UI types

From Lojban
Jump to navigation Jump to search

And Rosta:

UI types

True interjections

English examples are 'wow', 'ouch' etc. They are mere vocalizations of what could equally well be expressed by gesture, facial expression, or whatever. They therefore don't contribute to the logicosemantic form expressed by the rest of the sentence.

To this, xorxes has said:

It seems to me that lexicalized interjections will inevitably end up acquiring logical/semantic properties. For example, take {ua}, "discovery". It sounds like a true interjection, but I wouldn't mind being able to say something like:

la alis cu morji le du'u ua le ckiku cu cnita le rulpatxu

Alice remembers that (discovery!) the key is under the flower-pot.

My response to this is that interjections can be used empathetically, especially in narrative:

"He fell out of bed -- ouch!"

"She opened the envelope and, wow!, found a cheque for $50."

Illocutionary operators

These encode the illocutionary force of the utterance -- a question, command, wish, request, hope, speculation, assertion, and so forth. They participate in the logical form of the sentence: they have scope over the entire propositional content, where the propositional content is what is asserted, requested, hoped for, etc. Material outside the scope of the illocutionary operator is presupposed (indeed, IMO this is the definition of what presupposition is), or, IMO equivalently, is Grice's 'conventional implicature'.

Ideally illocutionary operators would not be in UI, since they are a variety of predicate. At any rate, they are sensistive to scope. In subordinate bridi they would still express illocutionary operators:

la alis jinvi ledu'u e'o ko'a cliva

Alice believes that he leaves, and I hereby-request that he does so.

I hereby-request that he leave, which Alice believes.

Assuming left to right scope, and perhaps default assertive illocutionary force.

la alis cusku lesedu'u e'o ko'a cliva

I hereby-request that, as discussed by Alice, he leave.

But analogy with xukau suggests that UIkau could be used to restrict the illocutionay force to the local bridi.

la alis cusku le se du'u e'o kau mi cliva

Alice expressed a request that I leave.

and beyond UI:

la alis cusku le se du'u ko kau cliva

Alice expressed a command that I leave.

xorxes comments:

I did experiment with that for a while some time ago, but I was not very satisfied with the results. I think there is a difference in saying whether John gave the flowers or not: la alis cusku le se du'u xu la djan pu dunda lo xrula la meris, she said either that he did or that he didn't, and saying words to the effect of asking a question. Hmmm, maybe something like:

la alis cusku le se du'u pau xukau la djan pu dunda lo xrula la meris

Alice said (question) whether John gave flowers to Mary.

la alis cusku le se du'u ju'a xukau la djan pu dunda lo xrula la meris

Alice said (assertion) whether John gave flowers to Mary.

If pau and ju'a are illocutionary operators then given what I have said, they too would need to be paukau, ju'akau. However, I accept that the xukau analogy is treacherous. The treacherousness arises from the fact that bare xu expresses an illocutionary operator (pau) plus what can informally be called a 'WH quantifier', while Qkau expresses only the WH quantifier. So Qkau cannot have a compositional meaning and is majorly fucked up. (I have proposed experimental NU that handle indirect questions better, so the case for Qkau is that it is entrenched, not that it is necessary.)

I suppose that all Q words could be seen as abbreviations for Qpau. Then Q could express the WH quantifier and pau the illocutionary operator. And kau would be a kind of zi'oish cmavo that means "don't insert a pau here". That would make Qkau compositional, but would nix my suggestion for e'okau, kokau etc.

Presuppositional elements

These are elements that are not illocutionary operators, are part of logical form, and are outside the scope of the illocutionary operator. An example would be {ku'i}, 'but', and the other discursives and the evidentials.

These are partially sensitive to scope, in that nothing has scope over them (-- they necessarily have maximally wide scope), but they themselves may have scope over only, say, a single phrase within the sentence. However, as with illocutionary operators, there is a way in which we might want to restrict them to subordinate bridi in a variety of indirect speech.

direct speech: Alice said "I, however, am leaving"

'semi direct': Alice expressed a proposition that is expressible as "She, however, is leaving".

direct speech: Alice said "Would that I were leaving"

'semi direct': Alice expressed a proposition that is expressible as "Would that she were leaving".

What xorxes is (I think) seeking is a way to express 'semi direct' speech: it must be like a quotation in restricting illocutionary and presuppositional elements to within its scope, but unlike a quotation in allowing anaphora and other sorts of binding to cross its boundaries. Currently it seems to me that the best way to achieve this is to use a new LU that is pertransitable by binding relations.

