The research reported here was derived from the hypothesis that hyperventilation contributes to the decrement in performance observed in test-anxious students. From this point of view, students identified as test-anxious would be expected to hyperventilate to a greater extent than non-test-anxious students when confronted with the stress of testing. The experiment reported here tested this hypothesis by continuous capnographic monitoring of end-tidal CO2 and respiration frequency of 16 high- and 16 low-test-anxious boys and girls (ages 12-14 years) before and during tests of math and word-recall memory under conditions of high- and low-stress (i.e. 'strong' motivational instruction versus 'weak' motivational instructions). Consistent with predictions, high test-anxious students displayed lower levels of end-tidal CO2 (under the high-stress condition) and faster respiration frequencies than low test-anxious students. Both high- and low-test-anxious students scored higher on the math test under high-stress conditions, but differences between recall scores were not significant. Collateral data revealed a positive relationship between scores on the Nijmegen Hyperventilation Questionnaire and the Revised Suinn Test Anxiety Behavior Scale, and a negative relationship between the questionnaire scores (self reports of frequency of symptoms of hypocapnia) and drop in level of end-tidal CO2 during testing, i.e. high-test-anxiety group reported a greater frequency of symptoms of hyperventilation and a larger drop in level of end-tidal CO2 during testing than low-test-anxiety group.