and often incorporated them into his speech. This is similar to what Clark found in her study of Adam. Wagner-Gough hypothesizes that patterns do not directly evolve into creative rule-governed language: "It is quite clear that there is no transfer between some imitations and subsequent free speech patterns. For example, a learner may say 'My name is Homer' in one breath and 'He Fred' in another, the former being a memorized pattern and the latter the learner's own rule" (p. 71). Thus, Wagner-Gough supports position 2.

The most complete study of routines and patterns in child second language acquisition is L. Fillmore's doctoral dissertation (Fillmore, 1976), an exhaustive examination of the speech produced by five acquirers of English as a second language in an English-speaking kindergarten. Unlike Wagner-Gough, Fillmore comes out strongly for position 3:

... the strategy of acquiring formulaic speech is central to the learning of language: indeed, it is this step that puts the learner in a position to perform the analysis which is prerequisite to acquisition. The formulas... constituted the linguistic material on which a large part of their (the children's) analytical activities could be carried out.... Once they were in the learner's repertory, they became familiar, and therefore could be compared with other utterances in the repertory as well as with those produced by other speakers (p. 640).

All the children studied by Fillmore used routines and patterns very early and very heavily: "The most striking similarity among the spontaneous speech records of the five children was the acquisition and use of formulaic expressions. All five quickly acquired repertoires of expression which they knew how to use more or less appropriately, and put them to immediate and frequent use (p. 640). Including only the clearest cases of formulaic expressions, Fillmore calculated that their use ranged from 52 per cent to 100 per cent of the total number of utterances at the early stages, down to a low of 37 percent in the most advanced performer at the end of the year. Two children, in fact, remained nearly completely dependent on routines and patterns even at the end of the year.

Routines and patterns evolved into creative language in a manner not unlike that reported by R. Clark (1974). Larger units were broken into smaller units, routines became patterns, and parts of patterns were "freed" to recombine with other parts of patterns. This break-up

93