the sentence. Each Lojban attitudinal has that power.

Try an experiment. Take any short Lojban sentence that you can understand the grammar of. Take say 3 or 4 different attitudinals expressing a variety of emotions. For each attitudinal, and for each word position, insert the attitudinal and try to figure out what it means.

Here try:

 mi dunda ti do
 I give this to-you.

with attitudinals chosen from .iu (love) .oi (complaint) .ui (happiness) .uu (pity) .u'u (regret) .ue (surprise) .auro'u (sexual desire)

For each attitudinal, there are five positions. Try interpreting the effects in the sentence of one or two of these attitudinals. A brave soul can try two attitudinals in different places in the sentence, which is also permitted. e.g.

.ui mi dunda ti do
Happily, I give this to you.

mi .ui dunda ti do
I'm so happy it was ME who gave this to you.

mi dunda .ui ti do
I'm GIVING this to you, and happily (Did you think I could charge you for it?)

mi dunda ti .ui do
I'm giving THIS (my dream gift for you) to you.

mi dunda ti do .ui
I'm giving this to YOU (who makes me so happy)

(This exercise is a good way to practice and learn attitudinal words, if you limit yourself to a small number at a time.)

Now of course in this sentence, all positions in the sentence would allow you to grammatically add a tagged sumti. A tagged sumti in an odd position can add emphasis to other adjacent words too, and by convention often seems to emphasize the previous word like an attitudinal does. But this added emphasis is quite minor, and open to a wider variation of interpretation than the corresponding English, since other reasons besides emphasis can justify where a tagged sumti is inserted:

ca le cabdei mi dunda ti do
Today, I give this to you.

mi ca le cabdei cu dunda ti do
I, today, give this to you.
(Another day, someone else?)

mi dunda ca le cabdei ti do
I GIVE today this to you.
(Another day I might take it?)

mi dunda ti ca le cabdei do
I give THIS today to you.
(Another day something else?)

mi dunda ti do ca le cabdei
I'm giving this to YOU today
(Another day someone else?)

It's trivial to change the sentence to one where this isn't so:

mi dunda le xunre cukta do
I give the red book to you.

where a free modifier/sumti inserted after xunre would violate Carter's proposed constraint. (But I want to say how much I love books that are red when I tell you about my gift. Who are you to tell me I'm not allowed to do so?)

Comments on the Tense System

by Greg Higley

Below are a few short comments on the tense system. But I would first like to congratulate John Cowan and any others who worked on it. It is brilliantly designed, flexible, and fascinating! It took me no time at all to understand it, with one exception which I have noted below.

One thing that I think should be pointed out more clearly is that the new usage of selma'o VA is going to alter the way it is used as sumti tcita. (I am not assuming you don't already realize this: I just think it should be made more clear to those who might not.) Remember that it is no longer the spatial analog of selma'o PU. FAhA is the proper spatial analog of PU, while ZI is the analog of VA. As you well know, "zu'avi" means "a short distance left": "vi" means "a short distance [from the origin, in the direction specified, if any]". Therefore, "vi le tcadu" doesn't mean "in the city" but "a short distance from the city". The spatial relation analogous to "ca" is "bu'u", which, along with "ne'i" is probably best for "in/at":

"bu'u le tcadu"
"in the city";
"bu'uvi le tcadu/vi le tcadu"
"a short distance from the city";
"bu'uva le tcadu/va le tcadu"
"a medium distance from the city";
just as "ca le djedi" means "in the day"

- in all of these examples we could have used "ne'i" as well as "bu'u", although they aren't always interchangeable.

One thing that you may consider changing is "te'e" "bordering". I suggest putting this in selma'o VA, where it might prove more useful. (Although I could be misinterpreting its meaning.) Can "te'e" be used to mean "touching/in contact with"? There is currently no cmavo assigned to indicate when two things are actually in contact except for this one. The problem with it is that it only indicates that they are bordering, and not where they are bordering. As a member of VA, we could then have such constructs as "ni'ate'e" "bordering below, i.e. on (/in contact with) the bottom of", or "ga'ute'e" - "on top of". (Leaving it "as is" really doesn't help. "ni'ate'e", in the current definition means "[origin] [down] [bordering]": "bordering a place below ...", which could mean "on the bottom of", but probably doesn't in most cases.) This, to my mind, would complete VA very nicely. We would have: "te'e" "in contact with/touching"; "vi" "a short distance from"; "va" "a medium distance from"; "vu" "a long distance from". Perhaps a new, shorter cmavo could be chosen for this function, if any are left.

I'm having a little difficulty using logical connectives with tense constructs, especially long ones. To solve my problem: Which binds more tightly, the connectives or the modifiers of the words connected, e.g. in "pujeba zi do" we have "®pu je ba¯ zi" or "pu je ®ba zi¯"?

How the hell do you use "zo'i", "ze'o", and "fa'a", by the way? They all appear to represent orientation. Am I right in assuming that "zo'izu'a" means "to the left of a place oriented towards me" and "zu'azo'i" means "on my left, oriented towards me"? Just wanted to be sure.

Is it possible to bind a temporal and spatial tense more tightly together so that we can indicate position at a certain time? In the sentence

la ivan. pu ti'a zutse le stizu
Ivan sat behind me in the chair

does "ti'a" refer to where you were at the time, or to where you are now, or even where you will be? Is "ti'a" tied to "pu"? Maybe a word order convention could be useful here. A temporal construct appended to the end of a spatial construct would link them in time, and a temporal construct placed before a space construct would be independent. Thus "ba ti'a pu zutse" would mean "will sit behind where I sat". We can still have our vagueness if we like: "pu ti'a" with no following time marker makes "ti'a" vague as to time. "pu ti'a ca" would mean, of course, "behind me then".

Is there another way to do this that I've overlooked? Logical connectives won't do it, perhaps "bo" will. I think my suggestion is more flexible. In the case of a logical connective, there is exactly that: logical connection, which is usually independent of time. "pujeti'a" says nothing about the "time" of "ti'a", it just says "both before in time and behind in space" - not necessarily simultaneously.

[Lojbab tackles some of these questions:]

I believe that "vi" still works as well as it has in the past. It is true that "bu'u" is the counterpart of "ca", but most often when we say "at" in reference to space locations, we do not strictly mean coincident in location. "vi" means a short, possibly very short distance (i.e, approximating 0), and can therefore be translated as "at" as easily as "near". It doesn't necessarily mean that you are adjacent, but context will usually include this as a plausible interpretation. In fact, "va" is probably a better word for "near, but not at". "zi" and "ze'i" and "ve'i" also work to indicate very small distances and/or areas/intervals. "bazi" can therefore mean "immediately", when referring to an impending action.

When you are dealing with something of significant size, like a city, there is always the question of where you measure from. If from the city center, then places technically "in" the city are merely "near". Tense information is, of course, vague, and if you want greater accuracy, you need a separate predicate.

It is true that some of the members of FAhA may be more exact about location than "vi" or "va", and some of your alternatives would work quite well in place of "vi". There are several members of FAhA that are more specific analogs of "vi", including "ne'i", "pa'o", "ne'a", "te'e", "re'o" (which includes touching), as well as "bu'u". But "vi" works fine, if a bit more vaguely.

Logical connectives have the largest scope within tense constructs, so that "pujebazi" will group as "pu je bazi".

"ze'o" and "fa'a" and the like are pure directions, and generally intended to be associated with motions as well. As a sumti tcita, of course, "fa'a" can indicate a direction without motion: "fa'a le zdani" (over towards the house), "mo'ifa'a le zdani" (while moving towards the house).

As I just said above, the tense system is not intended to express extremely complex ideas. If it is critical to you to distinguish between where I am now and where I was in the past, in deciding whether something is "behind", then you need at least two term phrases, and shouldn't be trying to load all of it onto the selbri tense. Try something like

la .ivan. pu zutse ti'a mi
Ivan sat behind me.
where the fact that "ti'a" appears after the "pu" means that you are already set into the past. For sitting behind where I am now, I would want to be more explicit about the tense contrast:
la ivan. ti'a mipeca pu zutse Ivan, behind the present me, sat.

ko'a stizu

[A comment on a usage issue, from Lojbab:]

John Cowan had labelled the use of "ko'a" in such a sentence as

ko'a stizu

as being 'incorrect' where "ko'a" has not been assigned. This is misleading, since we teach such usage in introductory lessons before relative phrases with "goi" have been taught.

If "ko'a" has not been defined, then using "ko'a" risks confusion. The appropriate answer then is "ko'a ki'a stizu", which for novices has to be answered with "ko'a du ti". We would prefer people to use the vague usage "ko'a stizu" than to overuse "du" as "ko'a du le stizu" which new Lojbanists will (and do) quickly acquire the malglico and very incorrect non-Lojbanic "du" = English "is".

So I favor people using undefined "ko'a" at the start. It is a relatively unserious error that is easily correctable and usually communicative. As opposed to the alternative, which if theoretically more correct is risky of bad pedagogy.

Hmmm. Perhaps "zo ko'a sinxa le stizu" is within a lesson 1 or lesson 2 student's grasp, in which case it should replace the sloppy form.

But "ko'a stizu" is always grammatical, and there's the possibility that the speaker defined it before the listener came in, in which case "ko'a ?ki'a stizu" is still the appropriate response.

Questions On Logical Connection

Colin Fine:

A simple question of semantics:

In a logical connection with an unspecified sumti, are the branches of the connection to be construed with the same value for the sumti or are they independently unspecified? i.e. If

mi klama la lidz. .e la bratfrd.
I go to Leeds and Bradford.

is true, it follows that

mi klama la lidz. zo'e .ije mi klama la bratfrd. zo'e
I go to Leeds from somewhere. And I go to Bradford from somewhere.

but does the stronger claim follow that

su'oda zo'u mi klama la lidz. da .ije mi klama la bratfrd. da?
For some (place) x, I go to Leeds from x. And I go to Bradford from (the same) x.

[John Cowan replied:]

Hitherto this point has been discussed but not settled. I believe that pragmatics dictates the 'independently unspecified' interpretation, and that to get the same value an explicit "da" is needed.

I think this example shows clearly why not. "klama" actually has five places, so

mi klama la lidz. .e la bratfrd.


mi klama la lidz. e. la bratfrd. zo'e zo'e zo'e

If this is construed as

mi klama la lidz. da de di .ije mi klama la bratfrd. da de di

in order to be sure that the origin (da) is the same in both bridi, then we are put in the silly position of insisting that the route ("de") must also be the same for both destinations! Thanks for providing this example.

Mark Shoulson:

Lately I have taken to trying to think of how to translate English expressions that I hear on the radio into Lojban's structure (not necessarily the words; my vocabulary isn't that big and I can't flip through lists whilst driving.) One struck me this morning and led to a little thought about some of Lojban's connectives. This is a pretty basic question and I'm positive it's been dealt with before (I can't remember reading about it anywhere in Lojban's literature, but I think it's there somewhere). Anyway, it was a commercial for some clothing sale, and it was saying how they have "clothes for men and women". Now. Do they mean "clothes for men as well as clothes for women" or "clothes which may be worn both by men and women"? I think these are plausible ways of handling these readings in Lojban:

lo taxfu be lo'e nanmu .e lo'e ninmu
clothes for men and clothes for women; not necessarily that the same clothes be for both.
lo taxfu be lo'e nanmu ku jo'u lo ninmu
unisex clothes, for both sexes.

Is this a legitimate distinction between ".e" and "jo'u"? ".e" is a logical connective, and I imagine it as asserting the relevant bridi twice, as it were, once for each of its arguments, with no connection in between. "mi .e la djan. klama" means that "I and John go/come, not necessarily that we do so together or at the same time or having anything to do with one another", while "mi jo'u la djan. klama" implies more of a connection, while "mi joi la djan. klama" implies that we worked on it as a team, so the action could really only be said to have been accomplished by both of us in concert.

I realize that this example is open to other methods, including relative clauses and the like. Also, note that you could argue that unisex clothes are not for "men AND women" but rather for "men OR women", and require the use of ".a", the inclusive-OR or some such. What are the opinions of you folks out there?

lo taxfu be lo'e nanmu .e lo'e ninmu

seems to me be equivalent (modulo existence) to

da poi taxfu lo'e nanmu .e lo'e ninmu

which expands to

da poi ge taxfu lo'e nanmu gi taxfu lo'e ninmu

i.e. something which is clothing for men and is clothing for women. This seems to me to mean strictly unisex clothing. I think '.a' will do quite happily for clothing that will do for a man, a woman or both.

I am not yet familiar with the non-logical connectives (I'm suddenly assimilating two or three years' worth of language development in a very short time .ue), but I would have thought that the "jo'u" example would mean what more like what you said for ".e".

I have a question: are "joi" and "jo'u" permissible when there group/mixture in fact contains only one of the connectands? Does "lo ninmu joi nanmu" imply that there are members of both sexes in the group?

Lojbab: Well, actually, I think more that it implies a hermaphrodite human mass; i.e. it exhibits properties of both genders simultaneously, as per a mass (with "lo" as the gadri, the thing itself need not be a mass). Compare the classic Loglan example (translated into Lojban vocabulary) "lo xunre joi xekri bolci", "a red-and-black ball", which is neither red, nor black, but a combination of the two. It would not be correct to call something a "red-and-black ball" in English unless there was some element of both colors on the ball.

Causality in Lojban

[A discussion between Lojbab, And Rosta and Jim Carter led to the following formulation of a significant and cohesive portion of Lojban semantics.]

Lojban embeds several varieties of expressions of causality. JCB originally analyzed Loglan causality as being of four types. Further analysis during the development of Lojban has identified other expressions of causality that are embedded in the language design.

1. rinka (ri'a) is principally physical causation, but has pragmatically tended to be a catch all for causations that don't fit other categories. This is historical, because JCB used rinka's equivalent for general causation. See below for our solution.

(I push) rinka (Jack falls).

2. sarcu (sau) is 'necessary', 'rinka' implies nothing about necessity.

3. mukti (mu'i) deals with motives and their (potentially) resulting actions.

mukti = x1 motivates activity x2 on the part of agent x3

We've decided that English has no good word for the x2 of mukti.

It is the motivated action. The activity may or may not take place but is at least achievable in the mind of the agent x3.

(I want money) mukti (I work)

4. krinu (ri'u) is explanatory causation; the x1 is the reason and the x2 is the thing explained.

(giraffes eat from trees) krinu (therefore) (giraffes have long necks)

5. nibli (ni'i) is logical entailment. S entails T when there is a logic (a list of logical transformations or theorem steps or applications of definitions of words) that starts with S and ends with T.

6. jalge (ja'e) indicates "result". It is a reversed direction causal that serves as the generic of causation, thus freeing rinka for its more limited meanings.

x1 is the result of x2 (x2 is the cause of x1)

7. zukte (zu'e) helps distinguish motives from goals

x1 acts at x2 to achieve x3

Note that the basic claim of "zukte" is that an action is taken in order to achieve the goal. "mukti" operates in the world of mental reality, and implies a relation between a motive and a motivational result. There is a weaker inference that the motivational result actually takes place; the person motivated might be unable to do what he is motivated to do.

8. By contrast, for simple agentive causation, use gasnu (gau)

x1 is agent in action x2

"gasnu" is closely related to "zukte" but does not imply any purpose or goal on the part of the agent.

[Another contrast with mukti, zukte and gasnu might be troci ("try"), which implies an agent and an activity. The activity may not take place, but is at least attempted. It is not clear with troci that there is any motive or goal beyond the attempt itself.

9. sarcu mentioned above, can express "necessary conditions". "sufficient conditions" may either be curmi (permit/allow) or banzu (sufficient). Of course, the entailment of nibli covers logical sufficiency, a fairly limited variety.

Note that the Lojban logical equivalent of "if...then", unlike the phrase in English, does not imply causality (Example: "if you water the plant, then it will grow"). In Lojban, it is undesirable to infer causation from such a statement. In Lojban, such an if-then is represented as "not a or b", from which causation is simply not inferable. The Lojban sentence ends up being equivalent to:

Either you water the plant, or the plant does not grow.

Since logical OR is reversible, this means the same as

Either the plant does not grow, or you water the plant.

On le and lo and Existence

[Another Lojban List discussion led to the following explanation by Lojbab:]

'le' relations or abstract events are specific products of the speaker's mind and hence must exist only in that mind. The description is a label and need not be accurate.

'lo' and 'loi' claim only that the described thing/event is something that actually fits the description, but doesn't claim that any such thing exists.

le ninmu cu nanmu
"The woman is a man." might be a statement about a male transvestite.
"lo ninmu cu nanmu" would not apply to a male transvestite unless you assert that a male can actually be a woman.

You thus can make statements 'le' and 'lo' about 'non-existent' things like unicorns. Want(x,y) in English makes no implication about existence of either x or y; e.g. The unicorn wants a maiden, where x is non-existent but y is (potentially) existent.

In Lojban you assert or reject existence through the use of quantified variables (da, de, di) which implicitly or explicitly invoke a prenex:

da zo'u da djica lo broda
For some x, x wants a maiden

"da" presumably excludes unicorns, since they do not exist in our universe

With or without a prenex, you can use a restrictive relative clause:

da poi danl,iunikorni cu djica lo ninmu

which is false, because the x1 is the empty set.

"lo x" can thus be taken as equivalent to "if x exists, then some x".

But this formulation has a problem. By the rules of symbolic logic, any conditional statement with a false antecedent is true. And thus a statement using "lo" where the referent is a nonexistent x will be a true statement, including the contradictory:

lo danl,iunikorni cu zasti
Unicorns exist.

Alas, all attempts to analyze "lo" run into some such problems, but the result is a useful shorthand regardless. Thus, we retain "lo" as a useful part of the human Lojban, while realizing that good 'logical' usage would be to use "da poi ...".

If there is a way out, it is to state that something exists because we can conceive of it, and it has the properties we attribute to it in our conception. This approach works around the conditional aspect of "lo", but no doubt is unsettling philosophically. Of course in the world of logic, things often 'exist' that don't apply in the 'real world', so this might be the best approach.

A Heated Exchange?


If I write:

2 +

you know there is something missing ... you yearn for another number, to complete the expression. The same with a Lojban expression:

mi klama
I come/go.

is incomplete. In Lojban, you yearn for a destination, departure, path, and means. ...

Art Protin comments and Bob Slaughter responds in italics:

I hate to have to say this so strongly, (and Bob please don't be offended,) but I find to be totally without merit, bogus, the comments offered [above].

While I can easily accept that we need a far different model to think about Lojban than the one we use for thinking about English, I reject any suggestion that

mi klama

is in any way incomplete. The image that I construct in my mind is small corresponding to the small amount of data provided, and it has "hooks" where I might attach additional data like the destination.

Then obviously you haven't learned to think in Lojban. :) Perhaps the phrase "is not fully completed" instead of "incomplete" might make more sense here. You may not yearn for them, but you know there are unanswered items, because the "hooks" are far more explicit in Lojban than English.

Other dialog/monolog is required to elevate that "slot" to any greater prominence. If that piece of the whole picture becomes both important and unspecified, I will inquire as I would for any other data I need to satisfy my view of that picture.

Bingo!! The unladen "hooks" are meant to be filled, or questioned. Lojban is a dialog-based language, rather than a monolog-based language like Standard Written English. I can see where a speaker of English "sees" "I come" as a fully completed sentence with no unknown information, but all speakers will know the speaker of "mi klama" could've said something but consciously didn't. Hmmm, imagine what that means for Lojbanistani politicians.....

I see no reason to provide any members of any relation (predicate) that are not relevant to the discourse. That the provided members can/do have the designated role in some instance of the relation is all that the language can express. We might, with sufficient dialog and experience, be relatively certain that we know exactly which instance is being described, but there can be no guarantees. (This is not a property of just Lojban but of human communication in general.)

But I might see a need for me to know something you said, so I will ask. But, it is the assumption of accurate and inaccurate assumptions that Lojban brings to the front of its conversation mode. By knowing there is unspecified data, and emphasizing it, we change the form of communication. Rather than pontification and counter-pontification discourse, we should have fully interactive dialog discourse.

Language Goals

Following are essays related to the goals of the language, most of them dealing with aspects of the application of Loglan/Lojban to the Sapir-Whorf Hypothesis. However, for the computer-inclined, we include a report on a new project using Lojban for artificial intelligence/natural language processing.

Lojban and Metaphysical Bias

(a discussion between And Rosta [not-indented] and Lojbab [indented])

Does one necessarily wish to avoid metaphysical bias? I would always wish to be able to say that something is at the "back" or "front" of my mind, or that I am in "high" or "low" spirits. I might wish to avoid distinguishing recipients from destinations and treat them as the same thing; I might want to treat possession as a kind of location, say.

In Lojban, we want to remove metaphysical bias when possible. It isn't always. The examples you have selected are examples that we will be trying to eliminate (at least in translation to Lojban), because they are English biased figures of speech, and it is not necessarily universal that all cultures consider "high" spirits to be better than "low" ones, or that the 'mind as queue' metaphor is superior to the 'mind as stack' one.

I follow the cognitivist doctrines of George Lakoff and his colleagues, of Jackendoff and of Langacker. (These are very simply expounded in Jackendoff's review article of Lakoff's new book in the June 1991 Language.) This doctrine maintains that certain things are conceptualized only metaphorically. Metaphors whose vehicles are space and the body predominate, and are used to conceptualize more abstract things. Some of these metaphors are claimed to be grounded in universal human cognition, and others to be dependent on culture.

We therefore could, maybe, draw the following conclusions:

(1a) Lojban's aim (of removing metaphysical bias) is doomed to fail.

The goal is to 'minimize' it, not remove it. For situations where one or more roughly equivalent methods exist to express something, but each is biased in some way, we try to allow all of them. If we must be arbitrary among several choices, we choose a single way, but are prone to choosing a non-English way to counter the tendency for English biases to creep in.

(1b) Lojban's aim flies in the face of the way we really think and is therefore a hindrance to thought.

This we will find out. The problem is that certain concepts are always metaphorized because we have no primitive non-metaphor to express them in NLs. Thus we have a chicken and egg problem. Lojban will try for a different egg.

Now even if metaphorless Lojban is possible, why is one supposed to avoid metaphor? My English-biased conceptual metaphors are the way I think.

Not if you are trying to communicate to someone from Thailand who does not know your metaphors. In an earlier book, Lakoff noted that not all cultures shared the same metaphors (e.g. "up" is "future" or "up" is "past", I think was one dichotomy). I prefer a language that says that future is future and makes no links with 'up' or 'down'.
(Remember that the goal of Lojban involving Sapir-Whorf means that as much as possible we must reduce and/or identify all sources of bias that would affect 'world-view' - which to me is a very similar concept to 'metaphysics'.)

Sapir-Whorfian Thoughts?

In response to a question from James Meritt, Lojbab said the following:

There is no evidence yet of Lojban providing thoughts that are unthinkable in English, but the constraints of English syntax do tend to make thinking in certain ways more difficult. It would be a long time before we truly came up with an example that unambivalently is uniquely Lojbanic.

Hmm. I'll have to amend this. In our discussions of the last week or two regarding Lojban property abstracts, it has become pretty clear that while it is possible more or less to define what is taking place using English words, I think it accurate to say that most of them have no English equivalent in any meaningful sense.

For example, "loika melbi" translates as "Beauty" the abstract concept and "leka lemi speni cu melbi" more roughly as "my wife's beauty" but more accurately as "the properties that make it true that my wife is beautiful [by some standard to some observer]". But with most predicate words, there is no English equivalent for the property abstract. For example, "loika klama" would translate as "Going-ness" if that were an English word - already hard to grasp, while its counterpart:

leka mi klama le zarci
the properties that make it true that I go to the store

conveys no sense of exactly what sort of properties these might be - we would tend in English to start thinking in terms of causes, which is not what the Lojban means, because "Going-ness" is just not an English concept. But constructs like this are rather easy to express in Lojban, and in some cases are virtually obligatory. A particular "Going-ness" for example is the property that is being compared when we say that I go to my local Safeway more than you do.

Whether use of statements like this in Lojban means that anything new and different in human thinking will arise as a result of this implication, is what is still not clear.

