Does ego depletion reduce judgment adjustment for both internally and externally generated anchors?

Highlights

• We test the sticky anchor hypothesis of ego depletion for both external and internal cues.

• We exclude alternative explanations, such as social mechanisms.

• We use a pre-registered, high-powered study design.

• Neither self-generated nor experimenter-provided anchors were affected by ego depletion.

Abstract

Ego depletion is a state in which people prefer to avoid mental effort, therefore possibly leading to increased reliance on heuristics. Effortful thinking has been shown to help reduce anchoring effects, in which people form social judgments by adjusting from an initial value (the anchor). We therefore predicted that ego depletion would reduce the amount of adjustment from an initial anchor, leaving the final judgment relatively close to the anchor value. In contrast to previous research by Banker et al. (2017), we excluded alternative explanations, such as social mechanisms. In particular, we tested adjustment from internally generated and externally provided anchors. We theorized that judgment adjustment processes are the same for both internal and external cues. The results showed that neither self-generated nor experimenter-provided anchors were affected by ego depletion, thus leaving social mechanisms as the prime alternative explanation. The data also showed that susceptibility to anchoring is not a trait because reactions to different anchors were not substantially intercorrelated. Further, providing externally or internally generated anchors did not make a fundamental difference. However, the loss of data was higher with self-generated anchors. Overall, the results suggest that researchers can confidently rely on experimenter-generated anchors.

Introduction

Ego depletion refers to a state in which a person's capacity for effortful control has been strained and the person prefers effortless to effortful thinking. This state typically occurs after a person has engaged in effortful cognitive activity, for example by suppressing one's emotions or making many decisions (Baumeister & Vohs, 2016b, p. 75). Resisting anchoring effects (e.g., not letting a numerical value bias one's estimation) can be considered effortful thinking. Based on these premises we hypothesized that ego depletion should lead to stronger anchoring effects.

Anchoring and adjustment is a heuristic process proposed by Tversky and Kahneman (1974). When trying to estimate an unknown value, the person may start from a relevant known value (the anchor) and then make adjustments. Much research has found that there is a general tendency for the adjustment to be insufficient, so that the judgment remains closer to the anchor than it should be. Today, several accounts for anchoring exist. For example, the original account suggests that people typically start with the anchor value and adjust the value until a reasonable estimate is achieved (e.g., Epley & Gilovich, 2001). Other accounts, such as the selective accessibility model (Mussweiler & Strack, 1999), suggest that an anchor primes knowledge that in turn makes the anchor plausible. For some time, these two accounts were believed to be specific to internally generated anchors (i.e., adjustment only occurs for numbers one thinks of at first when confronted with a question) and externally provided anchors (i.e., priming only occurs for random numbers given by the experimenter). However, it has been shown that both mechanisms work for both, internally and externally generated anchors (Chaxel, 2014; Simmons, LeBoeuf, & Nelson, 2010).

Anchoring emerged as relevant to ego depletion in an investigation by Banker, Ainsworth, Baumeister, Ariely, and Vohs (2017). Preliminary studies using the Dictator Game, in which one person was given a cash reward to divide between self and another, found that depleted Dictators kept more money for themselves, which could indicate selfishness (selfishness hypothesis). However, when the procedure was refined so that the money was first given to the other person and the dictator could then take as much as he or she wanted, depleted dictators were exceptionally unselfish. The implication was that depleted dictators tended to leave the money wherever it was. If it had initially been given to them, they kept more of it, but if it had initially been given to the other player, they left it there. Banker et al. (2017) concluded that ego depletion leads to stronger influence of external cues (sticky anchor hypothesis of ego depletion). This is consistent with other evidence that ego depletion creates a general attitude of passivity, in which the self becomes less likely to take active control and engage in proactive mental activity (Vonasch, Vohs, Pocheptsova Ghosh, & Baumeister, 2017).

We sought to extend that research in two key respects. First, anchoring effects in that study by Banker et al. (2017) were not operationalized in the usual manner, which involves having the person generate the anchor internally and arbitrarily (e.g., by using the numbers from one's birthday) and estimating whether some parameter is higher or lower. Instead, an amount of cash was allocated to self or other. The social and monetary aspect may have introduced motivational factors that interacted with the ego depleted state, such as attitudes toward the partner, guilt or pride over one's allocation, and wishing to make a good impression on the experimenter.

Thus, although Banker et al. (2017) interpreted their findings in terms of anchoring, two large questions remained open because they did not use any of the standard anchoring procedures. The first question is whether the interpersonal context was responsible for the effects: Participants were not simply estimating a parameter, as in most of the anchoring studies, but instead were re-allocating money to themselves and another person. Obviously, the interpersonal aspect introduces concerns beyond simple judgment and estimation, such as fairness, moral obligation, and cooperation. The second question is whether Banker et al.'s findings were – if at all attributable to anchoring effects – specific to externally generated anchors or whether they can be replicated with internal anchors. Hence Banker's findings point to the importance of testing whether ego depletion would increase the anchoring effect in the standard paradigms that involve just one person estimating a parameter based on an internally generated anchor.

