Mental simulation and argument

Abstract

We examine how opinion on a controversial real-world issue shifts as a function of reading relevant arguments and engaging in a specific mental simulation about a future, fictional state of affairs involving the target issue. Individuals thought either counterfactually about a future event (“if only X had not happened …”) or semifactually about it (“even if X had not happened …”). In Experiment 1, as expected, individuals became more in favour of a course of action (the electronic tagging of children) after reading relevant supporting arguments. They also became more in favour after they thought counterfactually how a negative future event could have been avoided if only electronic tagging had not been banned some years earlier. The effects of argument relevance and type of mental simulation were additive. Using a similar procedure but a different target issue (whether the UK should join the Economic Monetary Union), Experiment 2 confirmed that the type of mental simulation exerts a substantial effect on the impact of relevant arguments. Much weaker effects arose when the mental simulation was not relevant to the target issue. We set and discuss our results in terms of dual-processing theory.

Notes

We thank the editor, David Over, and three anonymous referees for constructive comments on an earlier version of this paper.

Exactly what kinds of mental representations are involved? One proposal (see McCloy & Byrne, 2002) is that individuals understand a counterfactual of the form “if only p, q” (e.g., If only I had caught the train, I would have made the meeting”) by keeping in mind a conjectured possibility of p and q (catching the train and making the meeting) and the presupposed fact, not-p and not-q (not catching the train and not making the meeting; see also Johnson-Laird & Byrne, 1991; Rios, Madruga, & Byrne, 2001). In contrast, individuals understand a semifactual of the form “even if p, not-q” (e.g., Even if I had caught the train, I would have missed the meeting”) by envisaging the conjectured possibility of p, not-q (catching the train and missing the meeting) and the presupposed fact, not-p and not-q (not catching the train and missing the meeting). The scenarios used in the present studies allow the same analysis except that the presupposed world is set in the future as is the conjectured possibility. For a discussion of this proposal see Over, 2004.

In Experiment 1, the same pattern of findings obtains if we restrict the analyses to performance in the baseline condition (n = 20) in which there were no relevant prior arguments and there was no relevant mental simulation: individuals endorsed 83% of their current thoughts as own arguments, endorsement varied with absolute affective value (44%, 86%, and 92% for absolute affective values of 0, 1, and 2, respectively), and there was a significant correlation between summed absolute affective value and rated argument strength, r = .528, p = .02, with the number of arguments partialled out. In Experiment 2, the baseline condition yielded a similar pattern. Current thoughts were endorsed as own arguments on 93% of the occasions, and endorsement was lower for current thoughts rated zero in an affective value (62%) and increased for absolute affective values of 1 and 2 (97% and 96%, respectively). The correlation of summed absolute affective value and argument strength was significant r = .526, p = .021, with the number of arguments partialled out.

There is substantial collinearity between affective value of the current thoughts and the affective value of the ones endorsed as own arguments (Pearson rs were .948 and .965 for Experiments 1 and 2, respectively). Simple correlations of these indices with opinion shift in Experiment 1 showed r = .115, p = .255 and r = .216, p = .031 for current thoughts and own arguments, respectively. The comparable figures for Experiment 2 were r = .185, p = .065 and r = .196, p = .051.

Logistic regression of the participants' yes/no decisions (Experiment 2) in the relevant mental simulation condition (N = 40), showed that prior opinion, regret, and own arguments were all significant predictors on 1-tailed tests (min p = .086/2). Type of mental simulation was not significant (p = .40).