The evidence for rationalisation, which motivates the target article, is exaggerated. There are two sources for this. First, it is an outcome of structural features of the experimental psychology tradition, which isolates effects using experimental control, and then uses null-hypothesis significance testing to establish their reality (i.e., their non-zero size), neglecting to gauge their importance (Stafford Reference Stafford2014). The second source of this exaggeration is the rhetoric of psychologists, who make hay out of emphasising the supposedly irrational aspects of our behaviour, de-emphasising reason-responsiveness (Stafford Reference Stafford2015).
With respect to the target article, we can see this in the motivating section (1.1. “Rationalization”). The concept of rationalisation is presented as pervasive (“people rationalize all the time”), solidly evidenced (“exhaustively documented”), and leading to “gross errors,” which are “stubbornly irrational.”
A study recruited as a key illustration of rationalisation is Sharot et al. (Reference Sharot, Velasquez and Dolan2010), but when we look at this example, we see that the largest effect reported in this paper was an average within-subject change of ~0.07 on a 6-point scale (experiment 1, see Fig. 1, t(20) = 2.4, p < 0.03), so the rationalisation manipulation produced a mean shift of ~1% in people's judgements. Hardly gross or stubborn.
Other examples cited by the target article are similar – showing small movements in people's ratings of belief, rather than flips from one belief state to another. This contrasts with the rhetorical portrayal of rationalisation. Participants in Brehm's (Reference Brehm1956) study showed an average change of ~0.9 on an 8-point scale (i.e., an ~11% between-groups shift, p < 0.01, with n = ~30 in each group). The experiment design, analysis, and presentation of results presented by Vinckier et al. (Reference Vinckier, Rigoux, Kurniawan, Hu, Bourgeois-Gironde, Daunizeau and Pessiglione2019) does not make a simple estimation of effect size for rationalisation obvious, but according to their Fig. 3, it looks like the within-subject effect size of choice (i.e., of rationalisation) is ~0.15 of a standardised (z) score. A “small effect,” as classically determined (Cohen Reference Cohen1992).
Some might view this as unproblematic. A reasonable view is that rationalisation is indeed common and commonly produces “gross errors,” but the methods of experimental psychology mean we can only hope to consistently capture rationalisation in the proxy form of meagre shifts on rating scales. But it is also a reasonable view, I believe, that the extant evidence for rationalisation does not support the grand claims for the power of the phenomenon. It is not enough that an idea be intuitively plausible.
The target article also invokes classic and widely known studies in psychology as evidence of rationalisation. It is not possible to review all of them in this commentary, but it is instructive to look at couple of salient examples. Cognitive dissonance is the first citation of the target article (Festinger Reference Festinger1962). The foundational demonstration of cognitive dissonance is Festinger and Carlsmith (Reference Festinger and Carlsmith1959). Taking the largest effect reported in this study, participants’ ratings of “enjoyability of task,” this showed a mean shift of 1.40 on an 11-point scale (i.e., ~13% change between groups, p < 0.03, t(38) = 2.22, with n = 20 in each group).
Another foundational contribution is Nisbett and Wilson (Reference Nisbett and Wilson1977), for which (surely) the most discussed study is Wilson and Nisbett (Reference Wilson and Nisbett1978, experiment 2), in which shoppers were asked to choose between four identical pairs of stockings. Famously, shoppers preferred the stockings on the right (p < 0.025, n = 52), but gave reasons other than position for their preference. Here my issue is not the size of the effect, but of its interpretation. While the explanations offered could be due to rationalisation, failure to report reasons is not the same as their inaccessibility. Stafford (Reference Stafford2014) argues that, where participants are ignorant of the conditions by which experimenters analyse their behaviour, failure to report those conditions is wholly compatible with rational choice.
Note my argument is not about the reproducibility, or not, of the evidence base, but rather of its interpretation.
It is telling that other motivating examples evoked by Cushman are fictional and/or psychologically exceptional (amnesia in the Bourne Identity, split-brain patients). While these are highly suggestive of the human potential of rationalisation they cannot be taken as evidence that rationalisation is pervasive in ordinary cognition.
I have argued that errors due to rationalisation are often small, rather than gross. The interpretive slippage between experimental effects at single time points and unusual edge cases (like split-brain patients) means that the evidence for the persistence (“stubbornness”) of rationalisation is simply unclear. Further, while the target article notes that reasons are a factor in driving action, alongside rationalisation, it is possible that some evidence presented as demonstrating rationalisation could actually be showing rational behaviour (so, for example, in Wilson & Nisbett Reference Wilson and Nisbett1978, it could be viewed as rational to give and defend an implausible answer if a psychologist asks you an impossible question such as which pair of identical stockings you prefer).
Where, then, does this leave the Cushman “representational exchange” account? I am not claiming that rationalisation doesn't exist, only that the evidence psychology has produced in support of it is far less strong than is commonly supposed. There is still – potentially – something for Cushman's account to explain. My criticism highlights an opportunity: One test of the value of Cushman's account is if it can provide experimentalists with the leverage to produce better evidence of rationalisation; that is, the value of the account could be in allowing us to predict when and how rationalisation will be at its strongest and most divergent from simple rationality in which actions are motivated by consciously accessible reasons.
Perhaps, guided by better theory, experimentalists will be able to generate manipulations that show rationalisation in the lab consonant with our intuitions of its importance in our everyday lives.