Going beyond this, I can say that there are a lot more perhaps more obscure things that can be easily said in Lojban, but which defy English translation. Lojban does after all, allow and almost encourage the expression of "grammatical nonsense", of the "green ideas sleep furiously" variety, but even weirder. People can indeed wrap their minds around such nonsense (for this English example, I have seen proposed places where it might actually be meaningful), but it can reasonably be said that the Lojban equivalents go far beyond what anyone will ever understand in English translation. Whether a Lojbanist thinking in Lojban will 'understand' such statements in the sense that we can understand "green ideas sleep furiously" is also presumable, but as yet unverifiable.

Metacognition-friendly Languages

On the conlang mailing list, Zack Smith asked:

Has anyone heard of any languages that specifically support and facilitate metacognitive thought?

Or, put differently,

Has anyone designed a language for critical, self-effacing thinkers, for the facilitation of processes which lead to success in thought?

For those who are unfamiliar with this topic, metacognition is essentially "thought about thought". It's what most successful thinkers do to remove biases, limitations of thought, mistakes due to the failings of human memory systems, etc. With metacognition comes many freedoms in thinking, living, feeling, creating, etc.


I'd like to think that Lojban has many of the things Zack is looking for in a language for metacognitive thought. We certainly have a lot of the features he mentioned in his posting. So perhaps he might like to investigate Lojban. Specifically:

  1. Lojban has a predicate grammar, and rephrasing your thoughts in a predicate formation tends to require you to think a little more carefully about what you are trying to say.
  2. Lojban has an extremely powerful and flexible tense system, stretching rather beyond what natural languages do in our desire to encompass the full scope of what natural language is capable of expressing in tense.
  3. Lojban separates the speaker's emotional attitudes from the statements being expressed, and all "emotion-loaded" words are marked.
  4. Lojban has a grammar for metalinguistic discussion which is distinct from the regular grammar, allow you to express metalinguistically relevant information, again separate from the main statement.
  5. Lojban has a set of evidentials, again distinct from the main grammar, which allows you to succinctly indicate how you come to make a statement (deduction, hearsay, definition or assumption, etc.)
  6. Most of these markers that convey metalinguistic information can be attached/focused at the sentence level, phrase level, or individual word level, as appropriate.
  7. Lojban handles certain constructs commonly associated with logic in the manner that predicate logic does; thus we have no confusion between OR and XOR, and distinguish clearly between causal if-then and implicational if-then (we also have embedded in the language several kinds of causality). Quantification and negation work as they do in predicate logic, hopefully reducing the types of errors that can result from misapplying these features of logic.
  8. Lojban deals clearly with multiple levels of abstraction, as are often involved in even simple natural language expressions, with each level being clearly distinguished from the others, and specific constructs for "raising" objects from one level of sentence abstraction structure to another.

Ralph Dumain commented:

[#8] is the only feature you mention that particularly impresses me as significant for dealing with questions of "world-view". Could you give a few examples or refer me (us) to the proper locations in the Lojban literature that explain this feature?

[Lojbab: This is the entirety of what we call "sumti-raising", dealt with extensively in JL16.]

In further discussion, Jeff Prothero commented [in italics] and Zack responded:

My basic conclusions are more or less:
(1) One could do a measurably better job on a language for thinking type purposes, but the pay-back doesn't justify it for an individual or small group...

I don't accept this assertion, for several reasons.

1. I believe that the human mind can always use a little exercise.

I subscribe to the idea that the mind is like the muscular system: The individual should exercise it regularly, and not miss any major areas..

My tools to this end are self-analysis and a language oriented toward clarity and self- (re)directing thought; though I do other things as well.

2. The mind makes mistakes, either because of emotional factors, various and sundry conditionings, sociological factors, or because the human memory system just isn't very reliable.

Hence, a continual untangling of memories and thoughts is necessary. Freud suggested that dreams fulfill this purpose to an extent. I'm no Zen master, but I prefer to not leave things up to dreams alone. There is no free lunch when it comes to mental health, clarity of mind, or clarity of ideas. The more one borrows, the more one accepts blindly, the worse one is.

My language exists to aid me in performing such clarification work.

(2) Most of the really high-payoff ideas fail due to human limitations...

I've definitely encountered trade-offs with my language, and I agree that some features will fail to get into -any-language because they're too unusual, too expensive to use, or too difficult to learn to use.

For example, I wanted at one point to expand the number of auxiliary pronouns (this, that) from three or so to about 8. The idea was that eventually, using them would translate to the speaker/thinker having one conceptual chunk in short-term memory, into which the speaker would reference whichever items he needed.

Without going into the details, the problems were that (A) I couldn't get anyone to even try to learn my language, (B) the feature itself was unusual, (C) the feature required a considerable shift in thinking, and (D) I estimated that the feature wouldn't be used often enough to justify its existence.

(3) The syntactic stuff isn't all that interesting, most of the important design decisions are in the conceptual vocabulary. Working on this is way beyond the current capabilities of any small group ... but if one thinks about it, finding/fashioning an effective set of concepts for understanding self/thought/universe is more or less what the entire international scientific machine is working on.

The first sentence is right on the money, but the rest I think is incorrect. Metacognition, the aim of my language, is the idea that one can improve on thought itself, not thought about any given topic. The reason is that there are considerable similarities in the reasoning that a stock-broker does in his life and that reasoning that any other person does. The points I'm expressing are key to the topic of creating a metacognitive language. They are:

  1. Patterns of successful thought are universal, regardless of who is thinking them, where, when, or what the topics of reasoning are. [Lojbab comments: I'm not even sure that the standard of what constitutes 'successful thought' is universal. Nora adds: As a critical thinker, Zack should take another look at his assumptions. For example, on the job I am very good at analyzing details, but I sometimes have problems with the 'big picture'. These two levels of reasoning can give different answers on the same topic, and yet either can be successful given the appropriate situation.]
  2. The same holds for mistakes of reasoning, be they biases, mental sets, etc. No one is immune from ignorance, fear, stupidity, etc.
  3. And, because the mechanisms of successful thought are universal and identifiable, the basic concepts on which they are based can be coded and regularly used.

The work is in finding the concepts, like you suggested.

For instance, imagine going through a dictionary, word by word, and working out the precise semantic structure of each word, then trying to identify core concepts are strategies for constructing words and phrases out of those core concepts, and still have it be nice to hear and easy to use. Yikes.

However, a scan of an entire dictionary isn't necessary. I think that the examination of a small group of words can provide considerable food for thought.

The question is, what is the essence of natural phenomena? What do the processes of creating art, writing papers, brokering stocks, running a business, being a responsible politician, being a corrupt politician, or making toast for oneself all have in common? The answer, I think, lies in the studies of psychology, evolution, formal logic, problem solving, etc.

(4) While natural languages tend to be uninterestingly different clones of each other, the space of possible languages is much larger than the little volume they cluster in, and one can have languages which make it possible to think thoughts one wouldn't come up with in natural language, thoughts which can't even be expressed in natural language.
At least one such language has been fashioned... we call it "mathematics"... an easy point to giggle and dismiss, but worthy of more serious attention than that...
On a more mundane level, I think you might find Jim Carter's 'Guaspi' the best laid out conlang aimed in the general direction you're going. Loglan has (imho) an excess of hair. Jim's language strips out most of the irrelevancies, not too far in spirit from some of my own efforts.
I suspect if you put some time in studying, you'll come to share my conclusion that if one strips out the uninteresting restrictions and syntax hacks from a language, there's nothing interesting left at the syntactic/superficial level... one is left with more or less the spoken equivalent of Lisp, with a simple universal parsing mechanism that doesn't commit one to anything, and all the interesting content relegated to the selection of concepts/functions to use in the vocabulary.

I took my first gander at a Loglan manual today, and I think that you're assertion that Loglan has much unnecessary fluff could be correct. For instance, why encode colors? In my language, I permit the speaker to import context-relevant data such as names of colors. I want to give the speaker/thinker the stuff of thought, not the fluff. I'm not creating culture here. [Nora comments: This is OK for individual thought, but not if you want to be understood by others, especially those of a different culture.]

As far as the Lisp argument above goes, I'll have to study some more. I don't understand how you envision language being used, and how a non-fuzzy logic language e.g. Lisp could usefully serve a sentient organism or machine. I don't agree that all verbal expressions (in any language) break down into logic, unless perhaps it's fuzzy logic. Even if they do, that's like proving 1+1=2 in logic - it takes several hundred pages - so why bother... [Nora rebuts: Because the very fact of doing so shows the questioning/examining of assumptions - the very thing Zack professes to want.] It's like talking about Hitler's taste in clothing rather than his crimes against humanity. Different subjects, different purposes, no?

Also - I'm not interested in taking out all of the hacks of a language - shortcuts and approximations are fine by me, so long as the language forces the speaker to note their existence in some way. A Lisp-like conlang is an extreme, impractical solution, especially for those of us who dislike Lisp.

The middle-road solution is to allocate one's semantic information as usefully and well as possible.

I want to allocate it for the metacognitive information - e.g., the estimations of utility or arguability, the estimations of linkage between expression and goals, the assessments of a speaker's emotional state, the protocol for cooperation between speakers, the statements of problems and tentative solutions, the acknowledgement and removal of bias or limitations of thought, the estimation of correlation, etc.

In terms of computer languages, I'm going for C++ rather than Lisp. I want the core concepts to be related to actions, events, objects, abstractions, scripts; not merely relations or predicates. Logic is essential, but I see it as one part of a larger process.

Then again, I've considered making my language only an extension to other languages, perhaps it's more like Objective-C than C++. ...

[On further questioning as to his goals, Zack replied:]

My project is still in its initial stages, compared to Loglan. The goals have evolved to include...

1. Choices of basic concepts which propitiate human thought.

Here I ask, how do humans think, i.e. what basic data types does the brain appear to process? For instance, some AI'ers assert that all thought is, or should be, based on predicate knowledge. I find this inaccurate and limiting.

By "basic concept" I refer to a notion which is semantically atomic, a sort of original or root class of semantic notions. These concepts would be used to form more complex words and sentence structures.

For example, if "market" and "protection" are basic concepts, then the word "democracy" would be based on both of those words, since democracy is the protection of a market in which the commodities are variations on kinds of government.

Note that my objective is not to rewrite Webster's, but to get critical words pinned down, then import context-specific words verbatim by attaching a prefix and suffix.

2. Choices of basic concepts which propitiate self-examining thought. This essentially would include the ideas of :

A. The encoding of metacognitive information, e.g. for the objective assessment of arguments between speakers.
B. The identification of information which describes the functioning of information processing systems such as humans and mammals.
For instance, it's easy to assert that Freud's "repression", or "setting aside and ignoring" (that was the original German meaning) is not only likely, but necessary. Any information processing system (e.g. the human mind) which must interact constantly with its environment whilst juggling multiple conflicting goals certainly must ignore some important data (repression), and certainly misfiles other important data (hence dreams and sudden recollections/ideas as means of bringing misfiled or repressed data to consciousness).
Any serious metacognitive language must take these effects into account, because they are psychological phenomena which affect the arguments and ideas presented by the speaker.
By the same token, biases, mental sets, and other phenomena must be identified through the language. They affect any such system.
C. Means of forcing the speaker to think before he expresses himself, in particular, to translate his concepts into terms of the workings of his own psychology.
This is a critical point; here is an example argument for such a feature: Suppose a boy hits his brother without provocation.
Ask yourself, what is the most useful means of teaching that boy the error of his action?
Besides using basic operant conditioning (no TV for You!), I suggest the following:
You ask him to explain his own experience in choosing to act as he did. Don't just ask for justification; ask for self-examination, then push him toward self-modeling, to cause him to realize that he himself is a collection of feelings and people, that he is responsible for all of it.
I'm not suggesting that one force him to identify his "evil side"; rather, the phenomenon that causes a behavior, e.g. insecurity, rivalry, etc.
This language must force the speaker to look within himself for the causes of his making expressions or the content of those expressions. The reason is simple: This is the path toward clarification of one's own thoughts and toward metacognition.
I'm not certain, but I think that Loglan doesn't force this analysis. It permits prefixing of phrases to improve discourse between speakers (e.g. it has prefixes for "suppose that", "for example", "is it that", etc.), but causes aren't characterized.
D. Evolutionary processes, and the phenomena which determine or found them, must be tightly coded. Considerably more real phenomena can be explained with evolutionary logic (or simulated with genetic algorithms, probably) than meets the eye. Evolution is critical to who we are, why we are here, what we will become...
E. Basic logic, or fuzzy logic, should be supported (tightly coded).

...[Later, Zack expanded upon his earlier statement: "... a mechanism for expressing tense information"]

This is neither critical, nor did I intend it to seem so. It is important, though.

I've found that of all places, lack of clarity in tense information is the most contagious, i.e. it affects to other types of information in one's expressions. Once tense becomes unclear, it's all lost. One can't speak about events, actions, activities, what have you, with any precision if tense info is imprecise.

Richard Kennaway responded with some ideas:

One of the main faults I find with woolly writing or speaking, especially in technical talks at conferences, is a lack of attention by the speaker to making clear the reason that he is saying what he is saying. It seems to me that it is impossible to understand an utterance unless one understands the reasons for making it.

Thus I would like to see means provided in the language for easily expressing such meta-information throughout one's speech. Off the top of my head, here are a few communication modes that might be worth expressing explicitly, rather than leaving them unspoken as is usually done:

  • background information to indicate what area of knowledge the speaker is dealing with, what knowledge he expects his audience to know, and to help those in the audience less acquainted with the subject.
  • new knowledge that he wishes to impart.
  • a summary of what he has just explained in detail.
  • a summary of what he is about to explain in detail.
  • an informal description intended purely to convey informal insight, rather than a precise statement.
  • smalltalk (utterances whose literal meaning hardly matters at all, the starter motor of social intercourse, not to be confused with the main engine).
  • a question to which a definite answer is required, in contrast to...
  • a question which is part of a conversation, which the hearer need not rigidly stick to in formulating a response.

There's an interesting book by Deborah Tannen, called That's not what I meant!, suggesting that a lot of miscommunication is due to people being unaware of the different modes of communication that they and others are using. Perhaps mechanisms encouraging the explicit marking of such modes would help.

I have only glanced through her later book, called (I think) You just don't understand!, in which she claims to correlate these different modes with gender. I suspect that it is just a repackaging of old wine in a trendy new gender-polarised bottle.

Lojbab comments:

Per Richard's comments. I think that a language used for solely for introspection, whether metacognitive or otherwise, is going to be significantly different from one used for interaction. So much of the problems of communication between people stem from things such as what Richard mentions (all of which I believe are covered in the Lojban design, but optionally). But few are really relevant to the problems Zack seems to refer to in self-analysis of his own thought.

The philosophers in the Lojban community (most of whom are not on the computer nets), may have something useful to say about Zack's ideas, and what (if anything) Lojban has to support his ideas.

le lojbo se ciska (cont.)

I (Lojbab) don't have many complaints about Nick's work in the following two stories. They were not passed by an independent editor, but Nick indicated that they had been reviewed on the computer nets a couple of times, and that he had made changes appropriately. Alas, he had not checked the text with a parser (only some minor errors), and he had two non-existent gismu in the second tale, one of which rquired guesswork to figure his intent since it was not a simple typo. But the texts are readable, and my formatting rules that failed to handle Nick's coffeehouse text are probably satisfactory for this text. All comments are from me.

Two Greek Folk Tales translated by Nick Nicholas

I. melu la xrist. na.enai la pacrux. seljdadji da li'u

=.ika'u la pacrux. klama la xrist. gi'e bacru ®lu ?pe'ipei ?xu do jinvi ledu'u leti cange bakplixa goi ko'a xriso li'u¯

=.i ®lu !pe'i go'i li'u¯ selba'u la xrist.

=.i ®lu do srera (to'i la pacrux. spuda toi) =.i le kakpa cu me !cai !ba'e mi !sa'e =.i mi'o fau lenu do na krici lenu go'i cu .!e'u klama ca le cermurse leko'a cange poi ko'a tsise'a[1] =.i do vi le cange cu !ba'a zgana lenu ko'a me mi li'u¯

ni'o ca le bavlamdei ke clira clira la xrist. joi la pacrux. klama le cange po ko'a gi'e se mipstu loi stani =.i le kakpa cu !ba'e sutra klama gi'enai kruce jdaxanmu'u gi'e lasna le bakni le te plixa gi'e co'a renro lei tsiju

=.i ®lu .!e'o ko zgana .!u'a (to'i la pacrux. bacru toi) =.i ko'a cu !sai me mi =.i ko'a ni'i le !da'i nu ko'a me do cu jdaxanmu'u pu lenu co'a gunka li'u¯ ®lu le kakpa cu !ja'o to'e depcni fi lemu'e mulgau lenu tsise'a =.i ko denpa lemu'e midydo'i =.i ca ri ko'a co'a citka =.i do ca zgana lenu ko'a jdaxanmu'u li'u¯

=.i midydo'i =.i ko'a co'a citka gi'enai jdaxanmu'u

=.i ®lu .!e'o ko zgana .!u'a (to'i la pacrux. bacru toi) =.i ko'a ni'i le !da'i nu ko'a me do cu jdaxanmu'u pu lenu citka =.i do caki na ji'u darlu =.i ko'a me !cai mi li'u ¯ ®lu .!e'o ko denpa =.i go ko'a mo'u citka gi'enaicabo jdaxanmu'u gi ko'a me do .!e'a li'u¯ =.i ko'a mutce citka gi'e mutce pinxe gi'enaiba'obo jdaxanmu'u gi'eji'a .!uero'a cladu gaxykafke =.i la xrist. bacru ®lu ko'a .!ainai ca .!e'a me do li'u¯ =.i la pacrux. cu bacru ®lu .!ienai na go'i =.i ko'a .!ainaicai me ko li'u¯

Neither Christ nor the Devil wants him.

Once the Devil went to Christ and said "Pray tell, do you think that plougher is a Christian?" "I do." "You're wrong", the Devil answered, "the plougher is all mine. If you don't believe me, let's go to his farm next dawn when he's ploughing. There you'll see he's mine."

Very early the next day, Christ and the Devil went to the plougher's farm and hid in some branches. The plougher hastened to the farm, didn't make the sign of the cross, attached the bulls to the plough and started sowing. "See?" said the Devil. "He's mine. If he was yours, he'd make the sign of the cross before working."

"The plougher is impatient to finish sowing. Wait for midday. Then he'll eat. You'll see him making the sign of the cross then." It became midday. The plougher started eating and didn't make the sign of the cross.

"See?" said the Devil. "If he was yours, he'd make the sign of the cross before eating. You can't argue anymore. He's all mine." "Wait. If he finishes eating and doesn't make the sign of the cross, he's yours." The plougher ate a lot, drank a lot, didn't make the sign of the cross, and to top it all off, let off a huge fart! Christ said "Now, you can have him." The Devil said "No, you have him!"

  1. I would probably use "tsipe'a" (seed-spread) or "tsifai" (seed-distribute) rather than "seed-insert", though my knowledge of farming is not particularly noteworthy.

II. (untitled)

=.ika'u pukiku le prenu goi ko'a cu mutce nelci lenu kelci loi kelkarda =.i ko'a ze'i cusku fi leko'a speni fe ®lu .!e'u vi'ecpe la xrist. mu'i lenu friti lo midydo'i sanmi ra li'u¯

=.i la xrist. cu te cusku le sego'i gi'e frasku ®lu mi .!ai klama li'u¯

=.ike'unai ca le midydo'i la xrist. noi se kansa ro leri tadni cu klama =.i leko'a speni bazi lenu viska ri joi ra cusku ®lu le nanba na banzu .!u'u .!oiro'a li'u¯

=.i la xrist. cusku ®lu .!i'a ja'a go'i =.i ti cavi nanba =.iseni'ibo ti .!o'o bazivi se citka mi'o li'u¯

=.i nicygai le jubme =.i zutse mu'i lenu citka =.i la xrist. cestoldapma le nanba =.i ri banzu tu'a lei citka gi'e .!u'a dukse

Once there was a man who loved playing cards. One day, he said to his wife, "Invite Christ here so we can offer him lunch." Christ was told this and responded "I'll go." So, at noon Christ, accompanied by all his student, came there. The man's wife, upon seeing them, said: "Oh, there won't be enough bread!" Christ said: "I think there will. This is the bread we've got, so this is what we'll eat." The table was spread, and they sat to eat. Christ blessed the bread. It was enough - more than enough for those present!

no'i la xrist. ba cpacu loi vanju mu'i lenu pinxe kei gi'e te preti fo ko'a fe lenu ko'a djica lenu la xrist. dunda dakau ko'a =.i lei tadni cu cusku ®lu dunda tu'a .!e'usai le cevzda li'u¯

=.i ku'i ko'a cusku fi la xrist. fe ®lu mi ponse lo plisytricu noi se klama zo'e ja'e lenu citka lei plise =.iseki'ubo mi djica lenu ro klama .!i'anai be le tricu cu se lasna fi ri li'u¯

=.i la xrist. cusku ®lu ledo seldji ca'a !do'a mansa li'u¯

=.i la xrist. ba cpacu le remoi kabri =.i cusku ®lu do djica lenu mi dunda ?ma do li'u¯

=.i lei tadni cu cusku fi ko'a fe ®lu ko bacru .!e'ucai ®lu dunda tu'a le cevzda li'u¯ li'u¯

=.i ko'a cusku ®lu .!ai na'e go'i =.i mi djica lenu mi jinga fo ro nu mi'a kelci loi kelkarda li'u¯

=.i la xrist. cusku ®lu ledo seldji ca'a !do'a mansa li'u¯

=.i la xrist. ba cpacu le cimoi kabri =.i ®lu do djica lenu mi dunda ?ma do li'u¯

=.i ko'a bazi cusku ®lu tu'a le cevzda li'u¯

=.i la xrist. cusku ®lu ledo seldji ca'a !do'a mansa li'u¯

=.i la xrist. baza cliva =.i ko'a co'a kelkarda kelci =.i ko'a jinga fi ro kelkansa =.i la xrist. kucyga'a se sfacatra =.ipujecajebabo ko'a kelci .!ue.i'enairu'e

Christ then took wine to drink, and asked the man what he wanted Christ to give him. The students said "Ask for the kingdom of heaven!" But he said to Christ: "I have an apple tree, which people always come and eat apples from. So I want anyone who goes to the tree to get stuck onto it." Christ said "As you wish, so it will be done." Christ took a second cup, and said "What do you want me to give you?" The students told him "Say 'Give me the kingdom of heaven!'" He said "No; I want to win every time I play cards." Christ said "As you wish, so it will be done." Christ took a third cup. "What do you want me to give you?" He then said "The kingdom of heaven." Christ said "As you wish, so it will be done." Christ left, later on, and the man started playing cards. He won over everyone he played with. Christ was crucified, and the man kept on playing!

ni'o la xrist. klagau lo notcrida noi cusku fi ko'a fe ®lu la xrist. klagau mi ti mu'i lenu mi lebna do =.i lenu do kelci cu banzu .!u'i =.i lenu do jmive cu sisti .!uo li'u¯

=.i ko'a cusku ®lu .!i'a go'i =.i .!!e'odo'a ko citka su'o plise =.ibabo mi klama li'u¯

=.i le notcrida cu klama mu'i lenu citka kei gi'e se lasna =.i lego'i cu cpesku ®lu ko .!e'ocai klama ja'e lenu to'e lasna mi li'u¯

=.i ko'a cusku ®lu mi klama do punaijeca .!ai.u'i .!ionairu'e lenu mi !ga'i djica li'u¯ gi'e di'i kelci =.i ko'a ca lenu mo'u se cinri lenu kelci cu klama le notcrida gi'e cusku ®lu mi ca to'e lasna do gi'e .!i'a klakansa do li'u¯

=.i ko'a joi le notcrida cu klama fo le daptutra gi'e viska la xades. noi se kansa pare se jdadapma =.i ko'a cusku ®lu .!e'u mi'o velji'a kelci =.i .!e'u ge mi te jinga gi'o roroi vi stali gi mi jinga gi'o cpacu leti se jdadapma li'u¯

=.i la xades. zanru =.i ri joi ko'a co'a kelci =.i ko'a ba cusku ®lu li ci pi'i mu du li pamu =.i li pamu su'i pa du li paxa .!u'a =.iseni'ibo .!e'o ko dunda le se jdadapma mi li'u¯

=.i ko'a lebna le se jdadapma gi'e klama le cevzda

Christ sent an angel, who told him "Christ sent me to take you away. You've played enough! Your life is over." He said "Fine. Do go and have some apples. Then I'll come with you." The angel went to eat, and got stuck. He begged the man: "Please come and get me off here!" He said "I'll come to you, but not before I feel like it!", and kept on playing. When he got bored of playing, he came to the angel and said, "I'll get you off the tree, and will come along with you now." They went past Hell, and saw Hades with twelve damned people. He said "I'll gamble with you! If you win, I stay here forever; if I win, I get these damned people." Hades approved, and they started playing. He then said "Three by five makes fifteen, plus one makes sixteen! So give me those damned." He took the damned and went to heaven.

no'i la xrist. ca lenu ko'a joi le drata cu klama ra cu cusku ®lu mi cpedu lenu do noi pamei cu klama mi =.i do mo'ifa'avi klagau .!ue lo du'emei li'u

=.i ko'a cusku ®lu mi !si'a ca lenu mi do vi'ecpe mu'i lenu mi friti le midydo'i sanmi do cu cpedu lenu do noi pamei cu klama mi =.i do klagau ku'i lo pacimei .!oiro'a =.i mi ne pa'a ca .!o'inai klagau lo pacimei li'u¯

=.iseni'ibo!zo'o la xrist. zanru tu'a ropaci klama

When the man and the others came, Christ said "I asked you, one person, to come to me. You've brought too many people here!" He said "And when I invited you to offer you lunch, I asked you, one person, to come to me. But you brought thirteen! So I'm bringing you thirteen too." ERGO, Christ let all thirteen in.

