Verbal memory and verbal fluency in adolescents with schizophrenia spectrum disorders


  • Nils Inge Landrø phd,

    Corresponding author
    1. Center for the Study of Human Cognition, Department of Psychology, University of Oslo,
    Search for more papers by this author
  • Torill Ueland phd

    1. Sogn Centre for Child and Adolescent Psychiatry, and
    2. Section for Psychosis Research, Division of Psychiatry, Ullevål University Hospital, Oslo, Norway
    Search for more papers by this author

*Nils Inge Landrø, PhD, Department of Psychology, University of Oslo, Box 1094, Blindern 0317, Oslo, Norway. Email:


Aim:  Although impaired verbal memory and verbal fluency are frequently found in adults with schizophrenia, there has been a paucity of studies investigating adolescents with schizophrenia. Thus, the aim of the present study was to investigate the main subcomponents of verbal memory and verbal fluency in adolescents with schizophrenia spectrum disorders.

Methods:  Verbal learning and memory and verbal fluency was assessed in 21 adolescents with schizophrenia spectrum disorders (mean age, 15.4 years) compared with 28 healthy adolescents (mean age, 15.1 years).

Results:  The patient group performed significantly below healthy controls on measures of learning, delayed recall and on a frequency estimation task. No differences between the groups were found for measures of recognition, retention, implicit memory, or susceptibility to interference. Although they had impaired delayed recall the patients remembered most of what they actually learned. The patient group was impaired on phonological and semantic fluency, but there were no differences between the groups with respect to clustering or switching on the fluency tasks, when controlling for total output. There was no disproportionate impairment in semantic, as compared to phonological fluency, in the patient group.

Conclusions:  Adolescents with schizophrenia spectrum disorders exhibit impairments in verbal learning and verbal fluency, which might have an impact on the individual's everyday functioning.

IMPAIRED VERBAL MEMORY1 is well documented in adults with schizophrenia and occurs over and above normal general intellectual functioning2,3 and in neuroleptic-naïve first-episode patients.4 Memory is a multifaceted function consisting of various subprocesses such as learning, storage, recall and recognition. Based on findings of impaired recall along with intact recognition, it has been suggested that patients with schizophrenia show a retrieval deficit.5 Others have reported impaired memory in schizophrenia under both free recall and under cued recall conditions, which could indicate a storage deficit.6 In general, however, the bulk of evidence indicates a primary deficit in the acquisition or learning of verbal material in adult patients with schizophrenia.7,8 Verbal fluency, frequently referred to as a semantic memory task, is another area consistently found to be impaired in schizophrenia samples.9,10 In two meta-analyses, including studies in which patients with schizophrenia and healthy controls completed both letter (phonemic) and category (semantic) fluency tasks, it was concluded that patients with schizophrenia were differentially deficient on category fluency.11,12 This finding is striking because numerous studies have demonstrated that in healthy controls semantic fluency may be an easier task than letter fluency. Furthermore, fluency performance is multifactorial, and two important components of fluency performance are clustering (i.e. the production of words within semantic or phonemic categories), and switching (i.e. the ability to shift efficiently to a new subcategory).13 Adult patients with schizophrenia generate significantly fewer total words, cluster-related words and switches than healthy controls on both phonological and category fluency tasks.14

Only a few neuropsychological studies of adolescents with schizophrenia exist,15–18 and none of these undertook a systematic analysis of the various subcomponents of verbal memory. Impaired phonological fluency has been documented in one previous study of adolescents with schizophrenia,15 while, to our knowledge, no previous study of adolescents has included a measure of semantic fluency. Adolescents are an interesting group to investigate for several reasons. Research indicates that earlier age of onset is related to a poorer prognosis and outcome,19 and some studies also indicate poorer cognitive functioning.20 Further, progressive volumetric reductions of hippocampal regions have been documented in adolescents with schizophrenia,21 and impairments in verbal learning and memory are associated with earlier onset of illness.22 A significantly increased prevalence of neurological soft signs has also been reported in adolescents with first-episode psychosis as compared to healthy controls.23 Finally, a closer description of deficient and intact subcomponents of verbal memory and verbal fluency could potentially contribute to the development of more effective rehabilitation interventions.

