- Email: [email protected]
Cortical and subcortical influences on clustering and switching in the performance of verbal fluency tasks
Neuropo'chologia, Vol.
~
Pergamon
PII: S0028-3932(97)00153-X
36, No. 4. pp. 295 304, 1998 i ~ 1998 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0028 3932"98 $19.00+0.00
Cortical and subcortical influences on clustering and switching in the performance of verbal fluency tasks ALEXANDER I. TROSTER,* JULIE A. FIELDS,* JULIE A. TESTA,? ROBERT H. PAUL,t CARLOS R. BLANCO,t KAREN A. HAMES? DAVID P. SALMON~ and WILLIAM W. BEATTYt§ * Department of Neurology, University of Kansas, Medical Center, 3901 Rainbow Blvd, Kansas City, KS 66160, U.S.A.; t Department of Psychiatry and Behavioral Sciences, University of Oklahoma Health Sciences Center, P.O. Box 26901, Oklahoma City, OK 73190, U.S.A.; ++Alzheimer Disease Research Center, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093, U.S.A. (Received 6 June 1997; accepted l0 October 1997)
Abstract- Impairments on lexical and semantic fluency tasks occur in both cortical and subcortical dementia. Recent reports that
the average size of phonemic and semantic clusters is reduced in Alzheimer's disease (AD), but not in Parkinson's disease (PD) could support the hypothesis that in AD verbal fluency deficits arise from degraded memory storage while in PD the same impairments result from defective retrieval. In the present study, patients with AD, PD with dementia, or Huntington's disease produced fewer words, fewer switching responses and smaller semantic cluster sizes. Patients with multiple sclerosis, regardless of whether or not they were demented, produced fewer words and switching responses, but normal size clusters, and patients with PD without dementia performed normally on all fluency measures. These results indicate that reductions in cluster size on verbal fluency tests are best interpreted as changes in the efficiency of access to lexical and semantic memory stores. The findings are also consistent with the idea that patterns of cognitive impairment may differ among diseases that result in subcortical dementia. ~ 1998 Elsevier Science Ltd. All rights reserved. Key Words: Alzheimer's Disease; Parkinson's Disease; Multiple Sclerosis; lexical memory; semantic memory.
two-process model involving: (1) a search for semantic subcategories followed by (2) an output mechanism to produce as many words as possible from the subcategories [21, 44]. Troyer et al. [43] recently devised a method to quantify these processes as switches and average cluster size. For normal subjects, switching and clustering contributed equally to performance on a semantic fluency task while switching was more strongly related to word production on the phonemic fluency task. They concluded that cluster size is a measure of the ability to access words within phonemic and semantic subcategories while switching is a measure of the ability to shift efficiently from one subcategory to another. By this account a reduction of cluster size could reflect either a breakdown in the integrity of memory stores or simply a reduction in the efficiency of access to these stores. Impairments in word production on fluency tests accompany focal lesions of the frontal and temporal lobes [13, 32, 35] as well as neurodegenerative diseases [1, 4,
Verbal fluency tests have been used extensively in neuropsychology to determine the status of patients' lexical (phonemic) and semantic memories. On these tests subjects must generate words that begin with a particular letter or are exemplars of a certain category. The usual dependent variable for both tests is the number of correct words; repetitions, intrusions, and other types of rule violations are sometimes recorded as well. Several investigators have observed that words tend to be produced in semantic clusters on semantic fluency tasks and in phonemic clusters on phonemic fluency tasks. The clusters are defined by bursts of words over time that are semantically or phonemically related [21, 38, 44]. The temporal pattern of responding suggests a
§Correspondence should be addressed to: Dr W. W. Beatty, University of Oklahoma, Health Sciences Center, Department of Psychiatry, P.O. Box 26901, Oklahoma City, OK 73190, U.S.A. Tel.: 405/271-2474; Fax: 405/271-6236. 295
296
A.I. TrOster et al./Verbal fluency
7, l l, 33, 39]. In Alzheimer's disease (AD), which is considered a prototype of cortical dementia, deficits are usually observed on both phonemic and semantic fluency tasks, but the deficits are larger in magnitude on semantic fluency tests [1 l, 30]. The patterns of loss are distinctive; mildly demented AD patients are impaired in producing specific exemplars of semantic categories but appear to retain knowledge of words that designate superordinate classes [40]. These data, together with studies of naming and other aspects of semantic memory in AD, suggest that the structure of this memory system is disrupted, perhaps irreversibly, in this disease [12, 23]. Others [31] maintain that the content of semantic memory in AD is intact, but the knowledge is simply inaccessible. Martin and colleagues [27, 28] have argued that semantic knowledge about objects and categories is organized in a hierarchical manner with the more general aspects at the top and the more specific features at the bottom. In their view, AD patients suffer a progressive "bottom-up" breakdown in the hierarchical organization of semantic knowledge. Analyses of switching and clustering by AD patients on verbal fluency tasks were reported in two recent studies [9, 42]. In one study, mean cluster size on the semantic fluency task and the importance of clustering for predicting word generation was reduced for the AD patients. Cluster size for the AD patients was reduced on the phonemic task but not significantly. However, no relationship between cluster size and mental status was observed [9]. In a second study, Troyer and Moscovitch [42] compared clustering and switching on phonemic and semantic fluency tasks in patients with AD or Parkinson's disease (PD). On both tasks, PD patients generated fewer correct words and made fewer switches than control subjects. AD patients also generated fewer words but made fewer switches only on the semantic task. However, unlike AD patients, the PD subjects had normal cluster sizes on both phonemic and semantic tasks. Parkinson's disease is considered to be a form of subcortical dementia [15, 17] in which the major cause of memory deficits, including lowered output on fluency tests, is believed to be retrieval failure rather than degraded storage. If cluster size is assumed to reflect mainly the integrity of lexical and semantic memory stores, then these [9, 42] results can be taken to support the idea of Martin [27] and others [12, 23] about the status of semantic memory in AD. Further, the findings of Troyer and Moscovitch [42] can be regarded as support for the clinical cortical subcortical distinction in dementia, because an effect on cluster size would not be predicted in PD. PD has extremely variable effects on cognition and it is not clear that the AD and PD patients in the Troyer and Moscovitch [42] study were of similar dementia severity. Brown and Marsden [10] have questioned the concept that there are distinctive patterns of cognitive impairments in the cortical and subcortical dementias, noting
that many of the claimed differences disappeared when dementia severity was controlled. In Experiment 1, we re-examined possible differences in patterns of clustering and switching between AD and PD patients. To provide rigorous control over dementia severity, two groups of PD patients were tested: One matched the AD patients in overall mental status, while performance of the other PD group approximated the performance of controls on a standard mental screening examination. If average cluster size is regarded as a measure of the integrity oflexical and semantic memory stores, then only AD patients should exhibit reduced cluster size and the magnitude of this effect should be correlated with dementia severity [27]. On the other hand, if cluster size primarily measures efficiency of access to lexical and semantic memory, then cluster size should be reduced in both cortical and subcortical dementia. Again, the magnitude of the effect should be related to dementia severity. Brown and Marsden [10] also questioned the idea that subcortical dementia is a homogeneous construct, noting that the differences in patterns of cognitive impairment among the various diseases that are supposed to result in subcortical dementia are almost as great as the differences between AD and any of the diseases considered to produce subcortical dementia. To address this issue, in Experiments 2 and 3, clustering and switching were studied in groups of patients with multiple sclerosis (MS) or Huntington's disease (HD).
Experiment 1 Subjects
All participants were literate residents of the greater Kansas City area and spoke English as their first language. Normal control participants ( N = 30) were recruited by advertisements and contact with caregiver support groups. They were community or retirement centre residents, free of neurologic or psychiatric illness. None were taking medications with CNS effects. None had histories of developmental disorders, substance abuse or head trauma. Patients with AD, PD non-demented (PDND) and PD demented (PDD) (N = 30/group) were recruited from among participants in a longitudinal study of neurodegenerative disease. Parkinson's disease patients were also recruited from an academic medical centre's movement disorders clinic. All patients underwent extensive medical and neurologic workups. No patient had medical illness (e.g. COPD) which might compromise cognition. No patient had history of neurologic disorders other than AD or PD (including head trauma with loss of consciousness). Patients were free of psychiatric illness (including substance abuse), but history of depression in PD, given its prevalence, was not an exclusionary criterion. The diagnosis of PD was made by a neurologist
A. I. Tr6ster et al./Verbal fluency on the basis of the presence of two of three cardinal signs (bradykinesia, tremor, rigidity) and responsiveness to Ldopa therapy. The diagnosis of dementia in PD was based on Cummings and Benson's [16] criteria, and motor signs antedated appearance of cognitive difficulties. Patients who had end-of-dose motor fluctuations were always tested in the " o n " phase. Patients with AD were diagnosed as having "probable" AD, using the NINCDS A D R D A criteria [26]. AD patients with extrapyramidal signs were excluded from this study. Participants or their caregivers provided written informed consent after a thorough explanation of the procedures which were approved by the local Institutional Review Board (IRB).
Procedure
As a part of a larger battery of neuropsychological tests that required about 3 h to administer, the subjects received the Dementia Rating Scale (DRS) [29], the Boston Naming Test (BNT) [24], a letter (FAS) fluency test and a semantic fluency test. On the letter fluency test, subjects were allowed 60 s to generate as many words as possible that began with each of the letters, f, a and s. Proper names and numbers were not allowed. On the semantic fluency task, subjects were allowed 60 s to name as many animals as possible. Responses were recorded on audio tape and transcribed verbatim for later analysis by the Fluency method of Troyer et al. [43]. For each fluency test the number of correct words (i.e. total words minus repetitions and rule violations), the mean cluster size, and the number of switches were determined according to the scoring rules provided by Troyer et al. In brief, for phonemic fluency, clusters were defined as groups of successively generated words that began with the same first two letters (e.g. fire, financial), differed only in a single vowel sound (e.g. sit, set), rhymed (e.g. strip, ship) or were homonyms (e.g. see, sea). Homonyms were counted only if the participant indicated that the two words were different exemplars during the word generation task. On the semantic fluency task, clusters were defined as successively generated words that belonged to the same semantic subcategory (e.g. pets, farm animals, birds, herbivores, African animals). These categories are not mutually exclusive. Cluster size was counted beginning with the second word (i.e. a two-word cluster was counted as a cluster size of one) and the mean cluster size was determined separately for phonemic and semantic fluency. Switches were counted as the number of transitions between clusters, including single words (i.e. cluster size = 0). Following the procedure of previous studies [9, 43], repetitions and rule violation errors were included in the calculation of cluster size and switches. To determine interrater reliabilities, two raters scored all of the protocols for switching and clustering responses independently. Initial analyses demonstrated that the
297
sizes of the correlations were similar for each of the four groups. When computed for all 120 subjects, the average r = 0.968 (Range = 0.919 0.993). Normal subjects sometimes appear to use alphabetic strategies for retrieving words on tasks (e.g., absent, accurate, add for words that start with "a" or buffalo, cow, dog, elephant for animals). Protocols for all subjects were scored for alphabetic strategies, but the alphabetic strategies were not considered clusters.
