Working Memory and Spoken Language Comprehension in Young ...

2 downloads 0 Views 229KB Size Report
memory resources that could otherwise be used for sto- rage. Thus the association between reading span and comprehension is proposed to reØ ect language ...
INT ERNATION AL JOU RN AL OF PSYC HO LOGY, 19 99, 34 (5/6), 364± 373

Working Mem ory and Spoken Language Comprehension in Young Children Anne-M arie Adam s and Lorna Bourk e University of M anchester, UK

Cathe rine Willis Liverpool Hope University College, UK

This study has two theoretical dim ensions: (a) to explore which co mponents of Baddeley ’s (1986) working memory model are associated with children’s spoken language co mprehension, and (b) to co mpa re the extent to which measures of the components of this fraction ated model and an index of a unitary model (listenin g span) are able to predict indiv idual differen ces in spoken language co mprehension. Correlational analyses revealed that within a group of 66 4- and 5-year-old children both listenin g span and phonolo gical memory, but not visuospatial memory, were associated with vocabula ry know ledge and spoken language co mprehension. How ever, of the proposed measures of central executive function Ð dual task coordin ation, sustained attentio n, verbal ¯ uencyÐ only the latter was related to children’s ability to understand spoken language. Hierarchical regression analyses indicate d that variance in vocabula ry knowle dge was best explain ed by phonolo gica l memory skills, whereas individu al differen ces in spoken language co mprehension exhibite d unique and independent associations with verbal ¯ uency. Cette e tude s’inte resse a deux aspects the oriques distincts: (a) ve ri® er quelles co mposantes du modeÁ le de me moire de travail de Baddeley (1986) sont associe es aÁ la co mpreÂhension du langage parle chez les enfants et (b) comparer la valeur de pre diction, au niveau des diffe rences individu elles dans la co mpre hension du langage, des diffe rentes mesures des co mposantes du modeÁ le de me moire de travail et d’ un indice de modeÁ le unitaire (e mpan d’ e coute). Des analyses corre lationnelles montrent que dans un groupe de soixante-six enfants de 4 et 5 ans, l’ empan d’ e coute et la me moire phonolo gique sont associe s aÁ la connaissance du vocabula ire et aÁ la co mpreÂhension du langage. Par contre, la me moire visuo-spatiale n’ est pas relie e aÁ ces deux habilete s. Cependant, par mi les mesures de l’ unite de gestion centrale propose es (coordination de double taà che, attention soutenue et ¯ uidite verbale) seule la ¯ uidite verbale est relie e aÁ la capacite de co mpreÂhension de langage chez les enfants. D es analyses de reÂgression hie rarchique indique nt que la variance dans la connaissance du vocabula ire est mieux explique e par les habilete s de me moire phonolo gique, alors que les diffe rences individu elles dans la co mpre hension du langage montrent des associations spe ci® ques avec la ¯ uidite verbale.

C hildren learn to talk and to understa nd lang uage at w idely different rate s. However, despite a wealth of information abo ut the cogn itive processes in skilled lang uage processing th ere h as been co mparatively little investigation of the cognitive processes implicated in the development of exp ressive and recep tive lan guage skills. T he present stu dy aimed to integrate research exa mining the relationship betw een children’s phono logical me mory and their expressive language develop ment, w ith research into individu al differences in ad ult comprehension, propo sed to re¯ ect variation in working me mory capacity. Since these va rious stud ies re¯ ect two traditions within working me mory research, the models of working me mory that und erlie them, and the assessment task s derived fro m these, w ill be considered next.

D ane man and her colleagu es proposed that individual differences in language comprehension re¯ ect individual differences in working memory capacity (D ane man & C arp enter, 19 80). T heir preferred measure of working memory is the reading span, or co mplex span task. Various versions of this task exist; however, all share the basic premise that th e subject is requ ired to process a series of sentences, and then to recall their ® n al words. T his task was d esigned to directly model the postulated functions of working me mory, the simultaneous processing and storage o f infor matio n. Tota l me mory capa city is not propo sed to var y between ind ividuals, w ith comprehension differences arising as a result of variation in the ef® ciency w ith which info r mation can be processed (D an eman & Tardif, 1987). Inef® cient processing is

Requests for reprints should be addresse d to Anne-M arie Ada ms, Depa rt ment of Applied Psycholog y, Liverpool John M oores Univers ity, True man Building, 15± 21 Webster Street, Liverpo ol, L3 2E T, UK . This research was supporte d by a research grant fro m the Econo mic and Social Research Council . The authors would like to thank the children and the teachers who gave so willingly and patiently of their time to assist in the study.

q

1999 International U nion of Psychological Science

WOR K IN G M EM ORY AND CH IL DR EN’ S CO M PREHEN SIO N

considered to dema nd mo re of th e limited working me mory resources that cou ld otherw ise b e u sed for storage. T hus the association between reading span a nd co mprehension is proposed to re¯ ect language p rocessing differences that lead to impaired short-te r m retentio n of the sentence-® n al words. T his interpretation of the task and its association w ith co mprehension has not, however, been un animously accepted (E ngle, C an tor, & C arullo, 1992). O ne of the most serious criticisms of this view is th at an association between reading span and co mprehension is poten tially theoretically vac uous, since essentially what has been de monstrated is th e language co mprehension element in bo th task s. Furth er to this individual differences in verbal processing skills may not best characterise individu al differences in working me mory abilities. C o mparable associations have been obtained between a co mp lex span task that did n ot require lan guage p rocessing skills (arith metic p rocessing co mbined w ith word recall) a nd reading co mprehension (En gle et al., 1992 ). T he model of working memory developed to acco unt fo r these a nd related ® ndings differentiated long-te r m me mo ry (the ``p ermanent’ ’ know ledge store), wo rking me mory (the proportio n of this that can be simultan eously activate d through relatively automatic processes), an d short-te r m me mory (a subset of working mem ory th at is th e current focus of atten tion) (C onway & E ngle, 1 994). Individ ual differences in wo rking me mory capacity are proposed to re¯ ect differences in general atte ntio nal resources that have implications for the ef® ciency of inhibition processes (En gle, 1996 ). T hese inhibition processes ® lter out activated but irrelevant long-te r m me mory infor mation th at wo uld otherwise congest the limited capacity working me mo ry. T he principal aim of the current study wa s not, how ever, to choo se between these theories, but to co mpare such theories, (here ter med ``unitary’ ’ models of working memo ry) w ith an alternative v iew o f working me mory resources, B addeley’s (1986) ``fr action ate d’ ’ model. Th e present study investigated whether the association between co mplex span and adu lt co mprehension was also eviden t in young children, an d w hether such unitary models provid e a more coherent account than a ``fra ctionate d’ ’ model. D aneman an d Blennerha sset (1984 ) tried to measure the listening span perfor mance of young children a nd relate this to their listen ing comprehension skills. It proved imp ossible, however, to prevent th e children fro m imitatin g the entire sentence rath er th an recalling on ly the sentence-® nal words. T hey therefore merely de monstrated th at children’s ability to repeat sentences was related to th eir ability to u ndersta nd language. In an atte mpt to resolve such dif® culties the p resent study e mployed a sentence co mpletio n task (Towse, H itch, & H utton , 1998 ), in w hich the processing co mpon ent requires that th e child sup ply the missing ® nal wo rd of a spoken sentence. Such self-generatio n may help children to identify and recall the sentence-® nal words.