The first aim of the present study was to assess verbal learning, delayed recall and recognition in adolescents with schizophrenia. Specifically, we hypothezised that the patients would exhibit impaired learning ability. To further investigate verbal memory functioning in schizophrenia a paradigm including interference, and subtasks tapping an automatic aspect of verbal memory, as well as implicit memory, were included. The second aim was to examine verbal phonemic and semantic fluency and their underlying processes by estimating clustering and switching. We hypothezised that there would be a disproportionate impairment in semantic compared to phonological fluency in the patient group.



Twenty-one adolescents with schizophrenia spectrum disorders and 28 healthy controls participated in the present study. The patient sample was recruited from an inpatient ward for psychotic patients at Sogn Centre for Child and Adolescent Psychiatry in Oslo, Norway. Inclusion criteria for entering the study were: age between 12 and 18 years, a diagnosis of a schizophrenia spectrum disorder, IQ > 70 and no evidence of organic brain disease. Diagnoses were based on the Structured Clinical Interview for DSM-IV Axis I disorders24 and patient case records. Diagnostic assessments were performed by an experienced clinical psychologist. The subdiagnoses of the patient sample were: disorganized, n = 5; paranoid, n = 6; undifferentiated, n = 5; schizoaffective, n = 3; schizotypal personality disorder (SPD), n = 2. For the two patients with SPD the Structured Clinical Interview for DSM-IV Axis II disorders was used for diagnostic assessment.25 Six participants were not using medication, while the remaining 15 were receiving antipsychotic medication. Seven patients were using atypical (olanzapine and risperidone) and five were using typical antipsychotic medication only. One patient was using both types. One patient was using an atypical together with lithium, and one, a typical together with methylphenidate. Six patients had never previously been hospitalized for psychotic symptoms, 12 patients had been hospitalized once and three had been hospitalized twice. Psychiatric symptoms were assessed using the 24-item Brief Psychiatric Rating Scale (BPRS)26 and psychosocial functioning using the Global Assessment Scale (GAS)27 by two independent raters. Inter-rater reliability intraclass corrections (ICC)28 was 0.92 for the BPRS total score and 0.94 for the GAS. Full-scale IQ was measured using the Wechsler Intelligence Scale for Children–Revised (WISC-R)29 or the Wechsler Adult Intelligence Scale–Revised (WAIS-R)30 for subjects aged >16 years. The mean full-scale IQ in the patient group was 91 ± 17 (range, 70–127). Scores from the subtests Similarities and Picture Completion are listed in Table 1 because they were the only IQ estimates obtained from the control group. Three patients did not have Norwegian as their first language, but all had been living in Norway for at least 5 years and spoke good Norwegian.

Table 1.  Subject characteristics
 Schizophrenia group
n = 21
Control group
n = 28
  • *

    P < 0.0001;

  • **

    P = 0.004.

  • WISC-R, Wechsler Intelligence Scale for Children-Revised; WAIS-R, Wechsler Adult Intelligence Scale-Revised.

Male/female (ratio)13/8 15/13 
Age (years)15.40.815.11.0
Similarities (WISC-R/WAIS-R)*
Picture Completion (WISC-R/WAIS-R)8.72.511.33.2**
Medicated subjects (%)71.4 NA 
Global Assessment Scale30.78.8NA 
Brief Psychiatric Rating Scale54.711.6NA 
No. previous hospitalizations1.00.79NA 

The adolescents in the healthy comparison group were recruited on a voluntary basis from a local school. The controls all attended regular school classes at normal grade levels. They were screened for psychiatric symptoms using the Youth Self-Report (YSR).31 The YSR is a self-report questionnaire for the assessment of psychopathology in children and adolescents. It contains eight problem syndrome scales: withdrawal, somatic complaints, anxious depressed, social problems, thought problems, attention problems, delinquent behaviour, and aggressive behaviour. The total score for all scales was included and participants with raw scores in the clinical range were excluded (n = 3). As an estimate of general intellectual functioning, participants were assessed on the subtests Similarities and Picture Completion using the WISC-R29 or the WAIS-R,30 when aged >16 years. These two subtasks are considered reliable and valid estimates of IQ30 and are highly correlated with verbal IQ and performance IQ, respectively. All participants provided written informed consent. The study was approved by the Regional Committee for Medical Research Ethics and the Norwegian Data Inspectorate.

