More MS news articles for April 2002

Subjective Complaints, Verbal Fluency, and Working Memory in Mild Multiple Sclerosis

Appl Neuropsychol 8(4):204-210, 2002.
Karyn Matotek, Michael M. Saling, Department of Psychology, University of Melbourne, Melbourne, Australia; Peter Gates, Leslie Sedal, Department of Neuropsychology, St. Vincent's Hospital, Melbourne, Australia


Patients with mild multiple sclerosis (MS) regularly report subjective complaints characterized by generalized cognitive inefficiency. A feature of these complaints is reduced verbal fluency, for example, losing the thread of conversation. Mild MS patients and controls were compared on subjective complaints, verbal fluency, and working memory, and the possible role of working memory was investigated. As predicted, subjective difficulties and verbal fluency each correlated with working memory. Subjective difficulties and verbal fluency were also related. Within the control group, the subjective difficulties were associated only with depression. On the other hand, in the MS group, subjective difficulties were unrelated to affective state but were correlated with working memory. For the patients, partialing out depression and anxiety did not influence any of the correlations involving subjective difficulties, verbal fluency, or working memory. These findings support the proposal that working memory impairment underlies subjective complaints of mild MS patients and that verbal dys-fluency is part and parcel of this phenomenon.


In clinical practice, we regularly encounter a group of multiple sclerosis (MS) patients who are mildly physically impaired, relatively young, and able to lead an active lifestyle but who describe cognitive difficulties. These complaints are nonspecific in nature, suggesting reduced cognitive efficiency in daily life, such as, "Sometimes I am forgetful" or "I find it harder to add numbers in my head." These are associated with subjective accounts of dysfluency, for example, "I have mental blocks for names of people or things I know well," "I lose the thread of conversation," and "Sometimes I can't finish a sentence when I'm talking."
Mild MS sufferers generally perform relatively well on standard neuropsychological assessment, apart from mildly reduced performance on tests of long-term memory (Callanan, Logsdail, Ron, & Warrington, 1989; Feinstein, Kartsounis, Miller, Youl, & Ron, 1992; Klonoff, Clark, Oger, Paty, & Li, 1991; Kujala, Portin, & Ruutiainen, 1996). It is unlikely, therefore, that verbal-conceptual processing or immediate memory impairments account for subjective complaints of verbal dysfluency.

In fact, the literature suggests that the nature of cognitive dysfunction in mild MS is difficult to characterize (Dujardin, Donze, & Hautecoeur, 1998; Grant, McDonald, Trimble, Smith, & Reed, 1984; Grigsby, Ayarbe, Kravcisin, & Busenbark, 1994; Van den Burg, Van Zomeren, Menderhoud, Prange, & Meijer, 1987). One currently emerging view is that working memory deficits account for the early cognitive deterioration in this group (Foong et al., 1997; Grigsby et al., 1994; Litvan et al., 1988; Paul, Beatty, Schneider, Blanco, & Hames, 1998; Pelosi, Geesken, Holly, Hayward, & Blumhardt, 1997; Rao et al., 1993; Ruchkin et al., 1994).

This is interesting in the light of recent research that explicates the role of working memory in important cognitive functions, including verbal fluency (Daneman, 1991). Speaking involves a highly complex and skillful coordination of processing and storage requirements. Speakers must plan what to say and temporarily store the plans until ready to execute them as words, phrases, and sentences. At any instant, individuals may be planning what to say next while concurrently executing what was planned moments earlier. Daneman (1991) argued that working memory capacity is an important source of individual differences in verbal fluency. She found that a measure designed to tax the processing and storage functions of working memory during sentence production (the Speaking Span Test) correlated with measures of verbal fluency (e.g., number of words spoken per minute).

Given the co-occurrence in the clinical setting of complaints of generalized cognitive inefficiencies and subjective language dysfluency, it is possible that they emanate from a common root, possibly a dysfunction in working memory. Put another way, language dysfluencies may represent a marker of an underlying working memory problem in mild MS patients with generalized subjective complaints.

