I sought out that study [1]. They seemed to control for anxiety levels in the participants, so the results are still interesting. However, they also refer to some prior art:

> Our findings initially appear different from those of some other studies. Fyhri and Phillips (2013; Phillips et al., 2011) found that risk taking in downhill bicycling, measured through riding speed, did not simply increase when a helmet was worn; rather, the people who normally cycled with a helmet took fewer risks when riding without one. Why did the participants in Fyhri and Phillips’s study who were not habitual helmet users not react to wearing a helmet with increased risk taking, as our experiment might suggest they would? Clearly more work is needed to definitively pin down all the mechanisms here, but for now, we speculate that the difference might be related to considerable variations between the two studies’ procedures. Fyhri and Phillips greatly emphasized the physicality of their task (“to increase the difference in measures between the helmet-on and -off conditions, all participants were instructed to cycle using one-hand in both conditions”; p. 60), which provides a direct link between the action (bicycling) and the condition (helmet wearing) that was absent in our study. Moreover, that study used a repeated measures design, in which participants were aware they were riding a bicycle both with and without a helmet. This could have meant that behavior changed through mechanisms different from those seen here, where participants took part only in one condition and were not aware of any manipulation, nor even that they were specifically wearing a safety device.

So, in the past other studies have concluded that when the helmet has an _actual_ potential impact on safety, they do not induce greater risk-taking behavior. This fits right in with my personal experience, where I don't even consider that fact that I'm wearing my helmet during a ride.

1. https://journals.sagepub.com/doi/full/10.1177/09567976156207...