Demographic, psychometric and clinical characteristics of the two groups are given in Table 1. t-Tests were used to analyze differences between the groups for continuous measures and χ2 tests for categorical variables. There were no significant differences between the groups with respect to age or the female/male ratio. The patient group performed within the normal range on the estimated IQ measures Similarities and Picture Completion, but scored significantly below healthy controls (t(47) = 4.1, P < 0.0001; t(47) = 3.1, P = 0.004, respectively).

Assessment of verbal learning and memory

Verbal learning and memory were assessed using a multi-trial, free-recall paradigm. The task involves the presentation of 12 words to be remembered, followed by the immediate recall of as many of the words as possible. On trials 2–8 the subject was reminded only of words that were missed on the preceding trial, but required to again recall all words on the list during the next trial (a selective reminding procedure).32 If all of the words were recalled before the eight trials were completed, this part of the task was terminated. In such cases all of the remaining trials were scored as 12 correct responses. A sum score of all correctly recalled words over the eight trials of list A quantified the rate of learning.

Interference effects were measured by using another 12-word list (list B) once, after completing the learning trials of list A. After recall of this new list, recall from list A was again requested immediately. This measure is called short delay recall. Proactive interference was examined by comparing recall performance on the first trial of list A with the recall performance of list B.

Delayed free recall was evaluated by requesting recall of list A after 45 min of doing other cognitive tasks. This measure is called long delay recall. Retention rate was estimated by calculating the proportion of words retained from list A over the 45-min delay period, as compared to the number of words retained on the trial presented immediately after list B.

Following the delayed free-recall task, a recognition task was administered. The 12 words from list A were interspersed with 48 distracters, the task being to respond ‘yes’ if the word was on list A and ‘no’ if the word was not. Six of the distracters were from list B. The recognition score was defined as the number of hits minus the number false alarms.

A frequency of estimation task, presumably reflecting an automatic memory process,33 was also performed. The basis for this task was disguised in the recognition task. Of the 42 distracters, which were not on list B, some were presented once, others twice, three, four and up to five times. After the recognition task, pairs of words were presented to the subject, who was requested to indicate which of the two words had been presented most frequently during the test. The score was the number correct answers.

Finally, a stem completion memory task, presumably tapping implicit memory, was given. This task consisted of completing word-stems with the 12 words from list A. The subjects were told to fill in the remaining letters making up the first word that came to mind. The score was the number of completed stems matching words from list A.

The stimulus words in the two lists (A and B) were common one- or two-syllable Norwegian words such as elv (river), kniv (knife), hus (house), and tog (train).

Phonological and semantic fluency

For the phonological fluency tasks participants were instructed to generate words beginning with f or a, excluding proper names and variants of the same word (e.g. the same word with suffixes). For the semantic fluency task the participants were instructed to generate names of animals. Each trial lasted 60 s. The total number of words generated, excluding perseverative errors and intrusive errors, were obtained for each fluency test. Then mean cluster size and number of switches were calculated. The scoring rules for clustering and switching were identical to those proposed by Troyer et al.13 In short, phonemic clusters consisted of successively generated words including perseverations and intrusions beginning with at least the same first two letters (e.g. arm and art), differing only by a vowel sound (e.g. sat, soot, sight, and sought), rhyming (such as sand, stand) or homonyms, that is, words with two or more different spellings. On the semantic tasks, clusters were defined as groups of successively generated words belonging to the same semantic subcategory, such as farm animals, pets, animals from different continents, and various zoological categories. Cluster size was counted, beginning with the second word in each cluster, and mean cluster size was calculated for the phonemic and semantic tasks. Phonemic and semantic switches were calculated as the number of transitions between clusters, including single words.