Results Table 1 summarizes the demographic characteristics of the tbur groups as well as their average performance on the DRS and the BNT. There were no significant differences among groups for age or education (Fs (3, 116) < 1.27, n.s.), but the distribution of men and women was not proportional across groups (Z2 ( 3 ) = 11.46 P < 0.01). There were more men than women in both PD groups while the reverse was true for the AD group. To examine the consequences of this confound, preliminary Group × Gender analyses were performed on all dependent variables. These analyses revealed no significant effects of Gender or the Groups × Gender interaction on any variable (all Fs < 2.18, Ps >0.05). To facilitate description, the data were combined for male and female subjects. One way analyses of variance revealed significant differences among groups on the BNT and all scales from the DRS (Fs (3, 116)> 6.85, Ps <0.001). Subsequent analyses with Duncan Range Tests showed that the AD and the PDD groups differed from the control and P D N D groups on all measures. In addition, the AD patients performed more poorly than the PDD patients on the BNT and the Memory Scale from the DRS. Conversely, the PDD patients performed more poorly than the AD patients on the Initiation/Perseveration Scale from the DRS. One way ANOVAs on the verbal fluency data (Table 2) revealed significant differences among groups on all of the measures except the number of switching responses per correct word on the Animals task (Fs (3,116) > 4.10, Ps < 0.001). Subsequent comparisons with Duncan Range tests revealed that the AD and the PDD groups generated fewer correct words and made fewer switching responses than either the control or P D N D groups, which did not differ. AD patients produced smaller clusters than subjects from the other three groups on the phonemic (FAS) task. There was no significant difference among controls and either group of PD patients on this measure. On the semantic clustering measure (Animals), the PDD and the AD groups produced smaller clusters than controls. Performance by the P D N D group was intermediate and not significantly different from that of any of the other groups. On the phonemic task, the average number of switching responses per correct word was greater for the AD
298
A. I. Tr6ster et al./Verbal fluency Table 1. Demographic and clinical performance in experiment 1: mean (S.D.) Controls M/F Age Education BNT DRS Attention Initiation/Perseveration Construction Conceptualization Memory Total
AD
14/16 70.8 (7.0) 13.9 (2.4) 55.5(2.8)
PDD
11/19 69.7 (5.9) 13.0 (1.7) 37.4(13.4) "`b'c
35.2(1.9) 35.9 (2.2)
32.9(3.3) "'8 26.4 (6.5) "'8c
5.8 (0.5)
5.3 (1.0) a'b
36.2 (2.6) 24.0(1.1) 137.0(4.3)
29.6 (6.0) "'b 12.9(3.2) ~'b'c 107.4(13.4) ~b
PDND
23/7 72.2 (6.8) 13.6 (2.1) 43.5(9.8) ~8
19/11 70.8 (7.4) 14.0 (2.3) 52.2(5.1)
32.4(3.8) "b 23.2 (6.2) "`8 4.8 (1.6) a'b 28.5 (4.9) "'b 17.4(3.8) ~''b 107.4(14.0) ab
35.1 (2.8) 34.5 (3.2) 5.9 (0.4) 35.5 (2.7) 22.7(2.5) 133.7(5.9)
Significantly different from controls, P < 0.05. b Significantly different from group PDND, P < 0.05. c Significantly different from group PDD, P < 0.05.
Table 2. Verbal fluency performance in experiment 1: mean (S.D.) Controls FAS No, Correct Cluster Size Switches Switches/Word Animals No. Correct Cluster Size Switches Switches/Word
AD
PDD
PND
35.3(12.4) 0.36(0.20) 26.2 (8.4) 0.75 (0.12)
20.7(10.5) ~'b 0.16(0.14) a'b'c 20.0 (9.3) a'bx 0.98 (0.17) a
15.2(7.9) "`8 0.26(0.17) 13.5 (6.2) a'b 0.93 (0.26)"
33.0(14.3) 0.32(0.20) 26.8 (9.5) 0.85 (0.17)
17.8 (4.1) 1.28 (0.54) 7.6 (2.6) 0.43 (0.10)
7.6 (4.6) ~'b 0.82 (0.60) ~ 3.8 (2.9) "'b 0.47 (0.23)
7.0 (3.8) ~'b 0.83 (0.67)" 3.1 (2.1),.b 0.41 (0.20)
16.3 (5.3) 1.13 (0.74) 7.7 (2.3) 0.51 (0.20)
Significantly different from controls, P < 0.05. b Significantly different from group PDND, P < 0.05. c Significantly different from group PDD, P < 0.05.
and P D D groups than for controls. Performance by the P D N D groups was intermediate and not significantly different from that of any other group. T o examine the relationship of clustering with severity of dementia, correlations were c o m p u t e d between D R S Total and both measures of average cluster size. For the P D D group both correlations were significantly different from zero (r = 0.58, P < 0.001 for p h o n e m i c fluency; r = 0.52, P < 0.01 for semantic fluency). F o r the A D group, both correlations were positive, but only the correlation with p h o n e m i c clustering attained significance (r = 0.38, P < 0.05 for phonemic fluency; r = 0.17, n.s. for semantic fluency). N a m i n g was not systematically related to either measure of cluster size for either group (absolute rs < 0,10, n.s.). F o r the A D group the n u m b e r of switching responses was highly correlated with dementia status as indexed by D R S Total (r = 0.67, P < 0.001 for p h o n e m i c fluency; r = 0.71, P < 0.001 for semantic fluency). The pattern was similar for the P D D group, but the correlations were not as high (r = 0.34, P < 0.10 for p h o n e m i c fluency, r = 0.49, P < 0.01 for semantic fluency).
Analyses of the use of alphabetic search strategies revealed that only one control subject appeared to use such a strategy on the phonemic fluency task. On the semantic task 11 controls, 14 P D N D patients, 7 P D D patients and 6 A D patients produced 1-2 strings each of animal names that could have been generated by an alphabetic strategy. However, 8 of the groups of names generated by controls, 6 by P D N D patients, 1 by P D D patients and 2 by A D patients were also part of semantic clusters. These findings indicate that alphabetic strategies contributed little to the verbal fluency p e r f o r m a n c e of any of the groups.
Discussion W h e n equated for dementia severity, the P D D and A D patients showed similar reductions in the n u m b e r of correct words and the n u m b e r of switching responses on b o t h fluency tasks. Although average cluster size on the p h o n e m i c fluency task for the P D D patients was similar to that of the control and P D N D subjects, this was not
A. I. Tr~3ster et al./Verbal fluency the case for the semantic fluency task. On this task, the average cluster size for the A D and the P D D groups was significantly smaller than for controls, but there was no difference in the average cluster size for A D and P D D groups on the semantic task. This occurred despite the fact that the P D D patients were not as severely impaired as the A D patients on the BNT, a classic measure of semantic memory. Overall, these results do not support the idea that impairments in word generation on semantic fluency tasks in AD and P D D arise from fundamentally different processes (i.e., retrieval failure for PD, degraded storage for AD). This idea might still apply to word generation on phonemic tasks~ on which cluster size was reduced for A D patients, but not for P D D patients, Although the results of Experiment 1 suggest that the influence of A D on semantic clustering may be more closely related to the presence of dementia than to any unique aspect of the neuropathology of AD, this conclusion must be offered tentatively. Because PD and A D generally become evident at about the same age and both diseases are relatively common in elderly persons, there is no satisfactory way, in life, of ensuring that patients with PD who are also demented do not also suffer from AD. The force of this argument for the present study is attenuated by the observed differences between the P D D and A D groups in naming (BNT) and in the patterns of impairment on the DRS. As would be expected from previous findings [34], the P D D patients showed greater impairment on the lnitiation/Perseveration scale while the A D patients showed greater impairment on the Memory Scale. These differences were small in magnitude and only apply to group comparisons. MS and H D also can give rise to subcortical dementia (16); most patients exhibit symptoms before the age of 40, at an age when A D is quite rare. In Experiments 2 and 3, we examined the influences of clustering and switching on the verbal fluency performances by MS and H D patients.
Experiment 2 MS is a demyelinating disease of the central nervous system (CNS). Although lesions can occur anywhere in the CNS, there is a preponderance of damage in the periventricular white matter and the long sensory and motor tracts [19]. Patients with MS consistently produce fewer correct words on both phonemic and semantic fluency tests [2, 4-6, 8]. On tests of confrontation naming, some MS patients exhibit mild impairment, but more serious aphasic disturbances are quite rare [2]. The most common explanation for the verbal fluency deficits in MS is that they arise from difficulties in retrieving information from lexical and semantic memory [37] which are compounded by slowed information processing and dysarthria [2]. If these ideas are correct and average cluster size measures the integrity of lexical and semantic
299
memory, then the reduction in the output of correct words by MS patients should be accompanied by reduction in the number of switches, but the average cluster size should not be affected. Experiment 2 provided a test of these predictions.
Method Subjects. The participants were 133 patients (36 men, 97 women) with clinically definite MS [36] and 63 (20 men, 43 women) healthy control subjects. All were literate and spoke English as their first language. Subjects in both groups met the exclusionary criteria described in Experiment 1. Forty-eight patients and 38 controls were from North Dakota and 76 patients and 25 controls were from Oklahoma. Preliminary analyses revealed no significant differences associated with gender or region for patients or controls so the data were pooled across these variables. There is no simple examination for identifying dementia in MS. Very few MS patients score below the range of normals [2, 3] on tests like the MMSE [20], but a significant proportion of MS patients score below the 5th percentile of controls on three or more cognitive domains, one operational definition of dementia [16]. Impairment on the BNT is a good predictor of this sort of broad cognitive impairment for MS patients [3]. Accordingly, the MS patients were divided into two groups based on their BNT scores: <50 (N = 26) and 50 or higher (N = 107). The former group scored more than 2 S.D. below the mean of controls on the BNT and scored below the 5th percentile of controls on an average of 4.5 out of the 7 cognitive domains (vocabulary, attention, memory, information processing speed, abstraction problem solving, naming, visuo-spatial perception) studied by Beattv et al. [8]. Procedure. As part of a larger battery of neuropsychological tests lhat lasted approximately 2 h, all subjects received the BNT [24], the phonemic fluency test (FAS) and the Animals test. These tests were administered and scored as described in Experiment 1. Data for the fluency tests were previously reported [5, 6] in terms of numbers of correct words generated, but analyses of switching and clustering were not included in the previous reports. All subjects provided written informed consent after a thorough explanation of the procedures which were approved by the local IRB, The M S patients averaged 3.7_ 2.8 on the Ambulation Index (AI) [22], a measure of physical disability that is highly correlated ( r = 0.96) [4] with Kurtzke's [25] Expanded Disability Status Scale (EDSS). Because the interrater reliabilities for both switching and clustering measures were consistently high for all four groups in Experiment 1, only 30 protocols, selected at random, from each of the MS and control groups were scored independently by two raters in Experiment 2. The sizes and patterns of correlations were similar for both groups; the average correlation was 0.969 (Range: 0.922-0.997).
