gadri report, aug 2003

From Lojban
Jump to: navigation, search
  • Nick Nicholas reporting.
    • This is the half-remembered outcome of five months of mayhem on jboske, September 2002 through January 2003. The participants in that debate will be surprised to see this proposal; I am pretty much giving them what they asked for, without many of the incredible kludges I was coming up with to express the required meanings within baseline constrains. Basically, while it is just barely possible to do so within the existing baseline, the resulting constructions are so cumbersome that noone would use them - and the Lojban gadri system is broken enough that no cumbersomeness should impede its repair, especially given that gadri are compulsory, and And's attempts at repair (e.g. lo'edu'u) immediately strike the uninvolved Lojbanist as awkward.
    • On the other hand, I am committed to keep the disruption of the existing gadri system to a minimum, and in fact am currently suggesting only two new cmavo, and no major realignment of existing gadri.
    • This is, of course, a summary, and the elaborate background formalisms I was devising - the Ontology in particular - are nowhere to be seen here; I'm not sure I even need to include it in the final proposal or any supporting paedagogy, though it did help clarify my thoughts. Conversely, I am presuming that the reader has some basic notion of the issues that came up on jboske, so the audience for this is not the BPFK in general yet. To explain the background issues involved, and intension in particular, at a level approachable to all Lojbanists, essentially requires me to write two chapters of Lojban for Intermediates. And I will do so for the final proposal. I just won't do it this sunny morning.
    • There are several issues in here that need to go to poll; I indicate my preferences, but a poll will need to finalise things. Since the solution is conservative, and preserves much of the ambiguity of the existing system, with extra optional disambiguation, it will likely not be the only proposed solution; there will likely be a continuation of the Excellent Solution vs. Kludgesome Solution debate on jboske. I think my solution is the one with the best chance of preserving as much backwards compatibility as we can afford, while fixing things. (That things need fixing is beyond dispute, by the way.) It will not likely preserve backward compatibility with usage, as opposed to the prescription, but xod's Road To Damascus conversion on this point is I believe correct: there are some points of Lojban grammar where usage has been so confused, anglocentric and inconsistent that it is not worth preserving compatibility with usage. Often enough, the deviation of usage from the baseline prescription is telling us something is wrong with the prescription. In this instance, I think it is telling us Lojbanists really have no idea how gadri work, in the absence of good paedagogy - which I am committed to repairing. (Repairing gadri is in fact the major point of the forthcoming Lojban for Intermediates - the rest of the logical convolutions of Lojban are what would be left for Lojban for Uebergeeks.) But the final decision on that is up to the commission.

Problem 1. Collective vs. Stuff.

Lojban has two essential categories distinguished by its gadri: the individual and the mass. Though this distinction exists in English, it is not as prominent as the number distinction, and the result has been that the Lojban mass, which overlaps with the plural, has been used often as a surrogate plural. This is a fault of teaching and transference, rather than the prescription, and needs repair.

There is a second problem with the Lojban mass/individual distinction that has emerged. loi is ambiguous, and conflates two different kinds of 'mass': the collective and the stuff. The collective is a group of individuals, considered together as an entity: there is no question that the individuals' entity is preserved within the collective. (The difference between the collective and a plurality of individuals -- the couple and the two people - is that there is some salient property characteristic not of any one individual, but of the collective as a whole; the classic Lojban example is people carrying a piano. The property need not belong to the totality of the group, but it does need to emerge beyond the individuals in the group.)

The stuff - which is what semanticists traditionally call mass - is also an entity distinct from the individual, but it is not guaranteed to consist of individual entities; it refers to the physical substance making up the entity, rather than to any distinct individuals. So the Beatles are a collective of humans, and consist of four individual humans. The Beatles' toes are stuff of humans: they are not constituted of any individual humans, but they are part(s) of what make up humans. Forty beans are a collective of beans, a "bunch of beans". The bean dip made from those forty beans is no longer a bunch of beans, since the individual beans are no longer preserved; but the stuff that the beans were made of is still there, so the bean dip counts as a mass of beans.

