- Email: [email protected]
Category-Specificity: What is the Question?
FORUM ON “METHODOLOGICAL CRITIQUE ON CATEGORY SPECIFICITY” CATEGORY-SPECIFICITY: WHAT IS THE QUESTION? John C. Marshall1 and Jennifer M. Gurd1,2 (1University Department of Clinical Neurology, Oxford, UK; 2Psychology Department, University of Hertfordshire, Hatfield, UK)
There are very good reasons, metaphysical, evolutionary and developmental (Gelman, 1990), why the human mind should distinguish between living creatures and inanimate objects. It would certainly be difficult for us to do without the distinction (c.f., autism) despite the fact that cultures may draw the line differently (c.f., animism) and that some `things’ may have a borderline status (c.f., viruses). Keith Laws’ (2005) fascinating critique adopts no view on the psychological nature of the living/nonliving distinction; rather, as he states, Laws is just deeply unconvinced by the current evidence for a double-dissociation between performance on naming exemplars of the two categories after neurological disease or trauma. The criticisms that Laws makes of the current literature appear well-taken. His figure 1 is a seemingly horrifying set of examples of how the failure to use appropriate control data can distort the theoretical interpretation of patients’ differential performance on two semantic categories. There are, no doubt, many areas of neuropsychology other than the study of category effects that have committed similar sins. The presence of ceiling effects when normal subjects are tested with exactly the same material as patients is ubiquitous. To increase significantly the difficulty level of the material may result in floor effects in the patients’ performance; to use different materials or different experimental paradigms for patients and controls would raise further interpretive problems of their own. But does Laws’ fastidiousness risk throwing the baby out with the bathwater? There is, one might argue, a sense in which the existence of ceiling effects in normal subjects somehow strengthens the significance (in the non-statistical meaning of the term) of the patient’s deficit: Consider, for example, patient T.O.B. of McCarthy and Warrington (1988). After left temporal damage, this senior civil servant could only define the spoken word “pig” as an “animal”; by contrast, his definition of “lighthouse” was “Round the coast, built up, tall building, lights revolve to warn ships”. Although this case does not fall within the purview of Laws’ critique (the dissociation involves defining, not naming), it does raise many of the same issues: What could possibly be an appropriate set of control data for such a singular inability to give reasonably precise definitions of “rhinoceros” Cortex, (2005) 41, 861-862
(Patient: “animal”), “dolphin” (Patient: “a fish or bird”) and “peacock” (Patient: “common name, can’t place it”) in the context of being able to provide more than adequate definitions of inanimate objects? Choosing relatively commonplace stimuli makes it plausible to assume that T.O.B. would have known the above words prior to his brain damage, and hence to regard his current performance as indeed pathological. With rarer words, the patient’s premorbid interests will likely determine to a considerable extent the observed dissociations. Thus with respect to the stimuli used in Experiment 2b of Laws et al. (2005), one of the current authors doesn’t know his armadillo from his platypus but could name or define a demijohn and a tantalus without hesitation. There is accordingly a case to be made for not choosing such stimuli in patient studies and simply living with ceiling effects in the control subjects. Neuropsychology has often relied on the method of extreme cases: That is, the study of patients who are seriously impaired on tasks that healthy adults find trivially easy. In optimal viewing conditions and with no pressure to respond quickly, normal subjects do not make semantic errors when reading individual words; nor do they draw clockfaces that only include the numbers 12, 1, 2, 3, 4, 5 and 6. Nonetheless, investigations of neurological patients who do perform in the above ways can throw light upon the functional architecture of the normal reading system (Marshall and Newcombe, 1966) and of the normal mechanisms of spatial attention (Marshall and Halligan, 1993). Such considerations suggest that one could profitably broaden the question that underlies Laws’ critique. One cannot doubt that normal people can and do distinguish between living creatures and inanimate objects; neither can one doubt (on pain of dualism) that the ability to make that distinction depends upon the central and peripheral nervous system (and the environment in which the owner of that nervous system finds him/herself). Accordingly, the deeper theoretical question is whether knowledge pertaining to animates and knowledge pertaining to inanimates is represented in the brain in such a way that (relatively) focal brain damage can impair either one of those categories (relatively) independently of the other. When generalized in this way, the issue is not
862
John C. Marshall and Jennifer M. Gurd