365

T he impetus fo r Dane man and C arpenter’s adoption of the co mplex span task w as the desire to develop a task that re¯ ected both the storage and processing functions of working me mory. H ow ever, o ther researchers incorporated th e con trast b etween processing and storage into their models an d explicitly e mployed th ese distin ctions to derive more constrained models of wo rking me mory. O ne such model, h ere ter med the ``fr actionate d’ ’ model, is the working memory model of Bad deley (1986) . W ithin this model storage capab ilities are d isseminate d to the ph onological loo p and the visuo spatial sketch pad, speci® cally designed to store phono logical and visuospatial infor mati on respectively. T he processing fun ction of working me mory is ascribed to a separate compo nent ter med th e central executive, which provides the interface w ith info rmation in long-te r m memory and also coordinate s the distributio n of the limited resources th roughou t the me mory syste m. A lthough both classes of theories of working me mory h ave speci® ed its importan ce in high er-level cognitive fu nctions, and have actively explored th e relation betw een working memory and lang uage, in contrast to the wo rk ou tlined above, research w ithin the latter tradition has focused on the relationship betw een language and the storage cap abilities o f the model, particularly phonological memory skills (see B addeley, Gath ercole, & Papagn o, 1998, for a review ). A ssociations have been identi® ed between children’s ph onological memory an d their vocabu lary knowledge (G athercole, Willis, E mslie, & B addeley, 19 92), w ith better ph onological me mory b eing associated with better vocabulary kn ow ledge. L inks have also been demonstrated between children’s phonological me mory a nd their expressive language skills (B lake, Au stin, Can on, L isus, & Vaughan, 199 4), an d it h as p roved po ssible to differentiate the spoken lan guage pro® les of you ng children grouped in ter ms of th eir phono logical memory skills (A dams & Gath ercole, 1995). Although evidence of phonological me mory involvement in adu lt co mp rehension is weak and it is generally accep ted th at und ersta nding is predo mina ntly achieved w ithout the need for short-te r m pho nological storage (C aplan & Waters, 19 90), a different picture may b e evident during the develop ment o f co mp rehension skills. Four-ye ar-old children classi® ed as h aving relatively poor phon ological me mory abilities exh ibited i mp aired perfor mance on a sta ndardised task of spoken language co mprehension (W illis, 1996). T he evidence fro m unitary models of working me mory sug gests that the working memory resources implicate d in co mprehension may exten d beyon d the ph onological do ma in, how ever. Independent evid ence also exists that children’s spoken language develop ment (G erken, 1991), and perfo r mance (C ara mata & Schw artz, 19 85) may b e constrained by across-do main processing limitations. T he current stu dy w ill therefore consider wheth er associations between working memory a nd receptive langu age skills are con® ned to the phon ological

366

ADAM S, BOUR K E, W IL LIS

co mponent of B addeley’s model, o r whether th ey are also fou nd with visuospatial me mory, and perhaps particularly w ith central executive fun ctions. O ne facto r th at is likely to have contributed to the distin ct focus o f previous investigations of the working memory/language relationship is the imprecise characterisation of the moda lity-free central executive. H ow ever, recent advan ces in ascribing speci® c fun ctions to the central executive, and the d evelop ment of approp riate assess ment techn iques (B addeley, 1996) may affo rd the o pportunity to exa mine w hether ind ividual differences in central executive function are also related to spoken language ab ilities. In th e present study tests modi® ed for use w ith children which re¯ ect th ree po stulate d central executive functio ns were exa mined in order to d etermine whether associations exist between children’s receptive lang uage ab ilities and cen tral executive skills. Th e executive fu nctions exa mined were the search a nd retrieval of infor matio n fro m lon g-ter m me mory, dualtask coordination, and sustained atten tion. Verbal ¯ uency was used to assess long-te r m me mory search an d retrieval processes. T he dual-tas k paradigm of B add eley, D ella Sala, G ray, Papag no, and Spinnler (19 97a), which co mpares the relative detriment in perfor mance between levels achieved when conducting two tasks in isolation to levels achieved when they a re condu cted simultaneously, was taken a s an index of the central executive function of coordinatin g the simultan eous execution of two task s. T he ® nal executive fun ction assessed was sustain ed atte ntion to response task (SA RT) using th e task d esigned by Robertson, M anly, Andrade, Badd eley, and Yiend (1997) . T his task, developed to identify ad ults w ith impaired fron tal lobe fun ction, requires the inh ibition of a highly p ractised respon se to a relatively rare targe t. T he hypothesis tested is th at individual differences in such central executive functions are a ssociated w ith individual differences in children’s ab ility to u nderstan d spoken language. Th e research q uestions add ressed in the p resent study can therefore be su m marised a s follows. Are ind ividua l differences analyses able to con® r m previous evidence fro m gro up stud ies of associations between spoken langu age co mprehension skills and ph onological memory abilities? A re these associations limited to the p honological co mponent o f wo rking memory or are they also associated w ith visuospatial me mory? Since the mechanisms postulate d to for m the basis o f the association tend not to include a visuospatial co mpon ent (Ada ms & G ath ercole, 1995) this relationship w as not expected to be signi® can t. F urther questions included wheth er, like adults, children’s language co mprehension skills are associated w ith co mplex span perfor mance a nd w hether these associations are stronger than those fo und w ith simple span perfor mance? F inally, is the index of ``unitary’ ’ working memory (co mplex span) a better p redictor of individua l differences in language skills th an p honological me mory or central executive fu nctions?

METHOD Phase One: Screening Participants C hildren, excluding any who were receiving speech therapy or who had a history of hearing problems, were recruited fro m schools in the N orth -West of E nglan d. T hree hu ndred and ten children (147 male and 1 63 female), age d between 47 and 6 3 months (mean age 58.28 mo nths), prod uced data for each measure in the initial ph ase of testing .