Scope summary

  • True interjections are not part of the logical structure expressed by the rest of the sentence.
  • Phrases can be inside or outside the scope of illocutionary operators; logically, an illocutionary operator has scope over a bridi. This can perhaps be handled by the usual left-to-right scope rule. In addition, the operators can be 'quotativized' by placing them within the scope of a LU.
  • Phrases can be inside or outside the scope of presuppositional elements, which can have scope over any sort of phrase. I'm not sure how this scope is best expressed syntactically. In addition, the presuppositionals can be 'quotativized' by placing them within the scope of a LU.

Added clarification on scope

In a structure like {su'o da ro de su'o di}, scope works in a simple left-to-right way such that each word has scope over everything that follows it (within the bridi): so {ro de} has scope over {su'o di}, is outside the scope of what it precedes (su'o di) and inside the scope of what it follows (su'o da). But with something like "ASSERTED: p or, in-other-words, q", q is within the scope of "in other words" (chosen here as an example of a presuppositional element) but "in other words" is not within the scope of what it follows. The logical structure is rather something like "(ASSERTED: p or q) and q is in-other-words". What makes illocutionaries and presuppositionals special is that nothing can have scope over them. (That is arguable: in "Is it dinner time yet, because I'm really hungry", the logical structure seems to be "Because I'm really hungry: ASKED-WHETHER: it is dinner time". But I don't think Lojban is robust enough to handle this sort of stuff merely by means of UI and word order.)


Thanx. This clarifies a bunch of issues and takes steps toward solutions for most of them.pc

We need to be sure that the various forms are distinctive so that UI that function in more than one area (if there are such) will be unambiguous in a particular use. pc


And: I suppose that all Q words could be seen as abbreviations for Qpau. Then Q could express the WH quantifier and pau the illocutionary operator. And kau would be a kind of zi'oish cmavo that means "don't insert a pau here". That would make Qkau compositional, but would nix my suggestion for e'okau, kokau etc.

  • That's very clear. I think that's what made me abandon kau for that function, though I had not been able to articulate the reason.
  • Is there any reason why illocutionary operators could not be restricted by default to the local bridi in abstractions?

And Rosta replying to the question:

In the case of Q(pau) and ko there is a practical reason, because they conflate two logical elements -- an illocutionary operator (question/command) plus 'WH'/'you'. Currently it is the latter element, WH/you, that determines which bridi the word is placed in. However, for the straightforward exx, such as e'o and a'o, I think the interpretation is this

lakne fa ledu'u a'o mi klama

"Would that (as is likely) I come!"

"(DESIRED: I come) and (likely: I come)"

rather than this

ko'a tavla ledu'u a'o mi klama

NOT: "She said she hopes I'll come"

RATHER: "Would that (as she says) I come!"

To get the reading "She said what could be expressed as 'would that And comes'", i.e. where something has scope over the illocutionary operator, I have suggested "cusku LU a'o". IOW, the only way for something to have scope over an illocutionary or presuppositional is to somehow embed it within some (quasi)quotative context.

  • Bingo! Though I am no so sure about {Qpau}: WH? (not WH<) is an illocutionary operator.

pc estivates (just barely, technically)

And Rosta: re Q(pau). This involves the illocutionary element -- 'TELL-ME', or whatever, plus the machinery involving sets of answers that you & xorxes worked out. I am using 'WH' as a shorthand for the machinery involving sets of answers. So in "She knows who went", we have 'WH' involved (the sets of answers apparatus) but not the illocutionary element.


xorxes: I understand the description of what presumably is the status quo. What I don't see is a reason to prefer this over local scope in abstractions. Is it just that the rule is easier to state ("always main bridi scope")? All the examples seem to give things one would hardly ever say, whereas local scope often gives things one does want to say.

And Rosta: I am deducing the scope rules from the logical character of illocutionaries and presuppositionals. An illocutionary encodes not a predicate, which is something that has a truth value when its arguments are bound, but an action (of assertion, hoping, command, etc.). Utterances are actions but bridis aren't; bridis consist of things that have truth values. So an illocutionary encodes the kind of action that the utterance is. Ergo illocutionaries always have utterance-level scope. Because quotatives allow one utterance to be embedded within another, quotatives are the way to restrict illocutionaries to a subpart of the utterance. As for presuppositionals, they by their very nature encode the fact that they are outside all the truth-conditional content: perforce, that then takes them to utterance-level.

If it is true that I am describing the status quo (& I'm not convinced I am!), then for once the status quo is right! If your preferred system is more useful, then we need to work out a logically-coherent story for it, and a way of encoding it.

xorxes: I'm not sure that your ergo follows.

