real and sustained language use. In the first two of the three Krashen et al. studies, the exposure measure included a self-rating of how much English the subject spoke each day, but as Krashen and Seliger (1976) point out, this estimate may not have been true of the entire time the S spent in the second language environment: some may have spent a fair amount of time in the United States before attempting to use English regularly. In the third study of this series, only "years in an English-speaking country" was considered. A significant number of subjects who did not use the language regularly may have affected the results for the sample. Thus, the Upshur, Mason, and Carroll subjects appear to have been involved in an intensive, daily, and often demanding second language environment. The Krashen et al. subjects may have varied much more with respect to the amount of real communicative use they make of their second language.
While the characteristics of utilized primary linguistic data (termed "intake" in recent years) have not been determined in detail,* mere "heard language" is probably insufficient input for the operation of a language acquisition device at any age. The difference between "heard language" and "intake" is emphasized in Friedlander, Jacobs, Davis, and Wetstone (1972), who examined the linguistic environment of a child who at 22 months was judged to be nearly as fluent in Spanish as she was in English. The child heard Spanish primarily from her father. This input, according to Friedlander et al., made up only 4 per cent of the child's total "heard language" but was 25 per cent of the language directed at the child. This confirms the hypothesis that the relevant primary linguistic data are those which the acquirer is actively involved with: the total linguistic environment is less important.
The result of the studies reviewed here can all be considered as consistent with hypothesis I. The Upshur, Mason, and Carroll studies are direct evidence, while the Krashen et al. series may be interpreted as showing that acquisition from the informal environment requires regular and intensive language use. Hypothesis II, however, also receives no counter-evidence from any of the studies. The correlations between years of formal study and proficiency found by Krashen et al. are reliable and are consistent with Carroll's interpretation of his data.