This experiment addressed the opinion prevailing among researchers that people are poor at producing random binary sequences. Participants tried to produce sets of sequences of outcomes of imaginary coin tosses that could not be distinguished statistically from sets expected from actual coin tossing. The results generally support the conclusion that people are not very good at this task, although the distributional properties of the sets of sequences produced are qualitatively similar to those expected of sets produced by a random process. The results do not support the common finding that people consistently produce substantially more alternations and fewer repetitions than would be produced by chance, nor do they provide evidence of the pervasive operation of a gambler's fallacy manifesting itself in a tendency for an alternation to increase with the length of a preceding run.