[YG Conlang Archives] > [engelang group] > messages [Date Index] [Thread Index] >
Martin Bays, On 15/09/2012 03:59:
* Thursday, 2012-09-13 at 14:30 +0100 - And Rosta<and.rosta@hidden.email>:I am open to discussing how I as a user would interpret stuff. I just want to maintain -- and have my interlocutors maintain -- a clear conceptual distinction between that and the formal specification of the language.OK, I think I see what you mean. You want to consider defining the language as a two-step process - first, some syntactic translation into a "logical form", and then giving a semantics (formal or informal) for that. Since you doubt the plausibility/desirability of giving a *formal* semantics for the logical form, you consider only the first syntactic operation to be part of the formal specification of the language. Is that right? It does seem reasonable. (Although I think the clearest way to give an informal semantics might be to try to give a formal semantics!)
That's right. The reason I doubt the desirability of the formal semantics is that there is a risk of that leading to closing down the range of meanings that the language can express.
A user could be of any of these three sorts: A. The user construes things as types that have subtypes, using l to point to the type, and s& r to quantify over subtypes. Every subtype has its own subtypes so there are uncountably infinitely many types. B. The user takes it that the world against which truthconditions are evaluated contains individuals and categories, these are different things, and it is clear which is which. The user uses l- to point to individuals and s- and r- to quantify over members of categories. C. The user takes it that the world against which truthconditions are evaluated contains individuals and categories, these are different things, but it is not clear which is which -- is an apparent individual in fact different similar-looking individuals? Are apparently different individuals in fact the same individual? This user would avoid referring to individuals, instead so as to be on the safe side always quantifying over members of (possibly singleton) categories with -s and r- and avoiding l-.I'll leave this here for now. I don't understand it yet, but I'm hoping to after some of the issues above are cleared up. One thing which might help, though: could you confirm my suspicion (based on vague recollection of what you've said in the context of lojban) that in all cases, you would have e.g. types of cats of various levels and (in B and C) individual cats all satisfying mlt_?In Type-A ontology, there aren't levels, only types, and types satisfy mlt_.Let's stick with A. So with "la mlta Pa", we're forced by the presupposition to interpret it in a (quantifier-domain restricting) situation containing just one of the various cat-types. Depending on the context (including the hints given by P), this could naturally be "cats in general", or "black cats in general", or "this cat here", or "the noon-yesterday-stage of this cat here", or various other things. Is this right?
Yes, but it's not the main part of the story, for me. The UoD may be one which the only cat is Tiddles, e.g. a story world, but (now that I've given the matter more thought) I don't see the UoD as shrinking as appropriate to ensure that there is only one. That is, if Tiddles is sitting on the windowsill, and I say "la mlta li [windowsill]i [sitting on]aki", I don't think I am temporarily shrinking the UoD to contain only Tiddles; rather, I think I'm performing some sort of singularization -- myopic singularization, massification, whatever suits the context -- on all catdom in the UoD. The evidence is that if Tiddles is on the windowsill and Felix is by the fire, this can be described as "la mlta li je [windowsill]i [fireplace]i [sat at]aki".
In particular, there's no way to get at other cat-types in P, not even at the proper subtypes of a, because we have to interpret Pa in a situation in which a is the only cat-type. Correct?
Ah, I see an inadvertent ambiguity in "only one". I meant "exactly one", as in "the cardinality of the set of cats (individuated by a given body of criteria) is One", rather than "by any body of individuating criteria, the cardinality of the set of cats is nothing but One". So -- with different individuating criteria -- you can get at the subtypes of a.
I'm still not sure I understand what you mean by "using [...] s& r to quantify over subtypes", though.
I meant that "sa/ra mlta" is interpreted as quantifying over subtypes of the type Mlt.
Do you mean something like using "re mneka Pe" in the scope of a binding of 'a' to a cat-type (which would have to have been done with something other than "la mlta"!)?
It wasn't what I'd meant, but "la mlta re mneka bcde" works.
In Types B& C, I'd have thought that only individuals and not categories satisfy mlt_, but I'm open to being corrected on that -- if you tell me that in Type-B/C ontology, various ways of generalizing over cats are themselves cats, even tho they're not individual cats, I will listen with interest and (perhaps bemused) credence.This is a side-issue, but: there's no a priori reason that mlt_ shouldn't refer to both individual cats and types of cat, much like (as xorxes has been known to be fond of pointing out) in English we can say "the black cat is a cat" and mean "all black cats are cats". I would however expect the dictionary entry to indicate whether or not this was the case (if there aren't language-universal rules for deciding), which is perhaps difficult if you want to cater to A-types.
Only certain ontologies recognize a distinction between types and individuals. So the loglang, IMO, needs either to be ontologically "secular", recognizing no specific ontology, or else to be ontologically "ecumentical", with means of catering to many diffeent incompatible ontologies. --And.