The psychometric properties and predictive validity of the Depression Change Expectancy Scale (DCES), a modification of an expectancy scale originally developed for patients with anxiety disorders, were examined in two studies. In Study 1, the 20-item scale was administered along with a battery of questionnaires to a sample of 416 dysphoric undergraduate students and demonstrated good internal consistency. A two-factor solution most parsimoniously accounted for the variance, with one factor containing all pessimistically worded items (DCES-P) and the second containing all optimistically worded items (DCES-O). The DCES-P showed patterns of correlations with other measures of related constructs consistent with hypothesized relationships; the DCES-O showed similar, but weaker, relationships with the other measures. Multilevel modeling was used to examine the predictive utility of the DCES in a clinical sample of 63 adults (Study 2). Improved depressive symptoms (over 6 weeks) were strongly associated with optimistic expectancies but were unrelated to pessimistic expectancies for change. The DCES appears to be a promising measure of expectancies for improvement among individuals with depressive symptoms.