This study evaluates a working hypothesis held by a number of second-language researchers that second-language learners progress in their acquisition of target language structure by observing regularities in their input, implicitly forming hypotheses, testing those hypotheses against further input, and revising some while dropping others as a result of their fit with the input. The study considers four possible sources for variability encountered in learner language: (1) the situation(s) in which particular forms are produced; (2) the learner's encoding and decoding capabilities; (3) the target language itself, whereby systematicity at one level may leave variability at another; and (4) the analyst and procedural decisions that may effect the perception of variability. The study concludes that the analyst's task must match in complexity that of the language learner if the fullness of the learner's accomplishments is to be captured accurately.

In this paper empirical data are presented that demonstrate significant differences in the manner in which two commonly used experimental tasks evaluate developing linguistic competence in adult second language (L2) learners. Results indicate that production tests principally evaluate a learner's developing structural competence in the L2. On the other hand, comprehension tests provide a less direct measure of structural competence and are significantly influenced by pragmatic context.

Fifty-one adult Spanish speakers at three levels of ESL ability were tested in both their elicited imitation (production) and act-out (comprehension) of complex sentences that were structurally identical. The stimuli varied in terms of the pre- and post-posing of a subordinate adverbial when clause and in terms of the direction of pronoun anaphora (forward and backward). Results from the production test indicate a significant effect of directionality, i.e., a preference for forward pronouns in post-posed clauses at the intermediate level. Results from the comprehension test did not show a significant directionality constraint at any level. They did, however, demonstrate a significant effect due to the use of a biasing pragmatic lead. Such findings are comparable to those found in early first language acquisition (Lust, Loveland, & Kornet, 1980). Implications of these findings for experimental methodology are discussed.

This paper investigates the theoretical and applied domains of pragmatic failure. With respect to theory, it further clarifies pragmatic failure both in native and non-native speech, and with respect to the applied domain compares request realizations of native and non-native speakers in terms of length of utterance. In discussing the results of this comparison, a number of hypotheses are put forward concerning the ways in which deviation from native norms of utterance length might be a potential cause for pragmatic failure.

The data were collected within the CCSARP (Cross-Cultural Speech Act Realization Pattern) project involving seven different languages and dialects (Blum-Kulka & Olshtain, 1984). The data were collected from both native and non-native speakers of each of the languages. The analysis of responses across several languages revealed a systematic difference in length of utterance used to realize speech acts by non-native speakers as compared to native speakers. The types of pragmatic failure that might be linked to being a non-native speaker are examined, thus continuing a line of research focusing on the pragmatic aspects of interlanguage (Blum-Kulka, 1982; Thomas, 1983; Edmondson et al., 1984).

This paper reports the results of a study which investigated (a) the separate effects that differences in learners' contact can have on various aspects of proficiency, and (b) the combined effects that differences in informal contact and instructional variation can have on improvement in proficiency. Forty-eight adult learners from three intermediate-level ESL classes participated in the study.

To measure differences in learners' informal contact, a language contact questionnaire was administered. The results revealed both quantitative and qualitative differences in learners' out-of-class contact with the second language. When these differences were examined in relation to learners' performance on seven proficiency measures, correlational analysis revealed that while learners' performance on some measures was related to differences in amount of contact, it was related on other measures to differences in type of contact.

To determine whether differences in contact interacted with instructional variation to produce differences in improvement in proficiency, learners' pre-and post-test scores were examined in relation to contact and class in an analysis of co-variance. The results indicated that learners' informal contact interacted with differences in instruction to produce variation in improvement on two proficiency measures.

The findings are discussed in relation to the need for more class-room-centered research to investigate both the separate and combined effects of learner and instructional variables on second language proficiency.

The free written recall task is used in much current L2 reading research (Bernhardt, 1983; Carrell, 1983, 1984a, 1984b; Connor, 1984; Lee & Ballman, in press). An examination of the work using this task as a measure of comprehension reveals variation in design and findings: measurement of quantity recalled, method of textual analysis, the language used to recall, the level of the learners, and pre-reading instructions given. The purpose of this paper is: (1) to elaborate the differences between studies using a free written recall task to measure L2 reading comprehension; and (2) to report the results of experimentation that attempts to test the significance of the language of recall (native versus target) and the effect of knowing, prior to reading, that the subjects would be required to recall the passage. Results showed that passage recall is significantly better when done in the subjects' native language than in the target language. Also, a significant interaction was found for prereading instructions and level of the learner. Implications for research design are discussed.