In the dictator game paradigm studied by Banker et al. (2017), the initial allocations were determined by the experimenter (i.e., assigning all the money initially to either the participant/dictator or to the non-dictator partner). The sticky anchor hypothesis of ego depletion as described by Banker et al. (2017) builds upon the theory that “explicit self-reports reflect automatic associations to a greater extent when people do not have the motivation or the cognitive capacity to retrieve additional information from memory” (Hofmann, Gawronski, Gschwendner, Le, & Schmitt, 2005, p. 1380). However, so far the hypothesis has been tested only for external cues (Banker et al., 2017; Haan & van Veldhuizen, 2015). We will extend the research by using internal cues. In sum, we will choose experimental setups that resemble the anchoring paradigms and are typically used to test whether depletion has an effect on anchoring while excluding social mechanisms and including internal cues in the form of self-generated anchors.

The intended contribution of this investigation is as follows: We build upon Banker et al. (2017), who considered only one mechanism of anchoring, namely the selective accessibility model (p. 2), and we will briefly show that the observed effects would be predicted on the basis of other mechanisms, such as anchoring and adjustment (e.g., Epley & Gilovich, 2001), too. Second, we examine ego depletion's effects on susceptibility to anchoring effects on the basis of a classic anchoring paradigm. This way we will provide a cleaner test of the prediction that ego depletion affects anchoring. Third, we include self-generated anchors, which we consider to be internal cues. We believe it to be important to corroborate the sticky anchor hypothesis for internal cues: if self-generated anchors, that is, internal cues are affected by ego depletion, too, this would open new research avenues and hypotheses, for example it would predict stronger IAT-patterns for depleted participants. On the other hand, if the effect turns out to be limited to the interpersonal context studied by Banker et al., our broad hypothesis would be falsified. That would suggest that the role of ego depletion would be to alter susceptibility to socially motivational factors rather than simply decreasing the internal process of effortful adjustment.

Finally, by combining two lines of research we aim to learn more about how both of them (i.e., ego depletion and anchoring) work. Note that both fields are penetrated by theoretical questions: Anchoring effects – although their existence is undisputed – are subject to numerous theories, which seem to be exchanged arbitrarily in order to explain new findings. Ego depletion is subject to theoretical and methodological criticism. There is uncertainty about its existence despite a plethora of studies (e.g., Baumeister, Tice, & Vohs, 2018; Hagger, Wood, Stiff, & Chatzisarantis, 2010). Following failed replication attempts (e.g., Hagger et al., 2016), heated discussions about the existence of the phenomenon have started (Baumeister & Vohs, 2016a), leading to fears of large file drawers of unpublished and unsuccessful attempts to demonstrate ego depletion effects (Friese, Loschelder, Gieseler, Frankenbach, & Inzlicht, 2018, p. 12). Recent attempts to answer the question on the methodological boundary conditions have resulted in pre-registered articles: For example, Radel, Gruet, and Barzykowski (2019) chose optimal conditions for the effect but found inconclusive evidence of the effect, whereas a high-powered meta-analysis found a small effect (Dang et al., 2019). Some nonsignificant findings may be due to using a very weak manipulation. Pre-registered studies with longer manipulations by Sjåstad and Baumeister (2018) found very large effect sizes on the manipulation check and medium to large effects on the dependent variable.

If we can confirm our hypothesis we will provide evidence for the existence of the ego depletion effect and – because this is a registered report – enlarge the body of pre-registered ego depletion research (e.g., Dang, Liu, Liu, & Mao, 2017). The effect of ego depletion on the susceptibility to anchoring effects is a sufficient condition for the existence of ego depletion. This means that if the hypothesis is not confirmed, the present design will, however, not permit us to conclude that the ego depletion effect does not exist. It could also mean that ego depletion does exist, but it does not alter the anchoring and adjustment process. In the case of null-results, we plan to conduct inferiority tests to determine whether our study's effect is significantly smaller than the effects reported by Banker et al. (2017).

There are several competing theories that seek to account for anchoring effects, and more than one may be correct (e.g., Simmons et al., 2010). Apart from selective accessibility as a possible mechanism predicting anchoring effects (p. 2), there is a host of other mechanisms that have been proposed (e.g., Tversky & Kahneman, 1974) and each mechanism has been supported by some findings while not being able to account for others (e.g., the proposed mechanism numeric priming does not account for the finding that anchoring the number of calories of French fries does not anchor how many grams of fat they contain, Frederick & Mochon, 2012). As the present study will focus on the role of ego depletion in anchoring, we will briefly discuss how ego depletion may affect each of the proposed mechanisms.

