Cultural Differences in Correspondence Bias Are Systematic and Multifaceted

In previous work (Miyamoto & Kitayama, 2002), we showed that when Americans read a political essay written by someone under social constraints, they ascribe the attitude corresponding to the essay content to the essay writer even when the essay is unpersuasive and thus nondiagnostic of attitudes. However, under such conditions, Japanese do not show this correspondence bias as strongly.

Was this finding replicated in Many Labs 2 (Klein et al., 2018, this issue)? This question was not addressed in Many Labs 2. Instead, the authors assessed heterogeneity across various samples, while carrying out an exploratory comparison between WEIRD (Western, educated, industrialized, rich, and democratic) and less WEIRD samples, which yielded no difference. Aside from the fact that the measures of heterogeneity confounded between-culture heterogeneity with within-culture heterogeneity, the comparison of WEIRD and less WEIRD samples may have masked certain meaningful dimensions of cross-national variability because it neglected differences in culture, history, language, and the like. Moreover, as we had pointed out a priori, cramming numerous studies into a single 30-min session could change the effect. We worried that expecting to work on various tasks could change how participants would approach the tasks, even from the very first task.

Despite these concerns, however, it was important to use the Many Labs 2 data to test for a difference between American and Japanese samples—the focus of our original study. We first confirmed that the essay used in Many Labs 2 was equally low in perceived persuasiveness in the two groups, F(1, 1661) = 0.008, p = .927 (a requirement to test the cross-cultural difference in correspondence bias). ¹ We thus proceeded to test the expected cultural difference in correspondence bias and did find that the magnitude of the correspondence bias was significantly greater among Americans (n = 1,586) than among Japanese (n = 75), F(1, 1656) = 7.04, p = .008, d = 0.63, a result replicating the original finding.

Although this comparison of American and Japanese samples replicated the key finding from the original study, the magnitude of correspondence bias in samples from other Eastern and Western countries showed seemingly anomalous patterns. To begin to make sense of these patterns, it is important to recognize that cultural differences in correspondence bias can result from some confounding variables. In what follows, we illustrate this point by presenting the results of hierarchical linear modeling we conducted to test two potential culture-level moderators of correspondence bias. The first factor is perceived persuasiveness of the stimulus essay. In our original study, when the essay was made persuasive, we did find a strong correspondence bias in both Japan and the United States. In Many Labs 2, a stronger correspondence bias was observed in countries where the essay was perceived to be more persuasive, b = 0.91, df = 28, p = .04, which suggests the importance of equating perceived persuasiveness across cultures when comparing the magnitude of correspondence bias.

Second, correspondence bias is a bias to favor the essay writer’s stated position over the base rate in the population. Imagine that you know that there is a fair consensus of opinion on capital punishment in your country. Now someone is stating her position on capital punishment under social constraints. Do you take her statement literally or discount it in favor of the consensus in the society? We submit that, everything else being equal, the position stated under social constraint is more likely to be discounted when there is a clear consensus regarding the issue at hand than when there is none. To explore this possibility, we computed two measures of consensus for each culture included in Many Labs 2: (a) the standard deviation in the estimate of the average student’s attitude (the smaller the standard deviation, the greater the consensus) and (b) the deviation of the estimated average student’s attitude from the scale’s midpoint, labeled neither (the greater the deviation, the clearer the consensus). Both measures moderated the magnitude of correspondence bias in the predicted directions, bs = 1.49 and −0.49, dfs = 28, ps < .001, respectively. Correspondence bias is stronger where consensus is weaker, or, alternatively, it is weaker where consensus is stronger. Further, both measures indicated that consensus was stronger in the Western European and Northern European countries compared with the other countries, t(28) = 3.14, p = .004, d = 1.36, and t(28) = 5.76, p < .001, d = 2.21, respectively, and the first measure of consensus partly accounted for the difference in the magnitude of correspondence bias between the Western European and Northern European countries and the other countries, b = 1.38, df = 27, p = .002, though the second measure of consensus did not, b = −0.32, df = 27, p = .155.

In the age of globalization, psychology has remained largely “WEIRD.” We hope our Commentary will help all psychological scientists reflect on the complexity and nuances involved in cross-cultural research. Any effort must be informed with thorough analyses of each and all of the cultural contexts involved. Given this challenge, all data must be taken as a stepping-stone toward ever-higher levels of theoretical understanding of the phenomenon at issue.

Supplemental Material

MiyamotoOpenOracticesDisclosure – Supplemental material for Cultural Differences in Correspondence Bias Are Systematic and Multifaceted

Supplemental material, MiyamotoOpenOracticesDisclosure for Cultural Differences in Correspondence Bias Are Systematic and Multifaceted by Yuri Miyamoto and Shinobu Kitayama in Advances in Methods and Practices in Psychological Science

Supplemental Material

MiyamotoSummaryTables – Supplemental material for Cultural Differences in Correspondence Bias Are Systematic and Multifaceted

Supplemental material, MiyamotoSummaryTables for Cultural Differences in Correspondence Bias Are Systematic and Multifaceted by Yuri Miyamoto and Shinobu Kitayama in Advances in Methods and Practices in Psychological Science