Representational and executive selection resources in ‘theory of mind’: Evidence from compromised belief-desire reasoning in old age

Abstract

Effective belief-desire reasoning requires both specialized representational capacities—the capacity to represent the mental states as such—as well as executive selection processes for accurate performance on tasks requiring the prediction and explanation of the actions of social agents. Compromised belief-desire reasoning in a given population may reflect failures in either or both of these systems. We report evidence supporting this two-process model from belief-desire reasoning tasks conducted with younger and older adult populations. When task inferential complexity is held constant, neither group showed specific difficulty with reasoning about mental state content as compared with non-mental state content. However, manipulations that systematically increase executive performance demands within belief-desire reasoning caused systematic decreases in task performance in both older and younger adult groups. Moreover, the effect of increasing executive demands was disproportionately greater in the older group. Regression analysis indicated that measures of processing speed and inhibition contributed most to explaining variance in accuracy and response times in the belief-desire reasoning tasks. These results are consistent with the idea that compromised belief-desire reasoning in old age is likely the result of age-related decline in executive selection skills that supplement core mental state representational abilities, rather than as a result of failures in the representational system itself.

Introduction

Our capacity to understand the actions, reactions and interactions of other social beings, termed ‘theory of mind’, has been studied extensively over the last 20 years, most prolifically from a developmental perspective. Recently, the capacities underwriting social cognition have begun to receive serious attention more broadly in cognitive neuroscience, including the documentation of possible brain areas supporting ‘theory of mind’ via neuro-imaging (see Gallagher and Frith, 2003, Saxe et al., 2004 for recent reviews) and neuropsychological case studies of acquired and developmental brain lesions (e.g. Baron-Cohen et al., 1985, Fine et al., 2001, Frith et al., 1991, Rowe et al., 2001, Channon and Crawford, 2000, Apperly et al., 2004). Remarkably, despite more than 20 years of cognitive developmental research on ‘theory of mind’, the first simple models of successful belief-desire reasoning have only recently been proposed (Friedman and Leslie, 2004a, Friedman and Leslie, 2004b, Friedman and Leslie, 2005, Leslie and Polizzi, 1998, Leslie et al., 2004, Leslie et al., 2005).

In the current paper, we outline briefly the assumptions of a recent model of belief-desire reasoning, and test predictions of this model in young and older adults. The model has been developed to account for results from behavioral belief-desire reasoning tasks largely undertaken with young children as participants, in which children are required to predict or explain the actions that social agents take in terms of their underlying mental states. For example, children might be required to predict where a story character might look for some desired object x, given that she believes it is in location y.

Inference tasks such as this might seem trivially simple from an adult point of view, but for young children, under certain conditions, they can prove quite challenging. For example, in one prominent task children watch a character, Sally, hide a marble in one location before leaving the scene, whereupon the marble is switched to another location in her absence. In predicting Sally's search on returning to the room children younger than 4 years of age often fail to take into account her false belief, and predict search at the location containing the object—an error that is overcome by the time that typically developing children reach age five (Baron-Cohen et al., 1985, Wellman et al., 2001). Thus, one major contributor to the ease or difficulty of belief-desire reasoning is whether the content of the belief to be attributed is true (easier) or false (more difficult), and this fact has proved to be one key finding that theories of the development of belief-desire reasoning have attempted to explain (Bloom, & German, 2000).

There is considerable debate about the nature of the mechanisms that support ‘theory of mind’ skills. However, theories in cognitive development have generally agreed that effective belief-desire reasoning requires at least two kinds of process. First, there must be representational resources that capture mental state knowledge itself. Second, there must be processes that allow this knowledge to be deployed in actual prediction and explanation tasks. Different theorists place more or less emphasis on the relative importance of each kind of process in explaining the developmental data (Leslie et al., 2004, Leslie et al., 2005, Scholl and Leslie, 1999, Scholl and Leslie, 2001, Wellman et al., 2001).

According to the framework adopted here, ‘theory of mind’ capacities are in part based on a representational system that allows mental state concepts such as belief and desire to be represented as such. This representational system is provided by a specialized reliably developing modular neuro-cognitive mechanism that supports the early capacity to attend to and then learn about mental states (Baron-Cohen, 1995, German and Leslie, 2000, German and Leslie, 2001, German and Leslie, 2004, Leslie, 1994b, Leslie, 2000, Leslie et al., 2004, Roth and Leslie, 1998, Scholl and Leslie, 1999). Reasoning over the representations provided by this basic representational system also requires a species of executive function. These supplementary processes serve to select particular mental state contents to be attributed to social agents in a given belief-desire reasoning problem.

The idea is that in a given situation, perceptual descriptions of a social agent's behavior act as inputs to the specialized representational system which provides a number of candidate mental state attributions. The executive selection process then decides among these candidate mental state contents. For example, in the false belief task, the system might provide two possible contents for Sally's belief: first, that she believes the marble is in location 1 (where she left it) or second, that she believes it is in location 2 (where it currently is). A further assumption of this model of belief desire-reasoning is that the selection process has a default content. Because people's beliefs are most often true, the selection process will always attach most initial weight to the true-belief content: the content consistent with the state of the world as it appears to the attributer of the belief (e.g. the participant).