Data analysis

Statistical analyses were performed using the SPSS for Windows Release 9.0 (SPSS, Chicago, IL, USA). Two multivariate analyses of variance (MANOVA) were performed, first with group as the between-group factor, and the main verbal memory measures (verbal learning, delayed recall, recognition, frequency estimation and stem completion) as dependent factors, and second, with the fluency measures as within-group factors. Follow-up univariate analyses of variance (anova) were performed on the various tests.

Interference effects, retention and clustering and switching were analyzed using separate univariate ANOVA. Analyses of clustering and switching were repeated with total output as a covariate (ANCOVA). Associations between the verbal learning and memory measures and the fluency measures were examined on correlational analyses, as was the relationship between neuropsychological performance and severity of symptoms and psychosocial functioning.


Analyses of the verbal learning and verbal memory measures

Means and standard deviations for all test results are given in Table 2. MANOVA indicated an overall group difference over the six verbal memory tasks, with patients scoring lower than controls (Wilks λ, F(5,43) = 4.8, P = 0.001). There was also a significant diagnosis by verbal memory measures interaction (Wilks λ, F(5,43) = 5.7, P = 0.0001). Subsequent anovas indicated significant differences between the groups in verbal learning (F(1,47) = 22.2, P < 0.0001), short delay recall (F(1,47) = 8.6, P = 0.005), long delay recall (F(1,47) = 6.2, P = 0.016), and frequency estimation (F(1,47) = 5.9, P = 0.019). There were no significant differences between the groups on the recognition task or the implicit memory task. A more detailed analysis showed that patients had fewer hits than controls, whereas they did not make more false alarms. Repeated measures ANOVA, however, indicated no statistical significant interaction between group and response type (hits vs false alarms).

Table 2.  Verbal memory and fluency measures
VariableSchizophrenia groupControl group
Verbal memory     
 Verbal learning64.816.481.27.4<0.0001
 Short delay recall8.02.910.02.00.005
 Long delay recall8.
 Frequency estimation9.11.610.21.40.019
 Stem completion4.
Verbal fluency     
 Phonological fluency total15.65.421.35.9<0.001
 Phonological clustering0.
 Phonological switching10.04.314.84.2<0.0001
 Semantic fluency total15.04.518.84.80.008
 Semantic clustering0.
 Semantic switching9.

Although the patients performed below controls on both short- and long-delay recall, there was virtually no loss of information over this time interval in either group. The results for the patient group and the controls on the first trial on list A were 8.0 ± 1.3 and 9.0 ± 1.3, respectively, and the results for the two groups on list B were 8.0 ± 1.6 and 10.0 ± 1.2. Repeated measures analysis of variance, with list A first trial and list B as within-group variables, showed no significant group × condition interaction, indicating no group difference in susceptibility to proactive interference.

Analyses of verbal fluency measures

MANOVA for the verbal fluency tasks showed an overall group difference, with patients scoring lower than healthy controls (Wilks λ, F(6,42) = 2.9, P = 0.017). There was no significant diagnosis × fluency measures interaction, indicating that the patients did not perform disproportionately worse on semantic than on phonological fluency. Follow-up anovas showed significant differences between the groups on phonological fluency (F(1,47) = 11.9, P < 0.001), semantic fluency (F(6,42) = 7.6, P = 0.008) and number of switches in phonological fluency (F(1,47) = 15.1, P < 0.0001). There were no significant differences between the groups with respect to number of switches in semantic fluency or in mean cluster size on either of the fluency tasks. When the significant difference between the groups in number of switches in the phonological fluency task was analyzed using total output as a covariate, the difference was no longer significant.

Other analyses

There were no significant correlations between the impaired verbal learning and memory subtasks (learning rate, delayed recall and frequency estimation) and the two fluency tasks in either group.

In the schizophrenia group there were no significant correlations between verbal learning and delayed recall on the one hand and severity of symptoms (BPRS) and psychosocial functioning (GAS) on the other. Frequency estimation and GAS, however, were significantly correlated (P < 0.05). There were also significant correlations between phonological fluency and BPRS (P < 0.05) and between phonological fluency and GAS (P < 0.05). Semantic fluency was not correlated with severity of symptoms or psychosocial functioning.


