which he had learned by imitation. Hatch's comment neatly summarizes the situation: "Quite clearly two separate and very distinct strategies were running side by side. After week 12 it became increasingly difficult to separate out imitations since Paul's rule stages moved so fast that he quickly caught up with the language as it was spoken by the children in the playground" (p. 31).

The picture Hatch describes for Paul resembles, in reverse, what one may see in recovery patterns in aphasia. Alajouanine (1956) notes that when propositional speech returns, "fixed phrases" may disappear. The automatic speech is immune to the ungrammaticality of the aphasic's developing language.

The relationship of analytic and gestalt speech in Paul is again that predicted by position 2: the two modes are independent and the analytic mode eventually predominates, with gestalt speech primarily serving only as a short-cut, a pragmatic tool to allow social interaction with a minimum of linguistic competence.

Hatch suggests a reason why second language acquirers may use more routines and patterns than first language acquirers. She emphasizes the second language performer's greater capacity to remember longer utterances: "The person (L2 acquirer) brings with him a great capacity to create language by rule formation. At the same time he is capable of storing, repeating, and remembering large chunks of language via imitation. He can repeat them for use in an appropriate situation. While he is still at the two-word stage in rule formation, he can recall and use longer imitated sentences" (p. 33)

In another case history, Hakuta (1974) reported on the linguistic development of Uguisu, a Japanese-speaking 5-year-old acquiring English as a second language in informal situations in the United States. Hakuta reports that he found evidence for "a strategy of learning on the surface structure level: learning through rote memorization of segments of speech without knowledge of the internal structure of those speech segments (p. 287).

Hakuta's study is mostly concerned with patterns (as opposed to routines): "segments of sentences which operate in conjunction with a movable component, such as the insertion of a noun phrase or a verb phrase", and the evidence he provides for the existence of patterns in his corpus is quite similar to that provided by Brown (1973) in his

91