M aterials and Procedure P honolo gical M emory Tests 1. Nonword repetition. T he Children’s Test o f N onwo rd Repetition (G ath ercole & B add eley, 1 996), in which the child is encou raged to copy the ``fu nny made-u p wo rds’ ’ that the experimenter has said, w as presented to each child. The number o f ph onemically correct repetition s was recorded (m axi mu m 5 40). T he mean score on this test was 29 .19 (SD 5 6.14) ranging fro m 11± 40 items correct. 2. M em ory span for words and digits. Stimuli for the wo rd span task w ere 12 pho nologically dissimilar, singlesyllable words, appropriate fo r children o f this age (cake, pig, box, fork , b ag, bed, sun, egg, mous e, coat, d uck , dog ), and for the digit span task stimuli w ere the digits fro m 1± 9 (excluding the bisy llabic d igit 7). Presentation lists ranged in length fro m two to six ite ms w ith three lists at each list length. Items w ere presented at the rate of one item per second . C orrect repetition of two of the three lists at a g iven length resulted in th e len gth of the next list being increased by one item. Testing ceased when the child failed two o f the th ree lists at a given length . M emory span (for bo th words a nd digits) w as scored a s the maxi mu m length at which the child was able to repeat two lists correctly. M ean me mory span for words was 3.30 (SD 5 0.69 ) an d for digits was 3.78 (SD 5 0.8 0). 3. C o mpos ite pho nological me mory score. A ll three measures of ph onological memo ry w ere highly intercorrelated, signi® cant at the level P , .001. N onword repetition w as correlated w ith both word spa n and digit span (r 5 .478, and r 5 .444 respectively), an d both measures of span w ere also signi® cantly correlated (r 5 .536). To produce a co mposite measure of phonological memory, the scores on each of the three phonological memory measures were converted to a z score, a nd the mean of th ese scores was calculated. This val ue is referred to as the phono logical me mory score. N onverbal A bility. Four subscales (Object assembly, B lock design, M azes, a nd Animal pegs) fro m the Wechsler Preschool an d Primary Scale of Intelligence (Wech sler, 1990) provided a Perfo rma nce IQ for each child. M ean

WOR K IN G M EM ORY AND CH IL DR EN’ S CO M PREHEN SIO N

Perfo r ma nce IQ , subsequently referred to as the nonverbal ab ility score, was 105. 92 (SD 5 15. 19) ran ging fro m 69± 143.

Phase Tw o: Assessm ent Participants C hildren w ere selected to ta ke part in the second ph ase of th e study on the basis of their no nverbal ability and their phonological memo ry skills. All 310 children were ranked according to their compo site phono logical me mory score, and every third child in th is ranking to ok part in Phase Two of the study. Th e data a re reported fro m 66 of these children fro m who m co mplete data sets are currently ava ilable. T he mean age of these children was 6 0.65 (SD 5 3.05 months) rangin g betw een 54 ± 66 months. D ue to the number of tests to be ad ministered, Phase Two testing was conducted in two sessions between 4 to 8 weeks apart. In the ® rst session visual short-te r m memory and language skills were assessed. In the second session the children were given a listening span ta sk and th ree ta sks to assess central executive functions. W ith in each session the order of presentati on of the tests w as counterbalanced a cross children.

M ater ials and Procedure Visuo-s patial memory ab ility 1. C orsi blocks. S hort-ter m me mory for visuospatial infor mation was assessed using C orsi blocks (Isaacs & Vargha-K hade m, 1989), in w hich the child must replicate a sequence tap ped o ut on a subset of rando mly spa ced blocks. F ive sequences at each length fro m an initial length of two blocks were constructed. S uccessfu l completio n of th ree sequences presented at a given length resulted in the number of blocks in the next sequence being increased by o ne block. Testin g stop ped w ith failure on th ree sequences at a given length . S pan was recorded as the maxi mu m sequence length at w hich three sequ ences were correctly reprod uced. 2. Visual patter n span . Short-ter m memory for visual patterns w as assessed using a p aper and pencil version of the task developed by W ilson, Scott, and Power (1987) . T his task consisted of a series of matrices in w hich half the cells w ere rando mly ® lled. Co mplexity in the task was indexed as th e number o f ® lled boxes w ith the initial level of co mplexity being two ® lled boxes (a 2 3 2 matrix). F ive matrices were availab le for presentation at each level of co mplexity. T he children’s task was to recall the p osition of the ® lled cells in th e matrices, making th eir responses in a book let of blank matrices correspond ing in size to each of the targe t patte rns to be recalled. E ach child was given a set of practice trials (two atte m pts at 2 3 2 and two atte mpts at 2 3 3 matrices) following the same procedure as the experimental trials. C ontinuation and cessatio n criteria were th e same a s those adop ted fo r C orsi blocks. S pan was calculated as the numb er of

367

® lled cells at w hich three matr ices w ere correctly recalled. 3. L istening spa n. In this ta sk the experimenter read ou t a series of sentences selected fro m th ose in Towse et al. (199 8) an d th e child was asked to supp ly the missing word. A t the end of a set of sentences the children w ere asked to recall, in the correct serial o rder, the words that they h ad supplied to co mplete the senten ces. Initial set size w as two sentences (recall two ® nal wo rds) w ith th ree sets of senten ces availab le at each span length . Testing procedu re followed that adop ted for th e word/digit span tasksÐ if the child correctly repeated two of th e th ree sets of sentences at a given length then the number of sentences in the next set increased by one. T he children, however, still found this task dif® cult, with some repeating the entire senten ce rath er than just the ® nal wo rds. Su ch comp lete imitations were scored as an incorrect response. A ssessing span as the number of senten ces in which th e child correctly recalled the ® nal words in two sets of sentences produced a very limited rang e o f scores (0± 2). A further score was therefore co mputed as per the absolute span procedure o f E ngle, C arullo, an d C ollins (1991) in order to re¯ ect the variation between children more accurate ly. A bsolute span w as calculate d as the total number of words recalled in correct trials. T hus if a child recalled two of th e three wo rds at set size two and on e at set size three their absolute span sco re would be 7 [(2 3 2) 1 (3 3 1)]. C redit was not given for correct words in partially correct trials. T his absolute span score is reported in the following analyses. C entral E xecutive. T he central executive functions measu red in the present study w ere: search and retrieval fro m long-te r m me mory; dual-ta sk coordination; a nd sustained atten tion. 1. Search and retrieval from long-te rm me mory (verbal ¯ uency). The nu mber of exemplars fro m the two categories of anim als and food and drink that th e child was able to p roduce within a 30-s econd time limit was recorded. A practice category, items of clothing, was ® rst presented to all the children to con® r m that they und erstoo d the task requirements. Perfor mance on the two measu res o f verbal ¯ uency was highly associated , r 5 .429 , P , .001, and in the fo llowing analyses the mean of these two verbal ¯ u ency measures is reported. 2. D ual-ta sk coordina tion . T his task was based on the du al-tas k paradigm of B addeley et al. (1997a ), modi® ed fo r use w ith children. T he two tasks are a track ing task and me mory span for digits. F irst a conservative measure of the children’s digit span, the longest list length at which they were able to repeat th ree out of three lists correctly, was o btain ed. T he children w ere subsequently required to recall lists of digits at this span length for a period of 30 second s. Perfor mance was measured as the percentage of correctly repeated sequences (span single). For the tracking task each child w as presented with an 2 A 4-size sheet of paper on which were 80 3 2c m boxes linked by a line to for m a path . T he children w ere asked