A main finding is the robust deficit evident in verbal learning in adolescents with schizophrenia spectrum disorder, corroborating and elaborating the results of a few other studies of memory in adolescents with schizophrenia.17,34 These studies, however, did not systematically investigate various subcomponents of verbal learning and memory. The patient group also performed significantly below healthy controls on a measure of delayed recall, while recognition memory was intact. A similar pattern of verbal learning and verbal memory impairments have been found in adult chronic9 and first-episode35,36 populations. Although the patients exhibited impaired delayed recall, as compared to controls, previously learned material does not appear to fade over time more than in healthy controls. This is in accordance with studies of adult patients with schizophrenia.7,37 Lack of correlations between the verbal learning and verbal recall measures and symptom severity and psychosocial functioning indicate that this aspect of cognitive functioning could be an independent pathology dimension in adolescents with schizophrenia spectrum disorder.

There was a between-group difference on the frequency estimation task, presumed to tap an automatic process in verbal memory, with patients scoring poorer. Although the strong claim of automaticity has come into question,38 it is clear that frequency estimation is accomplished with a similar high degree of accuracy by subject groups that vary in recall ability.32 Although one should be careful about drawing conclusions based on a single study, this finding indicates that impaired verbal memory might extend beyond effort-demanding tasks in adolescents with schizophrenia spectrum disorders.

The lack of significant between-group differences on the stem completion task, tapping implicit memory, is accordance with most studies of adult patients with schizophrenia. In a study by Perry et al. patients with schizophrenia showed impaired performance on recall of words, but they performed normally on a word stem priming test.39 In another study both patients with schizophrenia and first-degree biological relatives of schizophrenia patients exhibited worse free recall than controls, but both groups had intact repetition priming on a subsequent lexical decision task.40

Although both the frequency of estimation task and the stem completion task might be classified as reflecting automatic processes, the results differed. There are obvious differences, however, in task demands posed by the two tests. A possible explanation for this dissociation could be accounted for by postulating a continuum of tests, going from those requiring primarily perceptual information to those requiring knowledge of meaning.41 Thus, these results indicate that basic perceptual comparison processes are unaffected in adolescents with schizophrenia. In contrast, more conceptually driven processes, although automatic in nature, seem to be impaired.

The present results did not confirm that adolescents with schizophrenia spectrum disorders are more prone to proactive interference than controls. Because interference may be viewed as an index of distractibility, this finding is somewhat surprising. Nevertheless, the present data are in line with other studies reporting no differences between adult patients with schizophrenia and healthy controls on parameters reflecting proneness to interference.37,42

A common assumption has been that verbal memory deficits in schizophrenia can be attributed to an associative organizational disturbance resulting in inefficient encoding.43 Impaired learning and free recall in the context of intact recognition memory is in accordance with this. It has also been argued that a lack of false recognition may be due to failures in semantic activation.44 A recent meta-analysis reported the effect size for recognition memory, as measured by false alarms, to be smaller than as measured by hit rate.45 The present study confirms this. Repeated measure of analysis, however, showed no statistical significant interaction between group and hits versus false alarms, indicating that there is no significant difference between these two recognition task indices in adolescents with schizophrenia spectrum disorder compared to controls.

Corroborating studies of adults with schizophrenia, the adolescent patient group was impaired on both phonological and semantic fluency tasks. Impaired phonological fluency has been documented in one previous study of adolescents with schizophrenia,15 while, to our knowledge, no previous study of adolescents has included a measure of semantic fluency. Contrary to many studies of adult patients11 and contrary to our hypothesis, however, the results did not indicate a disproportionate impairment in semantic compared to phonological fluency in the patient group.

Consistent with the present results, impaired switching in phonological fluency in adult patients with schizophrenia has been reported previously,46 although total output was not controlled for in that study. Also in line with the present results, it has been found that patients with first-episode schizophrenia had similar mean cluster sizes as healthy controls on fluency tasks.47 A study by Elvevåg et al. used a different procedure in defining clustering, and also failed to find any difference in number and size of clusters between adult patients with schizophrenia and healthy controls.48 In another study it was found that patients with schizophrenia generated significantly fewer total words, cluster-related words and switches than healthy controls on both phonological and category fluency tasks.14 In that study, however, a disproportionate impairment in semantic fluency resulted from a differential deficit only in clustering.

