Working memory, intelligence and knowledge base in adult persons with intellectual disability

Research in Developmental Disabilities 23 (2002) 105±118

Working memory, intelligence and knowledge base in adult persons with intellectual disability H. Numminena,*, E. Serviceb,1, I. Ruoppilaa a

Department of Psychology, University of JyvaÈskylaÈ, P.O. Box 35, 40351 JyvaÈskylaÈ, Finland b Department of Psychology, University of Helsinki, P.O. Box 13, 00014 Helsinki, Finland

Received 10 May 2001; received in revised form 26 October 2001; accepted 13 December 2001

Abstract Previous studies have suggested that performance in working memory (WM) tasks is de®cient in all etiologies and at all levels of intellectual disability (ID). Knowledge about WM structure, cognitive processes re¯ected in WM tasks, or the long-term memory contribution to WM capacity in ID is, however, not satisfactory. In the present study, WM capacity, WM task requirements, as well as effects between WM, skills, knowledge base, and intelligence were explored in two groups with matched ¯uid intelligence: adult persons with ID and normally developing children aged 3±6 years. The ID Group performed equally well as the children in WM tasks based on familiar semantic information and were signi®cantly better on all measures re¯ecting skills and knowledge base. The Child Group performed better in phonological and visuo-spatial WM tasks including nonsemantic information, respectively. In particular, it appeared that the groups differed in their WM performance although they were matched for ¯uid intelligence. We hypothesize that the ID Group depended more on knowledge support from long-term memory whereas the Child Group could bene®t more from ef®cient online WM processes. # 2002 Elsevier Science Ltd. All rights reserved. Keywords: intellectual disability; working memory; long-term memory; ¯uid intelligence

1. Introduction Working memory (WM) refers to a cognitive system responsible for temporary storage and manipulation of information during cognitive activities. According to *

Corresponding author. Tel.: ‡358-9-8273337; fax: ‡358-9-8273337. E-mail addresses: heli.numminen@surfeu.® (H. Numminen), elisabet.service@helsinki.® (E. Service). 1 Co-corresponding author. 0891-4222/02/$ ± see front matter # 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 8 9 1 - 4 2 2 2 ( 0 2 ) 0 0 0 8 9 - 6

106

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

the in¯uential WM framework by Baddeley and Hitch (1974), WM comprises the Central Executive, and at least two sub-components, specialized to modality speci®c information processing: the Phonological Loop for verbal and phonological information, and the visuo-spatial Sketchpad for visual and spatial information. Lately, a modality-neutral storage component, the Episodic Buffer, has been put forward to deal with abstract representations of events (Baddeley, 2000). According to the WM framework, the importance of WM lies in its multiple connections with cognitive operations rather than memory per se. In general populations, these connections have been established in many studies: the Central Executive has been suggested to be related, for instance, to action planning (Baddeley, 1996; Norman & Shallice, 1980), ¯uid intelligence (Duncan, Emslie, Williams, Johnson, & Freer, 1996; Engle, Tuholski, Laughlin, & Conway, 1999), and reasoning (Kyllonen & Christal, 1990). The Phonological Loop has been reported to be involved, for example, in vocabulary acquisition and language development in general (see Gathercole & Baddeley, 1993 and Baddeley, Gathercole, & Papagno, 1998 for reviews), and in foreign-language acquisition (Cheung, 1996; Service, 1992). The visuo-spatial Sketchpad has been found to be related, for instance, to geographical orientation and spatial planning skills (e.g., Logie, 1995), respectively.TheconnectionsbetweencognitiveactivitiesandWMare suggested to be reciprocal, as skills and cognitive functions relying on the knowledge base of long-term memory have been found to affect WM performance (e.g., Hulme, Maughan, & Brown, 1991). In verbal WM tasks, semantically familiar items are easier to remember than unfamiliar items. Reciprocal effects are also reported in developmental studies, as the Phonological Loop and the growing body of lexical knowledge have been found to develop in reciprocal co-operation. It has been proposed that after 5 years of age, vocabulary size begins to determine Phonological Loop capacity rather than vice versa (e.g., Gathercole, Willis, Emslie, & Baddeley, 1992; Gathercole, Service, Hitch, Adams, & Martin, 1999). Most previous studies have suggested that all persons with intellectual disability (ID) share restricted capacity for tasks requiring involvement of short-term memory or WM, irrespective of level of handicap, or etiology of ID (e.g., Hulme & Mackenzie, 1992; Pulsifer, 1996). Some studies have also reported that in ID WM capacity lags behind general intellectual performance (e.g., Hulme & Mackenzie, 1992; Mackenzie & Hulme, 1987; Marcell & Weeks, 1988; Pulsifer, 1996; but see Vicari, Carlesimo, & Caltagirone, 1995 for contradictory results). In addition, the WM de®cit related to ID has been found to increase with development (e.g., Hulme & Mackenzie, 1992), and be qualitatively different in distinct ID etiologies (e.g., Jarrold & Baddeley, 1997; Numminen, Service, Ahonen, & Ruoppila, 2001; Vicari et al., 1995; Wang & Bellugi, 1994), and individuals (Numminen, Service, Ahonen, Korhonen, Tolvanen, Patja, & Ruoppila, 2000). WM performance has, in turn, been found to be related to many everyday cognitive skills in ID, such as vocabulary (e.g., Grant, Karmiloff-Smith, Gathercole, Paterson, Howlin, Davies, & Udwin, 1997); Jarrold & Baddeley, 1997), and reading and writing ability (e.g., Numminen et al., 2000). Current

