Typical structural brain abnormalities found in the patient population tested. Coronal sections of structural MRI scans from three patients are shown. From left to right, sections at the levels of the splenium, rostral hippocampus, and frontal lobes are shown; the left hemisphere is displayed on the left. Rostral temporal cortex was severely degenerated in all cases. 

Contexts in source publication

Context 1

... performance reduction observed in the lexical decision experiment (average d 0 = 1.5 for the patients as compared with average d 0 = 3.4 for matched healthy participants). Importantly, however, and constituting a novel finding, there were significant differences in lexical decisions on words from different semantic categories, which were absent in matched healthy control subjects (significant Word Category × Group interactions). In SD patients, semantic word groups matched for a range of important psycholinguistic criteria led to significantly different processing success: Hit rates and d 0 values were lower for the COLOR condition than for the FORM condition, and action words in the FACE category produced lower hit rates and d 0 values than those in the ARM condition. Because of careful matching of other properties that affect lexical decision accuracy, we can assert with some confidence that these differences are best explained in terms of semantic features and corresponding category-specific topographies of the underlying cortical circuits. These findings have important implications for theories of semantic and conceptual processing and representation. SD has been considered an across-the-board deficit in semantic processing (Patterson et al., 2007; Saffran & Schwartz, 1994). As mentioned in the Introduction section, previous group studies have failed to uncover category specificity of semantic deficits in SD in terms of the most commonly assessed distinctions: living versus nonliving concepts (Lambon Ralph, Graham, et al., 1998; Lambon Ralph et al., 2007) and objects versus actions (Cotelli et al., 2006). Our current findings suggest that the absence of any category differences in previous studies of SD might be due to the grain size of the categories under study. A contrast between the broad categories of living and nonliving things can be motivated on the basis that they are innate to the human cognitive system (Caramazza & Mahon, 2003). By contrast, if category differences result from differential grounding of semantics in sensory/motor versus functional knowledge (Warrington & McCarthy, 1983), then more fine-grained subdivisions of semantic space are required separating, for example, large artifacts from tools, color from form, and action knowledge relating to different body parts (Barsalou, 2008; Pulvermüller, 2005; Humphreys & Forde, 2001; Farah & McClelland, 1991; Warrington & McCarthy, 1987). Con- sistent with this suggestion, the differential degrees of deficit in lexical decision revealed in SD patients by the present investigation cannot be explained by any domain- general principle. Although the larger categories of visually related and action-related words did not afford a well-matched contrast, it is nevertheless observable that the patients were not differentially impaired on one or other of these broad domains. The generally reduced performance on abstract words might suggest a general category difference (abstract vs. concrete), but as not all psycholinguistic confounds (e.g., word frequency) could be ruled out for this contrast, this conclusion is not fully endorsed by the data. Where a degree of category specificity emerged in the present work was between subtypes of action words and between subtypes of visually related words. This pattern of results supports a distributed circuit model of lexico-semantic networks according to which knowledge of object color (relative to knowledge of object form) and knowledge about actions relating to the face and articulators (compared with those relating to hand and arm) draw more heavily on anterior-temporal and adjacent frontal structures. Considering the first of these contrasts, despite the fact that color processing areas in the occipito-temporal cortex are posterior to areas spe- cializing in form analysis (Corbetta, Miezin, Dobmeyer, Shulman, & Petersen, 1991; Zeki et al., 1991), recent studies have indicated that processing of color concepts involves anterior-temporal structures (Miceli et al., 2001; Mummery, Patterson, Hodges, & Price, 1998), whereas form-related concepts and words activate posterior-lateral fusiform gyrus and, in addition, dorsolateral pFC (Moscoso Del Prado Martin et al., 2006; Pulvermüller & Hauk, 2006). Comparison of mildly and severely affected patients indicated that the difference between color and form word impairment was more pronounced at early (mild) stages of the disease, whereas later on the difference diminished, perhaps suggesting posterior spreading of the lesion. With regard to the second contrast: Action words and concepts elicit semantic somatotopy along the fronto-central cortex, with more ventral activation to FACE words and more dorsal activation to ARM words (Hauk et al., 2004; Pulvermüller