## Design Motivated by a relatively high rate of suspicion (37%) and a desire to be as honest with participants as possible, we changed the threat manipulation from explicit feedback (always fabricated, and by definition inconsistent with participants’ gender identity about half the time) to a recollection task in which participants will be asked to describe a time they behaved in a feminine or masculine manner. We also changed the writing task from advice to one's self in the pre-manipulation task, and advice to an imagined other in the post-manipulation task, to a dialogue between a woman and man (half of the conversation written before the manipulation and the remaining half finished after). This is meant to (1) make the pre- and post-measure writing tasks more consistent, (2) test the language style effect in a different type of writing task, and (3) gather behavioral data on how participants depict a typical woman and man, in terms of their respective language patterns. The tasks in this study will also explore different indirect effects of gender threat on social processes; where Study 1 tried to get at an interpersonal, advice-giving processes, Study 2 will explore social representations. Other than changing the gender threat manipulation and the writing task, procedures are the same as in Study 1, including number and source of participants. ## Measures To further explore the voting effects observed in Study 1, we expanded both the general political orientation and voting intention items. To the voting intention items, we specifically added questions about proposed laws that are partisan but not particularly relevant to gender roles (e.g., lowering taxes). For efficiency, we excluded measures that were not as relevant to the condition or main outcomes in Study 1 (the gender and feminism components of the sexism measure, and the systemizing-empathizing quotient). The advice confidence items and gender norms measures (CFNI and CMNI) were also dropped as they were parts of Study 1 that are not in Study 2 (advice giving and gender feedback respectively). Finally, we added a few identification with gender items to test whether the gender manipulation would have a stronger effect on people who identify more strongly with their gender (along the same lines as [Terry & Hogg, 1996][3]). ## Hypotheses Following from Study 1, we expect that (1) women in the unthreatening condition will use a less feminine style and less feminine content in their second than their first writing task, and (2) that men in the threatening condition will endorse more conservative policies. We expect the latter effect to be present for all voting intention items but stronger for voting on policies that are relevant to gender roles, such as bathroom laws or gay marriage. [3]: https://dx.doi.org/10.1177/0146167296228002