This paper compares the foregrounding strategies of native speakers of English and advanced learners of ESL. Fifteen native speakers and thirty-five advanced learners produced on-line (play-by-play) descriptions of the unfolding action in an animated videotape. A methodology for the quantitative analysis of discourse production is described which permits the explicit identification of foreground-background information and its interaction with both native speaker and interlanguage syntaxes. Results show that: (1) Native speakers and learners exhibit one universal discourse strategy in retreating to a more pragmatic mode of communication under the severe communicative stress of on-line description, as revealed through the systematic loss of the coding relations ordinarily holding between foreground-background information and clause independence/tense-aspect; (2) Learners also exhibit a learner-specific general communication strategy in which significant events are described but nonsignificant events avoided.

While the recent explosion of work on L2 discourse has provided researchers with a mass of data on the development of communicative abilities, we find that most of this work is predicated on a single (perhaps even curious) assumption: that all discourse, of which L2 discourse is taken as a subcategory, is intended by its speakers to be informative to some interlocutor (i.e., transmits a message via an acoustic or graphic conduit) and that the production of discourse entails the formulation of strategies on the part of speakers in order to maximize this transmission of data. That discourse is not communicative (i.e., sending and receiving information) by nature has, in fact, been long known, but overlooked, especially in L2 research. Malinowski pointed out as early as 1935 that:

Language is an activity the function of which is not an expression of thought or communication of ideas … the neglect of the obvious has often been fatal to the development of scientific thought. The false conception of language as a means of transferring ideas from the head of the speaker to that of the listener has, in my opinion, largely vitiated the philological approach to language (Malinowski, 1935, 9).

This paper investigates the question of whether the Wave Model of linguistic change can furnish a valid idealization for the patterning of variation in interlanguage systems and, thereby, a scale for the determination of their degree of naturalness. Using cross-sectional data from 162 French-speaking adult learners, the paper analyzes the variability in the evolution of the English possessive determiners HIS and HER. The main findings are as follows:

1. The Wave Model furnishes a valid idealization for acquisitional change.

2. The developmental continuum reveals that L2, acquisition appears to involve a greater scope of variability. Change may be initiated in a more marked environment at a point when the rule change process is not yet fully completed in the less marked environment. Nonetheless, the proportion of learners showing variable application of the mature rule in only one environment at a time as opposed to those with variable application in both is more than 2:1.

3. Less natural systems evidence rule blockage and/or compartmentalization of the marked environment; rule conflict between noncontiguous parts of the rule, and violation of markedness constraints.

4. Premature adoption of gender marking on the possessive determiners or, more generally, failure to approach the target language rule with an unmarked, initial hypothesis is one important source of unnaturalness in rule variability.

The data analysis leads up to a number of other issues in interlanguage research such as inter-learner variation, the ‘acquisition-learning’ distinction, and the relationship between the Simple Codes Hypothesis and unsuccessful L2, acquisition.

This paper reports on the influence of formal instruction on L2 acquisition in an instructional experiment with child learners. The main tendency of the findings is that a structure can only be learned under instruction if the learner's interlanguage has already reached a stage one step prior to the acquisition of the structure to be taught.

My hypothesis for an explanation suggests that the teachability of L2 structures is constrained by the same processing restrictions that determine the developmental sequences of natural L2 acquisition: since the processing procedures of each stage build upon the procedures of the preceding stage there is no way to leave out a stage of the developmental sequence by the means of formal teaching. Following such a processing capacity approach I reject the assumption that the constraints on teachability can be explained on the basis of linguistic input.

Gundel and Tarone (1981) studied the acquisition of pronouns by Chinese and Spanish speaking adults learning English. On the basis of this study it was proposed that when L1 and L2 differ in the relative order of object pronoun and verb, the acquisition of direct object pronouns proceeds in three stages: (1) direct transfer of the native language pattern; (2) zero anaphora, i.e. no overt form at all, in direct object position; (3) the correct L2 pattern. The present paper presents further evidence for this proposal, and hence for a hypothesis-testing view of second language acquisition, on the basis of longitudinal data from English speaking children learning French in the Toronto French Immersion Program.