A meta-analysis assessing the magnitude of sleep changes from baseline in placebo-treated (psychological and pharmacological placebo) and untreated groups issued from independent trials was conducted. Comparisons were then performed to assess if the magnitude of sleep changes in the placebo control groups were significantly different than those of the untreated group. Medline, PsychInfo and Current Contents databases (1990-2004) were searched for primary insomnia treatment studies using a randomized controlled parallel-group design. Effect sizes were computed for each end-point variable based on subjective (patient-reported) and polysomnographic measures. Thirty-four studies (n = 1392 subjects) met inclusion criteria; twenty-three used a pharmacological placebo (n = 1163), four used a psychological placebo (n = 81), and seven used a waitlist condition (n = 148). Between-group comparisons were performed using a random effects model analysis. Significant pre-post changes were observed in the pharmacological placebo condition on several sleep parameters, both on objectively and subjectively measured outcomes [objective and subjective sleep onset latency (SOL) and total sleep time (TST) and subjective wake after sleep onset]. Although a tendency was observed for objective SOL, only the changes on subjective SOL and TST in the pharmacological placebo condition were significantly different from the corresponding changes in the untreated group. No differences were significant for the psychological placebo groups. Although the present findings suggest that sleep may significantly change in response to a pharmacological placebo, conclusions remain tentative because of possible confounds that may arise when comparing groups issued from different trials. Further research directly comparing placebo groups and untreated groups from the same randomized trials remains necessary.