368

ADAM S, BOUR K E, W IL LIS

to start at th e b eginning and to follow th e path along by puttin g a mark in each of the boxes. Th ey were told that they should try and do this as q uickly as possible w hilst making sure that th ey didn’t miss any of the b oxes out. A shortened version of the task for med a practice trial. Perfo rmance was assessed as the number of b oxes that had been marked w ithin the time limit o f 3 0sec (tracking single). Th e children were then requ ired to cond uct both tasks simultaneously (sp an du al and tracking dual). T he relative detriment in perfor ma nce over b oth tasks was calculate d using the mu score ou tlined in B addeley et al. (1997a ). Th is measure provides a single measure of dualtask perfor mance as a percentage o f single-task perfor man ceÐ a low score represents a large decrement and a high score a small decrement in perfo rman ce. 3. Sustained attent ion to respons e task (SA RT). In this child’s version of the task, rando m sequ ences of animals, dog , frog, duck, an d tortoise, w ere presented singly at various location s on a co mputer screen for 0.5sec each. T he locatio n of the animals and the delay betw een successive stimulus p resentations in the sequ ence w as rando mised (ranging fro m 1 to 2sec). C hildren were asked to press a co mputer key as soon as th ey saw o ne of th ese animals appear. T he ta rget animal (a dinosaur) was included in the sequence in a pre-® xed quasirando m fashion. T he children were instructed not to press the space bar when th e dinosaur app eared o n the screen. T hey were asked to b e as q uick as th ey could but to be careful that they didn’t press when they shouldn’t. T here were 60 stimulus events, in which the dinosaur appeared 5 times. A short practice trial was presented until the child was able to press for the other animals but inhibit their response to the dinosau r. For each child, target accuracy was recorded as the nu mber of times on w hich they successfully refrained fro m pressing the key when the dinosaur appeared (m axi mum 5 5). Lan guage A ssessment 1. Vocabulary co mprehension. T he B ritish Picture Vocabulary Scale (D unn, D unn, W hetton, & Burley, 1997), in w hich the child must choose which of a set of pictu res represents the word spoken by the exp erimenter, was given to each child. Raw scores are reported, which re¯ ect b oth the level reached and the number of errors made. 2. Language comprehension. Th e co mprehension scale of the Reyn ell D evelop mental L anguage Scales (E dw ards et al., 1997), a stan dard test of lan guage d evelop ment, was used to assess the children’s spoken language co mprehension skills. Toys and a picture bo ok are e mp loyed, w ith the child being asked to d emonstrate und erstanding of w hat th e experimenter said either by pointin g to a pictu re or by d emonstrating the actions w ith the toys. U ndersta nding of a rang e o f linguistic co mponents a nd constructions is tested including: single words, agen ts and actio ns, sub ject-ve rb-o bject constructio ns, adjectives, locative relations, passives and p ost-mod ifyin g clauses. The ® n al section of the test assessed the chil-

dren’s ab ility to draw inferences based on real world knowledge. All th e ite ms in th e test were presented to each child, resulting in a maxi mum score of 62.

RESULTS Descriptive Statistics D escriptive statistics for each o f the measures assessed in Phase Two are show n in Table 1. L evels of perfo rmance in me mory are co mparable to p revious stud ies (A da ms & G ath ercole, in press; Siegel, 19 94) and in line w ith standardized scores for language abilities. A lthough each of the central executive task s seemed to be sensitive to individu al differences among st the children, the standard on two of the tasks merits some co m ment. T he mean target accuracy in th e SART ta sk of 44% (2.2/5 correct) co mpares with 70% (17.4/ 25 correct) for frontal lobe patients and 84% (21/25 correct) for adult controls (Robertson et al., 1997). T he children are therefore perfo rming worse than patients w ith impaired executive function. D ual-task coordination perfo r ma nce also seemed to be p oor, w ith perfo r ma nce decrements in both tasks under dual-ta sk conditions. M u scores ranged fro m 29± 126% , co mparable to the ran ge reported by B addeley et al. (1997a , 54± 120% ) for adult control participants. However, the children’s mean mu score of 71% is below the mean of the young adult controls (92 % ) an d TABLE 1 Descriptive Statistics for Nonverbal Ability, Phonological and Visuospatial Memory Measures, Listening Span, Central Executive Skills, and Language Abilities (N 5 66) Test Nonverbal ability Verbal me mory N onword repetition (max 40) Word span D igit span Visuosp atial me mory C orsi blocks Pattern span Listenin g span task Listening span Ab solu te span score C entral executive skills Verbal ¯ uency A nimals Fo od and drink M ean verbal ¯ uency Sustained attention Target acc. (max 5) D ual-task coordination Span sin gle (% accuracy) Span dual (% accuracy) Trackin g sin gle (max 80) Trackin g dual (max 80) M u score Language comprehe nsion skills Vocabulary comprehension Lan guage co mprehension

M ean (SD )

R ange

107.01 (14.29)

73± 134

28.42 (6.68) 3.36 (0.74) 3.74 (0.79)

11± 40 2± 5 2± 6

3.06 (1.21) 2.30 (0.76)

0± 5 0± 3

0.30 (0.72) 1.56 (1.85)

0± 2 0± 7

5.73 (1.87) 5.57 (1.98) 5.65 (1.63)

2± 11 1± 11 2± 10

2.20 (1.25)

0± 5

79 62 39 22 71

(23.10) (34.40) (8.46) (8.15) (22.78)

55.79 (10.55) 53.32 (3.72)

0± 0± 22± 6± 29±

100 100 60 48 126

34± 80 45± 60

WOR K IN G M EM ORY AND CH IL DR EN’ S CO M PREHEN SIO N

further mo re similar to that prod uced by dysexecutive patients (77% , Ba ddeley, D ella Sala, G ray, Pap agn o, & Sp innler, 199 7b). Two possibilities may explain the children’s poor perfor mance on these tasks. Either these task s are not va lid measures o f central executive function in children of this age or th ese function s are as yet not consistently ava ilable to these children. T hese issues are considered furth er in the discussion.