We did not perform the Bonferroni correction to correct for multiple comparisons and Type I errors. In general, the results of studies such as the present one are tempered by the relatively small group sizes. This entails a reduction in statistical power, which may result in Type II errors. It is particularly in the combination of small study groups that multiple variables and strategies that correct for Type I errors can result in a failure to produce significant and potentially clinically meaningful results.

Although it might be a controversial issue, we have not co-varied for IQ despite the difference of the IQ measures. In many research contexts it is common to match patients and controls on IQ, education and sometimes on socioeconomic status. Matching for these psychometric and demographic characteristics can, however, lead to erroneous conclusions in the case of schizophrenia. Such analyses increase the likelihood of committing Type II error. It could also be argued that because patients with schizophrenia are intellectually compromised, controlling for IQ is in actuality the same as removing the effect of schizophrenia.49 The reason for this is that the existence of the disease in an individual in itself impedes education, impairs cognitive abilities measured by IQ tests, and produces downward drift in socioeconomic status. Thus, equating patients and controls on variables that may be disrupted by the process or outcome of the disease itself may lead to biased comparisons, that is, contrasting overachieving patients with underachieving controls.

Because patients were receiving a variety of medications and some were not medicated, this could be a confounding variable. Results are, however, still inconclusive regarding the effects of both atypical and typical antipsychotic medication on cognitive functioning.50 A recent study indicated that antipsychotic medication, independent of whether typical or atypical, leads to small improvements in cognitive functioning.51 Also, previous studies of neuroleptic-naïve patients have documented cognitive deficits.4 Thus, taken together it seems unlikely that the differences between medicated patients and controls could be explained by the use of medication or medication type, although this cannot completely be ruled out. In contrast, it is well known that anticholinergic medication can interfere with learning and memory in both healthy subjects and neurologically impaired populations.52 None of the subjects in the present study, however, was receiving anticholinergic medication.

The heterogeneity of the group may be considered a limitation, particularly the inclusion of subjects with SPD. In contrast, research shows that cognitive deficits are also evident in patients with other psychotic disorders,36 as well as in subjects with SPD.53 The two SPD subjects exhibited better cognitive functioning than the mean of the schizophrenia sample, but the data were also re-analyzed excluding the two SPD subjects and the results were the same.

Another limitation is that we had no direct assessment of the duration of illness. The majority of patients, however, were undergoing their first or second hospitalization and for most this was their first episode of frank psychosis, thus indicating that they were at least relatively early in the phase of illness.

Although impaired cognition in schizophrenia has long been acknowledged it is only recently that the functional consequences of such impairments have been recognized. Pertaining to impaired verbal memory, this has been found to have a significant impact on social problem solving and community and daily activities.54 This implies that deficits in learning may have an impact on the individual's capacity to participate in social interactions, make decisions and perform a job efficiently. This poses a strong argument for targeting these impairments for interventions such as cognitive remediation. More specifically, training adolescents with schizophrenia to apply semantic encoding strategies may elevate verbal learning abilities. Due to the early onset of illness and the consequence that this may have for the future prospects of these adolescents, this may be particularly vital for this group.

In conclusion, these results suggest that the core problem for adolescents with schizophrenia spectrum disorders lies in the acquisition phase, that is, they learn less to begin with. Although the patients also show deficient long-term memory they do in fact remember most of what they actually learned. Contrary to many studies of adult patients, there is not a disproportionate impairment in semantic as compared to phonological fluency among adolescents with schizophrenia spectrum disorders.


This study was supported by funds from the Norwegian Research Council, grant 14268/320, and the National Council for Mental Health/Health and Rehabilitation, grant 2001/2/0003. Additional funding was provided by Sogn Centre for Child and Adolescent Psychiatry, Department of Psychology, University of Oslo, and Regional Centre for Child and Adolescent Mental Health, regions east and south.