Insights from semantic dementia on the relationship between episodic and semantic memory
Introduction
Over the last 25 years there has been considerable controversy amongst neuroscientists over the cognitive and neural organisation of long-term memory [38], [51], [55], [56], [57], [58]. In one of the earliest, and most influential, models of long-term memory, Tulving proposed the fractionation of memory into two distinct types, episodic and semantic [55], [56]. Episodic memory refers to our repository of personally experienced events, the retrieval of which requires conscious recollection of the specific temporal-spatial setting of an episode from the past. By contrast, semantic memory applies to our store of culturally-shared general knowledge about the world (e.g. the meaning of words, objects, concepts, facts and people). Unlike episodic memory, this type of information does not require recollection of when and where it was initially learnt.
In Tulving’s original conception of long-term memory [56], episodic and semantic memory were considered psychologically and neurologically distinct, a dichotomy reflecting the way in which the human brain is supposed to acquire, process and store information. It was initially thought that patients with amnesia, who show impaired new episodic learning but spared semantic knowledge [35], [40], [61], [64], supported the fractionation of long-term memory, although recent studies have challenged this view [8], [11], [62], [65].
To account for the evidence against a simple dissociation between episodic and semantic memory in amnesia, Tulving [38], [55], [57], [58] revised his model, suggesting that episodic memory is a subsystem of semantic memory and is, therefore, dependent upon the integrity of semantic knowledge (see Fig. 6a). The most recent instantiation of his theory (termed SPI, standing for Serial encoding, Parallel storage, and Independent retrieval) expands on this hierarchical view. In this model, there are four major categories of cognitive memory system: perceptual representation; semantic; working; and episodic memory. There are three crucial premises to the model: (1) information is encoded into systems serially, with encoding in one system dependent upon output from the previous stage; (2) information can be stored in different systems in parallel; and (3) information in different systems can be retrieved independently without any effects on retrieval of information from other systems. This view explains why a patient with a deficit of episodic memory (amnesia) may still be able to retrieve semantic information that was acquired earlier in life [57].
Our study addresses one prediction that has remained constant throughout the revisions to Tulving’s model of long-term memory, i.e. that “… 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) [57]. This prediction stems from the fact that in the SPI model, encoding of material is serial, that is, the registration of information in episodic memory is contingent upon the output from semantic memory, which in turn is dependent upon input from perceptual systems. Although there is already some neuropsychological evidence that calls into question the strictly serial nature of Tulving’s hypothesis [7], [9], [27], the claim that episodic memory is dependent upon semantic memory, and that one should never see patients with normal episodic memory for events and/or stimuli that they fail to comprehend, has never been directly challenged or tested.
In this article, we document precisely this ‘impossible’ side of the double dissociation in new learning in patients with the disorder of semantic dementia [22], [48], and propose an account of the cognitive and neural architecture of long-term memory which reflects the differential contributions of semantic and perceptual inputs to the hippocampal complex.
Tulving’s prediction about the dependence of new learning on semantic memory has never been specifically tested due to the fact that, until recently, there was little evidence (apart from some early reports [45], [63]) that patients could present with a selective impairment to semantic memory. This situation has been rectified by a newly described syndrome termed semantic dementia [22], [48], also called progressive fluent aphasia [32] or the temporal variant of frontotemporal dementia [4], [6], [24], [33], [34], which results in a progressive, relatively selective deterioration of semantic memory. It was Pick [41] (English translation in [10]) who, a century ago, first noted that patients with neurodegenerative disease could exhibit a focal cognitive deficit such as impaired language. Some 90 years after this initial description, Mesulam [32] reported six patients with slowly progressive aphasia, some of whom exhibited fluent and articulate speech which notably contained few content words. This anomic pattern, which is typically present early in semantic dementia, has been shown to reflect a progressive breakdown in the central store of semantic memory affecting both verbal and non-verbal aspects of conceptual knowledge about objects, people, facts, concepts, and the meanings of words.
Patients with semantic dementia perform poorly on any task which requires semantic knowledge, such as picture naming, category fluency (i.e. generating exemplars from semantic categories such as household items), word-picture matching, defining and drawing concepts in relation to their names, selecting an appropriate colour for a black and white line drawing (e.g. red for a tomato) and sorting words or pictures according to a pre-specified criterion (e.g. living or non-living). By contrast, performance on tests measuring other cognitive domains, such as the phonological and syntactic aspects of language, non-verbal problem-solving, working memory and visuo-spatial and perceptual abilities, is relatively unaffected, even at relatively late stages of the disease [19], [22], [23], [48], [49].
