Developmental Changes in Semantic Verbal Fluency

This article was downloaded by: [Vrije Universiteit Amsterdam] On: 14 December 2011, At: 02:12 Publisher: Psychology Press Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Child Neuropsychology Publication details, including instructions for authors and subscription information: http://www.tandfonline.com/loi/ncny20

Developmental Changes in Semantic Verbal Fluency: Analyses of Word Productivity as a Function of Time, Clustering, and Switching a

b

b

P. P. M. Hurks , D. Schrans , C. Meijs , R. Wassenberg c

M. Feron & J. Jolles

b d

, F. J.

a b d

a

Department of Psychology and Neuroscience, University Maastricht, The Netherlands b

Department of Psychiatry and Neuropsychology, University of Maastricht, The Netherlands c

Youth Health Care of the Municipal Health Center, Maastricht, The Netherlands d

Department of Psychiatry and Neuropsychology, Academic Hospital Maastricht, The Netherlands Available online: 06 Apr 2010

To cite this article: P. P. M. Hurks, D. Schrans, C. Meijs, R. Wassenberg, F. J. M. Feron & J. Jolles (2010): Developmental Changes in Semantic Verbal Fluency: Analyses of Word Productivity as a Function of Time, Clustering, and Switching, Child Neuropsychology, 16:4, 366-387 To link to this article: http://dx.doi.org/10.1080/09297041003671184

PLEASE SCROLL DOWN FOR ARTICLE Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions This article may be used for research, teaching, and private study purposes. Any substantial or systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to anyone is expressly forbidden. The publisher does not give any warranty express or implied or make any representation that the contents will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims, proceedings,

Downloaded by [Vrije Universiteit Amsterdam] at 02:12 14 December 2011

demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in connection with or arising out of the use of this material.

Child Neuropsychology, 16: 366–387, 2010 http://www.psypress.com/childneuropsych ISSN: 0929-7049 print / 1744-4136 online DOI: 10.1080/09297041003671184


DEVELOPMENTAL CHANGES IN SEMANTIC VERBAL FLUENCY: ANALYSES OF WORD PRODUCTIVITY AS A FUNCTION OF TIME, CLUSTERING, AND SWITCHING P. P. M. Hurks,1 D. Schrans,2 C. Meijs,2 R. Wassenberg,2,4 F. J. M. Feron,3 and J. Jolles1,2,4 1

Department of Psychology and Neuroscience, University Maastricht, The Netherlands, Department of Psychiatry and Neuropsychology, University of Maastricht, The Netherlands, 3Youth Health Care of the Municipal Health Center, Maastricht, The Netherlands, and 4Department of Psychiatry and Neuropsychology, Academic Hospital Maastricht, The Netherlands 2

We investigated age-related improvement in semantic category verbal fluency (VF) in 309 Dutch schoolchildren attending first to ninth grade. Quantitative analyses of number of correct responses as a function of time as well as qualitative analyses of clustering and switching were conducted. Overall, Dutch VF task performance, i.e., number of correct responses over 60 seconds, was not established before mid-adolescence. This is in line with previously published studies, using VF number of correct responses over 60 seconds as the main outcome measure and examining VF task performance across other cultures and languages (e.g., Italian, French, Hebrew). Next, mean cluster size, a measure of lexicosemantic knowledge, was not established until at least grade 3. In contrast, performance on the VF outcome measures “number of switches/clusters” was established at least 4 years later. Qualitative and quantitative Design Fluency (DF) outcome measures support the notion that the numbers of switches/clusters are valid measures of higher order cognitive functions, such as strategy use and cognitive flexibility. In line of this, VF number of correct responses during 16–60 seconds, a measure of controlled information processing, is established at least 2 years later (i.e., grades 7–8) than number of correct responses during the first 15 seconds time slide, a measure of automatic processing. Finally, environment, i.e., the level of parental education, primarily affected automatic and lexico-semantic knowledge. No effects of sex on VF performance were found. These data suggest that the alternative scoring methods of VF tasks can be used to acquire knowledge on development of lower and higher order cognitive functions in healthy children and the influence of the environment on it. Keywords: Children; Fluency; Controlled and automatic processing.

Worldwide, verbal fluency (VF) tasks are frequently used in a clinical setting as well as in research. In general, VF tasks are operationalized as the number of words produced, usually within a restricted category and within a given time limit (Lezak, 2004). Performance on this type of task is thought to depend on the integrity of different cognitive Address correspondence to P. P. M. Hurks, Department of Neuropsychology and Psychopharmacology, University of Maastricht, P.O. Box 616, NL - 6200 MD Maastricht, The Netherlands. E-mail: Pm.Hurks@ maastrichtuniversity.nl © 2010 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business


VERBAL FLUENCY IN CHILDREN AGED 6–15 YEARS

367

functions, e.g., the ability to initiate search and to retrieve data from the lexicon, an efficient strategy use and attention, in adults (Martin, Wiggs, Lalonde, & Mack, 1994; Monsch et al., 1994; Rosser & Hodges, 1994) as well as in children (Barkley, 2006; Klenberg, Korkman, & Lahti-Nuuttila, 2001). The exploring of the potential use of this type of task in pediatric and developmental neuropsychological research has only begun recently (Hurks et al., 2006; Riva, Nichelli, & Devoti, 2000). For one, cross-sectional studies have shown that word fluency improves with increasing age to at least 13 years (Klenberg et al., 2001; Koren, Kofman, & Berger, 2005; Levin, Song, Ewing-Cobbs, Chapman, & Mendelsohn, 2001), indicating that normal VF task performance is not “established” until mid-adolescence. However, when reviewing these developmental studies, two major limitations came to mind. First of all, these studies often lacked to examine the influence of age-extrinsic biological factors (e.g., sex) and psychosocial factors (e.g., socioeconomic background, educational background, and/or profession of the caregiver) on maturational differences in VF task performance. This is unfortunate, since recent findings suggest that these factors may explain subtle interindividual differences in patterns of global cognitive development (Hurks et al., 2006; Klenberg et al., 2001; Korkman, Kemp, & Kirk, 2001; Lupien, King, Meaney, & McEwen, 2001; McCulloch & Joshi, 2001; Rietveld, Dolan, van Baal, & Boomsma, 2003; Roberts & Bell, 2002; Seymour, Aro, & Erskine, 2003) and therefore quite possible also of VF task performance. A second shortcoming relates to the outcome measures included in these studies. Most studies previously defined VF task performance as the total number of words generated within 60 seconds. However, some authors (e.g., Hurks et al., 2004, 2006; Troyer, 2000) have claimed that this scoring method does not provide information about the specific cognitive mechanisms underlying poor test performance. Indeed, recent experiments have shown that studying (a) the pattern of correct responses as a function of time and/or (b) measures of systematic organization of information, such as clustering of words, enables a more direct insight in the processes (and connections between them) underlying VF task performance in adults (Troyer, 2000) as well as in children (Hurks et al., 2004, 2006). Although the search of alternative VF task scoring methods has primarily focused on adult populations, both alternative scoring methods will be discussed here within a developmental as well as in an age-extrinsic context. The first alternative scoring method is based on the model of lexical organization (Crowe, 1996; Smith & Claxton, 1972). This model states that there are two types of stores activated while performing a VF task, namely (a) a long-term store (“topicon”) that is readily accessible and contains common words, and (b) a more extensive lexicon that is searched after the topicon is exhausted. This model hypothesized that during the first 15–20 seconds of a VF task, a ready pool of common words (i.e., the topicon) is available and is automatically activated for production. As time passes, this pool becomes exhausted and production becomes more effortful or controlled, less productive, and more dependent on executive functions (Crowe, 1998). Successful performance on a VF task thus seems to depend on the effectiveness of both automatic and controlled processing. Although this method of analyzing correct responses as a function of time has led to a number of interesting observations within the field of adult neuropsychology (e.g., Crowe, 1998; Monsch et al., 1994; Rosser & Hodges, 1994; Troyer, 2000), to our knowledge, this alternative scoring method has only rarely been applied within a developmental context. For one, some of us (Hurks et al., 2004, 2006) indeed found support for the model of lexical organization in children. Secondly, Hurks et al. (2006) found that the effects of environment,


