Academia.eduAcademia.edu
OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN Chapter 27 COGNITIVE SCIENCE AND EXPL ANATIONS OF PSYCHOPATHOLO GY Kelso Cratsley and Richard Samuels Introduction The past two decades have witnessed a striking convergence in the interests and methods of cognitive science and psychiatry. On the one hand, cognitive psychologists and cogni- tive neuroscientists have increasingly sought to model various forms of psychopathol- ogy; in part because they promise to illuminate our understanding of normal cognition, but also because such phenomena are now viewed as intriguing in their own right. At the same time, researchers within psychiatry have increasingly come to adopt the methods and assumptions of the cognitive sciences—especially cognitive neuropsychology and cogni- tive neuroscience—with an attendant commitment to a particular pattern of explanation which depends upon the construction of mechanistic models (Broome and Bortolotti 2009; Kendler 2008, Kendler et al. 2011). The study of psychopathology has thus become an important facet of the cognitive sciences, and the cognitive sciences have, in turn, exerted an important influence on many regions of psychiatry. While these developments have led to significant research findings, the application of a cognitive scientific approach to psychopathology is still very much in its infancy, and the scope and limits of the research strategy remain to be determined. The aim of this chapter, then, is to explore the prospects for a developed cognitive science of psychopathology. In “Core explanatory assumptions” we outline the core theoretical assumptions of much cog- nitive science with a specific focus on the sorts of explanatory strategies that cognitive scien- tists typically seek to produce. In “Explaining psychopathology” we consider the extension of these explanatory strategies to the study of psychopathology. Then, in “Some cognitive accounts of psychopathology” we briefly illustrate how these strategies have been applied to two specific pathological phenomena: Autism Spectrum Disorder and Major Depressive Disorder. Finally, in “Two challenges to the cognitive science of psychopathology,” we discuss 27_Fullford_C27.indd 413 4/10/2013 8:28:36 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 414 Summoning Concepts a pair of challenges to standard cognitive scientific approaches to psychopathology, one that focuses on the pervasive role of dissociative data in theory construction, and another, more “philosophical” challenge which purports to establish in-principle limits on the sorts of explanations of psychopathology that cognitive scientists have sought to provide. Core Explanatory Assumptions: Information Processing and Mechanism Though cognitive scientists do not share a single vision of how explanation ought to pro- ceed, large regions of the field cleave to a set of familiar assumptions about the sorts of models that ought to be developed. For heuristic purposes, we divide up these assumptions into two related families of commitments. The first concern the idea that cognitive proc- esses and capacities depend on information processing. The second concern the idea that cognitive explanations are in some appropriately broad sense mechanistic. Neither family of assumptions has gone unchallenged in cognitive science. So, for example, advocates of situated or dynamical approaches sometimes reject them (Chemero 2009). For present pur- poses, however, we are concerned with research that adheres to these assumptions. And in what follows, we sketch them in a bit more detail and identify a range of explanatory strat- egies to which they give rise. In practice these different strategies are seldom pursued in isolation from each other, but are instead combined in different admixtures within indi- vidual research programs. Nevertheless, it will be useful to separate them out since in doing so we will be better placed to appreciate the range of strategies available in understanding psychopathology. Information processing Amongst the core assumptions of much cognitive science is that cognitive capacities –vision, reasoning, language production, memory, and so on—depend upon information process- ing of some sort. This general assumption has been articulated in a range of different ways, invoking different notions of information and different conceptions of the sorts of process- ing that is involved (see, e.g., Piccinini and Scarantino 2011). But in practice cognitive sci- entists typically assume that the relevant processes involve representations: roughly, physical states and structures have semantic contents in that they mean something or denote aspects of the world (Clark 2000; Thagard 2012; Von Eckardt 1993). Further, it is widely assumed that the relevant sort of processing is in some sense computational (Cummins and Cummins 2000; McDonald and McDonald 1995). This much is true of the sorts of “classical” computa- tional models that dominated much early cognitive science (Fodor and Pylyshyn 1988), but it is also true of mainstream connectionist research as well (Smolensky 1988). Further, even amongst researchers who are less explicit in their computational assumptions, there is a per- vasive tendency to characterize cognition in terms of representations and operations ther- eon. Such assumptions are, for example, widespread in developmental psychology (Carey 27_Fullford_C27.indd 414 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 415 2009), linguistics (Pinker 1994), and cognitive neuroscience (Gazzaniga 2009), even though much of this research does not provide any very perspicuous computational characteriza- tion of the phenomena under discussion. Suppose, as most cognitive scientists do, that psychological capacities and processes involve the sort of information processing we sketched earlier. Then it will be possible to explain cognitive capacities and processes in a range of different ways, or, roughly equiv- alently, to characterize them at a number of different levels of description. First, it will be possible to describe cognition at what has variously been called the intentional, semantic or knowledge level (Newell 1990; Pylyshyn 1984). Roughly put, we can explain what peo- ple do with reference to the semantic contents of their representational states—for exam- ple, what they believe and what their goals are—and the meaningful connections that hold between such states. Thus we might describe the cognitive development of infants in terms of their learning new concepts or beliefs on the basis of perceptual experience, and we might explain the decision-making behavior of an adult in terms of their preferences and their beliefs about the world. Such explanations abstract away from the computational (and neu- robiological) details of psychological processes, but can nonetheless be exceedingly useful for making sense of human behavior and cognition. Second, if an information processing account of cognition is correct, then it will be pos- sible to describe psychological processes in more computationally perspicuous terms—at what is sometimes called the representational/algorithmic level (Marr 1982). Commitment to such a level of description in exceedingly natural in light of the following broadly held assumptions: The semantic contents of mental states are somehow encoded by information carrying states and structures of the brain, what are sometimes called representational vehicles. Such representational vehicles are created, utilized, and transformed by the computational processes in which they figure. Under these assumptions, it is plausible to suppose that psychological capacities and proc- esses can be explained by characterizing the computations they involve and the properties of representational vehicles that allow for them to figure in such computations. To take one well-known example, it is common to characterize aspects of visual percep- tion—such as binocular vision—by providing some account of the representational vehicles involved, and a computational characterization of the systems that allow for the exercise of the capacity—for example, the extraction of depth information from binocular disparity (Marr 1982). As with intentional level description, this strategy is exceedingly common- place in cognitive science. Mechanistic commitments So far we have sketched some widespread assumptions about cognition, and highlighted the way in which they give rise to two commonplace explanatory strategies. We now turn to another commitment, also widespread amongst cognitive scientists, which concerns the mechanistic character of cognition. To a first approximation, it is widely supposed that 27_Fullford_C27.indd 415 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 416 Summoning