Semantics – for boys and girls
The schools in semantic studies are numerous, and the dialogue between them is not always easy, in spite of the fact that the metalanguage may be similar. Most structural theories of language distinguish lexical and grammatical (syntactic) semantics, but I don’t think this distinction is specific of the structural approach and should depend on the fact, whether a linguist treats syntactic structures as bilateral units having both their form and meaning, or not. Some radical structuralists try to dissolve grammar in the lexicon and claim that grammatical semantics is nothing more than a gross extrapolation of lexical constraints imposed on the use of this or that item: if these constraints are similar for different lexical items, one may – they argue – speak of ‘lexical rules’. Such a trend is – as far I can judge – typical, say, for Moscow lexical school and its affiliates. On the other hand, another group of linguists draw the line between semantic taxons, as e.g. predication of Action, State, Activity, Result etc. basing on the distribution of morphosyntactic markers: in their mind, an active phrase can never be synonymous to a passive phrase, a nominative-accusative phrase can never be synonymic to an impersonal phrase etc. etc. Under such an approach semantic categories largely duplicates the distribution of syntactic and morphological categories. Meanwhile, there is recent trend in generative syntax towards claiming that syntax and semantics are really two sides of the same module of language and that semantic rules should actually be applied before (a great deal of) syntactic rules, not afterwards. Theories of this type are not wholly formalized yet, but when they will be formalized, they will probably arrive at the same two extremes, where two competing structural schools are.
I have never considered myself an accomplished structuralist and thought that the above mentioned polemic extremes do not concern me directly. My own conception of meaning is that syntactic (or: non-lexical) semantics exists and is not a projection of lexical and is not just an implementation of lexical concepts. Non-lexical semantics, in my view, should concern such problems as, for instance predicate classification a la Vendler or a la Aristoteles, quantification issues, subject-predicate and predicate-argument relations and the mapping of predicate types and predicate- argument relations onto syntactic structure.
Last year I was embarrassed by a distinguished colleague of mine, who told me that syntactic explanations, that do not fall back upon semantic categories, are boys’ syntax. Did she rather say ‘Boy! …Syntax‘ ? If my first guess was right, I would like to know, what ‘girl’s syntax’ or ‘mature syntax’ is about: probably it is kinda theory, that tends to explain every formal feature with some underlying semantic mechanism and claims that syntax and semantics are actually two sides of one and the same module of language.
I understand, why Minimalist Grammars need functional projections, theta-roles, case assigning, feature checking and feature licensing. I realise that feature-driven movement into Spec positions is a very useful invention, which allows for formalizing all these things. I can even at times believe (when I wake up in a happy mood) that aspectual projection AspP is located at a given place in clausal structure and follow the advantages of such an analysis. I can eagerly assume that Topic and Focus are maximal projections with a semantic load of their own; I have indeed offered a variant of this theory myself, see the page DIALOGUE 2008. I am not puzzled, when I hear that NOM is actually the same as TENSE. But I fail to understand, why e.g. Agent, Patient or Experiencer must be analyzed in terms of syntactic trees. Considerations like ‘it is most economic‘ do not touch me a slightest bit. First of all, nobody has demonstrated, how such a Syn/Sem module really works, since all accounts are semi-formal at their best, not formal. For the second, the notion of economy (which is dear to my ill heart) should be better restricted with isolated modules in syntax responsible for a strictly limited number of operations on trees and subtrees, otherwise it is just a fashionable word.
For those naive souls, who still insist on the existence of Lexicon, I can add that I believe both in lexical and in grammatical semantics. Contrary to radical proponents of lexicalist semantics, I think that grammatical semantics is basing on its own principles and its rules are not just a gross generalization of constraints on the use of this or that lexical item, as some structuralists have claimed. At the same time, I don’t think that semantic categorization simply duplicates the choice of formal means and that sentences Syntax bores me and I am tired of syntax convey different categorial meanings just because they represent ‘different constructions’. If ‘being tired of Syntax’ is a Stage-level predicate, and ‘to bore smb.’ is not, we need a more developed semantic apparatus, which would not stick to trivial features only, such as e.g. +/-underlying animacy, nom/obl. case marking on the subject etc. etc.
If you think this is highly speculative without examples, try the embedded page SEMANTIC TEST, and you will get facts from two languages.
Lexical vs non-lexical semantics
This perspective is discussed in some of my publications focused on Russian grammar and semantic classes. The papers were for the most part published in the series ‘Logical Analysis of Natural Language’ edited by Nina D. Arutyunova.
This paper discusses Stage-level predicates (предикаты состояния, statives, in the Russian tradition) and predicates of evaluation. I argue that we need to distinguish at least three categories of inactive one-place nominal predicates and may not classify them under one and the same cover term ‘statives’:ss-and-se1
The following paper discusses the links between modal operators and lexical items grouped in synonymic lists or ‘semantic fields’.