conclude that the varietyof grammars in human languages cannot possibly be regimented by afinite parameterization.
Culicover goes further and argues against the conception of parameter triggering in language acquisition. Chomsky's
(1981) strategy is to derive superficial differences among languages from some“deep” or“abstract” parameter.
Culicover observes that the learner has to structure the input properly in order to trigger the parameter, which then
establishes the proper grammar. Thenhe asks: Oncethechildstructures theinput properly, whatfurther workis there
for an“abstract” parameter to do? (Tomasello 2000a makes a similar argument.) Culicover therefore advocates
specifying choices among grammars in relatively superficial structural terms, not in terms of elaborate invisible
structures that the child needs to f-know in advance.
The view of grammar and of grammar acquisition urged here offers an attractive realization of the alternative
suggested by Culicover and by Ackerman and Webelhuth. We have construed rules of grammar as lexical
constructions of more or less specificity or generality, involving partial phonological, syntactic, and/or semantic
structures and containing one or more typed variables. They are learned by extracting general variables from
previously stored items.
Now, we can ask, what guides the extraction of these patterns? The poverty of the stimulus still applies: for any
collectionof input data, there are numerous ways it could bestructured intopatterns. Yetchildren prettymuch allpick
the same way. What favors this way over others? The answer in Chapter 4 was“Universal Grammar.”But we did not
specifyhowUniversal Grammar might affect choice of 1-rules.
Suppose that Universal Grammar consists of a collection of skeletal fragments of 1-rules built into lexical memory;
Ackerman and Webelhuth call them “grammatical archetypes.”And suppose that, in the course of acquisition,
alternative 1-rules are emerging simultaneously in competition (i.e. in traditional terms, the learner is faced with a
choice between two competing generalizations). If one of these emerging 1-rules inherits structure from existing
fragments of UG—in exactly the same way that an idiom inherits structure from a phrase-structure rule—then that
one will emerge more strongly (come to be activated more easily or more quickly) and extinguish the other.
Moreover, inheritance is not absolute; it tolerates partial violations, for instance in irregular verbs and idioms. Thus
under this conception, patterns that deviate to some extent from UG could be tolerated, if the primary linguistic data
demanded it. Thus we arrive at a nice approach to“markedness”phenomena: the“unmarked”case is the one
prespecified by UG, and“marked”rules deviate fro mthe un marked case qualitatively in just the way irregular verbs
deviate fro mregular for ms. In this way UG can be seen to“shape”the emerging grammar