The stuff/collective distinction is one that English is capable of: "some steak" vs. "some steaks". Normally however it is left to grammar: a noun is assumed either to be a count noun or a mass noun, and not ambiguous. Outside of linguists' example sentences (the "universal grinder"), people aren't mass nouns in English: you don't normally "step in some human". Conversely, if mass nouns are treated as count nouns, there is the understanding that what is being counted is a portion of the stuff, rather than a true individual of the stuff: "a water" is not a molecule of water, but a portion of water of quite variable size, whereas "a human" is a bounded entity, that cannot be arbitrarily added to the way you can pour more water into a glass. (I attempted in my Ontology to set more formal rules for when something counts as an "atom" or not relative to a predicate, but we can assume common sense for now.)

Lojban is by design agnostic as to the mass/count noun distinction: any brivla ca be used to denote either a mass or an individual. Part of the stated rationale is that the decision on whether something is a mass or an individual is culture-specific, and indeed varies within the history of a given language: e.g. pease used to be a mass noun in English, before being reanalyzed as 'pea - peas'. This equivalence is notional: taken to its logical conclusion, it would license lo nanmu being interpreted as "some human flesh" (a portion of humanity), since that interpretation is consistent with mass predicates - cf. lo djacu. The interpretation is not licit, and Lojban needs an explicit statement of gismu defaults, or at the least some way of disambiguating the candidate senses.

The Lojban expression of masses with loi is also agnostic as to the distinction between collective and stuff. This differentiation is bound with the foregoing: the default mass for a count noun is a collective, and for a mass noun is stuff, but outside of gismu defaults Lojban currently has no good way of disambiguating the two. Even the gismu for mass - gunma, jorne, kampu - are ambiguous.

There are several possible disambiguations within the confines of the current baseline; none are particularly succinct or intuitive in my opinion. The radical proposal has been made of having distinct gadri for collectives and stuff, and doing away with their conflation in Lojban as masses. I believe this will cause excessive disruption in the existing language, and since context usually supplies the necessary disambiguation, it is unnecessary. Instead, I propose that LAhE be used to disambiguate: loi remains ambiguous between stuff and collective, while LAhE1 loi and LAhE2 loi disambiguate into stuff and collective.

The remaining issue is which LAhE to use. There is already a LAhE in place corresponding to loi, lu'o. We could keep lu'o ambiguous between stuff and collective, and adopt two new LAhE to do the disambiguation. In my opinion this is unnecessary, and lu'o should unambiguously say that the referent is stuff, while a new LAhE states that the referent is collective. It will be necessary to survey existing usage of lu'o, but I suspect this has been minimal.

I believe some machinery should also be put in place to indicate the cardinality of collectives, outside of using mei - e.g. LAhE2 loi re lo remna. I have made extensive such proposals on jboske, though they would need to be revisited.

Sets

I had ignored sets in my travails; Nick and John 2003-02-03 alludes to the proper handling of sets as distinct from collectives, but that is a matter not as urgent as the others I'm discussing here.

Extensions and Intensions

A definition first. Extensionally defined notions are those for which you can enumerate the referents. The default in Lojban, and in logical understandings of language in general, is for predications to be understood extensionally: that is what prenexes are for. So "there exists X such that a(X)" means that there is a specific, particular X of which a() holds. Which means things like, X can be distinguished from all other X: it is a particular individual. If it's the kind of thing that can have a name, it makes sense for it to have a name. You may not happen to know their identity, but there is no dispute that they have one. So if you say "I talked to a doctor", it makes sense to interpret that sentence by substituting "a doctor" with a particular doctor's name: "I talked to Dr Fred" is a possible interpretation of "I talked to a doctor".

For intensionally defined notions, on the other hand, enumeration is impossible: the sentence does not make sense if you enumerate the referents, because it deals with an underlying abstract notion which the referents would match. So if you say "I'm looking for a doctor -- any doctor", the "a doctor" cannot have a distinct identity: it does not make sense to say "I'm looking for a doctor, any doctor - and his name is Dr Fred". The sentence "I'm looking for a doctor, any doctor" is not true of any particular doctor X, so it cannot sensibly have a prenex attached to it. Rather, it is an assertion to do with doctors in general.

Problem 2. Unique.

