Preliminary Notes on Checking Theory
Chomsky's recent work (Chomsky 1993, 1994), which he refers to as his 'Minimalist Program', incorporates for the first time a distinct morphological component, now fleshed out in considerable detail by Sells (1995). Chomsky claims that words emerge fully derived and inflected in syntax where they must be 'checked' against the functional categories at Logical Form (LF) within their 'checking domain', generally, the specifer-head relation. The checking domain is that of the functional categories of the Principles and Parameters model. Chomsky no longer claims that lexical morphemes are inserted into these positions; rather, functional positions contain only morphosyntactic features (phi-features) which are used in checking lexical output for appropriateness. At LF, the lexical content of each morpheme of the lexical head is 'checked' against the categories of the next functional projections associated with that head.
In order for the lexicon to generate derivationally and inflectionally preformed words, it must contain some sort of morphological component which knows how to attach prefixes, suffixes, infixes, and how to generate reduplicated and revowelled forms. Bound and free grammatical morphemes presumably continue to be listed as regular lexical items in the dictionary since without grammatical functions in the lexical feature inventory, there will be nothing to check a derivation at LF against. Such a lexical morphological component will grossly overgenerate, however, making a complex filter necessary.
Checking Theory is just such a filter. It is a complex filter in the sense that it will have to contain all the paradigmatic relations of a genuine MS-component, very similar to that found in LMBM. Checking theory is a redundancy model of the LMBM MS-module in that its job is to review words already generated in the lexicon and compare the affixes they already contain with the conditions on those suffixes, conditions which appear as features in the projection above it. Such a process would be essentially the same as reading L-features in the derived lexeme and I-features in the functional projections above it prior to affixation, then copying the appropriate affixes, derivational and inflectional, onto the stem for the first time, all in one process. Checking theory thus concedes an autonomous morphological theory like LMBM. Furthermore, it distinguishes lexemes from grammatical morphemes and distinguishes affixation from the conditions on affixation, a kind of Separation Hypothesis.
Checking Theory remains to be fully specified as of this writing so that a detailed critique is not possible. However, there do seem to be several non-trivial potential problems with this approach which render it less appealing than LMBM.
The Minimalist approach posits inflectional morphology in the lexicon. Matthews (1972), Anderson (1982, 1992), Aronoff (1994), Beard (1995) have provided a wealth of arguments against such an organization of language. To the extent that productive inflection is a property of syntax, Checking Theory runs against the grain of Chomsky's own modularism.
As it currently stands, Checking Theory seems to require two MS-modules, one in the lexicon and one at LF, to account for the same phenomena that lexeme-base morphologies describe with one. Fully inflected words are derived in the lexicon. This will require either a crude morphological component capable of null morphology, reduplication, Semitic revowelling, and the like, while ignoring functions and ordering, or a fully equipped morphological module incorporated into the lexicon itself. In the former instance, two components are required to do the work of one. In the latter, a fully fledged morphological component in the lexicon would make Checking Theory redundant, since all inflected words would emerge from the lexicon well-formed.
The major fault in this approach, however, is the fact that the conditions on affixation are situated after the operations that they condition. Adding affixes to a stem prior to contact with their conditions implies, for example, that speakers add the Agreement suffix -s to verb stems for phonological reasons and that its association with [3rdSg] is an entailment of the phonological act. However, this cannot be the case; speakers select Agreement morphemes for reasons that only emerge in syntax and use the suffix -s to express those reasons. There are no empirical or theoretical grounds for avoiding the conclusion that grammatical functions condition morphological realization.
If the lexicon contains an impoverished model of morphology, which randomly assembles fully inflected words, it would have to belong to a language organ incapable of learning from its own errors, one which has no means of recognizing the economic loss in gross overgeneration and filtering out erroneous derivations. It is more likely that the brain induces inflectional rules as current morphological theory assumes. Of course, filters are a weak theoretical device on independent grounds, since empirical evidence of them is difficult if not impossible to find.
These problems are not trivial and it is difficult to see how they may be overcome without assuming a position very close to that of LMBM. Halle and Marantz (1993) and Noyer, for example, have recently moved specifically in this direction. Nor are the problems mentioned here the only ones facing Checking Theory; see Speas (1991) for others. Moreover, the detail provided by Sells in his article in Linguistic Inquiry does not address these problems adequately.
|LMBM Contents||Word Syntax||Linguistic Fun||Other Sites||Dictionaries||Grammars|