MRC Cognition and Brain Sciences Unit

You are in:  Home » History of the Unit

5. MEMORY AND KNOWLEDGE: DMS SECTION PROGRESS REPORT

Overview

The central goal of the Dementia, Memory and Semantics (DMS) programme of the CBU's Memory & Knowledge Group is to analyse the functional and neuroanatomical structure of long-term memory and language from evidence about disorders of these cognitive abilities, especially in patients with neurodegenerative disease. Additional sources of evidence in our multi-approach research programme include studies of normally functioning individuals (both behavioural and neuroimaging experiments); neuropsychological studies of individuals with static rather than progressive brain lesions; and the development of computational, connectionist models of cognitive function. The group is perhaps best known for effectively combining detailed neuropsychological assessments with neuroanatomical investigations on the one hand, and connectionist modelling on the other. Examples of these two combinations that have resulted in high-profile and influential work over the last decade include articles on the neuropsychology and neuroanatomy of semantic dementia (Hodges, Patterson, Oxbury & Funnell, 1992), on temporal patterns of impaired retrograde memory in Alzheimer's disease vs. semantic dementia (Graham & Hodges, 1997), and articles on connectionist models of reading and its disorders (Plaut, McClelland, Seidenberg & Patterson, 1996) and of the relationship between comprehension and naming (Lambon Ralph, McClelland, Patterson, Galton & Hodges, 2001). Our connectionist enterprise derives from Karalyn Patterson's long-term collaboration with the well-known modelling group in the USA led by Jay McClelland, Mark Seidenberg and David Plaut and supported by joint NIMH programme grants, with the third successive 5-year funding period starting in 2002. The facts that our neuropsychological studies are exceptionally well grounded in neurological/ radiological assessments, and that we are able to study whole case-series of patients with disorders especially germane to important theoretical issues, are attributable to John Hodges' position as Professor of Behavioural Neurology and to his MRC-funded programme of clinical research in the Cambridge University Department of Neurology. When the CBU's new programme of research under the directorship of William Marslen-Wilson was established in 1997, the DMS programme based jointly at the CBU and in Neurology was singled out by the MRC as critical to the future goals of the CBU and its role in a broader Cambridge-wide initiative in neuroscience.

We wish to emphasise three sine qua non for the success of the DMS programme that derive from its joint CBU/Neurology base. First, our main patient populations are those with neurodegenerative diseases – Alzheimer's disease, frontotemporal dementia, corticobasal degeneration, etc. Unlike patients with static lesions from cerebrovascular accident or tumour resection etc., who often improve or at least remain stable post-insult and are not necessarily 'ill', our patients steadily deteriorate. They require substantial and regular medical care, and their families require clinical attention and counselling, all of which is provided by the services associated with the clinics organised by John Hodges (the weekly Memory and Cognitive Disorders Clinic and the twice-monthly Early Dementia Clinic). The second vital aspect of this coordination is the extent to which our research benefits from the extensive clinical work-up that the patients receive. This benefit comes in the form of both clinical neuropsychological tests and neuroradiological investigations – typically structural MRI which, for neurodegenerative cases, needs to be repeated longitudinally. The third advantage provided by the symbiosis of the two programmes comes at the level of inter-disciplinary development: our approach fosters appreciation of theoretical work by clinicians, and of clinical methods by academic researchers. The benefits gained by the students and postdoctoral scientists who work in this cross-disciplinary environment endure well beyond the period of time that these individuals spend in our laboratory and percolate out into the research field in general.

As reviewed below, the central goal of DMS project M1 has been to advance our understanding of the internal organisation and neural basis of semantic memory, from studies of patients with semantic dementia (SD) and Alzheimer's disease (AD), functional imaging investigations of normal participants performing semantic tasks, and the intact and 'lesioned' behaviour of a new connectionist model of semantic memory developed by Timothy Rogers. Project M2 has addressed the relationships amongst forms of long-term memory (semantic, episodic, autobiographical) that are, at least in some theories, distinctly different. The primary approaches have been neuropsychological studies of retrograde memory in patients with SD, frontal dementia or AD; a novel functional imaging study of autobiographical memory in normal participants; and also an extensive programme of work on anterograde memory (new learning) in various patient populations. Project M3 consists of the DMS research on language. We have attempted to account for patterns of impairment in SD and AD on language tasks, such as those that tax receptive and expressive vocabulary, which are necessarily affected by semantic impairment. Another major component has addressed the performance of semantically or phonologically impaired patients on language tasks, such as reading aloud or inflecting verbs, for which our interactive models of language make specific predictions.

M1: The organisation and neural basis of semantic memory and knowledge

Scientific Direction: Hodges (45%), Patterson (35%), K. Graham (20%)
MRC-supported scientists: Kellenbach (100%), T. Rogers (60%), Bak (30%), Lee (20%) Grant-supported: Lambon Ralph (50%) Research support: Drake (34%), Erzinclioglu (33%), Wilkinson (33%)
Grant-funded research support: Everitt (100%), Donald (100%), Hearn (100%) MRC-supported students: Bozeat (100%), (50%), Simons (10%)
Grant-supported students: Clague (50%)

Background

"'Fodor's First Law of the Nonexistence of Cognitive Science': the more global a cognitive process is, the less anybody understands it" (Fodor, 1983, p.107).

What could be more global in human cognition than semantic memory, encompassing – as it is meant to do – our knowledge of all aspects of the physical and mental world plus the words that we use to talk about it? By Fodor's law, then, no-one should understand semantic memory. In fact, when he offered this assessment 20 years ago, it was probably appropriately pessimistic; but we believe that a more optimistic assessment would be appropriate today, and that this achievement is in large part attributable to the kinds of neuropsychological and modelling approaches that have characterised our work on semantic memory over the last quinquennium.

Because so much of this project has relied on studies of SD and/or AD, we begin with a brief account of these disorders. SD is now recognised as forming part of the spectrum of diseases referred to as frontotemporal dementia (FTD), a neurodegenerative disease associated with non-Alzheimer pathology at post-mortem. The syndrome of SD results from profound bilateral, asymmetrical atrophy of the anterior temporal lobe, more often with greater neuronal loss on the left side (L>R) but sometimes with more severe right-sided abnormality (R>L). Although semantic deficits occur in disorders other than SD (e.g. AD, herpes simplex virus encephalitis, following traumatic brain injury, etc.), the relative purity of the semantic breakdown in SD makes it ideal for studying (a) the cognitive architecture of semantic knowledge, (b) the impact of this dissolution on other cognitive processes, and (c) the neural basis of semantic knowledge. Our early papers on SD have become widely cited classics of the neuropsychological literature and have helped to promote world-wide interest in this syndrome. In addition to our more specific, theoretically governed research on this syndrome over the last quinquennium which is reviewed below, we have published a number of review papers and chapters with the goal of continuing to inform both neurological and cognitive communities about the nature and implications this disorder (e.g. Garrard & Hodges, 1999, 2000; Hodges, in press; Hodges & Miller, 2001a, b; Patterson & Hodges, 2000, 2001).

Our neuropsychological studies of semantic memory also involve patients with AD. In contrast to SD, the predominant cognitive deficit in AD is impairment of episodic memory, reflecting the fact that the medial temporal lobe (MTL) carries the brunt of pathology. Impaired performance on semantic tests such as category fluency and concept definitions are, however, also well-recognised features of AD. Some of our earlier studies (e.g. Hodges and Patterson, 1995; Hodges, Patterson, Graham & Dawson, 1996; Lambon Ralph, Patterson & Hodges, 1997) were influential in promoting the prevailing (though still not unanimous) view that these deficits in AD reflect genuine disruption to, rather than impaired retrieval from, semantic memory. We have also contributed to the characterisation of several now well recognised atypical presentations of AD (Galton, Patterson, Xuereb & Hodges, 2000).

M1.1 How many semantic systems?

M1.1.1 Objects vs. words

Our working hypothesis has been that when semantic knowledge about a concept (such as a zebra or a potato peeler) is activated in response to seeing a real exemplar of the concept, or by hearing or reading the concept's name, or by the intention to speak its name or to describe the object, it is the same distributed network of information that is activated. There will, of course, be important principled differences in the full activation patterns under these different conditions, resulting from links between semantic representations and brain regions more critically involved in processing words vs. objects as stimuli, and of processes needed to activate speech vs. action as responses. But with regard to central conceptual knowledge, we think in terms of a single distributed network. Therefore, although patients with central semantic deficits are most prominently impaired in producing and comprehending words, we predicted that they would also demonstrate abnormal performance on appropriate semantic assessments with pictures or real objects as stimulus materials and with non-verbal actions as responses. This theoretical position contrasts with a hypothesis of separate word-, object- and action-based semantic systems that can be independently damaged (e.g. Papagno & Capitani, 2001).

Our studies of semantic memory have involved the use of a battery of tests designed in the early 1990s and then updated about five years later (in collaboration with Dr Peter Garrard, an MRC Training Fellow at Addenbrooke's Hospital working with our group from 1996-1999). The battery assesses the status of conceptual knowledge about the same set of concrete-concept target items across a range of tasks differing in the modality of input and output. In support of the hypothesis of a modality-general semantic network, patients with SD showed striking parallels in concept definitions produced in response to pictures of these objects and (on a separate occasion) to the corresponding object names (Lambon Ralph, Graham, Patterson & Hodges, 1999b). As always in this disorder, success was strongly modulated by stimulus frequency or familiarity; but there were no cases even approaching a classical dissociation with respect to stimulus modality in either direction. Most patients' performance was mildly better in the picture condition, a difference that we attribute to the relationship between stimulus surface form and meaning which is wholly arbitrary in the case of words but partially systematic for objects or pictures. This difference is well captured by our connectionist models of conceptual knowledge (Rogers et al., submitted-a). The only two patients who produced more information in response to words than pictures both had R>L atrophy, and we attribute their pattern to disruption of visual/structural processing that seems to depend critically on right-hemisphere temporal lobe structures (Kellenbach, Hovius & Patterson, submitted).

The view that there is a specialised (and more robust) semantic system relating to objects and actions still enjoys wide support (e.g. Lauro-Grotto, Piccini & Shallice, 1997). We have tackled the issue in our research with SD patients in several additional ways. To begin with, we designed a series of entirely non-verbal semantic tasks involving (a) the matching of different pictures on the basis of conceptual knowledge, (b) the matching of pictures of objects to their characteristic sounds, and (c) delayed copying of drawings of familiar objects (Bozeat et al., in press-a; Bozeat, Lambon Ralph, Patterson, Garrard & Hodges, 2000). These experiments demonstrate not only consistent deficits in SD on all of these non-verbal tasks but also, as predicted by our hypothesis, extremely high correlations between degree of deficit on verbal and non-verbal semantic tasks. Results from the copying tasks are particularly striking. SD patients have normal ability to copy a line drawing of a familiar object that is in front of them; but when asked to study a drawing and to reproduce it after a brief filled delay, their responses both simplify and complicate the target drawing in ways that fit our understanding of the nature of the deficit. That is, after imposition of a short delay, the patients' drawings become much more prototypical: camels lose their humps and ducks acquire four legs. This suggests a broadened 'basin of attraction' for frequent and typical exemplars of a category, which can be observed in the verbal domain when the patients name camels and ducks as "dog" or "animal" and in the non-verbal domain when they draw camels and ducks that look like dogs.

M1.1.2 Do actions speak louder (or more successfully) than words?

In view of persistent claims in the literature that "action semantics" can be selectively preserved when other conceptual knowledge is compromised, we have explored the ability of patients with SD to use everyday objects (such as pencil sharpeners, match boxes, potato peelers and screwdrivers). A series of experiments designed to evaluate the ability of SD patients to use real exemplars of these objects, in relation to the patients' naming of and other kinds of knowledge about the exact same items, has demonstrated (a) that the patients invariably have (familiarity-modulated) deficits in real object usage which reflect degraded knowledge rather than impairment in praxic abilities or mechanical problem solving skills, the latter two being strikingly well preserved in SD (Hodges, Bozeat, Lambon Ralph, Patterson & Spatt, 2000; Hodges, Spatt & Patterson, 1999); (b) that degree of success/failure in object usage can be strongly predicted by performance on naming and semantic matching tests for the same items (Bozeat, Lambon Ralph, Patterson & Hodges, in press-b); (c) that aspects of action which are systematically related to object affordances, although restricted in scope, are relatively preserved under semantic impairment (Bozeat et al., in press-b); and (d) that – as originally demonstrated by Snowden, Griffiths & Neary (1994) – the patients are markedly more successful at using specific exemplars of common objects with which they have personal and recent experience than equally good exemplars provided by the experimenter (Bozeat, Lambon Ralph, Patterson & Hodges, 2002). The precise interpretation of this latter effect is debated, but our interpretation is that it reflects relatively preserved stimulus-specific procedural and episodic memory in SD rather than genuine enhancement of conceptual knowledge by everyday experience. The opposite pattern to that seen in SD – i.e. severely impaired object use due to disrupted praxis and mechanical problem-solving ability, despite preserved knowledge of how objects are meant to be used – occurs in patients with corticobasal degeneration, a degenerative disorder involving basal ganglia, frontal and parietal cortices (Spatt, Bak, Bozeat, Patterson & Hodges, 2002).

M1.1.3 Object recognition

A related topic of debate in the cognitive neuropsychological literature has been the extent to which conceptual knowledge is necessary to recognise an object as familiar, i.e. something that has been encountered before. Some theorists (e.g. Humphreys, Riddoch & Quinlan, 1988) have argued in favour of a pre-semantic structural description system that can function normally to identify a visual stimulus as a familiar object even if the semantic system itself is degraded, whereas our more interactive view of object processing predicts impaired recognition under these conditions. Consider the object decision tests, which have become standard for assessing this issue, in which participants are asked to judge whether pictures represent real or non-real objects. We predicted and have now confirmed that object-decision success by SD patients depends crucially on the relative plausibility/ typicality of the real and non-real objects in the stimulus set (Hovius, Kellenbach, Graham, Hodges & Patterson, in press; Rogers, Lambon Ralph, Hodges & Patterson, submitted-b). Particularly dramatic were the results from the latter study in which the patients were asked to make two-alternative forced-choice object decisions in two conditions: one where the more typical visual form for an animal (e.g. having a tail, not having a hump) applied to the real item in the pair (e.g. a lion with/without a tail; a donkey without/ with a hump), and the other where typicality was higher for the non-real animal (a gorilla without/with a tail; a camel with/without a hump). The SD patients achieved excellent performance in the former condition but were significantly impaired (to varying degrees across patients, which correlated with severity of semantic deficit) in the second condition. These results support our hypothesis that, even if there is a pre-semantic structural description system, its interaction with semantic representations is essential to its normal function.