The controversies on jboske started with pc grumpily discussing the issue of how to translate "I like chocolate". The problem with the existing means of saying it, using loi, is that they all involve implicit quantifiers, and thereby are all extensionally defined. If it is true of piro loi cakla, then you are claiming you like every single piece of chocolate in existence - which cannot be true. If it is true of pisu'o loi cakla, then it is true even if you like only Aztec chocolate and chili powder, but not chocolate in general, which is too weak. We could make it be pirau loi cakla, but even that is extensionally defined: we would expect the speaker to be able to identify explicitly the quantities of chocolate they like, and submit them to some sort of quantitative test to identify whether they are a raumei. That way madness lies.

The proper solution to this problem is the same as that for generic plurals, such as "I like dogs" and "Dogs make good pets". And the solution we have alighted on is to take the generic plural, "dogs", not to be an extensionally defined set of dogs (Fido, and Rover, and Azor, but not Fifi...), but rather an abstraction away from all dogs to a Mr Dog, such that the differences between individual dogs don't matter: you don't get into the game of extensionally defining Mr Dog, because as far as you're concerned, all dogs are the same - at least, as far as the statements "I like dogs" or "dogs make good pets" are concerned. This is not to say that there cannot be exceptions to the statement: of course there can, and you can still dislike Fifi. But "I like dogs" is not a statement of Fido, Rover, and Azor ("there exist at least 3 dogs, F, R, and A, such that I like them.") Rather, it is a statement about dog-kind, the same way "I like people" is a statement about humankind.

There is an analogy to be made with individuals, and "stages" of individuals - snapshots of individuals at a particular time. The predicate "X wears nappies" is true of Nick-at-age-2 and not true of Nick-at-age-32. So in one sense, Nick-at-age-2 and Nick-at-age-32 are distinct referents: something true of one is not necessarily true of the other. But ordinarily, we don't treat them as distinct: we treat them as manifestations of the same underlying individual, Mr Nick, and we ignore any properties that differentiate between them. Likewise, the generic statement "Dogs make good pets" ignores the distinction between Fido, Rover, and Azor; it even ignores the exceptions like Fifi, just as "Nick lives in Melbourne" ignores Nick-early-July-2003 who was in Sydney. (And the statement "Nick lives in Melbourne" would be true even in early July 2003.) We aren't extensionally counting stages of Nick; we are asserting that the stages add up to the one individual, and we make assertions of that individual.

This construct turns out to be very handy once you think of it. If I say "Nick and And ate fish and chips", I'm not asserting that they eat the same meal, extensionally defined: the quantity of fish and chips in And's newspaper is distinct to that in mine. I am asserting, though, that the distinction between the two quantities is irrelevant to my purposes: as far as I'm concerned, And's meal and my meal are the same thing. We are both eating Mr Fish & Chips. The same can be said for, say, the New York Times: lo karni peme'e la niuiorktaimys. is a specific newspaper that you physically pick up at the newsagents, while Mr New York Times, the generalisation from all the specific newspapers, is what people usually refer to when they speak of the paper. (There are further generalisations to be made, across issues and so on.)

And although Lojbanists have been slow to realise it, the "Mr" intension - the Kind, or the Unique, is the proper treatment of event abstractions. le nu limna is *not* swimming: it is a particular swim, at a particular time and place. If you like swimming, just as if you like chocolate, you are not interested in enumerating the particular swims you like. Rather you are asserting that for your purposes, all swims are the same thing - Mr Swim - and *that* is what you like. The same goes for du'u, and any number of other Lojban abstractions. As we will see below, it is as good a candidate as any for handling intensions in general in Lojban.

Since "Mr" involves turning off extensions - i.e. quantifiers - I spent a long time on jboske trying to express it using existing gadri and non-quantifiers like tu'o in quantifier place. Unfortunately tu'o ended up being claimed for more than one such function, and the results were too prolix to be acceptable. I say this for the benefit of conservatives, concerned that I have not given the existing resources of the language a proper run before going outside them: believe me, I have, and it won't fly. It's not that the current language can't do it, it's that it looks silly when it does, and this will end up encouraging Lojbanists to persist in their simple, easy to understand, wrong alternatives.

