Functional Neuroanatomy of the Semantic System: Divisible by What? C. J. Mummery Wellcome Department of Cognitive Neurology K. Patterson and J. R. Hodges Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/6/766/1758408/089892998563059.pdf by guest on 18 May 2021 MRC Applied Psychology Unit, Cambridge C. J. Price Wellcome Department of Cognitive Neurology Abstract ■ Studies of patients with brain damage suggest that speciªc a similarity judgment about the number of syllables in the brain regions may be differentially involved in repre- target and response words. All tasks were performed under senting/processing certain categories of conceptual knowl- two different stimulus conditions: names of living things and edge. With regard to the dissociation that has received the most names of artifacts. Judgments for both domains and both attrib- attention—between the domains of living things and arti- ute types activated an extensive, distributed, left-hemisphere facts—a debate continues as to whether these category-spe- semantic system, but showed some differential activation- ciªc effects reºect neural implementation of categories directly particularly as a function of attribute type. The left temporo- or some more basic properties of brain organization. The pre- occipito-parietal junction showed enhanced activity for judg- sent positron emission tomography (PET) study addressed this ments about object location, whereas the left anteromedial issue by probing explicitly for differential activation associated temporal cortex and caudate nucleus were differentially acti- with written names of objects from the domains of living vated by color judgments. Smaller differences were seen for things or artifacts during similarity judgments about different living and nonliving domains, the positive ªndings being largely attributes of these objects. Subjects viewed triads of written consistent with previous studies using objects; in particular, object names and selected one of two response words as more words denoting artifacts produced enhanced activation in the similar to a target word according to a speciªed perceptual left posterior middle temporal gyrus. These results suggest that, attribute (typical color of the objects) or an associative attrib- within a distributed conceptual system activated by words, the ute (typical location of the objects). The control task required more prominent neural distinction relates to type of attribute. ■ INTRODUCTION 1992; Sartori, Job, Miozzo, Zago, & Marchiori, 1993; Silveri & Gainotti, 1988; Warrington & Shallice, 1984). It has The issue of where and how the brain represents knowl- been suggested that this dissociation might reºect a bias edge about the world is a topic of continuing debate in in some variable such as familiarity that inherently favors cognitive neuroscience. Studies of normal subjects and everyday artifacts, and indeed not all of the apparent of patients with brain damage suggest that there may be reports of a deªcit for living things have survived strin- signiªcant distinctions in the cortical representation of gent stimulus matching (Funnell & Sheridan, 1992; Sar- different types of information. The most thoroughly tori, Coltheart, Miozzo, & Job, 1994). There are, however, documented conceptual distinction is between the do- some compelling demonstrations of this dissociation main of living things (including animals, fruits, vegeta- based on careful stimulus selection. Furthermore, the bles) on the one hand and manufactured artifacts (such existence of cases with the opposite pattern (i.e., a as tools, household objects, vehicles, etc.) on the other. deªcit for artifacts relative to living things: Caramazza & There are now numerous reports of patients, most with Shelton, 1998; Hillis & Caramazza, 1991; Sacchett & Hum- bilateral inferomedial temporal lobe damage secondary phreys, 1992; Warrington & McCarthy, 1987) rules out to herpes simplex virus encephalitis, with signiªcantly any simple familiarity imbalance as an explanation of a impaired semantic performance on living things relative selective living-things impairment. to artifacts (Funnell & Davies, 1996; Hart & Gordon, Two main classes of explanation have been advanced © 1998 Massachusetts Institute of Technology Journal of Cognitive Neuroscience 10:6, pp. 766–777 Downloaded from http://www.mitpressjournals.org/doi/pdf/10.1162/089892998563059 by guest on 02 October 2021 for the living thing-artifact dissociation. The most trans- activation in normal individuals, there are often substan- parent account suggests a categorical organization of tial differences between studies in the nature of stimulus semantic information in the brain; that is, living things materials, tasks, or both. Such diversity is not