On this theory, the standard false belief task is difficult because the selection process is required to overcome the default ‘true belief’ attribution. By hypothesis, executive inhibitory processes are required to manage this selection, in effect reducing the activation level of the true belief content to a point below that of the false belief content (which is the correct content to attribute given the situation). The proposal that inhibitory processes are involved in belief-desire reasoning has support from several converging sources.

First, inhibitory executive function is limited in the preschool period and beyond (Gerstadt et al., 1994, Diamond et al., 2002), and performance on belief-desire reasoning problems is strongly correlated with measures of executive inhibitory function (Carlson et al., 1998, Carlson et al., 2002, Carlson and Moses, 2001, Russell et al., 1991). Second, false belief performance in young children is affected by manipulations of the relative salience of the ‘true belief’ vs. ‘false belief’ (correct) content. For example, when attention is drawn to the initial location of the object in a false belief problem, children perform significantly better (Leslie et al., 2005, Siegal and Beattie, 1991, Surian and Leslie, 1999, Yazdi et al., in press).

Effective belief-desire reasoning thus depends critically on an intact representational system for belief and desire knowledge that comes online in the typical case in the first years of life (Leslie, 1987, Leslie, 1994a) and also on mechanisms of executive selection that develop only gradually over the preschool years. Therefore while successful performance requires both cognitive systems to be intact, compromised performance might result from inefficiencies in either system, or both. With colleagues, we have argued elsewhere that inefficient belief-desire reasoning that is seen in young children (e.g. failures in the false belief task) can be plausibly accounted for in terms of the gradual development of pre-frontal cortex supported executive selection systems, as demonstrated via the strategy of varying the difficulty of the selection demands in belief-desire reasoning problems described above (German and Leslie, 2000, German and Leslie, 2004, Leslie et al., 2004, Leslie et al., 2005, Yazdi et al., in press)¹.

In other populations with known social cognition problems, however, failures with the specialized belief-desire representational system itself have been implicated. This hypothesis has been advanced for the case of the Autism Spectrum Disorders (Baron-Cohen, 1995, Frith et al., 1991, Leslie and Roth, 1993, Leslie and Thaiss, 1992).

In the current paper we investigate recent claims of compromised belief-desire reasoning associated with cognitive aging (Maylor et al., 2002) with respect to the question of decline in frontal executive systems.

Individuals undergoing cognitive aging are an ideal population to test the notion that belief-desire reasoning requires executive selection processes because such individuals have been shown to suffer from decline in a range of executive capacities including aspects of inhibitory function (Hasher and Zacks, 1988, Kramer et al., 1994), as well as processing speed (Salthouse, 1996) as age increases. These declines have been captured under the ‘frontal’ hypothesis of aging (Lowe and Rabbitt, 1997, Raz, 2000). Moreover, whereas for preschool children, the claim that belief-desire reasoning failure might result from a failure to have yet acquired relevant specialized representational knowledge about mental states (Wellman et al., 2001) is at least plausible, it seems considerably less so for elderly adults².

Prior studies of ‘theory of mind’ in old age have provided mixed results. An initial study by Happé, Brownell and Winner (1998) used tasks requiring inferences from short stories and reported a sparing of mental state inferences compared to control tasks; their older group (mean age 73 years) outperformed a younger group (mean age 21 years) in the mental state inference condition. However, this advantage was not replicated by Maylor et al. (2002), who actually found an age related decline in the ability to successfully answer simple theory of mind inferences, as compared to control tasks. Moreover, the decline in mental state based inferences was shown to persist even in tasks with reduced memory demands, and taking into account measures of processing speed and executive function (see also, Sullivan, & Ruffman, 2004). Maylor et al. (2002) concluded that the declining theory of mind abilities observed in their study were consistent with both neuroimaging and neuropsychological evidence of a frontal locus for aspects of processing in theory of mind tasks (Frith and Frith, 1999, Saxe et al., 2004, Rowe et al., 2001) and the notion of frontal decline associated with aging (Lowe, & Rabbitt, 1997), but expressed puzzlement at the lack of association with their measures of executive function (the WCST), and the persistence of the problem after executive and speed of processing measures were accounted for (Maylor et al., 2002, p 481).

We note that one possible reason for the failure to find an executive association, noted above, might be an underestimation of the fractionation that could exist within executive frontal processes (Burgess et al., 1998, Goldman-Rakic, 1996, Miyake et al., 2000, Shallice and Burgess, 1991). Developmental evidence has suggested that the relationship between executive function and mental state reasoning may be quite specific to tasks within the executive function spectrum that draw on certain inhibitory components (e.g. response conflict rather than delay, Carlson et al., 2002). Thus, finding or not finding an association might depend critically on the selected measures of executive function chosen.