This study aims to investigate the relationship between the subjective complaints that we see in the clinical setting, verbal fluency, and working memory to elucidate the nature of the subtle cognitive deficits described by mild MS patients. In this study, tests of working memory and narrative fluency and a subjective difficulties questionnaire were administered to MS patients and controls matched according to age and education. It was predicted that mild MS patients would perform worse than controls on measures of subjective complaints, verbal fluency, and working memory. It was hypothesized that participants with a large number of subjective general cognitive complaints would demonstrate a worse performance on a test of working memory and verbal fluency than those without such complaints. Following Daneman (1991), it was also envisaged that there would be a positive relationship between a measure of verbal fluency and a test of working memory (the Speaking Span Test). The Narrative Fluency Test offers an opportunity to obtain qualitative information about the types of errors committed during verbal production. To this end, transcripts obtained during the Narrative Fluency Test were analyzed for the number of hesitations and various types of speech errors. Hesitations and other types of speech errors are indicative of planning and monitoring errors, which reflect limitations in working memory capacity (Daneman, 1991; Levelt, 1989). It was expected that patients would exhibit more errors than would controls.



Participants were 39 patients with a diagnosis of definite MS according to the criteria of Poser et al. (1983). All participants were rated as suffering from mild MS as determined by a score of less than 5 on the Kurtzke Expanded Disability Status Scale (Kurtzke, 1983). Experienced neurologists carried out these assessments. This was a community-based sample. They were included in the study if they demonstrated no visual impediment to reading and no evidence of dysarthria on neurological examination and on self-report. All were involved with home duties, work, or study. All patients were required to be proficient English speakers, with no history of any neurological disorder other than MS or alcohol or drug abuse. With the exception of one patient, none of the participants had a history of psychiatric disorder. This patient had been successfully treated for depression 2 years previously and reported that she was not suffering from depression at the time of the assessment.

There were 11 men and 28 women, reflecting the known preponderance of women in MS populations. There were 22 patients with relapsing/remitting MS, 8 with benign MS, and 9 with primary progressive dis-ease. Thirty-six patients were in remission at the time of assessment. The average time since symptom onset was 9.4 years, and the average number of years since diagnosis was 3.6. All patients had brain lesions typical of mild MS on MRI as judged by a neuroradiologist. Forty controls were selected from respondents to newspaper advertisements. Controls were excluded from the study if they had a history of alcohol or drug abuse, psychiatric disorder, or any neurological disorder. None of the controls had undergone neuropsychological assessment in the last year, and none exhibited any visual impediment to reading. There were 9 men and 31 women in the control group. Age of patients (M = 39.18, SD = 10.38) and controls (M = 36.18, SD = 9.78) did not differ significantly. Although the controls (M = 12.25, SD = 2.99) had 1 more year of education than the patients (M = 11.44, SD = 2.50), the difference did not reach statistical significance, and, as will be described, the two groups were well matched in terms of verbal skills. Consent forms were signed by each participant in the presence of a witness according to requirements set by the St. Vincent's Hospital, Melbourne, ethics committee.

Participants were part of a larger study assessing aspects of neuropsychological function, subjective complaints, and performance of a number of tests of working memory. Patients and controls did not differ on important aspects of cognitive function that are relevant to the study, namely, verbal skills, attention and concentration, simple speed of information processing, and executive function as assessed by the Wisconsin Card Sorting Test-modified version (WCST-m; Nelson, 1976). Separate analyses of covariance were conducted with age and education as covariates. For the Verbal Intelligence Quotient (VIQ) derived from the Wechsler Adult Intelligence Scale-Revised (WAIS-R; Wechsler, 1981), a nonsignificant group main effect was obtained, F = (1,75) 0.18, p = .671. When the groups were compared on the Attention/Concentration Index (ACI) of the Wechsler Memory Scale-Revised (WMS-R; Wechsler, 1987), the analysis yielded a nonsignificant group main effect, F = (1,75) 0.68, p = .412. The measure of speed of processing used was the time taken to perform the Mental Control tests of the WMS-R, and a nonsignificant group main effect was obtained, F = (1,75) 0.60, p = .441. Finally, the groups were compared on the WCST-m. The measures used were the number of perseverative errors and the total number of sets achieved. The means and standard deviations for patients and controls on these measures are shown in Table 1.

