Trait ascription studies have employed a number of response formats, varying the number of possible trait responses given on situation response. The proportion of trait responding that would occur with random responding matches closely the trait-responding frequency of actual subjects. It is hypothesized that subjects are conforming to demands implicit in the response format. This hypothesis was tested in a study that varied the number of both trait and situation options. Results showed significant effects of both manipulations, and trait responding was in all cases close to chance expectancy. It is concluded that the absolute frequencies of trait ascription in these studies are attributable to the methods used, and that no inferences can be drawn from them about natural tendencies in trait ascription frequency. The present results do not, however, call into question the validity of self/other differences in trait ascription.