Semantic Dementia Research Paper

Semantic dementia (SD), given its current syndrome label by Snowden et al. (1989), was first described around 1900 in some individual case reports by neurologists intrigued by the fact that degenerative diseases could yield such striking and selective cognitive impairments (Girling and Berrios 1997: translation of Pick 1904). The modern history of SD began in the 1970s, with two papers cataloguing the cognitive syndrome in impressive detail (Warrington 1975, Schwartz et al. 1979). Further substantial interest in SD and related conditions has been stimulated in part by the recent dramatic increase in both availability and sophistication of structural brain imaging (particularly Magnetic Resonance Imaging, MRI) and techniques for analyzing and quantifying the brain abnormalities in individual MRI scans.

A specific cognitive impairment combined with a clearly identified locus of neuroanatomical lesion, documented across a largeish number of patients, has traditionally been considered to licence the assignment of that cognitive function to that brain region. On this logic, the sine qua non of semantic memory is the anterior, inferior, lateral region of the temporal lobe, because this is consistently the location of the earliest and most severe atrophy in patients with the syndrome of SD (Mummery et al. 2000). Initially, this atrophy may be measurable just on one side, or at least is often very asymmetrical, but invariably there is bilateral temporal involvement with disease progression. There is also the important caveat, emphasized by recent studies of functional brain imaging, that areas which do not themselves reveal significant structural abnormalities may nonetheless fail to function adequately, presumably because of reduced input from the damaged areas to which they are normally connected. The first activational PET (Positron Emission Tomography) study of SD (Mummery et al. 1999) identified such an area in the posterior left temporal lobe. The consistent locus of atrophy in the anterolateral temporal lobe for these patients only establishes that this region is one essential component of the brain’s network for semantic memory.

2. The Core Syndrome Of Semantic Dementia

2.1 Memory

SD is one of the main presentations of fronto-temporal dementia (see Sect. 5) that together constitute the second commonest cause of dementing illnesses with an onset before age 65. The most prevalent cause of dementia, certainly in later life but even before 65, is Alzheimer’s disease (AD). In typical presentations of AD, semantic memory is eventually also disturbed; the earlier and more prominent cognitive consequence of AD, however, is a severe impairment of the capacity for new learning, or episodic memory. A sharp distinction between episodic and semantic memory, originally proposed by Tulving (1972), is almost certainly over-simplified; but it does seem to capture a salient difference between two types of long-term memory. One’s semantic knowledge of (for example) the Egyptian pyramids, comprising some facts about their appearance and about how, when, where, for what and by whom they were built, seems to stand in sharp contrast to one’s episodic or autobiographical memory of the experience of visiting the pyramids. One typically cannot retrieve much in the way of episodic or contextual details associated with learning the semantic facts about these structures, whereas such details of time, place, sights, sounds, etc., seem the very essence of memory for the autobiographical episode. An appropriate, though approximate, summary is that a patient with SD would remember at least some aspects of a trip to Egypt to see the pyramids, but would be unable to produce almost any conceptual knowledge about them.

There are at least two ways in which the notion of preserved episodic memory in SD should be qualified. First, there is solid experimental evidence that these patients have good ability to encode new memories of experiences based on perceptually rich information. When asked to look at photographs of paintings (Warrington 1975) or of objects like telephones, bicycles and zebras (Graham et al. 2000), followed by a larger set of such photos incorporating the earlier set plus some new ones, SD patients display good recognition memory for the items seen earlier. This is true even if the objects in the photos are ones on which the patients perform poorly in tests of conceptual knowledge. However, suppose that the recognition memory test employs perceptually altered materials: in the experiment with photos of objects, for example, the picture in the test phase would be of the same object type, such as a telephone, as that seen earlier, but represented in the test by a different exemplar of a telephone. In this case, the SD patients’ episodic memory is significantly impaired relative to normal control subjects. This result suggests that stimulus meaning normally supports new learning, especially when the perceptual characteristics of the material to be recognized do not, on their own, readily reinstate the earlier event. The same principle helps to explain SD patients’ poor recognition memory for written words, which constitute rather perceptually impoverished stimuli.