A bridi consists of a selbri and zero or more sumti. (A sumti can refer to bridis, among other things.)

Bridis can be used (together with other things) to make utterances. The same bridi can be used to make different types of utterance, so we use illocutionaries to indicate what type of utterance the bridi is being used to make.

  • It might be helpful to distinguish between Complete utterances and Incomplete utterances, where completeness is inversely proportional to the amount of information that the hearer *must* glork. Setting aside ineradicable incompleteness (such as arises from specific reference), we can say that an utterance like "lo gerku" is incomplete, not least because the hearer has to glork the rest of the bridi. But "lo gerku cu bacru" is also incomplete, because the hearer has to glork what sort of action the utterance performs -- idle speculation? question? hope? assertion? A logically Complete utterance, then, requires an illocutionary operator.

When I say: {ju'a la djan klama le zarci}, I am using the expression {la djan klama le zarci} which corresponds to a bridi, adorned with ju'a to make an assertion. When I say {a'o la djan klama le zarci}, I use the same expression {la djan klama le zarci}, this time adorned with {a'o}, to express a hope. With {e'o la djan klama le zarci} I make a request, and so on.

A sumti is used to refer to things. In particular, we may want to refer to some propositional content. When I say {ju'a la alis krici le du'u la djan klama le zarci}, I am using the same expression I used before {la djan klama le zarci}, but this time within a sumti. I don't use it to assert its propositional content. I use it only to refer to its propositional content.

But we may also want to refer to some propositional content plus the use it is put to, without actually using it ourselves at that time. i.e. refer to the assertion, the request, the hope (but not anyone's in particular) that can be made with that bridi.

  • To refer to some propositional content plus the use it is put to is to refer to an utterance. Of course we can refer to an utterance without using it ourselves at that time. That is what quotative devices are for, and I have said this already.

You say that marking any bridi with ju'a, including one within a sumti, indicates by itself that the speaker is making an assertion. It is not obvious to me that marking a bridi within a sumti with an illocutionary indicator must mean that the bridi's propositional content is suddenly being used with illocutionary force in addition to being referred to. The bridi is then doing double duty.

  • There isn't a distinction between 'referring' to propositional content and 'using' it. Or rather, the distinction is that a 'referred to' proposition is an argument of a (truth-valued) predicate, while a 'used' proposition is an argument of an illocutionary operator. So the double-duty is no more than a single proposition being an argument of two different elements. There is nothing weird in that: sumti that are arguments of more than one predicate are two a penny. An illocutionary operator isn't a feature of a bridi: rather it encodes that the speaker is performing an action of a certain kind, an action that applies to a bridi.
    • (Just thinking aloud:) I find it weird, and none of the examples sound natural to me. Perhaps it is because most arguments of truth-valued predicates cannot be used as arguments of illocutionary operators. A dog can be an argument of a truth-valued predicate, but it can't be an argument of an illocutionary operator. We can assert a proposition and we can pet a dog. In saying that sentence I referred to propositions and to dogs. But in actually doing those things (asserting and petting), I don't refer to propositions or dogs. So it is odd when a proposition is asserted and referred to in the very same act. It's as if by petting a dog I referred to it in the same act. I guess in a sense we can do that. Someone asks "which one is your dog?", then I pet my dog and by so doing I'm answering the question. But that would be like someone asking "what does Alice think?" and I answer "tomorrow it will rain". By asserting a proposition, I suggest that that is what Alice thinks, just as by petting the dog I suggested that that one was mine. So I can pet/assert something as an indirect way of referring to it, but I can't use a reference ("lo gerku", "lo du'u broda") to perform an act of petting/asserting them. So, no, I don't think that a proposition can be referred to (as an argument of a predicate) and illocuted by the very same act. It's weird. (End of thinking aloud.)
    • And Rosta: I think a dog can be an argument of an illocutionary operator: {doi rover}, {doi su'o gerku}, {doi le gerku}. {doi} is of course an illocution of identifying or defining the addressee. (Isn't it? Or is it truly just a mere vocative? If it is just a mere vocative, then imagine a cmavo similar to doi that does identify or define the addressee.)
    • xorxes: OK, let's say it is. But we don't use it at the same time as an argument of doi and an argument of a predicate. We don't say {mi nelci doi rover} "I like hey you rover!".
    • And Rosta: Well, {do doi} in {mi nelci do doi rover} seems to me rather redundant. It would be good to have a variety of {do} that can both refer to the addressee and simultaneously identify/define who the address is. But be that as it may, in {doi le nanmu ku noi broda}, le nanmu is an argument of broda.
    • xorxes: Right. It is interesting though that in this case it is first an argument of the illocutionary, and only secondarily an argument of a predicate. With UI within abstractions it would be the other way around. Something like "O that John would come, as you believe" sounds perfectly natural, but "You believe that ..." with an argument that is also performed doesn't. Am I just being hopelessly malrarna here?
    • And Rosta: grammatically malrarna but not pragmatically, because it gardenpaths you into thinking that "you believe" is being asserted. However, if "you believe" were preceded by a UI marking the absence of illocution, then I think the sentence would be more palatable. Also the fact that the subordinating predicate is 'believe' -- leads the hearer to be expecting us to move to the believer's perspective. But if the predicate is 'past' -- {balvi fa lenu ui do klama}, meaning "You came (as we both know), and I'm happy about the coming (but not about it being in the past) -- then it seems less egregious than with 'believe'.