There are two gadri in Lojban which already come close to "Mr", and several months were spent on jboske trying to work out whether they already do the job. My opinion is that they do not. One is the gadri which Lojban lore has already defined as doing the work of "Mr", loi. I won't revisit the original JCB definition right now; in fact, I'm refusing to review anything while writing this, because it was that painful. My recollection, though, is that it was somewhat confused. A more important problem with loi is that it becomes ambiguous between an intensionally and an extensionally defined notion: loi fi'uro finprcarka can be a particular shark, or Mr Any One Shark. The point of "Mr" is that there be no enumeration; but loi must admit extensional use, in that loi has implicit fractional quantifiers, and is used in extensional contexts (you can talk to loi mikce, and mean the particular doctors Dr Fred and Dr Jane, rather than the abstraction Dr Any Doctor). Rather than continue to overload loi, it is less confusing all round to move this concept somewhere else.

The second gadri is lo'e. There has been an undercurrent of using lo'e as an intensional article in Lojban, though the main proponent of this is Jorge. CLL lo'e is *not* an intensional article, but a statement of typicality: lo'e does generate an intension, but not the same intension as "Mr". Our alternatives are to reassign "Mr" to lo'e, and dispense with CLL lo'e - which would leave the well-defined le'e orphaned - or to leave CLL lo'e alone. Jorge has advocated the former, on the grounds that CLL lo'e is nowhere near as useful as the use he would put lo'e to. I, of course, took the opposing view; and And has been extolling the "Nicolaic lo'e" (for which the credit properly goes to John), which was my way of differentiating between CLL lo'e and "Mr": lo'e makes claims that are generic and characteristic of the referent, to the extent that they can be taken as exemplars of the entity - i.e. "Lions live in Africa". But claims that are not a characteristic property of the referent, that one would not think of as definitional to our mental model of the referent, should be made with "Mr" instead. "Lions live in Africa" is a claim in some way definitional of lions, or at least characteristic of them. "I like lions", or "I study lions", is not. Even though the referent is in a sense the same - Lion-kind - I think it dangerous to conflate the two kinds of claim. The syntax of the predications involved - x1 vs x2 position of lo'e - might seem a way out of this; but of course that is untenable in Lojban: lo'e cifno cu citka lo'e mirlrantelope is a claim characteristic of both lion and antelope.

Moreover, it is crucial to be able to generalise "Mr" beyond individuals, to the other kinds of entity Lojban can discuss: collectives, and stuff. (See the jboske thread on "xodium".) I discuss the need for this immediately below. And has proposed as a solution an entire Unique series of gadri, encompassing stuff, collective, and individual; this of course may well not prove enough, since it would also have to encompass veridical/nonveridical/name. I believe the proper solution is, once again, LAhE: LAhE3 loi marjrxodiumu, LAhE3 lo cifno, LAhE3 loi re lo remna (Couple-kind). It should be obvious by now why I would want such a LAhE3 to be as short as possible, and why I favour a monosyllable for the job - lau if it is available, xVV if not.

Problem 3. Substance.

The last month of the jboske discussion was spent with xod in particular arguing for a distinct treatment of statements about Substance (Water is..., Salt is...), and me unable to understand why that was not simply a statement about piro loi djacu, piro loi silna - i.e. why statements about substances are not simply statements about stuff. I must admit, my counterproposals at the time were not particularly compelling.

While xod still wants a distinct gadri for substances, I latch on to one of the statements I read in a formal semantics paper I read for hints on stuff vs. collectives: while statements about substances are tricky to resolve logically, they seem best handled as generic statements about stuff. So a statement like "Water is cool" is no different to a statement like "Dogs make good pets" or "Munich is pretty". "Munich is pretty" generalises over all time-slices of Munich, and glosses over the exceptions (say 1945), because all time-slices of Munich add up to the same thing. The paper explicitly said that statements about Water or Gold are akin to statements about Munich: you abstract away from all the physical quantities of Water (pisu'o loi djacu} to Mr Water (LAhE3 loi djacu), and make your generic statement about that. Some of those statements, of course, are already covered by lo'e, but not necessarily all, and it is not clear how lo'e spans the individual/mass divide, given that Lojban predicates are by definition ambiguous between the two.

Problem 4. Intensionals.

As you'll have guessed by now, the Unique/Kind type - the "Mr" - is the start to the solution of how to handle intensionals in Lojban. It is only the start, because intensionals are messy; and there is a problem, because until now Lojban has been inconsistent in its approach to intensionals.