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

107

knowledge of the reciprocal effects between WM, skills, and knowledge base in ID is still, however, only patchy. Also, somewhat contradictory hypotheses have been proposed for the relationship between WM and intelligence. Hulme and Roodenrys (1995) have argued that the verbal short-term memory de®cit found in ID, may be closely related to the general restriction in intellectual capacity. This suggests the hypothesis that matching groups for intelligence level should also equate their verbal short-term memory performance. The WM framework by Baddeley and Hitch leads to a somewhat contradictory hypothesis. According to this theory, WM does not comprise a unitary system, and each component may have separate connections to intelligence and independent relations to cognitive processes. This study addresses the second hypothesis. The majority of studies of immediate memory have assumed that WM tasks measure more or less the same memory components and cognitive functions both in ID and in the general population. However, in a previous study, we found the WM performance of persons with ID to consist of two rather than three components, which we called Phonological WM and General WM (Numminen et al., 2000). Not surprisingly, Phonological WM re¯ected the WM tasks selected to assess the Phonological Loop. It was not related to the intelligence measures used, or the tasks that loaded on the General WM component. General WM re¯ected both performance on tasks selected for visuo-spatial Sketchpad and Central Executive assessment (cf. Gathercole & Pickering, 2000), and it was signi®cantly connected to intelligence. As an exception to this pattern, the Digit Span forwards task was moderately connected to both components in the study group. We hypothesized that the results may signal more active involvement of controlled attention (i.e., the Central Executive) in visuo-spatial WM tasks and possibly also in the Digit Span forwards task in ID than in the general population, as well as independence of Phonological WM from general intelligence. Gathercole and Pickering (2000) have recently argued that simple pairing of WM tasks and components may not be fruitful in the evaluation of WM in young children. Their study of 6±7-year-old children suggested that the tasks for Visuospatial Sketchpad assessment that they used required strong involvement of the Central Executive in this age group. Engle and his colleagues (1999) propose, that activation of either rudimentary or more complex memory components in speci®c memory tasks depends on a person's age, intelligence, and developmental level as well as on the task itself. They imply, that even simple memory tasks may require strong involvement of controlled attention at lower developmental levels. As development proceeds, the need for controlled attention involvement decreases. According to many researchers, for instance Cowan (1995), Baddeley (1986), and Baddeley and Della Sala (1996), controlled attention is re¯ected in the Central Executive process of WM (cf. Baddeley, 1996; Baddeley & Hitch, 1974). It is also seen as a connecting factor relating WM to ¯uid intelligence (cf. Cattell, 1963). According to Swanson (1996), changes in the capacity of the central processor, i.e., the Central Executive, re¯ect most accurately the developmental changes and the individual differences in WM in childhood. We hypothesize, that in ID the Central Executive is more actively involved in WM tasks than in the general adult