Table 1. Mean (Standard Deviation) for VIQ, ACI, WCST-modified, and Mental Speed for Patients and Controls
Quotient/Index Patients Controls
VIQ 105.49 11.78 103.83 11.37
ACI 96.28 14.89 99.75 12.52
WCST-m (sets) 5.34 1.40 5.50 1.11
WCST-m (persev) 3.63 5.16 3.55 4.67
Mental Speed 33.83 15.67 30.11 11.37

VIQ Verbal Intelligence Quotent; ACI Attention/ Concentration Index; WCST-m (sets) number of sets obtained. WCST-m (persev) number of perseverative errors.


Verbal fluency. To measure verbal fluency, we utilized a storytelling procedure adapted from Dane-man (1991). Participants were asked to speak for 1 min about a drawing entitled "How the tenant of the top flat can enjoy all the amenities of a back garden" (Robinson, 1934). The following instructions were provided verbally:

I am going to show you a drawing. I would like you to familiarize yourself with it. I am then going to ask you to speak for one minute about what you see. You might like to describe what you see, comment on the drawing or make up a story about the drawing. I will tell you when to start and when to stop.

Responses were tape recorded and transcribed verbatim. The transcripts were scored according to number of words produced during the first minute of speaking time.

To obtain the qualitative data, the following features were recorded according to guidelines described by Maclay and Osgood (1959):

  1. The number of filled pauses, for example, "um" or "ah.
  2. "The number of repetitions of a word or letter, for example, "The [d]dog is... [is] barking."
  3. The number of false starts, for example, "The man [has]... is sitting in the comfortable chair."
In addition, it was noted during pilot testing that some MS patients failed to finish sentences before moving on to the next sentence, for example, "And there appear to be people admiring him for his.... Or perhaps they're even worried about his particular situation." The number of unfinished sentences was also recorded. To determine the interrater reliability, a randomly selected sample of 12 tapes (15% of the total sample) was transcribed, analyzed in terms of speech errors, and scored by a colleague blind to the group membership of the participants.

Subjective difficulties questionnaire. A questionnaire was devised to quantify subjective complaints typically offered by patients in clinical settings. The items were chosen from accumulated clinical protocols obtained from patients with MS over a number of years. The final scale consists of 11 items phrased as questions requiring a yes or no response. In the interests of minimizing response bias, half the questions are positive (yes indicates a difficulty), and the remaining are negative (no indicates a difficulty). Positive and negative questions were randomly ordered. Average internal consistency, as determined by Cronbach's alpha in this sample, was acceptable (.895). Participants were instructed as follows: "Almost everyone experiences problems with concentration and memory at times. Please answer the following questions with a yes or no depending on whether you have noticed any recent changes in your abilities."

Working memory. The Speaking Span Test was adapted from Daneman (1991). Participants were required to silently read a set of words displayed individually at the rate of one per second on a computer screen. At the end of each set, they were asked to use each word to generate aloud a separate sentence. There were three sets of words at each level, where the level could be two, three, four, five, or six words.

The test was administered but not scored by computer. Participants were given two sets of practice items. They were instructed that they could call up the next set of words when they were ready by pressing the button marked "next." The examiner recorded the total number of correct responses.

Measures of anxiety and depression. Affect can influence cognitive performance in general (Coughlan & Hollows, 1984) and working memory in particular (Channon, Baker, & Robertson, 1993; Darke, 1988; Rapee, 1993). A recent study found that depressed MS patients performed worse than normal controls and nondepressed MS patients on a test of working memory and other capacity-demanding tasks (Arnett, Higginson, Voss, Bender, et al., 1999; Arnett, Higginson, Voss, Wright, et al., 1999).