et al., 2000). Although SD in the very early stages is associated with damage restricted to the temporal lobes (Nestor et al., 2006) — and the present population also showed most massive degradation there (Figure 2) — inferior- frontal atrophy and hypometabolism emerge with advanc- ing disease (Drzezga et al., 2008; Mummery et al., 2000). The bigger numerical difference between action word category d 0 values seen in more severely affected patients of the present cohort is consistent with this observation. In conjunction with the semantic topography model, the functional deficit in inferior-frontal structures explains the specific deficit for face-related words. Note that not all brain models of category specificity explain the pattern of deficits found here in SD. For example, Martin and Chao (2001) have proposed that the processing of color semantics is located posterior to form semantics in posterior temporal cortex. This proposal on its own does not account for the present differential deficit for color words in SD patients, which, similar to earlier results from patients (Miceli et al., 2001; Mummery et al., 1998) and the abovementioned imaging work, suggest a specific contribution of anterior-temporal cortex to color concept processing. Also, previous theories of category specificity did not reach the fine grain size to dif- ferentiate action word subcategories according to their body part reference. The semantic topography model in Figure 1A and its proposed extension (Figure 5) seem to provide the best explanation available to date of the pattern of lexico-semantic deficits seen in SD. Still, the model needs further specification to more fully account for the semantics of a wider range of words and concepts (e.g., prepositions, Kemmerer & Tranel, 2003; Noordzij, Neggers, Ramsey, & Postma, 2008). In contrast to one prediction of the semantic topography model, SD patients ʼ performance on leg words was not significantly better than that on face words, although the suggested cortical distribution of leg-action concepts should, by hypothesis, include neuronal assemblies in motor and premotor areas, which are not specifically affected in SD. Although this lack of significance does not constitute a strong finding falsifying the semantic somatotopy model, it still suggests that semantic circuits may be more complex than the sketches presented in Figure 1A would indicate. More specifically, the tendency toward an impairment of leg-word processing in SD patients may suggest that the processing of these words critically depends on the integrity of neuronal assemblies in anterior- temporal or inferior-frontal cortex, not just on leg motor and premotor regions. Direct brain imaging evidence for anterior-temporal contributions to leg-word processing has recently been revealed by cluster analysis performed on fMRI data from healthy subjects (Pulvermüller, Kherif, Hauk, Mohr, & Nimmo-Smith, 2009). These results revealed cell assemblies for leg words with both dorsal fronto- central (including motor) and anterior-temporal cortex contribution, therefore explaining why SD patients ʼ performance on leg words was similar to that on the affected face words. The present pattern of results refutes the suggestion outlined in the Introduction section that differential activations to various semantic kinds might be epiphenomenal rather than reflecting the operation of differently distributed semantic networks. Because the data come from patients with the most severe specifically semantic deficit known to date, there cannot be any doubt that the function of the damaged circuits is semantic conceptual in nature. In addition, the consistency between degrees of deficit in SD for various semantic categories and the evidence from a range of regional functional activation studies in the healthy brain using well-matched word groups with different types of meaning support the semantic topography model. A semantic hub model that denied a role of category- specific semantic networks would be unable to account for the present pattern of results. Similarly, the semantic topography model as shown in Figure 1A would not account for the massive general impairment in SD of all semantic classes as demonstrated in the present and earlier studies. The best account of the data therefore results from adding a semantic hub to the semantic topography model (Figure 1A and B). This results in a hub-supported model of semantotopic circuits (Figure 5). According to this model, an intact semantic hub in temporal poles is necessary for the functionality of the semantic system, possibly in ways made explicit in the computational model of Rogers et al. (2004), but semantic conceptual computa- tions are being carried out in widespread brain areas. Importantly, these widespread areas are specific to different kinds of concepts and meanings, with form and face words drawing most heavily on anterior-temporal and inferior- frontal areas affected in SD. The general importance of anterior-temporal lobes in semantic processing may relate to their strategic placement as links between cortical and limbic structures (Pulvermüller & Schumann, 1994). A future target ...

