The preception and orienting response (OR) reinstatement hypotheses are alternative explanations for the reduced responding to predictable as compared to unpredictable aversive stimuli. To test differential predictions from both theories, 60 subjects were presented with 30 stimuli varying in intensity (60 dB(A) vs 100 dB(A)) and predictability (constant vs variable warning) in a 2 x 2 between subject design. Impact ratings, SCR and heart rate were recorded as dependent variables. According to the preception hypothesis a steep and early decrease of responding in the predictable 100 dB(A) condition was expected, whereas according to the OR reinstatement hypothesis a slower decrease with differences between the predictable and unpredictable stimuli at both intensities was hypothesized. To control for response interference only those trials were selected for the analysis for which the interval was the same in the variable and constant warning condition. Results revealed an intensity effect for the SCRs and impact ratings, but no effect of predictability. Although for the heart rate magnitude the intensity by predictability was found in favor of preception, this result appeared to be due to differences in sensitivity between groups during the warning interval. It was concluded that neither hypothesis proved to provide a valid account for the reduced responding to predictable aversive stimuli, but that the data seemed to be most consistent with a safety signal interpretation. Time estimation was considered to be a crucial variable. It is suggested that beyond mere signalling, additional beneficial effects of predictability can be demonstrated in studies where procedures are used which make time estimation unnecessary.