In this study, the level of anxiety and depression were assessed using the Beck Depression Inventory (BDI; Beck & Steer, 1987) and the State scale of the State Trait Anxiety Inventory (STAI-State; Spielberger, 1983).


Half the patients were given the WAIS-R, WMS-R, and WCST-m tests first, and half were given the questionnaires, the fluency test, and the working memory test first. The timing of test sessions was designed to minimize the effects of fatigue, and patients could request a break at any time within the session.


The MS and control groups were compared using a multivariate analysis of covariance (MANCOVA), with age and education as covariates, and the Subjective Difficulties Questionnaire, the Speaking Span Test, and the measure of verbal fluency as the dependent variables.

This analysis yielded a significant multivariate groups main effect, F = (1, 69) 2.35, p = .006, justifying separate examination of the univariate effects (Table 2). MS patients were worse than controls on the Subjective Difficulties Questionnaire, F = (1, 69) 24.93, p = .001; the Speaking Span Test, F = (1, 69) 14.99, p = .001; and the verbal fluency measure, F = (1, 69) 9.49, p = .003.

Table 2. Means and Standard Deviations for MS Patients and Controls on the Subjective Difficulties Questionnaire, Speaking Span Test, and Verbal Fluency
Test/Index Group p
MS Patients Controls
Subjective Difficulties 5.56 3.92 1.49 2.34 0.001
Speaking Span Test 2.31 0.47 2.92 0.72 0.001
Verbal Fluency 130.00 26.64 153.19 29.59 0.001

As predicted, subjective difficulties correlated with working memory (r = -.34, p <.01) and verbal fluency (r = .27, p <.05), and performance on the Speaking Span Test correlated with verbal fluency (r = .27, p <.05). Within the control group, subjective difficulties were associated only with depression (r = .38, p <.05). Within the MS group, on the other hand, subjective difficulties were unrelated to anxiety (r = .15, p >.05) and depression (r = .20, p >.05) but were correlated with speaking span (r = .33, p <.05). For the patients, partialing out depression and anxiety did not influence any of the correlations involving subjective difficulties, verbal fluency, or working memory.

The Narrative Fluency Test provided qualitative information about the types of errors committed by patients and controls during verbal production. The interrater reliability in terms of filled pauses (r = .95), repetitions (r = .41), false starts (r = .79), and words per minute (r = .94) was acceptable (p = <.01 for all correlations). The small number of incidents of unfinished sentences in this sample precluded analysis of interrater reliability.

To compare the groups, a t test was performed on the average percentage of speech hesitations and errors as a proportion of total words produced (see Table 3). As hypothesized, the patients performed more poorly than controls in terms of false starts, repetitions of a word or letter, and number of unfinished sentences. The finding regarding filled pauses approached significance with MS patients performing worse than controls. There were a relatively small number of unfinished sentences produced. Nevertheless, the incidence of unfinished sentences was nearly eight times greater in the patient group than in the controls.

Table 3. Mean Percentages and Standard Deviations as a Proportion of Total Words Produced and t Values for Patients and Controls on Types of Hesitations and Speech Errors on the Narrative Fluency Test
Type of Hesitation/Speech Error Patients Controls t p*
Filled Pauses 3.45 3.07 2.57 2.15 1.88 0.090
False Starts 1.63 1.68 0.46 0.64 3.63 0.001
Repetitions 1.59 1.50 0.54 0.66 3.51 0.001
Unfinished Sentences 0.23 0.44 0.03 0.13 2.47 0.014

* One-tailed.


Mild MS patients are usually relatively young and physically mobile and actively involved in work and family life. The relatively high demands placed on mild MS patients means that even subtle cognitive problems might impact on their ability to cope. Nevertheless, the cognitive deterioration in mild MS has been difficult to characterize, and this may affect the nature of information and advice given to patients in the clinical setting. For example, it is not always clear to what extent difficulties are related to affective state or organic causes.

