اثر رده معنایی با بدتر شدن نام پریشی در زوال عقل نوع آلزایمر افزایش می یابد

A large number of patients (n=72) with probable Alzheimer’s type dementia (DAT) and mild cognitive impairment (MCI) carried out a picture naming task which comprised stimuli from biological and nonbiological categories. The results were stratified into five ranges of overall naming ability. Every group except those with scores within the range of elderly normal individuals demonstrated better nonbiological naming than biological naming, an effect which increased with worsening impairment. In general, patients diagnosed with other dementia (n=15) did not fit well within the pattern of the DAT/MCI participants, except those known to have a significant semantic impairment. A category effect favoring nonbiological items appears to be robust and produce a predictable pattern across progressive levels of impairment in AD.

Semantic memory deficits are a major feature in many cases of dementia of the Alzheimer’s type (DAT). Several studies have undertaken to determine whether the deterioration of semantic memory in DAT has an equivalent effect across semantic categories. The results of these studies have been used to support theories concerning the representation of concepts in the brain. Contradictory results, however, have been reported. These variable results can be attributed both to differences in the basis on which stimuli were selected and to heterogeneity within the DAT population. We carried out the present study in order to clarify claims regarding the existence of category-specific naming deficits in DAT. In it we tested a large group of elderly individuals on a picture naming task with a unique set of balanced stimuli. By investigating the performance of individuals at successive ranges of anomia (related to degenerative dementia), we seek to resolve important theoretical issues regarding semantic memory deterioration. The majority of studies which have found a category effect in DAT, have found it to be in the direction of poorer accuracy on items from the biological category (e.g., animals, fruits, vegetables, flowers) than on items from the nonbiological category (e.g., clothing, vehicles, household objects). This effect is rarely as striking as the category specific deficits displayed by some acute lesion patients who can demonstrate nearly perfect performance on some categories and less than 50% accuracy on other categories. Patients with AD perform poorly overall on semantic tasks and so do not have “preserved” categories and the difference in accuracy between categories rarely attains more than 20%. An effect attributable to category, nevertheless, has been found on a variety of semantic tasks ( Daum, Riesch, Sartori, & Birbaumer, 1996; Garrard, Patterson, Watson, & Hodges, 1998; Laiacona, Barbarotto, & Capitani, 1998; Mauri, Daum, Sartori, & Riesch, 1994; Silveri, Daniele, Giustolisi, & Gainotti, 1991, in press, Fung et al., 2001). Silveri et al. (1991) tested 15 DATs on a confrontation naming task and on a verbal association task in which participants were asked whether an orally presented word was related to the picture presented. In both tasks the DATs demonstrated poorer performance on the items from the living category than from the nonliving category. Daum et al. (1996) found a similar category effect in 8 DATs on tasks of picture naming, probe questioning, and object decision. Mauri et al. (1994) compared the performance of one DAT subject, Helga, on a variety of tasks with the performance of a herpes encephalitis patient, Michelangelo, who also had a category specific deficit. They found that Helga’s performance profile was very similar to that of Michelangelo, and that the category deficit extended across domains (visual and verbal), and input and output modalities. Fung et al. (2001), tested 16 DATs on a semantic association task in which participants were asked to indicate by a key response which of two words was more closely related to a target word (e.g., lamb: sheep or goat?). The categories tested were animals, fruits and vegetables, clothing and furniture, tools, action verbs, and abstract nouns. Normal elderly participants demonstrated equivalent performance across categories, but were not at ceiling. The DATs performance, though impaired, was significantly better on the nonbiological and action word categories (accuracy of approximately 79%) than on the biological and the abstract word categories (accuracy of approximately 67%). Fung et al. (2001) also found that the same DATs demonstrated a significant category effect on a task of picture naming. The claim that DATs demonstrate better performance on nonbiological objects has not gone undisputed. Arguments raised against the claim are threefold. The first objection is that in many semantic tasks no category effect can be demonstrated. These tasks include category fluency (e.g., name as many animals as possible in a minute), member dominance (judgment of best category membership), and ranking tasks (Cronin-Golomb, Keane, Kokodis, Corkin, & Growdon, 1992; Hodges, Salmon, & Butters, 1992). Montanes, Goldblum, and Boller (1996) found, in fact, that DATs were better on sorting biological objects into categories (animals vs birds) than on sorting nonbiological objects (tools vs vehicles). While a true category-specific semantic deficit should be reflected in a pattern of deficits across a range of tasks (Chertkow & Bub, 1990), each of these tasks entails its own nonsemantic elements (for instance, the search procedure involved in semantic fluency task) which might obscure the 20% semantic