368

P. P. M. HURKS ET AL.

i.e., the effects of levels of parental occupational achievement, were most profound during the 16–60 seconds (and thus during controlled processing) of a VF task. In addition, the authors found no sex differences in relation to time on task. Unfortunately, only a limited age range was included in this study, i.e., children aged 8–9 years, whereas, as mentioned earlier, other studies have shown that word fluency, as measured over 60 seconds, continues to improve with increasing age until at least 13 years. For this reason, it would be interesting to investigate this type of automatic and controlled VF task processing within a broader age range. Therefore, our first aim is to evaluate the development of word retrieval on VF tasks in relation to time on task in healthy children aged 6–15 years and the influence of age-extrinsic factors, such as sex and parental education (a measure that is known to correlate highly with the level of occupational achievement discussed earlier) on this development. We hypothesize that automatic processes (i.e., occurring during 1–15 seconds of the VF task) are “established” earlier compared to controlled processes (i.e., 16–60 seconds). In addition, environment is believed to affect the development of both automatic and controlled processes, but mainly the latter ones. Finally, no sex differences are expected to be found on these qualitative and quantitative VF scores. Next, we will discuss measures of systematic organization of information as alternative VF task scoring methods. As mentioned earlier, one may generate information regarding the processes underlying VF task performance by looking at specific time intervals. However, looking at the content of the words produced may enlarge our current knowledge in this perspective as well (Koren et al., 2005). It has long been known that word production on fluency tasks, especially semantic fluency tasks, tends to occur in spurts, and words produced during such spurts are often semantically related or clustered (Abwender, Swan, Bowerman, & Connolly, 2001). Presumably not all words in that nested subset (or that are clustered) are recoverable, and a switch to a new nested search is initiated when the time interval between responses lengthens (Wixted & Rohrer, 1994). In line of this, Troyer, Moscovitch, and Winocur (1997) suggested that clustering and switching are two qualitative components of the verbal fluency process that determine output quantity. Cluster size depends on specific cognitive functions, such as lexico-semantic knowledge. In contrast, the ability to initiate clustering and to switch between clusters is assumed to involve higher order cognitive processes such as cognitive flexibility or set shifting. As will be shown later, discussion exists on whether semantic clustering is a product of strategy, or merely a consequence of spreading activation throughout a semantic network (e.g., Collins & Loftus, 1975). Troyer et al. (1997) claimed that both abilities are equally important for successful VF task performance. Analyses of the number of clusters (as a measure of switching) and mean cluster size (measuring the ability to name words within a cluster) have been successfully applied to various healthy and neurologically or psychiatrically impaired adult patients in order to further understand the nature of the cognitive impairment and have more specifically helped to clarify the relative contribution of executive functions and semantic stores to the performance on VF tests in these diverse, although primarily adult populations (Elvevag, Fisher, Gurd, & Goldberg, 2002; Ho et al., 2002; Koren et al., 2005; Troyer et al., 1997; Troyer, Moscovitch, Winocur, Alexander, & Stuss, 1998). Despite the potential of this alternative scoring method in studying the cognitive development of children, it has only instantly been included in studies examining normal cognitive development and its results are still inconsistent in outcome (Kavé, Kigel, & Kochva, 2008; Koren et al., 2005; Sauzéon, Lestage, Raboutet, N’Kaoua, & Claverie, 2004). While studying children aged 7–16 years, Sauzéon et al. (2004) found that the mean cluster size ratio (i.e., mean cluster size divided by the total number of words generated)



369

started to diminish significantly after 9–10 years. In contrast, no age-related change was observed in the number of clusters, as a measure of switching. According to Troyer et al.’s (1998) arguments stated above, these results on the one hand suggest that the age-related growth in VF task performance is not likely related to the development of higher order functions such as strategy use, since the number of clusters were unaffected by age. On the other hand, Sauzéon et al. claim that the change in semantic cluster size ratio indicates that the age difference in semantic fluency is more closely related to the clustering component, itself associated with lexico-semantic knowledge and semantic network exploration (Sauzéon et al., 2004). However, an age-related decrease, instead of an increase, in cluster size ratio sounds counterintuitive. As mentioned above, Sauzéon et al. divided the mean cluster size by the total number of words generated over 60 s in order to calculate the cluster size ratio. However, the authors also reported a significant effect of age on the total number of words. Children 11–16 years old produced more words over 60 s than 7- to 8and 9- to 10-year-olds. This age-related increase in number of words generated over 60 s could lead to a significant age-related decrease in terms of cluster size ratio, even if mean cluster size is unaffected by age. Unfortunately, no data was provided with regard to the uncorrected outcome measure “mean cluster size” and its potential relation to age. Therefore, one should be careful in interpreting Sauzéon et al.’s data on age and cluster size ratio. In addition, Koren et al. (2005) found a reversed effect: They found a significant age effect when studying mean number of clusters (note that they only included two age groups, i.e., third graders and fifth graders), but not in cluster size, indicating that older children (i.e., aged 10–11 years) only produced more clusters than younger children (i.e., aged 8–9 years). Koren et al. argued that the increase in number of switches and concomitant increase in fluency might be related to the development of controlled processes or executive functions. Finally, Kavé et al. (2008) found, while including participants aged 8–29 years, that all measures (i.e., total number of correct responses over 60 seconds, number of clusters, and cluster size) increased with age. These results are inconclusive and one limitation is that age-extrinsic biological and environmental factors were insufficiently taken into account (i.e., only Koren et al. considered one age-extrinsic factor, namely sex, who by the way found no sex differences on semantic clustering measures). The second aim of our study is therefore to examine interindividual differences in the development of qualitative VF output measures, i.e., cluster size and switching, as an addition to our first aim. Finally, with regard to the interpretation of these output measures, it should be stated that a recent spate of published studies has examined the predictive validity and neuropsychological correlates of these components of the verbal fluency process in adults. Although generally encouraging, these studies have not always yielded the expected results, and some of the assumptions underlying the interpretation of qualitative measures of fluency performance remain open to question. For one, positive correlations between clustering measures and total productivity, as also found in the present study, cannot be taken ipso facto as evidence that clustering necessarily leads to the production of more words or is a strategic process. The longer a list of randomly (i.e., nonstrategically) generated words is then the more likely it is that what appear to be meaningful clusters will appear by chance (Abwender et al., 2001). Data from a study on verbal fluency of people with schizophrenia are, however, consistent with the possibility that clustering is a function of strategy use. Compared with normal controls, people with schizophrenia generated fewer words and fewer semantic clusters on a semantic fluency test, although their mean


370


cluster size, which is believed to be a measure of specific cognitive functions such as lexico-semantic knowledge, was similar to those of the controls (Robert et al., 1997). The authors suggested that the people with schizophrenia produced fewer words in total as a result of defective strategy use. That is, when they did utilize a clustering strategy, they produced clusters as large as controls, indicating that specific cognitive functions as lexico-semantic knowledge are normal in this population. People with schizophrenia nonetheless were limited in their ability to initiate and use the clustering strategy (Abwender et al., 2001). However, as mentioned above, these data are collected in an adult, clinical population. Therefore, to test the construct validity in the present study, we compared the developmental profiles on quantitative and qualitative VF outcome measures with quantitative and qualitative Design Fluency (DF) test outcome measures. Performance on this DF test is believed to be mediated considerably by strategy use (Lezak, 2004) and was therefore chosen to provide an index of spontaneous use of strategy in a divergent task. We hypothesized that our scores, designed to capture the VF clustering and switching constructs more purely, would correlate with DF output, thereby supporting the validity of these outcome measures as indicators of strategy use on verbal fluency. In sum, in the current study, we sought to clarify the effects of age, sex, and the level of parental education (LPE) on both quantitative and qualitative measures of semantic verbal fluency tasks. METHOD Procedure This study uses data of a large, longitudinal research program that focuses on mechanisms underlying cognitive development in children. Within this program, crosssectional and longitudinal experiments are conducted, including large numbers of healthy children aged 6 through 15 (e.g., Hurks et al., 2006; Kalff et al., 2005; Kroes et al., 2002; Meijs et al., 2008; Wassenberg, Hendriksen, et al., 2008; Wassenberg, Hurks, et al., 2008). In the present study, children were enrolled by approaching schools for regular elementary and regular secondary education in Maastricht — a city in the southern region of the Netherlands — and the surrounding region. Schools that agreed to participate in this study received information packages that they were asked to hand over to the caregivers of all children in the following grades: K (within the Dutch school system, children frequenting this grade are aged 5–6 years), second grade (i.e., children aged 7–8 years), fourth grade (i.e., children aged 9–10 years), sixth grade (i.e., children aged 11–12 years), seventh (i.e., children aged 12–13 years), and eighth grade (i.e., children aged 13–14 years). Seventh and eighth grade are part of the Dutch secondary school system: ranging from lower secondary professional education to pre-university education. Besides information about the purpose of the study, the information packages contained a stamped return envelope, an informed consent letter, and a questionnaire on sociodemographic characteristics and the child’s medical history and milestone development, which we asked to be filled in if caregivers agreed to participate (i.e., phase 1). Children who had the Dutch nationality, used Dutch as their primary language, and frequented one of the appropriate grades were eligible for participation in the study. Children who used medications that could influence their cognitive functions, such as antihistamines (e.g., Ng et al., 2004) and psycho-stimulants, and children who previously had retained or skipped a grade were excluded from the study. Children with DSM-IV Axis I “Disorders Usually First Diagnosed in Infancy,