Concepts cognitive capacities—such as, language production, memory and visual perception— depend upon the operation of cognitive mechanisms: roughly, physically realized informa- tion processing systems that are hierarchically decomposable into functionally specifiable components, whose individual activities and mutual interactions are responsible for the production of cognitive phenomena (Bechtel 2008). So characterized, this mechanistic assumption constitutes a rough-and-ready piece of metaphysics—a claim about the sorts of entities that cognitive systems are.1 But it is a piece of metaphysics that has important meth- odological ramifications, and in what follows we spell these out in more detail. A first ramification of the mechanistic assumption is that any comprehensive account of cognition must ultimately explain how representations and computational operations are dependent on—or implemented by—physical structures, states and processes. For if cognition depends on the activity of physical systems, then presumably there should be some explanation of how these physical systems—paradigmatically parts of the brain2— are able to perform the relevant computational operations. Such descriptions are some- times said to comprise a distinctive implementation or physical level of description (Marr 1982; Pylyshyn 1984). Crudely put, theories couched at this level are analogous to descriptions of computational hardware in that they characterize those physical states and structures in virtue of which a system is able to perform information processing tasks of the relevant sort. A second ramification of the mechanistic assumption is that cognitive phenomena can be explained by affecting a decomposition of the system into subparts (Bechtel and Richardson 1993; Cummins 1975). Roughly put, the idea is that if cognitive capacities depend on hierar- chically decomposable physical systems, then it should be possible to explain the cognitive capacity of an organism by decomposing the relevant system into its parts and describing how their individual activities and mutual interactions give rise to the capacity in question. Indeed, the typical goal of such decompositions is to “break up” the system into parts so simple that they manifestly involve no intelligence at all (Fodor 1968). It is important to be clear, however, that there are two quite different sorts of decomposi- tional analyses to be found in cognitive science. These two approaches, though not always clearly distinguished, differ in the sorts of parts that they specify. The first approach, some- times called functional analysis (Cummins 1975), involves a decomposition of a process—a system performing a task—into its various component subprocesses.3 So, for example, one might decompose the process of visual perception into a range of subprocesses, such as those involved in extracting light intensity gradients from the retinal image, those involved 1 For example, the mechanistic assumption appears to rule out various metaphysical theses about the mind, such as substance dualism. But it is largely neutral with respect to such issues as whether some form of functionalism is true, and whether some version of the type-identity theory is correct. 2 Though see Clark (2008) for a defense of the view that much cognition depends on extra-neural physical structures. 3 Functional analysis is sometimes characterized not in terms of the decomposition of processes into subprocesses, but in terms of the decomposition of the system’s dispositions into subdispositions, or capacities into subcapacities. But in the present context such differences are not important. This is because any exercise of a psychological disposition or capacity that involves the exercise of subcapacities/ disposition will be a process, and any psychological process of this sort will be an exercise of a capacity/ disposition. Thus any given functional analysis can equally well be construed as an analysis of a process or as an analysis of a capacity/disposition. 27_Fullford_C27.indd 416 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 417 in determining depth discontinuities, and those involved in determining the spatial orienta- tion of objects (Marr 1982). Further, these various subprocesses will themselves typically be decomposed, in iterative fashion, so that the final analysis represents the overall process as a hierarchically organized informational process of the sort that is naturally represented as a flowchart or a program (Cummins 2000). For this reason functional analysis is exceedingly closely related to what we earlier called the algorithmic level. Specifically, such analyses are often a prelude to—and indeed ultimately represented as—an algorithmic level description of the system. The second sort of decomposition does not merely seek to decompose a system perform- ing a task into subprocesses, but aims to characterize the structures from which the system is composed. More specifically, such mechanistic decompositions aim to explain a phenom- enon (or capacity) by decomposing the system into its functionally salient components, and describing the individual components, the activities in which they engage, and how they interact with each other in order to produce the phenomena in question. Moreover, such structural components are themselves typically decomposed further, in iterative fashion, so that the final analysis describes the system as a hierarchically organized set of causally interacting structural components—i.e., a mechanism. Thus the present explanatory strat- egy advocates that in understanding the mind, cognitive scientists in effect pursue the same approach that might be adopted in explaining the operation of a complex artifact, such as an internal combustion engine or a digital computer. In our view, both functional analysis and mechanistic decomposition are legitimate explanatory enterprises. Indeed, we think that functional analysis is an important step toward developing mechanistic accounts of cognition. Nonetheless, the discovery of mental mechanisms is, as we see it, the central task of contemporary cognitive science. And many of the developments that have occurred over the past few decades—especially those associated with the emergence of cognitive neuroscience—are motivated largely by this explanatory goal. Though the identification of mental mechanisms gives rise to many kinds of issues, one that will figure prominently in later sections of this chapter concerns the extent to which the study of psychopathology can help illuminate the structure and organization of the human mind. In view of this, we propose to conclude our overview of cognitive science with a brief discussion of the methods most commonly deployed in drawing inferences about normal cognition from the study of pathology. Evidence for mechanistic hypotheses: The importance of pathology There are many different sorts of evidence that are relevant to the assessment of hypoth- eses about the structure and organization of the human mind, including chronomet- ric data, developmental evidence, neuroimaging data, and computational simulations. But arguably the major source of evidence, and one central to the topic of this chapter, comes from the study of pathology. For just as experimental interventions can help tell us about the structure of the mind, “natural” interventions, such as disease, nonstandard development, or accidental damage, can provide opportunities to discover properties of the mechanisms normally responsible for our mental capacities (Bechtel 2008; Bechtel and Richardson 1993). A focus on accidental damage is traditionally the business of 27_Fullford_C27.indd 417 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 418 Summoning Concepts cognitive neuropsychology, which has been primarily interested in building models of normal cognition by studying the deficits incurred by patients with acquired brain inju- ries (Caramazza 1986; Coltheart 1985; Ellis and Young 1988; Shallice 1988). And a rel- atively recent variation on this approach, cognitive neuropsychiatry, pursues a similar agenda by focusing on pathological conditions traditionally classified as psychiatric dis- orders (David 1993; Frith 2008; Halligan and David 2001). One central idea common to both cognitive neuropsychology and cognitive neuropsy- chiatry is that the functional organization of the human mind can be discerned by chart- ing associations between pathologies and performance on different cognitive tasks. In particular, it is commonplace to study double dissociations in order to discern the com- ponent mechanisms responsible for cognitive capacities. In brief, double dissociations are instances of paired single dissociations, where a pathological condition (or experimental