A Lojban-to-Prolog semantic analyzer

by Nick Nicholas

[For an extended class project related to his Masters degree work in Cognitive Science, Nick Nicholas has undertaken a project in natural language understanding of Lojban. This is a significant undertaking with great potential for Lojban's credibility given his likely success. Nick and John Cowan contributed ideas to his final project statement, included here. Also included are the reports on preliminary results that Nick has thus far presented.


The problem I have now is: how do I shoehorn this project, which could go on forever (especially with tanru) into something I can spend at most 80 hours on (and I'd prefer 60)? We will need to decide what domains of the language we'll have to leave out: this will need to work on a subset of the language. Of course, I could continue work on the project after this semester.

John Cowan:

I think that you should simply not worry about the internal semantics of tanru, or indeed anything about selbri internals except possibly a place-structure-affecting SE [essentially, one that converts the last component of the tanru at whatever level of nesting, the ter(ter(ter...tertau]. Here's a very sketchy draft of something I wrote once; it actually does stop in the middle of a sentence - I got dragged away to do something else and never went back - that should give some idea of what can be done.

Preliminary Notes for A Lojban Canonicalizer Draft 1.0

by John Cowan

1. Introductory

Lojban is a predicate language; that is, Lojban utterances are for the most part predications. Tools exist in the computer world to process rules and facts expressed in the form of predications, and to answer queries based on those rules and facts. A well-known example is Prolog. Prolog is isomorphic to a small subset of Lojban, but relatively simple processing techniques would suffice to render a much larger set of Lojban utterances Prolog-compatible.

A Lojban Canonicalizer (LC) program would manipulate Lojban utterances, previously parsed by the standard Lojban parser, to produce other Lojban utterances belonging to the Prolog-isomorphic subset. The basic techniques employed include:

  • stripping of metalinguistics
  • argument order standardization
  • semantic transformations
  • expansion of logical connectives

and others to be defined (or thought of) later. The rest of this document details the techniques above.

2. Stripping of Metalinguistics

This is the easiest topic. Lojban allows for a variety of methods for adding metalinguistic comments to mainstream text. There are UI indicators, SEI comments, and TO/TOI parenthetical remarks. All of these can simply be removed from the parsed text. It is forbidden for text at a lower metalinguistic level to refer to text at a higher level, so removal cannot lead to loss of information (although it may lead to loss of context).

3. Argument Order Standardization

The Lojban predication, or bridi, is delivered by the parser as a predicate, or selbri, preceded and/or followed by "terms". There are four kinds of terms: arguments, or sumti; tagged sumti, where the tag either specifies which (numerical) argument of the selbri is involved or indicates a "modal" sumti outside the regular argument structure; bare tags with unspecified sumti; and negation boundaries. In addition, there can be a "prenex" which specifies the quantification of bound variable sumti.

Argument order standardization will rearrange every bridi to get the sumti into a fixed order, either x1, x2, x3, ... selbri or x1, selbri, x2, x3 ... A look-up will be done against the dictionary database to determine how many sumti this selbri should have; any missing sumti will be replaced with the Lojban place-filler sumti, "zo'e". Modal sumti will be moved to the end of the bridi and placed into a canonical order (perhaps alphabetical by tag; the set of tags is potentially unbounded). A prenex will be created with appropriate default quantifications, and all negations will be moved to it.

4. Semantic Transformations

Like other natural languages, Lojban possesses a "deep structure", in the sense (without prejudice to any particular linguistic theories) that some utterances with very different grammar "mean the same thing", with differences of emphasis and the like. The argument-order standardization discussed above involves applying certain transformations which affect sumti. The type discussed here, however, involves the "redundant structures" of Lojban.

In pursuit of linguistic neutrality, Lojban features certain pervasive schemas of grammatical alternatives. The most pervasive by far is the afterthought vs. forethought opposition. In such structures as possessives, logical and non-logical connectives,


This is the final draft of my project proposal:

Project Proposal for 433-603: A Lojban-to-Prolog semantic analyzer.

In this project, we propose developing a semantic analyzer such that, given a text in a subset of the artificial language Lojban, the analyzer will extract information from the text, store it as Prolog clauses, and be asked simple questions on the text content (the questions and answers will both be in Lojban, rather than explicit Prolog queries/clauses). To make the analyzer useful for non-Lojban speakers, output will also be provided in a pidgin English, and phrase markers to the text syntactic structure may also be displayed, time allowing.

Lojban is an artificial language intended for human use, of the type exemplified by Esperanto and Interlingua. It differs from most such languages, in that it has been explicitly based on predicate logic. Predicates serve the role of verbs, predicates with preposed determiners serve the role of nouns, and predications serve as sentences.

There is a number of reasons why this project is of interest. Lojban is a simplified model of a natural language (NL), using predicate logic as its modelling mechanism. Predicate logic also underlies the Prolog into which Lojban text will be transformed by the analyzer. Therefore the task of transferring such information across from Lojban to Prolog will be considerably simpler than doing so for an NL. Lojban has already been shoehorned into a context-free grammar using YACC (this has involved some imaginative use of error recovery, but LALR(1) nature was retained). Thus the task of parsing Lojban text into identifiable grammatical constituencies has already been dealt with: problems in resolving syntactic ambiguity need not distract the analyzer programmer from the more important semantic issues.

Most of the semantic issues complicating logic-based knowledge representation of NL remain in Lojban: higher-order predicates; metalinguistic comments and attitudinals; the ambiguous semantic relationship between head and modifier in word compounds; the representation of numbers, prepositional phrases, relative clauses, non-logical connectives, negation, tense and modality; the distinction between "the" and "a" (echoed in the language's veridical and non-veridical determiners); the distinction between individual and collective plurals; subject-raising; and so forth.

In effect, a Lojban-to-Prolog semantic analyzer would be addressing many of the current issues in NLP knowledge representation, though biased towards predicate logic in the way it does so. The use of a simplified model of NL, and the way the model falls short of capturing NL nuances, will help the analyzer cover much ground quickly, and provide insights in similar analysis of NL proper. (It is claimed that the subset of Lojban implemented would fall short; the author believes the language itself, if it acquires a speech community, will match NL adequately in most usages of language). Less attention would need to be paid to syntactic issues than would be the case with NL. Given how Lojban grammar is structured, modular subsets of Lojban grammar can be implemented in stages in the analyzer. This means that results for simple phrases will become available a very short time into the project.

To keep the project manageable, a subset of the language will have to be considered; this is in line with the Lojban Canonicaliser proposed by John Cowan (see Enclosures. The Canonicaliser will need to be implemented as a preprocessor to what text the analyzer actually sees). Lexically, the subset of Lojban to be implemented will include roughly 500 predicates.

Grammatically, the subset is described as follows, to be implemented in incremental, independent stages:

  1. Simple predications with a known predicate, and with arguments without internal structure (Proper names, logical variables). No quantification other than existential; e.g. "mi prami da" - EXISTS X: LOVES(i, X).
  2. Non-veridical arguments (cf. English "the") based on predicates, with internal arguments; e.g. "mi catra le prami be le pulji" - KILLS(i, x) & LOVES(x, y) & POLICE(y): "I kill the lover of the policeman." Note: strictly speaking, the non-veridical determiner indicates that the entity the speaker has 'in mind' is described by the predicate it precedes, but not uniquely specified by it (cf. veridical determiners). Given the absence of pragmatic content at this early stage of the analyzer, making this distinction will be problematic (it is, after all, inherently ambiguous); it will be dealt with here exactly as NLP deals with the "the"/"an" distinction.
  3. Veridical arguments (cf. English "an") based on predicates, with internal arguments. e.g. "mi catra lo prami be lo pulji" - EXISTS X EXISTS Y: KILLS(i, X) & LOVES(X, Y) & POLICE(Y): "I kill a lover of a policeman."
  4. Resolution of logical connectives; e.g. "mi nelci do .e ko'a" --> "mi nelci do .ije mi nelci ko'a" - LIKES(i, you) & LIKES(i, x1): "I like you and him."
  5. Anaphora and cross-indexing. e.g. "le prenu\i cu prami ri\i" - PERSON(x) & LOVES(x, x): "The person loves him/herself."
  6. Restrictive and non-restrictive relative clauses; e.g. "mi nelci le prenu poi do xebni ke'a" - (EXISTS x: HATES(you, x)) & LIKES(i, x) & PERSON(x): "I like the person you hate."
  7. Higher order predicates; e.g. "lenu mi cadzu cu nandu" - DIFFICULT(event: WALKS(i)): "My walking is difficult."
  8. Prepositional phrases (other than tense and location); e.g. "mi naumau do nelci ko'a" --> "mi zmadu do leni da nelci ko'a" - EXCEEDS(i, you, quantity: LIKES(X, x1)): "I like him more than you do.", e.g. "lo catra ne sepi'o lo mrudakfu" --> "lo catra noi pilno lo mrudakfu" - EXISTS X EXISTS Y: KILLS(X, _) & USES(X, Y, event: KILLS(X, _)) & HAMMER_KNIFE(Y): "an axe-murderer".[1]
  9. Attitudinals; e.g. "mi .ui sidju do" --> "mi sidju do .ije mi gleki mi va'o lenu mi sidju do": HELP(i, you) & HAPPY(i, i) & CONTEXT((state: HAPPY(i, i), event: HELP(i, you)): "I (smile) will help you; I am happy to help you."
  10. Tense (including location), and prepositions of tense (including location). Also includes modality and event contours; e.g. "mi ba'o tavla" --> "lenu mi tavla cu ba'o zei balvi zo'e": AFTERMATH(event: talk(i, _, _, _), _): "I have spoken."
  11. Masses and sets as arguments; e.g. "loi remna cu sipna": "the mass of humans sleep" (Though it is not true at any given moment that: FORALL X: HUMAN(X) => SLEEPS(X))
  12. Non-logical connectors. e.g. "la gilbrt. joi la salivn. cu finti la mikadon." - INVENT(X, mikado) & JOINT_MASS(X, gilbert, sullivan): "G & S (as a joint unit) wrote The Mikado."
  13. Quantification (including numerical, as well as subjective quantifiers such as "enough" and "most"); e.g. "mu le ze mensi cu cucycau": "five of the seven sisters are barefoot".
  14. Negation. Contradictory and scalar. Use of prenexes; e.g. "mi naku ro prenu cu prami": NOT(FORALL X:PERSON(X), LOVES(i,X)); "mi ro prenu na prami": FORALL X:PERSON(X), NOT(LOVES(i,X))[2]
  15. Vocatives, imperatives, interrogatives, and speech protocol words; e.g. "doi skami la sinderelan. mensi ma fe'o": "O Computer: Cinderella is sister to whom? (End of transmission)."

Sections of Lojban Grammar not anticipated to be included in the model:

  1. The mathematical subgrammar of Lojban.
  2. Any analysis of word compounds.
  3. Metalinguistic comments.

The detail of coverage of some sections, particularly tense, will probably have to be curtailed due to time constraints. It is anticipated to have this project take at most 80 hours of work.

John Cowan:

One thing I would suggest is supporting universal quantification as well as existential, since Prolog directly handles universal quantification, whereas existential quantification (except when appearing only in the antecedent of a rule) has to be kludged by skolemization.

On a different note, I think you should consider supporting two additional things: universal quantification a la simple Prolog variables, and imperatives. It would be way cool if a "ko" triggered a look-up so that "ko ciska le broda" came out "print(le_broda)." or the like. That way actual Prolog programming in Lojban would be possible!

  1. Iain Alexander:
    (Really picky:) If you really want the "ka mrudakfu pilno" to be part of the "nu catra", I think it ought to be bound into the selbri with "be":
    lo catra be sepi'o lo mrudakfu

    Otherwise it could just as easily be

    lo te zgike pesepi'o lo grana
    The musician who uses a stick (for walking)
  2. Iain: mi ro prenu na prami:
    I think this is
    naku zo'u mi ro prenu prami

    i.e. the same as the previous example. You need

    mi ro prenu naku prami

Progress Report 1

Well, ladies and germs, this is what I can get my Lojban-PROLOG processor to do so far:

Input text: }mi nelci le klama be le zarci be le ckafi be'o bei le pulji bei le berti

Parser output, after going through LEX:

brivla nelci
brivla klama
brivla zarci
brivla ckafi
brivla pulji
brivla berti
mi nelci le klama be le zarci be le ckafi ku beho ku bei le pulji ku bei le berti ku beho ku vau

PROLOG output

[q(suho(1), _FIPFN, q(suho(1), _FIREF, q(suho(1), _FISES, ckafi(_FISES, _FISZG, _FISZH, _FISZI, _FISZJ), [], zarci(_FIREF, _FISES, _FITIV, _FITIW, _FITIX)), [], q(suho(1), _FIUAD, pulji(_FIUAD, _FIUUR, _FIUUS, _FIUUT, _FIUUU), [], q(suho(1), _FIVID, berti(_FIVID, _FIWCR, _FIWCS, _FIWCT, _FIWCU), [], klama(_FIPFN, _FIREF, _FIUAD, _FIVID, _FIWRT)))), [], nelci(mi, _FIPFN, _FIPFO, _FIPFP, _FIPFQ))]


Branched quantifiers:
E X : broda(X) ; brode(X) => su'o broda ku poi brode
There exists a broda, which brode's, such that...

(These are preferred in Linguistics to the normal plain "E X".

q(E,X,A(X),B(X),C(X)) = (E X: broda(X) ; brode(X)) (C(X))
There exists an A, which Bs, such that C.

E X:
   E Y:
      E Z: ckafi(Z); [] (zarci(Y,Z))
      ; [] (
            E W: pulji(W)
            ; [] (
                  E V: berti(V)
                  ; [] (klama(X,Y,W,V))
   ; [] (nelci(mi,X))

Progress Report 2: Further Lojban->Prolog: relative clauses

At the moment, if "ke'a" isn't there, it isn't assumed; it's pretty certain that, if I don't find "ke'a" there, I'll shove it into the first free place in the relative clause predication.

I've gotten numbers working too, but that's not that spectacular. I'm about to implement the "lo"/"le" distinction.

mi prami le prenu ku poi ke'a citka le cakla
brivla prami
brivla prenu
brivla citka
brivla cakla
mi prami le prenu ku poi keha citka le cakla ku vau kuho
[q(suho(1), _FIODG, prenu(_FIODG, _FIPZL, _FIPZM, _FIPZN, _FIPZO), q(suho(1), _FIRXG, cakla(_FIRXG, _FITTL, _FITTM, _FITTN, _FITTO), [], citka(_FIODG, _FIRXG, _FIRXH, _FIRXI, _FIRXJ)), prami(mi, _FIODG, _FIODH, _FIODI, _FIODJ))]


E X:
     (E Y:
          cakla(Y); [] (citka(X,Y))

Progress Report 3: Lojban->Prolog: conjunctions

mi .e ko'a cu prami ro lo nanmu gi'e xebni ro lo ninmu

Note: c(C,X,Y) means C(X,Y), where C is some binary conjunction. Here it is ".e", meaning AND.

[c(e, q(la, mi, [], [], c(e, q(ro, _FIPVD, nanmu(_FIPVD, _FIRSQ, _FIRSR, _FIRSS, _FIRST), [], prami(mi, _FIPVD, _FIPVE, _FIPVF, _FIPVG)), q(ro, _FISRZ, ninmu(_FISRZ, _FIUPM, _FIUPN, _FIUPO, _FIUPP), [], xebni(mi, _FISRZ, _FISSA, _FISSB, _FISSC)))), q(la, koha, [], [], c(e, q(ro, _FIPVD, nanmu(_FIPVD, _FIRSQ, _FIRSR, _FIRSS, _FIRST), [], prami(koha, _FIPVD, _FIPVE, _FIPVF, _FIPVG)), q(ro, _FISRZ, ninmu(_FISRZ, _FIUPM, _FIUPN, _FIUPO, _FIUPP), [], xebni(koha, _FISRZ, _FISSA, _FISSB, _FISSC)))))]


         All x (man x) loves(mi,x) ,
         All y (woman y) hates(mi,x)),
         All x (man x) loves(koha,x) ,
         All y (woman y) hates(koha,y))).

Note that I'm using iota quantification for names and anaphors; this is left in for ease of anaphor resolution later, and can be stripped out.

Progress Report 4: Prolog: event abstractions

mi nelci lenu ko'a banli ro lo xelso ku poi ke'a prami le gugde kei ku poi ke'a cafne .e la kserkes. gi'e zutse le stizu

(Takes 16 seconds to parse).

     q(suho(1), _FJKVW,
          q(ro, _FJLAZ,
            xelso(_FJLAZ, _FJLGI, _FJLGJ, _FJLGK, _FJLGL),
            q(suho(1), _FJLKF,
              gugde(_FJLKF, _FJLPO, _FJLPP, _FJLPQ, _FJLPR), [],
              prami(_FJLAZ, _FJLKF, _FJLKG, _FJLKH, _FJLKI)
            banli(koha, _FJLAZ, _FJLBA, _FJLBB, _FJLBC)
          _FJLSS, _FJLST, _FJLSU),
        cafne(_FJKVW, _FJLWO, _FJLWP, _FJLWQ, _FJLWR),
        nelci(mi, _FJKVW, _FJKUU, _FJKUV, _FJKUW)
     nelci(mi, [kserkses], _FJKUU, _FJKUV, _FJKUW)
   q(suho(1), _FJMDN,
     stizu(_FJMDN, _FJMID, _FJMIE, _FJMIF, _FJMIG), [],
     zutse(mi, _FJMDN, _FJMDO, _FJMDP, _FJMDQ)
         {E X
              A Y

                {E Z
                banli(koha, Y)
            nelci(mi, X)
         nelci(mi, [kserkses])
     E W:stizu(W);zutse(mi,W)



2nd baseline as of 23 June 1991, which is original baseline 20 July 1990 incorporating technical fixes 1-28. This version includes change proposals 1-32 to that baseline, excluding changes 21 and 28 which are assumed annulled.

Prepared by The Logical Language Group, Inc.  2904 Beau Lane, Fairfax VA 22031 USA  703-385-0273

In accordance with the Logical Language Group, Inc. policy, this material constitutes Lojban language definition materials and is hereby placed irrevocably in the public domain. Signed: Robert LeChevalier, President, The Logical Language Group, Inc.

We request the following when this material is used in derived works: state that derivation and that the material baseline is preliminary, and provide the name and address of the Logical Language Group, Inc. as a source of further bonafide information about the material and about Lojban. We ask that all users of this material verify to ensure that they are using the latest material. Barring unexpected major problems there will be no change to this material prior to completion of the Lojban dictionary later in 1993, whereupon a 3rd baseline will declared.

Explanation of notation:

All rules have the form:

name<number> = bnf-expression

which means that the grammatical construct "name" is defined by "bnf-expression". The number cross-references this grammar with the rule numbers in the YACC grammar. The names are the same as those in the YACC grammar, except that subrules are labeled with A, B, C, ... in the YACC grammar and with 1, 2, 3, ... in this grammar. In addition, rule 971 is "simple_tag" in the YACC grammar but "stag" in this grammar, because of its frequent appearance, and rule 32 is "free_modifier" in the YACC grammar but "free" in this grammar.