Results and Discussion Table 3 summarizes performance by both groups of MS patients and control subjects. Here it can be seen that the groups were comparable in average age and education, but, as expected, the MS patients produced fewer correct words on both the phonemic (FAS) and the
A. I. Triister et &/Verbal fluency
300
Table 3. Mean (S.D.) performance by MS patients and controls in experiment 2
Age Education BNT FAS Correct FAS Cluster Size FAS Switches FAS Switches/Word Animals Correct Animals Cluster Size Animals Switches Animal Switches/Word
Controls
MS Patients BNT > 50
MS Patients BNT < 50
46.5(16.1) 13.9 (2.3) 55.8 ( 3.3) 43.0(13.7) 0.35 (0.15) 32.3 ( 9.9 ) 0.77(0.18) 20.6 (5.0) 1.37 (0.57) 9.0( 3.7) 0.43 (0.14)
42.7 ( 9.3) 14.4 (2.2) 56.2 (2.8) 35.3 (12.7)” 0.34 (0.24) 25.4 (9.0)” 0.74(0.15) 17.9(5.5)” 1.36 (0.76) 7.7(3.5) 0.43(0.13)
43.6 (12.6) 13.3 (2.0) 44.2 (6.0)‘.b 21.6( 8.6)“,h 0.42 (0.3 1) 16.7 (7. 9)a.b 0.82 (0.29) 13.8 (7.7)“,h 1.27 (0.72) 5.8 (3.7)“,h 0.46(0.15)
a Significantly different from controls, P < 0.05 ’ Significantly different from MS patients with BNT 2 50, P < 0.05
193) > 14.21, were accompanied by reductions in the number of switching responses (Fs (2, 193) > 7.29, Ps < O.OOl),but there was no difference between groups in the mean cluster size or the number of switching responses per correct word for either fluency task (all Fs < 2.23). Subsequent analyses showed that the MS patients with normal BNT scores produced fewer correct words than controls on both fluency tests and fewer switching responses on the FAS task. Patients with impaired naming differed from controls and patients with normal naming in the number of correct words and switching responses on both tasks. Like the AD and PDD patients, the MS patients, especially those with global cognitive impairment, generated fewer words and made fewer switching responses than controls. Unlike the AD and PDD patients, the MS patients showed normal cluster sizes on both fluency tasks. semantic
(Animals)
tasks
(Fs
(2,
Ps < 0.001). These deficits in word generation
Experiment 3 HD is a hereditary disease of the CNS which is characterized by uncontrollable choreiform movements and progressive intellectual deterioration. The initial target of the disease process is the neostriatum, particularly the head of the caudate nucleus. As the disease progresses, cortical regions, especially those areas with extensive and reciprocal connections with the caudate also deteriorate [40]. For this reason, HD in its early stages is often considered a prototype of subcortical dementia [15]. Impairments on verbal fluency tasks by HD patients appear early in the course of the disease and worsen with disease progression [1, Ill. In most studies, deficits of comparable magnitude on phonemic and semantic fluency tasks have been observed, even in patients whose confrontational naming abilities appear normal [ 111.For this reason, fluency deficits in HD are typically attributed
to retrieval failure rather than to degradation of the lexical or semantic memory stores. If this explanation is correct and average cluster size is a measure of the integrity of semantic memory, then reduced word output should be associated with fewer switching responses, but the average size of both phonemic and semantic clusters should not be altered. Experiment 3 tested this hypothesis.
Method Subjects. The 24 HD patients (13 men, 11 women) were literate, spoke English as their first language and were residing at home in the San Diego, CA area at the time of testing. They were diagnosed by a board certified neurologist on the basis of a positive family history and the presence of choreiform movements. Exclusionary criteria were the same as in Experiments 1 and 2. The patients averaged 117.9 on the DRS (Range: 68-140). Other clinical and demographic data are reported in Table 4. The HD patients received the DRS, the BNT, the FAS and the Animals fluency tests as part of a larger battery that lasted about 3 h. The performance of the HD patients was compared to that of the control subjects from Experiment 2. The HD patients and their guardians provided written informed consent after a thorough explanation of the study procedures which were approved by the local IRB. Interrater reliabilities (determined as in Experiment 1) for the HD patients on the clustering and switching measures averaged 0.973 (Range: 0.9424.999).
Results and Discussion
As seen in Table 4, the HD and control groups were comparable in age and education. Although the proportion of females in the control group was higher than in the HD group (68 vs 46%) this difference was not significant (x2 (1) = 2.82). Comparison of scores on the BNT and the numbers of correct words and switching responses on the phonemic
A. 1. Tr6ster et al./Verbal fluency
301
Table 4. Mean (S.D.) performance by HD patients and controls in experiment 3
Age Education BNT FAS Correct FAS Cluster Size FAS Switches FAS Switches/Word Animals Correct Animals Cluster Size Animals Switches Animals Switches/Word
HD Patients
Controls
F(1, 85) for groups
50.2(12.9) 14.0 (2.4) 45.7 (9.5) 12.0 (7.0) 0.17 (0.21 ) 10.0 (5.6) 0.98 (0.59) 7.5 (3.6) 0.92 (1.1) 3.9 (2.8) 0.48 (0.24)
46.5(16.1) 13.9 (2.3) 55.8 (3.3) 43.0 ( 1 3 . 7 ) 0.35 (0. l 5) 32.3 (9.9) 0.77 (0.18) 20.6 (5.0) 1.37 (0.57) 9.0 (3.7) 0.43 (0.14)
1.2l 0.06 24.60*** 187.14"** 11.90"* 59.14"** 0.29 176.09"** 3.61 45.43*** 0.89
**P < 0.01, ***P < 0.001
and semantic fluency tasks revealed highly significant differences between groups. However, the number of switching responses per word was not significantly different on either task. The average cluster size was significantly smaller for H D patients than for controls on the phonemic fluency task. On the semantic fluency task, a similar effect was observed but the trend was only marginally significant (P < 0.10) because of the highly variable performances among H D patients. Inspection of individual scores showed that three H D patients had mean semantic cluster sizes of 3.0 or greater while the remaining patients attained an average semantic cluster size of 0.52. Because parametric statistics were inappropriate, the data were reanalysed using non-parametric techniques. The median cluster size (1.27) was determined for the control group and the frequency of subjects above, or equal to or below that score was determined for each group. Significantly fewer H D patients than control subjects attained average cluster sizes above the control median (Z2 (1) = 7.62, P < 0.01). For the H D group, the correlations of dementia severity (DRS Total) with mean cluster size were 0.26 for phonemic clusters and 0.03 for semantic clusters. Neither correlation approached significance. By contrast, DRS Total was positively related to the numbers of words correctly generated and the numbers of switching responses on both fluency tasks (rs > 0.51, Ps < 0.05). An analysis of the possible use of alphabetic strategies, conducted as in Experiment 1, indicated that one patient appeared to use such a strategy on the phonemic task and 7 patients may have used an alphabetic strategy on the semantic task. In 6 of these cases, the alphabetic strategy was also part of a semantic cluster. Like the AD and PDD patients tested in Experiment l, the H D patients showed smaller average cluster sizes on both phonemic and semantic fluency tasks. The performance of the H D patients on the clustering measures contrasts sharply with that of the MS patients studied in Experiment 2. Despite slightly greater impairments in naming, the low BNT-MS patients had significantly
larger phonemic clusters (0.42 vs 0.17, t(48)= 3.16, P < 0 . 0 1 ) and semantic clusters (1.27 vs 0.92, Z2 (l) = 5.99, P < 0.05) than the H D patients.
General Discussion Table 5 provides a summary of the major results of the three experiments. The present findings replicate earlier reports [9, 42] that the average size of both phonemic and semantic clusters of words generated on fluency tasks by AD patients is reduced compared to the cluster sizes of age- and education-matched controls, but in contrast to an earlier report [42], when the severity of dementia for PD and AD patients was carefully equated, both groups exhibited comparable reductions in the size of their semantic clusters. This result underscores the importance of controlling dementia severity in comparisons of cortical and subcortical dementia [10]. The entirely normal performance of the P D N D patients indicates that the changes in semantic cluster size, in particular, and more generally on other measures of verbal fluency, are confined to the dementia of PD and are not characteristic of this disease in general. Results of another study are consistent with this inference [38]. All of these conclusions must be tempered by the impossibility of ensuring, without autopsy, that any individual demented PD patient did not also suffer from AD. Because of differences in the patterns of cognitive impairment of the PDD and AD patients described earlier, co-morbidity cannot provide a unitary explanation of the similarity of performances by the PDD and AD patients on the semantic fluency task. The H D patients had higher scores on the DRS (P < 0.05) and the BNT (P < 0.01) than the AD patients. Yet, the two groups showed quantitatively similar deficits in word generation, switching and in the sizes of their phonemic and semantic clusters. Because the H D patients averaged almost 20 years younger than the AD patients, it is not likely that many of the H D patients also suffered from AD. Therefore, if reductions in cluster size are taken
A. I. Tr6ster et al./Verbal fluency
302
Table 5. Summary of results by patient group and dementia status Semantic Fluency
Phonemic Fluency
AD,D ~ HD,D ~ PD,D" PD, ND . MS, D b MS, ND b
Correct
Switches
Cluster Size
Correct
Switches
Cluster Size
~ ,!, ,L
~, $ $
,l, ,[, --
+ ,L ,l,
,L ,[, ,L
,L ,[, ,L
.
.
$
.
.
,L
.
.
.
E
,[,
Dementia status based on diagnosis and DRS score. bDementia status inferred from BNT score.
as evidence of degradation of lexical and semantic memory as Martin et al. suggest occurs in AD, then the erosion of semantic knowledge evidently can occur in subcortical as well as in cortical dementia. If this is so, then the widely reported distinction that the memory deficits in subcortical dementias like H D and PD arise mainly from defective retrieval mechanisms while the amnesia of AD reflects principally storage defects [15, 39] requires reexamination, Alternatively, average cluster size might simply be regarded as an index of the ease of access to lexical or semantic memory. From this perspective, the similar reductions in average size of semantic clusters for the AD, PD and H D groups present no challenge to conventional theories of semantic memory in dementia. This conclusion is probably the most defensible because there is now ample evidence that AD patients exhibit defective retrieval of information from semantic memory as well as erosion of memory stores [18, 3 l]. Differences in cognitive profiles between patients with subcortical and cortical dementia are often most apparent in the early stages of the diseases. Because the average DRS scores of the present AD and PDD groups were in the moderately demented range, it might be argued that differences in cluster size, particularly for the semantic clusters, would have been detected if mildly demented patients had been studied. Two observations argue against this hypothesis. First, significant differences between the PDD and AD groups were observed in the patterns of impairment on the DRS subscales and on the BNT. Second, the H D patients averaged 10.5 points higher than the AD patients on the DRS, yet they showed similar patterns of reduction in the size of phonemic and semantic clusters. Regardless of their scores on the BNT, the MS patients exhibited reduced word generation and switching on both fluency tasks but normal cluster sizes. The finding that the MS patients with low BNT scores and global cognitive impairment showed normal cluster sizes appears to support the idea of Brown and Marsden [10] that the patterns of cognitive impairment in the subcortical dementias are
heterogeneous. Because it was not possible to equate the PDD, H D and MS patients with low BNT scores for severity of dementia, this conclusion must remain tentative. Troyer et al. [43] defined switching as a measure of the ability to shift from one phonemic or semantic subcategory to another and conceptualized switching as an executive function related to the integrity of the frontal lobes. A serious problem in using the absolute number of switching responses to analyse the verbal fluency performance of brain damaged patients is that the number of switching responses is strictly limited by the total verbal output. Thus, any sizeable reduction in the number of words generated will always be associated with a reduction in switching, a phenomenon observed in all of the patient groups examined in the present study. We attempted to avoid this problem by analysing switching in proportion to total word output. Expressed in this way, there were no significant differences between patients and controls in Experiments 2 and 3. In Experiment 1, the AD and PDD groups had higher relative switching scores than controls or P D N D patients. If this proportional measure is taken as the appropriate metric for switching, then the conclusion would be that the switching component of executive function on verbal fluency tasks is intact in AD, HD, PDD and MS. Such a conclusion is not warranted because the proportional measure of switching is also seriously flawed. Given equivalent word output, subjects who employ phonemic and semantic clustering will necessarily make fewer switching responses. The inherent negative relationship between switching and clustering probably explains why the relative switching scores of the AD and PDD patients were significantly elevated. The most salient fact about the performance of dementia and other brain-damaged patients on verbal fluency tasks is that total output is severely reduced. Analyses of switching and clustering do not address this issue directly and therefore at best can provide only a partial account of verbal fluency performance. What is needed is a new approach that confronts the
A. I. Tr~3steret al./Verbal fluency phenomenon of reduced behavioural output more directly. Because tests of verbal fluency place a premium on rapid responding which is generally reduced in dementia and many other types of brain damage, including measures of information processing speed in future models of verbal fluency should be considered. The absence of a consistent relationship between dementia severity and the average size of the semantic clusters remains puzzling. Regardless of whether one assumes that cluster size is a measure of the integrity of semantic memory stores or only of access to these stores, a positive correlation between mental status and cluster size would be predicted. Considering the AD, PD, and HD groups from the present study and the AD group from our earlier study [9], the average r for the correlation of dementia severity and average semantic cluster size was 0.17 (Range = -0.04-0.52). The problem cannot be attributed to a restriction in the range of scores on the DRS because the magnitude of the correlation for the HD patients, the group with the largest range (68-140), was 0.03 while the size of the correlation for the PDD patients, the group with the smallest range (75 122) was 0.52. Because each of these correlational analyses had relatively low power, we pooled the AD, PDD and HD groups and compared average semantic cluster size for the 28 patients with DRS scores above 120 and the 56 patients with scores of 120 and below. The small difference (0.95 vs 0.83) did not approach significance (F < 1). Finally, the failures to obtain the expected relationships cannot have arisen because of a restriction in the range on the cluster size variable. The ranges of scores for the three dementia groups on the measure of average cluster size were larger than those for controls. A more likely explanation is that the low output of the HD, AD, and PDD patients on the Animals fluency task results in estimates of average cluster size that are unstable. The bimodal distribution of semantic cluster size stores for the HD patients probably is one illustration of this phenomenon. Use of a semantic fluency task on which the dementia patients produced a greater number of words [7] could remedy this problem.