108

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

population. However, we also propose that differences of controlled attention involvement may also have been found in comparisons between persons with ID and normally developing children with equal intelligence. Long-term memory contribution to WM has rarely been taken into consideration in ID. It may, however, be useful, as WM studies usually compare teen-aged or adult ID groups to normally developing child groups. This situation may lead to differences in skills and the long-term knowledge base, usually favoring the older ID groups. For instance, the larger vocabulary of older persons with ID may signi®cantly affect performance, hindering the detection of the differences in the on-line processes of WM. The aim of the present study was to explore if adult persons with ID share equal WM capacity and pro®le with normally developing children matched for ¯uid intelligence. This addresses the hypothesis that matching for intelligence does not necessarily result in abolition of all WM differences. We were also interested to ®nd out, how possible skill and knowledge base differences that must exist between groups of very different ages, affect WM performance. We hypothesized, that differences in WM capacity may exist even between groups of matched intelligence. These ®ndings may re¯ect differences in connections between WM components and intelligence, differences in the cognitive processes re¯ected in the WM tasks, or differences in the skills and knowledge base between the groups. This factor was investigated using correlational and logistic regression analysis. 2. Method The ID Group was selected from a population-based sample (n ˆ 253) participating in a 36-year follow-up study. The principles of the base sample selection and data gathering have been reported previously (Numminen et al., 2000). Ten men and 14 women with unknown etiology of ID, a sub-sample of the sample reported by Numminen et al. (2000), were selected for the present study. Unknown etiology was de®ned as idiopathic ID. Criteria for selection for the ID Group for this study were: (1) chronological age under 60, (2) absence of severe neurological signs, (3) ability to follow test instructions, (4) ability to communicate verbally, (5) absence of acute psychiatric disorders or dementia, and (6) high level of everyday functioning, operationalized as noninstitutional living and attendance of working or other activities. Persons in the ID Group were matched for their individual ¯uid intelligence scores (estimated by Raven's Coloured Matrices series A, Ab, and B, Raven, 1956), to normally developing children: 13 boys and 11 girls recruited from a local day-care center. The chronological ages of the children ranged from 3 to 6. Raven's test was selected for the matching procedure because of the differences in chronological age, environmental background, and life-history of the two groups (Carpenter, Just, & Shell, 1990). The WM tasks were chosen to represent the original triarchic division of the WM framework by Baddeley and Hitch (1974). The Digit Span backwards task

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

109

was selected to measure Central Executive functions (cf. Baddeley, 1996). The Nonword Repetition task, the Nonword Span task, and the Digit Span forwards task were selected to measure Phonological Loop functions (cf. Gathercole & Baddeley, 1993). The Corsi Blocks task was selected to measure visuo-spatial Sketchpad functions (cf. Logie, 1995). The length of the longest sequence recalled (span) was used as a measure of capacity in all tasks except the Nonword Repetition task in which the number of total points received (number of successful trials) was scored. Answering time was not limited and performance speed had no effect on scoring in any of the tasks. The Digit Span backwards task was presented and scored as recommended in the WAIS-R manual (Wechsler, 1981). The Finnish Nonword Repetition task (cf. Ceponiene, Service, Kurjenluoma, Cheour, & NaÈaÈtaÈnen, 1999) was constructed from nonwords of four to ®ve syllables. The nonwords corresponded to Finnish language words regarding phonotactic structure, and pronunciation. For attentional reasons, the tasks were slightly different for adults and children. For the ID Group 40 nonwords were presented as pairs. For the Child Group 20 nonwords were presented one at a time, one in each pair from the list of pairs. The nonwords were presented from tape and the participant was asked to immediately repeat them. Responses were taperecorded for later scoring. One point was given for each perfectly repeated nonword. Recurring problems in articulation or order of the recall had no effect on scoring. The Finnish Nonword Span task (modi®ed from Service, 1998) was based on a pool of 12 recurring four-phoneme nonwords (consonant±vowel±consonant± vowel, e.g., /tepa/, /raku/, /hine/). The nonwords were phonotactically legal in Finnish. Participants were allowed to read or hear the list of nonwords in alphabetical order beforehand. The nonwords were presented in sequences, starting from sequence length two. Sequence length increased by one after every ®ve attempts. The nonwords were presented at a one-per-second rate from tape and the participants recalled them immediately. A correct response was scored for a fully repeated sequence in which all phonemes were correct and the nonwords in correct order. Testing was ®nished when the participant failed to correctly repeat at least three of the ®ve sequences of one sequence length. Span was estimated as one below the last sequence length tested and was increased by half a point for two correct repetitions of the last sequence length. Recurring problems in articulation had no effect on scoring. The Digit Span forwards task was presented and scored as recommended in the WAIS-R manual (Wechsler, 1981). The only exception was that two-digit sequences were added, and the task was begun from two instead of three in order to prevent a possible ¯oor-effect in the task. The Corsi Blocks task (e.g., Milner, 1971) included 15 black wooden blocks spaced equidistantly on a white board. The test administrator pointed at a (sequence of) block(s), the rote of which the participant was told to immediately imitate. The task was begun from sequence length one and the sequences were determined in advance so that they appeared to be random. Subsequent