  1. Names in lower case are grammatical constructs.
  2. Names in UPPER CASE are selma'o (lexeme) names, and are terminals (i.e. they have no internal grammar, but are replaced by any of the Lojban words in that "selma'o".
  3. Concatenation is expressed by juxtaposition with no operator symbol.
  4. "|" represents alternation (choice).
  5. "[]" represents an optional element.
  6. "&" represents and/or ("A & B" is the same as "A | B | A B").
  7. "..." represents optional repetition of the construct to the left. Left-grouping is implied; right-grouping is shown by explicit self-referential recursion with no "..."
  8. "()" serves to indicate the grouping of the other operators. Otherwise, "..." binds closer than &, which binds closer than |.
  9. "#" is shorthand for "[free ...]", a construct which appears in many places.
  10. "//" encloses an elidable terminator, which may be omitted (without change of meaning) if no grammatical ambiguity results.
text<0>  =   [NAI] [(CMENE ... #) | (indicators & free ...)] [joik-jek] text-1
text-1<2>  =   [(I [jek | joik] [[stag] BO] #) ... | NIhO ... # ] paragraphs
paragraphs<4>  =   paragraph [NIhO ... # paragraphs]
paragraph<10>  =   paragraph-1 [I [jek | joik] # [paragraph-1] ...
paragraph-1<11>  =   paragraph-2 [I [jek | joik] [stag] BO # paragraph-1]
paragraph-2<12>  =   utterance | [prenex | tag] TUhE # text-1 /TUhU#/
utterance<20>  =   ek # | gihek # | quantifier | NA | term ... /VAU#/ | prenex | relative-clauses | links | linkargs | sentence
prenex<30>  =   term ... ZOhU #
sentence<40>  =   bridi-tail | sentence-1
sentence-1<41>  =   term ... [CU #] bridi-tail | gek sentence-1 gik sentence | prenex sentence
bridi-tail<50>  =   bridi-tail-1 [gihek [stag] KE # bridi-tail /KEhE#/ tail-terms] ...
bridi-tail-1<51>  =   bridi-tail-2 [gihek # bridi-tail-2 tail-terms] ...
bridi-tail-2<52>  =   bridi-tail-3 [gihek [stag] BO # bridi-tail-2 tail-terms]
bridi-tail-3<53>  =   selbri tail-terms | gek-bridi-tail
gek-bridi-tail<54>  =   gek bridi-tail gik bridi-tail-3 | tag KE gek-bridi-tail /KEhE#/ | NA # gek-bridi-tail
tail-terms<71>  =   [term ...] /VAU#/
term<81>  =   sumti | (tag | FA #) (sumti | /KU#/) | termset | NA KU #
termset<83>  =   NUhI gek term ... /NUhU#/ gik term ... /NUhU#/ | NUhI term ... /NUhU#/ joik-ek # term ... /NUhU#/
sumti<90>  =   sumti-1 [(ek | joik) [stag] KE # sumti /KEhE#/] ...
sumti-1<91>  =   sumti-2 [joik-ek sumti-2] ...
sumti-2<91>  =   sumti-3 [(ek | joik) [stag] BO # sumti-2]
sumti-3<93>  =   sumti-4 | gek sumti gik sumti-3
sumti-4<94>  =   [quantifier] sumti-5 [relative-clauses] | quantifier selbri /KU#/ [relative-clauses]
sumti-5<96>  =   (LAhE # | NAhE BO #) [relative-clauses] sumti /LUhU#/ | KOhA # | lerfu-string /BOI#/ | LA CMENE ... # | (LA | LE) sumti-tail /KU#/ | LI mex /LOhO#/ | ZO any-word # | LU text /LIhU/ # | LOhU any-word ... LEhU # | ZOI any-word anything any-word #
sumti-tail<111>  =   [sumti-5 [relative-clauses]] sumti-tail-1 | relative-clauses sumti-tail-1
sumti-tail-1<112>  =   [quantifier] selbri [relative-clauses] | quantifier sumti
relative-clauses<121>  =   relative-clause [ZIhE relative-clause] ...
relative-clause<122>  =   GOI term /GEhU#/ | NOI sentence /KUhO#/
selbri<130>  =   [tag] selbri-1
selbri-1<131>  =   selbri-2 | NA selbri
selbri-2<132>  =   selbri-3 [CO # selbri-2]
selbri-3<133>  =   selbri-4 ...
selbri-4<134>  =   selbri-5 [joik-jek selbri-5] ...
selbri-5<135>  =   selbri-6 [(jek | joik) BO # selbri-5]
selbri-6<136>  =   tanru-unit [BO selbri-6] | [NAhE #] guhek selbri gik selbri-6
tanru-unit<150>  =   tanru-unit-1 [CEI # tanru-unit-1] ...
tanru-unit-1<151>  =   tanru-unit-2 [linkargs]

tanru-unit-2<152>  =   BRIVLA # | GOhA [RAhO] # | KE selbri-3 /KEhE#/ | ME sumti /MEhU#/ [MOI] # | (number | lerfu-string) MOI # | NUhA mex-operator | SE # tanru-unit-2 | JAI [tag] tanru-unit-2 | any-word (ZEI any-word) ... | NAhE # tanru-unit-2 | NU [NAI] # [joik-jek NU [NAI] #] ... sentence /KEI#/
linkargs<160>  =   BE term [links] /BEhO#/
links<161>  =   BEI term [links]
quantifier<300>  =   number /BOI#/ | VEI mex /VEhO#/
mex<310>  =   mex-1 [operator mex-1] ... | FUhA rp-expression
mex-1<311>  =   mex-2 [BO operator mex-1]
mex-2<312>  =   operand | [PEhO] operator mex-2 ... /KUhE#/
rp-expression<330>  =   rp-operand rp-operand operator
rp-operand<332>  =   operand | rp-expression
operator<370>  =   operator-1 [joik-jek operator-1] ...
operator-1<371>  =   operator-2 | guhek operator-1 gik operator-2
operator-2<372>  =   mex-operator # | KE operator /KEhE#/
mex-operator<374>  =   SE # mex-operator | NAhE # mex-operator MAhO mex /TEhU#/ | NAhU selbri /TEhU#/ | VUhU
operand<380>  =   operand-1 [(ek | joik) [stag] KE # operand /KEhE#/] ...
operand-1<382>  =   operand-2 [joik-ek operand-2] ...
operand-2<383>  =   operand-3 [(ek | joik) [stag] BO # operand-2]
operand-3<385>  =   quantifier | lerfu-string /BOI#/ | NIhE selbri /TEhU#/ | MOhE sumti /TEhU#/ | JOhI mex-2 ... /TEhU#/ | gek operand gik operand-3 | (LAhE # | NAhE BO #) operand /LUhU#/
number<812>  =   PA [PA | lerfu-word] ...
lerfu-string<817>  =   lerfu-word [PA | lerfu-word] ...
lerfu-word<987>  =   BY | any-word BU | LAU lerfu-word | TEI lerfu-string FOI
ek<802>  =   [NA] [SE] A [NAI]
gihek<818>  =   [NA] [SE] GIhA [NAI]
jek<805>  =   [NA] [SE] JA [NAI]
joik<806>  =   [SE] JOI [NAI] | interval | GAhO interval GAhO
interval<932>  =   [SE] BIhI [NAI]
joik-ek<421>  =   joik # | ek #
joik-jek<422>  =   joik # | jek #
gek<807)  =   [SE] GA [NAI] # | joik GI # | stag gik #
guhek<808>  =   [SE] GUhA [NAI] #
gik<816>  =   GI [NAI] #
tag<491>  =   tense-modal [joik-jek tense-modal] ...
stag<971>  =   simple-tense-modal [(jek | joik) simple-tense-modal] ...
tense-modal<815>  =   simple-tense-modal # | FIhO selbri /FEhU#/
simple-tense-modal<972>  =   [NAhE] [SE] BAI [NAI] [KI] | [NAhE] time & space & CAhA [KI] | KI | CUhE
time<1030>  =   ZI & time-offset ... & ZEhA [PU [NAI]] & interval-modifier
time-offset<1033>  =   PU [NAI] [ZI]
space<1040>  =   VA & space-offset ... & space-interval & (MOhI space-offset)
space-offset<1045>  =   FAhA [NAI] [VA]
space-interval<1046>  =   ((VEhA & VIhA) [FAhA [NAI]]) & FEhE interval-modifier
interval-modifier<1050>  =   interval-property & ZAhO
interval-property<1051>  =   number ROI [NAI] | TAhE [NAI]
free<32>  =   SEI # [term ... [CU #]] selbri /SEhU/ | SOI sumti [sumti] /SEhU/ | vocative selbri [relative-clauses] /DOhU/ | vocative relative-clauses sumti-tail-1 /DOhU/ | vocative CMENE ... # [relative-clauses] /DOhU/ | vocative [sumti] /DOhU/ | (number | lerfu-string) MAI | TO text /TOI/ | XI number /BOI/ | XI lerfu-string /BOI/ | XI VEI mex /VEhO/
vocative<415>  =   (COI [NAI]) ... & DOI
indicators<411>  =   [FUhE] indicator ...
indicator<413>  =    (UI | CAI) [NAI] | Y | POhA | DAhO | FUhO

The following rules are non-formal:

word<1100>  =   [BAhE | PEhA] any-word [indicators]
any-word  =   "any single word (no compound cmavo)"
anything  =   "any text at all, whether Lojban or not"
null<1101>  =   any-word SI | utterance SA | text SU FAhO is a universal terminator and signals the end of parsable input.

06/01/93 Lojban baseline rafsi list

The Logical Language Group, Inc., 
2904 Beau Lane, Fairfax VA 22031-1303 USA

In accordance with the Logical Language Group, Inc. policy, this material constitutes Lojban language definition materials and is hereby placed irrevocably in the public domain. Signed: Robert LeChevalier, President, The Logical Language Group, Inc.

We request the following when this material is used in derived works: state that derivation and that the material baseline is preliminary, and provide the name and address of the Logical Language Group, Inc. as a source of further bonafide information about the material and about Lojban. We ask that all users of this material verify to ensure that they are using the latest material. Barring unexpected major problems there will be no change to this material prior to completion of the Lojban dictionary later in 1993, whereupon a 3rd baseline will declared.

Lojban lujvo-MAKING

1. Long-form rafsi for gismu are derived directly from the gismu-form: the gismu itself for final position, and the gismu with final vowel replaced by 'y' for non-final position; cmavo have no long-form rafsi. As you will note below, many gismu have no short-form rafsi and must use the long forms in both initial and final positions.

Short form rafsi are derived from a limited set of possibilities:

C1V1C2C3V2 gismu have[1]

    CVC forms from  C1V1C2 or C1V1C3
    CVV forms from  C1V1V2 (C1V1'V2)
    CCV forms from  C2C3V2 or rarely C1C2V1

C1C2V1C3V2 gismu have

    CVC forms from  C1V1C3 or C2V1C3
    CVV forms from  C1V1V2 (C1V1'V2) or C2V1V2 (C2V1'V2)
    CCV forms from  C1C2V1

2. Any cmavo or other word may be incorporated into a lujvo independently of the rafsi system using the cmavo "zei": 'any-word' zei 'any-word' forms a brivla

3. All forms of lujvo built out of exactly the same component words/rafsi have identical meanings. There is no stigma attached to use of long forms, which can be especially useful when your audience is not familiar with the rafsi, and is not likely to be looking up words in a word-list. Long-forms also may be preferred to guessing when you are too lazy to use a list yourself, and you suspect that your audience will be using one - there is nothing like trying to interpret a lujvo when the unambiguously resolved components resolve into something totally strange.

4. The rules for building lujvo-forms are fairly simple.

  • Rules for Lojban word forms - The lujvo must be formed according to Lojban's word-formation rules. The constraints of Lojban word forms forbid any lujvo from ending in a consonant, so that words most commonly found in the final position of a tanru have been prioritized to have a rafsi that ends in a vowel. However, words found in initial positions often form better sounding combinations if their rafsi end in a consonant. (Also, because we usually recognize words by the consonants in them rather than the vowels, the rafsi of form CVV and CV'V are harder to memorize.
    Certain sounds are forbidden to occur next to each other (so-called 'impermissible medial' consonants), and must be separated by a 'hyphen'- sound, the "uh" of "sofa", represented in Lojban by the letter 'y' (this letter is found only as a hyphen, in lerfu, the words for letters of the alphabet, and along to represent the hesitation noise. It is thus not normally considered a 'V' is the C/V convention scheme. Indeed, "CyC" is considered a consonant cluster in Lojban morphology, albeit a hyphenated one). In addition, a CVV or CV'V rafsi at the beginning of any lujvo must either carry the penultimate stress, it must be 'glued' to the remaining rafsi with a syllabic 'r' or 'n' sound, or the rafsi falls off into a separate word, a cmavo. (In addition, a CVV or CV'V rafsi followed by another CVV or CV'V rafsi in a 2-term lujvo must have the 'r' or 'n' added, or the consonant cluster mandatory in any brivla in not present, and the rafsi break up into two separate cmavo.)
  • Multiple rafsi to choose from - Because of these rules, there is usually more than one rafsi usable for each gismu. The one to be used is simply whichever sounds best to the speaker/writer. There are many valid combinations of the possible rafsi. Any rafsi for a given word is equally valid in place of another, and all mean the same thing. There is an optional scoring component to the lujvo-making algorithm which attempts to systematically pick the 'best' one; this algorithm tries for short forms and tends to push more vowels into the words to make them easier to say. The Japanese, Chinese, and Polynesian speakers will prefer this; Russians have a different aesthetic, since they are used to saying consonant clusters. But these are not necessarily the criteria you will wish to use.
  • lujvo have ONE meaning - While a tanru is ambiguous, having several possible meanings, a lujvo (one that would be put into the dictionary) has one meaning. Just like gismu, a lujvo is a predicate which encompasses one area of the semantic universe, with one set of places. Hopefully this is the most 'useful' or 'logical' of the possible semantic spaces. A known source of linguistic drift in Lojban will be as Lojbanic society evolves, and the concept represented by a sequence of rafsi that is most 'useful' or 'logical' changes. At that time, it might be decided that we want to redefine the lujvo to assume the new meaning. lujvo must not be allowed to retain two meanings. So those that maintain the dictionary will be ever watchful of tanru and lujvo usage to ensure this standard is kept.
    One should try to be aware of the possibility of prior meanings of a new lujvo, especially if you are writing for 'posterity'. If a lujvo is invented which involves the same tanru as one that is in the dictionary, and is assigned a different meaning (including a different place structure), linguistic drift results. This isn't necessarily bad; it happens in every natural language. You communicate quite well in English even though you don't know most words in the dictionary, and in spite of the fact that you use some words in ways not found in the dictionary. Whenever you use a meaning different from the dictionary definition, you risk a reader/listener using the dictionary and therefore misunderstanding you. One major reason for having a standard lujvo scoring algorithm is that with several possible rafsi choices to consider, a dictionary is most efficient by putting the definition under the single most preferred form.
    You may optionally mark a nonce word that you create without checking a dictionary by preceding it with "za'e". "za'e" simply tells the listener that the word is a nonce word, and may not agree with a dictionary entry for that sequence of rafsi. The essential nature of human communication is that if the listener understands, then all is well. Let this be the ultimate guideline for choosing meanings and place structures for invented lujvo.
  • Zipf's law and lujvo - This complication is simple, but is the scariest. Zipf's Law (actually a hypothesis), says that the length of words is inversely proportional to their frequency of usage. The shortest words are those which are used more; the longest ones are used less. The corollary for Lojban is that commonly used concepts will tend to be abbreviated. Speakers will choose the shortest form for frequently expressed ideas that gets their meaning across, even at the cost of accuracy in meaning. In English, we have abbreviations and acronyms and jargon, all of which are words for complex ideas used with high frequency by a group of people. So they shortened them to convey the often-used information more rapidly.
    The jargon-forming interpretation of Zipf's Law may be a cause of multiple meanings of words in the natural languages, especially of short words. If true, it threatens the Lojban rule that all lujvo must have one meaning. The Lojbanist thus resigned accepts a complication in lujvo-making: A perfectly good and clear tanru may have to be abbreviated when made into a lujvo, if the concept it represents likely will be used so often as to cause Zipf's Law to take effect.
    Thus, given a tanru with grouping markers, abstraction markers, and other cmavo in it to make the tanru syntactically unambiguous, in many cases one drops some of the cmavo to make a shorter (incorrect) tanru, and then uses that one to make the lujvo.
    This doesn't lead to ambiguity, as it might seem. A given lujvo still has exactly one meaning and place structure. But now, more than one tanru is competing for the same lujvo. This is not as difficult to accept or allow for as it might seem: more than one meaning for a single tanru was already competing for the 'right' to be used for the lujvo. Someone has to use judgement in deciding which one meaning is to be chosen over the others. This judgement will be made on the basis of usage, presumably by some fairly logical criteria.
    If the lujvo made by a shorter form of tanru is already in use, or is likely to be useful for another meaning, the wordmaker then retains one or more of the cmavo, preferably ones that clearly set this meaning apart from the shorter form meaning that is used or anticipated. In Lojban, therefore, shorter lujvo will be used for a less complicated concept, possibly even over a more frequent word. If two concepts compete for a single rafsi sequence, the simpler concept will take a shorter form, and the more complex concept will have some indication of its more complex nature added into the word structure. It is easier to add a cmavo to clarify the meaning of a more complex term than it is to find a good alternate tanru for the simpler term.
    A good lujvo-composer considers the listener, and a good lujvo interpreter remebers the difficulties of lujvo-making. If someone hears a word he doesn't know, decomposes it, and gets a tanru that makes no sense for the context, he knows that the grouping operators may have been dropped out, he may try alternate groupings. Or he may try using the verb form of the concept instead of the first sumti, inserting an abstraction operator if it seems plausible. Plausibility is key to learning new ideas, and evaluating unfamiliar lujvo.
  1. Note: C and V in abbreviations of this sort stand for any Lojban consonant and vowel, respectively. The apostrophe is the Lojban "'", which is considered neither a consonant nor a vowel.)


The rules for the lujvo-making algorithm are stated formally. This may cause it to appear intimidating to a casual reader, and to seem harder than it really is to use. The following brief form, is more practical to learn.

  1. Find all rafsi forms for the component words. 5-letter forms can only occur in final position; 4-letter+y form, and CVC short forms can only occur in non-final positions. Make all possible combinations of the rafsi for the component words, keeping the order of the component words.
  2. Between any two impermissible medial consonants (see 5c of the formal algorithm), stick a y.
  3. Where there is a consonant triple formed where two rafsi join, if it is impermissible (see 5d of the formal algorithm), stick in a y where they join.
  4. If you have a CVV rafsi at the beginning, add in an r hyphen after it. An n is used instead if the letter after the hyphen is also an r. However, in a tanru with only 2 parts, do not add a hyphen when the second rafsi is a CCV-form.
  5. Always stick in a y hyphen after a 4-letter form, which is the gismu without its final letter.
  6. Perform the "tosmabru" test. Starting from the left, look for a sequence of 2-or-more CVC rafsi ending with (the first) hyphen 'y', or one-or-more CVC rafsi followed by an end-of-word CVCCV full word rafsi with a permissible medial as the CC. If either case occurs, look at each consonant pair, and if all of them are permissible initial consonant pairs, insert a 'y' hyphen between the first consonant pair.


The following is the official algorithm for generating Lojban lujvo (complex brivla, or predicate words), given a known tanru (metaphor) and a complete list of gismu (Lojban primitive roots) and their assigned rafsi (affixes). Note that Lojban does not require use of the optimal, or "best" form of a word. Poetic usage allows any of the valid word forms created by this algorithm to be used under appropriate circumstances.

Given an n-term tanru and the instruction to find the highest-scoring lujvo:

  1. For all terms except the final term, look up or generate all of the rafsi (3- and 4-letter forms). Three-letter forms will be of the structure CVC, CCV, CVV, or CV'V (the apostrophe is not counted as a letter in any Lojban rule). A standard gismu list gives the three-letter rafsi for each gismu and for each cmavo with an assigned rafsi. You can memorize the list also. This is not difficult if you use the language much: the set of possible rafsi for each word is limited, and because almost all possible rafsi have an assigned meaning, the more you know, the easier it is to learn the rest by elimination.
    • Given a CCVCV gismu C1C2V1C3V2, the CVC rafsi, if any, will be C1V1C3 or C2V1C3. The CVV/CV'V rafsi, if any, will be C1V1(')V2 or C2V1(')V2. The CCV rafsi, if any, will be C1C2V1. Very few gismu have both a CCV and a CVV/CV'V assigned.
    • Given a CVCCV gismu C1V1C2C3V2, the CVC rafsi, if any, will be C1V1C2. The CVV/CV'V rafsi, if any, will be C1V1(')V2. The CCV rafsi, if any, will be C1C2V2, or rarely, C1C2V1.
    • The rafsi for cmavo is assigned more arbitrarily. A CVV/CV'V form cmavo will often be its own rafsi, but when this isn't possible, the final letter is changed. A single letter, usually an arbitrary consonant, is added to a CV cmavo to make its rafsi.
    • The four-letter rafsi form for any gismu is formed by dropping the final vowel from the gismu (which is then effectively replaced by "y" in the lujvo).
  2. For the final term, look up or generate all of the three-letter rafsi, omitting any CVC-form rafsi since a lujvo cannot end in a consonant. Then, for this position only, add in the full gismu itself as a '5-letter rafsi'.
  3. Since most cmavo with rafsi have CVC rafsi and none has a 5-letter form, few cmavo can occur in the final position of a tanru used as the basis of a lujvo. cmavo in those positions are rare anyway, the exceptions being PA+MOI numbers. If a cmavo in any position has no rafsi, then it cannot be incorporated into the lujvo. Consider rephrasing or using "zei" to form an 'any-word' compound.
  4. Form all of the ordered combinations of these rafsi, one rafsi per corresponding term ordered in the sequence of their corresponding terms.
  5. Audible 'hyphens' may be necessary between some adjacent rafsi to make the word pronouncible, understandable, well-formed, and not prone to breaking up into two-or-more smaller words. Hyphens are never optional; they are not permitted in-between rafsi unless they are required. Right-to-left testing is recommended for reasons discussed below:
    1. If there are more than two terms, an initial CVV or CV'V rafsi will fall off and be heard as a separate cmavo. It must therefore be glued on with the letter 'r', which nominally stands in a syllable by itself. For example sai + zba + ta'u becomes sairzbata'u (syllabized as sai,r,zba,TA'u). If the initial rafsi is a CV'V, the 'r' may be joined onto the second syllable. Thus sa'i + zba + ta'u becomes sa'irzbata'u (syllabized as sa,'ir,zba,TA'u). If the first consonant of the second syllable is an 'r', the gluing 'hyphen' must be the letter 'n', instead of 'r' because doubled consonants are not permitted in Lojban. Thus sai + rai + ta'u becomes sainraita'u (syllabized as sai,n,rai,TA'u and NOT sain,rai,TA'u). 'n' is NOT permitted unless the adjacent 'r' forces it.

      If there are exactly two terms, and the initial term is a CVV or CV'V rafsi AND the final term is a 5-letter rafsi, an 'r' hyphen is needed as described above to prevent the initial rafsi from falling off into a separate CVV or CV'V cmavo. As above, an 'n' is used as glue if and only if an 'r' cannot be used. Thus sai + taxfu needs hyphen 'r' to become sairtaxfu (sai,r,TAX,fu). sai + ranji needs hyphen 'n' to become sainranji (sai,n,RAN,ji).

      If there are exactly two terms, and the initial term is a CVV or CV'V rafsi AND the final term is a CVV or CV'V rafsi, an 'r' hyphen is needed, because the lujvo is not well-formed, lacking a consonant cluster, and will fall apart into two CVV or CV'V cmavo. As above, an 'n' is used as glue if and only if an 'r' cannot be used. Thus sai + ta'u needs hyphen 'r' to become sairta'u (sai,r,TA,'u). sai + rai needs hyphen 'n' to become sainrai (SAI,n,rai). Note that hyphen in a syllable by itself is not counted in determining penultimate stress. However, if joined onto a vowel syllable as when ta'u + sai forms ta'ursai, the vowel syllable is counted and is stressed if penultimate (ta,'UR,sai).

      If there are exactly two terms, and the initial term is a CVV or CV'V rafsi AND the final term is a CCV rafsi, no hyphen is needed, because the lujvo is well-formed, having a consonant cluster, and penultimate stress falls on part of the CVV/CV'V rafsi, preventing it from falling off into a separate word. Thus sai + zba needs no hyphen 'r' to form saizba.

    2. Put 'y' after any 4-letter rafsi form (e.g. zbasysai). Do not count a syllable centered on this hyphen in determining penultimate stress. (e.g. ZBAS,y,sai or ZBA,sy,sai).
    3. Put 'y' at any proscribed C/C joint (impermissible medial consonant pair, e.g. nunynau). The following are the rules summarizing proscribed medials:

      Given that the consonant pair is defined as C1C2, that b, d, g, j, v and z are voiced consonants, c, f, k, p, s, t, and x are unvoiced consonants, and l, m, n, and r are nasal/liquid consonants.

      1. C1 cannot be the same as C2. e.g.
      2. If C1 is voiced, then C2 must either be voiced or nasal/liquid. If C1 is unvoiced, then C2 must be either unvoiced or nasal/liquid.
      3. Both C1 and C2 cannot be among c, j, s, or z.
      4. *cx, *kx, *xc, *xk, and *mz are not permitted.

      Do not count a syllable centered on this hyphen in determining penultimate stress. (e.g. NUN,y,nau or NU,ny,nau).

    4. Put y at any proscribed C/CC joint (e.g. nunydji). The following are the rules for proscribed triples:

      The first two consonants of a consonant triple in a Lojban brivla must be restricted as for permissible medial consonant pairs per the above. The second pair within the triple must be a permissible initial consonant pair. Since you cannot get a triple in a lujvo unless the latter two consonants are part of a CCV rafsi, testing the first two consonants per c) is sufficient for this part of the test. In addition, there are a few triples that meet the above conditions but are still not pronounceable so as to be easily and uniquely resolvable from other combinations. Hence they are also not permitted, and require a hyphen. These triples are:

      n,dj   n,dz   n,tc   n,ts

      Do not count a syllable centered on this hyphen in determining penultimate stress. (e.g. NUN,y,dji or NU,ny,dji).

    5. Test all forms starting with a series of CVC rafsi for "tosmabru failure", which means that the first CV will fall off into a separate cmavo, leaving the rest a valid lujvo. ("*tosmabru was a trial word that was found to so break up, and is used as the archetypal example of an invalid lujvo according to this rule.) This is a tricky rule, but not that common a circumstance, because the CV falls off only if a valid lujvo remains. The following are a set of simple short cuts to test for and correct all "tosmabru" situations. (The same situation with an apparent le'avla form remaining does not break up simply because such forms are forbidden to le'avla. This is the so-called "*slinku'i" rule for le'avla: if you stick a CV cmavo on the front of a le'avla and it forms a valid lujvo, then the le'avla is NOT valid.)

      If a series of rafsi has the pattern 'CVC ... CVC + X' , where no 'y' hyphens have been installed between any two of the CVC, there may be a "tosmabru" problem.

