X-Digest-Num: 104 Message-ID: <44114.104.570.959273824@eGroups.com> Date: Sat, 3 Apr 1999 16:04:38 -0300 From: "Pablo Stafforini" Subject: Re: semantics X-Yahoo-Message-Num: 570 Content-Length: 6776 Lines: 157 coi rodo You are right, Robin. The division between defining and typical features is essential to the analysis of this problem. I have some points that I would like to express, but they all depend on one aspect of lojban that is not very clear to me: Are lojban words defined by the typical ENGLISH features? When you say "...before going multilingual with the gismu list (which it is high time we did) we need to think carefully about our lexicography." I perceive that you are affirming my question. But, at the same time, this would be strange, since lojban has always aimed to be a "culturally neutral" language. More to come, after this point is clarified... co'o mi'e. pablov. > From: Robin Turner > > coi rodoi > > {.o'i} long theoretical piece - delete if you're not interested in > semantics! > > This current discussion on semantics has got me onto my favourite > hooby-horse, categorisation theory. Briefly, there are two opposing > interpretations of categorisation, with a few people like Adrienne Lehrer, > Ray Jackendoff and, of course, myself, sitting (sometimes > uncomfortably) in > the middle. In the one camp, we have what we can call "classical > semantics", which assumes that complex concepts can be broken down into > simple features, which constitute the "meaning" of the word, in > the sense of > its truth conditions (which is not always the way the word is used - such > things are removed to the lowly realm of pragmatics). > > For example, the English word "woman" refers to a category WOMAN, > having the > features [+HUMAN][+FEMALE][+ADULT]. In other words, if the statements > > H(x) > F(x) > A(x) > > are all true, then > > W(x) > > is true. Similarly, for Turkish KADIN, {H(x) ^ F(x) ^ ~V(x)} => K(x) , > where V -> "is a virgin". > > In the other camp, we have the cognitivist, fuzzy, prototype-based, "all > thought is metaphorical" people - George Lakoff, Mark Johnson and > the rest. > From this point of view, WOMAN is a category resting on a > prototypical idea > of "woman", maybe with metaphorical extensions and associations. > Categories > are fuzzy, and some members are more central to the category than others - > to give Eleanor Rosch's celebrated example, a robin is more of a bird than > an ostrich. > > However, as Anna Wierzbicka has pointed out, this doesn't explain why an > ostrich is still definitely a bird, and a bat is not. We > therefore need to > be careful when dealing with, in her words, "the fashionable > prejudice that > human thinking is 'fuzzy'." Looking at WOMAN, the feature [+ADULT] is > derived from a category, ADULT, that is fuzzy (except in strictly legal > terms) so you may sometimes be unsure as to whether to call a particular > female human a "woman" or a "girl", but the other two features are pretty > unfuzzy - an adult female bird is definitely not a woman, and when Captain > Kirk says of the Enterprise, "she is a beautiful woman, and I > love her," he > is being obviously and deliberately metaphorical (in this case in order to > confuse a hostile alien). The problem is that one cannot always, or even > often, deduce the boundaries of a category from its prototype. Margaret > Thatcher is, many would claim, some way removed from the > prototype of WOMAN > ({zo'o} or even of HUMAN), but, (zo'onai} she definitely _is_ a woman, > whereas many people who possess more of the features of a > prototypical woman > are excluded from the category because of the rather obvious fact > that they > are male. > > What I have proposed is a division of features into "defining" > and "typical" > features (similar to Lehrer's "obligatory" and "optional" features). Thus > WOMAN has the defining features [+HUMAN][+FEMALE][+ADULT] and a > whole bundle > of typical features, such as [-VIRGIN][+BREASTS] etc. etc. I also divided > features into those which may change status and those which are > pretty much > invariable (e.g. [+ADULT] is a weak defining feature, since there are some > situations where it doesn't apply. > > So how does all this apply to defining gismu, lujvo etc.? I would argue > that a gismu definition gives the defining features of a category, but not > its typical features, except perhaps in cases where one of the sumtiplaces > has a default value. Thus if I say {lo ninmu} I simply mean > [+HUMAN][+FEMALE] and am not making claims about anything else. If, > however, I say {lo'e ninmu} - typical woman- or {le'e ninmu} - > stereotypical woman, that's when things get really fuzzy and culturally > specific, because I am drawing on a a load of typical features > which are not > present in the gismu definition. {lo'e ninmu} is bound to have different > meanings in different cultures, even amongst native speakers of Lojban, if > such beings ever come into existence. And of course, if I say > {le ninmu}, I > could mean just about anything, though Gricean maxims demand that it be > something pretty closely related to {lo ninmu}, if not the same thing, and > Lojban etiquette demands that I mark it with {pe'a} if it is > definitely not > female or humanoid in any way. > > When it comes down to the lexicographical business of writing > authoritative > definitions for gismu, we will (as I think Pablo said) run into serious > problems when we move outside English. For those who weren't > around during > the infamous {djuno} debate - the definition of {djuno} is: > > x1 knows fact(s) x2 (du'u) about subject x3 by epistemology x4 > > Unfortunately, while the word "know" in English normally implies that what > you know is true, the equivalent words in many other languages do > not (e.g. > in Turkish you can say "dog^ru biliyorsam" - "?if I know rightly"). This > led to a massive string on whether you could use {djuno} for > something which > is false. > > What this indicates is that before going multilingual with the gismu list > (which it is high time we did) we need to think carefully about our > lexicography. One solution would be to adopt a feature-based analysis of > the gismu involved, using features which are, as far as possible, > consistent > across cultures. An alternative would be to use the Natural > Semantic Model > proposed by Anna Wierzbicka, which aims to define terms using a limited > number of universally accepted words (I think the current total > is 90). I'm > actually pretty sceptical about Wierzbicka's view of semantics, but > lexicographically speaking, NSM makes a lot of sense. > > None of this, of course, will provide us with rock-solid definitions which > are universally applicable - this kind of thing only existed in a > pre-Wittgensteinian universe. Nevertheless, I think a bit of semantic > analysis now might save us a lot of grief later. >