In a recent PET study (Kellenbach et al., submitted), normal subjects revealed significantly increased activation of left temporal-lobe regions when making decisions (to pictures of objects) requiring knowledge of object colour or encyclopaedic facts; but the posterior inferior temporal area (Brodmann Area (BA) 37) in the right hemisphere was the principal locus for differential activation in an object-decision condition. On the reasonable assumption that there are strong neural connections feeding back from more anterior areas to posterior regions of both temporal lobes (Gloor, 1997), we interpret the object-decision findings in SD as reflecting impaired communication between severely atrophied anterior regions and structurally largely-unaffected posterior regions that cannot now function normally owing to this reduced input. This fits with results from our functional activation PET study with SD patients (Mummery, Patterson, Wise, Vandenberghe, Price & Hodges, 1999): associative semantic decisions about words and objects, which probably rely principally on anterior temporal areas, also activated posterior left BA 37 in normal subjects but not in four SD patients.

M1.2 Categorical and/or hierarchical organisation of semantic memory?

M1.2.1 Semantic categories

One of the most intriguing and controversial findings in cognitive neuroscience has been that neurological patients may have disproportionately disrupted knowledge of particular semantic or linguistic categories. The best documented double dissociation is between living things (or natural kinds) vs. manmade items (or artefacts). Our own studies of AD patients have revealed a small, but consistent, advantage for naming of artefacts over natural kinds in the majority of cases (Garrard et al., 2001b). The discrepancy increases with disease severity and does not seem attributable to stimulus variables (word frequency, age of acquisition etc.), or to inherent difficulty since a few patients showed the opposite pattern. To explain these findings we have argued that the pathology in the majority of AD cases spreads out from the hippocampal formation into infero-lateral temporal neocortex, thus causing disproportionate disturbance of visually-based knowledge which may be critical for the more subtle differentiations required to identify natural kinds. By contrast, the rare AD cases with poorer performance on artefacts have had unusually severe parietal lobe involvement (Garrard, Patterson, Watson & Hodges, 1998; Garrard et al., 2001b). Many patients with SD also display a small advantage for artefacts over natural kinds (Garrard, Lambon Ralph & Hodges, 2002). The reason why so few SD cases are characterised by a major disadvantage for natural kinds remains mysterious, especially in light of the robust deficit for knowledge of sensory/perceptual aspects of conceptual knowledge in SD (Lambon Ralph et al., 1999b; Lambon Ralph, Patterson, Garrard & Hodges, submitted). In agreement with Tyler et al. (in press), our PET investigations of this issue in normal participants have failed to uncover significant or consistent activation differences associated with natural vs. artefact domains, and instead have revealed differential activations associated with feature type (perceptual vs. non-perceptual aspects of conceptual knowledge: Lee et al., 2002a; Mummery, Patterson, Hodges, Wise & Price, 1998).

M1.2.2 Linguistic categories

A second dissociation of long-standing interest in neuropsychology is between objects/nouns and actions/verbs. Modelled on the widely used Pyramids and Palm Trees test of associative semantic knowledge about objects (in which the subject must decide that the palm tree rather than the pine tree "goes with" an Egyptian pyramid), we designed a test of associative semantics for actions/verbs, called the Kissing and Dancing test. The most striking finding from joint administration of these two assessments has been disproportionately poor Kissing-and-Dancing performance in patients with motor neurone disease (MND) with dementia (Bak, O'Donovan, Xuereb, Boniface & Hodges, 2001b), and also in patients with the frontal variant of FTD (Bak & Hodges, in press). Post-mortem brain examination in a subset of the MND cases has shown severe neuronal loss in the inferior frontal lobe (BA 44/45), in keeping with the hypothesis that action/verb knowledge is associated with frontal cortical and/or basal ganglia regions (Bak & Hodges, 2001; Pulvermuller, Haerle & Hummel, 2000). It is not yet clear whether this dissociation is single or double: although some SD patients achieve better scores for actions than objects, the difference is less dramatic/consistent than the reverse contrast associated with frontal abnormalities. Furthermore, a different approach to the noun/verb question in SD attributed the slightly greater preservation of verbs than nouns in their impoverished spontaneous speech to the impact of stimulus/response frequency on all aspects of SD impairments (Bird, Lambon Ralph, Patterson & Hodges, 2000).

M1.2.3 "I never forget a face"

A third domain of potential categorical organisation relates to knowledge about people vs. objects. In collaboration with Dr Siân Thompson (MRC Training Fellow at Addenbrooke's Hospital, 2000-2003), we have devised a famous-person naming test of graded difficulty (parallel to Warrington's graded difficulty object naming test) and a new battery assessing semantic knowledge of famous people (modelled on our general object/animal battery). With these improved assessments, we have recently replicated (Thompson et al., submitted) earlier observations of an SD patient with R>L temporal atrophy and severe loss of knowledge about famous people in the absence of significant impairment on our general semantic battery (Evans, Heggs, Antoun & Hodges, 1995). The interpretation of this result with regard to right-temporal specialisation for knowledge of people is, however, complicated by the fact that patients with mainly left-sided pathology typically also perform poorly on person-based tests of knowledge (Hodges & Graham, 1998). A significant disruption to knowledge about famous people, which manifests mainly as difficulty in naming them, is also an early and consistent finding in patients with AD (Hodges & Greene, 1998). It remains to be determined whether the vulnerability of this knowledge is primarily another manifestation of the fact that more specific aspects of conceptual knowledge suffer first in progressive degradation of semantic memory. People are unique individuals, whereas knowledge of objects or animals relies on accumulated experience with many exemplars of the concept.

M1.2.4 Modelling

Apart from documenting the marked effects of specificity and prototypicality in patients with semantic deterioration, our investigations of these factors have centred on connectionist models in which these principles are inherent to the nature of representations. In a distributed semantic space, properties common to many semantically related items (such as four legs, two eyes and a tail for animals) benefit from mutually supportive connection weights. Atypical properties – those that serve to differentiate a given concept from its close semantic neighbours (e.g. penguins cannot fly; gorillas have no tail; zebras look very much like horses but happen to have stripes) – enjoy no such benefit and, under semantic degradation, suffer disproportionately. Thus, as already mentioned, patients with SD tend (a) to choose the gorilla with a tail (even though they can never have seen such a thing in real life) in a forced-choice object decision test; (b) to copy realistic line drawings of animals after a short delay by putting four legs on birds or snakes; (c) to name objects by assigning labels of the general category (zebra → "animal") or of more prototypical category coordinates (zebra → "horse"; one SD patient, after confidently responding "horse" to a picture of a zebra, pointed to the stripes and asked "but what are these funny things for?"); (d) to sort objects or words more successfully on the basis of general than specific distinctions; and so on. All of these features emerge spontaneously in a connectionist model of semantic memory (Rogers et al., submitted-a) which learns to translate between different surface representations such that it can simulate standard neuropsychological tasks such as object naming (visual representation → name), drawing to dictation (name → visual representation), concept definitions (picture or name → verbal attributes), naming to description (verbal attributes → name), etc.

Whereas models of semantic memory – connectionist or otherwise – have typically assumed that semantic representations consist of feature lists, a key aspect of this model is that the semantic representations have no explicit content: they are, in effect, the hidden unit representations that arise from learning about relationships between the content-bearing surface representations associated with stimuli and responses. Thus the attributes that people can list for concepts (e.g. "a zebra is a horse-like large African herbivore with black and white stripes") are not viewed as constituting or directly reflecting the explicit contents of semantic representations. These are instead features that the model learns to comprehend as input or to produce as output when it needs to translate between different surface forms.

One reason that we consider this to be a valuable way of construing semantic memory is that one no longer needs to confront the difficult question of whether there are separate semantic systems associated with different modalities or feature types. In this conception, the answer is both yes and no. The visual and verbal representations that enable processing of stimuli and generation of responses are separate (though interacting) sources of information; but they are not central semantic sub-systems. This fits well with recent functional imaging studies of normal subjects, in which we have demonstrated significantly different selective regions of activation associated with different components of object knowledge – e.g. the typical sizes, colours, sounds or actions associated with familiar objects. As predicted more than a decade ago (Allport, 1985), these selective activations are often in or near brain regions involved in sensory or motor experience with the attribute type. Thus colour and size judgements activated visual areas in the posterior inferior temporal lobe, while sound judgements activated the superior temporal gyrus, an auditory region (Kellenbach, Brett & Patterson, 2001); and responses to manipulable (in conrast to non-manipulable) objects activated a region of ventral pre-motor cortex (Kellenbach, Brett & Patterson, in press). We interpret these activations as reflecting the content-bearing representations associated with stimuli and responses. These representations are then linked by the abstract semantic representations that must learn how to put all of this content together such that each modality- or attribute-specific region can talk to the others.

M1.3 The neuroanatomical basis of impaired semantic memory as assessed by structural imaging

In our initial studies of SD, we observed consistent asymmetrical atrophy of the anterior temporal lobe, and suggested a key role for this region bilaterally in semantic processing. More recently, the results of convergent structural imaging methods have confirmed and refined this early observation. Our first quantitative study used voxel-based morphometry (VBM) to measure the extent of reduction of grey matter in specific cortical regions in six patients with SD (Mummery et al., 2000). All six cases had significant atrophy in the anterior infero-lateral temporal lobe on the left, though one revealed even more profound atrophy in the same region on the right (see Figure 1). One very early case, in whom abnormalities were as yet only measurable on the left side, also showed abnormal semantic performance only on the most challenging tests, which is consonant with an assumption of bilateral representation of conceptual knowledge.

Figure 1. Regions of significantly reduced grey matter density in each of six SD patients relative to a group of age-matched normals. From left to right: the left lateral surface, right lateral surface, left medial surface and ventral surface of the brain are depicted on a three-dimensional rendering of a standard MRI scan. The figures are thresholded at p < 0.001 (uncorrected) to show the extent of damage.

The results of a larger SD group study (n=18), using a volumetric method of tracing defined cortical regions on coronal MRI (in ANALYZE), again confirmed consistent and severe atrophy involving the temporal polar cortex and the fusiform, parahippocampal and inferior temporal gyri (Galton et al., 2001b). The degree of semantic impairment on a variety of measures (category fluency, naming, Pyramids and Palm Trees semantic association) correlated most highly with the volume of the left fusiform gyrus: the latter consists mainly of Brodmann Area 20 but with contributions from BA 35/36, the perirhinal cortex. In addition, the measures of expressive vocabulary (naming and category fluency) correlated significantly with degree of atrophy in other left temporal regions including temporal pole and inferior and middle temporal gyri. One unexpected finding was significant asymmetric hippocampal atrophy in SD which, on the left, was equivalent to or greater than that in the AD cases included in this study (see Chan et al., 2001 for a similar finding). Although this degree of left-sided hippocampal abnormality in SD was not predicted, it does fit some of our behavioural and cognitive findings. For example, whereas recognition memory for pictures of objects can be normal in SD (see Progress Report for M2), episodic memory/learning for verbal material (especially as measured by recall but even by recognition) is very impaired (K.S. Graham, Patterson, Powis, Drake & Hodges, 2002).

M1.4 Summary

Our findings in this domain over the last five years, from all of the main approaches employed (neuropsychology, structural/functional imaging and connectionist modelling), suggest that semantic memory is both one and many systems. Different modalities of input to/output from conceptual knowledge, and different aspects of such knowledge, place particular demands on processing that recruit partially specialised brain regions; but the core of semantic memory, which enables people to translate between different modalities and coordinate different aspects of knowledge so as to behave in the real world, is a single distributed system. Neurological damage to this system, associated mainly with bilateral anterior temporal atrophy, disrupts all coherent behaviour, including use of familiar objects that has often been considered a separate module. Aspects of knowledge that are shared amongst many related concepts are robust; aspects that are unusual deteriorate rapidly under semantic degradation, yielding a system that responds mainly to typicality.

Project M2: The behavioural and neural relationship of semantic memory to episodic and autobiographical memory

Scientific Direction: K. Graham (70%), Hodges (45%), Patterson (15%)
MRC-supported scientists: Lee (80%), Bak (40%) Research support: Erzinclioglu (34%), Drake (33%), Wilkinson (33%) Grant-funded research support: Powis (100%) MRC-supported students: Kropelnicki (100%), Simons (90%) Grant supported students: A.Graham, Clague (50%)

M2 is concerned with the organisation and neural basis of long-term memory, in particular the influential distinction, first proposed by Tulving (1972), between episodic and semantic memory. Episodic memory refers to personally experienced and temporally specific events, the retrieval of which is said to be associated with "autonoetic" conscious awareness (or "mental time travel", Tulving, 2001). Semantic memory, as discussed in detail in M1, refers to our store of representational knowledge including facts, concepts, and the meaning of words. Tulving initially proposed that episodic and semantic memory were cognitively and neurally separate systems, based on findings from patients such as the famous HM (Scoville & Milner, 1957), who suffered severe episodic memory deficits after surgical removal of structures in the MTL, but showed little evidence of semantic memory impairment. This led researchers to conclude that the neural basis of episodic memory was the MTL, and that damage to this region would result in a selective loss of episodic, but not semantic memory. Further studies in amnesia, however, challenged this view, and led to the development of more sophisticated models of long-term memory.

M2.1 Memory consolidation

One of the first problems with a simple fractionation between episodic and semantic memory was the fact that the episodic memory impairment seen in amnesia was often not complete. Some patients were able to retrieve memories from the remote past (e.g. childhood or early adulthood) despite their inability to recall recent events (Rempel-Clower, Zola, Squire & Amaral, 1996). This result implies a process of memory consolidation in humans, whereby new memories are initially dependent upon MTL regions, especially the hippocampus, but become consolidated over time in other brain regions (e.g. temporal neocortex). This account of how human long-term memories are acquired and stored has been termed the standard model (Squire, 1992), and a major focus of our research over the last five years has been the applicability of this model to profiles of remote memory, predominantly in patients with dementia.