      • If X is a CVCCV long rafsi with a permissible initial as the consonant cluster, then even a single CVC rafsi on the front requires a "tosmabru test" (as in tos + mabru which would break up into to + smabru). You are specifically testing here to ensure that the CV on the front does not fall off, leaving a lujvo composed of a series of CCV rafsi.
      • If X is any rafsi or partial-lujvo that causes a y hyphen to be installed between the previous CVC and itself by one of the above rules, and there are at least two CVC rafsi preceding, you must also test for "tosmabru" break up (as in tos + mab + bai which would have added a 'y' hyphen between the last two terms, and would break up into to + smabybai, where "smab" is a hypothetical 4-letter rafsi form). You are testing here to avoid the initial CV falling off to leave a lujvo with a spurious CCVC 4-letter rafsi form just before the X component. NOTE THAT THE RULES DO NOT DEPEND ON THERE ACTUALLY BEING RAFSI THAT WOULD MAKE THE BROKEN UP WORD POSSIBLE (smab- is not the 4-letter form for any gismu currently assigned, but the rules do not presume that the listener knows which rafsi are real - they are based ONLY on the forms if the words.)

      The "tosmabru" test is:

      Examine all the C/C joints between the CVC rafsi, and between the last CVC and the X term.

      If the ALL of those C/C joints, as well as the CC in X, if we are dealing with the CVCCV case for X, are "bridged" by permissible initials, listed in Section III or the back of the gismu list, then the trial word will break up into a cmavo and a shorter brivla ("tosmaktu" would thus be valid, unlike "tosmabru").

      If any C/C joint is unbridged, i.e., is impermissible as an initial CC, the trial word will not break up. It has passed the "tosmabru test".

      Only the first joint in a trial word needs to be unbridged in order to ensure resolvability. Thus: Install y as a hyphen at the first bridged joint if the "tosmabru" test fails (e.g. tosymabru).

      The 'lazy Lojbanist' "tosmabru test" is to add a hyphen any time you have a CVC rafsi followed by a CV... of 5-or-more letters, where the first C/C joint forms a permissible initial. This is NOT a correct algorithm - it will put in hyphens that are not necessary, thus resulting in words that are technically invalid. However, for nonce lujvo-making, if an unnecessary hyphen is present, the word can be successfully and unambiguously analyzed.

      On the other hand, if a "tosmabru" hyphen is omitted, the word is likely to be incorrectly analyzed.

      Note that the 'tosmabru test' requires all hyphens based on other rules to have been determined before conducting the test. This is why this step occurs last.

  6. Evaluate all combinations and select the word with the highest score, using some algorithm.


This algorithm was devised by Bob and Nora LeChevalier in 1989. It is not the only permitted algorithm, but it usually gives a choice that people find preferable. This is the algorithm encoded in the lujvo-making program sold by la lojbangirz. The algorithm may be changed in the future. Note that the algorithm basically encodes a hierarchy of priorities, preferring short words (counting an apostrophe as a half of a letter), then words with fewer hyphens, then words with fewer syllables and/or more vowels.

Values are attached to various properties of the lujvo. The score is the sum of these values.

  1. Count the number of hyphens (h), including 'y', 'r', or 'n'.
  2. Count the number of vowels (v) not including 'y'.
  3. Count the number of apostrophes (a).
  4. Count the total number of characters including hyphens and apostrophes (l).
  5. For each rafsi component, find the value in the following list. Sum this total (r):
         Cvv        (sai)         8
         CCVC       (zbas)        4
         CCV        (zba)         7
         -CCVCV     (-zbasu)      3
         CV'V       (ta'u)        6
         CVCC       (sarj)        2
         CVC        (nun)         5
         -CVCCV     (-sarji)      1

The score is then 32500 - (1000 * l) + (500 * a) - (100 * h) + (10 * r) + v In case of ties, there is no preference. This should be rare.

The following examples use the rafsi:

CVC = nun CCV = zba Cvv = nau, sai
CVCCV = sarji  CCVC- = zbas-  CV'V = ta'u

Stress is shown explicitly using capitalization in these examples. Being algorithmic (always penultimate), it does not have to be explicitly shown when these words are actually used.

   zba + sai                           ZBAsai
32500 - (1000 * 6) + (500 * 0) - (100 * 0) + (10 * 15) + 3 = 26653
   nun + y + nau                       NUNynau
32500 - (1000 * 7) + (500 * 0) - (100 * 1) + (10 * 13) + 3 = 25533
   sai + r + zba + ta'u                sairzbaTA'u
32500 - (1000 * 11) + (500 * 1) - (100 * 1) + (10 * 21) + 5 = 22115
   zba + zbas + y + sarji              zbazbasySARji
32500 - (1000 * 13) + (500 * 0) - (100 * 1) + (10 * 12) + 4 = 19524

rafsi list

 gismu                   berti ber          nort  briju bij          offi
 or                      h                        ce
 cmavo CVC   CCV    CVV  besna ben          brai  brito rit          Brit
 English keyword         n                        ish
                         betfu bef     be'u abdo  broda rod          pred
 bacru       ba'u   utte men                      icate var 1
 r                       betri bet          trag  brode       bo'e   pred
 badna         banana    edy                      icate var 2
 badri    dri       sad  bevri bev     bei  carr  brodi         predicate
 bajra baj          run  y                        var 3
 bakfu baf          bund bi biv        8          brodo         predicate
 le                      bi'i  biz          unor  var 4
 bakni bak          bovi dered interval           brodu         predicate
 ne                      bidju         bead       var 5
 bakri         chalk     bifce bic          bee   bruna bun     bu'a brot
 baktu         bucket    bikla bik          whip  her
 balji         bulb      bilga big          obli  bu bus      bu'i   word
 balni         balcony   ged                      to lerfu
 balre       ba'e   blad bilma       bi'a   ill   bu'a  bul          some
 e                       bilni bil          mili  selbri 1
 balvi bav          futu tary                     budjo buj     bu'o Budd
 re                      bindo bid          Indo  hist
 bancu bac          beyo nesian                   bukpu buk     bu'u clot
 nd                      binra         insure     h
 bandu bad          defe binxo bix     bi'o beco  bumru bum          fog
 nd                      me                       bunda bud          poun
 banfi         amphibian birje         beer       d
 bangu ban     bau  lang birka bir          arm   bunre bur     bu'e brow
 uage                    birti bit          cert  n
 banli bal     ba'i grea ain                      burcu    bru       brus
 t                       bisli bis          ice   h
 banro       ba'o   grow bitmu bim     bi'u wall  burna         embarrass
 banxa bax          bank blabi lab          whit  ed
 banzu baz          suff e                        ca'a  caz          actu
 ice                     blaci         glass      ally is
 bapli bap     bai  forc blanu    bla       blue  cabna cab          now
 e                       bliku    bli       bloc  cabra       ca'a   appa
 barda    bra       big  k                        ratus
 bargu bag          arch bloti lot   blo    lo'i  cacra         hour
 barja         bar       boat                     cadzu    dzu       walk
 barna       ba'a   mark bo bor        short      cafne caf          ofte
 bartu bar          out  scope link               n
 basna         emphasize bolci bol     boi  ball  cakla         chocolate
 basti bas          repl bongu bog     bo'u bone  calku cak          shel
 ace                     botpi bot     bo'i bott  l
 batci bat          bite le                       canci         vanish
 batke         button    boxfo bof     bo'o shee  cando cad          idle
 bavmi         barley    t                        cange cag          farm
 baxso         Malay-    boxna bon     bo'a wave  canja caj          exch
 Indonesian              bradi         enemy      ange
 bebna beb          fool bratu         hail       canko       ca'o   wind
 ish                     brazo raz          Braz  ow
 bemro bem     be'o Nort ilian                    canlu cal     ca'u spac
 h American              bredi red   bre          e
 bende bed     be'e crew ready                    canpa    cna       shov
 bengo beg          Beng bridi    bri       pred  el
 ali                     icate                    canre can          sand
 benji bej     be'i tran brife bif     bi'e bree  canti         gut
 sfer                    ze                       carce         cart
 bersa bes     be'a son

 carmi cam     cai  inte ciksi    cki       expl  ckini       ki'i   rela
 nse                     ain                      ted
 carna car          turn cilce cic          wild  ckire kir          grat
 cartu cat          char cilmo cim          mois  eful
 t                       t                        ckule kul     cu'e scho
 carvi cav          rain cilre    cli       lear  ol
 casnu    snu       disc n                        ckunu       ku'u   coni
 uss                     cilta cil          thre  fer
 catke       ca'e   shov ad                       cladu       lau    loud
 e                       cimde         dimension  clani    cla       long
 catlu    cta       look cimni         infinite   claxu       cau    with
 catni       ca'i   auth cinba         kiss       out
 ority                   cindu         oak        clika         mossy
 catra         kill      cinfo         lion       clira lir          earl
 caxno cax          shal cinje cij          wrin  y
 low                     kle                      clite lit          poli
 ce cec        in a set  cinki         insect     te
 with                    cinla         thin       cliva liv     li'a leav
 ce'i  cez          perc cinmo    cni       emot  e
 ent                     ion                      clupa cup          loop
 ce'o        ce'o   in a cinri       ci'i   inte  cmaci         mathemati
 sequence with           resting                  cs
 cecla cel     ce'a laun cinse cin          sexu  cmalu    cma       smal
 cher                    al                       l
 cecmu cem     ce'u comm cinta         paint      cmana       ma'a   moun
 unity                   cinza         tongs      tain
 cedra         era       cipni    cpi       bird  cmavo       ma'o   stru
 cenba    cne       vary cipra cip          test  cture word
 censa ces          holy cirko    cri       lose  cmene    cme  me'e name
 centi cen          .01  cirla         cheese     cmila       mi'a   laug
 cerda ced          heir ciska       ci'a   writ  h
 cerni cer          morn e                        cmima mim   cmi         
 ing                     cisma         smile      member
 certu    cre       expe ciste       ci'e   syst  cmoni    cmo  co'i moan
 rt                      em                       cnano       na'o   norm
 cevni cev     cei  god  citka    cti       eat   cnebo neb     ne'o neck
 cfari    cfa       init citno cit     ci'o youn  cnemu nem     ne'u rewa
 iate                    g                        rd
 cfika fik     fi'a fict citri cir          hist  cnici nic          orde
 ion                     ory                      rly
 cfila    cfi       flaw citsi         season     cnino nin     ni'o new
 cfine         wedge     civla civ          lous  cnisa nis          lead
 cfipu       fi'u   conf e                        cnita nit     ni'a bene
 using                   cizra ciz          stra  ath
 ci cib        3         nge                      co col        tanru
 ciblu    blu       bloo ckabu         rubber     inversion
 d                       ckafi kaf          coff  co'a        co'a   init
 cicna         cyan      ee                       iative
 cidja    dja       food ckaji       kai    qual  co'e  com     co'e unsp
 cidni cid          knee ity                      ecif bridi
 cidro    dro       hydr ckana    cka       bed   co'u        co'u   cess
 ogen                    ckape cap          peri  ative
 cifnu cif          infa l                        cokcu    cko       soak
 nt                      ckasu cas          ridi  up
 cigla cig          glan cule                     condi con   cno    coi  
 d                       ckeji kej   cke          deep
 cikna cik          awak ashamed                  cortu cor   cro         
 e                       ckiku kik          key   pain
 cikre         repair    ckilu       ci'u   scal  cpacu    cpa       get
                         e                        cpana         upon

 cpare par          clim dacru dac          draw  detri det          date
 b                       er                       dicra dir          inte
 cpedu    cpe       requ dacti       dai    obje  rrupt
 est                     ct                       dikca dic          elec
 cpina         pungent   dadjo daj          Taoi  tric
 cradi         radio     st                       diklo    klo       loca
 crane    cra       fron dakfu dak          knif  l
 t                       e                        dikni dik          regu
 creka cek          shir dakli         sack       lar
 t                       damba dab     da'a figh  dilcu         quotient
 crepu rep          harv t                        dilnu dil          clou
 est                     damri         drum       d
 cribe rib          bear dandu dad          hang  dimna dim          fate
 crida rid          fair danfu daf          answ  dinju dij     di'u buil
 y                       er                       ding
 crino       ri'o   gree danlu dal     da'u anim  dinko       di'o   nail
 n                       al                       dirba dib          dear
 cripu rip          brid danmo dam          smok  dirce       di'e   radi
 ge                      e                        ate
 crisa cis          summ danre       da'e   pres  dirgo dig          drop
 er                      sure                     dizlo diz   dzi         
 critu         autumn    dansu         dance      low
 ctaru         tide      danti dan          proj  djacu jac     jau  wate
 ctebi teb          lip  ectile                   r
 cteki tek     ce'i tax  daplu    plu       isla  djedi    dje  dei  full
 ctile         petroleum nd                       day
 ctino       ti'o   shad dapma dap          curs  djica    dji       desi
 ow                      e                        re
 ctuca    ctu       teac dargu dag          road  djine jin          ring
 h                       darlu       dau    argu  djuno jun     ju'o know
 cukla cuk          roun e                        do don      doi    you
 d                       darno dar     da'o far   donri dor     do'i dayt
 cukta    cku       book darsi         audacity   ime
 culno    clu       full darxi dax     da'i hit   dotco dot     do'o Germ
 cumki cum     cu'i poss daski         pocket     an
 ible                    dasni das          wear  draci         drama
 cumla cul          humb daspo    spo       dest  drani    dra       corr
 le                      roy                      ect
 cunmi         millet    dasri    sri       ribb  drata dat          othe
 cunso cun     cu'o rand on                       r
 om                      datka         duck       drudi rud   dru         
 cuntu       cu'u   affa datni         data       roof
 ir                      decti dec          .1    du dub      du'o   same
 cupra    pra       prod degji deg          fing  identity as
 uce                     er                       du'u  dum          brid
 curmi    cru       let  dejni dej          owe   i abstract
 curnu cur          worm dekpu         gallon     dugri dug          loga
 curve cuv          pure dekto dek          10    rithm
 cusku cus   sku         delno del     de'o cand  dukse dus     du'e exce
 express                 ela                      ss
 cutci cuc          shoe dembi deb          bean  dukti dut          oppo
 cutne cut          ches denci den     de'i toot  site
 t                       h                        dunda dud     du'a give
 cuxna cux     cu'a choo denmi dem          dens  dunja duj          free
 se                      e                        ze
 da dav   dza       some denpa dep     de'a wait  dunku duk     du'u angu
 thing 1                 dertu der     de'u dirt  ish
 da'a  daz          all  derxi    dre       heap  dunli dun     du'i equa
 except                  desku des          shak  l

 dunra dur          wint finti fin     fi'i inve  gapru gap          abov
 er                      nt                       e
 dzena    dze       elde flalu    fla       law   garna gar          rail
 r                       flani         flute      gasnu       gau    do
 dzipo zip     zi'o Anta flecu    fle       flow  gasta gat          stee
 rctican                 fliba    fli       fail  l
 facki fak     fa'i disc flira fir          face  genja gej          root
 over                    fo'a        fo'a   it-6  gento get     ge'o Arge
 fadni fad          ordi fo'e        fo'e   it-7  ntinian
 nary                    fo'i        fo'i   it-8  genxu gex          hook
 fagri fag          fire foldi    flo  foi  fiel  gerku ger     ge'u dog
 falnu fan          sail d                        gerna gen     ge'a gram
 famti         aunt or   fonmo fom     fo'o foam  mar
 uncle                   fonxa fon          tele  gidva gid     gi'a guid
 fancu         function  phone                    e
 fange         alien     forca    fro       fork  gigdo gig     gi'o 1E9
 fanmo fam     fa'o end  fraso fas          Fren  ginka gik          camp
 fanri         factory   ch                       girzu gir   gri         
 fanta         prevent   frati    fra       reac  group
 fanva         translate t                        gismu gim     gi'u root
 fanza faz          anno fraxu fax          forg  word
 y                       ive                      glare    gla       hot
 fapro fap   pro         frica fic          diff  gleki gek     gei  happ
 oppose                  er                       y
 farlu fal     fa'u fall friko       fi'o   Afri  gletu let   gle         
 farna far     fa'a dire can                      copulate
 ction                   frili fil          easy  glico gic   gli         
 farvi fav          deve frinu         fraction   English
 lop                     friti fit          offe  gluta    glu       glov
 fasnu       fau    even r                        e
 t                       frumu    fru       frow  gocti goc          1E-
 fatci fac          fact n                        24
 fatne fat     fa'e reve fukpi fuk     fu'i copy  gotro got          1E24
 rse                     fulta ful   flu          gradu       rau    unit
 fatri       fai    dist float                    grake    gra       gram
 ribute                  funca fun     fu'a luck  grana       ga'a   rod
 febvi feb          boil fusra fur          rott  grasu ras          grea
 femti fem          1E-  en                       se
 15                      fuzme fuz     fu'e resp  greku rek          fram
 fendi fed          divi onsible                  e
 de                      gacri       gai    cove  grusi rus          gray
 fengu feg     fe'u angr r                        grute rut          frui
 y                       gadri gad          arti  t
 fenki fek          craz cle                      gubni gub          publ
 y                       galfi gaf     ga'i modi  ic
 fenra fer     fe'a crac fy                       gugde gug     gu'e coun
 k                       galtu gal     ga'u high  try
 fenso fen     fe'o sew  galxe         throat     gundi gud          indu
 fepni fep     fei  cent ganlo       ga'o   clos  stry
 fepri         lung      ed                       gunka gun     gu'a work
 ferti    fre       fert ganra gan          broa  gunma gum          join
 ile                     d                        tly
 festi fes          wast ganse gas     ga'e sens  gunro gur     gu'o roll
 e                       e                        gunse         goose
 fetsi fet     fe'i fema ganti         testicle   gunta gut          atta
 le                      ganxo gax          anus  ck
 figre fig          fig  ganzu gaz          orga  gurni    gru       grai
 filso fis          Pale nize                     n
 stinian                 gapci gac          gas   guska guk          scra
 finpe fip     fi'e fish                          pe

 gusni gus     gu'i illu jecta jec     je'a poli  jinsa jis          clea
 mine                    ty                       n
 gusta         restauran jeftu jef          week  jinto         well
 t                       jegvo jeg     je'o Jeho  jinvi jiv     ji'i opin
 gutci guc          cubi vist                     e
 t                       jei   jez          trut  jinzi jiz          inna
 gutra         womb      h abstract               te
 guzme guz   zme         jelca jel          burn  jipci         chicken
 melon                   jemna    jme       gem   jipno jip     ji'o tip
 ja jav        tanru or  jenca jen          shoc  jirna         horn
 jabre         brake     k                        jisra         juice
 jadni jad     ja'i ador jendu jed          axle  jitfa jif          fals
 n                       jenmi jem     jei  army  e
 jakne         rocket    jerna         earn       jitro    tro       cont
 jalge jag     ja'e resu jersi       je'i   chas  rol
 lt                      e                        jivbu         weave
 jalna         starch    jerxo jex          Alge  jivna    jvi       comp
 jalra         cockroach rian                     ete
 jamfu jaf   jma         jesni jes          need  jmaji jaj          gath
 foot                    le                       er
 jamna jam          war  jetce       je'e   jet   jmifa         shoal
 janbe jab          bell jetnu jet     je'u true  jmina min          add
 janco jan          shou jgalu       ja'u   claw  jmive miv     ji'e live
 lder                    jganu    jga       angl  jo jov        tanru iff
 janli jal          coll e                        jo'e  jom          unio
 ide                     jgari       jai    gras  n
 jansu jas          dipl p                        jo'u        jo'u   in
 omat                    jgena    jge       knot  common with
 janta jat          acco jgina gin          gene  joi   jol     joi  in a
 unt                     jgira    jgi       prid  mass with
 jarbu         suburb    e                        jordo jor     jo'o Jord
 jarco       ja'o   show jgita git          guit  anian
 jarki jak          narr ar                       jorne jon     jo'e join
 ow                      jibni    jbi       near  ed
 jaspu jap          pass jibri jib          job   ju juv        tanru
 port                    jicla         stir       whether
 jatna       ja'a   capt jicmu    cmu       basi  jubme jub   jbu         
 ain                     s                        table
 javni    jva       rule jijnu jij          intu  judri         address
 jbama bam          bomb it                       jufra juf     ju'a sent
 jbari    jba       berr jikca jik          soci  ence
 y                       alize                    jukni juk          spid
 jbena    jbe       born jikru         liquor     er
 jbera jer          borr jilka jil          alka  jukpa jup          cook
 ow                      li                       julne       ju'e   net
 jbini bin     bi'i betw jilra         jealous    jundi jud     ju'i atte
 een                     jimca jic          bran  ntive
 jdari jar          firm ch                       jungo jug          Chin
 jdice    jdi       deci jimpe    jmi       unde  ese
 de                      rstand                   junla jul          cloc
 jdika         decrease  jimte jit          limi  k
 jdima       di'a   pric t                        junri jur          seri
 e                       jinci         shears     ous
 jdini din     di'i mone jinga jig     ji'a win   junta         weight
 y                       jinku         vaccine    jurme jum          germ
 jduli dul   jdu         jinme jim          meta  jursa jus          seve
 jelly                   l                        re
 je jev   jve       tanr jinru jir          imme  jutsi jut          spec
 u and                   rse                      ies

 juxre jux          clum kensa kes          oute  kruji ruj          crea
 sy                      r space                  m
 jvinu vin     ji'u view kerfa    kre       hair  kruvi ruv   kru         
 ka kam        property  kerlo ker          ear   curve
 abstract                ketco ket   tco          ku'a  kuz          inte
 kabri         cup       South American           rsection
 kacma         camera    kevna kev     ke'a cavi  kubli kub          cube
 kadno         Canadian  ty                       kucli         curious
 kafke         cough     kicne kic     ki'e cush  kufra kuf          comf
 kagni kag          comp ion                      ort
 any                     kijno kij          oxyg  kukte kuk          deli
 kajde    jde       warn en                       cious
 kajna         shelf     kilto       ki'o   1000  kulnu    klu       cult
 kakne       ka'e   able kinli kil          shar  ure
 kakpa         dig       p                        kumfa kum     ku'a room
 kalci         feces     kisto kis          Paki  kumte         camel
 kalri kar          open stani                    kunra kun          mine
 kalsa kas          chao klaji laj          stre  ral
 tic                     et                       kunti kut          empt
 kalte kat          hunt klaku kak          weep  y
 kamju         column    klama    kla       come  kurfa kur          squa
 kamni         committee klani       lai    quan  re
 kampu       kau    comm tity                     kurji kuj     ku'i take
 on                      klesi    kle  lei  clas  care of
 kanba         goat      s                        kurki         bitter
 kancu kac          coun klina    kli       clea  kuspe kup     ku'e rang
 t                       r                        e
 kandi kad          dim  kliru         chlorine   kusru kus          crue
 kanji kaj          calc kliti kit          clay  l
 ulate                   klupe lup     lu'e scre  labno         wolf
 kanla kal          eye  w                        lacpu lap   cpu         
 kanro       ka'o   heal kluza luz          loos  pull
 thy                     e                        lacri lac          rely
 kansa kan          with kobli kob     ko'i cabb  ladru lad          milk
 kantu       ka'u   quan age                      lafti laf          lift
 tum                     kojna koj     ko'a corn  lakne       la'e   prob
 kanxe kax          conj er                       able
 unction                 kolme kol     ko'e coal  lakse lak          wax
 karbi kab          comp komcu kom          comb  lalxu       la'u   lake
 are                     konju kon     ko'u cone  lamji lam     la'i adja
 karce         car       korbi kor     koi  edge  cent
 karda         card      korcu    kro       bent  lanbi         protein
 kargu         costly    korka kok          cork  lanci         flag
 karli         collar    kosta kos          coat  lanka         basket
 karni         journal   kramu         acre       lanli lal          anal
 katna       ka'a   cut  krasi    kra       sour  yze
 kavbu kav          capt ce                       lanme lan          shee
 ure                     krati       ka'i   repr  p
 ke kem        start     esent                    lante         can
 grouping                krefu ref     ke'u recu  lanxe lax          bala
 ke'e  kep     ke'e end  r                        nce
 grouping                krici    kri       beli  lanzu laz          fami
 kecti kec     ke'i pity eve                      ly
 kei   kez          end  krili         crystal    larcu lar          art
 abstraction             krinu rin     ki'u reas  lasna       la'a   fast
 kelci kel     kei  play on                       en
 kelvo       ke'o   kelv krixa kix     ki'a cry   lastu         brass
 in                      out                      latmo       la'o   Lati
 kenra ken          canc kruca kuc          inte  n
 er                      rsect                    latna         lotus

