Spared numerical abilities in a case of semantic dementia

Abstract

We report a case study of a patient (IH) with a progressive impairment of semantic memory affecting all categories of knowledge apart from numbers. Pictorial material was better understood than words, but was still severely impaired. The selective preservation of nearly all aspects of numerical knowledge suggested that this domain might have different neuropsychological status from other aspects of semantic memory.

Introduction

Progressive selective disorders of semantic memory result from cerebral degeneration mainly affecting the temporal lobes [28], [38], [73], [74], [75], [76], [81]. Patients with semantic dementia characteristically manifest a loss of meaning of words, and increasing difficulty in recognising objects and faces as the disease progresses. Their spontaneous speech, although fluent, is characterised by word finding difficulties, semantic errors, and often lacks content. Formal language tests reveal severe anomia, impaired comprehension of single words, reduced generation of exemplars on category fluency tests, and impoverished general knowledge. Conversely, non-semantic components of language, such as articulation, phonology and syntactic structure are usually well preserved. Patients are well orientated in time and space, and their visual perception and non-verbal problem solving abilities are spared, as are their day-to-day and recent autobiographical memories [77], [56].

Semantic dementia cases characteristically show atrophy of the inferolateral portion of temporal lobe [3], [36], [38], [73]. Recent functional imaging studies in healthy subjects reveal that semantic tasks activate a network of areas in the left temporal hemisphere, including the inferior temporal lobe, angular gyrus, and premotor frontal lobes with emphasis on the critical role of the left inferolateral temporal lobe in the network [46], [47], [57], [81].

The selective impairment of semantic memory was first documented by Warrington [83] who described three patients with progressive anomia and disorders in word comprehension. Detailed neuropsychological investigation revealed impairments in both receptive and expressive vocabulary, a reduced knowledge of a wide range of living and non-living things, with a specific impairment at the subordinate level. Since this seminal paper, a number of similar patients have been described [27], [38], [39], [40], [48], [73], [74], [75], [76], [88].

Although the majority of these patients have been labelled as having semantic dementia [37], [73], [74], [75], [76], other researchers use the term progressive aphasia to describe patients with either relatively spared comprehension [54] or severe deficits in both language comprehension and production [63], [80], [92]. In a recent paper, Garrard and Hodges [31] tried to distinguish semantic dementia from progressive non-fluent aphasia on the basis of clinical and neuropsychological features and discussed these disorders in terms of the organisation of knowledge.

The progressive degeneration of semantic memory results in an intricate pattern of selective impairments and preserved cognitive abilities. The disruption of semantic memory is neither absolute for a specific concept nor homogeneous across concepts. Patients may show fragmented knowledge of a word meaning, such that they understand broad category information, despite they have little information about a concept's specific properties [18], [83]. Furthermore, the patients can initially manifest disproportionate impairment or selective preservation of certain domains of information, such as abstract versus concrete nouns [83], animate versus inanimate objects [4], [48], [69], [71], [88], or the opposite [68], [86], [87]. Although explanations of these category dissociations remain controversial, they help to clarify the nature and organisation of semantic memory. For instance, the finding that the subordinate properties of a concept are more vulnerable than higher order information (e.g. dog→animal→living thing) has been interpreted as evidence for a hierarchical structure of semantic knowledge.

Previous studies have focused predominantly on the semantic components of language and the investigations have seldom included numerical knowledge. The isolation of numerical skills from other cognitive domains has been suggested by several case studies of patients with other clinical disorders. Preserved numerical ability has been reported in a profoundly aphasic patient [67], and in a demented one [66]. The opposite dissociation, namely the selective impairment of numerical skills, has also been documented [5], [33], [62]. These studies did not examine how numerical knowledge interacts with other aspects of semantic memory.

In recent years, several cognitive models have been proposed to describe the functional architecture of numerical system and calculation [9], [12], [20], [24], [49], [60], [61]. McCloskey, Caramazza and Basili's model [49], [50], provided useful framework for assessing number and calculation disorders and has been supported by several single case studies [19], [45], [50], [62], [84]. The model was based on number processing, calculation, and central semantic systems. The number processing components included cognitive mechanisms mediating comprehension and production of both Arabic numbers and number words. The calculation system involved components devoted to comprehension of arithmetical signs, retrieval of arithmetical facts, and the execution of arithmetical procedures. The model postulated that every numerical and calculation processes is accomplished via obligatory activation of central semantic representations, which specify the magnitude of numbers.

Cipolotti and Butterworth [12] proposed a modified version of this model, which incorporated asemantic transcoding processes in addition to the basic McCloskey architecture. The main modification consisted of three asemantic routes implicated in number processing but not involved in semantic representation of numbers.

A third model was proposed by Dehaene and Cohen [20], [21]. These authors suggested a modular number system based on three different codes: an auditory verbal code, a visual number form, and analogue magnitude representation. The first was specifically dedicated to verbal input and output, counting, and to the retrieval of memorised additions and multiplication facts. In this model, addition and multiplication problems were considered as stored verbal information. The visual Arabic number form was involved in the manipulation of Arabic numbers, the analogue magnitude code represented the quantity of a numeral and it was required to compare and to approximate quantities [20]. A further version of this model made explicit predictions regarding the functional and anatomical architecture of numerical processes in the left and right hemispheres [21]. For instance, it was claimed that the visual number form corresponded to a cascade of activation in the mesial occipito-temporal regions of both hemispheres. The analogue magnitude representation was said to be computed by areas of the parieto-occipito-temporal junction in both hemispheres, and the verbal word frame by the perisylvian regions of the left hemisphere.