A second qualification on this issue concerns memory age. Probed for information about world events or about significant autobiographical episodes in their own lives, SD patients have revealed good memory for events and people prominent over the last year or so, but degraded and vague knowledge of earlier events (Graham and Hodges 1997). This pattern is consonant with the neuroanatomical facts of SD. Extensive research with humans and other animals suggests that structures in the medial areas of the temporal lobe (particularly the hippocampal complex) are essential for establishing new memories and for keeping them retrievable for some period of weeks or months. This is why amnesia-inducing head injuries, which often affect hippocampal regions, tend to disrupt memory of events just preceding the accident. It also explains why patients with Alzheimer’s disease, to which medial temporal structures are especially vulnerable, are so poor at remembering recent events. The process whereby memories ‘consolidate’ in the brain seems to involve lasting changes in the more lateral structures of the temporal lobe, with decreasing reliance on the hippocampus for memory retrieval. SD is associated with significant atrophy in these lateral regions, but less so in the medial areas, thus fitting the observed relative preservation of very recent auto-biographical memory. This constitutes a qualification on a sharp semantic/episodic distinction because, while semantic memories are typically ‘old,’ and thus likely to be disrupted by the disease process of SD, autobiographical memories may be old or new, and their status in SD according to this account will be determined more by this factor of age than by the content quality of the memory itself.

2.2 Other Cognitive Features

Unlike patients with AD, those with SD are well oriented in time and space and find their way around with ease in familiar or even recently encountered (e.g., hospital) environments. They have good reasoning and problem-solving skills provided that the tasks do not require knowledge of specific concepts. Within the domain of language, they speak in well-formed sentences consisting of properly articulated words. The initial deficit that brings the patient into the clinic is almost always anomia: a difficulty in naming people, places, objects, and concepts. This early word-finding difficulty is typically followed, in both disease progression and prominence, by impaired word comprehension. Deterioration in both the production and comprehension of content-word vocabulary (nouns, verbs, adjectives) is significantly modulated by word frequency or familiarity, with more common words being more resilient. In SD speech, specific content words are replaced by more general and higher-frequency catchall terms: nouns such as ‘thing’ and ‘place,’ verbs such as ‘make’ and ‘do.’ One patient could only refer to each of the significant people in her life (husband, daughter, sister, etc.) as ‘my person.’

These prominent twin problems with producing and understanding the content words of speech might imply that semantic dementia is a deficit restricted to words. On the other hand, in studies where patients with SD have been rigorously assessed on semantic tests involving no verbal stimuli or responses (such as pairing up objects that are standardly found and used together, or photos of famous people that are linked, or demonstrating the conventional use of familiar objects, Hodges et al. 2000), the impairment almost invariably extends well beyond language-based tasks. This suggests that SD is a general conceptual deficit that disrupts knowledge of objects and people as well as the meanings of words. The abilities to produce and comprehend words are probably especially vulnerable even to mild degradation of underlying conceptual knowledge because of the arbitrary relationship be-tween meaning and word form. As discussed further in Sect. 3 and Sect. 4, the robustness of knowledge seems to be governed in part by the extent of consistency or systematicity between representations at different levels of the system. Names of things suffer because they are devoid of systematicity: nothing about the word ‘tiger’ gives a clue to the nature of the object that it labels, or to the similarity of that object to a lion.