The label of intensionals encompasses two distinct issues, which the literature (I say humbly, and probably erroneously) seems to conflate. We may refuse to enumerate -- to handle something extensionally - because the intensional context resists enumeration as a sensible interpretation; this is the case for "I'm looking for a Doctor -- Any Doctor". We may also refuse to handle something extensionally - to put in front of it a prenex saying "there exists x such that" - because the referent may or may not exist; that is the case for the sentence notorious from Montague Grammar, "I'm looking for a unicorn", where any interpretation involving a prenex "there exists a unicorn x such that..." must be dismissed as nonsensical, if unicorns don't exist. (If you don't like unicorns - and there are reasons not to, because people did think they existed - then plug in something more obvious, like "six-letter gismu".)

Since intensions (Mr X) need to be talked about even when they don't exist, one of the issues that needs to be resolved is whether, if there is no such thing as lo broda, there is still such a thing as LAhE3 lo broda - if unicorns don't exist, does Mr Unicorn exist? This matters, because ordinary Lojban predications still involve prenexes - "there exists X such that..." And's position is that they do, so you look for LAhE3 lo pavyseljirna. John's position is that they don't, and if lo pavyseljirna is meaningless, so is LAhE3 lo pavyseljirna. The way to get from non-existent entities to notions we can talk about, we agreed (and I have trouble reconstructing, but this is on Nick and John 2003-02-03) is through event abstractions - in particular the near-equivalents du'u and ka. As John has it, LAhE3 lo pavyseljirna does not exist, but le ka pavyseljirna does. That is was John means by reification: taking an abstract concept (referring to something that may or may not exist), and turning it into something concrete, which you can quantify over and insert into normal Lojban predications.

There are four major kinds of intensional context - predicates whose referents cannot be thought of extensionally (there exists a...) but intensionally. As we go down the list, you'll see that they shift from being intensional because they resist enumeration, to being intensional because the referents need not exist. The contexts are:

1. Wanting. Joey wants a doctor, any doctor (doctors exist). Joey wants a unicorn, the one called Fred (unicorns don't exist; Joey may or may not know that).

2. Seeking. Joey's looking for a doctor, any doctor (doctors exist). Joey's looking for a unicorn, the one called Fred (unicorns don't exist, but Joey probably doesn't know that).

3. Depicting. Joey's drawing a doctor (doctors exist, but the particular one Joey's drawing doesn't -- Joey's making her up. So the depiction isn't a property of any enumerable specific doctor). Joey's drawing a unicorn. (ditto)

4. Fearing. Joey fears the Oogy-Boogy-Man. (Joey's enumerating him alright, but he doesn't exist. Joey doesn't know that though.)

There's an added trait of intensions, definitional to them, which is that if X and Y are the same - but Joey doesn't know that - then Joey can believe something of X but not of Y in an intensional context. If Joey doesn't know Clark Kent is Superman, Joey can fear Clark Kent but not Superman - and it makes sense for me to say so, even though I do know they're the same. You can't do this in an extensional context: I can't say "Joey talked to Superman but not to Clark Kent", if I know they're the same person, because talking is extensional. "Joey realised that he talked to Superman but not Clark Kent" does work, though.

Now, the first solution to intensions is to insert an extra prenex . This properly resolves the existence problem, rather than the enumeration problem. "You want to talk to a unicorn" cannot be expressed as pa pavyseljirna goi da zo'u: do djica lenu do tavla da - because that asserts the unicorn really does exist. But the unicorn exists *only according to you*. We can convey that by placing the existential prenex inside the embedded abstraction: do djica lenu pa pavyseljirna goi da zo'u: do tavla da means something like "you want a situation in which there exists a unicorn, and you talk to it". By placing the claim of the existence of the unicorn inside the embedded abstraction, you're confining the claim of its existence to that abstraction, and to the world it portrays; it does not extend to the 'real' world.

This is a solution, and in fact Lojban makes it its official solution, by having prenexes *by default* nest inside embedded abstractions. It doesn't necessarily address the refusal to enumerate; indeed, the very fact it still contains a prenex means it does enumerate. We might be able to conventionally turn off enumeration/the assignment of identity if the prenex is embedded; we might need to fall back on something else.