Context 2

... semantic deficits (Lambon Ralph, Pobric, & Jefferies, 2009; Pobric, Jefferies, & Lambon Ralph, 2007). It is possible that the category-specific semantic circuits and the anterior-temporal semantic center inter- act in the processing of meaning in the brain (Figure 1B). The absence of any persuasive evidence for category specificity in SD may be related to methodological issues. The categories of living and nonliving things may be too broad to reveal a coherent dissociation pattern because they are intrinsically heterogeneous. For example, there are fun- damental differences in kind and, correspondingly, in the dissociation patterns after focal brain disease between subtypes of nonliving things, for example, tools, large man-made objects, and musical instruments (Warrington & McCarthy, 1987). A particular deficit for one subcategory could therefore be counteracted by better performance for other subcategories, and if specific fine subcategories of both action-related nonliving concepts and visually related living things were impaired, a large-category comparison would fail to reveal the specificity of the deficit. Similar concerns hold for any difference between nouns and verbs or between animals and tools. Guided by the semantic topography model (Figure 1A), we hypothesized that a degree of category specificity might emerge in SD if the investigation focused on fine- grained differences in semantic concepts. This should be especially true if the categories selected could be specifically mapped to the regions most damaged in SD, the anterior-temporal and the adjacent inferior-frontal cortex. The fine word categories whose specific fMRI activation maps are closest to the SD-affected regions are face- and articulator-related words, which have been demonstrated to activate inferior-frontal cortex (Hauk et al., 2004), and words with strong color associations, where fMRI suggests the importance of anterior-temporal cortex (Pulvermüller & Hauk, 2006). 1 A relevant contrast for the former is provided by words relating to hand and arm movements that, like face-related words, evoke general features of action but yield fMRI activations in more dorsolateral fronto-central regions. A contrast germane to the color category would be words relating to object form: Color, shape, or form is a visual feature of objects, but it yields fMRI activations in more posterior temporal areas. Semantic topography therefore predicts that, if all other relevant variables are balanced, SD patients ʼ performance in a lexico-semantic task should be more impaired for color words than for form words and more impaired for face words than for arm words. In contrast, a semantic hub model neglecting category-specific circuits would predict equally poor processing of all of these categories. A combination of semantic topography and hub models predicts general but also unequal degradation in SD. Knowledge of words from different categories was assessed in a lexical decision task, where the words were intermixed with word-like but meaningless pseudowords. The disadvantage of other more obviously semantic tasks, such as naming, is that different semantic categories tend to place differential demands on specific perceptual and cognitive processes. 2 Furthermore, the profound anomia characteristic of SD makes naming a poor choice for such an assessment: Naming scores close to or at zero, which are not uncommon in more advanced SD, may be dra- matic but would be uninformative in the current investigation. In lexical decision, stimuli can be controlled and matched for physical, cognitive, and psycholinguistic features, and the task stays constant over different conceptual categories whereas effects of semantic links can become manifest in behavioral responses. Note that not only behavioral evidence (Chumbley & Balota, 1984) but also neuro- imaging results indicate access to semantic knowledge in the lexical decision task when word and pseudoword stimuli are matched closely for orthographic and phonological characteristics, and the task instructions emphasize accuracy (see, e.g., Binder, Westbury, McKiernan, Possing, & Medler, 2005; Binder et al., 2003; Pulvermüller, Lutzenberger, et al., 1999; Pulvermüller, Mohr, & Schleichert, 1999). It is already well established that, with proper design of pseudoword foils, success in lexical decision is substantially reduced in SD patients relative to controls (Patterson, 2007; Rogers et al., 2004; Diesfeldt, 1992). If words and meanings are linked reciprocally by topographically specific circuits, a focal lesion should exacerbate the lexical decision deficit for words from semantic categories especially reliant on the affected circuits (Neininger & Pulvermüller, 2003; Pulvermüller, 1999). The patients participating in this study were 11 cases who presented to Neurology Clinics at Addenbrooke ʼ s Hospi- tal, Cambridge, and fulfilled standard criteria for a clinical diagnosis of SD (Hodges & Patterson, 