Abstract
Original article: Peters, K., & Fonseca, M. A. (2020). Truth, lies, and gossip. Psychological Science, 31, 702–714. doi:10.1177/0956797620916708
In the original article, the authors reported that they categorized 150 lies in avoidance rounds as positive. These included 40 lies stating that (a) the investor had sent a positive number of tokens and (b) the agent had returned more than 0% but less than 33% of received tokens. However, the authors now believe that these 40 lies do not easily fit into their typology, as it is not clear that these lies are more positive (or, indeed, more negative; i.e., more or less likely to lead to investment) than a true statement that the agent was avoided. They therefore reran their analyses, omitting these 40 lies. This Corrigendum is updating the values associated with these analyses as well as some passages describing them, all of which occur in the Results section. The authors feel that, if anything, omitting these lies strengthens their pattern of findings.
In the Descriptive Statistics subsection (p. 704), the second paragraph, fifth sentence, is being changed as follows, and the sentence in parentheses is being added: In avoidance rounds, when no tokens were sent, participants who told positive lies (n = 110) claimed that an average of 7.76 tokens had been sent and 13.39 had been returned, and participants who told negative lies (n = 166) claimed that an average of 8.07 tokens had been sent and 0 had been returned. (An additional 40 avoidance-round lies claimed that more than 0 but less than one third of tokens were returned; because these lies had an ambiguous valence, they were excluded from analyses that relied on this typology.)
In the Form and Function of Gossipers’ Lies subsection, the 11th and 12th paragraphs (p. 709) are being changed as follows: This analysis revealed that payoffs to trustworthy agents were significantly lower when their behavior was misrepresented than when it was either truthfully described—control: χ2(1) = 54.03, p < .001; competition: χ2(1) = 37.89, p < .001—or exaggerated—control: χ2(1) = 25.95, p < .001; competition: χ2(1) = 10.63, p = .001. Payoffs from truth and exaggeration did not differ—control: χ2(1) = 1.72, p = .190; competition: χ2(1) = 0.90, p = .344. The reverse pattern was evident for payoffs to untrustworthy agents. Here, payoffs were significantly higher when agents’ behavior was misrepresented than when it was either truthfully described—control: χ2(1) = 9.93, p = .002; competition: χ2(1) = 40.89, p < .001—or exaggerated—χ2(1) = 35.20, p < .001; competition: χ2(1) = 69.93, p < .001. Payoffs were higher under truth than exaggeration—control: χ2(1) = 25.36, p = .001; competition: χ2(1) = 14.46, p < .001. Thus, exaggeration is as effective as truth at achieving positive reciprocity and more effective at achieving negative reciprocity.
Also in the Form and Function of Gossipers’ Lies subsection, the beginning of the 13th paragraph (p. 709) is being updated as follows (note that the first and last sentences of this paragraph, reproduced here, are remaining the same): As a final step, we ran this same analysis for investor payoffs (Figs. 4b and 4d). When interacting with trustworthy agents, investors who acted on the truth received higher payoffs than those who acted on misrepresentation—control: χ2(1) = 34.77, p < .001; competition: χ2(1) = 21.20, p < .001. In the control condition, payoffs from exaggeration did not differ from payoffs from truth, χ2(1) = 0.02, p = .898, and were significantly higher than payoffs from misrepresentation, χ2(1) = 10.67, p = .001. In competition, payoffs from exaggeration were significantly lower than payoffs from truth, χ2(1) = 7.07, p = .008, and did not differ from payoffs from misrepresentation, χ2(1) = 0.57, p = .450. When investors interacted with untrustworthy agents, investor payoffs did not significantly differ on the basis of the content of the message in either the control condition, all χ2s(1) ≤ 0.85, p ≥ .357, or the competition condition, all χ2s(1) ≤ 0.57, p ≥ .449. Thus, truth (and, to a more limited extent, exaggeration) improves investor payoffs relative to misrepresentation when agents are trustworthy but does not when agents are untrustworthy.
Figures 4b and 4d are being replaced as well.
Results of the regression on mean payoff as a function of agent’s trustworthiness in the previous round, message type, and condition. Results are shown separately for analyses in which the dependent variable was trustworthy agents’ payoffs (a), investors’ payoffs from trustworthy agents (b), untrustworthy agents’ payoffs (c), and investors’ payoffs from untrustworthy agents (d). Error bars represent 95% confidence intervals.