A category-specific naming impairment after temporal lobectomy

~

Pergamon

0028 3932(95)00098~:~

Neuropsychologia, Vol. 34, No. 2, pp. 139 146, 1996 Copyright ~t~: 1996 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0028 3932/96 $15.00+0.00

A category-specific naming impairment after temporal lobectomy LYNETTE

J. T I P P E T T , * t

GUILA

GLOSSER~

and MARTHA

J. F A R A H t

tDepartment of Psychology, University of Pennsylvania, Philadelphia, PA, U.S.A.; and :~Department of Neurology, Graduate Hospital, 1 Graduate Plaza, Philadelphia, PA 19146, U.S.A. (Received 13 December 1994; accepted 3 June 1995)

Abstract--Unilateral temporal lobectomy patients and normal control subjects were tested in a speeded naming task with pictures of living and nonliving things that were equated for name frequency, familiarity, and visual complexity. Although right temporal lobectomy patients and normal subjects performed equally well with the living relative to nonliving things, left temporal lobectomy patients were disproportionately impaired at naming nonliving things. This result has several implications: First, it supports the existence of category-specific naming impairments. In particular, it undermines the proposal that living-nonliving dissociations are artifactual, resulting from the greater difficulty of living things. Second, it demonstrates an asymmetry in the neural representation of nonliving things, in favor of the left hemisphere. Third, it casts doubt on the hypothesis that the anterior temporal cortices are convergence zones that are particularly necessary for the naming of living things. Key Words: category-specific naming; temporal lobectomy.

as familiarity, word frequency, and visual complexity of the stimuli. By this account, sensitivity to these factors alone underlies the poor performance of some patients with living things relative to nonliving things. Proponents of this view have shown that patients who initially appeared to have a knowledge deficit for living things performed equally well with living and nonliving things when the aforementioned factors were carefully controlled [8, 24]. The importance of controlling or accounting for item difficulty when comparing patient performance on living and nonliving things cannot be denied. However, even when a variety of factors affecting difficulty were accounted for, two subjects performed significantly worse on living things, relative to nonliving things [6]. The studies in which naming difficulty for living and nonliving things were equated (as opposed to accounted for) were forced to use such a small stimulus set that they did not have sufficient statistical power to detect true differences in their subjects' abilities. In support of this possibility, Farah et al. [7] first replicated the null results of Funnell and Sheridan [8] with their two subjects, and then obtained significant differences between performance with living and nonliving items by simply quintupling the number of observations. A different approach to the question of whether category-specific knowledge impairments are artifacts of

Introduction

In recent years the categorical loss of semantic knowledge has received considerable attention, stimulated initially by the reports of Warrington and her colleagues [25-27]. These authors observed that different types of semantic knowledge could be dissociated after brain damage, along lines roughly corresponding to knowledge of the categories "living things" and "nonliving things". These observations were subsequently replicated with other patients, and reports of categoryspecific knowledge impairments, particularly for living things, are now numerous in the literature [e.g. 1, 6, 17, 18, 20, 21]. Despite the frequency of these reports, not all psychologists have been convinced that category-specific knowledge impairments exist. An alternative account is that the apparent categorical aspects of these data are simply artifacts of poorly controlled stimulus sets [e.g. 8, 9, 24]. These authors note that living items are often harder than nonliving items on dimensions important for the production and comprehension of language, such

*Address for correspondence: Department of Psychology University of Auckland, Private Bag 92019, Auckland, New Zealand. 139

140

L.J. Tippett et al./A category-specific naming impairment after temporal lobectomy