M2.1.1 Reverse temporal 'gradient' in SD

Until our recent work, the majority of research on memory consolidation had focused on patterns of remote memory in patients with nonprogressive damage to MTL regions, in particular asking whether such patients show a temporal gradient (recent < remote) in recall of semantic and autobiographical information. Our goal, however, has been to investigate the opposite, and previously unspecified, hypothesis that temporal neocortical damage would be more likely to impair recall of remote compared to recent memories. Patients with SD provide a unique opportunity to investigate this prediction, as they present with focal atrophy to temporal neocortical regions, with asymmetrical involvement of medial temporal structures (see M1.3). Our first experiment on this topic reported a double dissociation in the recall of personal events from the past (so-called autobiographical memory). While cases with SD showed better retrieval of personal events from recent life compared to childhood and early adulthood, a matched group of AD patients were better at retrieving memories from the remote compared to recent past (K.S. Graham & Hodges, 1997). Further studies revealed that the reverse temporal gradient in SD extends to remote semantic memory, including tests of event knowledge (K.S. Graham, Pratt & Hodges, 1998) and recognition/identification of famous names (Hodges & Graham, 1998). In addition, these investigations confirmed that patients with SD actually show a more step-like profile, as opposed to a gradient, in performance, with better recall of only very recent events from the last 2-3 years of their lives. These findings have received considerable attention from memory researchers around the world, and are strongly consistent with the standard model (Murre, Graham & Hodges, 2001).

M2.1.2 Multiple trace model

Notably, however, in the last few years the standard model has been contested. Nadel and Moscovitch (1997) noted that many patients with bilateral hippocampal damage show autobiographical memory deficits that are temporally extensive (Cipolotti et al., 2001); this pattern suggests that the hippocampus may play a more permanent role in the retrieval of all episodic memories regardless of their age. While this new memory consolidation model, termed the Multiple Trace Model (MTM), is relatively compelling, at least in terms of how it can account for the behavioural data from patients with focal MTL lesions, it has significant difficulty explaining the reverse step function documented in our patients with SD. In response to a commentary on this issue (K.S. Graham, 1999), Moscovitch and Nadel (1999) proposed a number of alternative explanations for the time-limited pattern in SD: (a) that there was only one reported patient showing this pattern on a sufficiently detailed autobiographical test; (b) that the profile may be due to strategic retrieval deficits caused by concomitant frontal pathology; and (c) that verbally-based testing of autobiographical memory may have exacerbated the degree of remote memory deficit. A series of investigations carried out in our Cambridge population have confirmed, however, the replicability of our initial finding across a variety of verbal and non verbal autobiographical tasks in both group and detailed single-case studies (K.S. Graham, Kropelnicki, Goldman & Hodges, in press; K.S. Graham, Lambon Ralph & Hodges, 1999a; K.S. Graham et al., 1998; Hodges & Graham, 2001; Nestor, Graham, Bozeat, Simons & Hodges, 2002). A further unpublished study by Anna Kropelnicki (MRC PhD student, 1999-2002) revealed that linguistic factors do not account for the time effect. These findings imply that remote autobiographical memories are truly inaccessible in patients with SD, and as discussed in detail in Nestor et al. (2002), lead one to question the exact role of the hippocampus in the MTM. For example, if the hippocampus acts as a pointer or indexer to the – presumably widely distributed – neocortical elements supporting the content of autobiographical events (as proposed by Moscovitch & Nadel, 1999), why can the hippocampus (which seems to be functionally preserved in SD, at least on the right) not activate components of remote episodic events dependent upon non-damaged neocortical regions (such as the occipital and parietal lobes, which may also be involved in autobiographical memory Maguire & Mummery, 1999; Rubin & Greenberg, 1998)?

M2.1.3 Relationship between autobiographical and semantic memory

One possible explanation consistent with the MTM is that the process of reconstructing autobiographical memories from the past is heavily dependent upon conceptual knowledge. Contrary to this hypothesis, however, is the case reported by Kitchener and Hodges (1999), who showed poor semantic memory, particularly for famous people but also for objects, yet performed well on tests of autobiographical memory. A further SD case study revealed no evidence that increasing the specificity of autobiographical cues improved accessibility to past personal events (K.S. Graham et al., in press), a result that is difficult to explain in terms of Moscovitch and Nadel's (1999) 'semantic' account.

To investigate the relationship between autobiographical and semantic memory in more detail, we carried out a series of neuroimaging studies of long-term memory at the Wolfson Brain Imaging Centre, Addenbrooke's Hospital. While autobiographical memory impairments in patients typically result from right-sided or bilateral temporal lobe lesions (Conway & Fthenaki, 2000; Kopelman & Kapur, 2001), neuroimaging investigations have highlighted a left-lateralised network of regions, including temporal pole, middle temporal gyrus, hippocampus, and medial prefrontal cortex (Maguire, 2001). There are, however, a number of problems with these latter investigations, such as the use of recognition, as opposed to recall, tasks (Maguire & Mummery, 1999) and insufficient length of time for autobiographical retrieval (Conway et al., 1999). To address these problems, our three PET studies employed tasks similar to those in our patient experiments (K.S. Graham & Hodges, 1997; Lee et al., 2002a; Nestor et al., 2002), which allowed us to test retrieval of semantic knowledge (of objects, animals and people) versus autobiographical memories in response to the same words. Consistent with the neuropsychological findings, we found clear evidence of bilateral anterior temporal lobe involvement in the retrieval of autobiographical memories (K.S. Graham, Lee, Brett & Patterson, submitted-b, see Figure 2).

Figure 2 Retrieval of autobiographical memories contrasted with recall of semantic facts revealed significant activations in the middle/inferior temporal gyri (BA 20), medial frontal cortex (BA 9/10) and inferior parietal lobe (BA 39/40). Reported in K.S. Graham et al. (submitted-b).

This finding implies that many neuroimaging experiments (Maguire & Mummery, 1999) have adopted – inadvertently – tasks that only require retrieval of autobiographical knowledge (e.g. I was on sabbatical at Berkeley), as opposed to a specific episodic event (e.g. I skied too fast round a corner at Lake Tahoe and ended up head first in a snow drift). Interestingly, the activations in the temporal lobe for autobiographical recall were virtually identical to those seen when recall of information about famous people (or famous events) was contrasted with retrieval of general semantic facts. This finding lends some support to the notion that recall of autobiographical memories may be dependent upon some aspects of semantic memory (as proposed by Moscovitch & Nadel, 1999), although it is important to note that while functional neuroimaging provides clues about the regions involved in a cognitive process, it cannot tell us which of these regions are critical for this behavioural task.

Our research on poor autobiographical memory in the context of normal semantic memory has included detailed studies of patient JM (Evans, Breen, Antoun & Hodges, 1996; Evans, Graham, Pratt & Hodges, in press), who suffered diffuse neocortical lesions after cerebral vasculistis, and of patients with the syndrome of transient epileptic amnesia (Manes, Zeman, Graham & Hodges, 2001). Patient JM is a good example of focal retrograde amnesia, with profound impairments in autobiographical memory in the context of normal anterograde and semantic memory. A follow-up study confirmed that she could acquire new autobiographical events despite her difficulties retrieving old memories from the past (Evans et al., in press). These studies reveal that at least two different processes can result in an isolated autobiographical memory deficit: (a) diffuse neocortical damage insufficient to cause loss of semantic memory but resulting in damage to cortico-cortical connectivity (as in JM), and (b) damage to strategic retrieval processes critical for autobiographical recall, as seen in some patients with frontal and thalamic lesions (Hodges & McCarthy, 1993), and in cases with frontal variant FTD (fvFTD, Nestor et al., 2002).

Further support for the last point has also come from recent re-analyses in patients with AD. Although many studies find evidence in support of a typical temporal gradient in AD (K.S. Graham & Hodges, 1997; Greene, Hodges & Baddeley, 1995), other experiments have failed to replicate this, even when using the same neuropsychological test (Nestor et al., 2002). A recent re-analysis of Greene and Hodges' AD data helped to resolve these discrepancies: while the majority of patients showed clear effects of time in autobiographical retrieval, a subgroup exhibited the flat profile predicted by Nadel and Moscovitch (1997). Strikingly, and contradictory to Nadel, Samsonovich, Ryan and Moscovitch (2000), there was no evidence that degree of remote memory impairment was related to performance on anterograde memory tests; in fact, the presence of an extensive and flat pattern was predicted by scores on tests of executive function (K.S. Graham, Goldman, Kropelnicki, Greene & Hodges, submitted-a). This intriguing finding suggests that the presence of extensive retrograde memory impairments in some non-progressive cases could, as in AD, be attributable to a frontally-based executive impairment.

M2.1.4 Role of the hippocampus in SD

As described in M1 above, our recent volumetric measurements of temporal lobe structures revealed that the degree of involvement of the left hippocampus in SD was largely indistinguishable from that seen in AD, although the SD patients had less MTL atrophy on the right side (Galton et al., 2001b). These findings initially seem problematic for the view that the relatively better recall of recent events in SD, and the poor memory in AD, is attributable to degree of atrophy in the hippocampi, and other MTL structures. Current cognitive and structural work, however, is aimed at investigating a number of possible explanations for these patterns, in particular whether these distinct profiles can be explained by the asymmetric nature of the pathology in SD. Structural work, carried out as part of John Hodges' clinical programme at Addenbrooke's Hospital, also indicates that differential involvement of anterior and posterior regions may be critical: Davies, Xuereb & Hodges (2002) found anterior hippocampal and entorhinal atrophy in SD with posterior involvement of these structures in AD. Parallel FDG-PET scanning has shown profound pan-limbic hypometabolism from a very early stage in AD (Nestor, Fryer, Smielewski & Hodges, submitted), while in SD the unilateral left-sided atrophy does not appear to affect the remainder of the limbic system (mammillary bodies, posterior cingulate, thalamic nuclei etc.).

M2.1.5 Cases with focal lesions to the MTL

While much of our research has focused on the patterns seen in SD and AD, we have also started addressing questions about memory consolidation, and the relationship between episodic and semantic memory, using cases with focal lesions to temporal lobe structures. Bak, Antoun, Balan & Hodges (2001a) reported two cases who suffered profound amnesia as a consequence of paraneoplastic limbic encephalitis, and noted that one patient showed no evidence of a remote memory impairment, despite significant difficulties on anterograde memory tasks. This type of pattern is not easily accommodated by the MTM, and further studies in patients with this rare condition are likely to be highly theoretically informative. In collaboration with Professor Narinder Kapur (Southampton University) we have also been investigating remote semantic memory in amnesia. Recent studies in cases with developmental amnesia (Vargha-Khadem, Gadian & Mishkin, 2001; Vargha-Khadem et al., 1997), reporting good learning of semantic knowledge despite profound episodic difficulties, concluded that there may be separate memory systems in the MTL, with the hippocampus and entorhinal cortex supporting the acquisition of context-dependent (episodic) and context-independent (semantic) memories, respectively. Kropelnicki, Graham, Kapur & Hodges (2002) tested this theory using a battery of tasks designed to investigate knowledge of vocabulary that had come into the English language at different time-periods over the past four decades. A group of adult amnesic patients, with bilateral MTL damage, demonstrated temporal gradients in their ability to define vocabulary, with a particular weakness in the decade (1970s) prior to their injury. In terms of vocabulary emerging after their neurological insult, there was evidence of good acquisition of gist (e.g. categorical) knowledge about the words, but the group performed less well than controls on a more difficult definitions task. There was no evidence that performance on memory tasks (especially recognition memory, a possible marker of non-hippocampal medial temporal damage) was related to vocabulary acquisition. These findings suggest that adult amnesics acquire new semantic knowledge via slow cortical learning, and has important implications for views about the role of MTL regions in new learning (see also M2.2).

M2.1.6 Summary

In conclusion, patients with SD and those with AD show distinct profiles in autobiographical and semantic memory retrieval: while most cases with AD show a standard temporal gradient (recent < remote), patients with SD typically recall more memories from the recent compared to the remote past. These patterns are problematic for the MTM, which does not predict such strong and replicable effects of time in these neurodegenerative diseases. Although Moscovitch and Nadel (1999) suggest that the pattern in SD, at least, might reflect the dependence of autobiographical retrieval upon semantic memory, some of our own neuropsychological studies are incompatible with this proposal (Graham et al., in press; Kitchener & Hodges, 1999). Our imaging study, however, did find remarkable overlap in the temporal lobe regions that were activated during retrieval of autobiographical memories and semantic knowledge of famous people and events, although it is not possible to know whether these regions are critical to these memory processes. While our data are most consistent with the standard model, it is clear that both views suffer from a serious lack of explanatory power: the standard model cannot account for patients who show extensive autobiographical deficits after bilateral hippocampal lesions, and the MTM is not able to explain profiles of remote memory in dementia. Further studies on this topic need to increase the specificity of the models, and to consider methods that will result in clearer predictions to be tested in neuropsychological populations.

M2.2 Is acquisition of new episodic memories dependent upon semantic memory?

Tulving (1995) has proposed that the registration of information in episodic memory is contingent upon access to and output from semantic memory, and writes, "a double dissociation between semantic and episodic memory is not possible and only single dissociations (impaired episodic memory and preserved semantic memory) can occur" (p.844). Over the last quinquennium, we have systematically investigated this topic in SD, and have demonstrated that episodic memory is not necessarily dependent upon normal functioning of the semantic system.