The evidence for rationalisation, which motivates the target article, is exaggerated. There are two sources for this. First, it is an outcome of structural features of the experimental psychology tradition, which isolates effects using experimental control, and then uses null-hypothesis significance testing to establish their reality (i.e., their non-zero size), neglecting to gauge their importance (Stafford Reference Stafford2014). The second source of this exaggeration is the rhetoric of psychologists, who make hay out of emphasising the supposedly irrational aspects of our behaviour, de-emphasising reason-responsiveness (Stafford Reference Stafford2015).
With respect to the target article, we can see this in the motivating section (1.1. “Rationalization”). The concept of rationalisation is presented as pervasive (“people rationalize all the time”), solidly evidenced (“exhaustively documented”), and leading to “gross errors,” which are “stubbornly irrational.”
A study recruited as a key illustration of rationalisation is Sharot et al. (Reference Sharot, Velasquez and Dolan2010), but when we look at this example, we see that the largest effect reported in this paper was an average within-subject change of ~0.07 on a 6-point scale (experiment 1, see Fig. 1, t(20) = 2.4, p < 0.03), so the rationalisation manipulation produced a mean shift of ~1% in people's judgements. Hardly gross or stubborn.
Other examples cited by the target article are similar – showing small movements in people's ratings of belief, rather than flips from one belief state to another. This contrasts with the rhetorical portrayal of rationalisation. Participants in Brehm's (Reference Brehm1956) study showed an average change of ~0.9 on an 8-point scale (i.e., an ~11% between-groups shift, p < 0.01, with n = ~30 in each group). The experiment design, analysis, and presentation of results presented by Vinckier et al. (Reference Vinckier, Rigoux, Kurniawan, Hu, Bourgeois-Gironde, Daunizeau and Pessiglione2019) does not make a simple estimation of effect size for rationalisation obvious, but according to their Fig. 3, it looks like the within-subject effect size of choice (i.e., of rationalisation) is ~0.15 of a standardised (z) score. A “small effect,” as classically determined (Cohen Reference Cohen1992).
Some might view this as unproblematic. A reasonable view is that rationalisation is indeed common and commonly produces “gross errors,” but the methods of experimental psychology mean we can only hope to consistently capture rationalisation in the proxy form of meagre shifts on rating scales. But it is also a reasonable view, I believe, that the extant evidence for rationalisation does not support the grand claims for the power of the phenomenon. It is not enough that an idea be intuitively plausible.
The target article also invokes classic and widely known studies in psychology as evidence of rationalisation. It is not possible to review all of them in this commentary, but it is instructive to look at couple of salient examples. Cognitive dissonance is the first citation of the target article (Festinger Reference Festinger1962). The foundational demonstration of cognitive dissonance is Festinger and Carlsmith (Reference Festinger and Carlsmith1959). Taking the largest effect reported in this study, participants’ ratings of “enjoyability of task,” this showed a mean shift of 1.40 on an 11-point scale (i.e., ~13% change between groups, p < 0.03, t(38) = 2.22, with n = 20 in each group).
Another foundational contribution is Nisbett and Wilson (Reference Nisbett and Wilson1977), for which (surely) the most discussed study is Wilson and Nisbett (Reference Wilson and Nisbett1978, experiment 2), in which shoppers were asked to choose between four identical pairs of stockings. Famously, shoppers preferred the stockings on the right (p < 0.025, n = 52), but gave reasons other than position for their preference. Here my issue is not the size of the effect, but of its interpretation. While the explanations offered could be due to rationalisation, failure to report reasons is not the same as their inaccessibility. Stafford (Reference Stafford2014) argues that, where participants are ignorant of the conditions by which experimenters analyse their behaviour, failure to report those conditions is wholly compatible with rational choice.
Note my argument is not about the reproducibility, or not, of the evidence base, but rather of its interpretation.
It is telling that other motivating examples evoked by Cushman are fictional and/or psychologically exceptional (amnesia in the Bourne Identity, split-brain patients). While these are highly suggestive of the human potential of rationalisation they cannot be taken as evidence that rationalisation is pervasive in ordinary cognition.
I have argued that errors due to rationalisation are often small, rather than gross. The interpretive slippage between experimental effects at single time points and unusual edge cases (like split-brain patients) means that the evidence for the persistence (“stubbornness”) of rationalisation is simply unclear. Further, while the target article notes that reasons are a factor in driving action, alongside rationalisation, it is possible that some evidence presented as demonstrating rationalisation could actually be showing rational behaviour (so, for example, in Wilson & Nisbett Reference Wilson and Nisbett1978, it could be viewed as rational to give and defend an implausible answer if a psychologist asks you an impossible question such as which pair of identical stockings you prefer).
Where, then, does this leave the Cushman “representational exchange” account? I am not claiming that rationalisation doesn't exist, only that the evidence psychology has produced in support of it is far less strong than is commonly supposed. There is still – potentially – something for Cushman's account to explain. My criticism highlights an opportunity: One test of the value of Cushman's account is if it can provide experimentalists with the leverage to produce better evidence of rationalisation; that is, the value of the account could be in allowing us to predict when and how rationalisation will be at its strongest and most divergent from simple rationality in which actions are motivated by consciously accessible reasons.
Perhaps, guided by better theory, experimentalists will be able to generate manipulations that show rationalisation in the lab consonant with our intuitions of its importance in our everyday lives.