3. The Cognitive And Neuroanatomical Progression Of Semantic Dementia

The disruption to conceptual knowledge in SD is eventually profound. A patient with severe SD (Hodges et al. 1992), during a test of yes no questions about category membership and other semantic features of various familiar objects and animals, was asked ‘Is a zebra an animal?’ She replied ‘Is a zebra an animal? I wish I could remember what an animal was.’ At earlier stages of the disease, however, this broad level of object categorization is one of the better-preserved aspects of conceptual knowledge. This is indicated by a large range of cognitive tests, including semantic judgments about pictures or names of objects at different levels of specificity. When patients are studied longitudinally, the conceptual deficit emerges earliest for specific details such as the stripes on a zebra or where it lives, while success in classifying a zebra as an animal is maintained longest. The same pattern is revealed by the patients’ responses in picture naming tasks, where they fail to name the particular exemplar but often provide either the general category name ‘animal’ or else a very high-frequency prototypical animal name such as ‘dog’ or ‘horse’ (Hodges et al. 1995). In another revealing anecdote, a patient responded to a picture of a zebra by saying ‘It’s a horse, isn’t it?’ Then, as she pointed to the stripes, she queried ‘What are these funny things for?’ An influential interpretation of this pattern of conceptual deterio-ration in SD, with important implications for the way in which such information is structured in the normal system, is the notion of disruption to elements in a similarity-based sub-symbolic connectionist network (McClelland et al. 1995). Features that are characteristic of many similar concepts (such as four legs on an animal) are more securely supported by connection weights across the network and are therefore more robust under damage than idiosyncratic features like the stripes of a zebra.

Another notable longitudinal feature of SD conerns the severity of the patients’ anomia in relation to the specific pattern of neuroanatomical progression. As noted earlier, although the inferolateral temporal atrophy characteristic of SD seems invariably to be bilateral beyond the earliest stages of the disease, it is often highly asymmetrical. The fact that semantic memory is disrupted by a predominance of either left or right damage is most consonant with the assumption that the semantic network is distributed bilaterally across both temporal lobes. Recent research demonstrates, however, that for approximately equated degrees of semantic deficit, the anomia is more severe in patients with more prominent left temporal atrophy than for those with greater right-sided abnormality (Lambon Ralph et al. 2001). The left-hemisphere elements of the distributed semantic network may be more crucial for activating the phonological system for speech production, which is typically strongly leftlateralised in the human brain.

4. Collateral Symptoms

Knowledge of word meanings is clearly central to the abilities to produce and comprehend informative connected speech, and also to the ability to name objects and people; these aspects of language competence therefore progressively deteriorate in patients with SD. By contrast, there are a number of more constrained language skills, involving the transformation of a word from one modality or form to another, that seem largely independent of word meaning. Two examples of a modality transformation are reading a word aloud (a transformation from orthography to phonology) or writing a word to dictation (phonology → orthography); and an example of a simple change in word form is generating the past tense of a verb from its present-tense form. If word meaning is not significantly implicated in these language skills, then patients with SD should perform well on these tasks. Do they? The answer is yes, and no. Their level of success depends on two aspects of the stimulus word to be transformed: whether it is a common or uncommon word, and whether the relationship between the stimulus and response form is consistent or inconsistent with that for other similar words in the speaker’s vocabulary. In the English language and in the domain of reading aloud, an example of an inconsistent or irregular word is _PINT, because the relationship between its spelling pattern and its pronunciation deviates from other similar and more typical words like _PRINT, _MINT and _HINT. Likewise, in English and in the domain of past-tense verb morphology, an atypical word is _THINK; unlike most verbs, and even unlike the phonologically similar verb _BLINK the past-tense of _THINK is not _THINKED but _THOUGHT.

For reasonably high-frequency words with a typical relationship between stimulus and response, patients with SD achieve excellent reading aloud, spelling to dictation and transformation of a verb from its present to past tense. Relative to normal age-and education-matched subjects, the patients show a significant but not too severe decrement for words that are either high frequency but irregular, or regular but lower frequency. On low-frequency irregular words, however, their average success is a dramatic 30–50 percent below normal on the three tasks (Patterson 2000). The patients’ errors to these difficult items reflect the well-learned generalizations: the overwhelmingly typical error is to pronounce a written word like ‘pint’ as if it rhymed with ‘mint;’ to spell a spoken word like ‘cough’ as ‘coff;’ and to turn the spoken present-tense verb ‘grind’ (in a sentence context like ‘Today I grind’ the coffee; yesterday I ‘—— the coffee’) into grinded. To the extent that parallel experiments have been per-formed in languages other than English, patients with SD have revealed precisely comparable phenomena. For example, the generalizations and inconsistencies of the relationship between orthography and phonology in the non-alphabetic writing systems of Japanese are quite different from those in alphabetic English; but Japanese written words do have varying degrees of consistency in this relationship, and Japanese patients with SD succeed in reading aloud the high-frequency consistent words, and fail on the low-frequency atypical words (Patterson et al. 1995).

