Pattern of response to antidepressants has been proposed as a method to identify patients whose improvement is more likely due to drug vs those whose improvement on drug is more likely to be a placebo effect. It is hypothesized that those with 'true-drug initial response pattern' are most likely to benefit from continuation treatment. The relationship between acute patterns of response and subsequent placebo-controlled continuation treatment with the antidepressant mirtazapine is examined. A total of 410 outpatients were treated openly with mirtazapine for 8-12 weeks. Patients who remitted in the acute phase were randomized to continue the same dose of mirtazapine or switched to placebo. Acute phase responders were classified as 'placebo initial response pattern' (early responders and nonpersistent responders) and 'true-drug initial response pattern' (delayed and persistent responders). Of those with a 'true-drug initial response pattern,' 10/40 (25.0%) relapsed with continuation mirtazapine, and 23/41 (56.1%) relapsed when switched to placebo. The difference (31.1%) is significant. Of those with a 'placebo initial response pattern,' 5/36 (13.9%) relapsed with continuation mirtazapine, and 12/39 (30.8%) relapsed with placebo substitution. This difference (16.9%) is not statistically significant. Moreover, the relapse rate for 'true-drug initial response pattern' patients switched to placebo (56.1%) was also significantly greater than for 'placebo initial response pattern' patients switched to placebo (30.8%). It has been suggested that patients with late onset and persistence are more likely to have improved because of drug. This hypothesis gains support from this study because of the different relapse rates of 'true-drug' responders on drug and placebo. The low relapse rate for patients with an acute placebo pattern switched to placebo suggests specific drug effect played a smaller role in their initial improvement.