 lazni         lazy      mabru mab          mamm  mentu met     me'u minu
 le'e  lem          the  al                       te
 stereotypical           macnu    cnu       manu  merko mer          Amer
 lebna leb     le'a take al                       ican
 lenjo len     le'o lens makcu       ma'u   matu  merli    mre       meas
 lenku lek          cold re                       ure
 lerci lec          late makfa maf          magi  mexco mex          Mexi
 lerfu ler     le'u lett c                        can
 eral                    maksi mak          magn  mi mib        me
 li'i  liz          expe et                       midju mij          midd
 rience abstract         malsi mas          temp  le
 libjo lib          Liby le                       mifra mif          code
 an                      mamta mam          moth  mikce mic          doct
 lidne       li'e   prec er                       or
 ede                     manci mac          wond  mikri mik          1E-6
 lifri lif   fri         er                       milti mil          .001
 experience              manfo         uniform    milxe    mli       mild
 lijda    jda       reli manku man          dark  minde mid     mi'e comm
 gion                    manri mar          refe  and
 limna lim          swim rence                    minji       mi'i   mach
 lindi lid          ligh mansa         satisfy    ine
 tning                   manti         ant        minli         mile
 linji lij     li'i line mapku map          cap   minra mir          refl
 linsi lin          chai mapni         cotton     ect
 n                       mapti mat          fit   mintu mit     mi'u same
 linto       li'o   ligh marbi    mra       shel  mipri mip          secr
 tweight                 ter                      et
 lisri lis          stor marce       ma'e   vehi  mirli         deer
 y                       cle                      misno mis     mi'o famo
 liste    ste       list marde mad          mora  us
 litce lic          lite ls                       misro         Egyptian
 r                       margu mag          merc  mitre    tre       mete
 litki lik          liqu ury                      r
 id                      marji maj     mai  mate  mixre mix   xre         
 litru       li'u   trav rial                     mixture
 el                      marna         hemp       mlana    mla       side
 livga         liver     marxa max          mash  mlatu lat          cat
 livla lil          fuel masno    sno       slow  mleca mec     me'a less
 lo'e  lom          the  masti       ma'i   mont  mledi led          mold
 typical                 h                        mluni lun          sate
 logji loj          logi matci         mat        llite
 c                       matli         linen      mo'a  mob          too
 lojbo lob   jbo         matne         butter     few
 Lojbanic                matra         motor      mo'i  mov          spac
 loldi lol     loi  floo mavji mav          oats  e motion
 r                       maxri    xri       whea  moi   mom     moi  ordi
 lorxu lor     lo'u fox  t                        nal selbri
 lubno       lu'o   Leba mebri meb          brow  mokca moc          poin
 nese                    megdo meg          1E6   t
 lujvo luv   jvo         mei   mem     mei  card  moklu mol     mo'u mout
 affix compound          inal selbri              h
 lumci lum     lu'i wash mekso mek     me'o MEX   molki    mlo       mill
 lunbe lub          bare melbi mel   mle          molro       mo'o   mole
 lunra lur          luna beautiful                morji moj     mo'i reme
 r                       meljo mej          Mala  mber
 lunsa lus          cond ysian                    morko mor          Moro
 ense                    menli men          mind  ccan
 mabla mal          dero mensi mes     me'i sist  morna mon     mo'a patt
 gative                  er                       ern
                                                  morsi    mro       dead

 mosra mos          fric nejni nen          ener  panka         park
 tion                    gy                       panlo       pa'o   slic
 mraji         rye       nelci nel     nei  fond  e
 mrilu    mri       mail nenri ner     ne'i in    panpi pap          peac
 mruli    mru       hamm ni nil        amount     e
 er                      abstract                 panra         parallel
 mu mum        5         nibli nib     ni'i nece  pante pat          prot
 mu'e  muf          poin ssitate                  est
 t-event abstract        nicte    cte       nigh  panzi paz          offs
 mucti mut          imma t                        pring
 terial                  nikle nik          nick  papri         page
 mudri mud          wood el                       parbi pab          rati
 mukti muk     mu'i moti nilce       ni'e   furn  o
 ve                      iture                    pastu pas          robe
 mulno mul     mu'o comp nimre         citrus     patfu paf     pa'u fath
 lete                    ninmu nim     ni'u woma  er
 munje muj     mu'e univ n                        patlu         potato
 erse                    nirna nir          nerv  patxu pax          pot
 mupli mup          exam e                        pe'a  pev          figu
 ple                     nitcu    tcu       need  rative
 murse         glimmerin nivji niv          knit  pelji    ple       pape
 g                       nixli nix   xli          r
 murta mur     mu'a curt girl                     pelxu pel          yell
 ain                     no non        0          ow
 muslo mus          Isla no'e  nor     no'e scal  pemci pem          poem
 mic                     ar midpoint not          penbi peb          pen
 mutce    tce       much nobli nol     no'i nobl  pencu pec     pe'u touc
 muvdu muv     mu'u move e                        h
 muzga muz          muse notci not     noi  mess  pendo ped     pe'o frie
 um                      age                      nd
 na nar        bridi     nu nun        event      penmi pen     pe'i meet
 negator                 abstract                 pensi pes     pei  thin
 na'e  nal          scal nu'o        nu'o   can   k
 ar contrary             but has not              perli per          pear
 nabmi nam          prob nukni nuk          mage  pesxu pex          past
 lem                     nta                      e
 nakni nak          male nupre nup     nu'e prom  petso pet          1E15
 nalci       na'i   wing ise                      pezli pez          leaf
 namcu nac     na'u numb nurma num          rura  pi piz        decimal
 er                      l                        point
 nanba nab          brea nutli nul     nu'i neut  pi'u  piv          cros
 d                       ral                      s product
 nanca       na'a   year nuzba nuz          news  picti pic          1E-
 nandu nad          diff pa pav        1          12
 icult                   pacna       pa'a   hope  pijne         pin
 nanla         boy       pagbu pag     pau  part  pikci         beg
 nanmu       nau    man  pagre    gre       pass  pikta         ticket
 nanvi nav          1E-9 through                  pilji       pi'i   mult
 narge nag          nut  pajni       pai    judg  iply
 narju naj          oran e                        pilka pil     pi'a crus
 ge                      palci pac          evil  t
 natfe naf     na'e deny palku pak          pant  pilno    pli       use
 natmi nat     nai  nati s                        pimlu pim     pi'u feat
 on                      palne         tray       her
 navni         neon      palta         plate      pinca         urine
 naxle nax   xle         pambe         pump       pindi pid          poor
 canal                   panci pan          odor  pinfu pif          pris
 nazbi naz   zbi         pandi pad          punc  oner
 nose                    tuate                    pinji         penis
                         panje         sponge

 pinka pik          comm pruce ruc     ru'e proc  remna rem     re'a huma
 ent                     ess                      n
 pinsi pis          penc pruni pun          elas  renro rer     re'o thro
 il                      tic                      w
 pinta pin          leve pruxi rux     ru'i spir  renvi rev     re'i surv
 l                       it                       ive
 pinxe pix          drin pu'i  pus          can   respa res          rept
 k                       and has                  ile
 pipno       pi'o   pian pu'u  puv          proc  ricfu rif   cfu         
 o                       ess abstract             rich
 pixra pir   xra         pulce puc     pu'e dust  rigni rig          disg
 picture                 pulji         police     usting
 plana         plump     pulni         pulley     rijno rij          silv
 platu    pla       plan punji puj     pu'i put   er
 pleji lej     le'i pay  punli pul          swel  rilti ril          rhyt
 plibu pib          pubi ling                     hm
 c                       purci pur   pru          rimni rim          rhym
 plini         planet    past                     e
 plipe pip     pi'e leap purdi pud          gard  rinci         drain
 plise         apple     en                       rinju       ri'u   rest
 plita pit          plan purmo pum     pu'o powd  rain
 e                       er                       rinka rik     ri'a caus
 plixa lix          plow racli         sane       e
 pluja luj          comp ractu         rabbit     rinsa         greet
 licated                 radno       ra'o   radi  rirci         rare
 pluka puk     pu'a plea an                       rirni rir          pare
 sant                    rafsi raf          affi  nt
 pluta lut     lu'a rout x                        rirxe       ri'e   rive
 e                       ragve rav          acro  r
 polje    plo       fold ss                       rismi ris          rice
 polno pol          Poly rakso         Iraqi      risna         heart
 nesian                  raktu       ra'u   trou  ritli       ri'i   rite
 ponjo pon     po'o Japa ble                      rivbi riv          avoi
 nese                    ralci rac          deli  d
 ponse pos     po'e poss cate                     ro rol        each
 ess                     ralju ral          prin  roi   rom     roi  quan
 porpi pop     po'i brea cipal                    tified tense
 k                       ralte       ra'e   reta  rokci rok     ro'i rock
 porsi por     poi  sequ in                       romge rog          chro
 ence                    randa rad          yiel  me
 porto pot          Port d                        ropno ron     ro'o Euro
 uguese                  rango rag          orga  pean
 prali pal          prof n                        rorci ror          proc
 it                      ranji       ra'i   cont  reate
 prami pam     pa'i love inue                     rotsu rot   tsu    ro'u 
 prane       pa'e   perf ranmi ram          myth  thick
 ect                     ransu         bronze     rozgu roz   zgu         
 preja pej     pe'a spre ranti ran          soft  rose
 ad                      ranxi rax          iron  ruble rub   ble         
 prenu    pre       pers y                        weak
 on                      rapli rap          repe  rufsu ruf          roug
 preti ret     rei  ques at                       h
 tion                    rarna rar          natu  runme rum          melt
 prije pij          wise ral                      runta         dissolve
 prina    pri       prin ratcu         rat        rupnu rup     ru'u doll
 t                       ratni rat          atom  ar
 pritu         right     re rel        2          rusko ruk     ru'o Russ
 prosa ros     ro'a pros rebla reb          tail  ian
 e                       rectu rec     re'u meat  rutni run          arti

 sabji sab          prov senta set          laye  skori    sko       cord
 ide                     r                        skoto kot     ko'o Scot
 sabnu         cabin     senva sev   sne          tish
 sacki         match     dream                    skuro       ku'o   groo
 saclu         decimal   sepli sep     sei  apar  ve
 sadjo    djo       Saud t                        slabu       sau    old
 i                       serti ser          stai  slaka         syllable
 sakci sak          suck rs                       slami         acid
 sakli sal          slid setca       se'a   inse  slanu         cylinder
 e                       rt                       slari sar          sour
 sakta sat          suga sevzi sez     se'i self  slasi las          plas
 r                       sfani         fly        tic
 salci    sla       cele sfasa    sfa       puni  sligu lig          soli
 brate                   sh                       d
 salpo       sa'o   slop sfofa    sfo       sofa  slilu    sli       osci
 e                       sfubu sub     su'u dive  llate
 salta         salad     si'o  siz          conc  sliri         sulfur
 samcu         cassava   ept abstract             slovo lov     lo'o Slav
 sampu sap          simp siclu sil          whis  ic
 le                      tle                      sluji    slu       musc
 sance    sna       soun sicni       si'i   coin  le
 d                       sidbo sib     si'o idea  sluni         onion
 sanga sag     sa'a sing sidju    dju       help  smacu         mouse
 sanji saj          cons sigja sig          ciga  smadi         guess
 cious                   r                        smaji    sma       quie
 sanli       sa'i   stan silka sik          silk  t
 d                       silna         salt       smani         monkey
 sanmi       sai    meal simlu    mlu       seem  smoka    smo       sock
 sanso         sauce     simsa    smi       simi  smuci muc          spoo
 santa         umbrella  lar                      n
 sarcu       sa'u   nece simxu sim     si'u mutu  smuni mun   smu         
 ssary                   al                       meaning
 sarji    sra       supp since         snake      snada sad          succ
 ort                     sinma       si'a   este  eed
 sarlu         spiral    em                       snanu nan          sout
 sarxe sax          harm sinso         sine       h
 onious                  sinxa    sni       sign  snidu nid          seco
 saske    ske       scie sipna sip          slee  nd
 nce                     p                        snime       si'e   snow
 satci         exact     sirji sir          stra  snipa nip          stic
 satre       sa'e   stro ight                     ky
 ke                      sirxo six          Syri  snuji nuj          sand
 savru sav   vru         an                       wich
 noise                   sisku sis          seek  snura nur     nu'a secu
 sazri saz          oper sisti    sti       ceas  re
 ate                     e                        snuti nut          acci
 se sel        2nd       sitna sit          cite  dental
 conversion              sivni siv          priv  so soz        9
 sefta    sfe       surf ate                      so'a  soj          almo
 ace                     skaci         skirt      st all
 selci    sle       cell skami sam          comp  so'e  sop          most
 selfu sef     se'u serv uter                     so'i  sor     so'i many
 e                       skapi kap          pelt  so'o  sos          seve
 semto    sme       Semi skari    ska       colo  ral
 tic                     r                        so'u  sot          few
 senci sec          snee skicu    ski       desc  sobde sob     so'e soya
 ze                      ribe                     sodna         sodium
 senpi sen          doub skiji sij          ski   sodva sod          soda
 t                       skina kin          cine  softo sof          Sovi
                         ma                       et

 solji    slo       gold steci tec     te'i spec  tance tac          tong
 solri sol          sola ific                     ue
 r                       stedu sed          head  tanjo         tangent
 sombo som     so'o sow  stela tel          lock  tanko         tobacco
 sonci son     soi  sold stero       te'o   ster  tanru       tau    phra
 ier                     adian                    se compound
 sorcu soc   sro         stici sic          west  tansi tas          pan
 store                   stidi sid     ti'i sugg  tanxe tax     ta'e box
 sorgu sog          sorg est                      tapla         tile
 hum                     stika tik          adju  tarbi         embryo
 sovda sov     so'a egg  st                       tarci tar          star
 spaji paj          surp stizu tiz          chai  tarla         tar
 rise                    r                        tarmi tam     tai  shap
 spali         polish    stodi    sto       cons  e
 spano san          Span tant                     tarti    tra       beha
 ish                     stuna sun          east  ve
 spati    spa       plan stura tur     su'a stru  taske         thirst
 t                       cture                    tatpi       ta'i   tire
 speni    spe       marr stuzi tuz   stu          d
 ied                     site                     tatru tat          brea
 spisa    spi       piec su'e  sup     su'e at    st
 e                       most                     tavla tav     ta'a talk
 spita         hospital  su'o  suz     su'o at    taxfu taf     ta'u garm
 spofu pof     po'u brok least                    ent
 en                      su'u  suv          unsp  tcaci cac          cust
 spoja poj     po'a expl ecif abstract            om
 ode                     sucta suc          abst  tcadu    tca       city
 spuda    spu       repl ract                     tcana         station
 y                       sudga sud          dry   tcati         tea
 sputu put     pu'u spit sufti    sfu       hoof  tcena ten          stre
 sraji raj          vert suksa suk          sudd  tch
 ical                    en                       tcica tic          dece
 sraku rak          scra sumji suj          sum   ive
 tch                     sumne         smell      tcidu tid          read
 sralo         Australia sumti sum     su'i argu  tcika         time of
 n                       ment                     day
 srana       ra'a   pert sunga sug          garl  tcila til          deta
 ain                     ic                       il
 srasu sas          gras sunla sul          wool  tcima tim     ti'a weat
 s                       surla sur          rela  her
 srera    sre       err  x                        tcini         situation
 srito         Sanskrit  sutra sut          fast  tcita         label
 sruma       ru'a   assu ta taz        that       te ter        3rd
 me                      there                    conversion
 sruri rur   sru         tabno tab          carb  temci tem     tei  time
 surround                on                       tenfa tef          expo
 stace sac          hone tabra         trumpet    nential
 st                      tadji         method     tengu teg     te'u text
 stagi         vegetable tadni tad          stud  ure
 staku tak          cera y                        terdi ted          eart
 mic                     tagji tag          snug  h
 stali    sta       rema talsa tal          chal  terpa tep     te'a fear
 in                      lenge                    terto tet          1E12
 stani         stalk     tamca         tomato     ti tif        this here
 stapa tap          step tamji taj          thum  tigni tig          perf
 stasu         soup      b                        orm
 stati         talent    tamne         cousin     tikpa tip          kick
 steba seb          frus tanbo       ta'o   boar  tilju tij          heav
 tration                 d                        y
                                                  tinbe tib          obey

 tinci         tin       tutra tut          terr  vitno       vi'o   perm
 tinsa         stiff     itory                    anent
 tirna tin          hear va vaz        there at   vlagi lag          vulv
 tirse tir          iron vacri var          air   a
 tirxu         tiger     vajni vaj     vai  impo  vlile vil          viol
 tisna tis          fill rtant                    ent
 titla tit          swee valsi val   vla          vlina         alternati
 t                       word                     on
 tivni tiv          tele vamji vam     va'i valu  vlipa    vli       powe
 vision                  e                        rful
 tixnu tix     ti'u daug vamtu vat          vomi  vo von        4
 hter                    t                        vofli vol     voi  flig
 to'e  tol     to'e pola vanbi vab          envi  ht
 r opposite              ronment                  voksa vok     vo'a voic
 toknu tok          oven vanci vac          even  e
 toldi tod          butt ing                      vorme vor   vro         
 erfly                   vanju van          wine  door
 tonga tog     to'a tone vasru vas     vau  cont  vraga    vra       leve
 tordu tor     to'u shor ain                      r
 t                       vasxu vax     va'u brea  vreji rej     vei  reco
 torni ton     to'i twis the                      rd
 t                       ve vel        4th        vreta    vre       recl
 traji       rai    supe conversion               ining
 rlative                 ve'e        ve'e   whol  vrici         miscellan
 trano         nitrogen  e space interval         eous
 trati         taut      vecnu ven     ve'u sell  vrude vud     vu'e virt
 trene ren     re'e trai venfu vef          reve  ue
 n                       nge                      vrusi vus     vu'i tast
 tricu ric          tree vensa ves          spri  e
 trina    tri       attr ng                       vu vuz        yonder at
 act                     verba ver     ve'a chil  vukro vur     vu'o Ukra
 trixe rix     ti'e behi d                        inian
 nd                      vi viz        here at    xa xav        6
 troci toc     toi  try  vibna vib          vagi  xabju       xa'u   dwel
 tsali    tsa       stro na                       l
 ng                      vidni         video      xadba xab          half
 tsani tan          sky  vidru vir          viru  xadni xad          body
 tsapi         seasoning s                        xagji         hunger
 tsiju    tsi       seed vifne vif          fres  xagri         reed
 tsina sin          stag h                        xajmi xam          funn
 e                       vikmi vim     vi'i excr  y
 tu tuf        that      ete                      xaksu xak          use
 yonder                  viknu vik          visc  up
 tubnu       tu'u   tube ous                      xalbo         levity
 tugni tug     tu'i agre vimcu vic     vi'u remo  xalka xal          alco
 e                       ve                       hol
 tujli tuj          tuli vindu vid          pois  xalni         panic
 p                       on                       xamgu xag     xau  good
 tumla tum     tu'a land vinji vij          airp  xampo xap     xa'o ampe
 tunba tub          sibl lane                     re
 ing                     vipsi vip          depu  xamsi xas          sea
 tunka tuk          copp ty                       xance xan     xa'e hand
 er                      virnu    vri       brav  xanka         nervous
 tunlo tul     tu'o swal e                        xanri xar          imag
 low                     viska vis     vi'a see   inary
 tunta tun          poke vitci vit          irre  xanto         elephant
 tuple tup     tu'e leg  gular                    xarci xac     xa'i weap
 turni    tru       gove vitke       vi'e   gues  on
 rn                      t                        xarju xaj          pig
 tutci    tci       tool                          xarnu         stubborn

 xasli         donkey    za'i  zaz          stat  zumri    zmu       maiz
 xasne         sweat     e abstract               e
 xatra       xa'a   lett za'o        za'o   supe  zungi zug          guil
 er                      rfective                 t
 xatsi xat          1E-  zabna zan     za'a favo  zunle zul          left
 18                      rable                    zunti zun     zu'i inte
 xazdo xaz   zdo         zajba zaj          gymn  rfere
 Asiatic                 ast                      zutse zut   tse         
 xe xel        5th       zalvi zal          grin  sit
 conversion              d                        zvati    zva       at
 xebni xen     xei  hate zanru zar     zau  appr
 xebro xeb   bro         ove
 Hebrew                  zarci zac     zai  mark
 xecto xet   cto         et
 100                     zargu zag     za'u butt
 xedja xej     xe'a jaw  ock
 xekri xek     xe'i blac zasni zas          temp
 k                       orary
 xelso xes          Gree zasti zat     za'i exis
 k                       t
 xendo xed     xe'o kind zbabu bab          soap
 xenru xer     xe'u regr zbani         bay
 et                      zbasu    zba       make
 xexso xex          1E18 zbepi    zbe       pede
 xindo xin          Hind stal
 i                       zdani    zda       nest
 xinmo xim          ink  zdile    zdi       amus
 xirma xir     xi'a hors ing
 e                       ze zel        7
 xislu xil     xi'u whee ze'e        ze'e   whol
 l                       e time interval
 xispo xip          Hisp ze'o  zev     ze'o outw
 anic                    ard
 xlali    xla       bad  zekri zer     zei  crim
 xlura    xlu       infl e
 uence                   zenba zen     ze'a incr
 xotli xol     xoi  hote ease
 l                       zepti zep          1E-
 xrabo rab          Arab 21
 ic                      zetro zet          1E21
 xrani       xai    inju zgana    zga       obse
 re                      rve
 xriso xis     xi'o Chri zgike    zgi  gi'e musi
 stian                   c
 xruba xub          buck zifre zif     zi'e free
 wheat                   zinki zin     zi'i zinc
 xruki xuk          turk zirpu zir     zi'u purp
 ey                      le
 xrula rul          flow zivle ziv   vle         
 er                      invest
 xruti    xru       retu zmadu    zma  mau  more
 rn                      zmiku    zmi       auto
 xukmi xum     xu'i chem matic
 ical                    zo'a  zon     zo'a tang
 xunre xun     xu'e red  ential to
 xurdo xur     xu'o Urdi zo'i  zor     zo'i inwa
 xusra xus     xu'a asse rd
 rt                      zu'o  zum          acti
 xutla xul          smoo vity abstract
 th                      zukte zuk     zu'e act
 bab    zbabu  soa big    bilga  obl  cag    cange  far  cim    cilmo  moi
         p                 iged               m                  st
 bac    bancu  bey bij    briju  off  caj    canja  exc  cin    cinse  sex
         ond               ice                hange              ual
 bad    bandu  def bik    bikla  whi  cak    calku  she  cip    cipra  tes
         end               p                  ll                 t
 baf    bakfu  bun bil    bilni  mil  cal    canlu  spa  cir    citri  his
         dle               itary              ce                 tory
 bag    bargu  arc bim    bitmu  wal  cam    carmi  int  cis    crisa  sum
         h                 l                  ense               mer
 baj    bajra  run bin    jbini  bet  can    canre  san  cit    citno  you
 bak    bakni  bov         ween               d                  ng
         ine       bir    birka  arm  cap    ckape  per  civ    civla  lou
 bal    banli  gre bis    bisli  ice          il                 se
         at        bit    birti  cer  car    carna  tur  ciz    cizra  str
 bam    jbama  bom         tain               n                  ange
         b         biv    bi     8    cas    ckasu  rid  col    co     tan
 ban    bangu  lan bix    binxo  bec          icule              ru
         guage             ome        cat    cartu  cha          inversion
 bap    bapli  for biz    bi'i   uno          rt         com    co'e   uns
         ce                rdered     cav    carvi  rai          pecif
 bar    bartu  out         interval           n                  bridi
 bas    basti  rep bof    boxfo  she  cax    caxno  sha  con    condi  dee
         lace              et                 llow               p
 bat    batci  bit bog    bongu  bon  caz    ca'a   act  cor    cortu  pai
         e                 e                  ually is           n
 bav    balvi  fut bol    bolci  bal  cec    ce     in   cuc    cutci  sho
         ure               l                  a set              e
 bax    banxa  ban bon    boxna  wav          with       cuk    cukla  rou
         k                 e          ced    cerda  hei          nd
 baz    banzu  suf bor    bo     sho          r          cul    cumla  hum
         fice              rt scope   cek    creka  shi          ble
 beb    bebna  foo         link               rt         cum    cumki  pos
         lish      bot    botpi  bot  cel    cecla  lau          sible
 bed    bende  cre         tle                ncher      cun    cunso  ran
         w         bud    bunda  pou  cem    cecmu  com          dom
 bef    betfu  abd         nd                 munity     cup    clupa  loo
         omen      buj    budjo  Bud  cen    centi  .01          p
 beg    bengo  Ben         dhist      cer    cerni  mor  cur    curnu  wor
         gali      buk    bukpu  clo          ning               m
 bej    benji  tra         th         ces    censa  hol  cus    cusku  exp
         nsfer     bul    bu'a   som          y                  ress
 bem    bemro  Nor         e selbri   cev    cevni  god  cut    cutne  che
         th                1          cez    ce'i   per          st
         American  bum    bumru  fog          cent       cuv    curve  pur
 ben    besna  bra bun    bruna  bro  cib    ci     3            e
         in                ther       cic    cilce  wil  cux    cuxna  cho
 ber    berti  nor bur    bunre  bro          d                  ose
         th                wn         cid    cidni  kne  dab    damba  fig
 bes    bersa  son bus    bu     wor          e                  ht
 bet    betri  tra         d to       cif    cifnu  inf  dac    dacru  dra
         gedy              lerfu              ant                wer
 bev    bevri  car cab    cabna  now  cig    cigla  gla  dad    dandu  han
         ry        cac    tcaci  cus          nd                 g
 bic    bifce  bee         tom        cij    cinje  wri  daf    danfu  ans
 bid    bindo  Ind cad    cando  idl          nkle               wer
         onesian           e          cik    cikna  awa  dag    dargu  roa
 bif    brife  bre caf    cafne  oft          ke                 d
         eze               en         cil    cilta  thr  daj    dadjo  Tao
                                              ead                ist