Correlational Analyses Within-domain Correlations Table 2 repo rts the correlatio ns between the phonological, visuospatial, listening span, and central executive me mory measures and vocabulary and lang uage co mp rehension scores. Th ere were signi® cant intercorrelations a mongst the three phonological memory measures, a nd the two visuospatial me mo ry measu res were also signi® cantly correlated. T here were, however, no signi® cant association s betw een absolute span and central executive skills. N onsigni® cant, indeed minimal, correlations were ob served betw een the assessments of central executive functions the mselves. T he i mp lication s of the relative independence o f the central executive measures w ill be considered further in the discussion. Vocabulary co mp rehension show ed a highly signi® can t association w ith language comprehension.

Across-dom ain Correlations C onsidering the relatio nship between the measures of ph onological me mory an d visuospatia l memo ry, Co rsi blocks bore no association w ith any phonological memory measure whereas visual patte rn span was associated on ly w ith no nword repetition skills. A bsolute span was related to n onword repetition and wo rd span, but not digit span . O f th e measures of central executive functio n, on ly verbal ¯ uency was associated w ith any o ther memory measure, being related to word span.

Hierarchical Regression s H ierarchical regression analyses were conducted to exa mine th ese associations in more detail, prim arily to consider whether any of the memory measu res bo re a un ique association w ith language skills. In th is technique var iables that are p roposed to predict var iance in a criterion variable are ad ded to the model in a theoretically constrained order. A t each stage in th is p rocess the a moun t of var iance in the criterion var iable that can be accounted for by variables n ot yet includ ed in the equation, or conversely the signi® cant ad ditional varian ce that th ey are able to exp lain when ad ded to the equation, can be ascertained. In this w ay, it is p ossible to identify variables that are able to make signi® cant a nd ind ependent con tributions to variance in the criterion var iable. In order to limit the numb er o f variables in these an alyses, wherever possible, co mposite measures of the mem ory con structs w ere employed. T hus in the same w ay that the co mposite measu re of phon ological me mory was constructed, a comp osite measure o f visuospatial memo ry was create d co mprising the mean of the z scores for C orsi blocks and visual patte rn span. Since the measures of central executive fun ctions were not related to each other, it w as no t considered viable to incorporate these scores into a co mposite measure a nd they w ere entered into the regression equations as separate variables. In the fo llowing an alyses age and nonverbal ability were entered ® rst to establish the degree of association between langu age an d me mory that w as ind ependent of these fac tors. C o mponents of the fractionate d model were entered next, w ith visuospatial me mo ry ® rst, since accounts of the association between limited cap acity resources and language develop ment do no t specify a visuospatial mechanism. C entral executive measures were entered next, to allow the contribution of phonological memo ry previously show n to be associated with lan guage skills to be evalu ate d independen tly o f this co mponent. A bsolute spa n as the index of un itary working me mory was entered as the ® nal var iable. A lternative

TABLE 2 Correlations between the Memory and Language Measures Assessed in Phase Two (N 5 Variables 1 2 3 4 5 6 7 8 9 10 11 12 13

1

A ge N onverbal ability N onword repetition Wo rd span D igit span C orsi blocks V isu al pattern span Verbal ¯ uency mean Sustained attention target acc. Dual task (mu) Absolute span Vocabulary co mprehension Lan guage comprehension * P ,

.05 ;

** P ,

.01.

1 .02 .10 .04 2 .04 .26* .20 2 .07 .14 .29* .15 .26* .21

2

1 .25* .18 .16 .24 .28* .28* 2 .05 .16 .31* .42** .45**

3

1 .55** .41** .11 .29* .14 2 .07 .09 .31* .46** .29*

4

1 .56** 2 .08 .08 .24* .01 .14 .27* .30* .30*

369

5

6

1 .07 .08 .10 2 .06 .12 .14 .27* .27*

1 .30* .04 .10 2 .04 .15 .27* .24

7

8

9

66) 10

11

12

1 .067 2 .0 0 .1 1

1 .39** .30*

1 .65**

1

2 .06 .00 .02 .17 .19 .13

1

2 .02 .02 .15 .21 .45**

1

2 .11 2 .09 2 .06 2 .07

370

ADAM S, BOUR K E, W IL LIS

orderings were a dopted to estab lish the degree of shared varia nce between signi® cant predictors.

suggests th at th e predictive power of absolute span may merely re¯ ect elem ents that it has in co m mon w ith phonological memory skills.

Vocabulary Comprehension as the Criterion Variable Lang uage Com prehensio n as the Criterion Variable

Th e ® rst regression equation (to p pan el of Table 3) was conducted w ith vocabulary co mprehension as the criterion var iable. B oth age and no nverbal ability made signi® cant contribution s to var iance in vocabulary knowledge. N eith er visuospatial me mory entered at the next step, nor any of th e measures of central executive fun ctions entered afte r th is were able to predict further varia nce in vocabulary know ledge. P honological me mory entered at step seven explained a signi® cant add itiona l 9.9% of var iance, w hile ab solute span entered at step eight fa iled to accou nt for a furth er signi® cant proportion of var iance in vocabulary know ledge. Reversing the order of the ® nal two steps in the regression equ ation (see botto m panel of Table 3), ab solute span entered at step seven accounted for an add ition al 4% of variance although this just failed to reach signi® cance (P 5 .056). In contrast, even w hen the effects of all the other variables h ad b een controlled, phono logical me mory remained a signi® cant predicto r, acco unting for a further 8% of the var iance in vocabulary know ledge. T his patte rn

T he ® rst regression equation (to p pa nel of Table 4) entered the independent variables in the same order as in the ® rst regression in Table 3; however, th is time lang uage comp rehension was the criterion var iable. N onverbal ability, but not age, predicted a signi® cant amount of varia nce in langu age co mprehension skills. A gain the co mposite measure of v isuo spatial memory w as unable to p redict signi® cant furth er var iance and the sa me was true of the central executive measures of d ual-task coordination a nd targe t accuracy in the sustained atte ntio n task. H owever, the central executive task of verbal ¯ uency predicted a n additional 13% of the var iance in lang uage co mp rehension. Neither ph onological me mory or absolute span entered after this could explain signi® cant proportions of var iance. In order to establish the degree of ind ependence of verbal ¯ uency fro m p honological me mory and absolute span, alternative orderings of th ese variable w ere considered .