371

Childhood, or Adolescence” (e.g., dyslexia) were not excluded from the study because the resulting selection would have resulted in a “supernormal” sample. Rather, we reasoned that if children were attending a school for regular education and were in the appropriate grade, they could be considered as developing within the “normal” range. We endeavored to include an equal number of boys and girls for each grade and also aimed to include children from various socioeconomic backgrounds. After inclusion in the study, the children were tested individually in a stimulus-free test room at their school by use of a neuropsychological test battery (i.e., phase 2). Testing was conducted by three well-trained graduate students majoring in developmental psychology and/or neuropsychology. An individual testing session took approximately 90 minutes and tests were all administered in the same order for each child. One year after the first neuropsychological assessment, all caregivers and children originally included in the study were tested again, individually, at their school (i.e., phase 3). Two tests, only conducted during phase 3, were the VF task and the DF test as described in the introduction. Here, our main focus was on grade differences (and interactions with age-extrinsic factors) in performance on this test (i.e., administered during phase 3 and therefore studied within a cross-sectional design). The entire study was approved by the ethical committee of the department of Psychology and Neuroscience at Maastricht University. Participants In phase 1, approximately 3000 information packages were sent out to children and their caregivers; 81.9% of them agreed to participate. Of the 892 children whose caregivers provided consent for the study, 431 were selected using the inclusion and exclusion criteria described above and conducted the baseline measurement (i.e., phase 2). After approximately 1 year, all the caregivers of the children who still attended one of the participating schools were contacted again to request further participation of their child in the study. In total, 333 caregivers gave consent for their child. Among these 333 children, n = 7 had repeated or skipped a grade between phases 1 and 2 and phase 3 and were thus excluded. Fifteen children were excluded because they were diagnosed with a condition known to affect cognitive functioning (e.g., epilepsy or ADHD) or because they started taking medication known to affect cognitive performance (between phases 1 and 2 and 3). Of the remaining participants, VF task data of 2 children were excluded from the final analyses, because of unreliable test administration or refusal of the child to comply with the test instructions (mainly because of fatigue). Table 1 shows the characteristics of the remaining 309 children. Instruments Semantic category verbal fluency test. Semantic category verbal fluency (VF) involves the recitation of examples of a given semantic category (here: to name as many animals as possible) in 60 seconds. To begin, the following instructions were read aloud: “Next, I’m going to give you one minute to tell me all the animals you can think of. They can be any kind of animals that you can think of, such as birds, fish, etcetera. Any questions? Go as fast as you can. Ready? Go.” Correct responses were recorded over (a) 1–15 seconds, (b) 16–30 seconds, (c) 31–45 seconds, and (d) 46–60 seconds. All responses were recorded verbatim, with incorrect responses subsequently excluded. Incorrect responses were: (a) words that were not an exemplar of the category “animals” and

372


Table 1 Group Characteristics of the Sample (N = 309). Grade

Variables Age Parental education


Sex Vocabulary subtest n

Mean SD low high girls boys Mean SD

1

3

5

7

8

9

7.32 0.36 30 35 37 28 10.4 3.8 65

9.36 0.41 23 37 30 30 10.3 3.5 60

11.40 0.35 25 34 38 21 10.4 3.0 59

13.46 0.31 13 31 23 21 9.9 2.1 44

14.35 0.33 12 35 19 28 10.2 2.1 47

15.47 0.33 15 19 13 21 9.5 2.1 34

chi-square /F value

p value

7.3

.201

9.5

.092

0.53

.754

(b) perseverations (e.g., repetitions of correct words and morphological variants [for instance when a child says cat and cats]). Next, three alternative scores were calculated: (a) mean cluster size, (b) number of switches, and (c) number of clusters. Mean cluster size. The method of analysis of cluster formation is still not standardized in literature as various definitions of clusters have been reported. In the present study, clusters were defined as groups of successively generated words belonging to the same zoological families (e.g., primates, birds, insects), largely in line with detailed scoring rules provided by Troyer (2000), Troyer et al. (1997), Robert et al. (1998), and Mahone, Koth, Cutting, Singer, and Denckla (2001). An overview of zoological families is provided in Appendix A. Guidelines were formulated for consistency sake but flexibility was allowed for the coding of associated words that did not fall under the list of predefined clusters. A cluster had to consist of three or more semantically associated words, as suggested by Robert et al. Cluster size was counted beginning with the first word in each cluster. For example, in the following list — walrus, fur seal, sea lion, monkey, hen, rooster, goose, whale, fly, cockroach, beetle, snake, pigeon, seagull, owl, canary — the number of words related by cluster is 13. All clustering scores for the fluency measures were scored twice to check scoring validity; discrepancies in scoring were handled between scorers. Number of switches. The number of switches between clusters of three words or more, between clusters, and words that were generated outside a cluster, and among those out of cluster words (as in Troyer et al., 1997; Kavé et al., 2008) was counted for every participant. Number of clusters. The number of clusters was calculated as the number of clusters generated within 60 seconds, without single words, in order to examine participant’s use of word association. As noted by Koren et al. (2005), the presence of single words may indicate that participants are in fact unable to utilize an associated strategy, and thus a measure that leaves out the single words is essential when focussing on the ability to produce related words. In line with Troyer et al.’s (1997) guidelines, incorrect responses (as defined earlier) were included in calculations of mean cluster size and number of clusters/switches


373


because any word that is produced provides information about the underlying cognitive processes regardless of whether or not it contributes to the total correct number of words generated. Finally, phonemic clusters generated within the semantic task were not scored, in line with Kavé et al. (2008). Reliability coefficients of American and Dutch versions of VF tests (i.e., the NEPSY subtest Verbal Fluency and the Dutch Word Fluency test) in a population of children are .76–.77 (Dekker, Mulder, & Dekker, 2007; Korkman, Kirk, & Kemp, 1998). Interrater reliabilities for clustering are believed to be good: Troyer et al. (1997) for instance reported reliabilities to be above .95. Design fluency test (Korkman et al., 1998). This test is considered to be a nonverbal analogue of verbal fluency tests. The test material consists of a large sheet with 70 squares, each with five dots, arranged in a structured array. The child is instructed to generate as many unique designs as possible in 1 minute by connecting two or more dots with straight lines. Participants could lift the pen from the page; that is, the designs did not have to be made with one, continuous stroke. Poor performance on this test is believed to stem from cognitive flexibility and/or failure to employ strategies for generating nonrepetitive designs (Lezak, 2004). Mean test-retest reliability is .59 (Korkman et al., 1998). Total number of unique designs over 60 seconds is recorded. Also, clustering, cluster size, and switching have been calculated in line with scoring methods of quantitative VF measures. A cluster had to consist of three or more associated patterns. Three types of clusters were defined, i.e., addition of an element (or straight line), deletion of an element, and rotation of a design. Addition is the strategy in which a participant adds an additional element to an existing form in a systematic fashion, meaning it is used in at least three consecutive designs. Deletion refers, contrarily, to the strategy in which a participant excludes a single element from an existing form in a systematic fashion. Rotation is the movement of elements around the main axis of the five points. Note, however, that, although the design fluency test has been implemented in many scientific studies, to the best of our knowledge, only limited research, and only in adults, has been done with regard to qualitative scoring DF systems (Goebel, Fischer, Ferstl, & Mehdorn, 2009). Again, all clustering scores for the fluency measures were scored twice to check scoring validity; discrepancies in scoring were handled between scorers. Vocabulary. This subtest of the Wechsler Intelligence Scales (WISC-Revised Dutch Version; De Bruyn et al., 1986) was used to estimate general verbal ability (Lezak, 2004). The examiner asked the child to explain the meaning of certain words. The complexity of the words increased with each item. Possible standard scores range from 1 to 19 (M = 10, SD = 3). The reliability coefficient for the Vocabulary subtest in a Dutch population of children is .86–.88 (De Bruyn et al., 1986). The WISC-R, rather than the WISC-III, was used because the latter became available in the Netherlands only after the study had started. Level of parental education (LPE). This variable was based on a full description of level of the attained parental education and was originally scored on a 8-point scale, ranging from elementary school (code 1) to PhD (code 8) (De Bie, 1987). For the purpose of the present article, this 8-point scale was recoded into two groups: lower LPE (including scores 1, 2, 3, & 4), and higher LPE (scores 5, 6, 7, & 8). When the LPE differed between mother and father or if data were missing for one caregiver, the highest available score was chosen. This system is similar to the International Standard Classification of Education (United Nations Educational Scientific and Cultural Organisation [UNESCO], 1997).