intervention) affects one particular capacity but not another. More specifically, suppose that we test at least two different patients (or groups) on at least two different tasks. In a double dissociation, a patient (or group) X is impaired on Task A but performs normally on Task B, whereas a patient (or group) Y shows the opposite pattern. In such situations, it is plausible to infer that the patients (or groups) differed in some cognitive variable that was differentially tapped by the two tasks. Moreover, the fact that the dissociation goes in both directions allows us to rule out task difficulty as an explanation of the data.4 As a consequence, it is often plausible to conclude that the best explanation of the pattern is that different cognitive mechanisms are involved in the performance of tasks A and B; and that these mechanisms are differentially impaired (and spared) in the patients (or groups). Archetypical cases of this kind of inference include the comparison of patients with dam- age to Broca’s area to patients with damage to Wernicke’s area, or patients with deficits in long-term memory to those with deficits to short-term memory. (For classic results see Shallice 1988.) The virtues of using dissociative data have been debated extensively elsewhere (Coltheart and Davies 2003; Dunn and Kirsner 2003; Davies 2010; Shallice 1988), and we defer any critical discussion to the section entitled “Two challenges to the cognitive science of psycho- pathology.” For the moment, however, two comments are in order. First, notice that double dissociations appear to provide prima facie evidence for the existence of cognitive mecha- nisms that are distinct, at least to the extent that they are susceptible to selective impairment (and sparing). Thus the study of dissociations appears to provide some basis for claims about the mechanisms responsible for our normal cognitive capacities.5 Second, and by the same token, double dissociations also appear relevant to understanding the cognitive incapaci- ties characteristic of different psychopathologies. For it is by establishing conclusions about deficit cases that one is able to infer conclusions about the normal case. It is to the issue of explaining psychopathology that we now turn. 4 Single dissociations may simply be the products of a single neural region, with two cognitive tasks sharing the same physical location, or the products of “resource artifacts” (Shallice 1988), where one task is impaired simply due to it being more difficult to carry out than the other. 5 These issues are often couched in terms of modularity (e.g., Davies 2005), though in the present instance the relevant notion of modularity is that of separate modifiability. This notion of modularity is significantly different from those deployed by Fodor (1983) and others (e.g., Barrett and Kurzban 2006). 27_Fullford_C27.indd 418 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 419 Explaining Psychopathology So far we have outlined a range of commonplace explanatory strategies in cognitive sci- ence that are routinely combined in various admixtures within different research programs. Specifically, we noted that cognitive capacities and systems can be characterized in the fol- lowing related ways: Intentional (semantic or knowledge) level descriptions Algorithmic level descriptions Implementation (physical) level descriptions Functional analyses Mechanistic descriptions. In this section we set out a range of different ways in which these various strategies might be extended to the case of psychopathology. Specifically, in “Three approaches to the explana- tion of psychopathology” we highlight three importantly different approaches to the expla- nation of psychopathology, and in “A menu of explanatory strategies” we show how these approaches cross-classify with the strategies outlined in “Core explanatory assumptions” in order to yield a menu of different approaches to psychopathology. As with much research in cognitive science, different explanatory strategies are deployed in different combinations. In “Some cognitive accounts of psychopathology: Autism and depression” we illustrate this point by considering a couple of well-known attempts to understand psychopathology in cognitive scientific terms. Three approaches to the explanation of psychopathology There is a number of importantly different ways in which one might apply the resources of cognitive science to the study of psychopathology, which differ in the assumptions they make about the relationship between explanations of normal cognition and those invoked in the case of pathology. In our view, three such approaches are especially important, what we call deficit-based models, input-based models and direct pathology models. Deficit-based models On a deficit-based approach to psychopathology, one starts with a proposal about how the relevant regions of cognition operate in the normal, non-pathological case and then seek to explain the phenomena associated with a psychopathological condition by citing respects in which the normal cognitive system is (selectively) impaired, paradigmatically as a con- sequence of environmental insult or genetic disorder. In the most extreme cases, deficits are explained by hypothesizing the complete breakdown of a given cognitive mechanism. To take a relatively uncontentious case, extreme forms of cortical blindness can be explained in terms of widespread damage to the visual regions of occipital cortex. Or to take another, more contentious example, one might explain prosopagnosia—the inability to recognize 27_Fullford_C27.indd 419 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 420 Summoning Concepts faces—in term of the total breakdown of face recognition mechanisms realized in the fusi- form gyrus (see recent articles in Young et al. 2008). But deficits need not be an all-or-nothing affair. It need not be the case that the system either works exactly as it should or is entirely inoperative. Instead, a mechanism can be damaged in particular respects and to varying degrees; and as a consequence, it may be that the explanatorily relevant deficit is more fine-grained and partial than complete breakdown. For example, it may be the case that in specific language impairment (SLI) only a highly dis- crete component of the language system is damaged, and perhaps only partially so (van der Lely and Marshall 2010). Indeed, the idea that various core pathological phenomena are a result of partial impairment has recently become quite common, for example, in research on delusions (Coltheart et al. 2011; Davies and Coltheart 2000). Input-based models A second approach to the explanation of psychopathology is what we call the input-based approach. Though less commonplace than deficit models, such an approach is still quite familiar from the literature. On this approach, the pathology is not entirely explained in terms of a deficit or breakdown of the normal system, but is instead explained, at least in part, with reference to the character of the inputs to normally functioning cognitive systems, inputs that are in some sense problematic or at least relatively unusual.6 What is the relationship between input-based and deficit models? Both approaches are parasitic on some prior conception of how relevant regions of normal cognition operate. But in contrast to deficit models, input-based models focus on the sorts of inputs that normally functioning systems receive. In some cases, which we might call pure input-based models, the divergence between normal and pathological cases is wholly explained in terms of the character of inputs that normally functioning systems receive. So, for example, in the lit- erature on depressive disorders, both early formulations of the learned helplessness account (Seligman 1975), as well as the social competition hypothesis (Price et al. 1994), maintain that some kinds of depression result from normally functioning cognitive systems receiving inputs that are in some way nonstandard (e.g., inputs that induce in the subject the percep- tion that they lack any control over their circumstances). Similarly, in research on addiction it is routine to explain drug dependencies in terms of the normally functioning reward sys- tem being “hijacked” by chemical stimulants (see, e.g., papers in Poland and Graham 2011). But it is important to note that input-based and deficit-based strategies can be combined in various ways so that the overall explanation of the pathology is a hybrid that recruits both input-based and deficit-based components. For instance, one sort of hybrid model proposes that abnormal inputs to normally functioning developmental mechanisms are responsible for producing malfunctioning cognitive mechanisms. So, for example, in some cases of language impairment, such as “feral” children growing up in highly impoverished linguistic environments, the deficits observed in adulthood are plausibly attributed to the character of the input that the child received in the course of development, as opposed to any malfunction of the learning mechanisms themselves. Nevertheless, the products of this 6 This approach supposes that some pathologies are strongly analogous to what computer scientists sometimes call garbage-in-garbage-out (or GIGO) phenomena. 