M2.2.1 Recognition memory for objects: a perceptual contribution

While our studies of autobiographical memory imply that cases with SD are able to acquire new memories, a number of other investigations have reported poor memory on standard measures of new learning, including recognition memory (Warrington, 1975). To address this controversy, our first study investigated recognition memory for pictures of real and non-real animals in patients with SD, AD and in control subjects (Graham, Becker & Hodges, 1997), and found a clear double dissociation. Our SD, but not AD, group showed similar levels of performance on this test compared to controls. In contrast, the SD patients were significantly impaired, compared to both control subjects and AD patients, on the study task in which they had to say whether the animals were real or not. Two further experiments have provided additional insights into the processes that support new learning in SD. K.S. Graham, Simons, Pratt, Patterson & Hodges (2000) found that a group of eight patients with SD showed normal recognition memory when the picture of the target item in the recognition memory task was identical to the item that had been seen at study. When the target item was changed between study and test (e.g. a round dial telephone was replaced with a push button telephone), however, the patients were significantly impaired (compared to control subjects) on items they were unable to name from a picture. This result suggests that loss of semantic knowledge only disrupts recognition memory when the target item is perceptually different from the studied item. This hypothesis was further supported by a case study assessing recognition memory for 'known' and previously familiar but now 'degraded' items. The only condition under which the patient showed poor recognition memory was perceptually different items for which the patient's conceptual knowledge was degraded. Recognition memory for all perceptually identical items and for 'known' perceptually different items was not significantly impaired.

M2.2.2 But what about faces?

This technique has since been extended to face recognition memory, in particular because previous studies in the literature suggested some vulnerability of this type of memory in SD (Hodges et al., 1992; Snowden, Neary & Mann, 1996). An MRC PhD student Jon Simons (1997-2000) showed that (a) patients with selective left temporal lobe atrophy were not significantly impaired on the faces component of the Warrington Recognition Memory Test; (b) a group of patients with predominantly right temporal lobe atrophy performed poorly on the test; (c) within this group, the status of the parahippocampal gyrus (which includes the perirhinal cortex) was predictive of performance (based on a temporal lobe rating scale developed by Galton et al., 2001a); and (d) like the object data described above, face recognition memory was affected by a change of picture (e.g. photographs of the Queen with and without a head-scarf), but only for famous people who were no longer 'known' to the patients (Simons, Graham, Galton, Patterson & Hodges, 2001a, see Figure 3).

Figure 3: (a) Example stimuli from Simons et al. (2001a). (b) The performance, as measured by d prime, of DM (a patient with semantic dementia) on the perceptually identical (PI) and perceptually different (PD) conditions of a recognition memory task (contrasting 'known' vs. 'degraded' (labelled as 'unknown') stimuli.

This study indicates two interesting points about the neural organisation of recognition memory for faces: first, a particular dependence upon the right temporal lobe, and second, involvement of non-hippocampal medial temporal regions. A further study explored more directly the relationship between knowledge and recognition memory using a modification of the Pyramids and Palm Trees Test (PPT, Simons, Graham & Hodges, 2002), and attempted to replicate the neuroanatomical findings from the previous investigation. In a study phase, subjects made the usual PPT semantic relatedness judgement, then 10 minutes later they were presented with an item from the test phase (pyramid) and a new semantically related foil (sphinx) and asked to choose the one they had seen before. We again demonstrated a double dissociation in the performance of patients with SD and AD. Similar to Simons et al. (2001a), the recognition memory performance observed in SD, but not in AD, was largely explained by the extent of atrophy affecting the parahippocampal gyrus (including perirhinal cortex) bilaterally.

M2.2.3 Multiple inputs to MTL

The results of this series of experiments are illuminating with regard to current cognitive and neural models of long-term memory. First, they show that perceptual information will support new learning, typically in conjunction with, but even in the absence of, meaningful input from the semantic system. A wider implication of this view is that the medial temporal based episodic system receives multiple sensory inputs from many disparate areas of the brain only some of which are damaged in SD. Second, our data provide some evidence that parahippocampal regions (in particular, perirhinal cortex) may be critical for recognition memory (see Hodges & Graham, 2001), a view consistent with emerging theories from the animal literature that different MTL regions play distinct roles in memory (Aggleton & Pearce, 2001). A PET neuroimaging study with normal participants (Simons, Graham, Owen, Patterson & Hodges, 2001b) examined two related issues: (a) which neural substrates are involved in recognition memory for people and objects, and (b) whether recognition memory for perceptually-identical vs. perceptually-different stimuli would differentially activate perceptual and semantic regions. Although the experiment did not support the latter hypothesis, the study confirmed that recognition memory for people activated a bilateral anterior temporal network, while recognition of object stimuli was more left-lateralised and involved posterior temporal regions. Notably, these patterns of activation – despite being produced on a very different type of memory task – mirror almost exactly those documented in our series of PET experiments investigating autobiographical and semantic memory (K.S. Graham et al., submitted-b; Lee et al., 2002a; Lee, Robbins, Graham & Owen, 2002b), and provide further, independent, confirmation of findings from our neuropsychological studies.

M2.2.4 Expanding the multiple inputs view to words

In contrast to their good performance on different forms of non-verbal anterograde memory, patients with SD typically perform very poorly on classic clinical tests of verbal memory, such as story recall and recognition (Hodges et al., 1992). The experiments discussed above help us to understand this finding. If new learning in SD relies heavily upon intact perceptual processes, then one would expect performance on word-based tests to be particularly poor since words, unlike pictures and faces, have little in the way of perceptually rich or distinctive information; consequently, learning verbal stimuli is likely to be almost entirely dependent on semantic encoding. Furthermore, unlike pictures, there is an arbitrary relationship between the phonological (and orthographic) forms of the word and the associated semantic representation. Our multiple inputs view predicts, therefore, that experiments based on the learning of 'known' and 'degraded' words would produce an advantage for 'known' items but that, unlike pictorial stimuli, the amount of learning even for 'known' words may still be impaired compared to that seen in control subjects.

We evaluated this prediction using a well-known test of verbal learning, recall and recognition (Welsh, Butters, Hughes, Mohs & Heyman, 1991). A group of SD patients were first pre-tested on the entire Snodgrass and Vanderwart (1980) corpus of 260 pictures to determine sets of 'known' and 'degraded' words (based on naming and word-picture matching) matched for word frequency and length. In the subsequent list-learning experiment, patients with SD showed very poor immediate and delayed recall of both 'known' and 'degraded' words but with an advantage for 'known' stimuli. There was even a significant impairment of yes-no recognition memory, particularly for items that were no longer known to the patients. Analysis of the performance of individual cases showed a very marked effect of disease severity: despite the overall 'known' > 'degraded' advantage, patients with mild semantic breakdown showed good recognition memory for both classes of stimuli, and the most severely semantically-impaired patients were poor at recognition memory in both conditions (K.S. Graham et al., 2002). These findings are consistent with another study in a single case of SD, in which we investigated whether relearning of "forgotten" vocabulary was possible via repeated exposure to pictures of concepts and their written labels. DM showed a remarkable ability to improve his word production (as measured using category fluency), although any benefit was quickly lost when he ceased practicing, thereby limiting the long-term usefulness of this particular strategy (K.S. Graham, Patterson, Pratt & Hodges, 1999c). Notably, DM's practice only improved his word production abilities: there was no evidence that DM's learning generalised within semantic category (e.g. he rigidly produced only items he practiced). In a follow-up study, we confirmed that DM was often unable to retrieve semantic information about items he produced accurately in category fluency (K.S. Graham, Patterson, Pratt & Hodges, 2001). These three studies suggest that extensive exposure (or rehearsal) to verbal stimuli is necessary for patients with SD to re-acquire new vocabulary, and that this learning does not seem to extend to semantic knowledge about the concepts represented by the words. The findings are in agreement with collaborative work carried out by Linda Clare, Barbara Wilson and John Hodges, in which it has been demonstrated that errorless learning of associations (e.g. face-name) can be beneficial in early AD (Clare, Wilson, Breen & Hodges, 1999; Clare, Wilson, Carter, Breen, Gosses & Hodges, 2000; Clare, Wilson, Carter & Hodges, 2001, see M5). DM's anomia meant that he rarely produced errors during practice, and it is therefore possible that his improvement also benefited – inadvertently – from an errorless approach.

M2.2.5 Familiarity versus recollection

It has recently been proposed that two functionally separate processes contribute to recognition memory: "recollection" (episodic retrieval of the original learning episode) and "familiarity" (recognition of the prior occurrence of an event without associated contextual retrieval). While the experiments described above have demonstrated preserved recognition memory for pictorial stimuli in patients with SD, one possible explanation for this pattern is that patients may be performing these tests using judgements of familiarity rather than recollecting the study episode (Tulving, 2001). This possibility is important, as it may be that Tulving's model – in which semantic processing is a necessary contributor to episodic memory – only relates to recollective memory. Given our findings that patients with SD can retrieve recent autobiographical experiences (K.S. Graham & Hodges, 1997), which must require retrieval of some contextual aspects of the event (e.g. temporal and spatial information), it seems unlikely that the good pictorial recognition memory in SD is purely due to familiarity; but this topic has not, to date, been examined systematically.

We developed two novel experimental paradigms that allowed us to study the contributions of familiarity and recollection to performance on tests of episodic memory in SD (Simons et al., in press). SD patients (varying from mild to severe) and controls were asked to name two sets of 30 line drawings presented 5 minutes apart. After a 15 minute delay, memory for the pictures and their sources was examined by asking subjects to indicate whether the pictures were unfamiliar or belonged to Set 1 or Set 2. Seven of the 10 patients showed normal item detection, confirming that the ability to discriminate familiar from unfamiliar items is preserved in the majority of cases with SD, even on a demanding episodic memory task. The three patients who showed deficits were the three most semantically impoverished cases, although again there was no evidence from item-specific analyses that performance on individual items was influenced by knowledge of the concept depicted by the picture. In the source discrimination component, three out of ten patients showed impairment when asked to decide which set a picture belonged to, although these were not the most semantically impaired cases, indeed two were among the five mildest patients. These findings, and similar results from a visual-visual associative memory task, confirm that many patients with SD have both excellent familiarity and recollection of studied items, and that both these processes presumably contribute to the good recognition memory evident in the disease.

The neural basis of familiarity and recollection is controversial, but, at least for familiarity, there is some evidence from animal lesion studies that neurons in perirhinal cortex, but not in the hippocampus, respond to re-presentation of a stimulus. These findings have led researchers to propose that familiarity is dependent upon a perirhinal cortex/dorsomedial thalamic nucleus system, while recollection requires a hippocampal/anterior thalamic memory system (Aggleton & Brown, 1999). Our previous studies demonstrating correlation of recognition memory with measurements of non-hippocampal medial temporal regions concur with these findings. There was, however, no significant relationship between volumetric measures of hippocampal loss and performance on the source discrimination component of our memory test. Instead, the source discrimination scores achieved by the SD group were significantly correlated with degree of impairment on a battery of frontal executive tasks. This implies that the recollective decision in our source monitoring task (predominantly discrimination of temporal order) may not be hippocampally-mediated. To explore this relationship further, we studied a group of patients with the frontal variant of fvFTD and found, as predicted, marked impairment in source discrimination with good item detection (Simons et al., in press).

M2.2.6 Recollection of spatial context

An additional set of studies carried out in collaboration with Dr Barbara Sahakian (Cambridge University), based on the paired associated learning paradigm (PAL) in the CANTAB battery, has also been informative with respect to the question of recollection in patients with SD. This task requires subjects to learn the spatial location of increasing numbers of complex visual patterns (up to 8), a process that is thought to be dependent upon the hippocampus. A clinical project, investigating the usefulness of the PAL task in early diagnosis of dementia, has confirmed that the PAL is exquisitely sensitive to very early stage AD (Swainson et al., 2001). To investigate whether this test would differentiate between different dementias, and whether patients with FTD could perform a spatially-demanding recollective task, we gave this task to patients with SD and fvFTD. Both groups performed relatively well (especially patients with fvFTD), although some cases with SD were unable to complete the eight item problem (Lee, Rahman, Hodges, Sahakian & Graham, submitted). The results of this experiment suggest that recollection is not a uniform process dependent upon the hippocampus but instead reflects the nature of the material being processed: spatial information does recruit hippocampal structures while tasks that require temporal judgement are more likely to be critically dependent upon frontal processes. This hypothesis is being explored in more detail by Dr Andrew Graham, a Wellcome Training Fellow at the CBU (2001-2004).

M2.2.7 Summary

Our new learning experiments in patients with SD revealed that although semantic breakdown impairs anterograde verbal recall, other aspects of episodic memory, including recognition of (identical) pictures of objects and faces, source discrimination and even associative learning of visual and spatial information, can function independently of semantic memory. These findings are contrary to Tulving's (2001) view that the acquisition of episodic memory depends upon an intact semantic system. Instead, we have proposed that while episodic memory typically draws upon multiple inputs from perceptual and semantic systems, even when semantic knowledge is degraded perceptual information can be sufficient to support some forms of new learning (Simons et al., in press).

In terms of the opposite dissociation – good semantic memory in the context of poor episodic memory - our studies of vocabulary in adult amnesics are inconsistent with recent findings in developmental amnesia that the acquisition of episodic and semantic memory are dependent upon different MTL systems (see M2.1.5). Instead, we have argued that, while some semantic learning can occur in amnesia (via slow cortical learning), this will be qualitatively different from and less detailed than normal conceptual knowledge. At present, therefore, our findings are consistent with the idea that there is a single MTL system supporting the acquisition of both episodic and semantic memories, although we acknowledge that current cognitive tests of memory may not be sufficiently sensitive to discriminate the distinct functions of regions within the MTL.

Project M3: The role of semantic knowledge and other factors in language and its disorders

Scientific Direction: Patterson (50%), Hodges (10%), K. Graham (10%)
MRC-supported scientists: T.Rogers (40%), Bak (40%)
Grant-supported scientists: Bird (100%), Lambon Ralph (50%)
Research support: Wilkinson (34%), Erzinclioglu (33%), Drake (33%)
Grant-funded research support: Ellis (100%)
MRC-supported students: Bozeat (100%)
Grant-supported students: Knott (100%), Cumming (100%)

Much of this programme of research, consisting of neuropsychological studies and some modelling, is based on our working hypothesis that SD represents a relatively pure disintegration of conceptual knowledge, and that all observed cognitive deficits relate to this central underlying impairment. This predicts that aspects of language transparently dependent on semantic knowledge – particularly receptive and expressive vocabulary – must be compromised in SD, and the challenge is then to characterise these deficits in detailed and theoretically informed ways, to relate them to affected brain structures, to capture them in computational models, etc. But there is another sort of challenge: if our assumption is correct, then any documented deficits in aspects of language that are not so obviously dependent on semantic knowledge should also be interpretable in terms of the impact of a degraded semantic system. Language tasks under this description that have been a focus of attention in our work over the last quinquennium are reading aloud, spelling to dictation, verb inflection and verbal short-term memory. Our review begins with the more clearly semantic aspects of language, and then moves on to these other less expected linguistic bed-fellows of a deteriorating semantic system. The review also mentions those "other factors" in the project title: we are attempting to understand some non-semantic (mainly phonological) contributions to language disorders by studying Broca's aphasia in patients with lesions from cerebrovascular accident (CVA), and nonfluent progressive aphasia (the form of language disturbance in SD being fluent progressive aphasia; a description of the contrast between these two forms of progressive aphasia can be found in Patterson, Graham, Lambon Ralph & Hodges, in press).