whether naming impairments after brain lesions support the double dissociation but whether any evidence at all supports the claim that knowledge of animates and of inanimates is differentially represented in discrete, punctate (possibly multiple) brain regions. One relevant source of evidence is clearly functional neuroimaging. Again, however, as in Laws’ analysis of the lesion studies, much of the evidence for category-specific activations has been weak and inconclusive (Price and Friston, 2002). The most reliable finding seems to be significant activation of left posterior middle temporal region when the tasks involve knowledge of tools (and body parts). Devlin et al. (2002) also suggest that, when word-retrieval and semantic decision tasks are used, the category of living things differentially activates medial aspects of the anterior temporal poles bilaterally. It is perhaps too early to foreclose on the issue of whether or not knowledge of animates and inanimates reliably localizes to distinct cortical regions, but Price and Friston (2002) have argued very reasonably that such knowledge may well be distributed across many interconnected regions: “For instance, animal-specific effects might result from increases in the coupling between two or more regions that also respond (in an uncorrelated way) to other object categories” (Price and Friston, 2002). If such a proposal were true, conceptual categories would be represented by the
functional integration of discrete, specialized brain regions. This form of localization of function would not give rise to obvious, clear-cut double dissociations when lesioned (Gurd and Marshall, 2003). Theoretically, however, localization of this nature is closer to traditional (Gallist) notions than it is to the wilder shores of connectionism. REFERENCES DEVLIN JT, MOORE CJ, MUMMERY CJ, GORNO-TEMPINI ML, PHILLIPS JA, NOPPENEY U, FRACKOWIAK RSJ, FRISTON KJ and PRICE CJ. Anatomic constraints on cognitive theories of category specificity. NeuroImage, 15: 675-685, 2002. GELMAN R. First principles organize attention to and learning about relevant data: Number and the animate/inanimate distinction as examples. Cognitive Science, 14: 79-106, 1990. GURD JM and MARSHALL JC. Dissociations: Double or quits? Cortex, 39: 192-195, 2003. LAWS KR. ‘Illusions of normality’: A methodological critique of category specific-naming. Cortex, 41: 842-851, 2005. LAWS KR, GALE TM, LEESON VC and CRAWFORD JR. When is category specific in Alzheimer’s disease? Cortex, 41: 452463, 2005. MCCARTHY RA and WARRINGTON EK. Evidence for modalityspecific meaning systems in the brain. Nature, 334: 428-430, 1988. MARSHALL JC and HALLIGAN PW. Imagine only the half of it. Nature, 364: 193-194, 1993. MARSHALL JC and NEWCOMBE F. Syntactic and semantic errors in paralexia. Neuropsychologia, 4: 169-176, 1966. PRICE CJ and FRISTON KJ. Functional imaging studies of category specificity. In Forde EME and Humphreys GW (Eds), Category Specificity in Brain and Mind. Hove, East Sussex: Psychology Press, 2002, pp. 427-447. John C. Marshall, Neuropsychology Unit, University Department of Clinical Neurology, Radcliffe Infirmary, Oxford OX2 6HE, UK. e-mail: [email protected]
There are very good reasons, metaphysical, evolutionary and developmental (Gelman, 1990), why the human mind should distinguish between living creatures and inanimate objects. It would certainly be difficult for us to do without the distinction (c.f., autism) despite the fact that cultures may draw the line differently (c.f., animism) and that some `things’ may have a borderline status (c.f., viruses). Keith Laws’ (2005) fascinating critique adopts no view on the psychological nature of the living/nonliving distinction; rather, as he states, Laws is just deeply unconvinced by the current evidence for a double-dissociation between performance on naming exemplars of the two categories after neurological disease or trauma. The criticisms that Laws makes of the current literature appear well-taken. His figure 1 is a seemingly horrifying set of examples of how the failure to use appropriate control data can distort the theoretical interpretation of patients’ differential performance on two semantic categories. There are, no doubt, many areas of neuropsychology other than the study of category effects that have committed similar sins. The presence of ceiling effects when normal subjects are tested with exactly the same material as patients is ubiquitous. To increase significantly the difficulty level of the material may result in floor effects in the patients’ performance; to use different materials or different experimental paradigms for patients and controls would raise further interpretive problems of their own. But does Laws’ fastidiousness risk throwing the baby out with the bathwater? There is, one might argue, a sense in which the existence of ceiling effects in normal subjects somehow strengthens the significance (in the non-statistical meaning of the term) of the patient’s deficit: Consider, for example, patient T.O.B. of McCarthy and Warrington (1988). After left temporal damage, this senior civil servant could only define the spoken word “pig” as an “animal”; by contrast, his definition of “lighthouse” was “Round the coast, built up, tall building, lights revolve to warn ships”. Although this case does not fall within the purview of Laws’ critique (the dissociation involves defining, not naming), it does raise many of the same issues: What could possibly be an appropriate set of control data for such a singular inability to give reasonably precise definitions of “rhinoceros” Cortex, (2005) 41, 861-862
