Memory impairment in children with language impairment

Transcription

1 DEVELOPMENTAL MEDICINE & CHILD NEUROLOGY ORIGINAL ARTICLE Memory impairment in children with language impairment GILLIAN BAIRD 1 KATHARINA DWORZYNSKI 2 VICKY SLONIMS 1 EMILY SIMONOFF 2 1 Guy's & St Thomas' NHS Foundation Trust, London, UK. 2 Institute of Psychiatry, King's College London, London, UK. Correspondence to Professor Gillian Baird at Newcomen Child Development Centre, Guy's & St Thomas' NHS Foundation Trust, Great Maze Pond, London SE1 9RT, UK. gillian.baird@gstt.nhs.uk PUBLICATION DATA Accepted for publication 17th July Published online 6th October LIST OF ABBREVIATIONS CELF Clinical Evaluation of Language Fundamentals C-LI Current language impairment CNRep Children's Test of Non-Word Repetition N-LI No language impairment NWR Non-word repetition P-LI Past language impairment WRAML Wide Range Assessment of Memory and Learning AIM The aim of this study was to assess whether any memory impairment co-occuring with language impairment is global, affecting both verbal and visual domains, or domain specific. METHOD Visual and verbal memory, learning, and processing speed were assessed in children aged 6 years to 16 years 11 months (mean 9y 9m, SD 2y 6mo) with current, resolved, and no language impairment using the Wide Range Assessment of Memory and Learning (WRAML), a standardized memory and learning test for children, and the Children s Test of Non-Word Repetition (CNRep), a test of phonological short-term memory. Fifty-one children (38 males, 13 females) with current speech and language impairment from 49 families were compared with 13 siblings (11 males, 2 females) with a past history of language impairment and 26 (15 males, 11 females) who had never had language impairment. RESULTS Children with current language impairment showed impairment in all verbal memory measures compared with children who had never had language impairment, and these impairments were still evident in children with a past history of learning impairment. Visual memory and learning were not impaired compared with children who had never had language impairment. The severity of verbal memory impairment correlated with the degree of language impairment. INTERPRETATION We concluded that in language impairment there is domain specificity of memory impairment affecting verbal processing. Language delays are common; according to one estimate, the prevalence among preschool children is , with a preponderance among males. 1 The cause or causes of persistence of language impairment have been much researched; it is likely that several processes, linguistic and cognitive, are relevant. Cognitive processing impairments involving speed 2 and working memory 3 are contributory separable, but interlinked, causes 4,5 correlated with academic achievement. 6,7 Whether such processing impairments are global or domain specific in language impairment is of theoretical and practical importance. There remains a debate about the domain specificity of working memory. Some models of working memory emphasize separable domain-specific resources for short-term storage with a domain-general processing component. 8 Others describe domain-specific verbal and visual memory resources 9 (visual memory refers to non-verbal information presented in the visual or spatial modality). Although many studies show a deficit in verbal memory in children with language impairment, results regarding visual memory vary. Any domain-specific deficit in working memory is of theoretical interest from the Baddeley model of working memory, and also of clinical importance if strengths can be utilized to compensate for difficulties. Verbal memory, both working memory and the shortterm phonological subcomponent measured by tasks such as reverse digit-span repetition and non-word repetition (NWR), shows a consistent association with the acquisition of vocabulary and grammar in typically developing children and impairment in children with language impairment. 3,4,10 12 Impairment of NWR is found in those with language impairment, even when the language disorder has resolved. NWR has been proposed as a marker for language impairment 13 and a strong genetic linkage has been found, specifically to chromosome Visual memory in those with language impairment has been investigated with variable results. Hoffman and Gillam 15 studied recall of digits and locations in a dual-processing task, and found that performance of a task involving central executive and short-term memory was poorer in children with language impairment than in typically developing children. Bavin et al. 16 found that children with language impairment showed poorer performance than an age-matched comparison group in pattern recall but not in memory for location in a spatial working memory task. Other researchers 17 found that the development of visuospatial short-term memory over 1 year was slower in 4-year-old children with language impairment than in a comparison group, but there was no difference in development using a visuospatial processing task. Archibald and Gathercole 3 showed that a substantial minority of children with language impairment have deficits in visuospatial memory. Subsequently, 18 they compared performance of tasks of ª The Authors. Journal compilation ª Mac Keith Press 2009 DOI: /j x 535

