2.1. The Diachronic Account

Gerrans and Stone (Reference Golan, Baron-Cohen and Golan2008, 125) characterize their view as returning to an early explanation of ToM by Baron-Cohen, Leslie, and Frith (Reference Baron-Cohen, Ring, Wheelwright and Joliffe1985). Baron-Cohen and colleagues explain ToM by positing a domain-general capacity called metarepresentation that enables ToM, as well as other capacities, in which a relationship between an object and a representation is represented. Gerrans and Stone's explanation of ToM is more complex than this early view: “ToM abilities depend, not on MR [metarepresentation], EF [executive function], language, or recursion per se, but on their developmental interaction with the low-level precursors previously described, e.g., gaze processing” (128). Gerrans and Stone are not clear about how appealing to development in addition to “low-level precursors” helps explains ToM. In the following paragraphs, I detail their argument for the diachronic account.

Gerrans and Stone first discuss developmental details with which no one involved in the debate over the explanation of the ToM deficit would disagree. They note that the capacity for ToM depends on the development of many lower-level cognitive mechanisms, such as face processing (Gerrans and Stone Reference Golan, Baron-Cohen and Golan2008, 126). The typically developing child will, in addition to these domain-specific systems, have domain-general systems that develop between the ages of 2 and 4; Gerrans and Stone are interested most in executive function, secondary representation, recursion, and metarepresentation. These “distance us from our primate ancestors … rather than the fact that we possess some domain-specific ToM mechanism and they do not” (127).

Next, Gerrans and Stone consider how the developmental aspect of their account, of which they provide minimal discussion, might explain persons with autism's performance on false-belief tests. They argue that the results from such tests are underdetermined, since persons with autism's failure of false-belief tests could result from either lower-level deficits, such as a joint-attention deficit, or higher-level deficits, such as a deficit in metarepresentation (Gerrans and Stone Reference Golan, Baron-Cohen and Golan2008, 129). However, they do think that a mark in their explanation's favor is that it predicts persons with autism's success on false-photograph tests and failure on false-belief tests.Footnote ² Gerrans and Stone argue that their account predicts this result because it posits lower-level deficits that affect the individual with autism on the false-belief task but not on the false-photograph task, as the following quotation illustrates: “Our reply to domain-specific theorists of ToM is that if children with autism have intact MR [metarepresentation], but impaired inputs to MR from impaired low-level domain-specific mechanisms (e.g., joint attention), then one would predict exactly this dissociation between false-photograph and false-belief performance. Individuals with autism perform well on the false-photograph task because the inputs to the metarepresentational system, in that task, are not impaired in autism” (131). Note that Gerrans and Stone say nothing about a diachronic developmental interaction in this explanation; instead, they appeal only to a lower-level deficit (joint attention).

There is a more general worry for Gerrans and Stone's diachronic account. Edouard Machery (Reference Rutherford, Baron-Cohen and Wheelwrightforthcoming) has recently criticized a similar developmental explanation of Williams Syndrome by Annette Karmiloff-Smith (Reference Kleinman, Marciano and Ault1998). Machery summarizes Karmiloff-Smith's related point as follows: “Brain areas are massively interconnected, so that developmental changes in one of these areas have consequences in numerous other areas. As a result, even minor developmental brain problems have cascading effects, which affect the development of the whole brain and of all cognitive capacities” (Machery, Reference Rutherford, Baron-Cohen and Wheelwrightforthcoming). Karmiloff-Smith's argument is similar to the diachronic account that I have sketched. On this account, the ToM deficit is best explained by a lower-level input deficit occurring early in development that results in “abnormal developmental interaction” between that input and higher-level systems (Gerrans and Stone Reference Golan, Baron-Cohen and Golan2008, 133).

Machery's worry is that, although this may seem plausible, it neglects that modularity is a key characteristic of biological development. As Machery notes, “in the context of developmental biology, ‘modularity’ refers to the property that the developmental pathway leading to the development of a trait is shielded from the changes that might happen to other developmental pathways” (Reference Rutherford, Baron-Cohen and Wheelwrightforthcoming). Given that we find modularity throughout biological development, it is highly plausible to expect that there is some degree of modularity in the architecture of the mind, even with capacities such as ToM. At this stage, we may infer that Gerrans and Stone's account is less prima facie plausible than an account that appeals to modularity in higher-level processes.