category effect in DAT. Alternatively, it may be that the category effect is only evident in tasks where object-specific information, is required. The second objection to the claims of category effects in DAT, is that, when effects are found, they are due to factors unrelated to the semantic category distinction itself. Tippett, Grossman, and Farah (1996) argued that Silveri et al.’s (1991) results on the naming and verbal association tasks could be attributed to a poor matching of stimuli between categories on measures of familiarity and frequency. Tippett et al. (1996) showed that with a single group of DATs they could demonstrate a category effect on the same stimulus list used by Silveri et al., but no effect on two other stimulus lists which had been more stringently matched for familiarity. In a similar vein, Montanes, Goldblum, and Boller (1995) found that DATs who demonstrated a category effect on naming black and white pictures did not show the same effect on pictures of colored “living” and “nonliving” objects. They also show that DATs show no category effect in picture naming when categories are matched for visual complexity. It is noteworthy, however, that in the Tippett et al. study DATs were significantly more accurate on the second and third lists indicating that the stimuli were probably more familiar than in list one. We note that in a few studies where the stimuli were such that the control participants could perform the task almost perfectly (Garrard et al., 1998; Hodges et al., 1992; Montanes et al., 1995), the DAT group did not demonstrate a category effect. The third type of objection to category effects questions either the universality of the category effect in DAT or the consistency of its direction. For instance, some studies have only found the effect within subgroups of patients and still others have found that multiple single case analyses sometimes reveal DAT participants who are more accurate for biological than nonbiological items. Montanes et al. (1995) failed to find a category effect in a picture naming task in a group of 25 participants. They found, however, that the test group was heterogeneous and that there was a subgroup of 4 participants who were disproportionately worse on verbal tasks than on visuo-constructional tasks. They noted that the effect of category was more likely to be brought out in the verbal deficit group. Laiacona et al. (1998) tested 26 DATs on a task of picture naming. A group analysis of their accuracy revealed no category effect, but single case analyses revealed that 11 participants did display better performance on one category than the other. Of the 11 participants, 8 showed the typical dissociation (better nonliving than living) and 3 displayed a dissociation in the opposite direction. Gonnerman, Andersen, Devlin, Kempler, and Seidenberg (1997) were interested in characterizing the progression of concept degradation in DAT. They present the longitudinal data of two DAT participants on a picture naming task carried out at six month intervals. It is shown that while one subject displayed consistently poorer performance on biological items, the other progressed from equally poor performance on both categories to considerably worse performance on artifacts, followed six months later by a large drop in performance on biological items. Gonnerman also showed that in two picture naming experiments, DAT participants failed as a group to demonstrate a category effect. They maintained, however, that in both experiments there was a pattern of better biological naming at mild stages of disease which crossed over to a nonbiological advantage at more severe levels of impairment. Their division of the patients into mild and severe stages was based on the patients’ accuracy in naming biological objects and in both experiments the mild DAT groups were small (n<6) and insufficient to statistically support their hypothesis of a crossover in effect. Nevertheless, these studies are intriguing and the claim of heterogeneity deserves to be carefully tested on larger sets of participants at varying degrees of impairment. Garrard et al. (1998) tested Gonnerman et al.’s (1997) proposed “early biological advantage” on a large group of DAT participants (n=58). They carried out both an items analysis and a multiple single case analysis of the participants’ aggregate naming scores. While the items analysis failed to reveal a category effect, the single case analyses showed that both the high scoring and the low scoring group (determined by median split) were more accurate on the nonbiological items than on the biological items. They similarly found that there was no interaction if the participants were median split by their MMSE score (mean of 19.9, range of 2–30). In analyses carried out using a “semantic index” which included the results from tasks of picture naming, word–picture matching and naming to description, they found that the category effect was in the same direction for both MMSE groups, although the effect size was larger in the low MMSE group. The Garrard et al. (1998) study represents an important challenge to the Gonnerman theory. Both the Garrard and the Gonnerman studies are limited, however, when it comes to providing a picture of semantic deterioration at different ranges of impairment. Whereas in the Gonnerman study, the test population was divided by visual inspection, in the Garrard study, high and low scorers are determined by median split. The results of median splits are dependent on the overall frequency distribution of the test population. In a less impaired population, Garrard et al.’s