2 visual short-term memory and working memory in 15 children with language impairment (aged 7 12y) and a comparison group matched for language and chronological age and found no deficit in visual memory in the former. They postulated domain specificity of verbal and visual short-term memory and a specific deficit in verbal short-term memory in children with language impairment. Variation in the definition of language impairment, the age of children, and the complexity of task within and across domains may explain some of the differences in the above results. To examine the specificity of visual and verbal memory and learning impairment in children with language impairment, we compared a large group of well-characterized children with language impairment with their siblings who had either no history of language impairment or a past, now resolved, history of language impairment, using a standard test of memory and learning, the Wide Range Assessment of Memory and Learning (WRAML). Speed of processing information, which typically increases with age and mediates the increasing verbal memory and learning, 19 is slower in children with language impairment than in typically developing children across domains and tasks. 2 The extent of slowing varies with the task. 20 We hypothesized (1) that children with language impairment, compared with children with no language impairment, would have specifically impaired verbal memory and learning but intact visual memory and learning; (2) that the degree of verbal memory deficit would correlate with the degree of speech language impairment; and (3) that speed of processing would be slower in the group with language impairment. METHOD Participants The participants were recruited for a study of genetic factors involved in language impairment. Families in which one or more children had language impairment were recruited through local clinicians and specialist language schools and by advertising through the Association for All Speech Impaired Children (Afasic). Ethics approval was granted by the Guy s Ethics Committee and informed consent was given by parents and, where appropriate, children. Eighty-seven families (252 children) were recruited. Parents were interviewed using the family history interview. 21 Children were included if they were aged 6 years to 16 years 11 months and if they or a sibling in the same age range had a language impairment as defined below. Children were excluded if they had a karyotype abnormality, active epilepsy, an adverse obstetric postnatal history (e.g. neonatal encephalopathy), sensory neural hearing loss, partial or complete visual loss, autism, any structural lesion contributing to speech difficulty, or a performance IQ (PIQ) of less than 80. Sixty-eight families with 152 eligible children (98 males, 54 females) met the criteria. Assessments Speech and language skills were assessed using the Clinical Evaluation of Language Fundamentals (CELF), preschool or school age (Pre-school CELF UK and CELF 3rd edition UK 22 ), and the Edinburgh Articulation Test. 23 Cognitive skills were assessed using the Wechsler Intelligence Scale for Children-III (WISC-III) 24 or the Weschler Preschool and Primary Scale of Intelligence (WIPPSI) revised. 25 Processing speed was calculated from the WISC subtests symbol search and coding. Memory and learning were assessed using the 20-item Children s Non-Word Repetition test (CNRep 26 ), the digit span subtest of the WISC, which incorporates reverse digit span and is a measure of working memory, and the WRAML. 27 The WRAML is constructed for the age range 5 to 17 years and taps both immediate and delayed recall. It consists of three scales verbal memory, visual memory, and learning each with a mean score of 100 and a standard deviation (SD) of 15, and each with three subtests with a mean standard score of 10. The verbal memory scale has the following three subtests: repetition of a random mix of numbers and letters (number letter), repetition of meaningful sentences of increasing length (sentence memory), and recollection of as many parts of two short stories as possible. In the visual memory scale there are the following three subtests: recollection of location pointing in finger windows, drawing memorized designs, and identifying altered elements in two pictures presented sequentially ( picture memory ). There are three subtests on the learning scale: learning a list of non-related words ( verbal learning ), memorizing the spatial location of an increasing number of pictures on a board ( visual learning ), and a paired associate learning task (sound symbol learning), recalling sounds associated with abstract figures. A verbal memory index, visual memory index, learning index, and general memory index standard score are all calculated. There is a delayed recall test consisting of four subtests, verbal learning, visual learning, sound symbol learning, and story memory, classified on a scale of 1 to 5. Normal values for WRAML are based on more than 2000 demographically representative children. 