Executive function plays a role in coordinating different perspectives, particularly when one’s own perspective is involved

Highlights

• In young children executive control figures equally in belief and desire ascription.

• Relations are strongest if the child’s own belief/desire contrasts the other.

• The coordination of different perspectives in general requires executive function.

• High inhibition demands are at place if one’s own perspective contrasts with others.

Abstract

While developmental experiments with children and elderly subjects, work with neuropsychological patients and adult experimental studies have consistently found close relations between executive function and theory of mind, the foundation of this relation still remains somewhat unclear. One prominent account holds that executive function is specifically involved in ascribing such mental states, paradigmatically beliefs, that aim at representing the world truly because ascribing such states requires inhibition of normative defaults (beliefs being true) (e.g. Sabbagh, Moses, & Shiverick, 2006). The present studies systematically tested for the role of executive function in different forms of mental state ascription as a function of the type of state ascribed (beliefs or desires) and the first person involvement of the ascriber (whether she herself has an attitude conflicting with one to be ascribed to someone else) in young children. The results reveal that (i) executive function is related not only to belief ascription but equally to desire ascription when both are matched in terms of logical complexity (such that two subjective attitudes have to be ascribed to two agents that are incompatible with each other). (ii) Both for desires and for beliefs, these relations are strongest in such tasks where the ascriber herself is one of the two agents, i.e. has a belief or desire herself that stands in contrast to that to be ascribed to someone else. All in all, these findings suggest that executive function figures in coordinating perspectives more generally, not only epistemic ones, and in particular in coordinating others’ and one’s own conflicting perspectives.

Introduction

Theory of mind research investigates how children come to explain and predict rational action by ascribing mental states, such as beliefs, desires and intentions (e.g. Perner, 1991). This perspective taking ability – in its explicit forms – emerges during the preschool years and depends crucially on the development of domain-general cognitive capacities. In addition to linguistic abilities, one such factor that has been found to be crucial for theory of mind (ToM) across the lifespan is executive function (EF). EF and ToM have been shown to be closely related in child development (e.g. Carlson and Moses, 2001, Perner and Lang, 1999), in adults (e.g. Qureshi, Apperly, & Samson, 2010), in aging (e.g. Bailey and Henry, 2008, Rakoczy et al., 2012) and in neuropsychological patients (e.g. Samson, Apperly, Kathirgamanathan, & Humphreys, 2005).

Most extensively this ToM-EF relation has been studied in early child development, where it has been documented to be very robust, both in cross-sectional and longitudinal designs (where EF at time 1 predicts ToM at time 2), holding across difference cultures, and even when extraneous factors such as chronological and mental age are controlled for (e.g. Carlson and Moses, 2001, Carlson et al., 1998, Rakoczy, 2010, Sabbagh et al., 2006). Importantly, the associations between ToM and EF appear to be rather specific: EF correlates with a variety of superficially different ToM tasks, but not with structurally similar tasks that do not require ToM reasoning (e.g. Sabbagh, Moses, & Shiverick, 2006).

But what accounts for this relation? What role might EF play in ToM reasoning? One prominent account views EF as fundamental to what has been considered the litmus test of ToM reasoning, namely the ascription of false beliefs (FB): “Executive functioning is required to reason only about representations that are intended to reflect a true state of affairs” (Sabbagh, Moses, et al., 2006, p. 1034). The idea behind this proposal is the following: beliefs, in contrast to other types of propositional attitudes such as desires, wishes or intentions, have so-called “mind-to-world direction of fit” (Searle, 1983). That is, they aim at truth, at representing the world as it is. They thus have a normative default (beliefs ought to be true), and ascribing false beliefs requires deviation from and inhibition of this default (Russell, 1996, Sabbagh et al., 2006). This proposal gets empirical support from several studies showing that EF is related to false belief tasks, but not to closely matched tasks without the involvement of truth-aiming attitudes. First, so-called “conflict inhibition” EF tasks (incorporating both inhibition and working memory demands) correlate with false belief tasks and with tasks requiring an understanding of false signs (that also aim at truth), but not with structurally analogous “false photo” tasks (Sabbagh, Moses, et al., 2006). The crucial difference in terms of logical structure of the tasks is that the latter do not involve truth-aiming false representations (outdated photos aren’t wrong… – see below and Appendix D).

Second, EF has been found to correlate with belief ascription specifically, as compared to other forms of attitude ascription: FB tasks, but not closely matched tasks involving ascribing simple desires or pretend attitudes, have been found to correlate with EF tasks (Moses, Carlson, Stieglitz, & Claxton, 2003; cited in Moses et al., 2005).