374


RESULTS


Statistical Methods The degree to which various VF components were associated was calculated for the entire sample (independently of age, sex, and LPE). Next, Table 2 displays means and standard errors of fluency performance for each grade. Performance on VF tasks as a function of time was evaluated statistically for differences in terms of (a) grade, (b) sex, and (c) LPE. Effects of these factors on VF number of correct responses were studied as a function of time by using a 4 × 6 × 2 × 2 (Time Intervals × Grade × Sex × LPE) General Linear Models (GLM) repeated measures analysis. In addition, effects of grade, sex, and LPE on VF as well as DF qualitative measures — i.e., number of switches, number of clusters, and mean cluster size — were studied by using separate 6 × 2 × 2 (Grade × Sex × LPE) GLM univariate analyses. Overall, nonsignificant factors were deleted from the final analyses to get the most optimal model. The critical value for rejecting the null hypotheses was defined at the more conservative level of p < .05. Post hoc corrections were conducted with the Least Significance Difference (LSD) algorithm.

Number of Correct Responses as a Function of Time: The Influence of Grade, Sex, and LPE GLM repeated measures analysis revealed that the main effect and/or interaction effects including the factor “sex” were nonsignificant. Thus, the factor “sex” did not influence VF number of correct responses over time and was deleted from the model. Therefore, a 4 × 6 × 2 (time intervals × grade × LPE) GLM repeated measures was run and revealed a significant interaction between time intervals and grade, F(15, 837) = 3.80, p < .001, ηp2 = .06, indicating that the pattern of words produced in the different time intervals was not identical for all grades (as shown in Table 2). During the 1–15 second interval, improvement (i.e., a significant increase in number of correct responses) was visible until grade 5 (i.e., aged 10–11 years). Thereafter, no further improvements or significant differences in automatic processing performance were found, time intervals × grade: F(5, 279) = 22.76, p < .001, ηp2 = .29. Next, during the 16–30 s interval, improvement was significant until grade 7 (i.e., aged 12–13 years), time intervals × grade: F(5, 279) = 5.21, p < .001, ηp2 = .20. During the 31–45 s interval, improvement was significant until grade 8 (i.e., aged 13–14 years), F(5, 279) = 7.04, p < .001, ηp2 = .11. Finally, during the 46–60 s interval, improvement was significant until grade 7, i.e., F(5, 279) = 5.62, p < .001, ηp2 = .09. In conclusion, automatic processing (as measured by 1–15 seconds) seems to function at an “established” level at least 2 years earlier than controlled processing (as measured by 16–60 seconds) in children. In addition, LPE affected performance on the VF tasks as a function of time, Time Intervals × LPE: F(3, 277) = 2.65; p = .049, ηp2 = .03. Independently of age, children of caregivers with a higher scaled education produced more words during the first two time slides than children of caregivers with a lower scaled education, 1–15s: F(1, 279) = 4.11, p = .044, ηp2 = .02 (Mlower LPE = 7.96, SD = 0.20 vs. Mhigher LPE = 8.47, SD = 0.16; 16–30s: F(1, 279) = 5.21, p = .023, ηp2 = .02 (Mlower LPE = 4.25, SD = 0.20 vs. Mhigher LPE = 4.83, SD = 0.16). This difference disappeared as time on task increases, 31–45s: F(1, 279) = 3.79, p = .053, ηp2 = .01 (Mlower LPE = 2.97, SD = 0.19 vs. Mhigher 2 LPE = 3.44, SD = 0.15); 46–60s: F(1, 279) = 1.52, p = .39, ηp < .001 (M lower LPE = 2.82,

375

1 3 5 7 8 9

Grade

12.86 (.57) 16.32 (.61) 18.74 (.62) 20.31 (.77) 22.01 (.76) 21.84 (.79)

# of VF Words

1.79 (.23) 2.25 (.25) 2.77 (.25) 2.92 (.31) 3.10 (.31) 2.91 (.32)

Mean VF Cluster Size

11.29 (.47) 14.14 (.49) 15.69 (.50) 16.42 (.58) 18.18 (.56) 17.12 (.65)

# of VF Switches 0.61 (.13) 0.93 (.14) 1.20 (.14) 1.44 (.16) 1.59 (.15) 1.76 (.18)

# of VF Clusters 5.75 (.26) 7.60 (.28) 8.79 (.28) 8.97 (.35) 8.73 (.34) 9.44 (.36)

# of VF Words Over 1–15s 3.02 (.26) 3.58 (.28) 4.53 (.28) 4.82 (.35) 6.05 (.34) 5.24 (.35)

# of VF Words Over 16–30s

Table 2 Means, Standard Error of Mean (in Parenthesis) of Fluency Variables By Grade.

2.16 (.24) 2.81 (.26) 3.08 (.26) 2.94 (.33) 4.23 (.33) 3.99 (.34)

# of VF Words Over 31–45s 1.93 (.22) 2.33 (.23) 2.34 (.24) 3.57 (.30) 3.01 (.29) 3.17 (.30)

# of VF Words Over 46–60s 8.74 (.39) 11.40 (.48) 14.22 (.54) 17.51 (.77) 18.59 (.86) 18.64 (.86)

# of DF Designs


0.48 (.15) 1.54 (.25) 2.26 (.26) 3.17 (.21) 2.82 (.27) 2.81 (.30)

Mean DF Cluster Size

8.32 (.37) 9.61 (.41) 10.96 (.47) 11.27 (.56) 12.09 (.55) 12.42 (.60)

# of DF Switches

0.18 (.06) 0.65 (.12) 1.15 (.15) 2.27 (.23) 2.25 (.28) 2.18 (.29)

# of DF Clusters

376


SD = 0.17 vs. Mhigher LPE = 2.63, SD = 0.13). It can be concluded that LPE influences primarily automatic processing.


Number of Clusters, Number of Switches, Mean Cluster Size: The Influence of Grade, Sex, and LPE For VF number of clusters, GLM repeated measures analyses revealed no significant main effect of sex and LPE or interaction effects including these factors. Thus, these factors were deleted from the model. A significant main effect for grade was found, F(5, 285) = 8.56, p < .001, ηp2 = .13. In terms of number of clusters, improvement was found until at least grade 7 (i.e., aged 12–13 years). For VF number of switches, GLM repeated measures analyses revealed no significant main effect of sex and LPE or interaction effects including these factors. Thus, these factors were deleted from the model. A significant main effect for grade was found, F(5, 285) = 23.21, p < .001, ηp2 = .29. In terms of number of switches, improvement was found until at least grade 8. Finally, with regard to VF mean cluster size (i.e., a measure of lexico-semantic knowledge), GLM repeated measures analyses revealed no significant main effect of sex or interaction effects including this factor. Therefore, a 6 × 2 (Grade × LPE) General Linear Models (GLM) univariate analysis was conducted. Although no significant interactions for grade and LPE were found, significant main effects were found for grade and LPE, i.e., F(5, 279) = 3.14, p = .009, ηp2 = .05 and F(1, 279) = 4.06, p = .045, ηp2 = .01, respectively. In terms of mean cluster size, age-related improvement was found until at least grade 3 (i.e., aged 8–9 years). Also, children whose caregivers had higher scaled educational levels tended to make longer clusters than did the children whose caregivers had lower educational levels (i.e., Mhigher LPE = 2.77, SD = 0.14 vs. Mlower LPE = 2.30, SD = 0.18). For DF number of designs over 60 seconds, GLM repeated measures analyses revealed no significant main effect of sex and LPE or interaction effects including these factors. Thus, these factors were deleted from the model. A significant main effect for grade was found, F(5, 284) = 45.15, p < .001, ηp2 = .44. In terms of number of designs, improvement was found until at least grade 7 (i.e., aged 12–13 years). With regard to DF number of clusters, GLM repeated measures analyses revealed no significant main effect of LPE or interaction effects including this factor. Therefore, a 6 × 2 (Grade × Sex) General Linear Model univariate analysis was conducted. Although no significant interactions for grade and sex were found, significant main effects were found for grade and sex, i.e., F(5, 277) = 24.51, p = .001, ηp2 = .307 and F(1, 277) = 4.84, p = .029, ηp2 = .02. In terms of number of clusters, age-related improvement was found until at least grade 7. Also, boys tended to make more clusters than girls (i.e., Mboys = 1.60, SD = 0.11 vs. Mgirls = 1.26, SD = 0.11). For DF number of switches, GLM repeated measures analyses revealed no significant main effect of sex and LPE or interaction effects including these factors. Thus, these factors were deleted from the model. A significant main effect for grade was found, F(5, 283) = 10.92, p < .001, ηp2 = .16. In terms of number of switches, improvement was found until at least grade 8. With regard to DF mean cluster size, GLM repeated measures analyses revealed no significant main effect of LPE or interaction effects including this factor. Therefore, a 6 × 2 (Grade × Sex) General Linear Model univariate analysis was conducted. Although no