 dak    dakfu  kni don    do     you  fed    fendi  div  gar    garna  rai
         fe        dor    donri  day          ide                l
 dal    danlu  ani         time       feg    fengu  ang  gas    ganse  sen
         mal       dot    dotco  Ger          ry                 se
 dam    danmo  smo         man        fek    fenki  cra  gat    gasta  ste
         ke        dub    du     sam          zy                 el
 dan    danti  pro         e          fem    femti  1E-  gax    ganxo  anu
         jectile           identity           15                 s
 dap    dapma  cur         as         fen    fenso  sew  gaz    ganzu  org
         se        dud    dunda  giv  fep    fepni  cen          anize
 dar    darno  far         e                  t          gej    genja  roo
 das    dasni  wea dug    dugri  log  fer    fenra  cra          t
         r                 arithm             ck         gek    gleki  hap
 dat    drata  oth duj    dunja  fre  fes    festi  was          py
         er                eze                te         gen    gerna  gra
 dav    da     som duk    dunku  ang  fet    fetsi  fem          mmar
         ething 1          uish               ale        ger    gerku  dog
 dax    darxi  hit dul    jduli  jel  fic    frica  dif  get    gento  Arg
 daz    da'a   all         ly                 fer                entinian
         except    dum    du'u   bri  fig    figre  fig  gex    genxu  hoo
 deb    dembi  bea         di         fik    cfika  fic          k
         n                 abstract           tion       gic    glico  Eng
 dec    decti  .1  dun    dunli  equ  fil    frili  eas          lish
 deg    degji  fin         al                 y          gid    gidva  gui
         ger       dur    dunra  win  fin    finti  inv          de
 dej    dejni  owe         ter                ent        gig    gigdo  1E9
 dek    dekto  10  dus    dukse  exc  fip    finpe  fis  gik    ginka  cam
 del    delno  can         ess                h                  p
         dela      dut    dukti  opp  fir    flira  fac  gim    gismu  roo
 dem    denmi  den         osite              e                  t word
         se        fac    fatci  fac  fis    filso  Pal  gin    jgina  gen
 den    denci  too         t                  estinian           e
         th        fad    fadni  ord  fit    friti  off  gir    girzu  gro
 dep    denpa  wai         inary              er                 up
         t         fag    fagri  fir  fom    fonmo  foa  git    jgita  gui
 der    dertu  dir         e                  m                  tar
         t         fak    facki  dis  fon    fonxa  tel  goc    gocti  1E-
 des    desku  sha         cover              ephone             24
         ke        fal    farlu  fal  fuk    fukpi  cop  got    gotro  1E2
 det    detri  dat         l                  y                  4
         e         fam    fanmo  end  ful    fulta  flo  gub    gubni  pub
 dib    dirba  dea fan    falnu  sai          at                 lic
         r                 l          fun    funca  luc  guc    gutci  cub
 dic    dikca  ele fap    fapro  opp          k                  it
         ctric             ose        fur    fusra  rot  gud    gundi  ind
 dig    dirgo  dro far    farna  dir          ten                ustry
         p                 ection     fuz    fuzme  res  gug    gugde  cou
 dij    dinju  bui fas    fraso  Fre          ponsible           ntry
         lding             nch        gac    gapci  gas  guk    guska  scr
 dik    dikni  reg fat    fatne  rev  gad    gadri  art          ape
         ular              erse               icle       gum    gunma  joi
 dil    dilnu  clo fav    farvi  dev  gaf    galfi  mod          ntly
         ud                elop               ify        gun    gunka  wor
 dim    dimna  fat fax    fraxu  for  gal    galtu  hig          k
         e                 give               h          gur    gunro  rol
 din    jdini  mon faz    fanza  ann  gan    ganra  bro          l
         ey                oy                 ad         gus    gusni  ill
 dir    dicra  int feb    febvi  boi  gap    gapru  abo          umine
         errupt            l                  ve         gut    gunta  att
 diz    dizlo  low                                               ack

 guz    guzme  mel jic    jimca  bra  jut    jutsi  spe  kez    kei    end
         on                nch                cies               abstracti
 jab    janbe  bel jif    jitfa  fal  juv    ju     tan          on
         l                 se                 ru         kic    kicne  cus
 jac    djacu  wat jig    jinga  win          whether            hion
         er        jij    jijnu  int  jux    juxre  clu  kij    kijno  oxy
 jad    jadni  ado         uit                msy                gen
         rn        jik    jikca  soc  kab    karbi  com  kik    ckiku  key
 jaf    jamfu  foo         ialize             pare       kil    kinli  sha
         t         jil    jilka  alk  kac    kancu  cou          rp
 jag    jalge  res         ali                nt         kin    skina  cin
         ult       jim    jinme  met  kad    kandi  dim          ema
 jaj    jmaji  gat         al         kaf    ckafi  cof  kir    ckire  gra
         her       jin    djine  rin          fee                teful
 jak    jarki  nar         g          kag    kagni  com  kis    kisto  Pak
         row       jip    jipno  tip          pany               istani
 jal    janli  col jir    jinru  imm  kaj    kanji  cal  kit    kliti  cla
         lide              erse               culate             y
 jam    jamna  war jis    jinsa  cle  kak    klaku  wee  kix    krixa  cry
 jan    janco  sho         an                 p                  out
         ulder     jit    jimte  lim  kal    kanla  eye  kob    kobli  cab
 jap    jaspu  pas         it         kam    ka     pro          bage
         sport     jiv    jinvi  opi          perty      koj    kojna  cor
 jar    jdari  fir         ne                 abstract           ner
         m         jiz    jinzi  inn  kan    kansa  wit  kok    korka  cor
 jas    jansu  dip         ate                h                  k
         lomat     jol    joi    in   kap    skapi  pel  kol    kolme  coa
 jat    janta  acc         a mass             t                  l
         ount              with       kar    kalri  ope  kom    komcu  com
 jav    ja     tan jom    jo'e   uni          n                  b
         ru or             on         kas    kalsa  cha  kon    konju  con
 jec    jecta  pol jon    jorne  joi          otic               e
         ity               ned        kat    kalte  hun  kor    korbi  edg
 jed    jendu  axl jor    jordo  Jor          t                  e
         e                 danian     kav    kavbu  cap  kos    kosta  coa
 jef    jeftu  wee jov    jo     tan          ture               t
         k                 ru iff     kax    kanxe  con  kot    skoto  Sco
 jeg    jegvo  Jeh jub    jubme  tab          junction           ttish
         ovist             le         kec    kecti  pit  kub    kubli  cub
 jel    jelca  bur jud    jundi  att          y                  e
         n                 entive     kej    ckeji  ash  kuc    kruca  int
 jem    jenmi  arm juf    jufra  sen          amed               ersect
         y                 tence      kel    kelci  pla  kuf    kufra  com
 jen    jenca  sho jug    jungo  Chi          y                  fort
         ck                nese       kem    ke     sta  kuj    kurji  tak
 jer    jbera  bor juk    jukni  spi          rt                 e care of
         row               der                grouping   kuk    kukte  del
 jes    jesni  nee jul    junla  clo  ken    kenra  can          icious
         dle               ck                 cer        kul    ckule  sch
 jet    jetnu  tru jum    jurme  ger  kep    ke'e   end          ool
         e                 m                  grouping   kum    kumfa  roo
 jev    je     tan jun    djuno  kno  ker    kerlo  ear          m
         ru and            w          kes    kensa  out  kun    kunra  min
 jex    jerxo  Alg jup    jukpa  coo          er space           eral
         erian             k          ket    ketco  Sou  kup    kuspe  ran
 jez    jei    tru jur    junri  ser          th                 ge
         th                ious               American   kur    kurfa  squ
         abstract  jus    jursa  sev  kev    kevna  cav          are
 jib    jibri  job         ere                ity        kus    kusru  cru

 kut    kunti  emp lig    sligu  sol  mab    mabru  mam  mij    midju  mid
         ty                id                 mal                dle
 kuz    ku'a   int lij    linji  lin  mac    manci  won  mik    mikri  1E-
         ersection         e                  der                6
 lab    blabi  whi lik    litki  liq  mad    marde  mor  mil    milti  .00
         te                uid                als                1
 lac    lacri  rel lil    livla  fue  maf    makfa  mag  mim    cmima  mem
         y                 l                  ic                 ber
 lad    ladru  mil lim    limna  swi  mag    margu  mer  min    jmina  add
         k                 m                  cury       mip    mipri  sec
 laf    lafti  lif lin    linsi  cha  maj    marji  mat          ret
         t                 in                 erial      mir    minra  ref
 lag    vlagi  vul lir    clira  ear  mak    maksi  mag          lect
         va                ly                 net        mis    misno  fam
 laj    klaji  str lis    lisri  sto  mal    mabla  der          ous
         eet               ry                 ogative    mit    mintu  sam
 lak    lakse  wax lit    clite  pol  mam    mamta  mot          e
 lal    lanli  ana         ite                her        miv    jmive  liv
         lyze      liv    cliva  lea  man    manku  dar          e
 lam    lamji  adj         ve                 k          mix    mixre  mix
         acent     lix    plixa  plo  map    mapku  cap          ture
 lan    lanme  she         w          mar    manri  ref  mob    mo'a   too
         ep        liz    li'i   exp          erence             few
 lap    lacpu  pul         erience    mas    malsi  tem  moc    mokca  poi
         l                 abstract           ple                nt
 lar    larcu  art lob    lojbo  Loj  mat    mapti  fit  moj    morji  rem
 las    slasi  pla         banic      mav    mavji  oat          ember
         stic      loj    logji  log          s          mol    moklu  mou
 lat    mlatu  cat         ic         max    marxa  mas          th
 lax    lanxe  bal lol    loldi  flo          h          mom    moi    ord
         ance              or         meb    mebri  bro          inal
 laz    lanzu  fam lom    lo'e   the          w                  selbri
         ily               typical    mec    mleca  les  mon    morna  pat
 leb    lebna  tak lor    lorxu  fox          s                  tern
         e         lot    bloti  boa  meg    megdo  1E6  mor    morko  Mor
 lec    lerci  lat         t          mej    meljo  Mal          occan
         e         lov    slovo  Sla          aysian     mos    mosra  fri
 led    mledi  mol         vic        mek    mekso  MEX          ction
         d         lub    lunbe  bar  mel    melbi  bea  mov    mo'i   spa
 lej    pleji  pay         e                  utiful             ce motion
 lek    lenku  col luj    pluja  com  mem    mei    car  muc    smuci  spo
         d                 plicated           dinal              on
 lem    le'e   the lum    lumci  was          selbri     mud    mudri  woo
         stereotyp         h          men    menli  min          d
         ical      lun    mluni  sat          d          muf    mu'e   poi
 len    lenjo  len         ellite     mer    merko  Ame          nt-event
         s         lup    klupe  scr          rican              abstract
 ler    lerfu  let         ew         mes    mensi  sis  muj    munje  uni
         teral     lur    lunra  lun          ter                verse
 let    gletu  cop         ar         met    mentu  min  muk    mukti  mot
         ulate     lus    lunsa  con          ute                ive
 lib    libjo  Lib         dense      mex    mexco  Mex  mul    mulno  com
         yan       lut    pluta  rou          ican               plete
 lic    litce  lit         te         mib    mi     me   mum    mu     5
         er        luv    lujvo  aff  mic    mikce  doc  mun    smuni  mea
 lid    lindi  lig         ix                 tor                ning
         htning            compound   mid    minde  com  mup    mupli  exa
 lif    lifri  exp luz    kluza  loo          mand               mple
         erience           se         mif    mifra  cod  mur    murta  cur
                                              e                  tain

 mus    muslo  Isl nim    ninmu  wom  pam    prami  lov  pip    plipe  lea
         amic              an                 e                  p
 mut    mucti  imm nin    cnino  new  pan    panci  odo  pir    pixra  pic
         aterial   nip    snipa  sti          r                  ture
 muv    muvdu  mov         cky        pap    panpi  pea  pis    pinsi  pen
         e         nir    nirna  ner          ce                 cil
 muz    muzga  mus         ve         par    cpare  cli  pit    plita  pla
         eum       nis    cnisa  lea          mb                 ne
 nab    nanba  bre         d          pas    pastu  rob  piv    pi'u   cro
         ad        nit    cnita  ben          e                  ss
 nac    namcu  num         eath       pat    pante  pro          product
         ber       niv    nivji  kni          test       pix    pinxe  dri
 nad    nandu  dif         t          pav    pa     1            nk
         ficult    nix    nixli  gir  pax    patxu  pot  piz    pi     dec
 naf    natfe  den         l          paz    panzi  off          imal
         y         nol    nobli  nob          spring             point
 nag    narge  nut         le         peb    penbi  pen  pof    spofu  bro
 naj    narju  ora non    no     0    pec    pencu  tou          ken
         nge       nor    no'e   sca          ch         poj    spoja  exp
 nak    nakni  mal         lar        ped    pendo  fri          lode
         e                 midpoint           end        pol    polno  Pol
 nal    na'e   sca         not        pej    preja  spr          ynesian
         lar       not    notci  mes          ead        pon    ponjo  Jap
         contrary          sage       pel    pelxu  yel          anese
 nam    nabmi  pro nuj    snuji  san          low        pop    porpi  bre
         blem              dwich      pem    pemci  poe          ak
 nan    snanu  sou nuk    nukni  mag          m          por    porsi  seq
         th                enta       pen    penmi  mee          uence
 nar    na     bri nul    nutli  neu          t          pos    ponse  pos
         di                tral       per    perli  pea          sess
         negator   num    nurma  rur          r          pot    porto  Por
 nat    natmi  nat         al         pes    pensi  thi          tuguese
         ion       nun    nu     eve          nk         puc    pulce  dus
 nav    nanvi  1E-         nt         pet    petso  1E1          t
         9                 abstract           5          pud    purdi  gar
 nax    naxle  can nup    nupre  pro  pev    pe'a   fig          den
         al                mise               urative    puj    punji  put
 naz    nazbi  nos nur    snura  sec  pex    pesxu  pas  puk    pluka  ple
         e                 ure                te                 asant
 neb    cnebo  nec nut    snuti  acc  pez    pezli  lea  pul    punli  swe
         k                 idental            f                  lling
 nel    nelci  fon nuz    nuzba  new  pib    plibu  pub  pum    purmo  pow
         d                 s                  ic                 der
 nem    cnemu  rew pab    parbi  rat  pic    picti  1E-  pun    pruni  ela
         ard               io                 12                 stic
 nen    nejni  ene pac    palci  evi  pid    pindi  poo  pur    purci  pas
         rgy               l                  r                  t
 ner    nenri  in  pad    pandi  pun  pif    pinfu  pri  pus    pu'i   can
 nib    nibli  nec         ctuate             soner              and has
         essitate  paf    patfu  fat  pij    prije  wis  put    sputu  spi
 nic    cnici  ord         her                e                  t
         erly      pag    pagbu  par  pik    pinka  com  puv    pu'u   pro
 nid    snidu  sec         t                  ment               cess
         ond       paj    spaji  sur  pil    pilka  cru          abstract
 nik    nikle  nic         prise              st         rab    xrabo  Ara
         kel       pak    palku  pan  pim    pimlu  fea          bic
 nil    ni     amo         ts                 ther       rac    ralci  del
         unt       pal    prali  pro  pin    pinta  lev          icate
         abstract          fit                el         rad    randa  yie

 raf    rafsi  aff rid    crida  fai  ruk    rusko  Rus  sel    se     2nd
         ix                ry                 sian               conversio
 rag    rango  org rif    ricfu  ric  rul    xrula  flo          n
         an                h                  wer        sen    senpi  dou
 raj    sraji  ver rig    rigni  dis  rum    runme  mel          bt
         tical             gusting            t          sep    sepli  apa
 rak    sraku  scr rij    rijno  sil  run    rutni  art          rt
         atch              ver                ifact      ser    serti  sta
 ral    ralju  pri rik    rinka  cau  rup    rupnu  dol          irs
         ncipal            se                 lar        set    senta  lay
 ram    ranmi  myt ril    rilti  rhy  rur    sruri  sur          er
         h                 thm                round      sev    senva  dre
 ran    ranti  sof rim    rimni  rhy  rus    grusi  gra          am
         t                 me                 y          sez    sevzi  sel
 rap    rapli  rep rin    krinu  rea  rut    grute  fru          f
         eat               son                it         sib    sidbo  ide
 rar    rarna  nat rip    cripu  bri  ruv    kruvi  cur          a
         ural              dge                ve         sic    stici  wes
 ras    grasu  gre rir    rirni  par  rux    pruxi  spi          t
         ase               ent                rit        sid    stidi  sug
 rat    ratni  ato ris    rismi  ric  sab    sabji  pro          gest
         m                 e                  vide       sig    sigja  cig
 rav    ragve  acr rit    brito  Bri  sac    stace  hon          ar
         oss               tish               est        sij    skiji  ski
 rax    ranxi  iro riv    rivbi  avo  sad    snada  suc  sik    silka  sil
         ny                id                 ceed               k
 raz    brazo  Bra rix    trixe  beh  sag    sanga  sin  sil    siclu  whi
         zilian            ind                g                  stle
 reb    rebla  tai rod    broda  pre  saj    sanji  con  sim    simxu  mut
         l                 dicate             scious             ual
 rec    rectu  mea         var 1      sak    sakci  suc  sin    tsina  sta
         t         rog    romge  chr          k                  ge
 red    bredi  rea         ome        sal    sakli  sli  sip    sipna  sle
         dy        rok    rokci  roc          de                 ep
 ref    krefu  rec         k          sam    skami  com  sir    sirji  str
         ur        rol    ro     eac          puter              aight
 rej    vreji  rec         h          san    spano  Spa  sis    sisku  see
         ord       rom    roi    qua          nish               k
 rek    greku  fra         ntified    sap    sampu  sim  sit    sitna  cit
         me                tense              ple                e
 rel    re     2   ron    ropno  Eur  sar    slari  sou  siv    sivni  pri
 rem    remna  hum         opean              r                  vate
         an        ror    rorci  pro  sas    srasu  gra  six    sirxo  Syr
 ren    trene  tra         create             ss                 ian
         in        ros    prosa  pro  sat    sakta  sug  siz    si'o   con
 rep    crepu  har         se                 ar                 cept
         vest      rot    rotsu  thi  sav    savru  noi          abstract
 rer    renro  thr         ck                 se         sob    sobde  soy
         ow        roz    rozgu  ros  sax    sarxe  har          a
 res    respa  rep         e                  monious    soc    sorcu  sto
         tile      rub    ruble  wea  saz    sazri  ope          re
 ret    preti  que         k                  rate       sod    sodva  sod
         stion     ruc    pruce  pro  seb    steba  fru          a
 rev    renvi  sur         cess               stration   sof    softo  Sov
         vive      rud    drudi  roo  sec    senci  sne          iet
 rib    cribe  bea         f                  eze        sog    sorgu  sor
         r         ruf    rufsu  rou  sed    stedu  hea          ghum
 ric    tricu  tre         gh                 d          soj    so'a   alm
         e         ruj    kruji  cre  sef    selfu  ser          ost all
                           am                 ve

 sol    solri  sol tap    stapa  ste  tis    tisna  fil  vas    vasru  con
         ar                p                  l                  tain
 som    sombo  sow tar    tarci  sta  tit    titla  swe  vat    vamtu  vom
 son    sonci  sol         r                  et                 it
         dier      tas    tansi  pan  tiv    tivni  tel  vax    vasxu  bre
 sop    so'e   mos tat    tatru  bre          evision            athe
         t                 ast        tix    tixnu  dau  vaz    va     the
 sor    so'i   man tav    tavla  tal          ghter              re at
         y                 k          tiz    stizu  cha  vef    venfu  rev
 sos    so'o   sev tax    tanxe  box          ir                 enge
         eral      taz    ta     tha  toc    troci  try  vel    ve     4th
 sot    so'u   few         t there    tod    toldi  but          conversio
 sov    sovda  egg teb    ctebi  lip          terfly             n
 soz    so     9   tec    steci  spe  tog    tonga  ton  ven    vecnu  sel
 sub    sfubu  div         cific              e                  l
         e         ted    terdi  ear  tok    toknu  ove  ver    verba  chi
 suc    sucta  abs         th                 n                  ld
         tract     tef    tenfa  exp  tol    to'e   pol  ves    vensa  spr
 sud    sudga  dry         onential           ar                 ing
 sug    sunga  gar teg    tengu  tex          opposite   vib    vibna  vag
         lic               ture       ton    torni  twi          ina
 suj    sumji  sum tek    cteki  tax          st         vic    vimcu  rem
 suk    suksa  sud tel    stela  loc  tor    tordu  sho          ove
         den               k                  rt         vid    vindu  poi
 sul    sunla  woo tem    temci  tim  tub    tunba  sib          son
         l                 e                  ling       vif    vifne  fre
 sum    sumti  arg ten    tcena  str  tuf    tu     tha          sh
         ument             etch               t yonder   vij    vinji  air
 sun    stuna  eas tep    terpa  fea  tug    tugni  agr          plane
         t                 r                  ee         vik    viknu  vis
 sup    su'e   at  ter    te     3rd  tuj    tujli  tul          cous
         most              conversio          ip         vil    vlile  vio
 sur    surla  rel         n          tuk    tunka  cop          lent
         ax        tet    terto  1E1          per        vim    vikmi  exc
 sut    sutra  fas         2          tul    tunlo  swa          rete
         t         tib    tinbe  obe          llow       vin    jvinu  vie
 suv    su'u   uns         y          tum    tumla  lan          w
         pecif     tic    tcica  dec          d          vip    vipsi  dep
         abstract          eive       tun    tunta  pok          uty
 suz    su'o   at  tid    tcidu  rea          e          vir    vidru  vir
         least             d          tup    tuple  leg          us
 tab    tabno  car tif    ti     thi  tur    stura  str  vis    viska  see
         bon               s here             ucture     vit    vitci  irr
 tac    tance  ton tig    tigni  per  tut    tutra  ter          egular
         gue               form               ritory     viz    vi     her
 tad    tadni  stu tij    tilju  hea  tuz    stuzi  sit          e at
         dy                vy                 e          vok    voksa  voi
 taf    taxfu  gar tik    stika  adj  vab    vanbi  env          ce
         ment              ust                ironment   vol    vofli  fli
 tag    tagji  snu til    tcila  det  vac    vanci  eve          ght
         g                 ail                ning       von    vo     4
 taj    tamji  thu tim    tcima  wea  vaj    vajni  imp  vor    vorme  doo
         mb                ther               ortant             r
 tak    staku  cer tin    tirna  hea  val    valsi  wor  vud    vrude  vir
         amic              r                  d                  tue
 tal    talsa  cha tip    tikpa  kic  vam    vamji  val  vur    vukro  Ukr
         llenge            k                  ue                 ainian
 tam    tarmi  sha tir    tirse  iro  van    vanju  win  vus    vrusi  tas
         pe                n                  e                  te
 tan    tsani  sky                    var    vacri  air