Gerrans and Stone (Reference Golan, Baron-Cohen and Golan2008) do not address this issue, but Gerrans (Reference Gerrans and Stone2002, 310–11) argues that we must determine on a case-by-case basis whether domain-general or domain-specific systems subserve abilities. I grant that we must do this on a case-by-case basis. Nonetheless, considering the ubiquity of modularity in biological development does inform our search for such systems; it directs us to look for modularity. Given the current state of autism research, the widespread modularity we find in biological development adds plausibility to modular explanations of abilities such as ToM.

As mentioned already, it is unclear that the diachronic aspect of Gerrans and Stone's (Reference Golan, Baron-Cohen and Golan2008) stated view is doing much work in their explanations of empirical findings. Regardless of whether their diachronic account is less plausible than a synchronic account, I will now defend my claim that when Gerrans and Stone attend to empirical work, they do not appeal to “developmental interaction” of any sort; rather, they explain such studies by positing only current lower-level deficits.

If Gerrans and Stone's (Reference Golan, Baron-Cohen and Golan2008) argument hinged on the diachronic component of their account, it would be unclear whether it was better supported than an account appealing only to lower-level deficits. They do not discuss whether the diachronic account explains the ToM deficit better than an account that appeals only to a lower-level deficit such as facial processing.Footnote ³ Regardless, when they are discussing empirical studies, they do not invoke their diachronic account to explain ToM deficits; thus, I suggest that the diachronic account is not doing the explanatory work they claim.

One such example mentioned already is their discussion of persons with autism's performance on false-belief and false-photograph tests (Gerrans and Stone Reference Golan, Baron-Cohen and Golan2008, 131). In their explanation of these phenomena, they appeal only to “impaired low-level domain-specific mechanisms” and not to any sort of developmental interaction. As they note: “We believe that deficits on ToM tasks arise essentially from deficits in lower-level input systems” (131). Gerrans and Stone (132) claim that appealing to low-level deficits explains persons with autism's performance more parsimoniously, but they provide no discussion of how such parsimony relates to the diachronic aspect of their account.

Gerrans and Stone (Reference Golan, Baron-Cohen and Golan2008, 132–33) also explain the performance of higher-functioning individuals with autism by appealing only to lower-level deficits. This explanation is discussed at length below, but it is worth noting that after discussing persons with autism's performance on false-belief and false-photograph tests, and after discussing the performance of higher-functioning persons with autism on other tasks, Gerrans and Stone claim that their developmental interaction account provides the best explanation for both phenomena, as the following quotation illustrates: “The weight of evidence indicates that the ToM impairments in autism … are a result of abnormal developmental interaction between low-level systems and higher-level, domain-general capacities, caused essentially by deficits in ToM precursors” (131, 133). Since the explanations they provide of both of the aforementioned phenomena rely only on their appeal to lower-level deficits, it is unclear why developmental interaction should play a role in explaining ToM deficits. Since Gerrans and Stone do not mention it at all in their explanations of these two crucial phenomena, it does not seem that the diachronic component of their account is what does the real explanatory work.

Even if it were not the case that the work in Gerrans and Stone's argument is being done by appealing only to current lower-level deficits, it seems that such an explanation of the ToM deficit still threatens the explanation that appeals to a domain-specific ToM module. That is, aside from being interested in interpreting Gerrans and Stone charitably, it is important for my argument that I deal with the possibility of explaining ToM deficits by appealing only to current lower-level deficits.

2.2. The Synchronic Account

Even if the likelihood of modularity in a higher-level process like ToM is increased by recognizing the near ubiquity of modularity in biological development, Gerrans and Stone's explanation of key studies in cognitive neuroscience still threatens a modular explanation of ToM deficits. Without relying on developmental interaction, Gerrans and Stone argue that studies of higher-functioning persons with autism where subjects demonstrate difficulty on ToM tasks can be explained by appealing only to lower-level deficits. In this section, I examine several such studies, including recent studies not discussed by Gerrans and Stone (Reference Golan, Baron-Cohen and Golan2008), and argue that they are best explained by appealing to a higher-level, domain-specific ToM module.Footnote ⁴

Gerrans and Stone (Reference Golan, Baron-Cohen and Golan2008, 132–33) discuss three case studies of individuals on the autism spectrum who were very high functioning; two were college students (a computer science student and a physics student), and the other was a mathematics professor who had won the Fields Medal (Baron-Cohen et al. Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999).Footnote ⁵ Gerrans and Stone discuss this study because it is a potential counterexample to their explanation of ToM deficits. This study is a potential counterexample because, given their careers, these subjects have an advanced ability “to represent relations between symbols and their objects (magnitudes, abstract spatial relations) and to perform recursive computations over these symbols” (133). Since on Gerrans and Stone's account ToM is supported by domain-general processes, individuals who are able to engage in activities requiring metarepresentation, recursion, and executive function, such as the activities in which these three subjects are able to excel, while being unable to perform well at activities requiring the representation of others’ mental states, would provide a counterexample to their explanation.

