The repetition-induced truth effect refers to a phenomenon where people rate repeated statements as more likely true than novel statements; a minority do the opposite—they reliably discount the validity of repeated statements

The truth revisited: Bayesian analysis of individual differences in the truth effect. Martin Schnuerch, Lena Nadarevic & Jeffrey N. Rouder. Psychonomic Bulletin & Review, October 26 2020. https://link.springer.com/article/10.3758/s13423-020-01814-8

Abstract: The repetition-induced truth effect refers to a phenomenon where people rate repeated statements as more likely true than novel statements. In this paper, we document qualitative individual differences in the effect. While the overwhelming majority of participants display the usual positive truth effect, a minority are the opposite—they reliably discount the validity of repeated statements, what we refer to as negative truth effect. We examine eight truth-effect data sets where individual-level data are curated. These sets are composed of 1105 individuals performing 38,904 judgments. Through Bayes factor model comparison, we show that reliable negative truth effects occur in five of the eight data sets. The negative truth effect is informative because it seems unreasonable that the mechanisms mediating the positive truth effect are the same that lead to a discounting of repeated statements’ validity. Moreover, the presence of qualitative differences motivates a different type of analysis of individual differences based on ordinal (i.e., Which sign does the effect have?) rather than metric measures. To our knowledge, this paper reports the first such reliable qualitative differences in a cognitive task.

General discussion

In this paper, we show a surprising finding. Although the truth effect is reliably obtained across many data sets, the effect itself is inconsistent across people. We are confident that in most experiments some people truly judge repeated statements as more valid than novel ones, while others truly judge them as less so. This effect is not just noise—the models indicate that this inconsistency occurs above and beyond trial-by-trial variation. What makes the finding surprising to us is that the result is in contrast to previous work with these individual-difference models. The modal result is that “everybody does”, that is, there are no qualitative individual differences in common cognitive effects such as Stroop and Flanker effects (Haaf and Rouder, 2017, 2019). In the repetition-induced truth effect, these differences exist, and they occur consistently across several data sets.

Does the presence of qualitative individual differences inform current cognitive theories of the truth effect? We think it should. A number of theoretical explanations have been proposed for the repetition-induced truth effect, for example, the recognition account (Bacon, 1979), the source-dissociation hypothesis (Arkes et al., 1991), the familiarity account (Begg et al., 1992), processing fluency (Reber & Schwarz, 1999), or the referential theory (Unkelbach & Rom, 2017). These accounts assume different underlying cognitive mechanisms, yet, they all make the same core prediction: repetition increases perceived validity. Unkelbach et al., (2019) summarize thusly: “No matter which mental processes may underlie the repetition-induced truth effect, on a functional level, repetition increases subjective truth” (p. 5). We argue, based on our analysis, that this statement is too general. In fact, we show what Davis-Stober and Regenwetter (2019) call the paradox of converging evidence: Across data sets, we find converging evidence that the statement holds on the mean level—yet, at the same time, we accumulate strong evidence that it doesn’t hold for everybody. Consequently, our results present converging evidence against theoretical positions that do not account for negative truthers.

This paper constitutes a first step by providing an answer to the fundamental question if there are qualitative individual differences in the truth effect. Having established such differences, the next step is to understand why they occur. One salient finding in this domain is that the overall truth effect can be reversed, that is, made negative, by certain experimental manipulations. Unkelbach and colleagues started with the proposition that easy-to-process statements are naturally more likely to be true (Unkelbach, 2007; Unkelbach & Stahl, 2009; see also Reber & Unkelbach, 2010; Unkelbach, 2006). In a set of creative experiments, these researchers reversed the correlation between fluency and truth, making difficult-to-read statements more likely to be true. With this correlation reversed, they observed a negative truth effect, that is, repeated statements, which are easier to process than novel statements, were now judged more likely to be false (but see Silva et al.,, 2016). One wonders if some participants have learned in their natural environment that ease-of-processing correlates with falseness, thus resulting in the observed qualitative individual differences.

Likewise, differences in memory ability might account for some of the individual differences patterns. We are most intrigued by the finding that there was evidence against individual differences in data sets where the interval between exposure and judgment lasted several days. Why would individual differences be attenuated or absent with increasing retention intervals? We suspect such a finding reflects an explicit memory-based effect (i.e., source recollection or memory for presented statements). As overall memory performance declines with increasing delay between exposure and judgment phase, these differences may diminish and, correspondingly, individual differences in the truth effect may disappear.

These post hoc explanations presented above are of course speculative. They form hypotheses to be addressed in future research. Based on our results, a promising way to examine the underlying mechanisms and possible covariates of individual differences in the truth effect is with a latent-class approach. Unlike correlational approaches, it relies on ordinal (i.e., In which direction is the effect?) rather than metric (i.e., How large is the effect?) measures. Given the strong evidence for qualitative individual differences in the majority of data sets, questions about who differs, when they differ, and why they differ are suitable to test and inform theories of the repetition-induced truth effect.