TABLE 3 Hierarchical Regressions: Vocabulary Comprehension as the Criterion Variable Step

Inclusio n O rder

1 2 3 4 5 6 7 8 7 8

Total % Var.

A dd. % Var.

F Value

Age N onverbal ability Visu al memory C entral executive: mu C entral executive: target acc. C entral executive: verbal ¯ uency Phonological memory Ab solu te span

7 24 25 26 27 29 39 41

7 17 1 1 1 2 10 2

4.75 13.95 0.70 1.45 0.87 1.20 9.43 1.94

Ab solu te span Phonological memory

33 41

4 8

3.81 7.30

P Value

,

.05 .01 n.s. n.s. n.s. n.s. , .01 n.s.

,

.056 (n.s.) , .01

TABLE 4 Hierarchical Regressions: Reynell Language Comprehension as the Criterion Variable Step

Inclusio n Order

1 2 3 4 5 6 7 8 6 7 8 6 7 8

Total % Var.

A dd . % Var.

F Value

P Value

A ge N onverbal ability V isual memory C entral executive: mu C entral executive: target acc. C entral executive: verbal ¯ uency P honological memory A bsolute span

5 25 25 25 25 38 42 42

5 20 0 0 0 13 4 0

3.06 16.66 0.08 0.03 0.49 12.25 3.61 0.37

A bsolute span P honological memory C entral executive: verbal ¯ uency

27 32 42

2 5 10

1.41 4.02 10.21

,

A bsolute span C entral executive: verbal ¯ uency P honological memory

27 39 42

2 12 3

1.41 11.59 2.91

,

n.s. .01 n.s. n.s. n.s. , .01 n.s. n.s.

,

n.s. .05 , .01 n.s. .01 n.s.

WOR K IN G M EM ORY AND CH IL DR EN’ S CO M PREHEN SIO N

T he middle panel of Table 4 shows that ab solute span, even when entered at step six, predicted only 2 % of variance in language co mprehension, w ith both phonological working me mory and verbal ¯ uency explaining signi® cant further p roportions o f var iance (5% and 10% respectively). Th is ® nding that verbal ¯ uency makes an independent contributio n to individual var iatio n in language co mprehension is h ighlighted by the patte rn of relationships found in the ® n al regression equation (bo tto m panel of Table 4). Reversing the order o f the ® nal two stage s in the previou s equation it is evident that the predictive power of verbal ¯ uen cy do es no t arise merely fro m shared varia nce w ith phono logical memory, since including ph onological me mory in th e ® nal stage, it is un able to predict a signi® cant a mount o f var iance furth er to that con tributed by verbal ¯ uency skills.

DISCUSSI ON Working Memory and Vocabulary Comprehension Individual differences in vocabulary k now ledge were related to children’s working me mory skills. T his was evident both w hen assessed by listening span a nd ph onological me mory tasks, although n ot generally w ith tasks assessing central executive fun ction, a nd only w eakly and inconsistently with visuospatial memory. H ierarchical regression a nalyses revealed that the link between vocabu lary comp rehension and phono logical me mory was signi® cant even when individu al differences in age, n onverbal ability, v isuo spatial memo ry, central executive function s, and listening span h ad been contro lled. S uch a distinct relatio nship con® r ms previous evidence o f substan tial link s between children’s vocabulary development and their phonological memory skills (G ath ercole et al., 1992). T he causal natu re o f this relationship has been the subject of much d ebate a nd the correlatio nal natu re of the present study is not infor mative in this regard. H ow ever, the mo st likely account now seem s to be th at the relationship is interactive; not on ly do phonological me mory skills in¯ uence long-te rm learning , but word knowledge affects ph onological me mory perfo r man ce (Badd eley et al., 1998). T he speci® city of these associations to the phono logical comp onent of working me mo ry have b een con® r med by the evidence presented here.

Working Memory and Language Comprehension A rather different picture e merged w ith language comprehension skills. L ang uage co mprehension was again associated w ith b oth listening span and phono logical me mory skills and not at all related to visuospatial me mory. H ow ever, a lthough bo th sustain ed atte ntio n a nd du al-task coo rdination continued to var y ind ependently of language co mprehension , verbal ¯ uency b ore a h igh

371

degree of association w ith this skill. T he prior demonstration of differences in the language co mprehension abilities of children grouped in ter ms of their p honological memory skills (W illis, 1996) was therefore also evident in this study at th e level of the ind ividual child. M ore detailed exploration of th is pattern of associations in hierarchical regressions revealed that altho ugh phonological memory shared a signi® cant degree of varia nce w ith language comprehension, this w as not independent of the var iance attr ibutable to differences in verbal ¯ uency skills. Verbal ¯ uency was able to predict a unique and signi® can t proportion of the var iance in language co mprehension. Previous evidence o f the relative ability o f co mp lex span co mpared to simple span to predict adult differences in reading co mprehension h as been equivocal. W hereas some studies have found th at co mplex but no t simple span is associated w ith measures of reading co mprehension (D aneman & C arp enter, 1980), others have show n that simple span is able to predict reading co mprehension (E ngle et al., 1991). It has been suggested that a simple span/co mprehension association may depend o n using a suf® ciently large po ol of sti muli. A lthough the present study used a pool of o nly 12 sti mu li, since the children made restricted progress throu gh th e possible co mbinations of the stimuli, relatively little repetition of ite ms occurred in the lists p resented to the children. T hus the task may w ell have functio ned for each child as though a larger po ol was availab le. A lthough the correlatio n betw een C orsi blocks a nd vocabulary co mprehension w as signi® can t it did n ot p redict a signi® can t proportion of va riance in th e hierarchical regression a nalyses. N either of the measures of visuospatial me mory were signi® can tly related to language co mprehension skills. Previous studies also sugge st that the relationship betw een visuospatial me mory skills and language abilities may be un clear. Ra ine, H ulme, C hadderton , and Ba iley (1 991) fou nd that co mp ared to controls, children w ith develop mental disorders of lan guage showed impaired me mory for sequences of pictured cards, whereas with probed recall, Ku shnir and Blake (1 996) found no such differences. Ada ms and G ath ercole (in press) recently noted associations between children’s expressive langu age skills and C orsi blocks but n ot visual pattern span. Kushnir and Blake (1996) propo sed th at the discrepancy betw een their ® ndings and those of Rain e et al. (1991) arose fro m the requirement in Raine et al.’s ta sk, but not theirs, to maintain serial o rder infor mation. T his is an intuitive account of the relatio nship b etween C orsi blocks but no t visual p atte rn span in this study, how ever, such an account wo uld p redict a signi® can t relation ship betw een C orsi blocks perfor mance and verbal span , which also requires me mory for serial order that w as not evident in this study. Th e reason for the tend ency for C orsi blocks to associate w ith so me measures of lan guage abilities remains uncertain.

