In the present study, we examined the stability of one measure of emotion, the emotion-modulated acoustic startle response, in an undergraduate sample. Using the acoustic startle paradigm on two different occasions, we measured stability of affective modulation of the startle response during and following the presentation of pictures selected to be of positive, negative, or neutral emotional valence. The two assessments were separated by 4 weeks. Two groups of subjects were compared: one group that viewed the same pictures at each assessment and a second group that viewed different pictures at the second assessment. We found that viewing different pictures at two assessments separated by 4 weeks yielded moderate stability of the emotion modulation of startle magnitude, whereas subjects who viewed the same pictures at both assessments showed poor stability. Furthermore, this difference was due to the stability of responses to high versus low arousal pictures, not to differences in valence.