Baron-Cohen et al. (Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999) administered three tests: a folk psychology test (ToM), a folk physics test (basic problem solving about the physical world), and an executive function test (Tower of Hanoi test). The first test is the most relevant to the present discussion. Subjects were shown 36 photographs of the eye region and asked to choose the word that best described what the person in the photograph was thinking or feeling (479); this is called the “reading the mind in the eyes” test (for discussion, see Baron-Cohen et al. Reference Baron-Cohen, Wheelwright, Hill, Raste and Plumb1997, Reference Bloom and German2001). The subjects performed well on the executive function and folk physics tests, but all exhibited deficits on the ToM test.

Gerrans and Stone argue that since the ToM test in Baron-Cohen et al.'s (Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999) study isolated the eye region, it is more likely that the subjects’ difficulty in identifying emotions was due to a lower-level deficit in facial or gaze processing: “all three had difficulty in inferring what someone was feeling, or paying attention to, from pictures of the eye region of the face (Baron-Cohen et al. [1999]), indicating a problem with lower-level domain-specific capacities for face and gaze processing rather than MR [metarepresentation]” (Gerrans and Stone Reference Golan, Baron-Cohen and Golan2008, 133). There is a problem with their interpretation of this study, and there are also related studies that weigh against Gerrans and Stone's interpretation. Note that, as with their explanation of performance on the false-belief tests mentioned above, Gerrans and Stone do not discuss how their diachronic account contributes anything to their explanation of the behavior of these subjects.

The problem with Gerrans and Stone's explanation is that they neglect to mention the control task administered to the subjects. To rule out a facial- or gaze-processing deficit, the subjects and the control group were asked to identify individuals’ gender by viewing only their eye region. On this task, the three subjects performed at the same level as the control group, but all three scored greater than one standard deviation below the mean of the control group on the ToM task (Baron-Cohen et al. Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999, 479–80). Since the subjects performed as well as the control group on a general facial-processing task not related to emotion, the ToM deficits exhibited by these subjects are not likely due to a lower-level deficit related to facial or gaze processing.

Surprisingly, Gerrans and Stone do not address this aspect of the study, but one can imagine how their response might go. They might argue that the control task of identifying gender from a photograph of someone's eye region does not decisively rule out that ToM is supported by a higher-level, domain-general process such as metarepresentation. Furthermore, they might grant that the control task does rule out that the subjects were generally impaired at facial processing, but it does not rule out that their poor performance on the ToM task could have been due to a more specific, lower-level impairment affecting only the facial processing of emotions. That is, though some elements of facial processing may be intact (i.e., processing of facial features sufficient enough to determine someone's gender), there is a domain-specific, lower-level component of facial processing that provides inputs of facial emotions.

This potential reply from Gerrans and Stone argues that the control task does not allow us to determine whether the ToM deficits in the Baron-Cohen et al. (Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999) study are due to the failure of a higher-level, domain-specific module (a ToM module) or of a lower-level, domain-specific module (an emotion-only facial-processing module) that is failing to provide inputs to a domain-general process (e.g., metarepresentation). Recent studies in facial processing that I will now discuss lend support to the former explanation over the latter.

Two recent studies (Lahaie et al. Reference Machery, Block and Adriaens2006; Humphreys et al. Reference Joseph and Tanaka2007) of facial processing by higher-functioning individuals on the autism spectrum suggest that, although some lower-functioning persons with autism have difficulty with facial processing, in higher-functioning individuals ToM deficits may be dissociated from facial-processing deficits. Lahaie and colleagues’ (Reference Machery, Block and Adriaens2006) study examined a group of higher-functioning persons with autism using two tests to determine whether their configural facial processing differed significantly from the control group and whether they responded to the priming of facial parts as the control group did. For the first test, Lahaie and colleagues used a standard design whereby faces were inverted and participants were asked to match upright and inverted faces. Face inversion is known to disrupt recognition of faces because it inhibits normal configural processing of facial features.