How is this pattern of performance to be under-stood? One view of word processing, based on net-works of parallel distributed processing, includes the premise that semantic representations of words interact with phonological and orthographic representations in all forms of lexical processing (Plaut et al. 1996). It further predicts that the interaction with semantic knowledge should have a somewhat more influential role precisely for this set of words that the patients find difficult. The prediction regarding word frequency stems from the fact that, in the networks responsible for these tasks, the strength of connections is heavily influenced by the frequency with which particular patterns are encountered and processed. The prediction regarding typicality of relationship between stimulus and response stems from the fact that the very same distributed network that is sensitive to individual word knowledge and frequency also captures similarities and generalizations about stimulus-response relationships across words in the vocabulary. In this framework, if a word benefits from neither strong connections in its own right (by virtue of being commonly encountered), nor from the strong shared connections that apply to words with typical mappings, its processing comes to depend more on additional activation from other parts of the language system, particularly word meaning. According to this theory, then, the SD patients’ pattern of successes and failures in these relatively ‘non-semantic’ language tasks can be explained in terms of the reduced activation available from semantic representations.

5. Semantic Dementia In The Context Of Other Related Syndromes

The book by Snowden et al. (1996) entitled Fronto-Temporal Lobar Degeneration distinguished three major syndromes that result from neurodegenerative diseases selectively affecting the frontal and/or temporal lobes of the human brain. How does SD relate to the other two syndromes, behaviorally and neuroanatomically?

Progressive aphasia, first described in the modern era by Mesulam (1982), is a disorder restricted to the domain of language, at least until late in disease progression. SD itself has sometimes, perhaps confusingly, been called progressive aphasia, because of its profound impact on both language production and comprehension; but this alternative syndrome of progressive aphasia is very different. SD gradually obliterates content word vocabulary, leaving fluent but anomic speech with good phonology and syntax. By contrast, the hallmark of nonfluent progressive aphasia, as it is more informatively called, is disrupted phonological and syntactic processing (Hodges and Patterson 1996). Progressive aphasic patients of this type lose the ability to produce connected sentences; their speech is increasingly restricted to short phrases or even to single content words, and these often contain phonological distortions. For example, one patient named a picture of an elephant as ‘feleftent:’ recognizable if the referent is known but obviously very distorted from the correct phonological word form. Neuroanatomically, nonfluent progressive aphasia seems to be associated with a less focal pattern of atrophy than SD. Both in i o structural imaging and post-mortem analyses reveal atrophy in the left hemisphere, including both superior temporal and posterior frontal regions around the sylvian fissure.

The third member of the triumvirate of conditions described by Snowden et al. (1996) is frontotemporal dementia, often called just frontal (or frontal lobe) dementia. Although an adequate understanding of frontal lobe function in human behavior and cognition is yet to be achieved, it seems clear that much of what can be called human problem solving, judgment, decision making, emotional and social appropriate-ness, etc., crucially depends on intact frontal lobes; and these are precisely the capacities that are eroded in frontal dementia. Although profound semantic disruption is characteristic of SD and not of frontal dementia, SD often does, especially with disease progression, begin to reveal some of the behavioral disturbances that are the hallmark of frontal dementia. This is perhaps not surprising in light of the fact that, after initial consistent abnormalities of the anterior temporal lobe, atrophy in SD typically spreads not only posteriorly within the temporal lobe but also anteriorly to inferior frontal regions (Mummery et al. 2000). The behavioral changes include a strong preference for fixed daily routine, often to the point where it virtually becomes a demand for ritual.