On a theoretical level, this account has been disputed by a prominent theory that views the role EF plays in ToM reasoning as much more general and pervasive. According to Leslie and colleagues, EF is involved in combining, coordinating and inhibiting meta-representations in ToM inferences very generally, both regarding beliefs and regarding desires (e.g. Friedman and Leslie, 2004, Leslie et al., 2005, Leslie and Polizzi, 1998).¹

On an empirical level, the studies showing specific correlations of EF only with false belief and false sign tasks involved some fundamental confounds. In contrast to the tasks that did not correlate with EF, the FB tasks (a) involved beliefs (or other representations aiming at truth), (b) presented so-called “perspective problems”, and (c) did so in a way that there was a conflict between the ascriber’s own self-perspective and the perspective to be ascribed to someone else.

Regarding (b), one more general theory about ToM development claims that what emerges around age 4 is a more general capacity to coordinate multiple subjective perspectives and to solve “perspective problems” (Perner et al., 2003, Perner and Roessler, 2012, Perner et al., 2002, Perner et al., 2005; see also Moll, Meltzoff, Merzsch, & Tomasello, 2013): These are formally defined as tasks requiring the coordination of the content of two representations (e.g. mental attitudes, pictures, sentences, etc.) whose content (e.g. p, q) cannot be joined by a simple conjunction (p and q) without being relativized to different perspectives. A good example of perspectives embodied in external (non-mental) representations are the visuo-spatial perspectives in pictorial depictions. Think of two objects A and B depicted from opposite sides of a room. The two contents (of the pictures) “A is in front of B” and “B is in front of A” cannot be combined by a simple conjunction (“A is in front of B and B is in front of A” is inconsistent). Only relativizing them to standpoints allows a conjunction: “A is in front of B as depicted in picture 1, but B is in front of A as depicted in picture 2”. In similar fashion, the false sign test (in which a sign post, supposed to indicate the location of an object O which is in B, in fact points to A) presents a perspective problem: one has to understand the perspective clash between what the sign “says” (O is in A) and what is the case (O is in B) (Perner and Leekam, 2008, Sabbagh et al., 2006, Sabbagh et al., 2006). The superficially similar false photo test (Zaitchik, 1990), in contrast, does not present a perspective problem. In this task, a photo is taken of an object O in A at time 1. At time 2, O is moved to B, and the crucial question is what the photo will show about O’s location. This does not present a perspective problem because the photo does not refer to the current real situation and misrepresents it in the way the false sign does. Rather it represents a past situation. Whereas the false sign is literally false, the photo is merely outdated² (Perner and Leekam, 2008, Sabbagh et al., 2006, Sabbagh et al., 2006) (see Appendix D for details).

Regarding mental representations, understanding conflicting visual perspectives (“A is front of B” as seen from standpoint 1 and “B is in front of A” as seen from standpoint 2) and ascribing false beliefs (“object O is in B” in reality, but the protagonist believes that “object O is in A”) present paradigm cases of perspective problems. In contrast, ascribing simple desires (e.g. Repacholi and Gopnik, 1997, Wellman and Bartsch, 1988), pretence attitudes, etc. does not (Perner et al., 2005, see Appendix D for details). Crucially, however, there are perspective problems that do not involve truth-aiming attitudes, such as incompatible desires held by different persons (Perner et al., 2005, Rakoczy et al., 2007) or even within one person (Choe, Keil, & Bloom, 2005). Such desires (say to the effect that p vs. non-p) are structured in a “mutually exclusive” (Moll et al., 2013) way, in the sense that only one can be fulfilled at the same time and that understanding them requires a relativization to standpoints (“p and non-p” is inconsistent; what is required is “person 1 desires p whereas person 2 desires non-p”). Work with such tasks suggests that EF might in fact be involved in such perspective problem tasks that do not involve belief ascription in the same way as in FB tasks (Rakoczy, 2010).

Regarding (c), one limitation of previous work is that all the studies suggesting a specific FB-EF relation have used standard FB tasks (Perner et al., 1987, Wimmer and Perner, 1983). In such tasks, the child has to ascribe a belief to a protagonist about a state of affairs regarding which she herself already has a contrasting belief – creating a situation with a strong “pull of the real” or “curse of knowledge” from the child’s perspective (e.g. Birch & Bloom, 2007). It is well possible that EF plays a crucial role particularly in such situations where the ascriber’s own perspective needs to be disregarded in judging another’s perspective (see, e.g. Apperly, Samson, & Humphreys, 2005). Two lines of evidence – on belief and desire reasoning – might be taken as suggestive of such a possibility:

First, support comes from studies with so called reality unknown false-belief tasks (Call & Tomasello, 1999). In these modified FB tasks (details see below) the subject herself is ignorant about the real location of the target at the moment when the character’s false belief has to be ascribed. The subject can only infer the real location when taking into account the false belief of a protagonist who (in contrast to the subject) has seen the hiding in one of two identical containers at time 1, has not witnessed the subsequent swapping of the containers at time 2, and indicates his (false) belief at time 3. That is, the subject initially has no belief of her own as to the location of the object but only reaches such a belief by taking into account and reasoning from the belief of the other person. From a theoretical point of view, two characteristics of these tasks are fundamental: In contrast to previous modifications of belief tasks (e.g. Wellman & Bartsch, 1988), (i) the subject initially does not have an own perspective and (ii) there is a conflict between reality and the protagonist’s subjective perspective (i.e. the task represents a true perspective problem). Empirically, this type of FB task has been found to be as difficult (or even more difficult) as standard FB tasks for children (Call and Tomasello, 1999, Figueras-Costa and Harris, 2001), quite likely because it poses other kinds of task demands in terms of working memory and inferential complexity (reasoning backwards from the other’s belief to reality; see also Robinson and Mitchell, 1995, Wertz and German, 2007). What matters for current theoretical purposes, however, is not so much the question of absolute levels of performance on standard vs. reality unknown FB tasks (which might be comparable but due to different reasons), but rather whether they involve different cognitive processes, in particular whether they relate differentially to EF.

Evidence for such differential involvement of EF in standard vs. reality unknown FB tasks comes from a neuropsychological case study by Samson, Apperly, Kathirgamanathan & Humphreys (2005) which suggests that ascribing beliefs to another person and inhibiting one’s own perspective are functionally dissociable in adults to some degree. A patient with severe EF deficits due to a right fronto-temporal lesion was unable to ascribe false beliefs (on standard FB tasks) to someone else only when he himself knew about the current state of affairs in question (where the target object actually was located) at that moment. Crucially, however, he succeeded on the modified reality unknown FB task. Using the same two types of FB tasks, Bailey and Henry (2008) showed that decline in EF in older age is related to declining performance in standard FB tasks, but not in the modified tasks without a self-perspective that needs to be inhibited.

A second line of evidence comes from a developmental study investigating the role of EF in ascribing incompatible desires to two persons (Rakoczy, 2010): In this study, EF was generally correlated with such desire ascription, but this correlation tended to be higher for ascribing desires when one of the persons was the ascriber herself than for ascribing such desires to two third persons.

In contrast to these data with elderly and neuropsychological populations, to date there are no developmental data on the role of self-perspective inhibition in the relation of EF to belief ascription in children. All in all, existing work thus remains inconclusive as to the exact developmental role EF plays in ToM reasoning – whether it relates to ascribing truth-aiming attitudes, perspective problems more generally, or specifically to such perspective problems with a self-perspective to be inhibited. The rationale of the present work therefore was to investigate the relation of EF to different forms of ToM reasoning by systematically varying both - the type of attitude to be ascribed (belief vs. desire) and the involvement of a self-perspective to be inhibited. To this aim, we used tasks of ascribing beliefs vs. incompatible desires in versions in which the ascriber did or did not have a belief/desire that contrasted with the one to be ascribed to a protagonist.

The logic is the following: If EF is indeed specifically involved in ascribing truth-aiming representations, there should be specific correlations of EF with FB tasks, but not with desire ascription tasks. If on the other hand, EF were involved more generally in solving perspective problems, there should be correlations of EF with both – ascribing false beliefs and incompatible desires. And finally, if EF was specifically involved in inhibiting self-perspectives in solving such perspective problems, these correlations should be most pronounced in cases where the ascriber herself has a belief or desire conflicting with that of the protagonist.

Study 1 tested this with a full 2 (kind of attitude to be ascribed: belief vs. desire) × 2 (self-perspective inhibition demands: high vs. low) design of ToM tasks whose relation to EF was tested. As this study yielded clear evidence, converging with previous findings, that EF was related to ascribing incompatible desires, particularly when there was a strong self-perspective to be inhibited, but ambiguous evidence regarding the ascription of beliefs, Study 2 investigated belief ascription with and without self-perspective more systematically in a variety of different FB task formats – including also a standard FB test format with high and low requirements of self-perspective inhibition (see Table 1).

On the basis of the theoretical background and the available empirical evidence we expected (1) positive correlations among the conflicting-desires and among the false-belief sub-tasks and (2) positive raw and partial correlations between the different ToM tasks and executive functioning tasks (3) in particular for those ToM tasks with high demands of self-perspective inhibition.