Acknowledgments--Supported by Grant HR3-005 from the Oklahoma Centre for Advancement of Science and Technology and by Grants AG-10182 to the University of Kansas Medical Centre and AG-05131 to UCSD from the National Institute on Ageing.
4.
5.
6,
7.
8.
9.
10.
1 I.
12. 13.
14.
15. 16.
References
17. 1. Barr, A. and Brandt, J., Word-list generation deficits in dementia. Journal of Clinical and Experimental Neuropsvchology, 1996, 18, 810-822. 2. Beatty, W. W., Cognitive and emotional disturbances in multiple sclerosis. Neurologic Clinics, 1983, 11, 189-204. 3. Beatty~ W. W. and Goodkin, D. E., Screening for
18. 19.
303
cognitive impairment in multiple sclerosis: An evaluation of the Mini-Mental State Exam. Archives of Neurology, 1990, 47, 297 301. Beatty, W. W., Goodkin, D. E., Hertsgaard, D. and Monson, N., Clinical and demographic predictors of cognitive performance in multiple sclerosis: Do diagnostic type, disease duration, and disability matter? Archit'es o["Neurology, 1990, 47, 305-309. Beatty, W. W., Goodkin, D. E., Monson, N. and Beatty, P. A, Cognitive disturbances in patients with relapsing-remitting multiple sclerosis. Archives o/ Neurology, 1989, 46, 1113--1119. Beatty, W. W., Goodkin, D. E., Monson, N., Beatty, P. A. and Hertsgaard, D., Anterograde and retrograde amnesia in patients with chronic progressive multiple sclerosis. Archiees of Neurology, 1988, 45, 611 -619, Beatty, W. W., Monson, N. and Goodkin, D. E., Access to semantic memory in Parkinson's disease and multiple sclerosis. Journal of Geriatric PLtvchiatO' and Neurology, 2, 159-168. Beatty, W. W., Paul, R. H., Wilbanks, S. L., Haines, K. A., Blanco, C. R. and Goodkin, D. E., Identifying multiple sclerosis patients with mild or global cognitive impairment using the Screening Examination for Cognitive Impairment (SEFCI). Neurology, 1995, 45, 718 723. Beatty, W. W., Testa, J. A., English, S. and Winn, P, A. Influences of clustering and switching on the verbal fluency performance of patients with Alzheimer's disease. Ageing, Neuropsychology and Cognition, in press. Brown, R. G. and Marsden, C. D., Subcortical dementia: The neuropsychological evidence. Neuroscience, 1998, 25, 363 387. Butters, N., Granholm, E. L., Salmon, D. P., Grant, I. and Wolfe, J., Episodic and semantic memory: A comparison of amnesic and demented patients. Journal ~[ Clinical and Experimental Neuropsychology, 1987, 9, 479-497. Chertow, H. and Bub, D. Semantic memory loss in dementia of the Alzheimer type: What do the various measures measure? Brain, 1980, 113, 397 417. Coslett, H. B., Bowers, D., Verfaellie, M. and Hellman, K. M., Frontal verbal amnesia: Phonological amnesia. Archit'es of Neurology, 1991, 48, 949-955. Cox, D. M., Bayles, K. A. and Trosset, M. W., Category and attribute knowledge deterioration in Alzhelmer's disease. Brain and Language, 1996, 52, 536 550. Cummings, J. L., Introduction. Subcortical Dementia. ed. J. L. Cummings. Oxford University Press, New York, 1990, pp 3-d6. Cummings, J. L. and Benson, D. F., Dementia: A Clinical Approach. Butterworths, Boston, 1983. Cummings, J. L. and Benson, D. F., Subcortical dementia: Review of an emerging concept. Archives (~['Neurolog)', 1984, 41, 871 879. Dopkins, S., Kovner, R., Rich, J, B. and Brandt, J., Access to information about famous individuals in Alzheimer's disease. Cortex, 1997, 33, 333-339. Ebers, G. C., Multiple sclerosis and other demyelinating diseases. In Diseases ~f the Nert,ous System:
304
20.
21.
22.
23.
24. 25. 26.
27.
28.
29. 30.
31.
A.I. Tr6ster et al./Verbal fluency
Clinical Neurobiology, eds A. K. Asbury, G. M. McKhann and W. I. McDonald. Saunders, Philadelphia, 1986, pp 1268-1281. Folstein, M. F., Folstein, S. E. and McHugh, P. R. "Mini-Mental State";: A practical method for grading the cognitive state of patients for the clinician, Journal of Psychiatric Research, 1975, 12, 189198. Gruenewald, P. J. and Lockhead, G. R., The free recall of category examples. Journal of Experimental Psychology: Human Learning and Memory, 1980, 6, 225-240. Hauser, S. L., Dawson, D. M., Lehrich, J. R., Beak M. F., Kevy, S. V., Propper, R. D., Mills, J. A. and Weiner, H. L., Intensive immunosuppression in progressive multiple sclerosis. New England Journal of Medicine, 1983,308, 173-180. Hodges, J. R., Salmon, D. P. and Butters, N., Semantic memory impairment in Alzheimer's disease: Failure of access or degraded knowledge? Neuropsychologia, 1992, 30, 301-314. Kaplan, N., Goodglass, H. and Weintraub, S., Boston Naming Test. Lea and Febiger, Philadelphia, 1983. Kurtzke, J. F., Rating neurological impairment in multiple sclerosis: An expanded disability status scale (EDSS). Neurolo#y, 1983, 33, 1444-1452. McKhann, G., Drachman, D., Folstein, M., Katzinan, R., Price, D. and Stadlan, E. M., Clinical diagnosis of Alzheimer's disease: Report of the NINCDSADRDA Work Group under the auspices of the Department of Health and Human Services Task Force on Alzheimer's disease. Neurology, 1984, 34, 939-944. Martin, A., Representations of semantic and spatial knowledge in Alzheimer's patients: Implications for models of preserved learning in amnesia. Journal of Clinical and Experimental Neuropsychology, 1987, 9, 191-224. Martin, A. and Fedio, P., Word production and comprehension in Alzheimer's disease: The breakdown of semantic knowledge. Brain and Language, 1983, 19, 124-141. Mattis. S., Dementia Rating Scale. Psychological Assessment Resources, Odessa, FL, 1988. Monsch, A. U., Bondi, M. W., Butters, N., Salmon, D. P., Katztnan, R. and Thai, L. J., Comparisons of verbal fluency tasks in the detection of dementia of the Alzheimer type. Archives of Neurology, 1992,349, 1253-1258. Nebes, R. D., Semantic memory dysfunction in Alzheimer's disease: Disruption of semantic knowledge or information-processing limitations? Neuropsychology of Memory. Second Edition. eds. N. Butters and L. R. Squire. Guilford Press, New York, 1992, pp 233 240.
32. Newcombe, F., Missile Wounds of the Brain. Oxford University Press, London, 1969. 33. Ober, B. A., Dronkers, N. F., Koss, E., Delis, D. C. and Friedland, R. P., Retrieval from semantic memory in Alzheimer-type dementia. Journal of Clinical and Experimental Neuropsychology, 1986, 8, 75--92. 34. Paolo, A, N., TrOster, A. I., Glatt, S. A., Hubble, I. P., and Koller, W. C., Differentiation of the dementias of Alzheimer's and Parkinson's disease with the Dementia Rating Scale. Journal of Geriatric Psychiatry and Neurology, 1995, 8, 184-188. 35. Perret, E., The left frontal lobe of man and the suppression of habitual responses in verbal categorical behaviour. Neuropschologia, 1974, 12, 323-330. 36. Poser, C, M., Paty, D. W., Scheinberg, L., McDonald, T., Davis, F. A., Ebers, G. C., Johnson, K. P., Sibley, W. A., Silberberg, D. H. and Tourtellote, W. W,, New diagnostic criteria for multiple sclerosis: Guidelines for research protocols. Annals of Neurology, 1983, 13, 227-231. 37. Rao, S. M., Leo, G. J. and St. Aubin-Faubert, P. On the nature of memory disturbance in multiple sclerosis. Journal of Clinical and Experimental Neuropsychology, 1989, 11, 699-1712. 38. Raskin, S. H., Sliwinski, M. and Borod, J. Clustering strategies on tasks of verbal fluency in Parkinson's disease. Neuropsychologia, 1992, 30, 95-99. 39. Salmon, D. P., Heindel, W. C. and Butters, N., Semantic memory, priming, and skill learning in Alzhelmet's disease. In Memo O, Functioning in Dementia. ed. L. B/ickman. Amsterdam, North Holland, 1992, pp 99-118, 40. Shoulson, I. Huntington's disease. In Diseases o/'the Nervous System: Clinical Neurobiology II, eds A. K. Asbury, G. M. McKhann and W. I. McDonald. Saunders, Philadelphia, 1986, pp 1259-1267. 41. Tr6ster, A. I., Salmon, D. P., McCuUough, D. and Butters, N., A comparison of the category fluency deficits associated with Alzheimer's and Huntington's disease. Brain and Language, 1989, 37, 500513. 42. Troyer, A. K. and Moscovitch, M., Clustering and switching on verbal fluency tests: Evidence from healthy controls and patients with Alzheimer's and Parkinson's Disease. Journal of the International Neuropsychological Society, 1996, 2, 11. 43. Troyer, A. K., Moscovitch, M. and Winocur, G., Clustering and switching as two components of verbal fluency: Evidence from younger and older healthy adults. Neuropsychology, 1997, 11, 138-146. 44. Wixted, J. T. and Rohrer, D., Analysing the dynamics of free recall: An integrative review of the empirical literature, Psychonomic Bulletin & Rev'iew, 1994, 1, 89-106.