27_Fullford_C27.indd 420 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 421 developmental process, the mechanisms for language production and comprehension, are also plausibly viewed as pathological. Thus the overall account of such profound linguistic deficits plausibly recruits both an input-based component and a deficit-based component. A second sort of hybrid model proposes that a normally functioning mechanism might be responsible for producing pathological symptoms because the inputs it receives are gener- ated by some other mechanism that is malfunctioning. Such proposals are well known from the literature on delusions. For example, according to Maher’s influential theory of delusion, although delusional beliefs are the output of systems for reasoning, these systems are in no way damaged or malfunctioning (Maher 1988). Instead, normally functioning reasoning systems produce delusional beliefs as a consequence of endeavoring to make sense of the bizarre sensory inputs that they receive from malfunctioning sensory systems. On this view, then, delusion is to be explained by combining a deficit-based account of sensory processing with an input-based account of reasoning processes. Direct pathology models The third and final kind of approach to understanding psychopathology is what we call a direct pathology model. In contrast to deficit and input-based models, direct pathology mod- els do not seek to explain pathology in terms of some deviation from normal conditions. Rather, researchers instead seek to model directly the pathological condition or process. Although direct pathology models are rather less common than the alternatives, it is still the case that a number of cognitive accounts of pathology conform to this explanatory pattern. This is especially so of accounts which hold that a given psychiatric condition does not result from some manipulation to normally functioning cognitive systems, but instead main- tain that the condition results from the existence of a stable polymorph—a distinct form of neurocognitive organization—within the relevant human subpopulation. So, for exam- ple, Annette Karmillof-Smith and her collaborators (D’Souza and Karmiloff-Smith 2011; Karmiloff-Smith 1992) have argued that for people with developmental disorders, such as Williams syndrome, the mature brain may be so different that their cognitive systems will need to be modeled in their own right, as opposed merely to characterized in contrast to neurotypicals. Such a view is also suggested by some of Baron-Cohen’s more recent work on autism (Baron-Cohen 2005). A menu of explanatory strategies In the “Core explanatory assumptions” section we identified a range of related explanatory strategies, commonplace in cognitive science. Then in “Three approaches to the explanation of psychopathology” we drew an orthogonal three-way distinction between approaches to psychopathology, approaches that differ in the assumptions that they make about the rela- tionship between the target pathology and normal cognitive processes. When combined, these distinctions yield a remarkably rich array of explanatory strategies. Indeed, even ignoring the hybrid approaches discussed, there are at least fifteen distinct strategies, each of which has been applied within research on psychopathology. Moreover, these various strategies are routinely combined in various ways by those inter- ested in psychopathology. For example, amongst those who adopt a deficit approach to a 27_Fullford_C27.indd 421 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 422 Summoning Concepts given pathology, it will be natural to seek: (a) an intentional characterization of both the normal process and the deficit, (b) to provide algorithmic and functional characterizations of the normal and pathological processes involved, and (c) to specify the cognitive mech- anisms on which such processes depend (and the brain organization that allows for the realization of such mechanisms). Indeed we suspect that there is a widespread consensus amongst cognitive scientists that all of these various explanatory tasks should ideally be per- formed in the course of comprehensively characterizing a disorder. As a matter of fact, how- ever, much extant research only roughly approximates this ideal, and it is quite uncommon for all of these strategies to receive equal weighting. In the next section we briefly review two examples that help illustrate the mixed—and often incomplete—nature of many cognitive models of psychopathology. Cognitive Accounts of Psychopathology: Autism and Depression In the previous section we provided a taxonomy of different cognitive scientific strategies for explaining and describing psychopathology. We noted that in general these strategies are not incompatible and that they are often combined within a single research program. These strategies have been applied to a broad range of psychopathologies, and we encourage read- ers to see other sources for further illustration (e.g., discussion of delusions in Davies and Egan, Chapter 42, this volume; Murphy 2006). In this section we very briefly sketch some well-known accounts of two forms of psychopathology that have long been the focus of cog- nitive scientific research: autism spectrum disorder (ASD) and major depressive disorder (MDD). In doing so, we make no attempt at being comprehensive; nor do we aim to adjudi- cate between competing proposals. Rather our aim is to illustrate the explanatory strategies outlined earlier. Autism spectrum disorder Although cognitive approaches to ASD have produced a range of competing hypotheses, the most common approach is one in which the social and communicative problems associated with the disorder are caused by impairment to a “theory of mind” mechanism or module: a system normally dedicated to processing information about beliefs, desires and other men- tal states (Baron-Cohen 1995; Frith 1989; Leslie 1994).7 Notice that the sort of model on offer here is deficit-based. As one might expect, then, much of the evidence comes from studies showing that people with ASD are significantly impaired on a range of social cognition tasks, including joint attention (following another’s gaze), the use of pretense, and the understanding of deception (Baron-Cohen 1995). But perhaps the most frequently cited evidence comes from studies of the so-called “false belief ” 7 For a recent commentary on research in this area see Frith (2012). 27_Fullford_C27.indd 422 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 423 task, where Leslie, Baron-Cohen, and others have argued that there is dissociative evidence for the selective impairment of a theory of mind mechanism (or ToMM). In particular, ToMM theorists argue that high functioning children with ASD perform poorly on this task compared to children with Down’s syndrome, even though this latter group have far more profound general cognitive deficits (Baron-Cohen et al. 1985). In elaborating their account of ASD, ToMM theorists operate at a number of descrip- tive levels. So, for example, they provide intentional descriptions of the relevant phenomena (e.g., detailing the supposed inability of patients to form accurate representations of men- tal states), and they also sketch the processing stages involved in, say, determining the con- tent of another’s belief. But perhaps the central focus of this approach is on the mechanistic decomposition of the systems responsible for “mindreading,” and in what follows we focus on this issue. Although ToMM theorists agree that there is a neurocognitive mechanism whose impair- ment is responsible for central aspects of ASD,8 there is considerable disagreement on a range of related issues. One such issue concerns how best to characterize the functional and neuroanatomical organization of ToMM. So, for example, it remains a topic of ongoing research how best to decompose ToMM into its component parts, and what neural regions are primarily responsible for our capacity to attribute mental states to ourselves and to oth- ers (Baron-Cohen 2005; Carruthers 2009; Saxe and Kanwisher 2003). Another issue concerns what mechanisms in addition to ToMM are implicated in “mindreading”. So, for instance, according to Baron-Cohen’s (1995) influential proposal, in