Theories and mechanisms that are put forward to explain anchoring effects include (1) confirmatory testing, (2) semantic priming, (3) numeric priming, (4) insufficient adjustment, and (5) scale distortion. The first two of these, (1) confirmatory testing and (2) semantic priming, are rooted in the selective accessibility model (Mussweiler & Strack, 1999) and are what Banker et al. (2017) featured in their research: Concerning confirmatory testing, “depletion may impair the individual's ability to consider options counter to the suggestion” (Wheeler et al., 2007, as cited in Banker et al., 2017) and may thereby lead to stronger anchoring effects. In this argument, depletion leads people to engage in more confirmatory information processing and less reasoning (Fischer, Greitemeyer, & Frey, 2008; Schmeichel, Vohs, & Baumeister, 2003). The argument dovetails with findings from research on anchoring: Considering the opposite, that is taking into account arguments that are inconsistent with the anchor, weakens anchoring effects (Mussweiler, Strack, & Pfeiffer, 2000).

Besides confirmatory testing, semantic (2) and numeric priming (3) have been proposed as mechanisms responsible for anchoring effects. Depleted participants may be more susceptible to all priming effects (e.g., Bertrams, Baumeister, Englert, & Furley, 2015). Also “priming selective accessibility (a) impacts the anchoring bias independently of the type of anchor and (b) interacts with effort in the same way across both sources of anchors” (Chaxel, 2014, abstract).

Originally, anchoring effects were referred to as (4) insufficient adjustment (Tversky & Kahneman, 1974). Epley and Gilovich (2001) later introduced the analogy of TOTE-units from cybernetics (pp. 392–393) suggesting that what occurs when people adjust is a finite loop of testing (Is this number a good estimate?), if no, operating (Take a bigger/smaller number), testing again, and exiting (Take this number as the final estimate) when the testing is positive. This mechanism requires effort (e.g., Simmons et al., 2010, p. 917). We suggest that taking the loop multiple times or undertaking big adjustments is effortful. In terms of the TOTE-model, depleted participants would make fewer steps or smaller steps away from the anchor and therefore produce a final judgment closer to the anchor, as compared to non-depleted participants. Note that Jia and Hirt (2016) could show that depletion “suspends the operation of the self-focused comparator mechanism”, or in our terms, people exit the TOTE-loops at an earlier stage when little adjustment has occurred.

Finally, (5) scale distortion theory suggests “a change in the way numeric labels are being used in a particular context unaccompanied by any change in the beliefs or mental representation of the entity being judged” (Frederick & Mochon, 2012, p. 131). Note that Frederick and Mochon (2012, p. 131) “expect that scale distortion would have a negligible effect for most of the items [with self-generated anchors] (Epley & Gilovich, 2001, 2004, 2005, 2006), [as the mechanism] is restricted to judgments in which a strongly associated (and clearly relevant) referent is spontaneously brought to mind” (Frederick & Mochon, 2012, p. 131). We see no reason why ego depletion should have any effects on anchors if only this mechanism were at work.

Considering the different mechanisms proposed to underlie anchoring effects, we argue that – except for scale distortion theory – all discussed theories and mechanisms should work for both self-generated and experimenter-provided anchors and derive the following hypothesis: Ego depletion leads to stronger anchoring effects, regardless whether the anchors were self-generated or provided by the experimenter.

The consistency of ego depletion effects across different types of anchors conforms with the sticky anchor hypothesis of ego depletion as proposed by Banker et al. (2017) but has not yet been tested with respect to internally generated anchors. In the case of null-results (no main effect of depletion on the strength of anchoring effects), we conjecture that the impact of ego depletion on the susceptibility to external cues as outlined by Banker et al. (2017) may depend on the social mechanisms that were present in their experiments and not on the internal adjustment and priming processes that are involved in anchoring effects. This would conform with results obtained by Haan and van Veldhuizen (2015). Note, that besides the social mechanisms resulting in this pattern, a false positive remains an alternative explanation. However, given the high power of Banker et al.'s results as shown by p-curve analysis (see Bankeretal_pcurve.txt at https://osf.io/2c75d) we deem this explanation very improbable. In the case of an opposite relationship, that is ego depletion leading to lower susceptibility to anchoring effects, we would conjecture that participants were less motivated to the degree that they gave random estimates or that participants became better at estimating due to unknown reasons. Such a finding would contradict the findings of Banker et al. (2017) and Epley and Gilovich (2001, Studies 2 and 3). Note that the primary purpose is to test the (one-tailed) sticky-anchor hypothesis. In terms of rejecting this hypothesis, it does not matter whether the effect cannot be found or is in the opposite direction. The secondary purpose is to conduct a follow-up test to test whether the effect predicted by the sticky anchor hypothesis depends on the type of anchor. If so, the sticky anchor hypothesis of ego depletion needs to be altered with respect to its universality. Possible explanations are different processing styles regarding internal and external stimuli.