M3.1 Naming and comprehension

Anomia is usually the most prominent presenting symptom of SD, and deficits in the comprehension of content-word vocabulary are rarely far behind. For example, when patients come to the neurology clinic, one of John Hodges's standard interview questions is "Do you have any hobbies?" Many SD patients, even at an early stage, reply "What's a hobby?". The deterioration of both expressive and receptive vocabulary is first and foremost a function of word frequency or familiarity (Lambon Ralph, Graham, Ellis & Hodges, 1998a). With regard to speech production, we demonstrated this effect by taking samples of narrative speech produced by normal individuals describing the "Cookie Theft" picture from the Boston Diagnostic Aphasia Examination (in which a boy in a kitchen is standing on a tipping stool trying to reach the cookies, and water from the sink is overflowing on to the floor). We then changed these samples by replacing all lower-frequency words with more common ones that would fit the context (e.g. stool was replaced by thing, overflowing by coming out, etc.). The distributions of word characteristics in these 'normal' narratives stripped of any lower-frequency nouns and verbs were good matches to the distributions of words in Cookie Theft narratives produced by SD patients, even on variables other than frequency (Bird et al., 2000).

Why are SD patients so profoundly anomic? The prevailing view in the literature on aphasia resulting from CVA is that naming difficulties can arise from several different underlying deficits: at a minimum, from disruption to knowledge of the concepts to be named (semantics), or to the representations of their names (phonology), or to communication between semantics and phonology (the semantics → phonology arrow). Disruption to phonological representations per se seems unlikely to be a major contributor to anomia in SD. Phonological errors (in which component phonemes of the target word are omitted or misplaced or exchanged, non-target phonemes are inserted, etc.) are common in most types of aphasia but very rare in SD naming or spontaneous speech; and the patients' virtually flawless repetition of single words likewise suggests uncorrupted phonological representations for speech production. In our earlier work, however, we did propose that anomia in SD could arise not only from semantic deterioration but additionally from insufficient activation of phonology by meaning. This was because we observed profound progressive anomia in a patient (FM) with only a mild and rather stable semantic deficit; indeed we labelled FM's pattern "progressive pure anomia" (K.S. Graham, Patterson & Hodges, 1995). Subsequent work, however, has led us to a somewhat different interpretation of these results.

All SD patients are anomic, but some are more anomic than others; and (with a nod towards George Orwell) it turns out that the 'less equal' ones in this instance are patients with greater atrophy on the left than the right (L>R). FM, whose atrophy had a strong left dominance, named 19% of a set of common objects at first presentation; a year later, when she was still functioning well in most aspects of cognitive testing and daily life, her naming had reduced to 6%. This suggested that the anomia in L>R cases reflected a deficit in semantics → phonology as well as in semantics itself.

A more satisfactory solution arose, however, from observations and hypotheses regarding the localisation of different aspects of language in the brain. Everything that we know about SD fits a hypothesis of bilateral temporal representation of semantics; but more than a century of research on aphasia suggests that speech production is left lateralised in virtually all right-handers and most left-handers. We therefore developed the hypothesis, and a corresponding computational model, in which semantic representations are bilaterally distributed, but phonological representations (a) are confined to the left hemisphere, and (b) receive much stronger inputs from their neighbouring semantic units on the left than from the more distant ones on the right (Lambon Ralph et al., 2001). As demonstrated in Figure 4, taken from that article, this model, when 'damaged' bilaterally but asymmetrically, produced an excellent fit to the relationship between naming and comprehension deficits in both R>L and L>R SD patients. FM's combination of naming and word-picture matching performance at one year post-presentation is almost perfectly captured by the point of sharpest inflection in the L>R simulation curve, with 88% correct word-picture matching but only 6% correct naming. In other words, even this extreme discrepancy between comprehension and naming scores can arise from a reduction in the neuron-like units comprising semantic representations, provided that these happen to be the semantic units most important for activating phonological representations

Figure 4. Comprehension vs. naming: direct comparison between grouped patient data and simulation results from a connectionist model. Patients: R>L = cases with more right than left temporal atrophy; L>R = cases with more left than right atrophy. Simulation: R>L = more right than left semantic unit damage; L>R = more left than right semantic unit damage.

M3.2 Two of the three R's: reading and riting

Our work on the impact of semantic degradation on 'non-semantic' language tasks began with a study of reading in SD patients, in whom we documented a consistent pattern of surface dyslexia (Patterson & Hodges, 1992; K.S. Graham, Hodges & Patterson, 1994). Surface dyslexia is a reading disorder characterised by a frequency-by-regularity interaction in accuracy of single-word oral reading. The majority of errors occur to lower-frequency words with atypical spelling-sound correspondences, and almost always take the form of 'regularised' pronunciations (pint pronounced to rhyme with "mint"; gauge pronounced "gawge", etc.). In contrast to the dual-route theory of reading, our view is that the association between semantic impairment and surface dyslexia is meaningful and causal. This account, developed over a number of years, is based on a connectionist model of word reading, not only in English (Plaut et al., 1996; Seidenberg & McClelland, 1989) but also in Japanese (Fushimi, Ijuin, Patterson & Tatsumi, 1999; Ijuin, Fushimi, Patterson & Tatsumi, 1999), in which the basic mechanism for translating a written word into a pronunciation is a network relying on learned correspondences between orthography and phonology at different sizes/levels. What is learned is not abstract symbolic grapheme-phoneme rules but rather connection weights based on statistical typicalities. Highly frequent orthographic sequences will be well learned by the network; and because knowledge about translation to phonology operates over both small and large chunks, words will be correctly and efficiently pronounced if they have either (a) typical component correspondences even if the whole word is not frequent (e.g. nave or cove), or (b) whole-word high frequency even if the component correspondences are not typical (e.g. have or move). Lower-frequency words with atypical correspondences, however, will not be well learned by this system, and thus something must counteract the tendency of the unimpaired network to produce regularised pronunciations of words like pint and gauge. Our hypothesis is that this additional source of word-specific constraint comes from word meaning, a part of the system already well established to differentiate between specific words. In the kind of interactive system that we postulate, meaning will be activated for all words, whatever their frequency or regularity characteristics; but this additional semantic activation will be especially important for error-free processing of the low-frequency exceptions. The hypothesis thus predicts that any serious disruption to semantics will lead to surface dyslexia, as it indeed seems to do in SD and other conditions mimicking it, such as the occasional traumatic brain injury causing severe left inferior temporal damage (Patterson & Behrmann, 1997). A more general version of this hypothesis suggests that all components of the ability to read (which is a late-acquired skill in the development of both the individual human child and the human species) are likely to depend on other cognitive functions, and thus that deficits in these reading components are always likely to be associated with disruption to other, earlier acquired functions. A case for a meaningful association of surface dyslexia with semantic deficits, of phonological dyslexia with malfunction of general language phonology, and of pure alexia with visual-processing deficits was made by Patterson & Lambon Ralph (1999).

As so often, AD requires a slightly more complicated story because semantic deterioration is only one part of the syndrome. Nevertheless, we have demonstrated a highly reliable increase in errors to lower-frequency irregular words as a function of the progression of AD, plus a striking increase in word-reading response times (RTs) over longitudinal assessments (Strain, Patterson, Graham & Hodges, 1998). When first assessed, our cohort of AD patients had average word-reading RTs equivalent to those of their age-matched controls (mean ≈ 650 ms for single-syllable words). About two years later (same patients, same words), correct reading responses to the three easiest word sets (high-frequency regular, low-frequency regular, high-frequency irregular) averaged around 860 ms, with correct RTs to the most difficult words (low-frequency irregular) even slower, ≈ 930 ms.

The ability to spell words to dictation receives the same treatment and prediction in our model, only more so, because the sound-to-spelling correspondences in English are even more unpredictable and one-to-many than spelling-to-sound correspondences for reading. The data support this prediction: all of the SD patients that we have studied have been surface dysgraphic, making many errors of spelling words as they sound (e.g. "giraffe" → jeraf, "tongue" → tung) (N.L. Graham, Patterson & Hodges, 2000); furthermore, this surface deficit is almost invariably revealed earlier and/or more severely in spelling than in reading. It is true that many normal adults are less skilled spellers than readers; but for some of our very dysgraphic SD patients we have evidence of excellent pre-morbid spelling. We were able to track spelling performance in FM, the profoundly anomic SD patient described above, for seven years (N.L.Graham, Patterson & Hodges, 2001). She started out as a pure surface dysgraphic case but, as her speech production and comprehension declined with SD progression, the 'control' of her spelling by phonology also loosened: at the end of this longitudinal study FM's responses in spelling to dictation bore virtually no relationship to what was dictated. The great majority of her productions were, however, still orthographically word-like and composed of letters in proportion to their frequencies of occurrence in the English vocabulary. These results demonstrate the preservation of bare bones orthographic knowledge, the production of which was no longer under any stimulus control.

M3.3 Verb inflections

M3.3.1 The impact of semantic impairment on inflecting verbs

The dual-mechanism (or "Words and Rules") account of language processes proposed by Pinker (1999) incorporates a symbolic rule system to deal with regular exemplars and an associative lexical memory to handle exceptions. In our contrasting connectionist account, a single complex procedure processes both regular and irregular items. Dissociations between success with these two classes are attributed to more general impairments in the semantic or phonological subsystems of language which, we claim, are unequally stressed by irregular and regular words. The arguments here are similar to those in reading, at least with regard to the impact of semantic impairment. That is, we predicted that SD patients would have a frequency-modulated selective difficulty with producing or recognising the correct past tense forms of irregular verbs, which they did (Patterson, Lambon Ralph, Hodges & McClelland, 2001: see Table 1 for their success in generating past-tense verb forms in a sentence completion experiment). The pattern of error types was especially consistent with the connectionist account, as was the fact that the degree of the irregular deficit across patients correlated with their performance on a synonym-judgement comprehension task on the same verbs. The slope of the function relating past-tense production to comprehension was 0.88 for the 50 irregular verbs and 0.08 for the 50 regular items.

Table 1: Mean percent correct for SD patients (n=8) in generating the past-tense forms of high- and low-frequency regular and irregular verbs.

High Freq Low Freq

Regular 98 96

Irregular 71 53

M3.3.2 The impact of phonological impairment on inflecting verbs

Although the research described next does not address the issue of the impact of semantic deficits on language processing, it seems sensible to include our investigation of the other side of the verb inflection issue here. Pinker's (1999) model predicts that the rule system responsible for the regular –ed inflection should be independently vulnerable to disruption by brain injury. Ullman, Pinker and colleagues (Ullman et al., 1997) published data demonstrating that one patient with Broca's aphasia was dramatically better at generating the past-tense of irregular than regular verbs in a sentence completion task, and that a larger number of similar patients who were too impaired to perform the generation task showed the irregular > regular pattern in reading past-tense verb forms. Our account of this side of the putative dissociation invokes phonological processes, with which Broca's aphasics have substantial difficulties. Compared to irregular past-tense verbs, regular past-tense forms are on average more phonologically complex (indeed, in an analysis by Burzio, 2002, they are frankly phonologically irregular) because they contain terminal consonant clusters like /vd/ and /gd/ (as in "loved" and "dragged") that never occur in irregular past-tense forms and indeed never occur in any words in English apart from past-tense regular verbs. In a study of 10 Broca's aphasic cases (Bird, Lambon Ralph, Seidenberg, McClelland & Patterson, in press), we demonstrated a significant irregular > regular advantage in three production tasks (sentence completion, repetition and reading) using materials like those in Ullman et al. In a subsequent experiment where we matched the consonant-vowel structure of regular and irregular past tense forms, the discrepancy in performance for these same patients completely disappeared in sentence completion and repetition. A remaining irregular advantage in reading was attributed to concreteness effects. Our interpretation of this and other evidence (McClelland & Patterson, in press) is that there is no compelling demonstration of an augmented deficit in producing the regular past tense that cannot be attributed to a phonological source.

M3.4 Auditory-verbal working memory (AVWM)

In our earlier research, we demonstrated that three SD patients, all with normal digit span and thus no prominent deficit in AVWM, had a high rate of phonological migration errors in immediate serial recall (ISR) of sequences of 3-4 unrelated words (Patterson, Graham & Hodges, 1994). For example, a sequence like "pencil, chicken, sword" might be repeated back as "sencil, sicken, pord". Lists for ISR were constructed for individual patients, with each list composed exclusively of words that were either 'known' or 'degraded' for that patient. So-called 'known' items were words that the patients could either still name from a picture, if they were concrete concepts, or words that they were still using appropriately in spontaneous speech; 'degraded' items were words that the patients failed to name and also failed to process correctly in word-picture matching or synonym judgement tasks. The fact that most of the ISR errors occurred to 'degraded' words, even with the two sets matched as closely as possible for factors such as word frequency, suggests that semantic knowledge about a word plays some important role in binding its phonological features, especially under conditions of high AVWM load. This phenomenon and hypothesis, and an extension of it to repetition of a single word after a short filled delay, were explored further in PhD research by Raymond Knott (Knott, Patterson & Hodges, 1997; 2000). These deficits may also be germane to our recent observations of poor word-list learning in SD. In a study in which lists of words (selected individually for each of seven patients as 'known' or 'degraded') were presented three times each (in different orders) and followed by free recall after each presentation, the patients' performance displayed a significant advantage for 'known' > 'degraded' but was sub-normal even for 'known' words (K.S. Graham et al., 2002). As in the ISR paradigm, they made many phonological blend errors on 'degraded' words.