addition to ToMM, the mindreading system contains highly specialized mechanisms for processing various sorts of perceptual information, including: an intentionality detector (ID) that interprets the movement of agent-like stimuli in terms of goals and desires; an eye-direction detector (EDD) that detects eye-like visual stimuli and tracks direction of gaze; and a shared attention mechanism (SAM), which takes inputs from ID and EDD and thereby enables the infant to work out whether they are attending to the same thing as another person (Baron-Cohen 1995).9 In contrast, Leslie et al. (2004) have argued that, in addition to whatever specialized modules there are, we must posit a rela- tively unspecialized “selection processor”—an executive system that is, amongst other things, responsible for inhibiting a default bias in normal adults toward attributing true beliefs to others. A third issue concerns the extent to which the phenomena associated with ASD are fully explained as a deficit to ToMM. While it has long been recognized that social defi- cits are central symptoms of this disorder (Wing and Gould 1979), more recent studies indicate that there are a range of other anomalies—for example, in executive function and memory (Happe and Frith 2006; Hill 2004). But the precise nature of the relationship between these anomalies and deficits to ToMM remains unclear. In particular, it remains a matter of active debate whether these anomalies are a downstream, developmental effect of ToMM impairment, or whether they are the products of other relatively independent deficits. 8 For example, there is neuroimaging evidence that areas normally activated during social cognition are underactive in individuals with ASD (Frith and Frith 2003). 9 For further additions and modifications see Baron-Cohen (2005). 27_Fullford_C27.indd 423 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 424 Summoning Concepts A final issue regarding the ToMM hypothesis concerns its reception amongst ASD researchers more broadly. In particular, its adequacy has been challenged on a number of grounds. Some have, for example, recruited evidence about the degree of functional, neu- roanatomical and developmental overlap between theory of mind capacities and executive function more broadly in order to argue that ToMM may not be a specialized, independ- ent system (Carlson and Moses 2001; Perner and Lang 1999). Similarly, Gerrans and Stone (2008) have argued that normal theory of mind capacities are not supported by a dedicated ToMM but rather are the product of the developmental interaction between relatively domain-general, high-level systems, such as those underwriting executive function, and domain specific modules for low-level perceptual processing (though see Adams 2011 for critical discussion). Major depressive disorder Though a wide array of factors have been implicated in the etiology of MDD, most current cognitive models generally adhere to a “vulnerability-stress” conceptualization, according to which depression results from the activation of cognitive biases by stressful life events. Perhaps the most influential version of this approach was originally proposed by Beck (1967, 1976, 1987), with a model based on the idea that an interplay of dysfunctional “schemas” (or belief frameworks) and negative life events can lead to a pattern of negatively biased apprais- als and thoughts. On this view, the characteristic signs and symptoms of MDD—including feelings of hopelessness and despair, suicidal ideation, and anhedonia—are downstream effects of these negatively valenced cognitive states. Notice that the sort of explanation on offer here differs in a range of respects from the models of ASD considered in the “Autism spectrum disorder” section. First, on the most natural interpretation, Beck’s model of MDD is substantially input-based in that the impact of negative life experiences is central to the explanation of depressive episodes. That is, the input-based approach takes depression to be the result of standard-issue cognition having fallen prey to the influence of stress and trauma over time. That said, whether the relevant cognitive “vulnerabilities” constitute a set of underlying deficits remains a point of active enquiry.10 In which case, it may be that a fully articulated version of the proposal will ulti- mately constitute a hybrid model. Second, Beck’s model does not appear to provide any mechanistic decomposition. Instead, the model seems primarily concerned with recruiting familiar concepts from cognitive science, such as the notion of a schema, in order to provide an intentional level description of the processes implicated in MDD. What results is a kind of functional anal- ysis—a decomposition of the process into intentionally characterized subparts—though one largely lacking in precise computational analyses. It should be noted, however, that in recent work Beck and his collaborators (Beck 2008; Clark and Beck 2010) have sought 10 There is a large literature on the cognitive profiles of individuals with MDD, detailing their divergence from the non-depressed on a range of cognitive tasks, most significantly in the domains of attention, memory, and aspects of executive functioning (Gotlib and Joorman 2010; Kircanski et al. 2012). This profile of behavioral variation may then signal underlying deficits of some sort. 27_Fullford_C27.indd 424 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 425 to develop an implementation level account of the neural structures involved in these processes. Two Challenges to the Cognitive Science of Psychopathology In the previous sections we outlined some of the main explanatory strategies to have emerged from cognitive science, and illustrated their application to the study of psycho- pathology. In this section we conclude by briefly considering two challenges to the pros- pects of a developed cognitive science of psychopathology. The first is methodological in character, and concerns the pervasive role of dissociative data in theory construction. The second challenge is rather more philosophical in spirit, and concerns whether cognitive scientific approaches to psychopathology will turn out to be subject to serious in-principle limitations. Challenge 1: The role of dissociative data As noted in the “Core explanatory assumptions” section, it is exceedingly common, espe- cially within cognitive neuropsychology, to use double dissociations in order to provide evidence for the existence of distinct neurocognitive mechanisms, and to identify the under- lying deficits responsible for neurological and psychiatric disorders. Indeed this method of dissociation (or MD) is arguably the main source of evidence for such hypotheses. In recent decades, however, this method has been a focus of much criticism, especially in its applica- tion of psychiatric disorders. In what follows, we outline three such criticisms (for a more comprehensive review see Davies 2010). Objection 1: The instability of symptoms A first worry with applying the MD to psychiatric disorders concerns the relative instability of psychiatric symptoms. While these considerations have not been extensively elaborated, the worry appears to concern the fact that many psychiatric symptoms—for example, sui- cidal and delusional ideation—tend to wax and wane in their occurrence, form, and sever- ity (e.g., see Young 2000 on delusions). In view of this, it may be that the MD cannot be used to support conclusions about the underlying deficits responsible for such symptoms. Specifically, the idea seems to be that if, on the one hand, symptoms are unstable whilst, on the other, neurocognitive mechanisms are stable, enduring structures, then dissociative data will not license inferences regarding which specific underlying deficit is responsible for the symptoms—i.e., which neurocognitive mechanism is damaged. Thus it may be necessary to revise, supplement, or even replace the MD in cognitive neuropsychiatry. Though symptom instability no doubt raises interesting methodological issues, we doubt that it poses a serious challenge to the use of dissociative data in the study of psy- chopathology. One obvious, and in our view convincing, response is that when studying 27_Fullford_C27.indd 425 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 426 Summoning Concepts psychopathology it may often be appropriate to characterize deficits in terms of partial impairment, as opposed to construing symptoms as products of completely impaired sys- tems (again, see work on delusions by Coltheart et al. 2011; Davies and Coltheart 2000). In this way, the fact that some psychiatric symptoms appear to wax and wane is explained by supposing that the dysfunctional informational products of cognitive deficits are