In previous studies of the face-inversion effect, persons with autism performed better than controls in recognizing inverted faces but similarly or worse than controls with upright faces, causing some researchers to argue that persons with autism have a deficit in configural facial processing (Lahaie et al. Reference Machery, Block and Adriaens2006, 31). In Lahaie and colleagues’ Reference Machery, Block and Adriaens2006 study, the group of subjects with autism was shown both faces with neutral expressions (to prevent a potential confound from faces showing emotion) and nonfacial stimuli, Greebles, that would also be inverted. Given previous studies, Lahaie and colleagues expected that the group of subjects with autism would not show an inversion effect, which would support the deficiency in configural face processing posited by previous studies (31). Instead, both the control group and the group of subjects with autism exhibited the face inversion effect, which supports the view that higher-functioning persons with autism do not have a deficit in configural facial processing; this finding is consistent with a similar study by Joseph and Tanaka (Reference Karmiloff-Smith2003).Footnote ⁶

The second test administered to the subjects in Lahaie and colleagues’ (Reference Machery, Block and Adriaens2006) experiment was designed to determine whether the group of subjects with autism would exhibit a priming effect similar to the control group. I will not discuss this second test in detail; the relevant result is that the group of subjects with autism performed similarly to the control group on the priming task, suggesting that individuals in both groups benefited from configural information being made available (Lahaie et al. Reference Machery, Block and Adriaens2006, 38). These findings support the view that configural facial processing is not impaired in individuals on the autism spectrum, but one might ask how configural facial processing is relevant to the “reading the mind in the eyes” test. It is relevant because Lahaie and colleagues found that, contrary to previous studies, subjects with autism looked at the eye region of the face more than other areas of the face. Not only did these subjects pay more attention to the eye region but, moreover, they processed the eye region better after being primed with that area than any of the other facial areas with which they were primed.

Their overall performance as equal with the control group and their enhanced ability to process the eye region suggest not only that these higher-functioning subjects with autism are not suffering from a lower-level facial-processing deficit but also that they process the eye region better than other facial regions. This lends validity to the methodology in the “reading the mind in the eyes” test and also weighs against Gerrans and Stone's appeal to lower-level facial-processing deficits to explain the ToM deficits in Baron-Cohen and colleagues’ study (Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999).Footnote ⁷ Recent experiments on auditory processing by persons with autism also lend support to the view that difficulty identifying emotions in spoken voices is not due to lower-level deficits.Footnote ⁸

Gerrans and Stone could respond to these studies in two different ways. First, they could argue that though the studies discussed above seem to indicate that higher-functioning individuals can have ToM deficits in the absence of lower-level deficits, this does not rule out the possibility of them having a higher-level deficit, such as with executive function. Second, they could accept the findings of these experiments but then posit a specialized lower-level module devoted only to processing emotional information in facial processing (and another such module for auditory processing; see n. 8). Neither of these responses will work.

The first reply will not explain the behavior of the three subjects in Baron-Cohen and colleagues’ (Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999) study. The subjects were tested not only for ToM using the “reading the mind in the eyes” test but also for executive function capabilities using the Tower of Hanoi test. Gerrans and Stone could not appeal to the domain-general process of metarepresentation that they posit, since these three individuals are clearly skilled at advanced representation and at recursive computations, given their professions, as Gerrans and Stone admit (Reference Golan, Baron-Cohen and Golan2008, 133). Furthermore, though it makes use of false-belief tests, a study of a subject with damage to the left amygdala suggests that ToM is dissociated from executive function (Fine, Lumsden, and Blair Reference Gerrans2001), and there are numerous other studies suggesting that young children diagnosed with autism do not exhibit deficits in executive function (e.g., Griffith et al. Reference Hill1999; Yerys et al. Reference Zaitchik2007).Footnote ⁹ As a result, it seems that Gerrans and Stone cannot appeal to a higher-level, domain-general process to explain the subjects’ performance in Baron-Cohen and colleagues’ (Reference Baron-Cohen, Wheelwright, Stone and Rutherford1999) study.

The second reply may seem plausible, but if Gerrans and Stone went this route they would strongly detract from the force of their argument. Gerrans and Stone emphasize that they are seeking a more parsimonious account of the ToM deficit by appealing to lower-level deficits. Their desire for parsimony is clear from the cautionary phrase that they co-opt: modulae non sunt multiplicanda praeter necessitate (Gerrans and Stone Reference Golan, Baron-Cohen and Golan2008, 122). To explain the behavior of subjects with autism in the visual and auditory processing tasks (see n. 8), Gerrans and Stone would need to posit a deficit in the module for emotion processing for vision and for hearing (while all other lower-level processing modules are intact). Rather than explaining these deficits by adding a module for each modality, these studies can be explained by appealing to a single higher-level, domain-specific ToM module that accounts for deficits in both visual and auditory tasks.