The second context has had a different Lojban solution offered: not the internal prenex, but John's reification: mi sisku leka pavyseljirna, understood as "I'm looking for a holder of the property of unicornhood, whether or not they exist" - because even if lo pavyseljirna doesn't exist, leka pavyseljirna does. This solution has been blown up: as I was reading an introduction to Montague Grammar, I found the author saying "of course, you can't do that, because what if you're looking for two unicorns". This is bad news, although we can sort of work around it with mi sisku leka pavyseljirna remei. But what if you're looking for some kryptonite? A better solution would be to allow the full range of gadri in here - collectives, stuff, non-veridicals - but to still have the reification making sure that you can speak sensibly of unicornhood without unicorns. The tentative solution is mi sisku lo seka pavyseljirna - but that would seem to imply that the property holder of unicornhood still exists. (This is why I counterproposed lo jai ka pavyseljirna, but that probably has the same problem.) Some combination of "Mr" and reification may yet save the day, or John may need to give up his extensionalism.

The other two contexts are not as obvious and have been ignored. I've bumped into a paper online that derides the attempts to deal with Depicting and Fearing in terms of propositionalism - addressing intensions by embedding an abstraction and having its prenex take care of the possibly non-existent referent. Its contention was that there was no obvious choice of predicate for such an embedded proposition to take. Lojban co'e and bu'a mean we may be able to ignore that. The other problem is that we aren't used to thinking of depicting and fearing in terms of embedded propositions - having to say mi te pixra lenu lo mikce cu co'e (or mi te pixra tu'a lo mikce), instead of mi te pixra lo mikce, is very unnatural. That is why Montague Grammar allows you to insert prenexes *between* sumti (or at least so I've heard reported.) It's not clear to me that what is drawn is Mr Doctor, either. John is more sanguine about this, though, in that what is drawn is lo simlu be lo ka co'e - and even if the lo co'e, say the unicorn, doesn't exist, le simlu does.

The directions in which solutions will move are:

  • Where the identity of the referent is immaterial - the prototypical intensions of wanting and searching - the Kind ("Mr") should be involved, although we *might* conventionally agree that if propositionalism is used, the Kind is understood.
  • Where the existence of the referent is an open issue, a reification is prudent; and since the existence may well usually be open, the reification should be a default strategy. Since natural languages don't bother with such paraphenalia, though, the reification devices have to be unobtrusive enough not to be rejected by Lojbanists: I do not think mi te pixra tu'a lo mikce meets that criterion, though perversely a contrast between mi te pixra lo mikce (a portrait) and mi te pixra lo simlu be leka mikce (a fictional doctor) just might.
  • While Propositionalism is time-honoured for intensions of wanting, it is clearly not the most general solution to intension, and we should have other devices available.
  • Intensions of seeking are not resolved by leka alone, and need a little more work for a general solution, which does not fall afoul of non-existence but which also allows one to seek different kinds of sumti.

Problem 5. De dicto/de re

This is a related problem to extension and intension, which Lojban hasn't really dealt with. The extension of "the president of the U.S." right now is Dubya. The intension of the expression, in Montague terms, is a function of the world and the time: it's a function mapping time to president - i.e. 1992-2000 => Bubba, 2001-2004 => Dubya.

Now, the utterance "in 1992, the president was a republican", made in 2003, is ambiguous. If "the president" is extensionally understood to mean "the president right now", Dubya, it is true. This interpretation is "de re" (of the object): you're using the label "president" to get to a concrete referent, Dubya, and "the president" means nothing but Dubya. If the "the president" is intensionally understood to mean "the president at that time", the interpretation is "de dicto" (of the saying): its interpretation varies in time, and does not refer to one fixed object: it is just a label. (Just as the intension is a function mapping time to president, not the president himself). Understood de dicto, "in 1992, the president..." refers to Bubba, and the statement is false.

So which of the two interpretations does ca li 1992 le merjatna cu cmima la republikana have? Does ca li 1992 le merjatna cu cmima la republikana have one interpretation and le merjatna ca li 1992 cu cmima la republikana the other? (That would be Lojbanic: the former would be Bubba and the latter Dubya, because what goes first into the prenex restricts what comes after.) If so, John has indicated CLL already needs some correcting. I do not think this is a problem the Mr intension can fix.