surprising and artifacts might be so genuinely different that knowl- and indeed should eventually assist progress toward a edge about the two domains is separately neurally im- full characterization of the relevant phenomena; but at plemented (Caramazza & Shelton, 1998; Laiacona, present, when so little of the experimental parametric Barbarotto, & Capitani, 1993). An alternative and now space has been explored, these major differences tend more widely accepted account argues that there are to impede neat comparisons and summaries. substantial differences in the types of semantic features In the limited number of functional imaging studies that constitute the core meaning of instances of these contrasting objects from the living thing/artifact do- Downloaded from http://mitprc.silverchair.com/jocn/article-pdf/10/6/766/1758408/089892998563059.pdf by guest on 18 May 2021 two broad categories. If living things are distinguished mains, some degree of consistency has emerged with primarily by their perceptual (mainly visual) features and respect to brain regions showing enhanced activation artifacts are more speciªed by functional/associative at- for artifacts (especially tools) relative to living things. tributes (how an object is used, where it is found, etc.), Experiments employing either picture naming or verbal the observed category effects may result from selective ºuency for semantic categories have produced greater disruption to neural regions specialized for these two activation for artifacts than for living things in left pos- types of featural information (Farah & McClelland, 1991; terior temporal or temporo-parietal areas (Damasio, Gainotti, Silveri, Daniele, & Giustolisi, 1995; Garrard, Pat- Grabowski, Tranel, Hichwa, & Damasio, 1996; Martin, terson, Watson, & Hodges, 1998; Saffran & Schwartz, Wiggs, Ungerleider, & Haxby, 1996; Mummery, Patterson, 1994; Warrington and Shallice, 1984). Consistent with Hodges, & Wise, 1996) and also in a left inferior frontal this conceptual division are reports of deªcits in percep- region (BA 44/6) (Grabowski, Damasio, & Damasio, 1997; tual relative to associative knowledge (Silveri & Gainotti, Martin et al., 1996). A further study by Perani et al. 1988). (1995), involving same-different judgments on pictures In addition to these two principal accounts of cate- of animals or tools rather than a naming task, revealed gory-speciªc effects, the literature offers several other left frontal activity for tools (albeit 24 mm distant) but proposals regarding this distinction. For example, Hum- no posterior temporal locus. The opposite comparison phreys and colleagues have drawn attention to impor- has tended to yield somewhat weaker and less consis- tant differences between the two domains in the degree tent results. When participants in the Damasio et al. and detail of visual processing required to distinguish (1996) study named animals (in comparison to naming one member of a category from another. Animals or tools), peak activation was found in the left inferior fruits and vegetables have more striking within-category temporal lobe, still relatively posterior but somewhat structural similarities than artifacts, and thus are thought anterior and medial to that observed for naming tools. to place greater demands on a presemantic “structural Mummery et al. (1996) observed enhanced bilateral an- description” system for object recognition (Forde, Fran- terior medial temporal activation when subjects gener- cis, Riddoch, Rumiati, & Humphreys, 1997; Humphreys, ated names of living things; both Martin et al. (1996) and Lamote, & Lloyd Jones, 1995). Furthermore, some recent Perani et al. (1995), however, reported more left unilat- accounts based on principles of distributed neural net- eral medial occipital activation associated with viewing works focus on differences between the domains in pictures of animals. degree and pattern of intercorrelation between different Even fewer functional imaging studies have contrasted elements of our knowledge about an object (Devlin, different types of semantic features. Vandenberghe, Price, Gonnerman, Andersen, & Seidenberg, 1998; Durrant- Wise, Josephs, and Frackowiak (1996) observed no selec- Peatªeld, Tyler, Moss, & Levy, 1997; Gonnerman, An- tive activation for either perceptual or associative simi- dersen, Devlin, Kempler, & Seidenberg, 1997; McRae, de larity judgments about triads of pictures or words. On Sa, & Seidenberg, 1997). It is by no means clear at this the other hand, when Martin, Haxby, Lalonde, Wiggs, and stage whether these various explanations constitute mu- Ungerleider (1995), asked subjects to imagine either a tually exclusive hypotheses about the conceptual basis perceptual attribute of an object (its color) or a func- of