I say that marking a bridi with ju'a marks it as being of type assertion (so if the bridi is posed, it will make the utterance an assertion and not a request), but that for a speaker to make the assertion they must pose the marked bridi itself. If they use the marked bridi within a sumti, they are not making the assertion, they only refer to it, just as by using the sumti {le du'u la djan klama le zarci} one refers to a propositional content but does not use it directly (one does not assert it, request it, hope for it, etc.)

  • The problem with this is that assertion is not a type of bridi. It is a type of action ('illocution') that applies to a bridi. But given that, let's see how we can formulate your preferred way:
    1. A syntactic bridi that contains no illocutionary word expresses a proposition.
    1. A syntactic bridi that contains an illocutionary word expresses an illocution.
    1. A subordinate bridi that contains an illocutionary word mentions an illocution.
    1. A matrix bridi that contains an illocutionary word (at matrix level) uses (i.e. performs) an illocution.
  • Now, this is not untenable, but it is not deducible from the rest of the grammar; instead, it requires the stipulations I've listed. My major objection would be that Stipulation (3) is not logically justified (-- there's no logical reason why occurrence in a subordinate bridi should trigger a shift from use to mention) and subverts the usual quotative mechanisms for making the use/mention distinction. My minor objection is that syntactic bridi would now at some times express propositions and at other times express illocutions; {du'u} would now mean "is the illocution or proposition <bridi>" -- not a very useful concept.
    • Isn't it also the case that a subordinate bridi that contains no illocutionary word mentions a proposition?
    • And Rosta: the use/mention distinction doesn't apply to propositions (or dogs, etc.). It applies only to utterances/utterabilia.
    • xorxes: Couldn't we say that it can only cause confusion with utterabilia? Using and mentioning a spoon are clearly different things, while using or mentioning a word may more easily be confused. Asserting something is different from requesting it. What is that something, which can be asserted, requested, and also mentioned, as I'm doing now (neither asserting nor requesting it)?
    • 'Doing/performing' and 'talking about' are clearer terms than the technical terms 'using' and 'mentioning'. We don't do or perform a spoon. But change the example to a dance, and yes, it is true that the confusion enters the picture when the function of the performables is to talk about things (or be in some way representational). Going back to the point about whether the proposition is mentioned, I don't think the 'performed'/'talked about' distinction applies to propositions (or spoons).
    • xorxes: Is there something that can be either asserted or talked about? If there is, what do we call it?
    • And Rosta: Propositions can be asserted or talked about. (Lots of things, such as dogs, can be talked about, but only propositions can be asserted.) But you can't perform a proposition; you can only perform an assertion (or other illocution) of a proposition.

The question now is, what can one do with an assertion, a request, a hope, other than assert it, request it, hope for it? If you can do nothing with an assertion but assert it, nothing with a request but request it, nothing with a hope but hope for it, then there is not that much point in being able to refer to it with a sumti, as long as we have a suitable reporting selbri: Alice says {ju'a broda} and we report {la alis xusra le du'u broda}, she says {e'o broda} and we report {la alis cpedu le du'u broda}, she says {a'o broda} and we report {la alis pacna le du'u broda}, the selbri already says what the user of the propositional content uses it for, so the propositional content need not be presented differently in each case.