occasion- ally over-ridden—or compensated for—by other processes.11 Objection 2: “Impure” cases A second challenge to the MD centers on the notion of impure cases, and the epistemic prob- lems that such cases generate (Van Orden et al. 2001). In brief, a single dissociation is pure when it involves damage to exactly one neurocognitive mechanism; otherwise it is impure. A double dissociation is pure, when it combines two pure single dissociations; otherwise it is impure. In order to appreciate the significance of the distinction, it is important see that pure and impure double dissociations have quite different inferential properties. If a double dissociation is pure, then it deductively follows that there are two distinct mechanisms, each of which is impaired with respect to a task. In which case, if one has strong evidence that a case is pure, then one also has good reason to conclude that there are two, distinct cognitive mechanisms, and that the observed pattern of performance is explained by their selective impairment (and sparing). The logical implications of impure dissociations are quite different. If a dissociation is impure, it does not deductively follow that there are two distinct mechanisms whose selec- tive impairment and sparing is responsible for the pattern of performance. Instead there is a range of alternatives that are compatible with the impure case. So, for example: 1. It is possible that the dissociable tasks depend at least partially on some shared resource. 2. It may be that different components within a single mechanism are differentially affected and that this is responsible for producing different patterns of performance. 3. It is even possible that the dissociable tasks are the products of some kind of continuous processing space that wholly lacks separate and identifiable mechanisms (Shallice 1988). But if this is so, if impure cases are consistent with all these possibilities, then having evi- dence of such a dissociation does not provide good reason to conclude that the pattern of performance is produced by the selective impairment of two distinct mechanisms. That is, impure cases neither provide strong evidence for distinct cognitive mechanisms nor for the underlying deficit responsible for a given pathology. What are we to make of these observations? The appropriate response may seem obvi- ous. If pure cases provide strong evidence and impure cases do not, then in using the MD we should restrict ourselves to pure cases. But this alone does not resolve the matter. First, we should expect the vast majority of dissociations to be impure since there is no reason to 11 And this also leaves space open for the study of the role of factors external to the impaired mechanism, such as environmental stressors, that might explain the more-or-less acute presentation of certain symptoms. 27_Fullford_C27.indd 426 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 427 suppose that the causes of neurocognitive insults—for example, blows to the head or chro- mosomal abnormalities—should respect the boundaries between cognitive mechanisms. In which case, if impure cases fail to provide a basis for dissociative inferences, then it would seem that the vast proportion of data that cognitive neuropsychologists have amassed is of little relevance to the task of discerning cognitive structure. A second and apparently more serious worry is that in order to pursue a policy of restrict- ing ourselves to pure cases, it would seem that we need some reasonably reliable method for determining which cases are pure and which impure. Yet (so the objection continues) there is no such reliable method; or at any rate, we have no good reason to suppose that there is. Indeed some go so far as to claim that there exists no noncircular way of determining whether a given case is pure or otherwise. For in order to know that a double dissociation is pure one must already have established what the dissociative evidence is supposed to show, viz., that there are two distinct mechanisms whose impairment is responsible for the pattern in performance (Van Orden et al. 2001). Thus critics conclude that the MD cannot support conclusions about the mechanistic structure of the mind or about the underlying causes of psychopathology. In our view, the distinction between pure and impure cases does raise problems for the MD, if only because it suggests that such data typically provide rather less support for con- clusions about neurocognitive structure than advocates have traditionally supposed. That said, we are inclined to think that the worries raised by impure cases are somewhat allayed by noting that the MD may well allow for respectable abductive inferences, even in the absence of any highly reliable means of discriminating pure from impure cases (Coltheart and Davies 2003; Davies 2010). This is because although possible alternative explanations will be routinely available, in many cases the best explanation—the simplest, most conserva- tive, most powerful explanation—may well appeal to the existence of functionally distinct neurocognitive mechanisms. Of course, the strength of any particular inference will depend upon the details of the case. For example, it will depend on the degree of dissociation between the disorders being compared and between the neural regions that are associated with these disorders. But if used judiciously, we are inclined to think that, despite the exist- ence of impure cases, the MD can still yield insight into the functional organization of the mind/brain and its pathologies. Objection 3: The problem of developmental dissociations Another potential challenge to the MD, especially as it extends to psychiatry, comes from the study of disorders that are thought to be the product of nonstandard development. Until quite recently, it was commonly assumed that dissociations arising from developmental psychopathology could provide good evidence for the existence of specialized cognitive systems. For example, as we saw in “Explaining psychopathology,” many researchers were led to infer that the impairments in autism were the product of a dysfunctional mechanism normally responsible for theory of mind. Similarly, Williams syndrome, a developmental disorder marked by impairments to a child’s spatial recognition capacities, has garnered considerable theoretical attention in large measure because there appear to be double disso- ciations between Williams and a range of other disorders, including: autism; prosopagnosia, which involves deficits in facial recognition; and SLI, which is characterized by syntactic impairments. Further, these dissociations have been taken to support conclusions regarding 27_Fullford_C27.indd 427 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 428 Summoning Concepts the existence of specific cognitive mechanisms whose selective impairment and sparing is responsible for the various disorders involved. Though such inferences have been challenged on a number of grounds, perhaps the most theoretically interesting concerns the developmental presuppositions implicit in the application of the MD to developmental disorders. Annette Karmiloff-Smith and her collaborators (D’Souza and Karmiloff-Smith 2011; Karmiloff-Smith 1992) have argued, for instance, that the application of the MD to developmental disorders presupposes that developmental anomalies tend only to affect relatively autonomous neurocognitive sys- tems, whilst leaving the rest of the brain “residually normal”. In contrast, Karmiloff-Smith maintains that the brain is more plausibly construed as a massively interconnected system that develops in holistic fashion, so that early emerging pathological disruption will tend to be amplified in the course of develop, and ultimately produce comprehensively atypical neural organization. But if this is so, if developmental disorders, such as Williams, probably result in comprehensively atypical brains, then it is no longer safe to assume that the signs and symptoms associated with such disorders are produced by selective impairments to an otherwise intact or “residually normal” brain. In which case, it would seem that the MD, at least as typically applied, fails to license conclusions about the existence of specific cogni- tive mechanisms or about the underlying causes of psychopathology (for further discus- sion see Machery 2011). Challenge 2: In-principle limitations? Let us turn to the second, more principled, of our challenges. Though the cognitive sciences have made considerable progress in explaining aspects of the human mind and its various disorders, no one would be so sanguine as to suggest that the project is complete, or even nearly so. On the contrary, success to date has been fragmentary; and it remains a largely open question