377

Figure 1 Mean z scores of fluency variables, mean cluster size, number of clusters, and number of switches on the VF/DF task by age group.

significant interactions for grade and sex were found, significant main effects were found for grade and sex, i.e., F(5, 277) = 18.17, p < .001, ηp2 = .247 and F(1, 277) = 4.13, p = .043, ηp2 = .02. In terms of mean cluster size, age-related improvement was found until at least grade 7. Also, boys tended to make longer clusters than girls (i.e., Mboys = 2.38, SD = 0.14 vs. Mgirls = 1.97, SD = 0.14). In order to present the data graphically on a single scale, individual results were transformed into z scores on the basis of the mean of each variable across all participants and then group means of z scores were computed and plotted (see Figure 1). As can be seen, all VF and DF measures showed an age-related improvement. Lexico-semantic knowledge (as measured by VF mean cluster size) seems to function at an “established” level at least 4 years earlier than higher order cognitive functions, such as strategy use (i.e., number of clusters and number of switches on both VF and DF). Validity of VF Outcome Measures As can be seen in Table 3, the total number of words generated over 60 seconds was significantly correlated with total number of clusters, number of switches, mean cluster size, and number of correct responses per 15-second time interval. Total number of clusters, number of switches, and mean cluster size were also significantly correlated with number of correct responses over each separate time interval. Qualitative and quantitative DF outcome measures correlated positively with almost all VF outcome measures. By use of hierarchical stepwise regression analysis, the power of the VF task measures “mean cluster size,” “number of switches,” and “number of clusters” on predicting

378

– .492** .840** .630** .701** .763** .624** .546** .503** .409** .275** .444** – .098 .672** .357** .339** .346** .253** .178** .234** .000 .245**

*Correlation is significant at the .05 level (two-tailed). **Correlation is significant at the .01 level (two-tailed).

# of VF words over 60s Mean VF cluster size # of VF switches # of VF clusters # of VF words over 1–15s # of VF words over 16–30s # of VF words over 31–45s # of VF words over 46–60s # of DF designs Mean DF cluster size # of DF switches # of DF clusters

# of VF Words Over 60s

Mean VF Cluster Size

– .155* .595** .671** .484** .455** .444** .323** .274** .361** – .448** .421** .454** .339** .289** .289** .135* .280**

# of VF # of VF Switches Clusters

Table 3 Within-Task Correlation Coefficients of Fluency Variables.

– .371** .191* .173* .389** .351** .262** .308**


– .347** .261** .408** .261** .270** .315**


– .145* .216** .207** .083 .215**


– .302** .251** .078 .343**


– .675** .662** .784**

# of DF Designs


– .110 .808**

Mean DF Cluster Size

– .065

–

# of DF # of DF Switches Clusters



379

VF number of correct responses over 60 seconds was tested, while including age, LPE, and sex as covariates in the model. In consecutive order, the number of switches (b = .97, SE = .02), the number of clusters (b = 2.16, SE = .07), and mean cluster size (b = .40, SE = .04) were significant (ps < .001) predictors of VF number of correct responses over 60 s. All relations were positive, e.g., an increase in number of switches caused an increase in total VF number of correct responses. In addition, by use of hierarchical stepwise regression analysis, the power of the DF task measures “mean cluster size,” “number of switches,” and “number of clusters” on predicting VF number of correct responses over 60 seconds was tested, while including age, LPE, and sex as covariates in the model. When we included only the DF task measures as independent factors in the model, the total number of designs was a significant predictor of VF number of correct responses over 60 seconds (b = .189, SE = .062). However, this effect disappeared when qualitative VF outcome measures were added to the model. Next, the power of the VF task measures “mean cluster size,” “number of clusters,” and “number of switches” on predicting VF number of correct responses over the four 15-second time indices were tested, while including age, LPE, and sex as covariates in the model. Again, (a) the number of switches (b = .26, SE = .03) and (b) the number of clusters (b = .71, SE = .10) were significant (ps < .001) predictors of VF number of correct responses over 1–15s. While these measures were included in the model, there was a tendency for mean cluster size to contribute to the model as well (p = .064). Comparable results were found for VF number of correct responses over 16–30 seconds. For the preceding time indices, the number of switches and number of clusters were again significant (ps < .001) predictors of VF number of correct responses. However, no tendencies were found for the mean cluster size. Adding DF outcome measures did not improve these models. Finally, because of the assumed relationship between the development of the ability to organize and retrieve words (i.e., VF number of correct responses over 60 s, 15-second time slides, number of switches and clusters, mean cluster size, DF number of designs over 60 s, number of switches and clusters, and mean cluster size) on the one hand and general ability on the other, separate Pearson correlations were calculated to study this relation. Low, but significant correlations for general ability (i.e., standard scores on the Vocabulary subtest) with measures of VF were found (see Table 4). In

Table 4 Correlation Coefficients of Fluency Variables and Grade, Sex, General Ability, and LPE.

# of VF words over 60s Mean VF cluster size # of VF switches # of VF clusters # of VF words over 1–15s # of VF words over 16–30s # of VF words over 31–45s # of VF words over 46–60s # of DF designs Mean DF cluster size # of DF switches # of DF clusters

Grade

LPE

Sex

Vocabulary subtest

.600** .242** .515** .361** .517** .439** .321** .279** .662** .477** .397** .549**

.164* .119* .146* .069 .123* .169* .129* −.006 .101 .111 .058 .095

.102 .106 .030 .110 .095 .080 .020 .071 .114 .155** -.038 .172**

.137* .126* .105 .116* .016 .163* .146* .040 .022 .030 .021 .010

*Correlation is significant at the .05 level (two-tailed). **Correlation is significant at the .01 level (two-tailed).

380


contrast, no correlations for general ability with DF measures were found. In sum, VF tasks would appear only to be marginally a language-related function but also related to other higher order cognitive functions, such as cognitive flexibility. Also, the influence of these functions seems to differ as a function of time on task. DF and VF tasks measure at least in part similar underlying neuropsychological functions, e.g., strategy use and cognitive flexibility.