2007). Figure 2 presents examples of structural MRI scans from three patients showing cortical atrophy mainly in rostral temporal cortex. All patients were native speakers of British English; 10 of 11 patients were right-handed and all were male — neither by design nor in accord with any sex bias in the incidence of SD, but merely by chance of those in the testable Cambridge SD cohort at the time. Their vision was normal or corrected to normal. Table 1 provides the results of standard testing as well as demographic details for each of the patients. The table, arranged in descending order of Addenbrooke ʼ s Cognitive Examination (ACE) scores to reflect the general cognitive status of these particular patients, reveals a considerable range of severity/stage of progression across the 11 cases, from mild (e.g., DB) to moderate (e.g., JM) to severe (IB, DCJ). The following description of the patients ʼ performance is made with reference to control data from a group of normal individuals of similar age and education (see Woollams, Cooper-Pye, Hodges, & Patterson, 2008). The nonsemantic tests, Rey Figure and digit span, produced performance within the normal range with only one or two exceptions among the more severe cases (Rey for DCJ, digit span for DG). Scores on the semantic and language tests are presented as proportions of the maximum possible score for naming (the standard 64-item picture naming test used in Cambridge), word – picture matching (the same 64 items with 10 picture choices per spoken word target), and the Camel and Cactus test (again, the same 64 items, where the target picture is to be matched to one of four alternative response pictures with which it is associated). The most mildly affected patient, DB, was within the control range on these easy naming and word-to-picture matching tests, but the other 10 patients were impaired on both, most substantially so, and all 11 cases had abnormal scores on the Camel and Cactus test of semantic association. Scores on the two fluency tests are given as proportions of the mean score from healthy control subjects. Three of the milder cases (DB, BC, and JW) were good at producing words on the basis of initial letter, but on the sensitive category fluency measure, all 11 patients, even the mildest, were markedly abnormal. For the lexical decision experiment, 10 male right- handed healthy individuals were selected from the MRC- CBU Participant Panel as suitable controls for the patients. For the two groups, mean ( SD ) age was 66.7 (6.9) years for the patients and 67.9 (6.4) years for the controls; mean ( SD ) years of education were 14.0 (3.3) years for the patients and 14.6 (3.6) years for the controls. Subjects were seated in front of a laptop computer with ∼ 1 m distance from the screen. Written words and pseudo- words were presented in large gray letters on a dark back- ground and remained on screen until a response was obtained. Between stimuli, a fixation cross appeared in the middle of the screen for an interval randomly varying between 1.5 and 2.5 sec. Subjects were instructed to look constantly at the fixation cross and to decide, as quickly and accurately as possible, for each stimulus whether it was a “ good meaningful English word ” or a “ meaningless letter string. ” In case of doubt, they were advised to guess. Our original intention was to use button press responses to obtain a measure of RT; but some of the more severely impaired patients had difficulty with the slightly “ dual-task ” nature of judging the lexicality of the letter strings while at the same time keeping track of the assignment of buttons to yes versus no responses. In the end, we therefore settled for the more natural verbal mode of responding ( “ yes ” or “ no ” or, if the patient preferred, “ word ” or “ no word ” ). As soon as a decision was expressed, the experimenter pressed a response button coding the type of response. As the experimenter ʼ s button presses only provide a delayed approximation of response times in the subjects, we primarily base our analysis on response accuracy. Instructions were first given in writing and then repeated verbally. A practice block then followed to familiarize participants with the task. Questions were answered, and more instructions were given after the practice block; practice was repeated if required. The experiment was started only after participants were comfortable with the task. After every 42 items, participants were asked whether they would like a rest, and if so, a break was given. The stimuli consisted of 210 meaningful English words and 210 orthographically well-formed pseudowords. Six different word groups, each consisting of 35 words, were selected as follows on the basis of semantic associations: words semantically related to (1) color knowledge (here- inafter COLOR), (2) form or shape (FORM), actions involv- ing (3) the face and articulators (FACE), (4) the arms and hands (ARM), or (5) the legs and feet (LEG), and (6) ...