whether—and to what extent—efforts to explain mental disorder will prove successful. In view of this, we propose to conclude this chapter by discussing considerations that some (e.g., Murphy 2006) have recently taken to suggest that the cognitive science of psychopathology may well be subject to serious limitations. The relevant considerations are familiar to cognitive scientists, and received their canoni- cal expression almost three decades ago in the final sections of Fodor’s Modularity of Mind (1983). In this work, Fodor famously advocates a conception of our mental architecture on which peripheral systems for motor control and low-level perception are modular in char- acter. That is, they are highly specialized and informationally encapsulated in the sense that they only utilize a restricted range of the information available to the organism as a whole. In contrast, Fodor maintains that central systems responsible for such “higher” cognitive tasks as reasoning and decision-making are radically non-modular, or unencapsulated: that they can utilize virtually any sort of information in the course of their computations. According to Fodor, this is likely to have important implications for the scope and limits of cognitive science. In particular, he maintains that the sort of cognitive science we currently have, one wedded to an information processing model, is unlikely to make much headway in providing models of highly unencapsulated central processes. Indeed Fodor goes so far as to claim “the limits of modularity are also likely to be the limits of what we are going to be able to understand about the mind, given anything like the theoretical apparatus currently 27_Fullford_C27.indd 428 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 429 available” (Fodor 1983, p. 126). In what follows, we propose to explore the implications of this Fodorian pessimism for the cognitive science of psychopathology. Reasons for Fodorian pessimism? What reasons are there for supposing that non-modular systems will likely prove recalcitrant to the explanatory strategies of cognitive science? Fodor provides two main considerations, and though clearly neither are conclusive, many have found them prima facie compelling. The first focuses on the prospects of providing implementational (e.g., neurobiological or neuroanatomical) descriptions of central systems. If a cognitive system is modular and, hence, processes only a restricted range of inputs, and bears limited informational relations to other systems, then it is reasonable to suppose that it will have a reasonably well-defined neural architecture: that it will be localized and have well-defined connections to other sys- tems. (This appears to be true in the case of early vision, for example.) In contrast, if cen- tral systems are, as Fodor claims, radically non-modular, bearing elaborate informational relations every which way, then they are also unlikely to exhibit a clearly articulated neural architecture. Further, the same will likely be true of the components of such mechanisms. As a consequence, according to Fodor, lesion studies, deficit studies, neuroimaging, and other strategies designed to aid in the identification of structurally characterizable units are unlikely to prove successful. The second consideration focuses on the prospects of providing algorithmic specifications of central processes. On the assumption that cognitive systems are information processing devices, we should expect to be able to specify the sorts of information that a cognitive sys- tem deploys and the sorts of computations involved in the processing of such information. In the case of low-level perceptual processes, the task of doing so appears tractable. We are, for example, able to determine with reasonable clarity what sorts of information are deployed in early vision, and how it is being processed. According to Fodor, however, the problem we confront in understanding central processes is that there appear to be few discernible con- straints on information processing. We are able to think about almost anything, the infor- mational relations appear to go every which way, and they are highly sensitive to context. In which case, the task of specifying such processes in information processing terms will likely prove an extraordinarily difficult one. Psychopathology in the shadow of Fodorian pessimism Though the case for Fodorian pessimism is far from overwhelming, it is suggestive enough to merit consideration of its implications. Let us suppose for the sake of argument, then, that it is correct, and ask what its implications are for a cognitive science of psychopathol- ogy. In our view they are likely to be profound. In particular, Fodorian pessimism has bleak implications for the prospects of cognitive models of psychopathologies that involve central processes. The point is perhaps clearest in the case of deficit-based models that seek to describe the pathology in algorithmic or mechanistic terms. Such models aim to explain how some path- ological phenomenon is produced by citing some abnormality—or malfunction—in those mechanisms or computational processes responsible for normal cognition. But such models presuppose some prior account of normal cognition. In which case if Fodorian pessimism 27_Fullford_C27.indd 429 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 430 Summoning Concepts is correct, then such deficit-based models will be not be forthcoming for pathologies that involve central cognition. This will simply be because there will be no mechanistic or algo- rithmic account of normal function to work from in the first place. Much the same is true of input-based models of psychopathologies that involve central processes. In such models one seeks to explain some pathological phenomena in terms of abnormal inputs to normally functioning mechanisms or processes. But again, such explanations presuppose a prior mechanistic account of normal function. In which case if Fodorian pessimism is correct about central processes, no such models of the psychopathol- ogies will be forthcoming. The issues are perhaps least clear in the case of direct pathology models. Rather than pre- supposing some account of normal cognition, such models instead proceed directly by con- structing an account of the pathology itself or of the mechanism(s) that are responsible for the pathology in question. It is harder to assess the implications of Fodorian pessimism for such approaches, in large measure because it remains unclear, for any pathology, how much of central cognition needs to be characterized in order to provide such models. As a conse- quence, one cannot argue directly from Fodorian pessimism to conclusions about the pros- pects of such models. For all that, if Fodorian pessimism is correct, we should expect that direct models of pathologies that crucially depend upon central cognition—for example, thought disorders and delusional disorders—will be hard to produce in much the same way as indirect (deficit and input-based) models are. Of course, none of this precludes the possibility of models of psychopathology that do not seek to provide mechanistic or algorithmic descriptions of the pathologies involved. But for present purposes, the point we wish to stress is that the search for such descriptions is very typically the ultimate goal of cognitive science. And what goes for cognitive science in general also goes for cognitive scientific explanations of psychopathology. In which case, if Fodorian pessimism is correct, this explanatory quest may turn out to be forlorn. References Adams, M. (2011). Modularity, theory of mind, and autism spectrum disorder, Philosophy of Science, 78, 763–73. Baron-Cohen, S. (1995). Mindblindness. Cambridge, MA: MIT Press. Baron-Cohen, S. (2005). The empathizing system: A revision of the 1994 model of the mindreading system. In B. Ellis and D. Bjorklund (Eds), Origins of the Social Mind, pp. 468–92. New York, NY: Guilford Publications Inc. Baron-Cohen, S., Leslie, A., and Frith, U. (1985). Does the autistic child have a ‘theory of mind’? Cognition, 21, 37–46. Barrett, H. C. and Kurzban, R. (2006). Modularity in cognition: framing the debate. Psychological Review, 113(3), 628–47. Bechtel, W. and Richardson, R. C. (1993). Discovering Complexity: Decomposition and Localization as Strategies in Scientific Research. Princeton, NJ: Princeton University Press. Bechtel, W. (2008). Mental Mechanisms: Philosophical Perspectives on Cognitive Neuroscience. New York, NY: Routledge. Beck, A. T. (1967). Depression: Causes and Treatment. Philadelphia, PA: University of Pennsylvania Press. 