results might not be replicated. The theoretical arguments which have been erected on the basis of the pattern of semantic dissolution in DAT are not trivial. On the one hand are arguments that category effects support the notion that different semantic categories (or the features which represent them) are localized in different focal brain areas (Daum et al., 1996; Garrard et al., 1998; Laiacona et al., 1998; Silveri et al., 1991). On the other hand are arguments that category effects indicate that all concepts are represented within the same distributed network but are distinguishable by their different patterns of connections (Devlin, Gonnerman, Andersen, & Seidenberg, 1998; Gonnerman et al., 1997). Clearly, however, before one can examine the theories, more empirical information is necessary. Here our goal is to elucidate the conflicting results with regards to the presence and direction of category effects in Alzheimer’s disease and to better understand the effect of categories within semantic memory. In order to do this we focussed on the naming abilities of DAT patients and stratified them on this basis. Our reason for stratifying participants by overall naming impairment and not by a more general measure of dementia severity such as MMSE is as follows. While semantic memory impairment is an inevitable consequence of DAT, the rate of semantic loss is variable across patients and any two patients who are judged to be at equivalent levels of disease progression according to dementia rating scales can demonstrate significant differences in their degree of semantic memory loss. Our primary goal being to understand the nature of category effects within a progressively deteriorating semantic system we chose to examine the category effect in patients grouped by performance on the naming task. In order to determine how specific the effect was to semantic impairment and DAT, we also carried out analyses which examined (1) the correlation between naming scores and dementia scores for DAT patients and (2) whether the category effect was present in patients diagnosed with other dementia. The choice of a naming task to investigate semantic memory requires a word of explanation. Performance on a naming task is not a reliable measure of semantic memory in all patient groups, particularly in aphasic patients where the difficulty is primarily one of phonological output. In DAT, however, there are many reasons to conclude that naming accuracy is indicative of the level of semantic impairment. First, phonemic cueing is only minimally helpful to participants with DAT suggesting that their problems are more profound than difficulty accessing an otherwise intact representation (Daum et al., 1996; Mauri et al., 1994). Second, there is a good correspondence between the items which individual participants name and cannot name, and performance on those same items using other semantic tasks which require a pointing or “yes–no” response (Chertkow & Bub, 1990; Hodges et al., 1992). Third, it has been found that to a large measure participants fail or succeed on the same items in test-retest (Chertkow & Bub, 1990). Finally, semantic cueing (e.g., picture of a ‘lion’—cue: ‘It is like a tiger’) is generally ineffective in naming tasks in DAT although it can help word retrieval for certain objects for which the subject demonstrates some knowledge through probe questioning (Chertkow & Bub, 1990). Picture naming accuracy is both the measure which was the most frequently used across studies to gauge the category effect in semantic impairment but also, as we outlined above, the most frequently disputed. In this study we used the Categorical Picture Naming Task (CPNT) by Chertkow, Murtha, Frederickson, and Whitehead (1999). It was developed to overcome some of the important weaknesses of other naming tasks in testing the effect of category (biological vs nonbiological). First of all, the items were selected so that elderly (normal) performance on the items was considerably less than perfect (83%). In this way it could be confirmed that there was no category bias hidden in a ceiling effect. Second, the CPNT was designed to stringently match items on variables that could be clearly (and appropriately) quantified. For example, items from the biological and nonbiological categories were matched pairwise for accuracy and familiarity based on norms collected from elderly normal individuals rather than from a young population. The quantifiable measures for what has been termed “visual complexity” are harder to come by. Snodgrass and Vanderwort (1980) asked participants to make visual complexity ratings of their set of pictures according to the “amount of detail or intricacy of line in the picture.” They found these ratings to be correlated to other measures such as familiarity, image agreement (i.e., how close a picture corresponded to one’s mental image of a concept) and name agreement. They discuss intelligently the possible reasons for these correlations but for the purposes of our investigation it is clear that many factors other than the number of lines in a picture are included in visual complexity ratings. It is possible that equating categories on the basis of visual complexity might lead away from the source of category effects, or result in attributing category effects to perceptual factors when cognitive explanations at a deeper level are merited.

The correlation between scores on the Boston Naming Test and the CPNT for the DAT & MCI participants was r=.83. The correlation between scores on the Category Picture Naming Test and MMSE for the same participants was r=.51.