(Patient: “animal”), “dolphin” (Patient: “a fish or bird”) and “peacock” (Patient: “common name, can’t place it”) in the context of being able to provide more than adequate definitions of inanimate objects? Choosing relatively commonplace stimuli makes it plausible to assume that T.O.B. would have known the above words prior to his brain damage, and hence to regard his current performance as indeed pathological. With rarer words, the patient’s premorbid interests will likely determine to a considerable extent the observed dissociations. Thus with respect to the stimuli used in Experiment 2b of Laws et al. (2005), one of the current authors doesn’t know his armadillo from his platypus but could name or define a demijohn and a tantalus without hesitation. There is accordingly a case to be made for not choosing such stimuli in patient studies and simply living with ceiling effects in the control subjects. Neuropsychology has often relied on the method of extreme cases: That is, the study of patients who are seriously impaired on tasks that healthy adults find trivially easy. In optimal viewing conditions and with no pressure to respond quickly, normal subjects do not make semantic errors when reading individual words; nor do they draw clockfaces that only include the numbers 12, 1, 2, 3, 4, 5 and 6. Nonetheless, investigations of neurological patients who do perform in the above ways can throw light upon the functional architecture of the normal reading system (Marshall and Newcombe, 1966) and of the normal mechanisms of spatial attention (Marshall and Halligan, 1993). Such considerations suggest that one could profitably broaden the question that underlies Laws’ critique. One cannot doubt that normal people can and do distinguish between living creatures and inanimate objects; neither can one doubt (on pain of dualism) that the ability to make that distinction depends upon the central and peripheral nervous system (and the environment in which the owner of that nervous system finds him/herself). Accordingly, the deeper theoretical question is whether knowledge pertaining to animates and knowledge pertaining to inanimates is represented in the brain in such a way that (relatively) focal brain damage can impair either one of those categories (relatively) independently of the other. When generalized in this way, the issue is not
862
John C. Marshall and Jennifer M. Gurd
whether naming impairments after brain lesions support the double dissociation but whether any evidence at all supports the claim that knowledge of animates and of inanimates is differentially represented in discrete, punctate (possibly multiple) brain regions. One relevant source of evidence is clearly functional neuroimaging. Again, however, as in Laws’ analysis of the lesion studies, much of the evidence for category-specific activations has been weak and inconclusive (Price and Friston, 2002). The most reliable finding seems to be significant activation of left posterior middle temporal region when the tasks involve knowledge of tools (and body parts). Devlin et al. (2002) also suggest that, when word-retrieval and semantic decision tasks are used, the category of living things differentially activates medial aspects of the anterior temporal poles bilaterally. It is perhaps too early to foreclose on the issue of whether or not knowledge of animates and inanimates reliably localizes to distinct cortical regions, but Price and Friston (2002) have argued very reasonably that such knowledge may well be distributed across many interconnected regions: “For instance, animal-specific effects might result from increases in the coupling between two or more regions that also respond (in an uncorrelated way) to other object categories” (Price and Friston, 2002). If such a proposal were true, conceptual categories would be represented by the
functional integration of discrete, specialized brain regions. This form of localization of function would not give rise to obvious, clear-cut double dissociations when lesioned (Gurd and Marshall, 2003). Theoretically, however, localization of this nature is closer to traditional (Gallist) notions than it is to the wilder shores of connectionism. REFERENCES DEVLIN JT, MOORE CJ, MUMMERY CJ, GORNO-TEMPINI ML, PHILLIPS JA, NOPPENEY U, FRACKOWIAK RSJ, FRISTON KJ and PRICE CJ. Anatomic constraints on cognitive theories of category specificity. NeuroImage, 15: 675-685, 2002. GELMAN R. First principles organize attention to and learning about relevant data: Number and the animate/inanimate distinction as examples. Cognitive Science, 14: 79-106, 1990. GURD JM and MARSHALL JC. Dissociations: Double or quits? Cortex, 39: 192-195, 2003. LAWS KR. ‘Illusions of normality’: A methodological critique of category specific-naming. Cortex, 41: 842-851, 2005. LAWS KR, GALE TM, LEESON VC and CRAWFORD JR. When is category specific in Alzheimer’s disease? Cortex, 41: 452463, 2005. MCCARTHY RA and WARRINGTON EK. Evidence for modalityspecific meaning systems in the brain. Nature, 334: 428-430, 1988. MARSHALL JC and HALLIGAN PW. Imagine only the half of it. Nature, 364: 193-194, 1993. MARSHALL JC and NEWCOMBE F. Syntactic and semantic errors in paralexia. Neuropsychologia, 4: 169-176, 1966. PRICE CJ and FRISTON KJ. Functional imaging studies of category specificity. In Forde EME and Humphreys GW (Eds), Category Specificity in Brain and Mind. Hove, East Sussex: Psychology Press, 2002, pp. 427-447. John C. Marshall, Neuropsychology Unit, University Department of Clinical Neurology, Radcliffe Infirmary, Oxford OX2 6HE, UK. e-mail: [email protected]