M3.5 Non-Fluent Progressive Aphasia

Progressive aphasia comes in two main syndromes or patterns: fluent but anomic speech associated with temporal-variant Pick's disease (SD), and nonfluent speech with phonological and syntactic abnormalities. The latter, known as non-fluent progressive aphasia (NFPA), appears to arise from several different degenerative aetiologies, principally (at least in our case series) AD (Croot, Hodges, Xuereb & Patterson, 2000) or corticobasal degeneration (N.L. Graham, Bak, Patterson & Hodges, submitted). Our work has concentrated far more on the fluent syndrome: apart from some basic comparisons of the main presenting features of the two syndromes (e.g. Hodges & Patterson, 1996; Patterson et al., in press), our published research on NFPA to date has consisted mainly of the work by Karen Croot, who was a PhD student with our group from 1994-1998. As well as characterising a number of cases of the atypical nonfluent aphasic presentation of AD (Croot et al., 2000), she studied word production by four patients with NFPA (Croot, Patterson & Hodges, 1998; Croot, Hodges & Patterson, 1999a) with the goal of relating their impairments in single-word production tasks like picture naming, word repetition and word reading to an interactive spreading-activation model of normal and impaired speech production (Dell, Schwartz, Martin, Saffran & Gagnon, 1997). This model has two independently manipulable parameters: the strength of the connections between nodes in the network, and the rate at which activation in the network, once initiated, decays. Abnormality in either of these parameters disrupts speech production; but the aphasic profiles associated with the two should be distinct with respect to severity of impairment in different speaking tasks (e.g. naming objects vs. repeating spoken words), error types, and other features. We observed examples of both aphasic profiles amongst our NFPA cases. Intriguingly, two cases representing one example of each type were brothers who subsequently came to post-mortem examination and were shown to have markedly different distributions of atrophy in peri- and extra-sylvian left-hemisphere regions.

M3.6 Summary

Our research on language has led to the hypothesis that semantics and phonology (and their interaction) are central to all aspects of language processing, and that essentially every pattern of language disorder can be traced back to disruption in one or other of these fundamental systems. Most of our research in the last quinquennium has concentrated on the wide-ranging impact of semantic disruption which – harking back to M1 – yields language performance dominated by typicality. This pattern of greater vulnerability for words with atypical transformations from spelling to sound, or stem to past tense, etc, is in essence an exaggeration of the normal pattern of processing. Phonological disorders, on the other hand, produce a more complicated range of language impairments. The incorporation of neuroanatomical hypotheses into our processing models, particularly that the semantic system is bilaterally represented but that phonology is left-lateralised, significantly improves success in capturing the range of impaired language performance.

AWARDS AND HONOURS

John Hodges:

• elected President of the British Neuro-Psychiatric Association 1998

• elected President of the World Federation of Neurology, Research Group on Aphasia and Cognitive Disorders 1998

• appointed Director of the Alzheimer's Research Trust Centre for Cambridge 1998

• elected Fellow of the Academy of Medical Sciences 2002

Karalyn Patterson:

• elected International Fellow of the American Psychological Society 2001

• nominated Fellow of the Academy of Medical Sciences (election 2003)

• keynote lecturer for the International Congress on Spoken Language Processing, Beijing, 2000

Kim Graham:

• elected member of the Memory Disorders Research Society 1998 (limited to 100 members internationally)

• invited speaker at the Royal Society of Edinburgh 2002

PUBLICATIONS

Refereed journals

BAK, T.H., Antoun, N., Balan, K.K., & HODGES, J.R. (2001a). Memory lost, memory regained: Neuropsychological findings and neuroimaging in two cases of paraneoplastic limbic encephalitis with radically different outcomes. Journal of Neurology, Neurosurgery and Psychiatry, 71, 40-47.

BAK, T.H., & HODGES, J.R. (2001). Motor neuron disease, dementia and aphasia: coincidence, co-occurrence or continuum? Journal of Neurology, 248, 260-270.

BAK, T.H., & HODGES, J.R. (in press). "Kissing and Dancing" - a test to distinguish the lexical and conceptual contributions to noun/verb and action/object dissociation. Preliminary results in patients with frontotemporal dementia. Journal of Neurolinguistics.

BAK, T.H., O'Donovan, D.G., Xuereb, J.H., Boniface, S., & HODGES, J.R. (2001b). Selective impairment of verb processing associated with pathological changes in the Brodman areas 44 and 45 in the motor neurone disease/dementia/aphasia syndrome. Brain, 124, 103-124.

BIRD, H., LAMBON RALPH, M.A., PATTERSON, K., & HODGES, J.R. (2000). The rise and fall of frequency and imageability: Noun and verb production in semantic dementia. Brain & Language, 73, 17-49.

BIRD, H., Lambon Ralph, M.A., Seidenberg, M.S., McClelland, J.L., & PATTERSON, K. (in press). Deficits in phonology and past-tense morphology: What's the connection? Journal of Memory & Language.

BOZEAT, S., LAMBON RALPH, M.A., GRAHAM, K. S., PATTERSON, K., Wilkin, H., Rowland, J., ROGERS, T.T., & HODGES, J.R. (in press-a). A duck with four legs: Investigating the structure of conceptual knowledge using picture drawing in semantic dementia. Cognitive Neuropsychology.

BOZEAT, S., LAMBON RALPH, M.A., PATTERSON, K., Garrard, P., & HODGES, J.R. (2000). Non-verbal semantic impairment in semantic dementia. Neuropsychologia, 38, 1207-1215.

BOZEAT, S., LAMBON RALPH, M.A., PATTERSON, K., & HODGES, J.R. (2002). The influence of personal familiarity and context on object use in semantic dementia. Neurocase, 8, 127-134.

BOZEAT, S., LAMBON RALPH, M.A., PATTERSON, K., & HODGES, J.R. (in press-b). When objects lose their meaning: What happens to their use? Cognitive, Affective and Behavioural Neuroscience.

Caine D., PATTERSON K., HODGES J.R., Heard R., & Halliday G. (2001). Severe anterograde amnesia, with extensive hippocampal degeneration in a case of rapidly progressive fronto-temporal dementia. Neurocase, 7, 57-64.

Calderon, J., Perry, R.J., ERZINCLIOGLU, S., Berrios, G.E., Dening, T., & HODGES, J.R. (2001). Perception, attention and working memory are disproportionately impaired in dementia with Lewy Body (LBD) compared to Alzheimer's disease (AD). Journal of Neurology, Neurosurgery & Psychiatry, 70, 157-164.

CLARE, L., WILSON, B.A., Breen, K. & HODGES, J.R. (1999). Errorless learning of face-name associations in early Alzheimer's disease. Neurocase, 5, 37-46.

CLARE, L., WILSON, B.A., CARTER, G., Breen, K., GOSSES, A., & HODGES, J.R. (2000). Intervening with everyday memory problems in dementia of Alzheimer type. Journal of Clinical and Experimental Neuropsychology, 22, 132-146.

CLARE, L., WILSON, B.A., CARTER, G., & HODGES, J.R. (2001). Long-term maintenance of treatment gains following a cognitive rehabilitation intervention in early dementia of Alzheimer type. Neuropsychological Rehabilitation, 11, 477-494.

CROOT, K., HODGES, J.R., & PATTERSON, K. (1999a). Evidence for impaired sentence comprehension in early Alzheimer's disease. Journal of the International Neuropsychological Society, 5, 393-404.

CROOT, K., HODGES, J.R., Xuereb, J., & PATTERSON, K, (2000). Phonological and articulatory impairment in Alzheimer's disease. Brain and Language, 75, 277-309.

CROOT, K., PATTERSON, K., & HODGES, J.R. (1998). Single word production in nonfluent progessive aphasia. Brain & Language, 61, 226-273.

CROOT, K., PATTERSON, K., & HODGES, J.R. (1999b). Familial progressive aphasia: Insights into the nature and deterioration of single word processing. Cognitive Neuropsychology, 16, 705-747.

EPSTEIN, R., Downing, P., & GRAHAM, K.S. (submitted). Scene representations in human parahippocampal cortex are viewpoint-specific.

Evans, J., GRAHAM, K.S., Pratt, K.H., & HODGES, J.R. (in press). The impact of disrupted cortico-cortical connectivity: A long term follow-up of a case of focal retrograde amnesia. Cortex.

Fushimi, T., Ijuin, M., PATTERSON, K., & Tatsumi, I.F. (1999). Consistency, frequency, and lexicality effects in naming Japanese Kanji. Journal of Experimental Psychology: Human Perception and Performance, 25, 382-407.

Galton, C.J., Gomez-Anson, B., Antoun, N., Scheltens, P., PATTERSON, K., Graves, M., Sahakian, B.J., & HODGES, J.R. (2001a). Temporal lobe rating scale: Application to Alzheimer's disease and frontotemporal dementia. Journal of Neurology, Neurosurgery and Psychiatry, 70, 165-173.

Galton, C.J., PATTERSON, K., GRAHAM, K.S., LAMBON-RALPH, M., Williams, G., Antoun, N., Sahakian, B.J., & HODGES, J.R. (2001b). Differing patterns of temporal atrophy in Alzheimer's disease and semantic dementia. Neurology, 57, 216-225.

Galton, C.J., PATTERSON, K., Xuereb, J.H., & HODGES, J.R. (2000). Atypical and typical presentations of Alzheimer's disease: A clinical, neuropsychological, neuroimaging and pathological study of 13 cases. Brain, 123, 484-498.

Garrard, P., & HODGES, J.R. (1999). Semantic dementia: Implications for the neural basis of language and meaning. Aphasiology, 13, 609-623.

Garrard, P., & HODGES, J.R. (2000). Semantic dementia: Clinical, radiological and pathological perspectives. Journal of Neurology, 247, 409-422.

Garrard, P., LAMBON RALPH, M.A., PATTERSON, K., & HODGES, J.R. (2001a). Prototypicality, distinctiveness and intercorrelation: Analyses of the semantic attributes of living and non-living concepts. Cognitive Neuropsychology, 18, 125-174.

Garrard, P., LAMBON RALPH, M.A., WATSON, P.C., Powis, J., HODGES, J.R., & PATTERSON, K. (2001b). Longitudinal profiles of semantic impairment for living and nonliving concepts in dementia of Alzheimer's type. Journal of Cognitive Neuroscience, 13, 892-909.

Garrard, P., PATTERSON, K., WATSON, P.C., & HODGES, J.R. (1998). Category specific semantic loss in dementia of Alzheimer's type: Functional-anatomical correlations from cross-sectional analyses. Brain, 121, 633-646.

GRAHAM, K.S., Goldman, W.P., KROPELNICKI, A., Greene, J.D.W. & HODGES, J.R. (submitted-a). Challenging the role of the hippocampus in extensive autobiographical memory deficits: Evidence suggesting critical involvement of frontal executive processes.

GRAHAM, K. S., KROPELNICKI, A., Goldman, W.P., & HODGES, J.R. (in press-a). Two further investigations of autobiographical memory in semantic dementia. Cortex.

GRAHAM, K.S., LAMBON RALPH, M., & HODGES, J.R (1999a). A questionable semantics: The interactions between semantic knowledge and autobiographical experience in semantic dementia. Cognitive Neuropsychology, 16, 689-698.

GRAHAM, K.S., LEE, A.C.H., BRETT, M., & PATTERSON, K. (submitted-b). The neural basis of autobiographical and semantic memory: New evidence from three PET studies.

GRAHAM, K.S., PATTERSON, K., & HODGES, J.R. (1999b). Episodic memory: New insights from the study of semantic dementia. Current Opinion in Neurobiology, 9, 245-250.

GRAHAM, K.S., PATTERSON, K., Pratt, K.H., & HODGES J.R. (1999c). Relearning and subsequent forgetting of semantic category exemplars in a case of semantic dementia. Neuropsychology, 13, 359-380.

GRAHAM, K.S., PATTERSON, K., Pratt, K.H., & HODGES, J.R. (2001). Can repeated exposure to 'forgotten' vocabulary help alleviate word-finding difficulties in semantic dementia? An illustrative case study. Neuropsychological Rehabilitation, 11, 429-454.

GRAHAM, K.S., PATTERSON, K., Powis, J., DRAKE, J., & HODGES, J.R. (2002). Multiple inputs to episodic memory: Words tell another story. Neuropsychology, 16, 344-358.

GRAHAM, K.S., Pratt, K.H., & HODGES, J.R. (1998). A reverse temporal gradient for public events in a single case of semantic dementia. Neurocase, 4, 461-470.

GRAHAM, K.S., SIMONS, J.S., Pratt, K.H., PATTERSON, K., & HODGES, J.R. (2000). Insights from semantic dementia on the relationship between episodic and semantic memory. Neuropsychologia, 38, 313-324.

GRAHAM, N.L., BAK, T., PATTERSON, K., & HODGES, J.R. (submitted). Language function and dysfunction in corticobasal degeneration.

GRAHAM, N.L., PATTERSON, K., & HODGES, J.R. (2000). The impact of semantic memory impairment on spelling: Evidence from semantic dementia. Neuropsychologia, 38, 143-163.

GRAHAM, N.L., PATTERSON, K., & HODGES, J.R. (2001). The emergence of jargon in progressive fluent dysgraphia: The widening gap between target and response. Cognitive Neuropsychology, 18, 343-361.

GRAHAM, N.L., Zeman, A., Young, A., PATTERSON, K., & HODGES, J.R. (1999). Dyspraxia in a patient with corticobasal degeneration: The role of visual and tactile inputs to action. Journal of Neurology, Neurosurgery, and Psychiatry, 67, 334-344.

Gregory, C., Lough, S., Stone, V., ERZINCLIOGLU, S.W., Martin, L., Baron-Cohen, S., & HODGES, J.H. (2002). Theory of mind in patients with frontal variant frontotemporal dementia and Alzheimer's disease: Theoretical and practical applications. Brain, 125, 752-764.