~
Pergamon
PII: S0028-3932(97)00153-X
36, No. 4. pp. 295 304, 1998 i ~ 1998 Elsevier Science Ltd. All rights reserved Printed in Great Britain 0028 3932"98 $19.00+0.00
Cortical and subcortical influences on clustering and switching in the performance of verbal fluency tasks ALEXANDER I. TROSTER,* JULIE A. FIELDS,* JULIE A. TESTA,? ROBERT H. PAUL,t CARLOS R. BLANCO,t KAREN A. HAMES? DAVID P. SALMON~ and WILLIAM W. BEATTYt§ * Department of Neurology, University of Kansas, Medical Center, 3901 Rainbow Blvd, Kansas City, KS 66160, U.S.A.; t Department of Psychiatry and Behavioral Sciences, University of Oklahoma Health Sciences Center, P.O. Box 26901, Oklahoma City, OK 73190, U.S.A.; ++Alzheimer Disease Research Center, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093, U.S.A. (Received 6 June 1997; accepted l0 October 1997)
Abstract- Impairments on lexical and semantic fluency tasks occur in both cortical and subcortical dementia. Recent reports that
the average size of phonemic and semantic clusters is reduced in Alzheimer's disease (AD), but not in Parkinson's disease (PD) could support the hypothesis that in AD verbal fluency deficits arise from degraded memory storage while in PD the same impairments result from defective retrieval. In the present study, patients with AD, PD with dementia, or Huntington's disease produced fewer words, fewer switching responses and smaller semantic cluster sizes. Patients with multiple sclerosis, regardless of whether or not they were demented, produced fewer words and switching responses, but normal size clusters, and patients with PD without dementia performed normally on all fluency measures. These results indicate that reductions in cluster size on verbal fluency tests are best interpreted as changes in the efficiency of access to lexical and semantic memory stores. The findings are also consistent with the idea that patterns of cognitive impairment may differ among diseases that result in subcortical dementia. ~ 1998 Elsevier Science Ltd. All rights reserved. Key Words: Alzheimer's Disease; Parkinson's Disease; Multiple Sclerosis; lexical memory; semantic memory.
two-process model involving: (1) a search for semantic subcategories followed by (2) an output mechanism to produce as many words as possible from the subcategories [21, 44]. Troyer et al. [43] recently devised a method to quantify these processes as switches and average cluster size. For normal subjects, switching and clustering contributed equally to performance on a semantic fluency task while switching was more strongly related to word production on the phonemic fluency task. They concluded that cluster size is a measure of the ability to access words within phonemic and semantic subcategories while switching is a measure of the ability to shift efficiently from one subcategory to another. By this account a reduction of cluster size could reflect either a breakdown in the integrity of memory stores or simply a reduction in the efficiency of access to these stores. Impairments in word production on fluency tests accompany focal lesions of the frontal and temporal lobes [13, 32, 35] as well as neurodegenerative diseases [1, 4,
Verbal fluency tests have been used extensively in neuropsychology to determine the status of patients' lexical (phonemic) and semantic memories. On these tests subjects must generate words that begin with a particular letter or are exemplars of a certain category. The usual dependent variable for both tests is the number of correct words; repetitions, intrusions, and other types of rule violations are sometimes recorded as well. Several investigators have observed that words tend to be produced in semantic clusters on semantic fluency tasks and in phonemic clusters on phonemic fluency tasks. The clusters are defined by bursts of words over time that are semantically or phonemically related [21, 38, 44]. The temporal pattern of responding suggests a
§Correspondence should be addressed to: Dr W. W. Beatty, University of Oklahoma, Health Sciences Center, Department of Psychiatry, P.O. Box 26901, Oklahoma City, OK 73190, U.S.A. Tel.: 405/271-2474; Fax: 405/271-6236. 295
296
A.I. TrOster et al./Verbal fluency
7, l l, 33, 39]. In Alzheimer's disease (AD), which is considered a prototype of cortical dementia, deficits are usually observed on both phonemic and semantic fluency tasks, but the deficits are larger in magnitude on semantic fluency tests [1 l, 30]. The patterns of loss are distinctive; mildly demented AD patients are impaired in producing specific exemplars of semantic categories but appear to retain knowledge of words that designate superordinate classes [40]. These data, together with studies of naming and other aspects of semantic memory in AD, suggest that the structure of this memory system is disrupted, perhaps irreversibly, in this disease [12, 23]. Others [31] maintain that the content of semantic memory in AD is intact, but the knowledge is simply inaccessible. Martin and colleagues [27, 28] have argued that semantic knowledge about objects and categories is organized in a hierarchical manner with the more general aspects at the top and the more specific features at the bottom. In their view, AD patients suffer a progressive "bottom-up" breakdown in the hierarchical organization of semantic knowledge. Analyses of switching and clustering by AD patients on verbal fluency tasks were reported in two recent studies [9, 42]. In one study, mean cluster size on the semantic fluency task and the importance of clustering for predicting word generation was reduced for the AD patients. Cluster size for the AD patients was reduced on the phonemic task but not significantly. However, no relationship between cluster size and mental status was observed [9]. In a second study, Troyer and Moscovitch [42] compared clustering and switching on phonemic and semantic fluency tasks in patients with AD or Parkinson's disease (PD). On both tasks, PD patients generated fewer correct words and made fewer switches than control subjects. AD patients also generated fewer words but made fewer switches only on the semantic task. However, unlike AD patients, the PD subjects had normal cluster sizes on both phonemic and semantic tasks. Parkinson's disease is considered to be a form of subcortical dementia [15, 17] in which the major cause of memory deficits, including lowered output on fluency tests, is believed to be retrieval failure rather than degraded storage. If cluster size is assumed to reflect mainly the integrity of lexical and semantic memory stores, then these [9, 42] results can be taken to support the idea of Martin [27] and others [12, 23] about the status of semantic memory in AD. Further, the findings of Troyer and Moscovitch [42] can be regarded as support for the clinical cortical subcortical distinction in dementia, because an effect on cluster size would not be predicted in PD. PD has extremely variable effects on cognition and it is not clear that the AD and PD patients in the Troyer and Moscovitch [42] study were of similar dementia severity. Brown and Marsden [10] have questioned the concept that there are distinctive patterns of cognitive impairments in the cortical and subcortical dementias, noting
that many of the claimed differences disappeared when dementia severity was controlled. In Experiment 1, we re-examined possible differences in patterns of clustering and switching between AD and PD patients. To provide rigorous control over dementia severity, two groups of PD patients were tested: One matched the AD patients in overall mental status, while performance of the other PD group approximated the performance of controls on a standard mental screening examination. If average cluster size is regarded as a measure of the integrity oflexical and semantic memory stores, then only AD patients should exhibit reduced cluster size and the magnitude of this effect should be correlated with dementia severity [27]. On the other hand, if cluster size primarily measures efficiency of access to lexical and semantic memory, then cluster size should be reduced in both cortical and subcortical dementia. Again, the magnitude of the effect should be related to dementia severity. Brown and Marsden [10] also questioned the idea that subcortical dementia is a homogeneous construct, noting that the differences in patterns of cognitive impairment among the various diseases that are supposed to result in subcortical dementia are almost as great as the differences between AD and any of the diseases considered to produce subcortical dementia. To address this issue, in Experiments 2 and 3, clustering and switching were studied in groups of patients with multiple sclerosis (MS) or Huntington's disease (HD).
Experiment 1 Subjects
All participants were literate residents of the greater Kansas City area and spoke English as their first language. Normal control participants ( N = 30) were recruited by advertisements and contact with caregiver support groups. They were community or retirement centre residents, free of neurologic or psychiatric illness. None were taking medications with CNS effects. None had histories of developmental disorders, substance abuse or head trauma. Patients with AD, PD non-demented (PDND) and PD demented (PDD) (N = 30/group) were recruited from among participants in a longitudinal study of neurodegenerative disease. Parkinson's disease patients were also recruited from an academic medical centre's movement disorders clinic. All patients underwent extensive medical and neurologic workups. No patient had medical illness (e.g. COPD) which might compromise cognition. No patient had history of neurologic disorders other than AD or PD (including head trauma with loss of consciousness). Patients were free of psychiatric illness (including substance abuse), but history of depression in PD, given its prevalence, was not an exclusionary criterion. The diagnosis of PD was made by a neurologist
A. I. Tr6ster et al./Verbal fluency on the basis of the presence of two of three cardinal signs (bradykinesia, tremor, rigidity) and responsiveness to Ldopa therapy. The diagnosis of dementia in PD was based on Cummings and Benson's [16] criteria, and motor signs antedated appearance of cognitive difficulties. Patients who had end-of-dose motor fluctuations were always tested in the " o n " phase. Patients with AD were diagnosed as having "probable" AD, using the NINCDS A D R D A criteria [26]. AD patients with extrapyramidal signs were excluded from this study. Participants or their caregivers provided written informed consent after a thorough explanation of the procedures which were approved by the local Institutional Review Board (IRB).
Procedure
As a part of a larger battery of neuropsychological tests that required about 3 h to administer, the subjects received the Dementia Rating Scale (DRS) [29], the Boston Naming Test (BNT) [24], a letter (FAS) fluency test and a semantic fluency test. On the letter fluency test, subjects were allowed 60 s to generate as many words as possible that began with each of the letters, f, a and s. Proper names and numbers were not allowed. On the semantic fluency task, subjects were allowed 60 s to name as many animals as possible. Responses were recorded on audio tape and transcribed verbatim for later analysis by the Fluency method of Troyer et al. [43]. For each fluency test the number of correct words (i.e. total words minus repetitions and rule violations), the mean cluster size, and the number of switches were determined according to the scoring rules provided by Troyer et al. In brief, for phonemic fluency, clusters were defined as groups of successively generated words that began with the same first two letters (e.g. fire, financial), differed only in a single vowel sound (e.g. sit, set), rhymed (e.g. strip, ship) or were homonyms (e.g. see, sea). Homonyms were counted only if the participant indicated that the two words were different exemplars during the word generation task. On the semantic fluency task, clusters were defined as successively generated words that belonged to the same semantic subcategory (e.g. pets, farm animals, birds, herbivores, African animals). These categories are not mutually exclusive. Cluster size was counted beginning with the second word (i.e. a two-word cluster was counted as a cluster size of one) and the mean cluster size was determined separately for phonemic and semantic fluency. Switches were counted as the number of transitions between clusters, including single words (i.e. cluster size = 0). Following the procedure of previous studies [9, 43], repetitions and rule violation errors were included in the calculation of cluster size and switches. To determine interrater reliabilities, two raters scored all of the protocols for switching and clustering responses independently. Initial analyses demonstrated that the
297
sizes of the correlations were similar for each of the four groups. When computed for all 120 subjects, the average r = 0.968 (Range = 0.919 0.993). Normal subjects sometimes appear to use alphabetic strategies for retrieving words on tasks (e.g., absent, accurate, add for words that start with "a" or buffalo, cow, dog, elephant for animals). Protocols for all subjects were scored for alphabetic strategies, but the alphabetic strategies were not considered clusters.
