English they spoke each day (on a scale of 1 to 4). Subjects with the same number of years spent in the country where English was spoken and the same report of speaking were considered to have the same exposure score.

Student samples differed somewhat. In Krashen and Seliger, subjects were registered in an intensive, 20 hour per week institute designed to prepare foreign students for study in American colleges. In Krashen et al., subjects were enrolled in a part-time extension program; these students were, on the average, older, and many were permanent residents or citizens of the United States. The measure of proficiency used in the first study was teacher ranking (which correlated significantly with local placement tests), and in the second study the Michigan Examination in Structure was used.

In the first study, six out of fourteen pairs of students matched for years of formal study of English were consistent with the hypotheses that more exposure meant more proficiency; that is, in only six cases did the student with more exposure show a higher ranking than his partner with less. Similarly, in the second study, more exposure was associated with a higher score in only ten out of twenty-one cases, which is consistent with the hypothesis that exposure has no consistent effect on second language proficiency, When students were matched for exposure scores, however, it appeared to be the case that more instruction did indeed mean more proficiency. In the first study, this was true of seven out of nine cases, and in the second it was true of eight of eleven cases, which in both studies was statistically significant.

Krashen, Jones, Zelinski, and Usprich (1978) arrived at similar results. Placement test scores for 115 students of English as a second language in an extension program were correlated with students' reports of years of formal study and years spent in a English-speaking country. The results (Table 1) confirm the conclusions of the studies described above: years of formal instruction reported is a better predictor of English proficiency than is time spent in an English-speaking environment. While exposure, here simply the report of years spent in an informal environment with no estimation of how much the S used the target language, was shown to have a significant effect, it accounted for relatively little of the variation in test scores.

44