27 Subtests have high internal consistency. The WRAML is moderately related to general cognitive ability. The verbal memory index is significantly related to verbal IQ and full-scale IQ, but not to PIQ. The visual learning index is significantly related to WISC PIQ. Definitions of language impairment are somewhat arbitrary; hence two approaches were used to define the dependent variable language impairment for analysis, one categorical and the other using language as a continuous variable in a dimensional analysis, with potentially more power to detect differences. In the categorical definition, children were classified into three groups on the basis of a current or past language problem, or never having had a language problem. Current language impairment (C-LI) was defined as a score >1.5SD below the mean for the child s chronological age (i.e. 77 on CELF total, expressive, or receptive score) and a PIQ on the WISC WPPSI of 80 or more, as used in the SLIC study. 28 Past language impairment (P-LI) was defined as a current CELF total, receptive, and or expressive 536 Developmental Medicine & Child Neurology 2010, 52:

3 score of less than 1.5SD below the mean and PIQ of 80 or more but language milestone delay in words (>24mo) or phrases (>33mo), and or a problem with articulation at 5 years such that the child was unintelligible to most people. All children in the P-LI and C-LI groups had been referred to speech and language therapy; their history was obtained from parents but, in nearly all cases, speech and language therapy reports were available. The category of no language impairment (N-LI) was defined as no history of language delay or unintelligibility, no current language impairment, and PIQ of 80 or more. Every family included one or more children with either a current or past language impairment and a sibling from any classification group. The dimensional measure of language impairment was calculated using the difference between PIQ and the total CELF score. Both the CELF and WISC have a mean of 100 (SD 15): a score of 0 indicates no discrepancy, and a positive score indicates better PIQ than language (i.e. the higher the score, the more pronounced the language difficulty). The C-LI group was also categorized by receptive or expressive language impairment, calculating a receptive minus expressive score from these two domains in the CELF. Those for whom the difference between receptive and expressive scores was less than 10 points were described as mixed, those with a 10- to 19-point discrepancy (in favour of receptive skills) were termed as having mild expressive disorder, and those with a discrepancy of 20 points or more were described as having severe expressive disorder. Analysis All analyses were undertaken in Stata 9.0 (StataCorp LP, TX, USA) and made use of the cluster option, which corrects standard errors to take account of the non-independence of data when multiple family members are included. 29 RESULTS Of the 152 children eligible for inclusion, 90 from 49 families completed all tests and are included in this analysis: 51 children were assigned to the C-LI group (38 males, 13 females), 13 to the P-LI group (11 males, 2 females), and 26 to the N-LI group (15 males, 11 females). The children with and without complete WRAML data did not differ significantly in age, full-scale IQ, verbal IQ or PIQ, or in total, receptive, or expressive language CELF scores. These sample sizes give us 80% power (a=0.05, two-tailed) to detect effect sizes of 0.68 (comparison of C-LI with N-LI), 0.88 (comparison of C-LI with P-LI), and 0.98 (P-LI vs N-LI). Table I shows the mean age, PIQ, and CELF score of each group. There was no significant difference in age or PIQ between the C-LI group and the N-LI and P-LI groups (age: F 1,48 =2.16, p=0.148; PIQ: F 1,48 =3.57, p=0.064). By definition, total CELF scores in the C-LI group were significantly lower than those in both the N-LI and P-LI groups (F 1,48 =97.2, p<0.0001, and F 1,48 =33, p<0.0001, respectively). In addition, the mean CELF score was significantly lower in the P-LI group than in the N-LI group (F 1,48 =8.64, p=0.005). Table I: Number and sex of children in each group and mean age, nonverbal IQ, and total CELF score Language groups N-LI P-LI C-LI Nr (M F) 26 (15 11) 13 (11 2) 51 (38 13) Age mean (SD) y:mo 10:2 (2:7) 10:2 (2:4) 9:4 (2:5) PIQ mean (SD) 109 (15.8) (15) (15.3) CELF mean (SD) 96.5 (9.2) 88.4 a (8.3) 72.2 b (10.2) a p<0.010 (compared with the current group). b p<0.010 (compared with both the N-LI and the P-LI groups). N-LI, never had language impairment; P-LI, past language impairment; C-LI, current language impairment; PIQ, performance IQ; CELF, Clinical Evaluation of Language Fundamentals. Visual and verbal memory and learning Table II shows the WRAML scores, which reveal clear group differences, with the verbal memory index mean standard score in the C-LI group approximately 11 points below that of the P-LI