The acquisition of the grammar in turn has to be divided into factors learned from the primary linguistic input and
factors due totheinitialstateoftheorganism. Determiningtheprimary linguisticinputdepends onwhat actual factors
inthelanguagelearner's environmentcountas input forlearning. Is therea dependenceonspecialspeechmodeswhen
talking to babies, on prosody, on statistical regularity, on prior understanding of context, and so forth? All of these
countas aspectsoftheinput, and may reducetheshare of complexityattributed tothechild's initialstate.As observed
in section 4.6, opponents of UniversalGrammar typicallyattribute more influence to the input than proponents—and
perhaps rightly so. But, I have insisted, the input alone is nevertheless likely to be insufficient: the complexity of
grammar still leaves a substantial gap for the initial state tofill in.
Turning to the initial state, we might artificially divide it into two factors: learning strategies and prespecification of
structure.Again there is a balancetobe struck: if thestructure of grammar can be more completely prespecified, there
is less work for the learning strategy, and vice versa.
Positions intheliterature tend toclusterat thetwoextremes. On onehand, connectionist theories, likemost empiricist
approaches, want littleor no prespecified structure, leaving all thework to the learning strategies. For that matter, they
typically want to minimize the complexity of the learning strategies as well. The question I have constantly raised here
is whether such an approach is adequate to the complexity of linguistic fact—that is, whether it can yield correct
grammars.
On the other hand, Principles and Parameters Theory (Chomsky 1981) and Optimality Theory both take the
prespecification ofgrammar totheextreme.TheybothconceiveofUniversalGrammar as containing alltheprinciples
necessary for all languages. In Principles and Parameters, the differences among languages are encoded as a set of
“parameters”or switches; all the learning strategy has to do isfind the appropriate“triggers”for these switches in the
input. In Optimality Theory, the differences among languages are encoded as different rankings of the universal
constraints; all the learning theory has to do isfind readily available cues in the input for constraint ranking.
The truth undoubtedlylies betweenthese twoextremes, Pollyannaish though thismay sound. Thereare many possible
intermediate positions, one of which we will explore in Chapter 6.
A cross-cutting division of the initial state is between those aspects that belong to a cognitive specialization for
language learning and those that belong to more general faculties of the f-mind, such as sociability, ability to
conceptualize the world, rhythmic analysis of temporal signals, and the ability to form hierarchical structures. Again,
theoretical linguists acknowledge the use of