Discussion

As noted elsewhere, the exact scope of buried (non-prenex) items - tense and negation and modals and quantifiers and God knows what all else - is going to continue to cause headaches in Lojban as they do in English. Since prenexing everything is not a possible sell, we need some unambiguous rules that are actually adhered to. At present the nearest thing to current practice and this goal is to say that all free-floating relevant critters go to the front in the left-to-right order of occurrence in the text and the selbri-bound pack (tense, negation, etc.) go to the extreme left (front) end of the prenex as a block in the order given. This is pretty much what we have, though CLL is not always clear about some parts of this and at least one point seems to deny some aspects of the last clause: pu na is apparently equated with na pu, rather than being particular and universal of the same matrix, and ca pu is not distinguished from pu ca. Of course, the problem of the end of scope remains.

It might be useful if someone reminded us of why this change of gadri is going on. I have not seen anything in the various proposals that is not already covered (albeit somewhat overlongly) in the present system, which is btw a system and ties in with other systems in the language, unlike the various alternatives so far.

  • pc, don't make me flame you. You started this whole sorry mess with your "I like chocolate" phrase, and how Lojban couldn't deal with it; without a Kind article, how do *you* propose to say it in baseline Lojban? I spent four painful months trying to accomodate baseline Lojban to the demands that came up, and couldn't without untenable cumbersomeness; if you can, then let's hear it. - nick, testily.
    • Well, looking through what I have on file it does not appear that I introduced "I like chocolate" - the first comment on it is a reply to someone else, though I can't reconstruct who. Nothing in this pile of things here and now seems to me to be about the chocolate problem - or at least the connection is long buried in the back story. I don't think I ever mentioned Kind - in what I gather is the present sense; but if I did and did not equate it with piro loi, then I apologize. As for chocolate, I seem to have stuck pretty closely to loias I understood it then (and, I think, now), though I did suggest we needed a new gadri for it since loiwas getting thoroughly mucked up with Stuff (which may have been a better way to go, but even that doesn't quite work). I did not at that point equate stuff with the constitutive group of the constituent parts of whatever the stuff is the stuff of. I might now swing back toward the other path I looked at, making "like" etc. into intensional contexts, though it does not yet seem to help much. But, come to that, I don't see how any of the rest of this helps either and it seems to have wandered so far afield that the good stuff (most of it, after all) that we had has been washed away and the replacements are higgledy-piggledy, don't cover basic situations, and don't have any obvious motivation. Maybe all I really want is to see what motivates these moves and how they are meant to help (and how the old stuff is preserved).

(Repairing gadri is in fact the major point of the forthcoming Lojban for Intermediates - the rest of the logical convolutions of Lojban are what would be left for Lojban for Uebergeeks.)

  • Personally I'd think it's better to plunge into the "Ubergeek" end and tidy up all the language semantics once and for all, than to start from the "Beginner"/"Intermediate" end i.e. create a mess and then try to deal with it as we move along. With the latter approach, it's hard to come up with a sane way of representing "I'm looking for a doctor -- any doctor" (which in symbolic logic will have to be contorted into something like LooksFor(Me, lambda x . Doctor(x] -- we're not talking about a specific object, but a non-specific object satisfying a specific predicate) -- User:tk1@
    • That is, of course, the way Lojban deals with sisku, but it deals with other similar cases by making reference to events rather than things: "I want a nail" becomes "I want it to be the case that there is a nail I have" or something like that. And there are probably other solution for other cases ("Get me a nail" seems to rely on the intensional quality of the imperative, for example). So one of the goals here is to find a single solution to cover all cases (although the fact that they are all alike in English does not seem a good reason to think they are all alike logically).pc
      • Judging from the number of sides (erm, proposals) in this gadri war, it seems the main problem isn't to find a single solution, but to decide which solution to use. :-( (To add more fuel to the fire, e'u ma'a vimcu ro gadri...) -- User:tk1@

Nick, many thanks for bringing up these - at least, for me! - highly complicated issues and making efforts to explaining the stuff in a more "paedagogical" way than it's usually done by others. I'm since long - and still am - struggling with this matter and would like to throw in my poor two cents: irrealis-realis vs. indistinct-distinct? and appreciate any help... --.aulun.