The mediating effect of reflective-analytic cognitive style on rational thought

Abstract

We investigate whether non-miserly cognitive styles mediate the effects of cognitive ability and thinking dispositions on rational thought. Specifically, we review relevant literature on two dimensions that define non-miserly cognitive style: reflection-impulsivity and analytic-intuitive. We use these two dimensions to identify a continuum of cognitive styles that vary from miserly (impulsive-intuitive) to non-miserly (reflective-analytic) and are congruent with tendencies to commit specific rational thinking errors. Further, we argue that this continuum, which we label reflective-analytic cognitive style, mediates the effect of cognitive ability and thinking dispositions on rational thought. We conduct an experiment testing our predictions and find evidence, via structural equation modelling, that reflective-analytic cognitive style does mediate the effect of cognitive ability (as measured by working memory capacity) and thinking dispositions (as measured by need for cognition and actively open-minded thinking) on responses to traditional rational thinking tasks.

Keywords:

• Analytic cognitive style
• reflection-impulsivity
• dual process theory
• working memory
• thinking dispositions

Acknowledgments

The authors thank Valerie Thompson (editor) and three reviewers, including Shira Elqayam, for providing valuable insights and significant direction in the development of this paper. The lead author thanks Shelby Hunt for encouraging the pursuit of truth, Roy Howell for his mathematical insights, and Arcade Fire for their poetic inspiration.

Disclosure statement

The authors assert that no potential conflicts of interest exist.

Notes

1 Although cognitive styles were initially considered to be stable attitudes and preferences, recent theory and empirical findings suggest that cognitive styles are malleable (Bendall, Galpin, Marrow, & Cassidy, 2016; Kozhevnikov et al., 2014; Zang, 2013). Alternatively, we find no theoretical arguments suggesting thinking dispositions are malleable, with the notable exception that need for cognition correlates with age and education (Cacioppo et al., 1996).

2 A similar argument is stated in Pennycook et al. (2014a, p. 197).

3 Further evidence that cognitive style and thinking dispositions are not interchangeable constructs comes from prior studies reporting that measures of cognitive style predict variance in rational thought incremental to thinking dispositions. For example, Toplak and colleagues report that the cognitive reflection test (CRT), which is one measure of analytic cognitive style, explains variance in rational thought that is incremental to thinking dispositions (Toplak et al., 2011, 2014). Although there are strong arguments that the CRT is a potent predictor of rational thinking because it is performance-based rather than self-reported (Pennycook et al., 2014a; Toplak et al., 2011; Trippas et al., 2015), the analysis reported by Toplak and colleagues supports the proposition that cognitive style measures, such as the CRT, are qualitatively different from those of thinking dispositions.

4 We note that the econometrics literature recommends using a log transformation of predictor variables to account for elastic (i.e., curvilinear) effects (Griffiths, Hill, & Judge, 1993). Thus, using a log transformation for response times is congruent with the reasoning provided in Baron et al. (2015) and recommendations from the econometrics literature.

5 The 20-item MFFT used in this study is available for future studies by contacting any of the authors.

6 We note that the direct effects models in Preliminary Experiments 1a and 1b, posted online at https://osf.io/2d4t9, yielded unique solutions. However, in both cases, the structural path from working memory to performance was statistically insignificant, reinforcing our conclusion that the mediation model is preferred over the direct effects model.

7 For example, we ran an SEM model in which the roles for thinking dispositions and cognitive style were flipped. The model yielded two nonsensical effects. The structural path from working memory capacity to thinking dispositions was negative (β = –.546 p = .055), suggesting that higher levels of working memory capacity would yield lower levels of thinking dispositions. Also, working memory capacity’s indirect effect on rational thinking errors via thinking dispositions was negative (β = –.447 p = .059), suggesting that higher levels of working memory capacity would yield more rational thinking errors.