DISCUSSION Our main aim was to investigate the subtle differences in the development of the interrelated cognitive processes underlying VF task performance. For this purpose, 309 children (aged 6–16 years) were tested within a cross-sectional, longitudinal study design. Our data revealed that, although young children are able to name some animals within 60 seconds (indicating that the cognitive skills underlying VF task performance emerge in early childhood), a significant period of development (i.e., until at least grade 7) occurs before the underlying functions are fully functional or established, as in most cognitive skills (Anderson, 2001). This result, including Dutch children, is in line with previously published studies that used VF number of correct responses over 60 seconds as the main outcome measure and that examined VF task performance across different cultures and languages (e.g., in Italian [Riva et al., 2000], French [Sauzéon et al., 2004], Hebrew [Koren et al., 2005]). One explanation for this protracted developmental trajectory of VF task performance may be that the semantic category “animals” is not easily exhausted in 1 minute. However, since number of correct responses seems to establish after grade 7 (i.e., 12–13 years), this does not explain our results to its full extent. An alternative — or additional — interpretation might be that the protracted development of VF task performance is correlated with the developmental trajectory of so-called higher order cognitive or executive functions (EFs). From the literature it is known that these EFs tend to improve gradually and are not fully “established” until mid-adolescence or early adulthood (Anderson, 2002; Korkman et al., 2001; Shute & Huertas, 1990; Wassenberg, Hendriksen, et al., 2008; Wassenberg, Hurks, et al., 2008). Indeed, our data analyses including alternative scoring methods corroborate this alternative interpretation at least in part. One of the primary outcome measures included in the present study, i.e., the number of switches/clusters, is potentially a measure of cognitive flexibility and/or ability to employ strategies for generating words: Both terms have previously been conceptualized as distinct EF domains (Anderson, 2001). Performance on the number of switches/clusters was not established here until at least grade 7/8 (i.e., 12–14 years), which is in line with previous studies on higher order executive function domains (as cited in Anderson, 2002) as well as with our results on quantitative and qualitative outcome measures of the design fluency test; a test that is believed to measure cognitive flexibility and/or the ability to employ strategies for generating nonverbal designs. This makes it hard to believe that VF measures of clustering and switching are merely a consequence of spreading activation throughout a semantic network. In addition, our data revealed that relative lower order cognitive functions, e.g., lexico-semantic knowledge (as measured by mean cluster size), are established at least 4 years earlier (i.e., at 8–9 years of age) than performance on VF measures switching and number of clusters. By the way, this last result may explain why Koren et al. (2005) did not find any difference in mean cluster size when comparing 8- to 9-year-olds



381

to 10- to 11-year-olds, as mentioned in the introduction. Although differential developmental trajectories were found for number of switches and cluster size, both outcome measures were positively and equally high correlated to and predictive of VF number of correct responses over 60 seconds. This adds to research conducted by Troyer et al. (1997) stating in adults an equal contribution of both components (i.e., number of switches and cluster size) to semantic fluency task performance over 60 seconds. In line with this, VF number of correct responses is not established until grades 7/8 (12–14 years). This was only true for number of correct responses during 15–60 seconds, and therefore controlled information processing, given that the productivity during 1–15 seconds is already established two years earlier (i.e., grade 5). This growth may be caused by executive control (which emerges around 11–13 years of age, according to Anderson, 2001), an expansion of the repertoire of strategies, the refinement of strategies, or a combination of all factors. According to this view, poor performance in semantic fluency tests (i.e., especially during the latter time slides) may provide warning of impairment to strategic retrieval, cognitive flexibility, or executive control. Further research in this area is needed (e.g., expanding the time interval, a more thorough analysis of response patterns); however, this relative new, alternative scoring method seems very promising in displaying with relative ease the processes underlying overall test performance and studying maturational differences in cognitive development. Also, future studies should further investigate whether this alternative scoring method is also suitable in a clinical practice, while examining, e.g., children with acquired or developmental neurological/psychiatric impairments and/or attention deficits. Previous results using comparable scoring methods are promising in this perspective (Hurks et al., 2004). Hurks et al. found that the total scores (over 60 seconds) on VF tasks did not reveal performance differences between healthy children and ADHD children. In contrast, however, process analyses, as defined in the present article, revealed that children with ADHD symptoms show a delay in the development of automating skills (i.e., performance over the 1–15 seconds of a VF task and more specifically of the initial letter fluency) for processing abstract verbal information. In addition, as mentioned earlier, most studies lacked to examine the influence of age-extrinsic biological factors (e.g., sex) and psychosocial factors (e.g., socioeconomic background, educational background, and/or profession of the caregiver) on maturational differences in performance on VF tasks. Indeed, the present study found that the level of parental education (LPE) appeared, in contrast to the sex of the child, to be a determinant of the child’s performance on semantic VF tasks (primarily the first time slides on task and thereby automatic processing) and of mean cluster size. No significant effects of LPE were found for number of switches/clusters. This indicates that environment primarily affects the automatic, lower cognitive lexico-semantic knowledge, and not so much the higher order cognitive functions. This is the first study that examines the influence of LPE on several different VF outcome measures. Using cluster analysis in the VF task revealed subtle performance differences (and underlying causes) as a function of LPE. In contrast, although clustering and switching have been investigated in clinical populations, few studies have addressed normal aging/developmental and sex differences in these component processes. In the present study, no sex effects nor sex × age interaction effects were found on the semantic VF task. This was in line with our expectations stated in the introduction; although the existing literature is inconsistent regarding sex differences in verbal fluency performance. These conflicting findings with regard to verbal abilities might for one be attributed to the variety of verbal tasks used; each probably involving different cognitive processes. Weiss and colleagues (2006) were for instance the


382


first to suggest that, in an adult population, men and women are using different processing strategies for verbal output on speeded word-generation tasks. However, these sex differences were found on phonological fluency tasks and/or measures of phonological clustering and not on the semantic category fluency test/semantic clustering, which was studied here. Semantic category fluency consists of words with related meaning. Phonological fluency consists of words starting with the same letter/sound or words that rhyme. The former is believed to occur more automatic, relying on common rules of categorization, whereas the latter is more laborious and relies on higher order cognitive functions (Ho et al., 2002; Hurks et al., 2006; Koren et al., 2005). Indeed, retrieval by letter appears to require the exploration of more subsets of words than the retrieval of exemplars of a given semantic category (Monsch et al., 1992; Ober, Dronkers, Koss, Delis, & Friedland, 1986; Sauzéon et al., 2004; Troyer et al., 1997). The differential mechanisms underlying test performance on these types of fluency might explain the inconsistency found. The idea that sex differences may interact with type of task is also supported by the fact that the current study did indeed reveal sex differences with regard to nonverbal design fluency, favoring boys. Next, the variability of behavioral performance may be the cause of conflicting results here. In the present study, only healthy children with average to high IQs were examined. Previous research has shown a strong relationship between IQ and performance-based tests of EF, such that these tests are less sensitive in children with average IQ and above (Mahone et al., 2002), and group effects may be stronger among samples with a wider range of IQ, or among samples of children with ADHD in which a wider range of learning and psychiatric comorbidities are allowed (Wodka et al., 2008). Indeed, Wodka et al. found, while studying a clinically referred group of children, that, when considering ADHD subtype and sex, children with the ADHD subtype less common for sex may demonstrate relative deficits on these process measures. Finally, the effect sizes of the sex differences found are known to be relatively small, which may account for finding no differences between males and females on either total words produced or use of switching and clustering (Lanting, Haugrud, & Crossley, 2009). Unfortunately, our study design has a limitation, in that it does not enable far-reaching generalizations. We did only include a language comprehension task (the vocabulary task) and no measures of broader language abilities (e.g., a standard test of semantic abilities and of general word retrieval abilities under a nonspeeded condition). This is of concern given the known impairments in VF tasks associated with specific language impairments. Next, as mentioned earlier, in literature, the method of analysis of cluster formation is still not standardized, as various definitions of clusters have been reported. In the current study, we choose to examine clustering simply based on zoological families. However, this type of scoring ignores the possibility for a subject to use everyday experience, and thereby “prototypes” (Rosch, 1973), “scripts,” and/or “schemata” (i.e., a mental framework for understanding and remembering information, Mandler, 1984) to organize his or her response in VF tasks. Examples of these alternative cluster strategies are: to recall animals present in a book recently read, to recall a visit to a little zoo, or to recall animals present in the family and relatives. Although new scoring variables have been introduced to address limitations in the conceptualization of clustering as a strategy (e.g., Abwender et al., 2001; Lanting et al., 2009), more research should be conducted to develop a new integrated system to score both objective and subjective categorizing strategies. In this context it would be interesting to include type of fluency tasks as well. As mentioned above, during a semantic fluency task activation of an initial and highly prototypical exemplar leads to automatic activation of closely related semantic neighbors. By



