LAUSR

Shanks et al. (2015) challenged the evidence that various forms of decision making can be influenced by romantic/mating primes. In their comment, Sundie, Beal, Neuberg, and Kenrick (2019) question both the meta-analysis and the 8 studies Shanks et al. reported, and describe an alternative p-curve analysis that they interpret as showing that romantic priming is a genuine phenomenon. In this reply, we comment on several contradictions in Sundie et al.’s article. First, they suggest that Shanks et al.’s replication experiments yielded different results from the original studies because we failed to appreciate the contextual sensitivity of romantic priming effects, but this argument rests largely on evidence from the very studies we were unable to replicate, and a wealth of other evidence suggests that social priming effects are largely invariant across samples and settings. Second, Sundie et al. criticize the selection rule by which Shanks et al. identified relevant priming studies, but then go on to include exactly the same set of studies in their p-curve analysis. Third, they criticize Shanks et al.’s selection of statistical results from these studies and propose a much wider selection, but then acknowledge that their selection process is poorly suited to assessing publication bias and p-hacking. Fourth, we show that their p-curve analysis, far from demonstrating that this literature is unaffected by p-hacking, in fact shows the exact opposite. Sundie et al. claim that Shanks et al.’s priming manipulation was demonstrably weak, but their argument is based on a confusion between different dependent measures. We conclude that romantic priming remains unproven, and urge researchers in this field to undertake high-powered preregistered replication studies.