Independent effects of relevance and arousal on deductive reasoning

ABSTRACT

Emotional content can have either a deleterious or a beneficial impact on logicality. Using standard deductive-reasoning tasks, we tested the hypothesis that the interplay of two factors – personal relevance and arousal – determines the nature of the effect of emotional content on logicality. Arousal was assessed using measures of skin conductance. Personal relevance was manipulated by asking participants to reason about semantic contents linked to an emotional event that they had experienced or not. Findings showed that (1) personal relevance exerts a positive effect on logicality while arousal exerts a negative effect, and that (2) these effects are independent of each other.

KEYWORDS:

• Reasoning
• emotion
• relevance
• trauma
• cognition

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 Nine participants suffering from clinical PTSD and 15 male participants were also tested because we initially intended to form PTSD and male comparison groups. We excluded these participants from the analyses reported in this paper due to the difficulty in completing these comparison groups.

2 Cohen’s d measure of effect size varied between 0.40 and 1.40 across five studies on the effect of emotional contents on logical reasoning (Blanchette & Campbell, 2012; Blanchette et al., 2007, 2014; Blanchette & Leese, 2011; Blanchette & Richards, 2004), with an average effect size of 0.77. With such effect size, groups of 20 participants yield a statistical power of 77%.

3 We ran the same analyses using the index [=z(P(“yes it’s valid” answer|Valid Problem)) − z(P(“yes it’s valid” answer|Invalid Problem))] reported by Heit and Rotello (2014) which are more robust against strategic effects than the traditional accuracy index used in this study. This analysis helps us determine whether differences are related to changes in reasoning process, rather than changes in response bias. With the index, both the interaction between content and abuse group, and that between content and accident group, were significant, respectively, F(2,162) = 3.58, p = .030, and F(2,162) = 3.57, p = .031. In addition, a Receiver Operating Characteristic (ROC) representation of the data (Fawcett, 2006) suggested that differences in performance between emotional contents were not due to trade-offs between ‘true positives’ (“yes it’s valid” answers to valid problems) and ‘false alarms’ (“yes it’s valid” answers to invalid problems). The latter ROC representation thus suggests that the interaction effects reported in this study are not the result of strategic differences in response bias between victims and controls. The data are available as supplemental material for additional analyses.

4 In an additional analysis, we excluded the ‘both abuse and accident’ participants and used Group as one independent variable with three levels (control, abuse, accident). This analysis included 77 participants, compared to 85 participants for the main analysis. The important interaction between Group and Content remained marginally significant, F(4,148) = 2.15, p = .078. This suggests that the interactions between groups and content were not solely driven by the ‘both abuse and accident’ participants.

5 Note that the syllogisms presented in Tables A1^{Table A2Table A3Table A4}–A5 are English translations of the French syllogisms used in the study.

Additional information

Funding

This research was funded by the Social Sciences and Humanities Research Council of Canada.