110

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

administration, ®nishing criteria, and scoring were identical to the Digit Span forwards task. The assessed aspects of skills and knowledge base included measures of reading, writing, and basic mathematical skills as well as a test of receptive Vocabulary (Ruoppila, 1964). The reading test included eight single words (nouns, adjectives, verbs, pronouns) and a meaningful text of 35 words. One point was given for each correctly read word. Recurring problems in articulation did not affect scoring. The writing test included the writing of one's name (®rst and last name), home town, and sex, spelling of six dictated words (nouns, adjectives, verbs, pronouns), and spontaneous writing elicited by a picture. The basic mathematical skills that were assessed included knowledge of the names of the numbers as well as their order, basic counting operations, tested by asking the participants to report the number of objects presented in a picture, ability to tell the time, knowledge of the age of persons, and knowledge of money and its use. Compound variables of Reading, Writing, and Mathematical skills were used in the analyses. The vocabulary test included 13 cards with four pictures each. As each card was shown to the participant, one to ®ve words were read aloud. The participants had to indicate by pointing to which of the four pictures each of the heard nouns or adjectives was related. A total of 68 test items were presented to the ID Group in a sequence unrelated to dif®culty. In the Child Group a shorter form of the test with a total of 30 test items was employed. To achieve comparable measures, the ®nal scores for the Child Group were extrapolated from the shorter test by dividing their original scores with 30 and multiplying the result with 68. In general, inter-item reliability was estimated by calculating Cronbach's alpha. As most WM measures had cut-off points, and the tests were not performed entirely, we assessed the reliabilities of these tests by correlations calculated between the length of the longest sequence recalled (span) and the total score (number of successful trials) attained during the test. In cases where the distributions of the variables were normal, differences between the groups were calculated by t-tests, whereas for those variables that were not normally distributed the nonparametric Mann±Whitney U-test for two independent samples was employed. Pearson's correlation coef®cients were computed for those variables that were normally distributed. For the remaining variables the reported coef®cients are Spearman's correlations. Partial correlations were calculated for Pearson correlations. Signi®cance levels of the differences between correlation coef®cients were calculated by using a z transformation. The correlative connections were further analyzed by logistic regression. 3. Results Normality of distributions as well as reliability estimates of the tests and tasks employed are presented in Table 1. Mean chronological age in the ID Group was 49.88 years (SD ˆ 3:76) and in the Child Group 5.50 years (SD ˆ 1:73). In the ID Group Raven was not related to

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

111

Table 1 Reliability estimates for the WM and knowledge base tasks in the study groups (ID: ID Group, CG: Child Group, N: normal distribution) Task

Distribution (ID)

Distribution (CG)

Reliability (ID)

Reliability (CG)

Intelligence Raven

N

N

.84

.73

WM tasks Digit Span backwards Nonword Repetition Nonword Span Digit Span forwards Corsi Blocks

N N N N N

N Not normal Not normal N N

.94a,** .90 .88a,** .87a,** .95a,**

.95a,** .88 .91a,** .97a,** .90a,**

Knowledge base Vocabulary Reading Writing Mathematical skills

N Not normal N N

N Not normal N N

.86 .94 .88 .80

.54 .90 .54 .74

a

Reliability estimated by correlation. p < :01.

**

chronological age (r ˆ :01, p ˆ 0:95), whereas in the Child Group the two were signi®cantly related (r ˆ :48, p < :05). As expected, chronological age was not related to any of the WM tasks in the ID Group but was related to all WM measures in the Child Group (Digit Span backwards: r ˆ :59, p < :01; Nonword Repetition: r ˆ :53, p < :01; Nonword Span: r ˆ :71, p < :01; Digit Span forwards: r ˆ :50, p < :01; Corsi Blocks: r ˆ :42, p < :05). Raven was related to Digit Span backwards (r ˆ :47, p < :05) and to Corsi Blocks (r ˆ :79, p < :01) in the ID Group but was not related to any of the single WM tasks in the Child Group. The sex of the participants did not affect any of the tasks employed in either group. Mean performance of the two groups in WM and everyday cognition tasks was compared (Table 2). The groups performed equally well in Digit Span forwards and backwards tasks. The Child Group performed signi®cantly better in the Corsi Blocks task, and in the Nonword Span task. Performance in the Nonword Repetition task could not be directly compared because of the different forms of the task, but the data suggest that the Child Group may perform better also in this task. As expected, the ID Group had better reading, writing, and basic mathematical skills and also scored better in the Vocabulary test than the Child Group. Correlation coef®cients were calculated among the WM tasks (Table 3). In addition, partial correlation coef®cients with chronological age controlled are shown to reveal the correlations after the age-based variation, affecting especially the children, is removed. Digit Span forwards was more related to Nonword Repetition ( p < :05) and Nonword Span ( p < :05) in the Child Group than in the ID Group. Corsi Blocks, with age partialled out, was more related to the Digit

112

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

Table 2 Descriptive statistics and results of the t-tests and nonparametric Mann and Whitney U-tests in WM and knowledge base tasks Task WM tasks Digit Span backwards Nonword Repetitiona Nonword Span Digit Span forwards Corsi Blocks Knowledge base Vocabulary Reading Writing Mathematical skills

