This investigation tested the hypothesis of humor effects on test anxiety to improve test performance. A pretest-posttest control group design was employed to determine differences between humorous and serious versions of the same test content. One graduate biostatistics course of 98 students participated. Based on three independent test administrations, ANCOVAs were computed to isolate the effects of humorous directions only, humorous items only, and the combination of both on emotional/physiological and worry/cognitive anxiety symptoms and biostatistics achievement. Humorous directions had a statistically significant (p < .05) impact on constructed-response item performance for the first test (descriptive statistics), with an effect size of .43. Multiple-choice test performance correlated negatively with the two pre-anxiety subscales (r = −.46, p < .001), explaining up to 21% of the variance. The limitations of very low pre-anxiety levels and very high test performance precluded any other significant effects. The contributions of the humor technique used in the study and the value of measuring situation-specific anxiety immediately before and after a real testing condition were discussed.