Most economic decisions are embedded in a specific social context. In many such contexts, individual choices are influenced by their observability due to underlying social norms and social image concerns. This study investigates the impact of choices being observed, compared to anonymity of choices, on risk taking in a laboratory experiment. I relate participants' investments in a risky asset directly to social norms for risk taking that are elicited in an incentivized procedure. I find that risk taking ist not affected by the choice being observed by a matched participant. Nor do investments follow elicited norms for risk taking more closely when observed. This holds when considering males and females separately. However, I provide strong evidence for gender-specific norms in risk taking. While these explain part of the existing gender gap in risk taking, males still "overshoot" by investing more than the norm dictates. This is particularly true for males being matched with a female participant.