Group Range

Mean (SD)

Test-value

ID CG ID CG ID CG ID CG ID CG

1.58 1.79 9.06 13.75 2.17 2.88 3.88 3.88 3.08 3.75

(1.25) (1.18) (4.02) (5.12) (.82) (.90) (.74) (.95) (1.25) (.79)

t (46) ˆ .59

52.08 33.24 30.61 1.96 8.25 1.96 26.75 18.79

(7.99) t (46) ˆ (8.18) (17.19) Zˆ (7.51) (4.95) Zˆ (2.03) (8.48) Zˆ (5.92)

ID CG ID CG ID CG ID CG

0±4 0±4 0±16.5 2±20 0±4 0±4 3±5 2±5 0±6 2±5 27±63 16±45 0±43 0±36 0±13 0±9 7±36 10±31

Signi®cance .55

Comparison not valid t (46) ˆ 2.86 t (46) ˆ .001 t (46) ˆ 2.21

.006** 1.00 .03*

8.07

.001**

5.30

.001**

4.10

.001**

3.34

.001**

a Nonwords were presented in pairs for ID and as single nonwords for CG. Results presented here were calculated for the same nonwords (i.e., only one in each pair for the ID Group). * p < :05. ** p < :01.

Span backwards in the ID Group than in the Child Group ( p < :05). The general pattern seen in both sets of correlation coef®cients is the same although the relatively small n results in fewer signi®cant partial coef®cients. It is also important to remember that age and ¯uid intelligence covaried for the children. Table 3 Intercorrelation coef®cients between different WM tasks Group

ID/CG

ID/CG

ID/CG

ID/CG

ID/CG

WM tasks

1

2

3

4

5

.34/ .01

.08/ .14 .63**/.49*

.15/.27 .03/.30 .03/.31

.59**/.02 .25/.31 .03/ .06 .10/ .10

1. 2. 3. 4. 5.

Digit Span backwards Nonword Repetition Nonword Span Digit Span forwards Corsi Blocks

.36/.31 .09/.49* .13/.48* .58**/.27

.61**/.67** .01/.49* .24/.47*

.04/.57** .03/.26

.11/.13

In the upper part of the table the coef®cients are partial correlation coef®cients (chronological age partialled out). The lower part of the table presents the raw correlations. Signi®cant differences between the coef®cients are marked by bold numbers. * p < :05. ** p < :01.

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

113

Analysis of logistic regression that attempted to predict group membership was employed to the explorative assessment of possible differences between the groups in multiple connections among different WM tasks as well as connections between WM tasks and mental age. The dependent variable was group (ID/Child) membership. All WM tasks, except the Nonword Repetition task, were included into the analysis as well as interaction variables generated from pairs of WM tasks and WM tasks paired with Raven. As a ®rst step, all variables were entered into the analysis and subjected to the backward Wald method with alpha level set at .05. The Corsi Blocks task and the Nonword Span task as well as the interaction variables of Corsi Blocks and Raven and Nonword Span and Raven signi®cantly discriminated between the two groups. As a second step, these variables were standardized and selected for a new logistic regression analysis (enter method). In addition, Raven was forced into the analysis to control for possible minimal differences in ¯uid intelligence between the groups. The regression model was highly signi®cant, Chi-square …5† ˆ 31:68, p ˆ :001. The model coef®cients are presented in Table 4. Interactions between the variables in the logistic regression were plotted to show the interaction of the Raven score with both Corsi Blocks and Nonword Span (Fig. 1). Poor performance in both Corsi Blocks and Nonword Span was always connected to ID. Good performance in both tasks was only achieved by children (Fig. 1). At a lower intellectual level, poor performance in the Corsi Blocks task, irrespective of Nonword Span performance, predicted belonging to the ID Group. In contrast, good performance in the Corsi Blocks irrespective of Nonword Span performance indicated belonging to the Child Group. At a higher intellectual level, poor Nonword Span performance, irrespective of Corsi Blocks performance, predicted belonging to the ID Group and good Nonword Span performance, irrespective of Corsi Blocks performance, predicted belonging to the Child Group. Thus, at a low intelligence level, the discriminating task between the groups was Corsi Blocks, at a higher intelligence level the discriminating task between the groups was Nonword Span. This analysis of a relatively small sample was checked by adding 22 participants with ID (®nal n of ID Group ˆ 46, sample of our previous study, Numminen et al., 2000). The resulting larger group, even if not matched with the Child Group, did not signi®cantly differ from it on Raven Table 4 Results of the logistic regression analysis, with Group membership (0 ˆ CG, 1 ˆ ID) as dependent variable Variable Raven Nonword Span Corsi Blocks Nonword Span and Raven Corsi Blocks and Raven Constant

B .29 1.59 1.38 2.04 2.01 0.42

Error of B

Wald

p

Odds ratio

.54 .63 .69 .85 .77 .48

.29 6.39 3.94 5.81 6.79 .78

.59 .01 .05 .02 .01 .38

1.34 .20 .25 .13 7.46

114

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

Fig. 1. Probability of belonging to the ID Group as a function of ¯uid intelligence in standard units and performance in the Corsi Blocks (CB) and Nonword Span (NS) tasks in the combined group of adults with ID and children. The (‡) symbols indicate persons performing above mean and the ( ) symbols persons performing below the mean of the combined group.

