Evidence on acoustic correlates of speakers’ actual and perceived sexual orientation is inconsistent. As suggested in previous studies, an investigation of contextual factors possibly influencing the expression and perception of sexual orientation is valuable. Under which conditions do speakers tend to stereotype themselves vocally in accordance with their sexual orientation? The purpose of the present study is to examine a language-related contextual factor that may help explain the contradictory findings: Text topic. Text topic is varied on three levels by making sexual orientation salient to different degrees: a) no explicit reference to sexual orientation (doing sports; control condition), b) lesbian/gay reference (attending lesbian/gay events), and c) reference to one’s own sexual orientation (ideas about the ideal partner). For this purpose, spontaneous speech recordings of lesbian/gay and straight women and men (n = 18 per group; German native speakers) were made for each of the three conditions by responding to one interview question for each condition. In this way, we used spontaneously produced connected speech (i.e., keeping text type constant), since this is where the greatest intra- and interindividual variations are to be expected. Female and male speakers were included since a) number of studies on acoustic and perceptual correlates of sexual orientation involving gender comparisons is low and b) sexual orientation more accurately judged for men compared to women (e.g., Kachel et al., 2019). We used a person perception approach for the present study, with a listener sample rating sexual orientation of the speakers. Compared to the examination of individual acoustic parameters, this approach offers a holistic measure of the extent to which speakers acoustically stereotype themselves. To increase the likelihood of self-stereotyping, only utterances that included an ego reference were selected. We used utterances of 72 speakers (36 male, 36 female) differing in sexual orientation. Three utterances for each speaker per topic were extracted (3-5s) and randomly assigned to one of three stimulus subsets to avoid high load for raters. Hence, each subset contained just one utterance for each speaker per topic. For few female speakers, only two utterances per topic could be extracted. This resulted in 637 stimuli in total (324 male, 313 female) almost equally distributed to the three subsets (212 or 213 utterances). To focus on acoustic-phonetic parameters only (and neglect utterance content), British raters were asked to judge speakers’ sexual orientation, since the likelihood of understanding the German utterances was minimized. They were randomly assigned to one of the three subsets and rated the corresponding stimuli on a 7-point scale (1 - lesbian/gay, 7 - straight). Stimuli were blocked by speaker gender; blocks and stimuli within blocks were presented randomly. Raters were asked to rely on their gut feeling and to answer as quickly as possible. Responses could be entered via mouse click within a 5-seconds time frame after stimulus offset (otherwise, the request "Please respond faster!" appeared on the screen). Prior to the experimental trials, a practice run with 5 stimuli for female and male speakers was performed, with stimuli which were not used thereafter.