Neuroradiological investigations in semantic dementia typically reveal focal atrophy of the antero-lateral aspects of one or both temporal lobes, especially the pole and inferior and middle temporal gyri (Brodmann areas (BA) 38/20) [21], [36], with sparing (at least at early stages of the disease) of the hippocampal complex (hippocampus proper, parahippocampal gyri and subiculum) [13], [16], [18]. In a recent functional neuroimaging study, in which regional cerebral blood flow was measured in four patients with semantic dementia while they performed a semantic decision task, the patients showed a significant reduction in activity in the left posterior inferior temporal gyrus compared to control subjects (BA 37) [36]. The imaging results suggest, therefore, a structural and functional disruption to the temporal lobes in semantic dementia. While there has been relatively little neuropathological data collected in the disease (although see [16], [28], [44], [46], [50]), a meta-analysis of published and unpublished information for 13 cases revealed that all patients had either classic Pick’s disease pathology or non-specific neuronal loss without characteristic Pick or Alzheimer histological markers [17].
The selective nature of the semantic memory impairment in semantic dementia presents cognitive neuropsychologists with a unique opportunity to investigate the cognitive and neural organisation of long-term memory [12], [13], [19], [21], [22], [23], [48], [49]. Recent experiments that have investigated autobiographical and semantic memory in semantic dementia have found evidence for a reverse temporal gradient (i.e. recent memories were better preserved compared to those from the more distant past) [13], [15], [18], [49]. These studies are important because they hint at the possibility that, at least at an early stage in the disorder, new learning may be normal, a finding which would have fundamental implications for our understanding of the nature of episodic and semantic memory.
In our first experiment, therefore, we investigated episodic and semantic memory in patients with semantic dementia and in amnesic patients with presumed early Alzheimer’s disease. In particular, we related the patients’ ability to name a picture of a familiar object during a study phase to their subsequent recognition memory for either the identical item or a perceptually different exemplar of the same object. The essence of our hypothesis is that new learning is normally based on a combination of sensory/perceptual information provided by the learning event and semantic information about the content of the event. If the target from the study phase of a recognition memory experiment is replaced at test by a perceptually different exemplar, this manipulation should reduce the usefulness of sensory/perceptual information and place more demands on input from semantic memory. We predicted, therefore, that patients with semantic dementia would show relatively preserved recognition memory in the perceptually identical (PI) condition but that their ability to select the target item in the perceptually different (PD) condition would be reduced relative to control performance as a result of the patients’ impaired conceptual knowledge of the items. By contrast, the patients with early Alzheimer’s disease should be impaired relative to controls but with equally poor performance on the PI and PD conditions of the episodic task.
Section snippets
Participants
Eight patients with a diagnosis of semantic dementia (mean age 62.6 years, SD=6.2), eight amnesic patients with presumed early Alzheimer’s disease (mean age 73.1 years, SD=6.5) and 18 elderly healthy control participants (mean age 67.4 years, SD=5.7) were tested on two specially designed experimental tasks tapping semantic and episodic memory.
All the patients tested in our study presented through the Memory Clinic at Addenbrooke’s Hospital, Cambridge, UK and have been tested longitudinally on
Materials
JH (aged 59 years) was initially given both a picture naming test and a comprehension test (spoken-word-to-picture matching) on all 260 line drawings of familiar objects from the Snodgrass and Vanderwart picture corpus [47]. On the basis of her success on these two tasks, 76 items were selected such that half (38) could be classified as concepts still ‘known’ by JH (items that she could name and comprehend) while the other half were now ‘unknown’ (i.e. previously familiar objects that she could
General discussion
In our two experiments, patients with semantic dementia showed impaired semantic knowledge yet relatively preserved episodic memory for perceptually identical items. These results refute Tulving’s hypothesis about the hierarchical nature of episodic and semantic memory: it is clearly not the case that the creation of a new episodic memory requires normal functioning of the semantic knowledge system. It was only possible to demonstrate direct effects of impaired semantic knowledge on new
Acknowledgements
We would like to thank all the patients and control participants who took part in this study. We are also very grateful to Gary Jobe for helping with the design of materials, Sasha Bozeat for assistance with test administration, Ian Nimmo-Smith for advice on the statistical analysis, and Endel Tulving for comments on his SPI model.
