Abstract
Self‐reported attitude data over the last decade have often been used as predictors of attained second language proficiency and for other purposes. Three possible sources of non‐random but extraneous vairance in self‐reported attitude data are considered: self‐flattery, response set, and the approval motive. It is suggested that some of the variance in verbal intelligence is common to variance in first and second language proficiency, some of which in turn may be common to the kinds of non‐random sources of variance in self‐reported data which are listed above. It is demonstrated that it is possible that self‐reported attitude measures may be surreptitious measures of verbal intelligence and/or language proficiency. To the extent that extraneous non‐random sources of variance in self‐reported attitude measures can be shown to exist, the measures must be assumed to be non‐valid.