In an investigation of the effects of stem structure and direction set on sentence completion responses, 160 undergraduate males were randomly assigned to one of four conditions: Rotter ISB with "feelings" instructions, Rotter ISB with "speed" instructions, Sacks SCT with "feelings" instructions, Sacks SCT with "speed" instructions. Eight clinical judges derived hypotheses from the completed protocols; two judges evaluated each protocol. Interjudge reliability was measured by Pearson product-moment correlations and percents of agreement. There was a significant effect for amount of stem structure; the structured Sacks SCT stems yielded more clinical hypotheses. Structured stems also elicited significantly more feeling words. No significant effect was found for instructional set, nor was there a significant relationship between stem structure or instruction set and the numbers of words in completions.