 vuz    vu     yon xol    xotli  hot  zug    zungi  gui  cmo    cmoni  moa
         der at            el                 lt                 n
 xab    xadba  hal xub    xruba  buc  zuk    zukte  act  cmu    jicmu  bas
         f                 kwheat     zul    zunle  lef          is
 xac    xarci  wea xuk    xruki  tur          t          cna    canpa  sho
         pon               key        zum    zu'o   act          vel
 xad    xadni  bod xul    xutla  smo          ivity      cne    cenba  var
         y                 oth                abstract           y
 xag    xamgu  goo xum    xukmi  che  zun    zunti  int  cni    cinmo  emo
         d                 mical              erfere             tion
 xaj    xarju  pig xun    xunre  red  zut    zutse  sit  cno    condi  dee
 xak    xaksu  use xur    xurdo  Urd                             p
         up                i                             cnu    macnu  man
 xal    xalka  alc xus    xusra  ass                             ual
         ohol              ert        bla    blanu  blu  cpa    cpacu  get
 xam    xajmi  fun zac    zarci  mar          e          cpe    cpedu  req
         ny                ket        ble    ruble  wea          uest
 xan    xance  han zag    zargu  but          k          cpi    cipni  bir
         d                 tock       bli    bliku  blo          d
 xap    xampo  amp zaj    zajba  gym          ck         cpu    lacpu  pul
         ere               nast       blo    bloti  boa          l
 xar    xanri  ima zal    zalvi  gri          t          cra    crane  fro
         ginary            nd         blu    ciblu  blo          nt
 xas    xamsi  sea zan    zabna  fav          od         cre    certu  exp
 xat    xatsi  1E-         orable     bra    barda  big          ert
         18        zar    zanru  app  bre    bredi  rea  cri    cirko  los
 xav    xa     6           rove               dy                 e
 xaz    xazdo  Asi zas    zasni  tem  bri    bridi  pre  cro    cortu  pai
         atic              porary             dicate             n
 xeb    xebro  Heb zat    zasti  exi  bro    xebro  Heb  cru    curmi  let
         rew               st                 rew        cta    catlu  loo
 xed    xendo  kin zaz    za'i   sta  bru    burcu  bru          k
         d                 te                 sh         cte    nicte  nig
 xej    xedja  jaw         abstract   cfa    cfari  ini          ht
 xek    xekri  bla zel    ze     7            tiate      cti    citka  eat
         ck        zen    zenba  inc  cfi    cfila  fla  cto    xecto  100
 xel    xe     5th         rease              w          ctu    ctuca  tea
         conversio zep    zepti  1E-  cfu    ricfu  ric          ch
         n                 21                 h          dja    cidja  foo
 xen    xebni  hat zer    zekri  cri  cka    ckana  bed          d
         e                 me         cke    ckeji  ash  dje    djedi  ful
 xer    xenru  reg zet    zetro  1E2          amed               l day
         ret               1          cki    ciksi  exp  dji    djica  des
 xes    xelso  Gre zev    ze'o   out          lain               ire
         ek                ward       cko    cokcu  soa  djo    sadjo  Sau
 xet    xecto  100 zif    zifre  fre          k up               di
 xex    xexso  1E1         e          cku    cukta  boo  dju    sidju  hel
         8         zin    zinki  zin          k                  p
 xil    xislu  whe         c          cla    clani  lon  dra    drani  cor
         el        zip    dzipo  Ant          g                  rect
 xim    xinmo  ink         arctican   cli    cilre  lea  dre    derxi  hea
 xin    xindo  Hin zir    zirpu  pur          rn                 p
         di                ple        clu    culno  ful  dri    badri  sad
 xip    xispo  His ziv    zivle  inv          l          dro    cidro  hyd
         panic             est        cma    cmalu  sma          rogen
 xir    xirma  hor zon    zo'a   tan          ll         dru    drudi  roo
         se                gential    cme    cmene  nam          f
 xis    xriso  Chr         to                 e          dza    da     som
         istian    zor    zo'i   inw  cmi    cmima  mem          ething 1
                           ard                ber

 dze    dzena  eld jge    jgena  kno  ple    pelji  pap  sni    sinxa  sig
         er                t                  er                 n
 dzi    dizlo  low jgi    jgira  pri  pli    pilno  use  sno    masno  slo
 dzu    cadzu  wal         de         plo    polje  fol          w
         k         jma    jamfu  foo          d          snu    casnu  dis
 fla    flalu  law         t          plu    daplu  isl          cuss
 fle    flecu  flo jme    jemna  gem          and        spa    spati  pla
         w         jmi    jimpe  und  pra    cupra  pro          nt
 fli    fliba  fai         erstand            duce       spe    speni  mar
         l         jva    javni  rul  pre    prenu  per          ried
 flo    foldi  fie         e                  son        spi    spisa  pie
         ld        jve    je     tan  pri    prina  pri          ce
 flu    fulta  flo         ru and             nt         spo    daspo  des
         at        jvi    jivna  com  pro    fapro  opp          troy
 fra    frati  rea         pete               ose        spu    spuda  rep
         ct        jvo    lujvo  aff  pru    purci  pas          ly
 fre    ferti  fer         ix                 t          sra    sarji  sup
         tile              compound   sfa    sfasa  pun          port
 fri    lifri  exp kla    klama  com          ish        sre    srera  err
         erience           e          sfe    sefta  sur  sri    dasri  rib
 fro    forca  for kle    klesi  cla          face               bon
         k                 ss         sfo    sfofa  sof  sro    sorcu  sto
 fru    frumu  fro kli    klina  cle          a                  re
         wn                ar         sfu    sufti  hoo  sru    sruri  sur
 gla    glare  hot klo    diklo  loc          f                  round
 gle    gletu  cop         al         ska    skari  col  sta    stali  rem
         ulate     klu    kulnu  cul          or                 ain
 gli    glico  Eng         ture       ske    saske  sci  ste    liste  lis
         lish      kra    krasi  sou          ence               t
 glu    gluta  glo         rce        ski    skicu  des  sti    sisti  cea
         ve        kre    kerfa  hai          cribe              se
 gra    grake  gra         r          sko    skori  cor  sto    stodi  con
         m         kri    krici  bel          d                  stant
 gre    pagre  pas         ieve       sku    cusku  exp  stu    stuzi  sit
         s through kro    korcu  ben          ress               e
 gri    girzu  gro         t          sla    salci  cel  tca    tcadu  cit
         up        kru    kruvi  cur          ebrate             y
 gru    gurni  gra         ve         sle    selci  cel  tce    mutce  muc
         in        mla    mlana  sid          l                  h
 jba    jbari  ber         e          sli    slilu  osc  tci    tutci  too
         ry        mle    melbi  bea          illate             l
 jbe    jbena  bor         utiful     slo    solji  gol  tco    ketco  Sou
         n         mli    milxe  mil          d                  th
 jbi    jibni  nea         d          slu    sluji  mus          American
         r         mlo    molki  mil          cle        tcu    nitcu  nee
 jbo    lojbo  Loj         l          sma    smaji  qui          d
         banic     mlu    simlu  see          et         tra    tarti  beh
 jbu    jubme  tab         m          sme    semto  Sem          ave
         le        mra    marbi  she          itic       tre    mitre  met
 jda    lijda  rel         lter       smi    simsa  sim          er
         igion     mre    merli  mea          ilar       tri    trina  att
 jde    kajde  war         sure       smo    smoka  soc          ract
         n         mri    mrilu  mai          k          tro    jitro  con
 jdi    jdice  dec         l          smu    smuni  mea          trol
         ide       mro    morsi  dea          ning       tru    turni  gov
 jdu    jduli  jel         d          sna    sance  sou          ern
         ly        mru    mruli  ham          nd         tsa    tsali  str
 jga    jganu  ang         mer        sne    senva  dre          ong
         le        pla    platu  pla          am         tse    zutse  sit

 tsi    tsiju  see zmu    zumri  mai  bu'i   bu     wor  cu'u   cuntu  aff
         d                 ze                 d to               air
 tsu    rotsu  thi zva    zvati  at           lerfu      da'a   damba  fig
         ck                           bu'o   budjo  Bud          ht
 vla    valsi  wor                            dhist      da'e   danre  pre
         d                            bu'u   bukpu  clo          ssure
 vle    zivle  inv ba'a   barna  mar          th         da'i   darxi  hit
         est               k          ca'a   cabra  app  da'o   darno  far
 vli    vlipa  pow ba'e   balre  bla          aratus     da'u   danlu  ani
         erful             de         ca'e   catke  sho          mal
 vra    vraga  lev ba'i   banli  gre          ve         dai    dacti  obj
         er                at         ca'i   catni  aut          ect
 vre    vreta  rec ba'o   banro  gro          hority     dau    darlu  arg
         lining            w          ca'o   canko  win          ue
 vri    virnu  bra ba'u   bacru  utt          dow        de'a   denpa  wai
         ve                er         ca'u   canlu  spa          t
 vro    vorme  doo bai    bapli  for          ce         de'i   denci  too
         r                 ce         cai    carmi  int          th
 vru    savru  noi bau    bangu  lan          ense       de'o   delno  can
         se                guage      cau    claxu  wit          dela
 xla    xlali  bad be'a   bersa  son          hout       de'u   dertu  dir
 xle    naxle  can be'e   bende  cre  ce'a   cecla  lau          t
         al                w                  ncher      dei    djedi  ful
 xli    nixli  gir be'i   benji  tra  ce'i   cteki  tax          l day
         l                 nsfer      ce'o   ce'o   in   di'a   jdima  pri
 xlu    xlura  inf be'o   bemro  Nor          a                  ce
         luence            th                 sequence   di'e   dirce  rad
 xra    pixra  pic         American           with               iate
         ture      be'u   betfu  abd  ce'u   cecmu  com  di'i   jdini  mon
 xre    mixre  mix         omen               munity             ey
         ture      bei    bevri  car  cei    cevni  god  di'o   dinko  nai
 xri    maxri  whe         ry         ci'a   ciska  wri          l
         at        bi'a   bilma  ill          te         di'u   dinju  bui
 xru    xruti  ret bi'e   brife  bre  ci'e   ciste  sys          lding
         urn               eze                tem        do'i   donri  day
 zba    zbasu  mak bi'i   jbini  bet  ci'i   cinri  int          time
         e                 ween               eresting   do'o   dotco  Ger
 zbe    zbepi  ped bi'o   binxo  bec  ci'o   citno  you          man
         estal             ome                ng         doi    do     you
 zbi    nazbi  nos bi'u   bitmu  wal  ci'u   ckilu  sca  du'a   dunda  giv
         e                 l                  le                 e
 zda    zdani  nes bo'a   boxna  wav  co'a   co'a   ini  du'e   dukse  exc
         t                 e                  tiative            ess
 zdi    zdile  amu bo'e   brode  pre  co'e   co'e   uns  du'i   dunli  equ
         sing              dicate             pecif              al
 zdo    xazdo  Asi         var 2              bridi      du'o   du     sam
         atic      bo'i   botpi  bot  co'i   cmoni  moa          e
 zga    zgana  obs         tle                n                  identity
         erve      bo'o   boxfo  she  co'u   co'u   ces          as
 zgi    zgike  mus         et                 sative     du'u   dunku  ang
         ic        bo'u   bongu  bon  coi    condi  dee          uish
 zgu    rozgu  ros         e                  p          fa'a   farna  dir
         e         boi    bolci  bal  cu'a   cuxna  cho          ection
 zma    zmadu  mor         l                  ose        fa'e   fatne  rev
         e         bu'a   bruna  bro  cu'e   ckule  sch          erse
 zme    guzme  mel         ther               ool        fa'i   facki  dis
         on        bu'e   bunre  bro  cu'i   cumki  pos          cover
 zmi    zmiku  aut         wn                 sible      fa'o   fanmo  end
         omatic                       cu'o   cunso  ran  fa'u   farlu  fal
                                              dom                l

 fai    fatri  dis gi'e   zgike  mus  ju'i   jundi  att  ku'u   ckunu  con
         tribute           ic                 entive             ifer
 fau    fasnu  eve gi'o   gigdo  1E9  ju'o   djuno  kno  la'a   lasna  fas
         nt        gi'u   gismu  roo          w                  ten
 fe'a   fenra  cra         t word     ka'a   katna  cut  la'e   lakne  pro
         ck        gu'a   gunka  wor  ka'e   kakne  abl          bable
 fe'i   fetsi  fem         k                  e          la'i   lamji  adj
         ale       gu'e   gugde  cou  ka'i   krati  rep          acent
 fe'o   fenso  sew         ntry               resent     la'o   latmo  Lat
 fe'u   fengu  ang gu'i   gusni  ill  ka'o   kanro  hea          in
         ry                umine              lthy       la'u   lalxu  lak
 fei    fepni  cen gu'o   gunro  rol  ka'u   kantu  qua          e
         t                 l                  ntum       lai    klani  qua
 fi'a   cfika  fic ja'a   jatna  cap  kai    ckaji  qua          ntity
         tion              tain               lity       lau    cladu  lou
 fi'e   finpe  fis ja'e   jalge  res  kau    kampu  com          d
         h                 ult                mon        le'a   lebna  tak
 fi'i   finti  inv ja'i   jadni  ado  ke'a   kevna  cav          e
         ent               rn                 ity        le'i   pleji  pay
 fi'o   friko  Afr ja'o   jarco  sho  ke'e   ke'e   end  le'o   lenjo  len
         ican              w                  grouping           s
 fi'u   cfipu  con ja'u   jgalu  cla  ke'i   kecti  pit  le'u   lerfu  let
         fusing            w                  y                  teral
 fo'a   fo'a   it- jai    jgari  gra  ke'o   kelvo  kel  lei    klesi  cla
         6                 sp                 vin                ss
 fo'e   fo'e   it- jau    djacu  wat  ke'u   krefu  rec  li'a   cliva  lea
         7                 er                 ur                 ve
 fo'i   fo'i   it- je'a   jecta  pol  kei    kelci  pla  li'e   lidne  pre
         8                 ity                y                  cede
 fo'o   fonmo  foa je'e   jetce  jet  ki'a   krixa  cry  li'i   linji  lin
         m         je'i   jersi  cha          out                e
 foi    foldi  fie         se         ki'e   kicne  cus  li'o   linto  lig
         ld        je'o   jegvo  Jeh          hion               htweight
 fu'a   funca  luc         ovist      ki'i   ckini  rel  li'u   litru  tra
         k         je'u   jetnu  tru          ated               vel
 fu'e   fuzme  res         e          ki'o   kilto  100  lo'i   bloti  boa
         ponsible  jei    jenmi  arm          0                  t
 fu'i   fukpi  cop         y          ki'u   krinu  rea  lo'o   slovo  Sla
         y         ji'a   jinga  win          son                vic
 ga'a   grana  rod ji'e   jmive  liv  ko'a   kojna  cor  lo'u   lorxu  fox
 ga'e   ganse  sen         e                  ner        loi    loldi  flo
         se        ji'i   jinvi  opi  ko'e   kolme  coa          or
 ga'i   galfi  mod         ne                 l          lu'a   pluta  rou
         ify       ji'o   jipno  tip  ko'i   kobli  cab          te
 ga'o   ganlo  clo ji'u   jvinu  vie          bage       lu'e   klupe  scr
         sed               w          ko'o   skoto  Sco          ew
 ga'u   galtu  hig jo'e   jorne  joi          ttish      lu'i   lumci  was
         h                 ned        ko'u   konju  con          h
 gai    gacri  cov jo'o   jordo  Jor          e          lu'o   lubno  Leb
         er                danian     koi    korbi  edg          anese
 gau    gasnu  do  jo'u   jo'u   in           e          ma'a   cmana  mou
 ge'a   gerna  gra         common     ku'a   kumfa  roo          ntain
         mmar              with               m          ma'e   marce  veh
 ge'o   gento  Arg joi    joi    in   ku'e   kuspe  ran          icle
         entinian          a mass             ge         ma'i   masti  mon
 ge'u   gerku  dog         with       ku'i   kurji  tak          th
 gei    gleki  hap ju'a   jufra  sen          e care of  ma'o   cmavo  str
         py                tence      ku'o   skuro  gro          ucture
 gi'a   gidva  gui ju'e   julne  net          ove                word

 ma'u   makcu  mat nai    natmi  nat  pi'a   pilka  cru  ri'e   rirxe  riv
         ure               ion                st                 er
 mai    marji  mat nau    nanmu  man  pi'e   plipe  lea  ri'i   ritli  rit
         erial     ne'i   nenri  in           p                  e
 mau    zmadu  mor ne'o   cnebo  nec  pi'i   pilji  mul  ri'o   crino  gre
         e                 k                  tiply              en
 me'a   mleca  les ne'u   cnemu  rew  pi'o   pipno  pia  ri'u   rinju  res
         s                 ard                no                 train
 me'e   cmene  nam nei    nelci  fon  pi'u   pimlu  fea  ro'a   prosa  pro
         e                 d                  ther               se
 me'i   mensi  sis ni'a   cnita  ben  po'a   spoja  exp  ro'i   rokci  roc
         ter               eath               lode               k
 me'o   mekso  MEX ni'e   nilce  fur  po'e   ponse  pos  ro'o   ropno  Eur
 me'u   mentu  min         niture             sess               opean
         ute       ni'i   nibli  nec  po'i   porpi  bre  ro'u   rotsu  thi
 mei    mei    car         essitate           ak                 ck
         dinal     ni'o   cnino  new  po'o   ponjo  Jap  roi    roi    qua
         selbri    ni'u   ninmu  wom          anese              ntified
 mi'a   cmila  lau         an         po'u   spofu  bro          tense
         gh        no'e   no'e   sca          ken        ru'a   sruma  ass
 mi'e   minde  com         lar        poi    porsi  seq          ume
         mand              midpoint           uence      ru'e   pruce  pro
 mi'i   minji  mac         not        pu'a   pluka  ple          cess
         hine      no'i   nobli  nob          asant      ru'i   pruxi  spi
 mi'o   misno  fam         le         pu'e   pulce  dus          rit
         ous       noi    notci  mes          t          ru'o   rusko  Rus
 mi'u   mintu  sam         sage       pu'i   punji  put          sian
         e         nu'a   snura  sec  pu'o   purmo  pow  ru'u   rupnu  dol
 mo'a   morna  pat         ure                der                lar
         tern      nu'e   nupre  pro  pu'u   sputu  spi  sa'a   sanga  sin
 mo'i   morji  rem         mise               t                  g
         ember     nu'i   nutli  neu  ra'a   srana  per  sa'e   satre  str
 mo'o   molro  mol         tral               tain               oke
         e         nu'o   nu'o   can  ra'e   ralte  ret  sa'i   sanli  sta
 mo'u   moklu  mou         but has            ain                nd
         th                not        ra'i   ranji  con  sa'o   salpo  slo
 moi    moi    ord pa'a   pacna  hop          tinue              pe
         inal              e          ra'o   radno  rad  sa'u   sarcu  nec
         selbri    pa'e   prane  per          ian                essary
 mu'a   murta  cur         fect       ra'u   raktu  tro  sai    sanmi  mea
         tain      pa'i   prami  lov          uble               l
 mu'e   munje  uni         e          rai    traji  sup  sau    slabu  old
         verse     pa'o   panlo  sli          erlative   se'a   setca  ins
 mu'i   mukti  mot         ce         rau    gradu  uni          ert
         ive       pa'u   patfu  fat          t          se'i   sevzi  sel
 mu'o   mulno  com         her        re'a   remna  hum          f
         plete     pai    pajni  jud          an         se'u   selfu  ser
 mu'u   muvdu  mov         ge         re'e   trene  tra          ve
         e         pau    pagbu  par          in         sei    sepli  apa
 na'a   nanca  yea         t          re'i   renvi  sur          rt
         r         pe'a   preja  spr          vive       si'a   sinma  est
 na'e   natfe  den         ead        re'o   renro  thr          eem
         y         pe'i   penmi  mee          ow         si'e   snime  sno
 na'i   nalci  win         t          re'u   rectu  mea          w
         g         pe'o   pendo  fri          t          si'i   sicni  coi
 na'o   cnano  nor         end        rei    preti  que          n
         m         pe'u   pencu  tou          stion      si'o   sidbo  ide
 na'u   namcu  num         ch         ri'a   rinka  cau          a
         ber       pei    pensi  thi          se         si'u   simxu  mut
                           nk                                    ual

 so'a   sovda  egg to'u   tordu  sho  xau    xamgu  goo  zu'e   zukte  act
 so'e   sobde  soy         rt                 d          zu'i   zunti  int
         a         toi    troci  try  xe'a   xedja  jaw          erfere
 so'i   so'i   man tu'a   tumla  lan  xe'i   xekri  bla
         y                 d                  ck
 so'o   sombo  sow tu'e   tuple  leg  xe'o   xendo  kin
 soi    sonci  sol tu'i   tugni  agr          d
         dier              ee         xe'u   xenru  reg
 su'a   stura  str tu'o   tunlo  swa          ret
         ucture            llow       xei    xebni  hat
 su'e   su'e   at  tu'u   tubnu  tub          e
         most              e          xi'a   xirma  hor
 su'i   sumti  arg va'i   vamji  val          se
         ument             ue         xi'o   xriso  Chr
 su'o   su'o   at  va'u   vasxu  bre          istian
         least             athe       xi'u   xislu  whe
 su'u   sfubu  div vai    vajni  imp          el
         e                 ortant     xoi    xotli  hot
 ta'a   tavla  tal vau    vasru  con          el
         k                 tain       xu'a   xusra  ass
 ta'e   tanxe  box ve'a   verba  chi          ert
 ta'i   tatpi  tir         ld         xu'e   xunre  red
         ed        ve'e   ve'e   who  xu'i   xukmi  che
 ta'o   tanbo  boa         le space           mical
         rd                interval   xu'o   xurdo  Urd
 ta'u   taxfu  gar ve'u   vecnu  sel          i
         ment              l          za'a   zabna  fav
 tai    tarmi  sha vei    vreji  rec          orable
         pe                ord        za'i   zasti  exi
 tau    tanru  phr vi'a   viska  see          st
         ase       vi'e   vitke  gue  za'o   za'o   sup
         compound          st                 erfective
 te'a   terpa  fea vi'i   vikmi  exc  za'u   zargu  but
         r                 rete               tock
 te'i   steci  spe vi'o   vitno  per  zai    zarci  mar
         cific             manent             ket
 te'o   stero  ste vi'u   vimcu  rem  zau    zanru  app
         radian            ove                rove
 te'u   tengu  tex vo'a   voksa  voi  ze'a   zenba  inc
         ture              ce                 rease
 tei    temci  tim voi    vofli  fli  ze'e   ze'e   who
         e                 ght                le time
 ti'a   tcima  wea vu'e   vrude  vir          interval
         ther              tue        ze'o   ze'o   out
 ti'e   trixe  beh vu'i   vrusi  tas          ward
         ind               te         zei    zekri  cri
 ti'i   stidi  sug vu'o   vukro  Ukr          me
         gest              ainian     zi'e   zifre  fre
 ti'o   ctino  sha xa'a   xatra  let          e
         dow               ter        zi'i   zinki  zin
 ti'u   tixnu  dau xa'e   xance  han          c
         ghter             d          zi'o   dzipo  Ant
 to'a   tonga  ton xa'i   xarci  wea          arctican
         e                 pon        zi'u   zirpu  pur
 to'e   to'e   pol xa'o   xampo  amp          ple
         ar                ere        zo'a   zo'a   tan
         opposite  xa'u   xabju  dwe          gential
 to'i   torni  twi         ll                 to
         st        xai    xrani  inj  zo'i   zo'i   inw
                           ure                ard