383

contrast, phonological fluency requires the processing of the phonemic characteristics of words according to a given rule (i.e., same first letter). The search process during the latter task is less automatic and necessitates the active generation of a new strategy. More than semantic fluency tasks, the phonological fluency task requires participants to make correct selections, to inhibit intrusions, and to keep a constant level of focused attention (Weiss et al., 2006). It would be highly interesting to see whether alternative, qualitative scores could display these hypothesized differences in underlying cognitive functions more thoroughly. Finally, the limits of interpretation of the current data with regard to construct validity should be noted here. As mentioned in the introduction section, much of the research on qualitative and quantitative outcome measures of word fluency tasks has been done in adults. Assumptions about skills underlying these process variables (i.e., the ability to “automatically” access previously learned semantic information) are possibly very different in adults (who supposedly have already established these skills) than they are for children—particularly younger children (who are still rapidly developing these skills). Ideally, in order to make assumptions about what these process variables are measuring in children who are still developing, we would have conducted a construct validity study. Unfortunately, our study did not include more pure measures of cognitive flexibility (e.g., Wisconsin Card Sorting test, or Trail Making test) to test for convergent construct validity. However, quantitative and qualitative measures of the Design Fluency task revealed comparable results with regard to developmental trajectories and VF measures correlated better with these DF measures than they do with basic vocabulary. According to Abwender et al. (2001), who included a DF task to support construct validity as well, these patterns of differential correlation go a long way in establishing construct validity for these subcomponents of the verbal fluency tasks. In conclusion, differential developmental trajectories were found for automatic processes and controlled processing in 6- to 15-year-old children, while using the VF task as a primary outcome measure. These data suggest that the alternative scoring methods of VF tasks can be used to acquire knowledge on interindividual differences (caused, e.g., by differences in parental education) in the development of higher order cognitive functions (e.g., executive functions, attention capacity) in healthy children. Original manuscript received March 17, 2009 Revised manuscript accepted January 29, 2010 First published online April 6, 2010

REFERENCES Abwender, D. A., Swan, J. G., Bowerman, J. T., & Connolly, S. W. (2001). Qualitative analysis of verbal fluency output: Review and comparison of several scoring methods. Assessment, 8, 323–338. Anderson, P. (2001). Measurement and development of executive function. Unpublished doctoral dissertation, The University of Melbourne, Victoria, Australia. Anderson, P. (2002). Assessment and development of executive function (EF) during childhood. Child Neuropsychology, 8, 71–82. Anderson, P., Anderson, V., Northam, E., & Taylor, H. (2000). Standardization of the contingency naming test for school-aged children: A new measure of reactive flexibility. Clinical Neuropsychological Assessment, 1, 247–273.


384


Barkley, R. A. (2006). Attention-deficit/ hyperactivity disorder: A handbook for diagnosis and treatment (3rd ed.). New York: The Guilford Press. Collins, A. M., & Loftus, E. F. (1975). A spreading-activation theory of semantic processing. Psychological Review, 82, 407–428. Crowe, S. F. (1996). The performance of schizophrenic and depressed participants on tests of fluency: Support for a compromise in dorso-lateral pre-frontal functioning. Australian Psychologist, 31, 204–209. Crowe, S. F. (1998). Decrease in performance on the verbal fluency test as a function of time: Evaluation in a young healthy sample. Journal of Clinical and Experimental Neuropsychology, 20(3), 391–401. De Bie, S. E. (1987). Standaardvragen 1987: Voorstellen voor uniformering van vraagstellingen naar achtergrondkenmerken en interviews [Standard questions 1987: Proposal for uniformization of questions regarding background variables and interview] (2nd ed.). Leiden, The Netherlands: Leiden University Press. De Bruyn, E. E. J., Van der Steene, G., Van Haasen, P. P., Coetsier, P., Pijl, Y. J., Spoelders-Claes, R., et al. (1986). Wechsler Intelligence Scale for Children - Revised (WISC-R). Lisse: Swets & Zeitlinger. Dekker, R., Mulder, J. L., & Dekker, P. H. (2007). De ontwikkeling van vijf nieuwe Nederlandstalige tests. Leiden: PITS. Elvevag, B., Fisher, J. E., Gurd, J. M., & Goldberg, T. E. (2002). Semantic clustering in verbal fluency: Schizophrenic patients versus control participants. Psychological Medicine, 32, 909–917. Goebel, S., Fischer, R., Ferstl, R., & Mehdorn, H. M. (2009). Normative data and psychometric properties for qualitative and quantitative scoring criteria of the Five-Point Test. The Clinical Neuropsychologist, 23(4), 675–690 Ho, A. K., Sahakian, B. J., Robbins, T. W., Barker, R. A., Rosser, A. E., & Hodges, J. R. (2002). Verbal fluency in Huntington’s disease: A longitudinal analysis of phonemic and semantic clustering and switching. Neuropsychologia, 40, 1277–1284. Hurks, P. P. M., Hendriksen, J. G. M., Vles, J. S. H., Kalff, A. C., Feron, F. J. M., Kroes, M., et al. (2004). Verbal fluency over time as a measure of automatic and controlled processing in children with ADHD. Brain and Cognition, 55, 535–544. Hurks, P. P. M., Vles, J. S. H., Hendriksen, J. G. M., Kalff, A. C., Feron, F. J. M., Kroes, M., et al. (2006). Semantic category fluency versus initial letter fluency over 60 seconds as a measure of automatic and controlled processing in healthy school-aged children. Journal of Clinical and Experimental Neuropsychology, 28, 684–695. Kalff, A. C., De Sonneville, L. M. J., Hurks, P. P. M., Hendriksen, J. G. M., Kroes, M., Steyaert, J., et al. (2005). Speed, speed variability, and accuracy of information processing in 5/6-year-old children at risk of ADHD. Journal of the International Neuropsychological Society, 11, 173–183. Kavé, G., Kigel, S., & Kochva, R. (2008). Switching and clustering in verbal fluency tasks throughout childhood. Journal of Clinical and Experimental Neuropsychology, 30, 349–359. Klenberg, L., Korkman, M., & Lahti-Nuuttila, P. (2001). Differential development of attention and executive functions in 3- to 12-year-old Finnish children. Developmental Neuropsychology, 20, 407–428. Koren, R., Kofman, O., & Berger, A. (2005). Analysis of word clustering in verbal fluency of school-aged children. Archives of Clinical Neuropsychology, 20, 1087–1104. Korkman, M., Kemp, S. L., & Kirk, U. (2001). Effects of age on neurocognitive measures of children ages 5 to 12: A cross-sectional study on 800 children from the United States. Developmental Neuropsychology, 20, 331–354. Korkman, M., Kirk, U., & Kemp, S. L. (1998). NEPSY. A developmental neuropsychological assessment. San Antonio, TX: The Psychological Corporation. Kroes, M., Kalff, A. C., Steyaert, J., Kessels, A. G., Feron, F. J., Hendriksen, J. G., et al. (2002). A longitudinal community study: Do psychosocial risk factors and child behavior checklist scores



