The relative predictive abilities of two indices of test anxiety were investigated. The galvanic skin response (GSR) and the Test Anxiety Scale for Children (TASC) were used as predictor variables for IQ and achievement test performance. The results of multiple linear regression analysis indicated that neither the TASC nor the GSR, combined over four achievement content area subtests, (administered in random order), were highly consistent predictors of the test performance. That the TASC failed to correlate, in a systematic way, with either GSR levels or with test scores casts some doubt on its validity as a measure of test anxiety. It is suggested that the effects of such variables as social desirability response, acquiescence, and defensiveness of response patterns, and two unexpected findings, the effects of test order on GSR arousal, and increased arousal across tests from the beginning to the end of testing be investigated. (PR)