Message-Id: <5371.9207042005@scott.cogsci.ed.ac.uk> From: Ivan A Derzhanski Date: Sat, 4 Jul 92 21:05:15 BST To: lojbab@grebyn.com Subject: Ancient History: adjuncts and complements Cc: cowan@snark.thyrsus.com, nsn@mullian.ee.mu.oz.au Content-Length: 3748 Lines: 89 > Date: Tue, 30 Jun 92 23:19:57 -0400 > From: lojbab@com.grebyn (Logical Language Group) > > Well, I've finally gone back to mail reading, and am finishing up April %^) You're doing significantly better than I am. :-) > Adjuncts and complements were discussed by Ivan and Nick in several > postings there. They seemed to agree that complements were like the > fixed places and adjuncts like the optional places. Exactly. > On reading this through carefully, I'm not sure that I buy it. In fact, > after writing this, I'm sure the theory is faulty for English as well. The theory is intended to be part of Universal Grammar, that is, to hold for all human languages. It is not a theory of English, though English evidence is often used for supporting it. > I suspect that, in Lojban at least, the complement/adjunct distinction > MUST be tied to a sumti, and not to the main bridi, in order to be realized. Er ... I beg your pardon? > In most NL's of course, the predicate is a predication about > the subject; i.e. x1. Thus the problem may truly be different for > Lojban in that the predicate stands completely apart syntactically from > the subject. Not really. In much of current syntactic work the subject is just another complement, and as such is opposed to the adjuncts. > I thus conclude that NLs tell us little or nothing about > defined vs. optional places on the main bridi. I don't see where this conclusion came from. > Lojban of course does not allow preposing of a complement in a question, > since we don't ask questions that way. No comparison is possible on > this point. True. Different tests are appropriate for different languages. > It might be that ALL non-subject objects (i.e. sumti), in a Lojban > predication, whether BAI-tagged or otherwise, are inherently adjuncts, > but that Lojban has simple no nounal transform of these. I still prefer to think that the BAI-tagged ones are adjuncts and the rest are complements, because the complements are the ones whose semantic interpretation is entirely determined by the predicate. > But, quoting Nick: > >Noun with complement can be turned into verb with complement: "student > >of physics" = "she studies physics". Can't do with adjuncts (single > >predication). I'm not sure what Nick means by this; it sounds counter-intuitive to me. > But by any definition equating "be" places to complements then violates > another 'rule': > Due to the recursion of N' -> N' ADJ, you can pile on many adjuncts > in arbitrary order, but only one complement: > "the student with long arms with short hair" > "the student with short hair with long arms" > *"the student of Physics of Chemistry" > > "le klama be le zarci bei ti" is a "complement" that violates this. Not in the least. {le zarci} is x2, {ti} is x3. > Indeed, I would question the truth of this rule for English in those > rare 3-place predicates of the language "the donor of the book to > charity" is two complements by all of Nick's rules except the one > forbidding two complements, <...> There is no rule which forbids two complements. What is forbidden is two complements with the same ro^le. In your case, the theme (the book) is a direct object, and the goal (charity) is an indirect object. The syntax of Lojban confirms the distinction just fine (you can have {broda [be] vi da [bei] vi de}, but not {broda [be] fi da [bei] fi de}. > But finally, the bit about nouns controlling kinds of complements but > not adjuncts would appear to be violated by Lojban, which can in theory > stick any possible BAI place into a 'noun phrase' <...> That's exactly what not controlling adjuncts means: the possibility of sticking any of them in.