385

at 5 years of age predict psychiatric diagnoses at a later age? Journal of the American Academy Child and Adolescent Psychiatry, 41, 955–963. Lanting, S., Haugrud, N., & Crossley, M. (2009). The effect of age and sex on clustering and switching during speeded verbal fluency tasks. Journal of the International Neuropsychological Society, 15, 196–204. Levin, H. S., Song, J., Ewing-Cobbs, L., Chapman, S. B., & Mendelsohn, D. (2001). Word fluency in relation to severity of closed head injury, associated frontal brain lesions, and age at injury in children. Neuropsychologia, 39, 122–131. Lezak, M. (2004). Neuropsychological assessment. New York: Oxford University Press. Lupien, S. J., King, S., Meaney, M. J., & McEwen, B. S. (2001). Can poverty get under your skin? Basal cortisol levels and cognitive function in children from low and high socioeconomic status. Development and Psychopathology, 13, 653–676. Mahone, E. M., Hagelthorn, K. M., Cutting, L. E., Schuerholz, L. J., Pelletier, S. F., Rawlins, C., et al. (2002). Effects of IQ on executive function measures in children with ADHD. Child Neuropsychology, 8, 41–51. Mahone, E. M., Koth, C. W., Cutting, L., Singer, H. S., & Denckla, M. B. (2001). Executive function in fluency and recall measures among children with Tourette Syndrome and ADHD. Journal of the International Neuropsychological Society, 7, 102–111. Mandler, J. (1984). Stories, scripts, and scenes: Aspects of schema theory. Hillsdale, NJ: Erlbaum. Martin, A., Wiggs, C. L., Lalonde, F., & Mack, C. (1994). Word retrieval to letter en semantic cues: A double dissociation in normal subjects using interference tasks. Neuropsychologia, 32, 1487–1494. McCulloch, A., & Joshi, H. E. (2001). Neighbourhood and family influences on the cognitive ability of children in the British National Child Development Study. Social Science & Medicine, 53, 579–591. Meijs, C. J., Hurks, P. P. M., Kalff, A. C., Slaats-Willemse, D. I., Rozendaal, N., & Jolles, J. (2008). Differential development of learning strategies on a Pictorial Verbal Learning Test (PVLT) in primary-school children. Child Neuropsychology, 15(3), 1–15. Monsch, A., Bondi, M. W., Butters, N., Salmon, D. P., Katzman, R., & Thal, L. J. (1992). Comparison of verbal fluency tasks in the detection of dementia of the Alzheimer type. Archives of Neurology, 49, 1253–1258. Monsch, A. U., Bondi, M. W., Butters, N., Paulsen, J. S., Salmon, D. P., Brugger, P., et al. (1994). A comparison of category and letter fluency in Alzheimer’s disease and Huntington’s disease. Neuropsychology, 8, 25–30. Ng, K. H., Chong, D., Wong, C. K., Ong, H. T., Lee, C. Y., Lee, B. W., & Shek, L. P. (2004). Central nervous system side effects of first- and second-generation antihistamines in school children with perennial allergic rhinitis: A randomized, double-blind, placebo-controlled comparative study. Pediatrics, 113(2), e116–e121. Ober, B. A., Dronkers, N. F., Koss, E., Delis, D. C., & Friedland, R. P. (1986). Retrieval from semantic memory in Alzheimer-type dementia. Journal of Clinical and Experimental Neuropsychology, 8, 75–92. Rietveld, M. J., Dolan, C. V., van Baal, G. C., & Boomsma, D. I. (2003). A twin study of differentiation of cognitive abilities in childhood. Behavior Genetics, 33, 367–381. Riva, D., Nichelli, F., & Devoti, M. (2000). Developmental aspects of verbal fluency and confrontation naming in children. Brain and Language, 71, 267–284. Robert, P. H., Lafont, V., Medecin, I., Berthet, L., Thauby, S., Baudu, C., et al. (1998). Clustering and switching strategies in verbal fluency tasks: Comparison between schizophrenics and healthy adults. Journal of the International Neuropsychological Society, 4, 539–546. Robert, P. H., Migneco, V., Marmod, D., Chaix, I., Thauby, S., Benoit, M., et al. (1997). Verbal fluency in schizophrenia: The role of semantic clustering in category instance generation. European Psychiatry, 12, 124–129. Roberts, J. E., & Bell, M. A. (2002). The effects of age and sex on mental rotation performance, verbal performance, and brain electrical activity. Developmental Psychobiology, 40, 391–407.


386


Rosch, E. H. (1973). Natural categories. Cognitive Psychology, 4, 328–350. Rosser, A., & Hodges, J. (1994). Initial letter and semantic category fluency in Alzheimer’s disease, Huntington’s disease, and Progressive Supranuclear Palsy. Journal of Neurology, Neurosurgery and Psychiatry, 57, 1389–1394. Sauzéon, H., Lestage, P., Raboutet, C., N’Kaoua, B., & Claverie, B. (2004). Verbal fluency output in children aged 7-16 as a function of the production criterion: Qualitative analysis of clustering, switching processes, and semantic network exploitation. Brain and Language, 89, 192–202. Seymour, P. H., Aro, M., & Erskine, J. M. (2003). Foundation literacy acquisition in European orthographies. British Journal of Psychology, 94(Pt 2), 143–174. Shute, G. E., & Huertas, V. (1990). Developmental variability in frontal lobe function. Developmental Neuropsychology, 6, 1–12. Smith, P. T., & Claxton, G. L. (1972, January). Lexical search and phonemic organisation in memory. Paper presented to the Experimental Psychology Society, London. Troyer, A. K. (2000). Normative data for clustering and switching on verbal fluency tasks. Journal of Clinical and Experimental Neuropsychology, 22, 370–378. Troyer, A. K., Moscovitch, M., & Winocur, G. (1997). Clustering and switching as two components of verbal fluency: Evidence from younger and older healthy adults. Neuropsychology, 11, 138–146. Troyer, A. K., Moscovitch, M., Winocur, G., Alexander, M. P., & Stuss, D. T. (1998). Clustering and switching on verbal fluency: The effects of focal-frontal and temporal-lobe lesions. Neuropsychologia, 36, 499–504. United Nations Educational Scientific and Cultural Organisation (UNESCO). (1997). International Standard Classification of Education (ISCED). Paris: UNESCO. Wassenberg, R., Hendriksen, J. G. M., Hurks, P. P. M., Feron, F. J. M., Keulers, E. H. H., Vles, J. S. H., et al. (2008). Development of inattention, impulsivity, and processing speed as measured by the d2 test: Results of a large cross-sectional study in children aged 7–13. Child Neuropsychology, 14, 195–210. Wassenberg, R., Hurks, P. P. M., Hendriksen, J. G. M., Feron, F. J. M., Meijs, C. J. C., Vles, J. S. H., et al. (2008). Age-related improvement in complex language comprehension: Results of a cross-sectional study with 361 children aged 5 to 15. Journal of Clinical and Experimental Neuropsychology, 30, 435–448. Weiss, E. M., Ragland, J. D., Brensinger, C. M., Bilker, W. B., Deisenhammer, E. A., & Delazer, M. (2006). Sex differences in clustering and switching in verbal fluency tasks. Journal of the International Neuropsychological Society, 12, 502–509. Wixted, J. T., & Rohrer, D. (1994). Analyzing the dynamics of free recall: An integrative review of the empirical literature. Psychonomic Bulletin and Review, 1, 89–106. Wodka, E. L., Mostofsky, S. H., Prahme, C., Larson, J. C. G., Denckla, M. B., & Mahone, E. M. (2008). Process examination of executive function in ADHD: Gender and subtype effects. The Clinical Neuropsychologist, 22, 826–841.


387

APPENDIX A: Semantic fluency Clusters on semantic fluency trials consisted of successively generated words belonging to the same zoological family, as specified below. Examples are listed for each zoological family; although listings are far from exhaustive (in line with Troyer et al., 2000).


Category

Description

Zoological name/Family

Examples

1

Fish

2

Birds

Chondrichthyes, Osteichthyes Aves

3 4 5 6

Reptiles Amphibians Pigs Camels

Sauropsida Amphibia Suidae, Tayassuidae Tylopoda

7

Deers, giraffes, and bovine

Ruminantia

8

Feline

Feliformia

9 10

Canine Bears

11

12

Marine mammals (without whales and dolphins) Weasels and skunks

Canidae Ursidae, Procyonidae, Ailuridae Odobenidae, Otariidae, Phocidae, Sirenia

13 14 15

Whales and dolphins Bats Insectivores:

16

Rabbits & Rodents:

17 18

Horses Primates

Equidae Primates

19

Pachyderm

20 21

Marsupial & platypus Insects

22

Shellfish, Mollusc, & Nettle animals

Proboscidea, Rhinocerotidae, Hippopotamidae, Tapiridae Marsupalia, Australosphenida opossum, koala, wombat, wallaby, platypus ant, beetle, cockroach, flea, fly, praying Chelicerata, Hexapoda, mantis Myriapoda, Annelida, Platyhelminthes Crustacea, Mollusca, oyster, mussel, lobster, scallop, clam Cnidaria

Mepitidae, Mustelidae Cetacea Chiroptera Insectivora, Macroscelidea, Xenarthra Lagomorpha, Rodentia, Scandentia

bass, guppy, salmon, trout budgie, condor, eagle, finch, kiwi, macaw, parrot, parakeet, pelican, penguin, robin, toucan, woodpecker alligator, turtle, snake, crocodile frog, salamander, toad pig, swine, peccary camel, llama, dromedary, alpaca, guanaca, vicuna antelope, caribou, eland, elk, gazelle, gnu, impala, moose, reindeer, giraffe, okapi, bison, buffalo, cow, musk ox, yak bobcat, cat, cheetah, cougar, jaguar, leopard, lion, lynx, mountain lion, ocelot, panther, puma, tiger coyote, dog, fox, jackal, wolf panda, bear, polar bear, raccoon manatees, dugong, walrus, fur seal, sea lion

weasel, skunk, badger, ferret, marten, mink, mongoose, otter, polecat, skunk dolphin, whale, beluga, narwhal aardvark, anteater, hedgehog, mole, shrew coney, hare, pika, rabbit, beaver, chinchilla, chipmunk, gerbil, gopher, groundhog, guinea pig, hamster, hedgehog, marmot, mole, mouse, muskrat, porcupine, rat, squirrel, woodchuck horse, donkey, zebra ape, baboon, chimpanzee, gibbon, gorilla, human, lemur, marmoset, monkey, orangutan, shrew elephant, rhinoceros, tapir