Numbers bear different meanings in different contexts [24], [29], [30]. For example, Fuson [29], [30] distinguished between numerical meanings (cardinal, ordinal, and measure), sequence meaning, and non-numerical meanings. The cardinal meaning of a number represents its magnitude, also known as numerosity. For instance, in the expression 4×3=12, four disjoint sets each with the numerosity of three, constitute the magnitude of 12. Sequence refers to a series rather than to quantity and considers number terms as an ordered list of words that can be recited as other familiar sequences, such as days of the week, months of the year, or letters of the alphabet [30], [32]. Non-numerical meanings refer to the use of numbers as labels, as in the expression ‘Channel 4’ or ‘Fiat Uno’ [6].

Solving calculation problems requires several different cognitive components. First, arithmetical facts (e.g. 5×2, or 6+3) are stored in long-term memory and are directly retrieved from there [1], [2], [8], [20], [51]. Second, procedural knowledge guides the execution of procedures and algorithms when solving multi-digit operations [51], [82]. Thirdly, conceptual knowledge underlines the understanding of arithmetical operations and principles [34], [35], [78].

Numerical abilities have rarely been investigated in patients with semantic dementia. In the majority of the studies reported, number and calculation skills are not mentioned at all (see the review by Westbury and Bub [92]) or they are just part of general neuropsychological assessments. In a few cases the only measure of numerical skills used is the patients’ performance in the arithmetic subtest of the WAIS [39], [40], [55]. Although the aim of this task is to give an indication of number abilities, other cognitive skills are needed, such as short-term memory, attention, planning and monitoring. Therefore a defective score in the arithmetic subtest of the WAIS may in fact reflect impairments in some of the other skills involved rather than in the patients’ numerical abilities. In other cases different measures have been employed. Among the four cases reported by Hodges et al. [39], one patient (Case 3) was able to perform serial seven subtractions and mental arithmetic quickly and accurately. However, no description was given of the tasks and procedures employed or the results obtained. The authors concluded by saying that “arithmetic was normal in the more mildly impaired cases and only became marginally reduced in the most deteriorated patient” (p. 1791). Hodges et al. [39] described another patient able to do simple calculation, and interpreted this ability as evidence of preserved meaning of numbers. Again no description of the methods or results was reported, and the authors did not specify their definition of the ‘meaning’ of numbers. In other cases, the authors considered general calculation abilities preserved on the basis of a limited range of numerical tasks that the patients were asked to perform (e.g. two-digit addition problems, or to complete numerical series, such 2, 4, 6…) without carrying out further investigations [73].

The only study specifically investigating numerical knowledge in semantic dementia concerned a patient with severe impairment of the semantic aspects of language and spared syntax and phonology [28]. Despite his severe semantic impairment, this patient was able to read and write Arabic numbers, and number names to dictation, at least at the beginning of the investigation. Though he had difficulty in processing arithmetical signs, he was able to perform simple and multi-digit arithmetical operations with error rates well within the range of normal subjects. The only exception was the patient's impaired knowledge of multiplication tables which declined significantly over the course of the investigation. Although this was the first study to report the deterioration of numerical knowledge in semantic dementia, there were several limitations to it. First, the author did not provide a full explanation of the methods used. Some background neuropsychological tests were not reported. For instance, the patient showed preserved reading abilities and severe word finding difficulties, but no indications were given regarding the tests, materials, and procedures used. Secondly, the numerical tests were only vaguely described, again without specifying the type and the number of stimuli used, and the analysis of errors performed. In the follow-up investigation, the author described some numerical abilities as having ‘declined significantly over the course of 18 months’ (p. 240) without providing specific information on the numerical abilities impaired and the level of decline. Thirdly, the author argued that there was a parallelism between calculation and language processing, claiming that the dissociation between intact arithmetical procedures and impaired retrieval of multiplication tables corresponded to that in language between preserved phonology and syntax and impaired retrieval of content words. Nevertheless, the dissociation between arithmetical procedures and tables was detected only in the follow-up study. Multi-digit multiplication problems (that especially require the use of arithmetical procedures) were not investigated at all, and the single-digit ones were not retested. Therefore little information was provided regarding the decline of arithmetical procedures, and it was not clear how the author could draw conclusions on this point.

In the present study we report a detailed investigation of numerical knowledge in a patient with semantic dementia. In line with the existing numerical models [11], [12], [20], [49], we examined both underlying processes necessary for comprehending and producing numbers and for carrying out calculation respectively. In addition to transcoding processing, we also looked at other numerical skills, such as number recognition and encyclopaedic numerical knowledge. This case provides further evidence of the heterogeneity of semantic memory, and examines the relationship between numerical and other domains of semantic knowledge.