OBJECTIVE To examine the effect of cognitive-behavioral therapy for insomnia (CBT-I) on the underreporting of sleep relative to objective measurement, a common occurrence among individuals with insomnia. METHODS Pre-treatment and post-treatment self-report measures of sleep were compared with those obtained from home-based polysomnography (PSG) in 60 adults (mean age = 69.17; 42 women) with comorbid insomnia. The self-report data were published previously in a randomized controlled trial demonstrating the efficacy of CBT-I compared with a placebo treatment. RESULTS Self-report measures significantly underestimated sleep at pre-treatment and CBT-I led to a correction in this discrepancy. There were no significant changes in PSG after CBT-I. Path analysis showed that an increase in an objective proxy measure of sleep quality (i.e., decreased stage N1 sleep) after CBT-I was significantly related to improvements in self-report of sleep, with full mediation by reductions in discrepancy. CONCLUSIONS This is the first CBT-I outcome study to analyze discrepancy changes and demonstrate that these changes account for a significant portion of self-report outcome. In addition, improved sleep quality as measured by a decrease in percentage of stage N1 sleep following treatment may be one mechanism that explains why sleep estimation is more accurate following CBT-I.