Results Table 1 summarizes the demographic characteristics of the tbur groups as well as their average performance on the DRS and the BNT. There were no significant differences among groups for age or education (Fs (3, 116) < 1.27, n.s.), but the distribution of men and women was not proportional across groups (Z2 ( 3 ) = 11.46 P < 0.01). There were more men than women in both PD groups while the reverse was true for the AD group. To examine the consequences of this confound, preliminary Group × Gender analyses were performed on all dependent variables. These analyses revealed no significant effects of Gender or the Groups × Gender interaction on any variable (all Fs < 2.18, Ps >0.05). To facilitate description, the data were combined for male and female subjects. One way analyses of variance revealed significant differences among groups on the BNT and all scales from the DRS (Fs (3, 116)> 6.85, Ps <0.001). Subsequent analyses with Duncan Range Tests showed that the AD and the PDD groups differed from the control and P D N D groups on all measures. In addition, the AD patients performed more poorly than the PDD patients on the BNT and the Memory Scale from the DRS. Conversely, the PDD patients performed more poorly than the AD patients on the Initiation/Perseveration Scale from the DRS. One way ANOVAs on the verbal fluency data (Table 2) revealed significant differences among groups on all of the measures except the number of switching responses per correct word on the Animals task (Fs (3,116) > 4.10, Ps < 0.001). Subsequent comparisons with Duncan Range tests revealed that the AD and the PDD groups generated fewer correct words and made fewer switching responses than either the control or P D N D groups, which did not differ. AD patients produced smaller clusters than subjects from the other three groups on the phonemic (FAS) task. There was no significant difference among controls and either group of PD patients on this measure. On the semantic clustering measure (Animals), the PDD and the AD groups produced smaller clusters than controls. Performance by the P D N D group was intermediate and not significantly different from that of any of the other groups. On the phonemic task, the average number of switching responses per correct word was greater for the AD
298
A. I. Tr6ster et al./Verbal fluency Table 1. Demographic and clinical performance in experiment 1: mean (S.D.) Controls M/F Age Education BNT DRS Attention Initiation/Perseveration Construction Conceptualization Memory Total
AD
14/16 70.8 (7.0) 13.9 (2.4) 55.5(2.8)
PDD
11/19 69.7 (5.9) 13.0 (1.7) 37.4(13.4) "`b'c
35.2(1.9) 35.9 (2.2)
32.9(3.3) "'8 26.4 (6.5) "'8c
5.8 (0.5)
5.3 (1.0) a'b
36.2 (2.6) 24.0(1.1) 137.0(4.3)
29.6 (6.0) "'b 12.9(3.2) ~'b'c 107.4(13.4) ~b
PDND
23/7 72.2 (6.8) 13.6 (2.1) 43.5(9.8) ~8
19/11 70.8 (7.4) 14.0 (2.3) 52.2(5.1)
32.4(3.8) "b 23.2 (6.2) "`8 4.8 (1.6) a'b 28.5 (4.9) "'b 17.4(3.8) ~''b 107.4(14.0) ab
35.1 (2.8) 34.5 (3.2) 5.9 (0.4) 35.5 (2.7) 22.7(2.5) 133.7(5.9)
Significantly different from controls, P < 0.05. b Significantly different from group PDND, P < 0.05. c Significantly different from group PDD, P < 0.05.
Table 2. Verbal fluency performance in experiment 1: mean (S.D.) Controls FAS No, Correct Cluster Size Switches Switches/Word Animals No. Correct Cluster Size Switches Switches/Word
AD
PDD
PND
35.3(12.4) 0.36(0.20) 26.2 (8.4) 0.75 (0.12)
20.7(10.5) ~'b 0.16(0.14) a'b'c 20.0 (9.3) a'bx 0.98 (0.17) a
15.2(7.9) "`8 0.26(0.17) 13.5 (6.2) a'b 0.93 (0.26)"
33.0(14.3) 0.32(0.20) 26.8 (9.5) 0.85 (0.17)
17.8 (4.1) 1.28 (0.54) 7.6 (2.6) 0.43 (0.10)
7.6 (4.6) ~'b 0.82 (0.60) ~ 3.8 (2.9) "'b 0.47 (0.23)
7.0 (3.8) ~'b 0.83 (0.67)" 3.1 (2.1),.b 0.41 (0.20)
16.3 (5.3) 1.13 (0.74) 7.7 (2.3) 0.51 (0.20)
Significantly different from controls, P < 0.05. b Significantly different from group PDND, P < 0.05. c Significantly different from group PDD, P < 0.05.
and P D D groups than for controls. Performance by the P D N D groups was intermediate and not significantly different from that of any other group. T o examine the relationship of clustering with severity of dementia, correlations were c o m p u t e d between D R S Total and both measures of average cluster size. For the P D D group both correlations were significantly different from zero (r = 0.58, P < 0.001 for p h o n e m i c fluency; r = 0.52, P < 0.01 for semantic fluency). F o r the A D group, both correlations were positive, but only the correlation with p h o n e m i c clustering attained significance (r = 0.38, P < 0.05 for phonemic fluency; r = 0.17, n.s. for semantic fluency). N a m i n g was not systematically related to either measure of cluster size for either group (absolute rs < 0,10, n.s.). F o r the A D group the n u m b e r of switching responses was highly correlated with dementia status as indexed by D R S Total (r = 0.67, P < 0.001 for p h o n e m i c fluency; r = 0.71, P < 0.001 for semantic fluency). The pattern was similar for the P D D group, but the correlations were not as high (r = 0.34, P < 0.10 for p h o n e m i c fluency, r = 0.49, P < 0.01 for semantic fluency).
Analyses of the use of alphabetic search strategies revealed that only one control subject appeared to use such a strategy on the phonemic fluency task. On the semantic task 11 controls, 14 P D N D patients, 7 P D D patients and 6 A D patients produced 1-2 strings each of animal names that could have been generated by an alphabetic strategy. However, 8 of the groups of names generated by controls, 6 by P D N D patients, 1 by P D D patients and 2 by A D patients were also part of semantic clusters. These findings indicate that alphabetic strategies contributed little to the verbal fluency p e r f o r m a n c e of any of the groups.
Discussion W h e n equated for dementia severity, the P D D and A D patients showed similar reductions in the n u m b e r of correct words and the n u m b e r of switching responses on b o t h fluency tasks. Although average cluster size on the p h o n e m i c fluency task for the P D D patients was similar to that of the control and P D N D subjects, this was not
A. I. Tr~3ster et al./Verbal fluency the case for the semantic fluency task. On this task, the average cluster size for the A D and the P D D groups was significantly smaller than for controls, but there was no difference in the average cluster size for A D and P D D groups on the semantic task. This occurred despite the fact that the P D D patients were not as severely impaired as the A D patients on the BNT, a classic measure of semantic memory. Overall, these results do not support the idea that impairments in word generation on semantic fluency tasks in AD and P D D arise from fundamentally different processes (i.e., retrieval failure for PD, degraded storage for AD). This idea might still apply to word generation on phonemic tasks~ on which cluster size was reduced for A D patients, but not for P D D patients, Although the results of Experiment 1 suggest that the influence of A D on semantic clustering may be more closely related to the presence of dementia than to any unique aspect of the neuropathology of AD, this conclusion must be offered tentatively. Because PD and A D generally become evident at about the same age and both diseases are relatively common in elderly persons, there is no satisfactory way, in life, of ensuring that patients with PD who are also demented do not also suffer from AD. The force of this argument for the present study is attenuated by the observed differences between the P D D and A D groups in naming (BNT) and in the patterns of impairment on the DRS. As would be expected from previous findings [34], the P D D patients showed greater impairment on the lnitiation/Perseveration scale while the A D patients showed greater impairment on the Memory Scale. These differences were small in magnitude and only apply to group comparisons. MS and H D also can give rise to subcortical dementia (16); most patients exhibit symptoms before the age of 40, at an age when A D is quite rare. In Experiments 2 and 3, we examined the influences of clustering and switching on the verbal fluency performances by MS and H D patients.
Experiment 2 MS is a demyelinating disease of the central nervous system (CNS). Although lesions can occur anywhere in the CNS, there is a preponderance of damage in the periventricular white matter and the long sensory and motor tracts [19]. Patients with MS consistently produce fewer correct words on both phonemic and semantic fluency tests [2, 4-6, 8]. On tests of confrontation naming, some MS patients exhibit mild impairment, but more serious aphasic disturbances are quite rare [2]. The most common explanation for the verbal fluency deficits in MS is that they arise from difficulties in retrieving information from lexical and semantic memory [37] which are compounded by slowed information processing and dysarthria [2]. If these ideas are correct and average cluster size measures the integrity of lexical and semantic
299
memory, then the reduction in the output of correct words by MS patients should be accompanied by reduction in the number of switches, but the average cluster size should not be affected. Experiment 2 provided a test of these predictions.
Method Subjects. The participants were 133 patients (36 men, 97 women) with clinically definite MS [36] and 63 (20 men, 43 women) healthy control subjects. All were literate and spoke English as their first language. Subjects in both groups met the exclusionary criteria described in Experiment 1. Forty-eight patients and 38 controls were from North Dakota and 76 patients and 25 controls were from Oklahoma. Preliminary analyses revealed no significant differences associated with gender or region for patients or controls so the data were pooled across these variables. There is no simple examination for identifying dementia in MS. Very few MS patients score below the range of normals [2, 3] on tests like the MMSE [20], but a significant proportion of MS patients score below the 5th percentile of controls on three or more cognitive domains, one operational definition of dementia [16]. Impairment on the BNT is a good predictor of this sort of broad cognitive impairment for MS patients [3]. Accordingly, the MS patients were divided into two groups based on their BNT scores: <50 (N = 26) and 50 or higher (N = 107). The former group scored more than 2 S.D. below the mean of controls on the BNT and scored below the 5th percentile of controls on an average of 4.5 out of the 7 cognitive domains (vocabulary, attention, memory, information processing speed, abstraction problem solving, naming, visuo-spatial perception) studied by Beattv et al. [8]. Procedure. As part of a larger battery of neuropsychological tests lhat lasted approximately 2 h, all subjects received the BNT [24], the phonemic fluency test (FAS) and the Animals test. These tests were administered and scored as described in Experiment 1. Data for the fluency tests were previously reported [5, 6] in terms of numbers of correct words generated, but analyses of switching and clustering were not included in the previous reports. All subjects provided written informed consent after a thorough explanation of the procedures which were approved by the local IRB, The M S patients averaged 3.7_ 2.8 on the Ambulation Index (AI) [22], a measure of physical disability that is highly correlated ( r = 0.96) [4] with Kurtzke's [25] Expanded Disability Status Scale (EDSS). Because the interrater reliabilities for both switching and clustering measures were consistently high for all four groups in Experiment 1, only 30 protocols, selected at random, from each of the MS and control groups were scored independently by two raters in Experiment 2. The sizes and patterns of correlations were similar for both groups; the average correlation was 0.969 (Range: 0.922-0.997).