Scores, t …68† ˆ :87, p ˆ 0:39. Inclusion of the additional participants did not in any way change the pattern of results in the logistic regression analysis. 4. Discussion The present study compared the WM capacity and pro®le of two groups, an adult ID group and a Child Group, of equal ¯uid intelligence but highly different chronological age. The groups performed equally well on both Digit Span tasks. Equal performance in the Digit Span forwards task was a contradictory ®nding compared to many previous studies (e.g., Hulme & Mackenzie, 1992; Marcell & Weeks, 1988), although similar results have also been reported (Vicari et al., 1995). According to the WM framework, equal performance in the Digit Span forwards task could re¯ect equal storage capacity of phonological short-term memory for both groups, as possibly neither of the groups is capable of active rehearsal or other complex processing in the task (cf. Gathercole & Hitch, 1993; Hulme & Mackenzie, 1992). According to the WM framework, as well as a previous study of ID (Vicari et al., 1995), equal performance in the Digit Span backwards task could be regarded to re¯ect similar capacity of the Central Executive in the two groups, respectively. We suggest, that the conclusion of equal capacity for phonological (Digit Span forwards) and executive (Digit Span backwards) WM in the two groups may be premature for several reasons. First, the digit span tasks differed in their pro®le of intercorrelations with other studied variables in the two groups. It is therefore

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

115

possible that the equality of performance in these tasks depended on differential involvement of WM components for the two groups. In contrast to many previous studies with general populations, the Digit Span forwards task did not correlate signi®cantly with other phonological WM tasks (see e.g., Gathercole & Baddeley, 1993 for a review) in the ID Group. This phenomenon was also seen in our previous study with a larger study group (n ˆ 46, Numminen et al., 2000). Perhaps this occurred because in Finnish many number words take much longer to articulate than for instance in English (three have three syllables). This may decrease the probability that they are phonologically coded, especially in lowspan populations, and increase the involvement of controlled attention in the task. Controlled attention in the groups may have been affected by the intelligence matching procedure (a measure of ¯uid intelligence). Especially differences in executive WM capacity may have been abolished. Secondly, differences in skills and knowledge base (vocabulary, reading, writing, and basic mathematical skills, favoring the ID Group) may also have affected the results by compensating for WM capacity in the ID Group as both the phonological shapes of digit words and their conceptual meanings could have been more familiar to the ID Group. The Child Group performed better in the Nonword Span task and in the Corsi Blocks task. Although the results in the Nonword Repetition task could not be directly compared, it appeared that the performance of the Child Group was superior also in this task compared to that of the ID Group. Nonword Span and Corsi Blocks share a serial nature and recall procedure with the Digit Span tasks and may encourage rehearsal activation. However, they also include nonlexical and nonsemantic information that is not supported by a semantic knowledge base. This seems to have accentuated differences between the groups. We suggest, assuming minimal use of rehearsal processes in both groups (cf. Gathercole & Hitch, 1993), that children had better capacity for such WM tasks that do not greatly bene®t from a well-developed knowledge base. This may signal more effective use of strategies in the Child Group, or qualitative differences in the mental representations of the memorized stimuli between the groups. These in turn, may re¯ect, for instance, a better capacity for auditory rehearsal processes in the Phonological Loop (cf. Hulme & Mackenzie, 1992). One could also think of the difference in terms of dependence on long-term memory. The children could be described as having more ``pure'' WM capacity. These data highlight problems with matching adults, or adolescent teenagers, with ID with young children of variable age. This is speci®cally important as the groups of persons with ID are generally compared with groups of normally developing children or adults with highly variable chronological age, educational background, and in child groups also high variation in the phase of WM development. The present study reveals the need for more accurate analysis of WM functions in ID, in which the developmental nature of WM capacity as well as the knowledge base of the compared groups have been taken into consideration. Exploratory evaluation of the connections between intelligence and the WM tasks suggested that the mean differences in Corsi Blocks and Nonword Span tasks between the groups, favoring the Child Group, depended on intelligence