differential difficulty or reflect real underlying differences between brain systems is to seek a double dissociation between knowledge of living and nonliving things. If there are also subjects who perform worse with nonliving things than with living things, this provides strong support for the hypothesis that semantic memory has dissociable components, rather than being an undifferentiated system which when damaged, fails harder items first. In contrast to the mounting number of reports of knowledge deficits for living things, only a small number of subjects with impaired knowledge of nonliving things have been reported. Warrington and McCarthy [25, 26] describe two cases (V.E.R. and Y.O.T.), both of whom had left hemisphere lesions (V.E.R. had an infarction of the left frontoparietal area, Y.O.T. an infarction in the left temporoparietal region). They were both described as globally aphasic and because of these language difficulties were tested primarily with matching-to-sample techniques. V.E.R. and Y.O.T. were relatively unimpaired in their knowledge of animals, flowers, and foods, but were impaired on objects. Both of these patients were also sensitive to the rate of presentation of the stimuli, such that with slower presentations much of the impairment went away. Hillis and Caramazza [13] described case J.J., who had suffered infarctions in the left temporal lobe and the left basal ganglia. At the time of testing J.J. had persisting impairments of the production and comprehension of language, although he had made significant improvements from earlier levels of performance. On a variety of tasks (oral and written naming, spoken and printed word/picture verification), J.J. showed a significant advantage for animals, compared with any other nonanimal categories, which in addition to body parts, clothing, furniture, and transportation also included vegetables, fruit, and foods. The selective sparing of animals was also evident when defining spoken words. Finally, Sacchett and Humphreys [19] described case C.W., who had a large frontoparietal infarct, nonfluent speech, and mild comprehension difficulties. On tests of naming and picture-word matching, C.W. was impaired in the categories of common artifacts and body parts. In contrast he showed little or no difficulty with animals, birds, insects, fruits, and vegetables. In this paper we add to this relatively small literature on category-specific impairments for nonliving things. We do so not by describing a new case, but rather by describing a population of subjects with such an impairment. In overview, we tested groups of rightand left-temporal lobectomy patients and control subjects in a speeded picture-naming task. The picture sets to be named were those selected by Funnell and Sheridan [8], in which the familiarity, name frequency, and visual complexity of living and nonliving items were matched. Our results have implications for the localization and lateralization of knowledge of nonliving things,

as well as generality and replicability of this form of knowledge deficit. Temporal lobectomy does not generally result in persistent anomia when testing is without time pressure [e.g. 11, 12]. However, subgroups of individuals without early risk factors for epilepsy may have such deficits [e.g. 15, 23]. There is also some evidence of impaired semantic processing in this subject population. Lansdell [14] found that left-temporal lobectomy patients were relatively impaired at selecting the appropriate word to complete a definition when given five possible words. As subjects were not required to generate the appropriate name, the poor performance of the left-temporal lobectomy group may refect an impairment of semantic memory. More direct evidence comes from Wilkins and Moscovitch [28], who tested left- and right-temporal lobectomy patients and control subjects on speeded naming and semantic classification tasks. The lefttemporal lobectomy group was significantly worse at both of these tasks, although they were unimpaired when asked whether the same stimuli were bigger or smaller than a chair. This suggests that speeded classifications per se were not impaired, and that the underlying deficit was one of semantic knowledge. However, none of these studies addressed the question of whether particular categories of knowledge were differentially affected. Our initial interest in the temporal lobectomy population was motivated by the hypothesis of Damasio and co-workers [e.g. 2-4] that the anterior temporal lobe is a "convergence zone" for neural representations, and is particularly important for knowledge of living things. Thus, we expected that if any category-specific impairment was found in this group, it would be an impairment for living, rather than nonliving, things. In the Discussion we will elaborate on the implications of our results for the "convergence zone" hypothesis.

Methods Subjects Experimental subjects. Subjects with temporal lobectomy were obtained from the Graduate Hospital Comprehensive Epilepsy Center. Thirty-one subjects participated, of whom 17 had undergone left temporal lobectomy and 14 had undergone right temporal lobectomy for the relief of intractable epilepsy. All subjects except one were right-handed, and all (including the left-handed subject) were left-hemisphere dominant for speech and language as demonstrated by intracarotid amobarbital test performed prior to surgery. The two groups did not differ in terms of years of education, presurgical IQ, presence of early risk factors for epilepsy, or time since surgery. Despite matching on these factors, the left temporal lobectomy group was overall more impaired at naming, achieving an average of only 35.8 (S.D. = 12.9) on the Boston Naming Test, compared with an average score of 52.5 (S.D.=9.4) for the right temporal lobectomy group. This difference is presumably a reflection of left hemisphere specialization for speech and language. Twenty subjects were

L. J. Tippett et al./A category-specific naming impairment after temporal lobectomy

141

Procedure

tested 3-6 weeks post-surgery, and the remaining 11 were seen in follow-up, 1 or more years after operation. The right temporal lobectomy subjects were older (mean=40.9 years) than the left temporal subjects (mean= 32.19), t (29)=2.202, P < 0.05. If anything, the older age of the right temporal group might have made the speeded task more difficult, but there was no evidence of this in the data. In general, patients received a standard anterior temporal lobectomy. In the nondominant hemisphere, the resection line measured 5-5.5 cm from the temporal tip, and in the dominant hemisphere, it measured 4.5-5 cm from the temporal tip. The amygdala and anterior 1.5-3 cm of hippocampus were removed, usually en bloc. Normal subjects. Sixteen normal subjects were also tested. These were undergraduate students at the University of Pennsylvania, whose age ranged from 18-26 years.