In this study, it was demonstrated that even in the mild stages of the disease, MS patients complained of more subjective complaints than the controls. They also had a smaller speaking span than controls and a lower score on verbal fluency. The poorer performance of patients cannot be explained by variation in verbal intelligence, basic attention, speed of information processing, or executive skills as assessed by the WCST-m. The poor performance of the MS group cannot be attributed to MS-related sensorimotor deficits, as verbal tests were administered, and patients were not affected by visual or articulatory changes. Although 3 patients were recovering from an exacerbation at the time of testing and 9 patients had progressive disease, they did not perform as outliers on any of the measures, and repetition of the MANCOVA with these cases excluded did not reduce the significance of the group difference, F(1, 62) = 3.11, p = .001, when the cases with progressive MS were removed and F(1, 66) = 3.16, p = .001, when those in exacerbation were removed from the sample.

The positive relationship between working memory and verbal fluency demonstrated here is in agreement with the previous findings of Daneman (1991) in a normal sample. Post hoc analysis of our data revealed that the problem lies not just with producing fewer words per minute. Rather, patients made about twice as many errors of production, for example, failure to complete a sentence and false starts. This indicates that MS patients may be making more errors than normal at a planning level of discourse production (Garrett, 1982; Holmes, 1988) as a result of reduced working memory capacity (Daneman, 1991; Howell, Saling, Bradley, & Berkovic, 1994; Just & Carpenter, 1992). Other possible reasons for poor performance are unlikely to account for the results given the normal performance of MS patients on tests of verbal-conceptual processing and short-term storage. In addition, patients did not produce more perseverative errors than controls on the WCST-m. In short, these results support the idea that capacity limitations in working memory underlie the verbal fluency problems experienced by MS patients.

As hypothesized, working memory was related to generalized complaints, supporting the notion that working memory dysfunction underlies the subjective cognitive difficulties commonly reported by individuals in the clinical setting. It is interesting to note, however, that the mechanism underlying the complaints of mild MS patients appears to differ from that of the controls. The major determinant of general cognitive inefficiencies in the control group was variation in affective state. For MS patients, on the other hand, affective factors did not account for the findings, raising the question of organic causes, presumably cerebral MS pathology.

This is in contrast to a recent study (Arnett, Higginson, Voss, Bender, et al., 1999) that found that of the three groups tested, namely, depressed MS patients, nondepressed MS patients, and normal controls, only the depressed patients were impaired on a test of working memory. The working memory test employed was different from that of this research, making it difficult to compare the studies. In any case, findings from this investigation are not clear-cut. The study failed, for example, to include neuroimaging data and a depressed control group. Thus, it was not possible to attribute the poor performance of the depressed MS group to psychological factors alone. Consistent with our findings, Arnett and coworkers stated, "Clearly, there are MS patients without current depression who experience significant deficits on working memory" (p. 554).

Our findings help elucidate the nature of the cognitive difficulties experienced by people with mild MS. Knowledge about the role of working memory in everyday problems can benefit patient education and assist the formulation of practical strategies to facilitate performance of work and home duties. Patients might benefit from discussion of ways to control environ-mental demands on working memory capacity during speaking and other tasks involving working memory.

This study also suggests that standard neuropsychological assessment, represented here by the Verbal Scale of the WAIS-R, and the Attention/ Concentration Index of the WMS-R, may be relatively insensitive to the subtle cognitive impairment experienced by patients with mild MS. It remains to be seen whether the Wechsler Adult Intelligence Scale-III (Wechsler, 1997), with its Working Memory Index, improves the sensitivity of assessment in this patient group.

Reprint Address

Requests for reprints should be sent to Karyn Matotek, Green-vale Medical Centre, 1 Greenvale Drive, Victoria, 3059, Australia.

© 2002 Lawrence Erlbaum Associates, Inc