HODGES, J.R., BOZEAT, S., LAMBON RALPH, M.A., PATTERSON, K., & Spatt, J. (2000). The role of conceptual knowledge in object use: Evidence from semantic dementia. Brain, 123, 1913-1925.

HODGES, J.R., & GRAHAM, K.S. (1998). A reversal of the temporal gradient for person knowledge in semantic dementia: Implications for the neural organisation of long-term memory. Neuropsychologia, 36, 803-825.

HODGES, J.R., & GRAHAM, K.S. (2001). Episodic memory: Insights from semantic dementia. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 356, 1423-1434.

HODGES, J.R., & Greene, J.D.W. (1998). Knowing people and naming them: Can Alzheimer's patients do one and not the other? Quarterly Journal of Experimental Psychology, 51A, 121-134.

HODGES, J.R., & Miller, B.L. (2001a). The classification, genetics and neuropathology of frontotemporal dementia (FTD). Introduction to the special topic papers: Part I. Neurocase, 7, 31-35.

HODGES, J.R., & Miller, B.L. (2001b). The neuropsychology of frontal variant FTD and semantic dementia. Introduction to the special topic papers: Part II. Neurocase, 7, 113-121.

HODGES, J.R., Spatt, J., & PATTERSON, K. (1999). What and how: Evidence for the dissociation of object knowledge and mechanical problem solving skills in the human brain. Proceedings of the National Academy of Sciences, 96, 9444-9448.

Hovius, M., KELLENBACH, M.L., GRAHAM, K.S., HODGES, J.R., & PATTERSON, K. (in press). What does the object decision task measure? Reflections on the basis of evidence from semantic dementia. Neuropsychology.

Ijuin, M., Fushimi, T., PATTERSON, K., & Tatsumi, I. (1999). A connectionist approach to Japanese Kanji word naming. Psychologia, 42, 267-280.

KELLENBACH, M.L., BRETT, M., & PATTERSON, K. (2001). Large, colorful, or noisy? Attribute- and modality-specific activations during retrieval of perceptual attribute knowledge. Cognitive, Affective & Behavioral Neuroscience, 1, 207-221.

KELLENBACH, M.L., BRETT, M., & PATTERSON, K. (in press). Actions speak louder than functions: The importance of manipulability and action in tool representation. Journal of Cognitive Neuroscience.

KELLENBACH, M.L., Hovius, M., & PATTERSON, K. (submitted). A PET study of visual and semantic knowledge about objects.

Kitchener, E.G. & HODGES, J.R. (1999). Impaired knowledge of famous people and events with intact autobiographical memory in a case of progressive right temporal lobe degeneration: Implications for the organisation of remote memory. Cognitive Neuropsychology, 16, 589-607.

Kitchener, E.G., HODGES, J.R., & McCarthy, R. (1998). Acquisition of post-morbid vocabulary and semantic facts in the absence of episodic memory. Brain, 121, 1313-1327.

KNOTT, R., PATTERSON, K., & HODGES, J.R. (2000). The role of speech production in auditory-verbal short-term memory. Neuropsychologia, 38, 125-142.

LAMBON RALPH, M.A., Cipolotti, L., & PATTERSON, K. (1999a). Oral naming and oral reading: Do they speak the same language? Cognitive Neuropsychology, 16, 157-169.

LAMBON RALPH, M.A., GRAHAM, K.S., Ellis, A., HODGES, J.R. (1998a) Naming in semantic dementia - what matters? Neuropsychologia, 36, 775-784.

LAMBON RALPH, M.A., GRAHAM, K.S., PATTERSON, K., & HODGES, J.R. (1999b). Is a picture worth a thousand words? Evidence from concept definitions by patients with semantic dementia. Brain and Language, 70, 309-335.

LAMBON RALPH, M.A., Howard, D., Nightingale, G., & Ellis, A.W. (1998b). Are living and non living category-specific deficits causally linked to impaired perceptual or associative knowledge? Evidence from a category-specific double dissociation. Neurocase, 4, 311-338.

LAMBON RALPH, M.A., McClelland, J.L., PATTERSON, K., Galton, C., & HODGES, J.R. (2001). No right to speak? The relationship between object naming and semantic impairment: Neuropsychological evidence and a computational model. Journal of Cognitive Neuroscience, 13, 341-356.

LAMBON RALPH, M.A., McClelland, J.L., PATTERSON, K., & HODGES, J.R. (2000). The relationship between semantic memory and speech production: Neuropsychology, neuroanatomy and a neural-net model. Higher Brain Function Research, 20, 145-156.

LAMBON RALPH, M.A., PATTERSON, K., Garrard, P., & HODGES, J.R. (submitted). Semantic dementia with category specificity: A comparative case-series study.

LEE, A.C.H., GRAHAM, K.S., Simons, J.S., HODGES, J.R., OWEN, A.M., & PATTERSON, K. (2002a). Regional brain activations differ for semantic features but not categories. Neuroreport, 13, 1497-1501.

LEE, A.C.H., Rahman, S., HODGES, J.R., Sahakian, B., & GRAHAM, K.S. (submitted). Associative and recognition memory for novel objects in dementia: Implications for diagnosis.

LEE, A.C.H., Robbins, T.W., GRAHAM, K.S., & OWEN, A.M. (2002b). Pray or prey? Dissociation of semantic memory access from episodic memory processes using positron emission tomography and a novel homophone task. Neuroimage, 16, 724-735.

Manes, F., Zeman, A., GRAHAM, K.S., & HODGES, J.R. (2001). Focal autobiographical amnesia in association with transient epileptic amnesia. Brain, 124, 499-509.

McClelland, J.L., & PATTERSON, K. (in press). Rules or connections in past-tense inflections: What does the evidence rule out? Trends in Cognitive Sciences.

Mummery, C.J., PATTERSON, K., HODGES, J.R., Wise, R.J.S., & Price, C.J. (1998) Functional neuroanatomy of the semantic system - divisible by what? Journal of Cognitive Neuroscience, 10, 766-777.

Mummery, C.J., PATTERSON, K., Price, C.J., Ashburner, J., Frackowiak, R.S.J., & HODGES, J.R. (2000). A voxel based morphometry study of semantic dementia: The relationship between temporal lobe atrophy and semantic dementia. Annals of Neurology, 47, 36-45

Mummery, C.J., PATTERSON, K., Wise, R.J.S., Vandenberghe, R., Price, C.J., & HODGES, J.R. (1999). Disrupted temporal lobe connections in semantic dementia. Brain, 122, 61-73.

Murre, J.M.J., GRAHAM, K.S., & HODGES, J.R. (2001). Semantic dementia: Relevance to connectionist models of long-term memory. Brain, 124, 647-675.

Nestor, P.J., Fryer, T.D., Smielewski, P., & HODGES, J.R. (submitted). Limbic hypometabolism in Alzheimer's disease and mild cognitive impairment.

Nestor, P.J., GRAHAM, K.S., BOZEAT, S., SIMONS, J.S., & HODGES, J.R. (2002). Memory consolidation and the hippocampus: Further evidence from the study of autobiographical memory in semantic dementia and the frontal variant of frontotemporal dementia. Neuropsychologia, 40, 633-54.

PATTERSON, K. (2000). The impact of semantic degradation on nonsemantic language abilities. Higher Brain Function Research, 20, 85-98.

PATTERSON, K., & LAMBON RALPH, M.A. (1999). Selective disorders of reading? Current Opinion in Neurobiology, 9, 235-239.

PATTERSON, K., LAMBON RALPH, M.A., HODGES, J.R., & McClelland, J.L. (2001). Deficits in irregular past-tense verb morphology associated with degraded semantic knowledge. Neuropsychologia, 39, 709-724.

PATTERSON, K., Okada, S., Suzuki, T., Ijuin, M., & Tatsumi, I. (1998). Fragmented words: A case of late-stage progressive aphasia. Neurocase, 4, 219-230.

Perry, R.J. & HODGES, J.R (2000) The fate of patients with questionable (very mild) Alzheimer's disease: Longitudinal profiles of individual subjects' decline. Dementia and Geriatric Cognitive Disorders, 11, 342-349.

Perry, R.J., WATSON, P.C. & HODGES, J.R. (2000) The nature and staging of attention dysfunction in early (minimal and mild) Alzheimer's disease: Relationship to episodic and semantic memory impairment. Neuropsychologia, 38, 252-271.

Price, C. J., Howard, D., PATTERSON, K., Friston, K.J., & Frackowiak, R.S.J. (1998). A functional neuroimaging description of two deep dyslexic patients. Journal of Cognitive Neuroscience, 10, 303-315.

Rahman, S., Sahakian, B.J., HODGES, J.R., Rogers, R.D., & Robbins, T.W. (1999). Specific cognitive deficits in mild frontal variant frontotemporal dementia. Brain, 122, 1469-1493.

Reich, S., CHOU, T-L., & PATTERSON, K. (in press). Acquired dysgraphia in Chinese: Further evidence on the links between phonology and orthography. Aphasiology.

ROGERS, T.T., Lambon Ralph, M.A., Garrard, P., BOZEAT, S., McClelland, J.L., HODGES, J.R., & PATTERSON, K. (submitted-a). The structure and deterioration of semantic memory: A neuropsychological and computational investigation.

ROGERS, T.T., Lambon Ralph, M.A., HODGES, J.R., & PATTERSON, K. (submitted-b). Object recognition under semantic impairment: the effects of conceptual regularities on perceptual decisions.

SIMONS, J.S., & GRAHAM, K.S. (2000). New learning in semantic dementia: Implications for cognitive and neuroanatomical models of long-term memory. Revue de Neuropsychologie. 10, 199-215.

SIMONS, J.S., GRAHAM, K.S., Galton, C.J., PATTERSON, K., & HODGES, J.R. (2001a). Semantic knowledge and episodic memory for faces in semantic dementia. Neuropsychology, 15, 101-114.

SIMONS, J., GRAHAM, K.S., & HODGES, J. R. (2002). Perceptual and semantic contributions to episodic memory: Evidence from semantic dementia and Alzheimer's disease. Journal of Memory and Language, 47, 197-213.

SIMONS, J.S., GRAHAM, K.S., OWEN, A.M., PATTERSON, K., & HODGES, J.R. (2001b). Perceptual and semantic components of memory for objects and faces: A PET Study. Journal of Cognitive Neuroscience, 13, 430-443.

SIMONS, J.S., Verfaellie, M., Galton, C.J., Miller, B.L., HODGES, J.R., & GRAHAM, K.S. (in press). Recollection-based memory in frontotemporal dementia: Implications for theories of long-term memory. Brain.

Spatt, J., BAK, T., BOZEAT, S., PATTERSON, K., & HODGES, J.R. (2002). Apraxia, mechanical problem solving and semantic knowledge: Contributions to object usage in corticobasal degeneration. Journal of Neurology, 249, 601-608.

STRAIN, E., PATTERSON, K., GRAHAM, N.L., & HODGES, J.R. (1998). Word reading in Alzheimer's disease: Cross-sectional and longitudinal analyses of response-time and accuracy data. Neuropsychologia, 36, 155-171.

Swainson, R., HODGES, J. R., Galton, C. J., Semple, J., Michael, A., Dunn, B. D., Iddon, J. L., Robbins, T. W., & Sahakian, B. J. (2001). Early detection and differential diagnosis of Alzheimer's disease and depression with neuropsychological tasks. Dementia and Geriatric Cognitive Disorders, 12, 265-280.

Thomas, R. M., Forde, E. M., Humphreys, G. W., & GRAHAM, K. S. (in press). The effects of passage of time on a patient with category-specific agnosia. Neurocase.

Thompson, S.A., GRAHAM, K.S., PATTERSON, K., Sahakian, B.J., & HODGES, J.R. (2002). Is knowledge of famous people disproportionately impaired in patients with early Alzheimer's disease? Neuropsychology, 16, 344-358.

Thompson, S.A., GRAHAM, K.S., Williams, G., PATTERSON, K., Kapur, N., & HODGES, J.R. (submitted). Dissociating person-specific from general semantic knowledge: Roles of the left and right temporal lobes.

Yang, J., Weng, X., Guan, L., Kuang, P., Zhang, M., Sun, W., Yu, S., & PATTERSON, K. (in press). Involvement of the medial temporal lobe in priming for new associations. Neuropsychologia.

Zeman, A.Z.J., Boniface, S.J. & HODGES, J.R. (1998). Transient epileptic amnesia: A description of the clinical and neuropsychological features in 10 cases and a review of the literature. Journal of Neurology, Neurosurgery and Psychiatry, 64, 435-443.

Books

Fara, P., & PATTERSON, K. (Eds). (1998). Memory: The 11th Annual Darwin Lectures. Cambridge: Cambridge University Press.

HODGES, J. R. (Ed.). (2001). Early-Onset Dementia: A Multidisciplinary Approach. Oxford: Oxford University Press.

ROGERS, T.T., & McClelland, J.L. (in press). Semantic Cognition: A Parallel Distributed Processing Approach. Boston, MA: MIT Press.

Theses

KNOTT, R.A. (1998). The impact of lexical and semantic impairment on short-term verbal recall.

SIMONS, J.S. (2000). Episodic learning in semantic dementia.

BOZEAT, S. (2002). Legs and Handles: Insights into object knowledge and object use from an analysis of their impairments.

Book Chapters & Contributions

Garrard, P., Lambon Ralph, M.A., & HODGES, J.R. (2002). Semantic dementia: A category-specific paradox. In E. Forde & G. Humphreys (Eds.), Category-Specificity in Brain and Mind (pp.149-179). Hove: Psychology Press.

GRAHAM, K.S., JOHNSRUDE, I.S. & Simons, J.S. (in press-b). The temporal lobes. In M. Aminoff & R.B. Daroff (Eds.), Encyclopedia of the Neurological Sciences. San Diego: Academic Press.

GRAHAM, K.S., Murre, J. M. J., & HODGES, J.R. (1999). Episodic memory in semantic dementia: A computational approach based on the tracelink model. In J. Reggia, E. Ruppin & D. Glanzman (Eds.) Disorders of Brain, Behavior and Cognition: The Neurocomputational Perspective. (pp. 47-65). New York: Elsevier.