References (66)
- et al.
A dissociation between perceptual explicit and implicit memory processes
Brain and Cognition
(1997) - et al.
The impaired learning of semantic knowledge following bilateral medial temporal-lobe resection
Brain and Cognition
(1988) - et al.
Long-term retention of computer learning by patients with memory disorders
Neuropsychologia
(1988) - et al.
A reversal of the temporal gradient for famous person knowledge in semantic dementia: implications for the neural organisation of long-term memory
Neuropsychologia
(1998) - et al.
Is semantic memory consistently impaired early in the course of Alzheimer’s disease? Neuroanatomical and diagnostic implications
Neuropsychologia
(1995) Residual learning capability in organic amnesia
Cortex
(1982)- et al.
Dissociations of language function in dementia: a case study
Brain and Language
(1979) - et al.
Retrograde amnesia and memory consolidation: a neurobiological perspective
Current Opinion in Neurobiology
(1995) - et al.
Memory beyond the hippocampus
Current Opinion in Neurobiology
(1997) The primate parahippocampal gyrus: new insights regarding its cortical connections
Trends in Neurosciences
(1982)
Recall of remote episodic memory in amnesia
Neuropsychologia
The entorhinal cortex of the monkey: I. Cytoarchitectonic organization
Journal of Comparative Neurology
Neuropathological staging of Alzheimer-related changes
Acta Neuropathologica
Changing faces: visual and non-visual coding processes in face recognition
British Journal of Psychology
Consensus statement. Clinical and neuropathological criteria for frontotemporal dementia. Lund and Manchester groups
Journal of Neurology, Neurosurgery and Psychiatry
Priming and recognition of transformed three-dimensional objects: effects of size and reflection.
Journal of Experimental Psychology: Learning, Memory, and Cognition
The temporal variant of frontotemporal dementia
Brain
Double dissociation between memory systems underlying explicit and implicit memory in the human brain
Psychological Science
On the relationship between senile cerebral atrophy and aphasia
History of Psychiatry
On the relationship between knowledge and memory for pictures: evidence from the study of patients with semantic dementia and Alzheimer’s disease
Journal of the International Neuropsychological Society
Differentiating the roles of the hippocampal complex and the neocortex in long-term memory storage: evidence from the study of semantic dementia and Alzheimer’s disease
Neuropsychology
Progressive pure anomia: insufficient activation of phonology by meaning
Neurocase
A reverse temporal gradient for public events in a single case of semantic dementia
Neurocase
Quantification of cortical atrophy in a case of progressive fluent aphasia
Brain
Semantic dementia
Charting the progression in semantic dementia: implications for the organisation of semantic memory
Memory
Nonfluent progressive aphasia and semantic dementia: a comparative neuropsychological study
Journal of the International Neuropsychological Society
Semantic dementia: progressive fluent aphasia with temporal lobe atrophy
Brain
Loss of semantic memory: implications for the modularity of mind
Cognitive Neuropsychology
The differentiation of semantic dementia and frontal lobe dementia (temporal and frontal variants of frontotemporal dementia) from early Alzheimer’s disease: a comparative neuropsychological study
Neuropsychology
Routes to object constancy: implications from neurological impairments of object constancy
Quarterly Journal of Experimental Psychology
Double dissociation of memory capacities after bilateral occipital-lobe or medial temporal-lobe lesions
Brain
Cited by (162)
(What) can patients with semantic dementia learn?
2024, NeuropsychologiaA memory system of a robot cognitive architecture and its implementation in ArmarX
2023, Robotics and Autonomous SystemsMissing links: The functional unification of language and memory (L∪M)
2022, Neuroscience and Biobehavioral ReviewsNo Pattern Separation in the Human Hippocampus
2020, Trends in Cognitive SciencesExpertise reversal effect: Cost of generating new schemas
2020, Computers in Human BehaviorConsistently inconsistent: Multimodal episodic deficits in semantic aphasia
2020, Neuropsychologia