fundamentally different mental representation than other kinds of language.

Routines and Patterns in First Language Acquisition

R. Brown (1973), in his study of first language acquisition, noted that some of his subjects' sentences were memorized wholes and patterns. He hypothesized that prefabricated routines in children were the result of very high input frequency of a structure that was, at that time, beyond the child's linguistic maturational level. We cannot improve on Brown and Hanlon's (1970, pp. 50-51) description of this phenomenon (see also Cazden, 1972, p. 110).

The parents of Adam, Eve, and Sarah did produce certain WH-questions at a very high rate in a period when children did not understand the structure of WH-questions. What happened then? The children learned to produce the two most frequently repeated WH-questions, What's that? and What doing? on roughly appropriate occasions. Their performance had the kind of rigidity that we have learned to recognize as a sign of incomprehension of structure: they did not produce, as their parent of course did, such structurally close variants as What are these? and Who's that? and What is he doing? When, much later, the children began to produce all manner of WH-questions in the pre-posed form (such as What he wants) it was interesting to note that What's that? and What are you doing? were not at first reconstructed in terms of the new analysis. If the children had generated the sentences in terms of their new rules they ought to have said What this is? and What you are doing? but instead, they, at first, persisted with the old forms.... We suggest that any form that is produced with very high frequency by parents will be somehow represented in the child's performance even if its structure is far beyond him. He will find a way to render a version of it and will also form a notion of the circumstances in which it is used. The construction will become lodged in his speech as an unassimilated fragment. Extensive use of such a fragment probably protects it, for a time, from a reanalysis when the structure relevant to it is finally learned.

Thus, routines appear to be immune to rules at first. This clearly implies that routines are part of a system that is separate from the process generating rule-governed, propositional language. It is also evidence that automatic speech does not "turn into" creative constructions. Rather, the creative construction process evolves independently. This is exactly position 2 as stated above.

Another indication that automatic speech forms are generated by a different process than creative construction is the fact that Brown's subject Adam produced many patterns (such as "It's a ______", and

86