Attach
by
Philip A. Bralich, Ph.D. and Derek Bickerton, Ph.D.ATTACH @
Philip A. Bralich and Derek Bickerton
1, Introduction. Since Chomsky (1957) linguists have been grappling with the problem of
providing a generative theory of syntax to describe the structure of human language, This
endeavor has resulted in a proliferation of syntactic theories each with a slightly different set of
assumptions about the correct characterization of syntax. Just to name a few there is the
Government and Binding theory of Chomsky (1981, 1986), Lexical Functional Grammar of
Bresnan (1982) and Kaplan and Bresnan (1982), Categorial Grammar in its many versions as
described in Oehrle, Bach, and Wheeler (1988), Lexicase of Starosta (1988), Generalized Phrase
Structure Grammar as described in Gazdar, Klein, Pullum, and Sag (1985), and Head-driven
Phrase Structure Grammar of Pollard (1984, 1985). However, to date no one theory has
emerged as a widely accepted formulation that strictly conforms to the requirements of an
explicit, generative grammar, nor has any one theory gained wide acceptance as a truly
representative model of the human faculty of language. Furthermore, none of these theories has
resulted in framework that can be fully implemented in a computer programming language.
‘The present paper demonstrates that by beginning with a theory of phrase structure which
is rigidly limited to a step by step process of attachment of lexical items into larger words,
phrases, and sentences, it is possible to provide a theory of syntax that is simpler and more
general than current theories of syntax and which does not require deep structure, movement,
deletion, or insertion. Further, the theoretical machinery required for this theory is much more
restricted than that required by current theories, The many miscellaneous theoretical
subcomponents such as c-command, barriers, subjacency, the Empty Category Principle,
Government, Proper Government, Lexical Government, etc, are also not required in the current
formulation. Finally, all the theoretical mechanisms proposed for this theory can be easily
implemented in a variety of programming languages.
The present framework requires only one empty category and seven theoretical
mechanisms: 1) a rule of attachment called Attach «, 2) a series of selectional restrictions and
subcategorizations that constrain the application of Attach o, and determine the category, bar
level, and lexical features of the item created by Attach , 3) a principle to account for the
assignment of bar levels in intermediate and maximal structures, 4) a description of the
completion of the process of attachment, 5) a description of theta role assignment and
interpretation’ 6) the principles which describe the coindexing requirements of non-referential
expressions, and 7) the morphological principle of locality (Bralich (1991)) called the Immediate
Dominance Condition. We also claim that these seven mechanisms with minor language
specific changes in parameters are the same in all languages and constitute the full set of
theoretical mechanisms necessary to account for the syntax of the languages of the world.
Interpretation’ of theta roles refers to the process by which N2s are associated with empty categories in
order to receive a theta role