In the current study, we tested a central prediction of the two part processing model of belief-desire reasoning—that executive selection is required as part of belief-desire reasoning—in a group of elderly participants (mean age=78) and a group of younger adult control participants (mean age=20), using two basic strategies. Firstly, we created matched minimally different inference tasks that either required a calculation based on the mental states of a social agent, or a calculation based on a physical system that had access to equivalent information. Critically, we improved on prior inference tasks used to assess ‘theory of mind’ vs. ‘control’ inferences, both with elderly participants (Happé et al., 1998, Maylor et al., 2002) and in early functional imaging studies (Fletcher et al., 1995, Gallagher et al., 2000), by designing true minimal pair story content.

Minimal pair designs aim to manipulate content between two conditions via the smallest possible change being made. Examples from cognitive development that have elucidated understanding of the acquisition of ‘theory of mind’ skills include the comparison between the false belief task and the false photograph task (Zaitchik, 1990), in which task structure remains constant and content is minimally altered. This pair of tasks was used to demonstrate the domain specific nature of the limitation showed in reasoning about representational content in children with autism spectrum disorders, who perform at ceiling with public representations such as photographs and maps, while performing poorly on problems involving belief (Charman and Baron-Cohen, 1995, Leslie and Thaiss, 1992). Another key example of a minimal pair design that has had great impact involves the addition of a single word to the standard false belief action prediction problem; if children, instead of being asked where Sally will look, are asked where she will look first, performance dramatically improves (Siegal and Beattie, 1991, Surian and Leslie, 1999, Yazdi et al., in press). To date, minimal pair designs such as these have not been utilized in the study of mental state reasoning in elderly populations; inspection of the stories used in the mental state vs. control conditions reveals many potential differences in content, structure, and complexity of the inferences. In the current manuscript, we attempted to more closely match the structure of the mental state and control problems such that a close minimal pair was formed.

In one battery of reasoning tasks, representational content was varied within a task structure held constant in terms of the number and complexity of the inferences required. In ‘mental state’ tasks a scenario was described and a social agent was introduced. A judgment was then required about the action that agent would take, given the scenario. In the control versions of the tasks, the same scenarios were used but the inference was required about whether a mechanical device (e.g. an alarm system) with access to equivalent information would be triggered or not. Critically, the amount of information required to make each inference was held constant. Because mental state stories involved an additional entity (the social agent), this feature was matched in the control inference by the inclusion of an additional variable that could affect the outcome (e.g. whether a power supply was functioning or not). An example of a minimal pair of inference tasks used in the experimental battery appears in Table 1.

The second strategy was to manipulate executive selection requirements of a basic belief-desire reasoning problem. In each case, the participant was required to predict the action that would follow from a particular belief-desire pair. The tasks differed according to factors empirically known to affect the ease or difficulty of the executive selection problem, based on studies conducted with preschool children. One factor, briefly discussed before, was that the tasks varied in terms of the truth or falsehood of the belief (TB vs. FB tasks). As noted, false belief problems are harder for preschool children than are true belief problems.

The second factor manipulated in this task battery was whether the actor had a desire to ‘approach’ or ‘avoid’ a target object. While children from age four typically solve false belief tasks where the character has the desire of approaching the object (e.g. Sally wants her marble, but thinks it is in the box) results show that this success does not extend to the case where Sally wishes to avoid something about the location of which she has a mistaken belief (Cassidy, 1998, Leslie and Polizzi, 1998, Leslie et al., 2005). Children do not reliably solve avoid desire false belief problems until age six (Friedman and Leslie, 2004a, Friedman and Leslie, 2004b). Indeed, from the preschool data across tasks that vary the truth and falsehood of the protagonist's belief, and whether the character has the desire to approach or avoid the object, there is a predictable developmental trajectory: TB approach>TB avoid>FB approach>FB avoid (Cassidy, 1998, Friedman and Leslie, 2004a, Leslie and Polizzi, 1998, Leslie et al., 2005). An example of a story in each of its four versions appears in Table 2.

We predicted that if belief-desire reasoning problems arise as a result of failures to meet the executive selection demands, our elderly participants would show a disproportionate increasing cost in accuracy and/or response time in solving the belief-desire reasoning tasks as executive performance demands increase in the order derived from the developmental trajectory.

Finally, in order to directly assess the hypothesis that executive selection capacities are related to belief-desire reasoning, a number of measures of cognitive function with no mental state content were included. These were chosen to provide a focused test of the idea that executive selection in belief-desire reasoning might draw on inhibitory processes (Friedman and Leslie, 2004a, Friedman and Leslie, 2004b, Leslie, 2000, Leslie et al., 2004, Leslie et al., 2005) and thus a variety of ‘response competition’ inhibition tasks were included. These comprised interference conditions of the Stroop task, the Hayling sentence completion task (Burgess, & Shallice, 1996), and the interference condition of the ‘day–night’ task (Gerstadt et al., 1994).

In addition, given claims that working memory might be one of the executive systems engaged by mental state reasoning (Gordon and Olson, 1998, Davis and Pratt, 1995), we assessed working memory via the backwards digit span task. Following Maylor et al. (2002) we also assessed ‘fluid’ processing speed and matched our participants in terms of ‘crystallized’ and ‘pre-morbid’ intelligence.