This study assessed the effects on second language learning of variations in homework written feedback that either suppressed student errors or made them salient. Eighty students from two college Spanish courses were randomly assigned to treatment groups for a six-week period. Performance data were collected before and after treatment, as well as from homework during treatment. Analysis of variance blocking on pretest revealed significant achievement increases for treated students independent of course membership. Planned comparisons pooled across courses showed achievement was consistently superior under salient error conditions and in particular with constructively critical feedback. Results support the notion that the written performance of students learning a second language can benefit most by focusing on homework errors in a motivationally favorable manner. Relevance of the findings for instructional theory and second language teaching are discussed.As foreign language teaching has gradually come under the influence of theory and research on language acquisition, teachers have come to expect that students learning a second language will make many errors. A recent review on error correction in foreign language learning (Hendrickson 1978) suggested that (a) systematic analyses of errors can provide useful insights into the process of language acquisition because errors are signals of the student's progress and success in language learning; (b) teachers should avoid correction strategies that might embarrass or frustrate students; and (c) the attitudes, motivation, personality, and past language learning history of the students must be taken into account when teachers provide feedback to students. Other writers (e.g. Page 1958, Clark 1975, Clifford 1979 have noted further that pointing cut errors need not be considered as punishment but rather as information feedback to motivate the learner. The signal aspect of the reinforcement rather than the reward is critical (see also Gardner and Lambert 1972).This view rests on the assumption that the more information students have about their responses the better they will understand why they make mistakes. Such understanding should aid the students' ability to correct mistakes and increase achievement (Kulhavy 1977). This array of propositions was examined