GRAHAM, K.S., PATTERSON, K., & HODGES, J.R. (1998). Semantic dementia and pure anomia: Two varieties of progressive fluent aphasia. In E.G. Visch-Brink & R. Bastiaane (Eds.), Linguistic Levels in Aphasiology. (pp. 49-68). San Diego: Singular Press, Inc.

HODGES, J.R. (in press). Semantic dementia: A selective disorder of semantic memory? In M. D'Esposito, (Ed.), Cognitive Neuroscience: A Neurological Perspective.

HODGES, J.R., Garrard, P., & PATTERSON, K. (1998). Semantic dementia. In A. Kertesz & D. Munoz (Eds.), Pick's Disease and Pick's Complex (pp. 83-104). New York: Wiley-Liss, Inc.

HODGES, J. R., & GRAHAM, N. L. (2001). Vascular dementias. In J. R. Hodges (Ed.), Early Onset Dementia: A Multidisciplinary Approach. (pp. 319-337). Oxford: Oxford University Press.

Kapur, N., & GRAHAM, K.S. (in press). Recovery of memory function in neurological disease. In A. Baddeley, B. Wilson & M. Kopelman (Eds.), Handbook of Memory Disorders. Chichester: Wiley.

PATTERSON, K. (2000). Phonological alexia: The case of the singing detective. In E. Funnell (Ed.), Case Studies in the Neuropsychology of Reading. (pp.57-83). Hove: Erlbaum.

PATTERSON, K., GRAHAM, N.L., Lambon Ralph, M.A, & HODGES, J.R. (in press). Varieties of silence: The impact of neurodegenerative diseases on language systems in the brain. In J.R. Pomerantz & M. Crair (Eds.), Integrative Neuroscience: From Molecules to Cognition. Cambridge: Cambridge University Press.

PATTERSON, K., & HODGES, J.R. (2000). Semantic dementia: One window on the structure and organisation of semantic memory. In L. Cermak, (Ed.), Revised Handbook of Neuropsychology: Memory and its Disorders (pp. 313-335). Amsterdam: Elsevier Science.

PATTERSON, K., & HODGES, J.R. (2001). Semantic dementia. In J.L. McClelland (Ed.), International Encyclopaedia of the Social and Behavioural Sciences (Behavioural and Cognitive Neuroscience Section). (pp. 3401-3405). New York: Elsevier Science.

ROGERS, T.T., & McClelland, J.L. (in press). A parallel distributed processing approach to semantic cognition: Applications to conceptual development. In G-S. L. Rakison (Ed.), Building Object Categories in Developmental Time: Proceedings of the Carnegie Symposium on Cognition, vol. 32. New Jersey: Lawrence Erlbaum Associates.

Zeman, A. & HODGES, J.R. (2000) Transient global amnesia and transient epileptic amnesia. In G.E. Berrios and J.R. Hodges (Eds.), Memory Disorders in Psychiatric Practice. (pp. 187-203). Cambridge: Cambridge University Press.

Conference Proceedings & Published Abstracts

Davies, R., Xuereb, J.H., & HODGES, J.R. (2002). The human perirhinal cortex in semantic memory: An in vivo and post mortem MRI study in semantic dementia, Alzheimer's disease and matched controls. Neuropathology and Applied Neurobiology, 23.

GRAHAM, K.S., Goldman, W., KROPELNICKI, A., Greene, J., & HODGES, J.R. (2002). Patterns of remote memory in Alzheimer's disease: Implications for models of memory consolidation. Ninth Annual Meeting of the Cognitive Neuroscience Society, 32.

GRAHAM, K.S., KROPELNICKI, A., Nestor, P., & HODGES, J.R. (2001). Semantic dementia: a challenge to the multiple trace model of memory consolidation? Eighth Annual Meeting of the Cognitive Neuroscience Society, 36.

GRAHAM, K.S., LEE, A.C.H., BRETT, M., OWEN, A., & PATTERSON, K. (2001). Taking sides: Left- and right-temporal activations during retrieval of long-term memories. Society for Neuroscience Annual Meeting, 639.10.

KROPELNICKI, A., GRAHAM, K.S., Kapur, N., & HODGES, J.R. (2002). Acquisition of vocabulary in patients with bilateral medial temporal lobe damage: gist but not specifics? Ninth Annual Meeting of the Cognitive Neuroscience Society, 112.

LEE, A.C.H., Robbins, T.W., GRAHAM, K.S., & OWEN, A.M. (2001). Dissociation of semantic and episodic memory using positron emission tomography and a novel homophone task. Society for Neuroscience Annual Meeting, 639.8.

Nestor, P.J., GRAHAM, K.S., & HODGES, J.R. (2001). A reversed temporal gradient for episodic memory is seen in semantic dementia but not "frontal variant" frontotemporal dementia. Brain and Cognition, 47, 296-300.

PATTERSON, K. (2000). How and where does the brain produce words? Evidence from language disorders in patients with brain disease. In Proceedings of the 6th International Conference on Spoken Language Processing (Vol 1, pp. 1-12). Beijing: China Military Friendship Publishers.

PATTERSON, K., LAMBON RALPH, M.A., BIRD, H., HODGES, J.R., & McClelland, J. L. (2000). Normal and impaired processing in quasi-regular domains of language: The case of English past-tense verbs. In Proceedings of the 6th International Conference on Spoken Language Processing. (Vol II, pp 20-25). Beijing: China Military Friendship Publishers.

ROGERS, T.T., & Griffin, R. (2002) Goal attribution without goal representation: A connectionist account of infants' early understanding of intentional actions. Proceedings of the 2002 International Conference on Child Studies, 19.

ROGERS, T.T., Lambon Ralph, M.A., HODGES, J.R., & PATTERSON, K. (2001). When recognition is dominated by regularity: Perceptual and lexical decisions under semantic impairment. Abstracts of the Psychonomic Society, 6, 74-75.

ROGERS, T.T., Lambon Ralph, M.A., HODGES, J.R., & PATTERSON K. (2002). When recognition is dominated by regularity: The effects of semantic impairment on perceptual decisions. Ninth Annual Meeting of the Cognitive Neuroscience Society, 108.

ROGERS, T.T., Lambon Ralph, M.A., Plaut, D.D., HODGES, J.R., & PATTERSON, K. (2002). Domain differences in semantic dementia: Implications for theories of category-specific deficits. Proceedings of the Experimental Psychology Conference, S7.1.

SIMONS, J.S., GRAHAM, K.S., Galton, C.J, Verfaellie, M., PATTERSON, K., & HODGES, J.R. (2001). Recollection-based memory in frontotemporal dementia. Eight Annual Meeting of the Cognitive Neuroscience Society, 34.

SIMONS, J.S., GRAHAM, K.S., PATTERSON, K., & HODGES, J.R. (1999). Semantic knowledge and episodic memory for famous faces in a case of semantic dementia. Proceedings of the British Psychological Society, 7, 45.

SIMONS, J.S., Verfaellie, M., GRAHAM, K.S., Galton, C.J., PATTERSON, K., & HODGES, J.R. (2001). Recollection-based memory in frontotemporal dementia. Eighth Annual Meeting of the Cognitive Neuroscience Society, 34.

Thompson, S.A., GRAHAM, K.S., Williams, G., PATTERSON, K., Kapur, N. & HODGES, J.R. (2002). Dissociating person and general semantic knowledge: Role of the left and right temporal lobes. Ninth Annual Meeting of the Cognitive Neuroscience Society, 111.

Others

GRAHAM, K.S. (1999). Semantic dementia: A challenge to the multiple trace theory of memory consolidation? Trends in Cognitive Sciences, 3, 85-87.

GRAHAM, K.S. & HODGES, J.R. (1999). Episodic memory in semantic dementia: Implications for the roles played by the perirhinal and hippocampal memory systems in new learning. Behavioural and Brain Sciences, 22, 452-453.

HODGES, J.R. (1998) Unraveling the enigma of transient global amnesia. Annals of Neurology, 43, 151-153.

HODGES, J.R. (1998) The amnestic prodrome in Alzheimer's disease. Brain, 121, 1601-1602.

HODGES, J.R. (2000). Man and his memory: The scientific basis of remembering and forgetting. Published talk (13th Sinclair House Debate).

LAMBON-RALPH, M.A., & GRAHAM, N.L. (2000). Mini Revew - Previous cases: Acquired phonological and deep dyslexia. Neurocase, 6, 141-178.

ROGERS, T.T. (submitted). Is there madness in the method? A commentary on Storms and Dirikx. Neuropsychology.

SIMONS, J.S., GRAHAM, K.S., & HODGES, J.R. (1999). What does semantic dementia reveal about the functional role of the perirhinal cortex? Trends in Cognitive Sciences, 3, 248-249.

SIMONS, J.S., & HODGES J.R. (2000). Mini review and previous cases on TGA. Neurocase, 6, 211-230.

TRANSFER TO HEALTH SERVICE

Contributions of the DMS group to the Health Service come in two forms:

(a) Like other MRC-CBU Research Groups, we are continually developing cognitive assessments that are subsequently adopted for use in Memory Disorder and Speech Therapy Clinics. One of the most important of these in recent years is the ACE (Addenbrooke's Cognitive Examination: Mathuranath, Nestor, Berrios, Rakowicz & Hodges, 2000), based on the universally used Mini-Mental State Examination but in an expanded format that enables improved differentiation of various types of dementia.

(b) People who care for sufferers of the more prevalent forms of dementia (especially Alzheimer's Disease) receive substantial support from organisations like the Alzheimer's Disease Society. Those who care for individuals with less common but equally distressing forms of dementing illnesses, like frontotemporal dementia (FTD), have no comparable source of information and support. The DMS research group therefore organises an FTD support group that meets at the CBU approximately 4 times/year. At each meeting, we provide a special speaker, either from within our group or outside, on a topic of importance to carers of these patients; and the carers then have the opportunity to talk both to us and to each other about their concerns. A newsletter is produced and sent to those who attend and all those unable to attend the meeting.

EXTERNAL GRANTS

1. National Institutes of Mental Health (USA) Programme Grant. Toward a model of normal and disordered cognition. 1.3.97 to 28.2.02. Total: £149,275 (KP) (NB this is the CBU portion of a much larger grant for a consortium of researchers led by Prof J L McClelland)

2. Alzheimer's Research Trust Centre Grant. 1.8.98 to 31.7.03. Total: £75, 000. (JRH)

3. MRC Programme Grant. The early diagnosis, differentiation and clinical course of dementia: Alzheimer's, frontotemporal and vascular types. 1.10.98-30.09.03. Total: £845, 000. (JRH)

4. MRC Training Fellowship (for Dr Sian Thompson). The neural basis of autobiographical and semantic memory. 1.4.00 to 31.3.03. Total: £105, 000 (JRH)

5. PSP Association Project Grant. Clinical and neuropsychological studies of PSP. 1.4.00 to 31.3.02. Total: £40, 000. (JRH/TB)

6. Alzheimer's Research Trust Ph.D. studentship (for Ms Fiona Clague). The early diagnosis of AD and related dementias using tests of people naming and cross-modal associative learning. 1.10.01 to 30.9.04. Total: £57, 000. (KG/JRH)

7. Wellcome Research Training Fellowship (for Dr Andrew Graham). Memory in frontotemporal dementia and the role of the ventromedial frontal cortex. 1.12.01 to 30.11.04. Total: £180, 165. (KG/JRH)

8. Wellcome Research Training Fellowship (for Dr Rhys Davies). The neural basis of semantic memory impairment in Alzheimer's disease and Fronto-temporal dementia: A pathological and in vivo study. 1.3.02 to 28.2.05. Total: £170, 000 (JRH)

9. Alzheimer's Research Trust Programme Grant. The functional contribution of medial temporal lobe regions to profiles of memory impairment in dementia: An interdisciplinary approach using neuropsychological and neuroimaging methodologies. 1.10.02 to 30.9.07. Total: £690,000. (KG/JRH)

10. National Institutes of Mental Health (USA) Interdisciplinary Behavioural Science Centre Grant. Toward a neurobiologically constrained framework for modelling human cognition. 1.10.02 to 30.9.07. Total: £158,040 (KP) (NB this is the CBU portion of a much larger grant for a consortium of researchers led by Prof J L McClelland)

11. Sarah Matheson Trust. PSP and multiple system atrophy. 1/8/02-30/04/04. £7,118

PRINCIPAL COLLABORATORS

Dr N Antoun (Addenbrooke's, Radiology)

Dr M Buckley (Oxford)

Dr T Bussey (Cambridge, Experimental Psychology)

Dr D Caine (University of Sydney)

Dr R Epstein (University of Pennsylvania)

Dr P Eslinger (Hershey Medical Centre, USA)

Dr T Fushimi (Tokyo Metropolitan Institute of Gerontology, Japan)

Dr D Gaffan (Oxford)

Dr C Gregory (Cambridge, Psychiatry)

Dr G Halliday (Prince of Wales Medical Research Institute, Sydney)

Dr M Ikeda (Ehime University Medical School, Japan)

Prof N Kapur (Southampton University)

Prof N Kroll (University of California, Davis)

Prof M Lambon Ralph (University of Manchester)

Prof J McClelland (Carnegie Mellon University, USA)

Prof M Macdonald (University of Wisconsin)

Prof B Miller (University of California at San Francisco)

Dr E Murray (NIMH)

Prof D Plaut (Carnegie Mellon University, USA)

Dr C Price (FIL, London)

Dr B Sahakian (Cambridge, Psychiatry)

Prof D Schacter (Harvard University)

Prof M Seidenberg (University of Wisconsin)

Dr J Simons (Institute of Cognitive Neuroscience, UCL)

Prof F Vargha-Khadem (Institute of Child Health, UCL)

Dr M Verfaellie (Boston University)

Dr J Xuereb (Cambridge, Pathology)

Dr A Yonelinas (University of California at Davis)

Other sections in the 1998-2002 report

1. SUMMARY

2. ATTENTION GROUP

3. COGNITION AND EMOTION GROUP

4. LANGUAGE AND COMMUNICATION GROUP

5. MEMORY AND KNOWLEDGE GROUP: DMS SECTION

6. MEMORY AND KNOWLEDGE GROUP: REHABILITATION SECTION

7. METHODS RESEARCH AND INFRASTRUCTURE GROUP