372

ADAM S, BOUR K E, W IL LIS

The Validity of the Central Executive O ne o f the ai ms of the present study w as to co mpare unitary a nd fractio nated working me mory models not only on the issues on which they most markedly diverge, the assertion of ind ependent mod ality-speci® c storage mechanisms, but also in the area in w hich they are most likely to converge, processing resource limitations and operation s ascribed to the central executive. A particular aim was to deter mine th e exte nt to which these models ma de relative contributions to the high er-level cognitive function o f language comprehension. Such p roposed theoretical convergence o f unitary models and central executive functio n was n ot supported by th e current empirical evidence, either in ter ms of associations betw een central executive functio ns and co mplex span assessment, or of comparable links with language co mprehension skills. Althoug h contrary to th e current d ata, Rosen and E ngle (1994, cited in En gle, 1996) showed that verbal ¯ uency was linked to perfo r mance in a complex span task . Perhaps a more serious problem arises when considering the theoretical valid ity of the cen tral executive construct itself. A ltho ugh previous work has demon strated so me degree of association between th e p ostulated measures o f executive fun ction a dopted in the present stu dy, i.e. between dual-task coo rdination and verbal ¯ uency (Badd eley et al., 1997b ), no signi® cant inter-correlations were observed in th e present data. The lack of signi® cant inter-relations between executive, fro ntal, an d working memory tasks has been noted previously (L ehto, 1996). F inding inconsistent associations b etween va rious simple and co mplex span measures and executive measures taken fro m neurop sychological research, L ehto concluded that th e eviden ce sup ported not a single executive system, but one w hose co mp onent functions cou ld be differentiated. D iscussing such evidence and the interfac e betw een his mo del and u nita ry models of working memory, Badd eley conceded th at more evidence w as n eeded to deter mine whether a single executive sho uld be postulated, o r whether var ious function s may diverge to reveal an executive co m mittee (B add eley, 1996 ). T he present ® ndings, exte nding previous w ithin-in dividual inconsistencies between perfor mance o n alternative executive tasks fro m adults to children, thus aligns mo re closely w ith the proposal of a central executive co m mittee, rather than a sing le executive director. It is of course also possible that these fu nctions may b e d ifferentially implicated in va rious higher-level cognitive tasks. To exten d the metap hor further, one of the co m mittee member’s primary responsibilities may be the selection a nd manipulatio n of infor mation ma inta ined in the longter m sto re of k nowledge (cf. Rosen & E ngle, 1994). T he evidence presented h ere, a ltho ugh only correlatio nal, is consistent w ith the prop osal that this executive function may have p articular implications for the co mprehension of spoken language in yo ung children. A lternative interpretatio ns of the differences w ithin individua ls across th e var ious executive tasks do exist,

however. D espite atte mpts to reduce the dif® culty o f the tasks, they may simply not b e valid indices of children’s central executive fun ctions. F urther work is needed to address this and re® ne executive tasks to render the m app licable for u se w ith children. T he challenge of measuring the degree of association b etween executive skills and lan guage processing in children may be furth er co mplicated by the conten tion that executive skills may not yet be fu lly developed in children of this age Ð an implication of th e late develop ment of the fro ntal cortex (Johnson, 1997). In line w ith this th e children’s patte rn of perfor mance, although re¯ ecting quite large individual differences, did attr ibute to children on both SART an d dual-task coordination, mean achieve ment levels that were co mp arable to patients w ith executive impair ment.

Verbal Fluency as a M easure of Working Memory The executive function revealed as the most likely candidate for the mechanism that might underlie the association betw een working me mory and langu age co mprehension wa s verbal ¯ uency. H owever, the p articular process that it re¯ ects remains unclear. O ne intuitive prop osal, that individu al differences in vocabulary know ledge may exp lain p erfor mance on both tasks, can , however, be discredited by the present data. If this w ere so, one wou ld exp ect the contributio n of word know ledge to verbal ¯ uency skills to be represented in a signi® cant relationship b etween verbal ¯ uency an d vocabulary comprehension, a relationship that w as no t evident in th is study. It is n ecessary, therefore, to identify an other aspect of verbal ¯ uency that may underlie its association w ith lang uage co mprehension abilities. Research by E ngle and his colleagu es exp loring the natu re of th e mechanisms underlying perfo r mance in complex span task s may provide so me insight into this issue. This work developed fro m h is accoun t of memory divided into long-te r m memory, working me moryÐ the activate d p roportion of long-ter m memo ryÐ and short-ter m memo ryÐ the proportio n of wo rking me mory th at is u nder the current fo cus o f atte ntion (Co nw ay & E ngle, 1 994). O n the b asis of a series o f exp eriments manipulating the ``fan size effect’ ’ they concluded that differences in wo rking memory capacity re¯ ected not differences in th e time to make long-ter m memo ry active but the time to con duct controlled p rocesses to search, retrieve, or to focus atte ntio n on particular aspects of long-ter m me mory that were already active in wo rking me mory. E ngle’s (199 6) mod el of working me mory, which he accepts b ears so me co mp arison w ith the cen tral executive, main tains that individual differences in working memory are due to individual differences no t in the amount of infor mation in long-te r m me mo ry that can be activated at any one time, but to ind ividual differences in atte ntional resources, which a re required to inhibit irreleva nt infor mation in working memory. A central element of this research w as that verbal ¯ u ency is intimately related to co mplex span assess ment at least partly

WOR K IN G M EM ORY AND CH IL DR EN’ S CO M PREHEN SIO N

because it re¯ ects the availab ility of such controlled atte ntional resources. A lth ough th e current correlatio nal data are insuf® cient to con® r m any particular account of the association between working memory and language co mprehension, they are nevertheless consistent with the proposal that verbal ¯ u ency may bear a particular association w ith children’s language processing skills since it re¯ ects th eir ability to conduct controlled and hence effortful processing of long -ter m infor mation. Speci® c processes may include activating th e semantic correlates of th e phon ological for ms th at they p erceive, manipu lating or atten ding to these when they are activate d in working memory, or inhibiting info rmation which, although activated , is irrelevant to the current co mp rehension task.