Testing began with practice items, which familiarized subjects with the task and also allowed the experimenter to find a suitable rate of picture presentation. The goal of this procedure was to find a pace at which subjects made errors, but were able to recover after each error in order not to commit long runs of errors. Each subject was required to name the practice stimuli, initially presented at the rate of one picture every 1.5 sec. The pictures were hand-turned by a highly practiced tester, paced by the beat of a metronome. If the subject made no errors and seemed to find the task easy, the speed was increased and the pictures presented again in a different order. This was repeated until the subject either made one or more errors, or was falling behind the rate of presentation of the pictures. The upper speed limit was one every 0.83 sec, as this was the fastest speed at which the tester could reliably turn the cards without fumbling. Once the rate of picture presentation was determined, each subject was presented with the 85 test items in three blocks. Each block of pictures began with a dummy item. There was a rest interval of 30 sec between blocks, or longer if the subject wanted more time. Whenever possible the stimuli were presented to each subject a second time, in a different order. Two orders of the stimuli were used and order was counterbalanced across groups of subjects. All responses were recorded on an auditory tape and later transcribed. Only accuracy data were obtained. Errors were defined as either nonresponses to stimuli, or responses other than the dominant name or acceptable synonym as classified by Snodgrass and Vanderwart [22].

Materials Line drawings from the corpus of Snodgrass and Vanderwart [22] were used, corresponding to the two sets of drawings used by Funnell and Sheridan [8] in their research on categoryspecific knowledge deficits. In their first set of 68 pictures [8, Appendix 2], living and nonliving items are matched-pairwise for familiarity and word frequency (Set 1 in this study). In their second set of 24 pictures [8, Appendix 3], the items are matched for familiarity, word frequency, and visual complexity (Set 2). In this study, the item "tree", which had been listed as a nonliving item by Funnell and Sheridan, was replaced by "key". Because some of the same items appear in the two sets of pictures, there is a total of 85 pictures. Each of these pictures was centered on a 6 x 4 in. white card. An additional 10 pictures, again comprising living and nonliving items matched for word frequency and familiarity, were used as practice items.

Results F i g u r e 1 a n d T a b l e 1 s h o w the a c c u r a c y o f each g r o u p at n a m i n g the 34 pairs o f living a n d n o n l i v i n g items c o m p r i s i n g Set 1. W h e r e a s c o n t r o l subjects a n d right temporal lobectomy patients perform approximately

100

80

•

Living

[]

Nonliving

60

o~

40

20

D Left

Temporal

Right

Temporal

Normal

Subjects

Group Fig. 1. Accuracy naming 34 pairs of living and nonliving pictures comprising Set 1 (from Appendix 2, Funnell and Sheridan, 1992) on the first administration of a paced naming task.

142

L.J. Tippett et al./A category-specific naming impairment after temporal lobectomy

Table 1. Mean accuracy at naming living and nonliving pictures comprising Set 1 (Administration 1) in a paced presentation Living (n = 34)

Nonliving (n = 34)

t-value (paired, two-tailed)

P

27.06 (S.D. = 5.91) 28.86 (S.D. = 3.72) 30.31 (S.D. = 2.52)

23.88 (S.D. = 5.92) 28.71 (S.D. = 3.63) 29.69 (S.D. = 2.24)

4.299

0.001

0.141

0.890

1.022

0.323

Group Left temporal (n = 17) Right temporal (n = 14) Normal subjects (n = 16)

equally with living and nonliving items, the left temporal lobectomy patients are impaired at nonliving relative to living items. This difference for the left temporal group was significant by matched pairs t-test over subjects, t (16) = 4.299, P < 0.001 (two-tailed). In contrast, neither of the other groups showed significant effects of picture category, P > 0. I. In order to verify that the difference between living and nonliving items for the left temporal group was general over the stimulus set, and not due to a small number of aberrant items in one or the other category, we also tested the difference over items, and again found it to be significant, t (33)= 2.595, P < 0.02 (two-tailed). In order to assess whether the pattern of results in the left temporal lobectomy group was significantly different from the pattern in the other two groups, an analysis of variance was performed, with factors Group (left temporal lobectomy, right temporal lobectomy, and normal control; a between-subjects factor), Category (living and nonliving; a within-subjects factor), and Order (picture order 1 and picture order 2; a betweensubjects factor). If the living-nonliving difference is significantly greater for left temporal lobectomy patients than for the other subjects, there will be a significant

interaction between Group and Category. This prediction was confirmed, F (2, 41)= 6.338, P