Sexual thoughts are sufficient to increase testosterone (T) in women, yet erotic films are not. A key confound in past studies is autonomy in stimulus selection: women choose the content of their sexual thoughts but films have been selected by researchers. We hypothesized that self-chosen erotic films, compared to researcher-chosen erotic films, would (1) increase women's self-reported arousal, enjoyment, and identification with stimuli, and decrease negative affect; and (2) increase T. Participants (N = 116 women) were randomly assigned to a neutral documentary condition or one of three erotic film conditions: high choice (self-chosen erotica from participants' own sources), moderate choice (self-chosen erotica from films preselected by sexuality researchers), or no choice (researcher-chosen erotica). Participants provided saliva samples for T before and after viewing the film in the privacy of their homes. Compared to researcher-chosen erotica, self-chosen erotica increased self-reported arousal and enjoyment, but also unexpectedly disgust, guilt, and embarrassment. Self-chosen erotica only marginally increased identification with stimuli compared to researcher-chosen erotica. Overall, film condition did not affect T, but individual differences in identification moderated T responses: among women reporting lower levels of identification, the moderate choice condition decreased T compared to the no choice condition, but this difference was not observed among women with higher identification. These results highlight the importance of cognitive/emotional factors like identification for sexually modulated T. However, self-chosen erotica results in more ambivalent rather than unequivocally positive cognitive/emotional responses, perhaps because stigma associated with viewing erotica for women becomes more salient when choosing stimuli.