Results and Discussion Table 3 summarizes performance by both groups of MS patients and control subjects. Here it can be seen that the groups were comparable in average age and education, but, as expected, the MS patients produced fewer correct words on both the phonemic (FAS) and the
A. I. Triister et &/Verbal fluency
300
Table 3. Mean (S.D.) performance by MS patients and controls in experiment 2
Age Education BNT FAS Correct FAS Cluster Size FAS Switches FAS Switches/Word Animals Correct Animals Cluster Size Animals Switches Animal Switches/Word
Controls
MS Patients BNT > 50
MS Patients BNT < 50
46.5(16.1) 13.9 (2.3) 55.8 ( 3.3) 43.0(13.7) 0.35 (0.15) 32.3 ( 9.9 ) 0.77(0.18) 20.6 (5.0) 1.37 (0.57) 9.0( 3.7) 0.43 (0.14)
42.7 ( 9.3) 14.4 (2.2) 56.2 (2.8) 35.3 (12.7)” 0.34 (0.24) 25.4 (9.0)” 0.74(0.15) 17.9(5.5)” 1.36 (0.76) 7.7(3.5) 0.43(0.13)
43.6 (12.6) 13.3 (2.0) 44.2 (6.0)‘.b 21.6( 8.6)“,h 0.42 (0.3 1) 16.7 (7. 9)a.b 0.82 (0.29) 13.8 (7.7)“,h 1.27 (0.72) 5.8 (3.7)“,h 0.46(0.15)
a Significantly different from controls, P < 0.05 ’ Significantly different from MS patients with BNT 2 50, P < 0.05
193) > 14.21, were accompanied by reductions in the number of switching responses (Fs (2, 193) > 7.29, Ps < O.OOl),but there was no difference between groups in the mean cluster size or the number of switching responses per correct word for either fluency task (all Fs < 2.23). Subsequent analyses showed that the MS patients with normal BNT scores produced fewer correct words than controls on both fluency tests and fewer switching responses on the FAS task. Patients with impaired naming differed from controls and patients with normal naming in the number of correct words and switching responses on both tasks. Like the AD and PDD patients, the MS patients, especially those with global cognitive impairment, generated fewer words and made fewer switching responses than controls. Unlike the AD and PDD patients, the MS patients showed normal cluster sizes on both fluency tasks. semantic
(Animals)
tasks
(Fs
(2,
Ps < 0.001). These deficits in word generation
Experiment 3 HD is a hereditary disease of the CNS which is characterized by uncontrollable choreiform movements and progressive intellectual deterioration. The initial target of the disease process is the neostriatum, particularly the head of the caudate nucleus. As the disease progresses, cortical regions, especially those areas with extensive and reciprocal connections with the caudate also deteriorate [40]. For this reason, HD in its early stages is often considered a prototype of subcortical dementia [15]. Impairments on verbal fluency tasks by HD patients appear early in the course of the disease and worsen with disease progression [1, Ill. In most studies, deficits of comparable magnitude on phonemic and semantic fluency tasks have been observed, even in patients whose confrontational naming abilities appear normal [ 111.For this reason, fluency deficits in HD are typically attributed
to retrieval failure rather than to degradation of the lexical or semantic memory stores. If this explanation is correct and average cluster size is a measure of the integrity of semantic memory, then reduced word output should be associated with fewer switching responses, but the average size of both phonemic and semantic clusters should not be altered. Experiment 3 tested this hypothesis.
Method Subjects. The 24 HD patients (13 men, 11 women) were literate, spoke English as their first language and were residing at home in the San Diego, CA area at the time of testing. They were diagnosed by a board certified neurologist on the basis of a positive family history and the presence of choreiform movements. Exclusionary criteria were the same as in Experiments 1 and 2. The patients averaged 117.9 on the DRS (Range: 68-140). Other clinical and demographic data are reported in Table 4. The HD patients received the DRS, the BNT, the FAS and the Animals fluency tests as part of a larger battery that lasted about 3 h. The performance of the HD patients was compared to that of the control subjects from Experiment 2. The HD patients and their guardians provided written informed consent after a thorough explanation of the study procedures which were approved by the local IRB. Interrater reliabilities (determined as in Experiment 1) for the HD patients on the clustering and switching measures averaged 0.973 (Range: 0.9424.999).
Results and Discussion
As seen in Table 4, the HD and control groups were comparable in age and education. Although the proportion of females in the control group was higher than in the HD group (68 vs 46%) this difference was not significant (x2 (1) = 2.82). Comparison of scores on the BNT and the numbers of correct words and switching responses on the phonemic
A. 1. Tr6ster et al./Verbal fluency
301
Table 4. Mean (S.D.) performance by HD patients and controls in experiment 3
Age Education BNT FAS Correct FAS Cluster Size FAS Switches FAS Switches/Word Animals Correct Animals Cluster Size Animals Switches Animals Switches/Word
HD Patients
Controls
F(1, 85) for groups
50.2(12.9) 14.0 (2.4) 45.7 (9.5) 12.0 (7.0) 0.17 (0.21 ) 10.0 (5.6) 0.98 (0.59) 7.5 (3.6) 0.92 (1.1) 3.9 (2.8) 0.48 (0.24)
46.5(16.1) 13.9 (2.3) 55.8 (3.3) 43.0 ( 1 3 . 7 ) 0.35 (0. l 5) 32.3 (9.9) 0.77 (0.18) 20.6 (5.0) 1.37 (0.57) 9.0 (3.7) 0.43 (0.14)
1.2l 0.06 24.60*** 187.14"** 11.90"* 59.14"** 0.29 176.09"** 3.61 45.43*** 0.89
**P < 0.01, ***P < 0.001
and semantic fluency tasks revealed highly significant differences between groups. However, the number of switching responses per word was not significantly different on either task. The average cluster size was significantly smaller for H D patients than for controls on the phonemic fluency task. On the semantic fluency task, a similar effect was observed but the trend was only marginally significant (P < 0.10) because of the highly variable performances among H D patients. Inspection of individual scores showed that three H D patients had mean semantic cluster sizes of 3.0 or greater while the remaining patients attained an average semantic cluster size of 0.52. Because parametric statistics were inappropriate, the data were reanalysed using non-parametric techniques. The median cluster size (1.27) was determined for the control group and the frequency of subjects above, or equal to or below that score was determined for each group. Significantly fewer H D patients than control subjects attained average cluster sizes above the control median (Z2 (1) = 7.62, P < 0.01). For the H D group, the correlations of dementia severity (DRS Total) with mean cluster size were 0.26 for phonemic clusters and 0.03 for semantic clusters. Neither correlation approached significance. By contrast, DRS Total was positively related to the numbers of words correctly generated and the numbers of switching responses on both fluency tasks (rs > 0.51, Ps < 0.05). An analysis of the possible use of alphabetic strategies, conducted as in Experiment 1, indicated that one patient appeared to use such a strategy on the phonemic task and 7 patients may have used an alphabetic strategy on the semantic task. In 6 of these cases, the alphabetic strategy was also part of a semantic cluster. Like the AD and PDD patients tested in Experiment l, the H D patients showed smaller average cluster sizes on both phonemic and semantic fluency tasks. The performance of the H D patients on the clustering measures contrasts sharply with that of the MS patients studied in Experiment 2. Despite slightly greater impairments in naming, the low BNT-MS patients had significantly
larger phonemic clusters (0.42 vs 0.17, t(48)= 3.16, P < 0 . 0 1 ) and semantic clusters (1.27 vs 0.92, Z2 (l) = 5.99, P < 0.05) than the H D patients.
General Discussion Table 5 provides a summary of the major results of the three experiments. The present findings replicate earlier reports [9, 42] that the average size of both phonemic and semantic clusters of words generated on fluency tasks by AD patients is reduced compared to the cluster sizes of age- and education-matched controls, but in contrast to an earlier report [42], when the severity of dementia for PD and AD patients was carefully equated, both groups exhibited comparable reductions in the size of their semantic clusters. This result underscores the importance of controlling dementia severity in comparisons of cortical and subcortical dementia [10]. The entirely normal performance of the P D N D patients indicates that the changes in semantic cluster size, in particular, and more generally on other measures of verbal fluency, are confined to the dementia of PD and are not characteristic of this disease in general. Results of another study are consistent with this inference [38]. All of these conclusions must be tempered by the impossibility of ensuring, without autopsy, that any individual demented PD patient did not also suffer from AD. Because of differences in the patterns of cognitive impairment of the PDD and AD patients described earlier, co-morbidity cannot provide a unitary explanation of the similarity of performances by the PDD and AD patients on the semantic fluency task. The H D patients had higher scores on the DRS (P < 0.05) and the BNT (P < 0.01) than the AD patients. Yet, the two groups showed quantitatively similar deficits in word generation, switching and in the sizes of their phonemic and semantic clusters. Because the H D patients averaged almost 20 years younger than the AD patients, it is not likely that many of the H D patients also suffered from AD. Therefore, if reductions in cluster size are taken
A. I. Tr6ster et al./Verbal fluency
302
Table 5. Summary of results by patient group and dementia status Semantic Fluency
Phonemic Fluency
AD,D ~ HD,D ~ PD,D" PD, ND . MS, D b MS, ND b
Correct
Switches
Cluster Size
Correct
Switches
Cluster Size
~ ,!, ,L
~, $ $
,l, ,[, --
+ ,L ,l,
,L ,[, ,L
,L ,[, ,L
.
.
$
.
.
,L
.
.
.
E
,[,
Dementia status based on diagnosis and DRS score. bDementia status inferred from BNT score.
as evidence of degradation of lexical and semantic memory as Martin et al. suggest occurs in AD, then the erosion of semantic knowledge evidently can occur in subcortical as well as in cortical dementia. If this is so, then the widely reported distinction that the memory deficits in subcortical dementias like H D and PD arise mainly from defective retrieval mechanisms while the amnesia of AD reflects principally storage defects [15, 39] requires reexamination, Alternatively, average cluster size might simply be regarded as an index of the ease of access to lexical or semantic memory. From this perspective, the similar reductions in average size of semantic clusters for the AD, PD and H D groups present no challenge to conventional theories of semantic memory in dementia. This conclusion is probably the most defensible because there is now ample evidence that AD patients exhibit defective retrieval of information from semantic memory as well as erosion of memory stores [18, 3 l]. Differences in cognitive profiles between patients with subcortical and cortical dementia are often most apparent in the early stages of the diseases. Because the average DRS scores of the present AD and PDD groups were in the moderately demented range, it might be argued that differences in cluster size, particularly for the semantic clusters, would have been detected if mildly demented patients had been studied. Two observations argue against this hypothesis. First, significant differences between the PDD and AD groups were observed in the patterns of impairment on the DRS subscales and on the BNT. Second, the H D patients averaged 10.5 points higher than the AD patients on the DRS, yet they showed similar patterns of reduction in the size of phonemic and semantic clusters. Regardless of their scores on the BNT, the MS patients exhibited reduced word generation and switching on both fluency tasks but normal cluster sizes. The finding that the MS patients with low BNT scores and global cognitive impairment showed normal cluster sizes appears to support the idea of Brown and Marsden [10] that the patterns of cognitive impairment in the subcortical dementias are
heterogeneous. Because it was not possible to equate the PDD, H D and MS patients with low BNT scores for severity of dementia, this conclusion must remain tentative. Troyer et al. [43] defined switching as a measure of the ability to shift from one phonemic or semantic subcategory to another and conceptualized switching as an executive function related to the integrity of the frontal lobes. A serious problem in using the absolute number of switching responses to analyse the verbal fluency performance of brain damaged patients is that the number of switching responses is strictly limited by the total verbal output. Thus, any sizeable reduction in the number of words generated will always be associated with a reduction in switching, a phenomenon observed in all of the patient groups examined in the present study. We attempted to avoid this problem by analysing switching in proportion to total word output. Expressed in this way, there were no significant differences between patients and controls in Experiments 2 and 3. In Experiment 1, the AD and PDD groups had higher relative switching scores than controls or P D N D patients. If this proportional measure is taken as the appropriate metric for switching, then the conclusion would be that the switching component of executive function on verbal fluency tasks is intact in AD, HD, PDD and MS. Such a conclusion is not warranted because the proportional measure of switching is also seriously flawed. Given equivalent word output, subjects who employ phonemic and semantic clustering will necessarily make fewer switching responses. The inherent negative relationship between switching and clustering probably explains why the relative switching scores of the AD and PDD patients were significantly elevated. The most salient fact about the performance of dementia and other brain-damaged patients on verbal fluency tasks is that total output is severely reduced. Analyses of switching and clustering do not address this issue directly and therefore at best can provide only a partial account of verbal fluency performance. What is needed is a new approach that confronts the
A. I. Tr~3steret al./Verbal fluency phenomenon of reduced behavioural output more directly. Because tests of verbal fluency place a premium on rapid responding which is generally reduced in dementia and many other types of brain damage, including measures of information processing speed in future models of verbal fluency should be considered. The absence of a consistent relationship between dementia severity and the average size of the semantic clusters remains puzzling. Regardless of whether one assumes that cluster size is a measure of the integrity of semantic memory stores or only of access to these stores, a positive correlation between mental status and cluster size would be predicted. Considering the AD, PD, and HD groups from the present study and the AD group from our earlier study [9], the average r for the correlation of dementia severity and average semantic cluster size was 0.17 (Range = -0.04-0.52). The problem cannot be attributed to a restriction in the range of scores on the DRS because the magnitude of the correlation for the HD patients, the group with the largest range (68-140), was 0.03 while the size of the correlation for the PDD patients, the group with the smallest range (75 122) was 0.52. Because each of these correlational analyses had relatively low power, we pooled the AD, PDD and HD groups and compared average semantic cluster size for the 28 patients with DRS scores above 120 and the 56 patients with scores of 120 and below. The small difference (0.95 vs 0.83) did not approach significance (F < 1). Finally, the failures to obtain the expected relationships cannot have arisen because of a restriction in the range on the cluster size variable. The ranges of scores for the three dementia groups on the measure of average cluster size were larger than those for controls. A more likely explanation is that the low output of the HD, AD, and PDD patients on the Animals fluency task results in estimates of average cluster size that are unstable. The bimodal distribution of semantic cluster size stores for the HD patients probably is one illustration of this phenomenon. Use of a semantic fluency task on which the dementia patients produced a greater number of words [7] could remedy this problem.