  • This is what quotative devices are for. As I've said above, it seems to me that a nonverbatim quotative that is transparent to binding relations would be the right tool for this job.
    • OK. Isn't that what {sedu'u <bridi>} is? Isn't it "is an utterance corresponding to the syntactic bridi <bridi>"?
    • And Rosta: That's what the baseline says. (I've always thought it a bit broken.) But anyway, the snag here is that you want the syntactic bridi to not correspond to a semantic bridi (instead it corresponds to an illocution). I think it is always better to keep the syntax close to the underlying logic, whenever possible, in which case sedu'u should mean "is an illocution applied to the proposition expressed by <bridi>".
    • xorxes: I'm confused. I want the syntactic bridi to correspond to a semantic bridi, and the uttering of a syntactic bridi to be an illocution (an utterance). But the illocution is the utterance of the matrix bridi, the sub-bridi referred to as a term would not itself be illocuted.
    • And Rosta: So du'u...kei contains a syntactic bridi, which corresponds to a semantic bridi. sedu'u refers to an illocution applied to that bridi. But since an illocutionary UI is not part of a semantic bridi, {sedu'u UI broda} would not refer to an illocution of type UI.
    • xorxes: I see. So {sedu'u UI broda} could refer to an illocution of any type, not necessarily one of type UI. Couldn't we say that UI always indicates the type of illocution that corresponds to its argument bridi, whether or not it is actually performed?
    • And Rosta: We could say that, but it would be a stipulation that doesn't follow from other principles. What follows from (pe'i) established principles is that the UI marks the speaker's illocutionary treatment of the proposition. That's why I prefer a LU, since it is already established that the nature of quotatives (in contrast to NU) to turn Use into Mention.

If we had a selbri "x1 makes/poses/expresses illocution x2", where x2 is propositional content plus illocutionary type, we would need to indicate the type of illocution somehow. I propose {le du'u UI broda} to be that: propositional content provided by broda and illocutionary type provided by UI. Anther possible predicate that use reference to illocutions might be "x1 shares with x2 the posing of illocution x3" (tugni)

  • For the reasons given above, I counterpropose {LU (UI) broda}.
    • Would that LU be equivalent to {la'e lu}, the meaning of the quoted words?
    • And Rosta: Not strictly equivalent, because the LU is nonverbatim and transparent to binding. As for loosely equivalent, then yes, if "words" are taken to be phonological strings, but the la'e/lu'e distinction is weak when we come to illocutionaries. If you take the word {gerku}, a pairing between sound /gerku/ and concept 'dog', then it is clear that the word is not a dog. But (switching to English) the word 'hello' (an illocutionary) is a greeting: find an utterance of the word, and you find a greeting. That is, an illocutionary word's meaning is a specification of the word's function -- a specification of what illocution the word performs.

xorxes:

OK, I think you've convinced me that UI can't do what I want inside NU abstractions.

Now, are these two equivalent, at least for illocutionaries:

  1. UI1 broda LE NU UI2 brode
  1. UI1 broda LE NU brode i UI2 brode

?

In other words, does an illocutionary inside an abstraction create a new utterance, as it were?

And Rosta: They have to be equivalent, I think, because brode is part of the proposition that UI1 has scope over, so in the first, both UI1 and UI2 have scope over brode. E.g. "COMMAND: you fetch the book that ASSERTED: is in the cupboard" -- asserting that the book is in the cupboard, so you can't satisfy my command by first placing a book in the cupboard and then fetching it. Or "COMMAND: you imagine that DENIAL: I am crazy" = "Imagine I'm crazy (but I'm not)".

xorxes: Do we have an illocutionary operator for DENIED? Is that what {ju'anai} should be? Are all of the following illocutionary operators?

e'a PERMITTED:

e'acu'i neither PERMITTED nor FORBIDDEN:

e'anai FORBIDDEN:

e'e EXHORTED:

e'o REQUESTED:

e'u SUGGESTED:

ai INTENDED:

au WISHED:

a'o HOPED:

ju'a ASSERTED:

ju'acu'i neither ASSERTED nor DENIED:

ju'anai DENIED:

ca'e DEFINED (content made true by the utterance):

ru'a POSITED:

ja'o CONCLUDED:

su'a GENERALIZED:

su'anai PARTICULARIZED:

ba'a EXPECTED:

ba'acu'i EXPERIENCED (now):

ba'anai REMEMBERED:

za'a OBSERVED (witnessed):

ti'e HEARD (not witnessed):

ka'u KNOWN (common knowledge):

se'o INTUITED:

pe'i OPINED:

Many of those seem to be varieties of ASSERTED.

And Rosta: The ba'a/se'o clump and ja'o/su'a are arguably presuppositional. For example, is a ti'e utterance a mere report of hearsay, or is it a claim, marked as being based on hearsay?