116

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

level. The Corsi Blocks task discriminated the groups at a lower intelligence level, and the Nonword Span task at a higher intelligence level. This may signal qualitative changes in the cognitive processes re¯ected in WM tasks in the groups, which may occur with ¯uid intelligence variation in the ID Group, and chronological age variation in the Child Group. We investigated correlations in the larger ID sample (n ˆ 46) of our previous study (Numminen et al., 2000), with similar distribution of ¯uid intelligence, and compared the bottom and top halves of a median-split based on the Raven scores. Corsi Blocks was more related to two executive WM tasks, Digit Span backwards ( p ˆ :017) and Complex Span (modi®ed from Daneman & Carpenter, 1980) ( p ˆ :035) in the poorer compared to the better half. It seems possible that in ID, controlled attention is required more often in visuo-spatial WM tasks, especially at lower intelligence levels. This may consume storage capacity for memorized stimuli in the ID in comparison to the children of equal ¯uid intelligence (cf. Swanson, 1996). This may re¯ect qualitative differences in WM processing between the two groups. It is also possible, that at lower intelligence level, compensatory functioning of the knowledge base (i.e., larger vocabulary) in the ID Group may be able to support performance in the Nonword Span task (cf. Gathercole, 1995). At a higher intelligence level, the compensatory functioning may be insuf®cient to overcome the superior performance of the Child Group, resulting, for instance, from more adequate rehearsal activation. These questions need further studies. To conclude, the present results suggest that although the two groups were matched for ¯uid intelligence their WM capacities differed in some sense. We ®nd it likely that the found differences stem from qualitative differences in the WM tasks employed. Those which were more independent of prior knowledge or semantic memory were better able to discriminate between the groups. It also seemed that executive WM skills were highly related to ¯uid intelligence in the ID Group, whereas in the Child Group the connection was strongly mediated by chronological age. The present pattern of results may re¯ect qualitative differences in WM processes between the two groups. These in turn may result from differences in skills and size and contents of knowledge base (favoring adults). There could also be variation in the involvement of controlled attention in different WM tasks and at different stages in development or levels of intelligence. Our results further support criticisms against mechanical pairing of WM tasks and cognitive processes in special populations and young children. Acknowledgments The study was ®nanced by the Academy of Finland (grants no. 752724, 39253, and 174080). We thank all study participants and persons who took part in the data collection. Special thanks to Asko Tolvanen who offered his superior help in the statistical analysis of the data, to Timo Ahonen who kindly commented on the manuscript, and to psychologists Lilli Ahonen, who collected the child data, and Marika Liimatainen, who constructed the original nonword repetition list.

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

117

References Baddeley, A. D. (1986). Working memory. Oxford: Oxford University Press. Baddeley, A. D. (1996). Exploring the central executive. Quarterly Journal of Experimental Psychology, 49A, 5±28. Baddeley, A. D. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Sciences, 11, 417±423. Baddeley, A. D., & Della Sala, S. (1996). Working memory and executive control. Philosophical Transactions of the Royal Society of London B, 351, 1397±1404. Baddeley, A. D., Gathercole, S., & Papagno, C. (1998). The Phonological Loop as a language learning device. Psychological Review, 105, 158±173. Baddeley, A. D., & Hitch, G. J. (1974). Working memory. In G. Bower (Ed.), The psychology of learning and motivation (Vol. 8, pp. 47±90). New York: Academic Press. Carpenter, P. A., Just, M. A., & Shell, P. (1990). What one intelligence test measures: A theoretical account of the processing in the Raven Progressive Matrices test. Psychological Review, 97, 404± 431. Cattell, R. B. (1963). Theory of ¯uid and crystallized intelligence: A critical experiment. Journal of Educational Psychology, 54, 1±22. Ceponiene, R., Service, E., Kurjenluoma, S., Cheour, M., & NaÈaÈtaÈnen, R. (1999). Children's performance on pseudoword repetition depends on auditory trace quality: Evidence from eventrelated potentials. Developmental Psychology, 35, 709±720. Cheung, H. (1996). Nonword Span as a predictor of second-language vocabulary learning. Developmental Psychology, 32, 867±873. Cowan, N. (1995). Attention and memory: An integrated framework. Oxford: Oxford University Press. Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450±466. Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). Intelligence and the frontal lobe: The organization of goal-directed behavior. Cognitive Psychology, 30, 257±303. Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999). Working memory, shortterm memory and general ¯uid intelligence: A latent variable approach. Journal of Experimental Psychology: General, 128, 309±331. Gathercole, S. E. (1995). Is nonword repetition a test of phonological memory or long-term knowledge? It all depends on the nonwords. Memory and Cognition, 23, 83±94. Gathercole, S. E., & Baddeley, A. D. (1993). Working memory and language. Hove: Lawrence Erlbaum Associates. Gathercole, S. E., & Hitch, G. J. (1993). Developmental changes in short-term memory: A revised working memory perspective. In A. Collins, S. E. Gathercole, M. A. Conway, & P. E. Morris (Eds.), Theories of memory (pp. 189±210). Hillsdale, NJ: Lawrence Erlbaum Associates. Gathercole, S. E., & Pickering, S. J. (2000). Assessment of working memory in 6-and 7-year-old children. Journal of Educational Psychology, 92, 377±390. Gathercole, S. E., Service, E., Hitch, G. J., Adams, A. -M., & Martin, A. J. (1999). Phonological short-term memory and vocabulary development: Further evidence on the nature of the relationship. Applied Cognitive Psychology, 13, 65±77. Gathercole, S. E., Willis, C. S., Emslie, H., & Baddeley, A. D. (1992). Phonological memory and vocabulary development during the early school years: A longitudinal study. Developmental Psychology, 28, 887±898. Grant, J., Karmiloff-Smith, A., Gathercole, S. E., Paterson, S., Howlin, P., Davies, M., & Udwin, O. (1997). Phonological short-term memory and its relationship to language in Williams Syndrome. Cognitive Neuropsychiatry, 2, 81±99. Hulme, C., & Mackenzie, S. (1992). Working memory and severe learning dif®culties. Hove: Lawrence Erlbaum Associates. Hulme, C., Maughan, S., & Brown, C. D. A. (1991). Memory for familiar and unfamiliar words:

