This exploratory field study examined Bandura's (1977) self-efficacy model to determine how well efficacy expectations predicted writing performance, and whether cognitive (deep processing) and affective (anxiety) variables were related to efficacy expectations. Other variables assessed with respect to efficacy and writing performance were race, sex, an English entrance exam (ACT) score, and locus of control. Subjects were college freshmen enrolled in introductory writing courses. The major findings were that (a) efficacy expectations predicted writing on phase 1 (beginning of writing course) data, but not phase 2 (end of course); (b) depth of processing, locus of control, and anxiety were related in varying degrees to amount of efficacy and to the accuracy of efficacy predictions of writing; (c) subjects significantly overestimated their writing performance, the discrepancy being even larger at phase 2. These results provide partial support for the construct validity of self-efficacy and suggest that cognitive and affective variables influence efficacy.