6. The Neuropathology Of Semantic Dementia

The underlying pathological cause of many neurodegenerative conditions can only be identified by analysis of brain tissue, which typically happens (if at all) only after the patient’s death. Hodges et al. (1998) summarized the findings for all 13 cases to that date which met two criteria: (a) the patient had been studied in sufficient detail in life to enable identification of a clear pattern of SD; (b) brain tissue was available for post-mortem analysis. In this set, there were no cases with the neurofibrillary tangles and neuritic plaques that constitute the pathological features of Alzheimer’s disease. All 13 cases had either the specific intraneuronal inclusions that typify Pick’s disease (Hodges 1994) or nonspecific degeneration neuronal loss. It is worth mentioning, however, that Alzheimer’s pathology has recently been identified post-mortem in a few patients who had profound semantic disruption, even if not the most prototypical presentation of SD (Galton et al. 2000). In these cases, the brunt of the AD pathology was found in the left anterior temporal lobe—atypical for AD but exactly where atrophy is centered in semantic dementia. It seems clear that cognitive deficits are determined more by the neuroanatomical locus of pathology than by its precise nature.

Bibliography:

1. Fodor J A 1983 The Modularity of Mind. MIT Press, Cambridge, MA
2. Galton C J, Patterson K, Xuereb J, Hodges J R 2000 Atypical and typical presentations of Alzheimer’s disease: A clinical, neuropsychological, neuroimaging and pathological study of 13 cases. Brain 123: 484–98
3. Girling D M, Berrios G E 1997 On the symptomatology of left-sided temporal lobe atrophy (translation of Pick A 1904 Zur Symptomatologie der linksseitigen Schafenlappenatrophie. Monatschrift fur Psychiatrie und Neurologie 16: 378–88). History of Psychiatry 8: 149–59
4. Graham K S, Hodges J R 1997 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 11: 77–89
5. 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: 313–24
6. Hodges J R 1994 Pick’s disease. In: Burns A, Levy R (eds.) Dementia. Chapman and Hall, London, pp. 739–53
7. Hodges J R, Garrard P, Patterson K 1998 Semantic dementia. In: Kertesz A, Munoz D G (eds.) Pick’s Disease and Pick Complex. Wiley-Liss, New York, pp. 83–104
8. Hodges J R, Graham N, Patterson K 1995 Charting the progression in semantic dementia: Implications for the organization of semantic memory. Memory 3: 463–95
9. Hodges J R, Patterson K 1996 Nonfluent progressive aphasia and semantic dementia: A comparative neuropsychological study. Journal of the International Neuropsychological Society 2: 511–24
10. Hodges J R, Patterson K, Oxbury S, Funnell E 1992 Semantic dementia: Progressive fluent aphasia with temporal lobe atrophy. Brain 115: 1783–806
11. 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: 1913–25
12. Lambon M A, McClelland J L, Patterson K, Galton C J, 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: 341–56
13. McClelland J L, McNaughton B, O’Reilly R 1995 Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychological Review 102: 419–43
14. Mesulam M M 1982 Slowly progressive aphasia without generalized dementia. Annals of Neurology 11: 592–8
15. 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
16. 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: Relationship between temporal lobe atrophy and semantic memory. Annals of Neurology 47: 36–45
17. Patterson K 2000 The impact of semantic degradation on ‘non-semantic’ language abilities. Higher Brain Function Research 20: 85–98
18. Patterson K, Suzuki T, Wydell T N, Sasanuma S 1995 Progressive aphasia and surface alexia in Japanese. Neurocase 1: 155–65
19. Plaut D C, McClelland J L, Seidenberg M S, Patterson K 1996 Understanding normal and impaired word reading: Computational principles in quasi-regular domains. Psychological Review 103: 56–115
20. Schwartz M F, Marin O S M, Saffran E M 1979 Dissociations of language function in dementia: A case study. Brain and Language 7: 277–306
21. Snowden J S, Goulding P J, Neary D 1989 Semantic dementia: A form of circumscribed cerebral atrophy. Behavioral Neuroscience 2: 167–82
22. Snowden J S, Neary D, Mann D M A 1996 FrontoTemporal Lobar Degeneration. Churchill Livingstone, New York
23. Tulving E 1972 Episodic and semantic memory. In: Tulving E, Donaldson W (eds.) Organization of Memory. Academic Press, New York
24. Warrington E K 1975 Selective impairment of semantic memory. Quarterly Journal of Experimental Psychology 27: 635–57