REFERENCES A da ms, A .-M ., & G athercole, S.E . (1995). Phonological working memory and speech product ion in pre-school children . Journal of Speech and Hearing Research, 38(2), 403± 414. A da ms, A .-M . & Gathercole, S.E . (in press). Limitatio ns in working memory: Implications for language develop ment. International Journal of Language and Co mmunication D isorders. Baddeley, A .D. (1986). Working me mory. O xford: Oxford University Press. Baddeley, A .D. (1996). Explorin g the central executive. The Q uarterly Journal of Experimental Psychology, 49A(1), 5± 28. Baddeley, A., D ella Sala, S., Gray, C., Papagno, C., & Spinnler, H. (1997a). Testing central executive functio ning with a pencil-a nd-pape r test. In P. Rabbitt (Ed.), M ethodolo gy of frontal and executive fun ction. H ove, U K: Psychology Press. Baddeley, A., D ella Sala, S., Gray, C., Papagno, C., & Spinnler, H. (1997b). Dual-task perfor mance in dysexecutive and non-dys executive patients with a frontal lesion. N europsychology, 11(2), 187± 194. Baddeley, A.D. G athercole, S.E ., & Papagno, C. (1998). The phonolo gical loop as a language learning device. Psychological Review, 105(1) 158± 173. Blake, J., Austin, W., Canon, M ., Lisus, A., & Vaughan, A . (1994). The relationship betw een memory span and measures of imitativ e and spontaneous language co mplexity in preschool children. International Journal of Behavioural D evelop ment, 17(1), 91± 107. Caplan, D., & Waters, G.S. (1990). Short-t er m memory and language co mprehension: A critical view of the psychological literatu re. In G. Vallar & T. Shallice (Eds.), N europsychological impairments of short-term memory (pp. 337± 389). Ca mbridge: Cambridge University Press. Caramata. S.M ., & Schwartz, R.G. (1985). Production of object words and action words: Evidence for a relationship between phonolo gy and semantics. Journal of Speech and Hearing Research, 28, 323± 330. Conway, A.R.A ., & Engle, R.W. (1994). Working memory and retrieval: A resource-dependent inhibitio n model. Journal of Experimental Psychology: General, 123, 354± 373. Dane man, M ., & Blennerhassett, A. (1984). H ow to assess the listenin g co mprehen sion skills of prereaders. Journal of Educational Psychology, 76(6), 1372± 1381. Dane man, M ., & Carpenter, P.A. (1980). Individ ua l differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19, 450± 466.

373

D ane man, M ., & Tardif, T. (1987). Working memory and reading skill re-examined. In M . Coltheart (Ed.), Attention and performance XII: The psychology of reading (pp. 491± 508). Hove, UK: Lawrence Erlbau m Associates Ltd. D unn, L., D unn, L., W hetton, C., & Burley, A. (1997). The British Picture Vocabulary Scale (2nd ed). W indsor, U K: N FER -Nelson. Edw ards, M ., Fletcher, P., G arman, D., Hughes, D., Letts, P., & Sinka, S. (1997). Reynell Develop mental Language Scales, (III): The U niversity of Reading Edition. W indsor, U K: N FER -Nelson. Engle, R.W. (1996). Working memory and retrieval: A n inhibitio n-resource approach. In J.T.E. Richardson (Ed.), Working me mo ry and human cognition (pp. 89± 119). N ew York: Oxford U niversity Press. Engle, R.W., Cantor, J., & Carullo, J.J. (1992). Individu al differences in working memory for co mprehension: A test of four hypothes es. Journal of Experimental Psychology: Learning M emory and Cognition, 18, 972± 992. Engle, R .W., Carullo, J.J., & Collins, K .W. (1991). Individ ual differen ces in working memory for co mprehension and follow ing directio ns. Journal of Educational Research, 84, 253± 262. G athercole, S.E ., & Baddeley, A.D. (1996 ). The Children’ s Test of Nonword Repetition . London: Psychological Corporation. G athercole, S.E ., W illis, C., E mslie, H., & Baddeley, A.D. (1992). Phonolog ical memory and vocabula ry develop ment during the early school years: A longitud inal study. D evelopmental Psychology, 28, 887± 898. G erken, L. (1991). The metrical basis for childre n’s subjectless sentences. Journal of M emory and Language, 30, 431± 451. Isaacs, E.B., & Vargha-Khadem , F. (1989). D ifferential course of develop ment of spatial and verbal memory span: A normative study. British Journal of Develop mental Psychology, 7, 377± 380. Johnson , M .H . (1997). Develop mental cognitive neuroscience. O xford: Blackwell Publishers. Kushnir, C.C., & Blake, J. (1996). The nature of the cognitiv e de® cit in speci® c language impair ment. First Language, 16(1), 21± 41. Lehto, J. (1996). Are executive functio n tasks depend ent on working memory capacity? Q uarterly Journal of Experimental Psychology, 49A(1), 29± 50. Raine, A ., H ulme, C., Chadderton, H ., & Bailey, P. (1991). Verbal short-term memory span in speech-disordered children : implications for articulatory coding in short-term memory. Child D evelop ment, 62, 415± 423. Robertson, I.H., M anly, T., Andrade, J., Baddeley, B.T., & Yiend, J. (1997). ``Ooops!’ ’ : Performance correlates of everyday attentio nal failures in trau matic brain injured and nor mal subjects. Ne uropsychologia, 35(6), 747± 758. Rosen, V., & Engle, R .W. (1994). Working me mory capacity and retrieval from long-term me mory. Unpublis hed manuscript. Siegel, L.S. (1994). Working memory and reading: A life-spa n perspective. International Journal of Behavioural Development, 17 (1), 109± 124. Towse, J., Hitch, G.J., & Hutton, U. (1 998). A re-evaluation of working memory capacity in children . Journal of M emory and Language, 39(2), 195± 217. Wechsler, D. (1990). Wechsler Preschool and Prima ry Scale of Intelligence (revised U K ed.). London: Psychological Corporation. W illis, C.S. (1996). Working memory and children’ s language processing. U npublish ed PhD thesis, Lancaster U niversity. W ilson, J.T.L., Scott, J.H., & Power, K .G. (1987). Developmental differences in the span of visual memory for pattern. British Journal of D evelop mental Psychology, 5, 249± 255.