Acknowledgments--Supported by Grant HR3-005 from the Oklahoma Centre for Advancement of Science and Technology and by Grants AG-10182 to the University of Kansas Medical Centre and AG-05131 to UCSD from the National Institute on Ageing.
4.
5.
6,
7.
8.
9.
10.
1 I.
12. 13.
14.
15. 16.
References
17. 1. Barr, A. and Brandt, J., Word-list generation deficits in dementia. Journal of Clinical and Experimental Neuropsvchology, 1996, 18, 810-822. 2. Beatty, W. W., Cognitive and emotional disturbances in multiple sclerosis. Neurologic Clinics, 1983, 11, 189-204. 3. Beatty~ W. W. and Goodkin, D. E., Screening for
18. 19.
303
cognitive impairment in multiple sclerosis: An evaluation of the Mini-Mental State Exam. Archives of Neurology, 1990, 47, 297 301. Beatty, W. W., Goodkin, D. E., Hertsgaard, D. and Monson, N., Clinical and demographic predictors of cognitive performance in multiple sclerosis: Do diagnostic type, disease duration, and disability matter? Archit'es o["Neurology, 1990, 47, 305-309. Beatty, W. W., Goodkin, D. E., Monson, N. and Beatty, P. A, Cognitive disturbances in patients with relapsing-remitting multiple sclerosis. Archives o/ Neurology, 1989, 46, 1113--1119. Beatty, W. W., Goodkin, D. E., Monson, N., Beatty, P. A. and Hertsgaard, D., Anterograde and retrograde amnesia in patients with chronic progressive multiple sclerosis. Archiees of Neurology, 1988, 45, 611 -619, Beatty, W. W., Monson, N. and Goodkin, D. E., Access to semantic memory in Parkinson's disease and multiple sclerosis. Journal of Geriatric PLtvchiatO' and Neurology, 2, 159-168. Beatty, W. W., Paul, R. H., Wilbanks, S. L., Haines, K. A., Blanco, C. R. and Goodkin, D. E., Identifying multiple sclerosis patients with mild or global cognitive impairment using the Screening Examination for Cognitive Impairment (SEFCI). Neurology, 1995, 45, 718 723. Beatty, W. W., Testa, J. A., English, S. and Winn, P, A. Influences of clustering and switching on the verbal fluency performance of patients with Alzheimer's disease. Ageing, Neuropsychology and Cognition, in press. Brown, R. G. and Marsden, C. D., Subcortical dementia: The neuropsychological evidence. Neuroscience, 1998, 25, 363 387. Butters, N., Granholm, E. L., Salmon, D. P., Grant, I. and Wolfe, J., Episodic and semantic memory: A comparison of amnesic and demented patients. Journal ~[ Clinical and Experimental Neuropsychology, 1987, 9, 479-497. Chertow, H. and Bub, D. Semantic memory loss in dementia of the Alzheimer type: What do the various measures measure? Brain, 1980, 113, 397 417. Coslett, H. B., Bowers, D., Verfaellie, M. and Hellman, K. M., Frontal verbal amnesia: Phonological amnesia. Archit'es of Neurology, 1991, 48, 949-955. Cox, D. M., Bayles, K. A. and Trosset, M. W., Category and attribute knowledge deterioration in Alzhelmer's disease. Brain and Language, 1996, 52, 536 550. Cummings, J. L., Introduction. Subcortical Dementia. ed. J. L. Cummings. Oxford University Press, New York, 1990, pp 3-d6. Cummings, J. L. and Benson, D. F., Dementia: A Clinical Approach. Butterworths, Boston, 1983. Cummings, J. L. and Benson, D. F., Subcortical dementia: Review of an emerging concept. Archives (~['Neurolog)', 1984, 41, 871 879. Dopkins, S., Kovner, R., Rich, J, B. and Brandt, J., Access to information about famous individuals in Alzheimer's disease. Cortex, 1997, 33, 333-339. Ebers, G. C., Multiple sclerosis and other demyelinating diseases. In Diseases ~f the Nert,ous System:
304
20.
21.
22.
23.
24. 25. 26.
27.
28.
29. 30.
31.
A.I. Tr6ster et al./Verbal fluency
Clinical Neurobiology, eds A. K. Asbury, G. M. McKhann and W. I. McDonald. Saunders, Philadelphia, 1986, pp 1268-1281. Folstein, M. F., Folstein, S. E. and McHugh, P. R. "Mini-Mental State";: A practical method for grading the cognitive state of patients for the clinician, Journal of Psychiatric Research, 1975, 12, 189198. Gruenewald, P. J. and Lockhead, G. R., The free recall of category examples. Journal of Experimental Psychology: Human Learning and Memory, 1980, 6, 225-240. Hauser, S. L., Dawson, D. M., Lehrich, J. R., Beak M. F., Kevy, S. V., Propper, R. D., Mills, J. A. and Weiner, H. L., Intensive immunosuppression in progressive multiple sclerosis. New England Journal of Medicine, 1983,308, 173-180. Hodges, J. R., Salmon, D. P. and Butters, N., Semantic memory impairment in Alzheimer's disease: Failure of access or degraded knowledge? Neuropsychologia, 1992, 30, 301-314. Kaplan, N., Goodglass, H. and Weintraub, S., Boston Naming Test. Lea and Febiger, Philadelphia, 1983. Kurtzke, J. F., Rating neurological impairment in multiple sclerosis: An expanded disability status scale (EDSS). Neurolo#y, 1983, 33, 1444-1452. McKhann, G., Drachman, D., Folstein, M., Katzinan, R., Price, D. and Stadlan, E. M., Clinical diagnosis of Alzheimer's disease: Report of the NINCDSADRDA Work Group under the auspices of the Department of Health and Human Services Task Force on Alzheimer's disease. Neurology, 1984, 34, 939-944. Martin, A., Representations of semantic and spatial knowledge in Alzheimer's patients: Implications for models of preserved learning in amnesia. Journal of Clinical and Experimental Neuropsychology, 1987, 9, 191-224. Martin, A. and Fedio, P., Word production and comprehension in Alzheimer's disease: The breakdown of semantic knowledge. Brain and Language, 1983, 19, 124-141. Mattis. S., Dementia Rating Scale. Psychological Assessment Resources, Odessa, FL, 1988. Monsch, A. U., Bondi, M. W., Butters, N., Salmon, D. P., Katztnan, R. and Thai, L. J., Comparisons of verbal fluency tasks in the detection of dementia of the Alzheimer type. Archives of Neurology, 1992,349, 1253-1258. Nebes, R. D., Semantic memory dysfunction in Alzheimer's disease: Disruption of semantic knowledge or information-processing limitations? Neuropsychology of Memory. Second Edition. eds. N. Butters and L. R. Squire. Guilford Press, New York, 1992, pp 233 240.
32. Newcombe, F., Missile Wounds of the Brain. Oxford University Press, London, 1969. 33. Ober, B. A., Dronkers, N. F., Koss, E., Delis, D. C. and Friedland, R. P., Retrieval from semantic memory in Alzheimer-type dementia. Journal of Clinical and Experimental Neuropsychology, 1986, 8, 75--92. 34. Paolo, A, N., TrOster, A. I., Glatt, S. A., Hubble, I. P., and Koller, W. C., Differentiation of the dementias of Alzheimer's and Parkinson's disease with the Dementia Rating Scale. Journal of Geriatric Psychiatry and Neurology, 1995, 8, 184-188. 35. Perret, E., The left frontal lobe of man and the suppression of habitual responses in verbal categorical behaviour. Neuropschologia, 1974, 12, 323-330. 36. Poser, C, M., Paty, D. W., Scheinberg, L., McDonald, T., Davis, F. A., Ebers, G. C., Johnson, K. P., Sibley, W. A., Silberberg, D. H. and Tourtellote, W. W,, New diagnostic criteria for multiple sclerosis: Guidelines for research protocols. Annals of Neurology, 1983, 13, 227-231. 37. Rao, S. M., Leo, G. J. and St. Aubin-Faubert, P. On the nature of memory disturbance in multiple sclerosis. Journal of Clinical and Experimental Neuropsychology, 1989, 11, 699-1712. 38. Raskin, S. H., Sliwinski, M. and Borod, J. Clustering strategies on tasks of verbal fluency in Parkinson's disease. Neuropsychologia, 1992, 30, 95-99. 39. Salmon, D. P., Heindel, W. C. and Butters, N., Semantic memory, priming, and skill learning in Alzhelmet's disease. In Memo O, Functioning in Dementia. ed. L. B/ickman. Amsterdam, North Holland, 1992, pp 99-118, 40. Shoulson, I. Huntington's disease. In Diseases o/'the Nervous System: Clinical Neurobiology II, eds A. K. Asbury, G. M. McKhann and W. I. McDonald. Saunders, Philadelphia, 1986, pp 1259-1267. 41. Tr6ster, A. I., Salmon, D. P., McCuUough, D. and Butters, N., A comparison of the category fluency deficits associated with Alzheimer's and Huntington's disease. Brain and Language, 1989, 37, 500513. 42. Troyer, A. K. and Moscovitch, M., Clustering and switching on verbal fluency tests: Evidence from healthy controls and patients with Alzheimer's and Parkinson's Disease. Journal of the International Neuropsychological Society, 1996, 2, 11. 43. Troyer, A. K., Moscovitch, M. and Winocur, G., Clustering and switching as two components of verbal fluency: Evidence from younger and older healthy adults. Neuropsychology, 1997, 11, 138-146. 44. Wixted, J. T. and Rohrer, D., Analysing the dynamics of free recall: An integrative review of the empirical literature, Psychonomic Bulletin & Rev'iew, 1994, 1, 89-106.