27_Fullford_C27.indd 430 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 431 Beck, A. T. (1976). Cognitive Therapy and the Emotional Disorders. New York, NY: International University Press. Beck, A. T. (1987). Cognitive models of depression. Journal of Cognitive Psychotherapy, 1, 5–37. Beck, A. T. (2008). The evolution of the cognitive model of depression and its neurobiological correlates. American Journal of Psychiatry, 165, 969–77. Broome, M. and Bortolotti, L. (2009). Psychiatry as Cognitive Neuroscience: Philosophical Perspectives. Oxford: Oxford University Press. Caramazza, A. (1986). On drawing inferences about the structure of normal cognitive systems from the analysis of patterns of impaired performance: the case for single patient studies. Brain and Cognition, 5, 41–66. Carey, S. (2009). The Origin of Concepts. Oxford University Press: Oxford. Carlson, S. M. and Moses, L. J. (2001). Individual differences in inhibitory control and children’s theory of mind. Child Development, 72(4), 1032–53. Carruthers, P. (2009). How we know our own minds: the relationship between mindreading and metacognition. Behavioral and Brain Sciences, 32, 121–82. Chemero, A. (2009). Radical Embodied Cognitive Science. Cambridge, MA: MIT Press. Clark, A. (2000). Mindware: An Introduction to the Philosophy of Cognitive Science. New York, NY: Oxford University Press. Clark, A. (2008). Supersizing the Mind: Embodiment, Action, Cognitive Extension. New York, NY: Oxford University Press. Clark, D. A. and Beck, A. T. (2010). Cognitive theory and therapy of anxiety and depression: Convergence with neurobiological findings. Trends in Cognitive Sciences, 14, 418–24. Coltheart, M. (1985). Cognitive neuropsychology and the study of reading. In M. I. Posner and O. S. M. Marin (Eds), Attention and Performance XI, pp. 3–37. Hillsdale, NJ: Erlbaum. Coltheart, M. and Davies, M. (2003). Inference and explanation in cognitive neuropsychology. Cortex, 39, 188–91. Coltheart, M., Langdon, R., and McKay, R. (2011). Delusional belief. Annual Review of Psychology, 62, 271–98. Cummins, R. (1975). Functional analysis. Journal of Philosophy, 72, 741–64. Cummins, R. (2000) “How does it work?” vs. “What are the laws?” Two conceptions of psychological explanation. In F. Keil and R. Wilson (Eds), Explanation and Cognition, pp. 117–45. Cambridge, MA: MIT Press. Cummins, D. D. and Cummins, R. (2000). Minds, Brains and Computers. Oxford: Blackwell. David, A. S. (1993). Cognitive neuropsychiatry? Psychological Medicine, 23, 1–5. Davies, M. (2005). Cognitive science. In F. Jackson and M. Smith (Eds), The Oxford Handbook of Contemporary Analytic Philosophy, pp. 358–94. Oxford: Oxford University Press. Davies, M. (2010). Double dissociations. Mind & Language, 25(5), 500–40. Davies, M. and Coltheart, M. (2000). Pathologies of belief. Mind and Language, 15(1), 1–46. D’Souza, D. and Karmiloff-Smith, A. (2011). When modularization fails to occur: A developmental perspective. Cognitive Neuropsychology, 28(3&4), 276–87. Dunn, J. C. and Kirsner, K. (2003). What can we infer from double dissociations? Cortex, 39, 1–7. Ellis, A. W. and Young, A. W. (1988). Human Cognitive Neuropsychology. Hove: Erlbaum. Fodor, J. (1968). Psychological Explanation. New York, NY: Random Rouse. Fodor, J. (1983). The Modularity of Mind: An Essay on Faculty Psychology. Cambridge, MA: MIT Press. Fodor, J. and Pylyshyn, Z. (1988). Connectionism and cognitive architecture: A critique. Cognition, 28, 3–71. 27_Fullford_C27.indd 431 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN 432 Summoning Concepts Frith, C. (2008). In praise of cognitive neuropsychiatry. Cognitive Neuropsychiatry, 13(1), 1–7. Frith, U. (1989). Autism: Explaining the Enigma (2nd edn 2003). London: Blackwell. Frith, U. (2012). Why we need cognitive explanations of autism. The Quarterly Journal of Experimental Psychology, 65(11), 2073–92. Frith, U. and Frith, C. D. (2003). Development and neurophysiology of mentalizing. Philosophical Transactions of the Royal Society B: Biological Sciences, 358(1431), 459–73 Gazzaniga, M. S. (2009). The Cognitive Neurosciences (4th edn). Cambridge, MA: MIT Press. Gerrans, P. and Stone, V. E. (2008). Generous or parsimonious cognitive architecture? Cognitive neuroscience and theory of mind. British Journal of Philosophy of Science, 59, 121–41. Gotlib, I. H. and Joorman, J. (2010). Cognition and depression: Current status and future directions. Annual Review of Clinical Psychology, 6, 285–312. Halligan, P. W. and David, A. S. (2001). Cognitive neuropsychiatry: towards a scientific psychotherapy. Neuroscience: Nature Reviews, 2, 209–15. Happe, F. and Frith, U. (2006). The weak coherence account: detail-focused cognitive style in autism spectrum disorders. Journal of Autism and Developmental Disorders, 36(1), 5–25. Hill, E. L. (2004). Executive dysfunction in autism. Trends in Cognitive Sciences, 8(10), 26–32. Karmiloff-Smith, A. (1992). Beyond Modularity: A Developmental Perspective on Cognitive Science. Cambridge, MA: MIT Press. Kendler, K. S. (2008). Explanatory models for psychiatric illness. American Journal of Psychiatry, 165, 695–702. Kendler, K. S., Zachar, P., and Craver, C. (2011). What kinds of things are psychiatric disorders? Psychological Medicine, 41, 1143–50. Kircanski, K., Joorman, J., and Gotlib, I. H. (2012). Cognitive aspects of depression. WIREs Cognitive Science, 3, 301–13. Leslie, A. (1994). ToMM, ToBY, and agency: core architecture and domain specificity. In L. Hirschfeld and S. Gelman (Eds), Mapping the Mind, pp. 119–48. Cambridge: Cambridge University Press. Leslie, A. M., Friedman, O., and German, T. P. (2004). Core mechanisms in ‘theory of mind.’ Trends in Cognitive Sciences, 8(12), 528–33. Machery, E. (2011). Developmental disorders and cognitive architecture. In P. Adriaens and A. De Block (Eds), Maladapting Minds: Philosophy, Psychiatry, and Evolutionary Theory, pp. 91–116. Oxford: Oxford University Press. Maher, B. (1988). Anomalous experience and delusional thinking: the logic of explanations. In T. F. Oltmanns and B. A. Maher (Eds), Delusional Beliefs, pp. 15–33. New York, NY: Plenum Press. Marr, D. (1982). Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. Cambridge, MA: MIT Press. Murphy, D. (2006). Psychiatry in the Scientific Image. Cambridge, MA: MIT Press. Newell, A. (1990). Unified Theories of Cognition. Cambridge, MA: Harvard University Press. Perner, J. and Lang, B. (1999). Development of theory of mind and executive control. Trends in Cognitive Sciences, 3(9), 337–44. Piccinini, G. and Scarantino, A. (2011). Information processing, computation, and cognition. Journal of Biological Physics, 37(1), 1–38. Pinker, S. (1994). The Language Instinct. New York, NY: William Morrow. Poland, J. and Graham, G. (2011). Addiction and responsibility. Cambridge, MA: MIT Press. Price, J., Sloman, L., Gardner, R., Gilbert, P., and Rohde, P. (1994). The social competition hypothesis of depression. British Journal of Psychiatry, 164, 309–15. Pylyshyn, Z. (1984). Computation and cognition. Cambridge, MA: MIT Press. 27_Fullford_C27.indd 432 4/10/2013 8:28:37 AM OUP UNCORRECTED PROOF – REVISES, Wed Apr 10 2013, NEWGEN COGNITIVE SCIENCE AND EXPLANATIONS OF PSYCHOPATHOLOGY 433 Saxe, R. and Kanwisher, N. (2003). People thinking about thinking people: the role of the temporo-parietal junction in ‘theory of mind.’ NeuroImage, 19, 1835–42. Seligman, M. E. P. (1975). Helplessness: On Depression, Development, and Death. San Francisco, CA: W. H. Freeman. Shallice, T. (1988). From Neuropsychology to Mental Structure. Cambridge: Cambridge University Press. Smolensky, P. (1988). On the proper treatment of connectionism. Behavioral and Brain Sciences, 11, 1–23. Thagard, P. (2012). Cognitive science. In E. N. Zalta (Ed.), The Stanford Encyclopedia of Philosophy (Fall 2012 Edition). [Online.] Available at: <http://plato.stanford.edu/archives/ fall2012/entries/cognitive-science/>. van der Lely, H. K. J and Marshall, C. (2010). Grammatical-specific language impairment: A window onto domain specificity. In J. Guendouzi, F. Loncke, and M. Williams (Eds), Handbook of Psycholinguistics and Cognitive Processing Perspectives in Communication Disorders, pp. 403–19. Hove: Psychology Press. Van Orden, G. C., Pennington, B. F., and Stone, G. O. (2001). What do double dissociations prove? Cognitive Science, 25(1), 111–72. Von Eckardt, B. (1993). What is Cognitive Science? Cambridge, MA: MIT Press. Wing, L. and Gould, J. (1979). Severe impairments of social interaction and associated abnormalities in children: epidemiology and classification. Journal of Autism and Develop- mental Disorders, 9, 11–29. Young, A. W. (2000). Wondrous strange: the neuropsychology of abnormal beliefs. Mind and Language, 15(1), 47–73. Young, A. W., de Haan, E. H. F., and Bauer, R. M. (2008). Face perception: a very special issue. Journal of Neuropsychology, 2(1), 1–14. 27_Fullford_C27.indd 433 4/10/2013 8:28:38 AM