118

H. Numminen et al. / Research in Developmental Disabilities 23 (2002) 105±118

Evidence for a long-term memory contribution to the short-term memory span. Journal of Memory and Language, 30, 685±701. Hulme, C., & Roodenrys, S. (1995). Practitioner review: Verbal working memory development and its disorders. Journal of Child Psychology and Psychiatry, 36, 373±398. Jarrold, C., & Baddeley, A. D. (1997). Short-term memory for verbal and visuo-spatial information in Down's syndrome. Cognitive Neuropsychiatry, 2, 101±122. Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more than) working memory capacity? Intelligence, 14, 389±433. Logie, R. H. (1995). Visuo-spatial working memory. Hove: Lawrence Erlbaum Associates. Mackenzie, S., & Hulme, C. (1987). Memory span development in Down's syndrome severely subnormal and normal subjects. Cognitive Neuropsychology, 4, 303±319. Marcell, M. M., & Weeks, S. L. (1988). Short-term memory dif®culties and Down's syndrome. Journal of Mental De®ciency Research, 32, 153±162. Milner, B. (1971). Interhemispheric differences in the localisation of psychological processes in man. British Medical Bulletin, 27, 272±277. Norman, D. A., & Shallice, T. (1980). Attention to action: Willed and automatic control of behavior. University of California at San Diego, CHIP Report 99. Numminen, H., Service, E., Ahonen, T., Korhonen, T., Tolvanen, A., Patja, K., & Ruoppila, I. (2000). Working memory structure and intellectual disability. Journal of Intellectual Disability Research, 44, 579±590. Numminen, H., Service, E., Ahonen, T., & Ruoppila, I. (2001). Working memory and everyday cognition in adult persons with Down's syndrome. Journal of Intellectual Disability Research, 45, 157±168. Pulsifer, M. B. (1996). The neuropsychology of mental retardation. Journal of the International Neuropsychological Society, 2, 159±176. Raven, J. C. (1956). Coloured progressive matrices: Revised order. London: Lewis and Co. Ruoppila, I. (1964). Test of passive vocabulary for 7±12-year-old. (Kuvasanavarastotesti 7±12vuotiaille). JyvaÈskylaÈ: Kasvatustieteiden tutkimuskeskus. Service, E. (1992). Phonology, working memory, and foreign-language learning. Quarterly Journal of Experimental Psychology, 45A, 21±50. Service, E. (1998). The effect of word length on immediate serial recall depends on phonological complexity, not articulatory duration. Quarterly Journal of Experimental Psychology, 51A, 283± 304. Swanson, H. L. (1996). Individual and age-related differences in children's working memory. Memory and Cognition, 24, 70±82. Vicari, S., Carlesimo, A., & Caltagirone, C. (1995). Short-term memory in persons with intellectual disabilities and Down's syndrome. Journal of Intellectual Disability Research, 39, 532±537. Wang, P. P., & Bellugi, U. (1994). Evidence from two genetic syndromes for a dissociation between verbal and visuo-spatial short-term memory. Journal of Clinical and Experimental Neuropsychology, 16, 317±322. Wechsler, D. (1981). Wechsler Adult Intelligence Scale-Revised (WAIS-R). New York: The Psychological Corporation.