group (F 1,48 =8.35, p=0.006) and approximately 20 points below that of the N-LI group (F 1,48 =68.06, p<0.0001). Scores in the P-LI and C-LI groups also differed significantly (F 1,48 =11.78, p=0.001). Of the subtests, sentence and number letter memory were significantly lower in the C-LI (F 1,48 =53.43, p<0.0001, and F 1,48 =30.69, p< respectively) and the P-LI (F 1,48 =10.89, p<0.002, and F 1,48 =4.04, p=0.05 respectively) groups than in the N-LI group. Story memory, although lower in the C-LI than in the N-LI group (F 1,48 =13.47, p=0.001), was not significantly different between P-LI and C-LI groups. The visual memory individual subtests did not significantly differ among any of the groups. The visual memory index was lower in the C-LI group than in the N-LI group (F 1,48 =4.88, p=0.032), but the score in the P-LI group was higher than that in the C-LI group (F 1,48 =4.69, p=0.035) and not different from that in the N-LI group. The total learning index score in the C-LI group was significantly lower than in the N-LI group (F 1,48 =12.96, p=0.001). This was probably due to differences in verbal learning, which was significantly lower in the C-LI group than in the N-LI group (F 1,48 =16.64, p<0.0001) and P-LI group (F 1,48 =6, p=0.018), and sound symbol learning, which was significantly lower in the C-LI group than in the N-LI group (F 1,48 =9.06, p=0.004); there was no group difference in visual learning. The mean scores for NWR and digit span are shown in Table III. Both showed a clear group effect, children in the C-LI group scoring significantly lower than those in the N-LI group (NWR: F 1,48 =29.47, p<0.0001; digit span: F 1,48 =26.52, p<0.0001). NWR was significantly lower in the P-LI group than in the N-LI group (F 1,48 =17.72, p=0.0001), whereas for digit span the P-LI group scored significantly higher than the C-LI group (F 1,48 =8.93, p=0.004). Correlation of memory with language impairment To address the prediction that the degree of verbal memory deficit is associated with severity of language impairment, the dimensional language impairment measure described above Memory and Language Impairment in Children Gillian Baird et al. 537

4 Table II: Wide Range Assessment of Memory and Learning (WRAML) group comparison and continuous discrepancy results WRAML index and subtest Mean (SD) score N-LI group P-LI group C-LI group Group differences Dimensional comparison using the predictor PIQ CELF total score (b-coefficient) a General memory index (13.6) (11.6) 92.3 (11.8) N-LI=P-LI b >C-LI b )0.07 Verbal memory index 98.7 (13.6) 89.2 (11.1) 77.7 (10.1) N-LI>P-LI b >C-LI b )0.32 b Visual memory index (12.6) (11.8) 98.4 (12.9) N-LI=P-LI>C-LI c 0.14 Learning index (14.2) (11.5) (13.8) N-LI=P-LI>C-LI b 0.06 Verbal memory subtests Story memory 11.6 (2.9) 11.3 (1.8) 9.2 (2.5) N-LI=P-LI>C-LI b )0.02 Sentence memory 9.11 (2.9) 6.7 (2.4) 4.9 (2.3) N-LI>P-LI b =C-LI c )0.08 b Number letter memory 8.7 (2.3) 6.9 (2.9) 5.7 (2.2) N-LI>P-LI c =C-LI b )0.06 b Visual memory subtests Picture memory 12.2 (2.6) 12.0 (2.7) 11.0 (2.8) 0.01 Finger windows 9.0 (2.0) 9.8 (2.7) 8.4 (2.8) 0.02 Design memory 10.8 (2.8) 11.2 (3.0) 10.3 (3.1) 0.04 Learning subtests Verbal learning 12.7 (2.4) 12.1 (2.0) 10.3 (2.8) N-LI=P-LI>C-LI b )0.02 Visual learning 12.5 (3.3) 11.1 (3.0) 11.2 (3.1) 0.03 Sound symbol 12.6 (2.9) 12.2 (2.7) 10.9 (3.0) N-LI>C-LI b 0.02 Recall scores (scaled 1 5) Verbal learning recall 3.6 (1.0) 4.1 (1.0) 3.7 (1.0) 0.01 Story memory recall 3.7 (0.7) 3.6 (1.2) 3.6 (1.1) 0.00 Sound symbol recall 3.6 (0.9) 3.5 (1.1) 3.2 (1.1) )0.01 Visual learning recall 3.8 (1.0) 3.8 (1.1) 3.8 (1.0) 0.00 Story memory recognition score 3.9 (0.8) 3.9 (1.0) 3.9 (1.3) 0.00 a In the last column a positive discrepancy score indicates a relative strength of WRAML performance compared with language skills, and a negative coefficient indicates that language problems are associated with poorer performance on the WRAML scale. b p< c p< N-LI, never had language impairment; P-LI, past language impairment; C-LI, current language impairment; PIQ, performance IQ; CELF, Clinical Evaluation of Language Fundamentals. Table III: Non-word repetition, digit span, and processing speed (mean [SD]): group comparison and continuous discrepancy Categorical group comparison Language status N-LI P-LI C-LI Group differences Coefficient of discrepancy Non-word repetition standard score 99.8 (11.0) 87.1 (10.5) 84.2 (16.6) N-LI>P-LI a =C-LI a )0.28 a Digit span task (WISC) 10 (3.0) 8.5 (2.4) 6.3 (2.3) N-LI=P-LI>C-LI a )0.05 a Processing speed index (15.2) 97.0 (12.0) 89 (16.7) N-LI>C-LI a 0.15 a p<0.01. N-LI, never had language impairment; P-LI, past language impairment; C-LI, current language impairment; WISC, Wechsler Intelligence Scale for Children. was regressed on WRAML (Table II, final column), NWR, and digit span (Table III, final column). Significant associations were found with the verbal memory index, sentence memory, number lettermemory,nwr,anddigitspan. Processing speed The WISC subscore of processing speed is shown in Table III. There was a significant group difference: C-LI>N- LI (F 1,48 =14.9, p<0.0001). Because one of the subtests used to calculate WISC processing speed (coding) is also a component of PIQ which was used for participant selection, the WISC subscore (perceptual organization index), which excludes coding,wasusedasacovariateinthisanalysis. Association of memory and learning with receptive or expressive problems Among the children in the C-LI group, the discrepancy score (difference between receptive and expressive scores) was less than 10 in 26 children (mixed group). Sixteen children had a mild expressive impairment (defined as a discrepancy score of standard points) and nine had severe expressive (discrepancy score 20 points; Table IV). There was a trend for all memory scores to be more impaired with greater receptive impairment. There were group differences in visual memory index between those with mild and severe expressive difficulties (t 23 =2.14, p=0.043), in learning index between the mixed and severe expressive groups (t 33 =2.28, p=0.029) and between those with mild and severe expressive difficulties (t 23 =2.14, p=0.043), and in the combined general memory index between the mixed and severe expressive groups (t 33 =2.29, p=0.028). Although the observation that memory improvements with age are smaller in children with language impairment could not be directly tested, we examined the association between age at assessment and memory scores in our sample. A posthoc stratified regression analysis was carried out using age and PIQ in the C-LI and N-LI groups to predict the WRAML 538 Developmental Medicine & Child Neurology 2010, 52:

5 Table IV: Wide Range Assessment of Memory and Learning (WRAML) scores (mean and SD) in children with receptive and expressive problems in the current language impairment group Group Mixed (n=26): discrepancy score <10 points Mild expressive difficulty (n=16): discrepancy score points Severe expressive difficulty (n=9): discrepancy score >20 points Significant group comparisons a Verbal memory index 75.8 (8.5) 78.8 (11.0) 81.3 (12.2) Visual memory index 97.7 (12.6) 95.1 (12.9) (12.9) Mild expressive<severe expressive Learning index 105 (13.8) (13.1) (12.1) Mixed=mild expressive<severe expressive General memory index 90.5 (9.8) 90.4 (12.8) (12.9) Mixed<severe expressive a p<0.05. general memory index as well as the three subcomposite indices (learning, visual, and verbal scores), while also covarying for CELF scores. The visual memory subcomposite index showed an increase with age in both C-LI and N-LI groups; the general memory index score increased with age in the N-LI group. However, group differences were apparent in the slope of the general memory index, which significantly decreased with age in the C-LI group (C-LI: b-coefficient=)1.22, p=0.020; N-LI: b-coefficient=0.97, p=0.100), probably because of the significant decrease in subcomponents of learning in the C-LI group (C-LI: b-coefficient=)1.81, p=0.002; N-LI: b-coefficient=1.14, p=0.130), whereas for verbal memory coefficients did not differ significantly (C-LI: b- coefficient=)0.97 p=0.090; N-LI: b-coefficient=0.89, p=0.260). DISCUSSION The purpose of this study was to establish whether there were specific differences in visual and verbal memory, learning, and speed of processing between groups of children with current, past, or no language impairment. All verbal memory measures were significantly impaired in the C-LI group compared with the P-LI and N-LI groups; these results concur with other studies of language impairment. 4,5,18 Visual memory and learning did not differ between children in the C-LI group and their siblings in the N-LI group, and this was also a finding of Archibald and Gathercole. 18 Verbal learning measured by word list learning was impaired, but less severely than verbal memory. A post-hoc finding was that the increase in general memory index with age was markedly different in the C-LI group compared with the N-LI group, apparently largely driven by verbal memory and learning. This might be a result of continuous interaction of language processing, learning, and memory development over time, but other explanations are also possible, including sample heterogeneity. Both the categorical analysis of language impairment with the C-LI, P-LI, and N-LI groups, and the dimensional approach using total CELF scores and discrepancy between language impairment and PIQ, converge in showing that indicators of verbal memory (NWR, digit span, sentence, and number letter memory) are all significantly impaired in children with language impairment, and that the degree of impairment correlates with the severity of language impairment. Thus, there is no evidence of a global memory and learning impairment in children with language impairment compared with their unaffected siblings. An important finding is that delayed recall on the WRAML across all groups was not significantly different in any aspect of memory and learning, suggesting that it is the immediate processing and short-term storage memory that is more of a problem in language impairment. Within the components of the WRAML verbal memory index, scores were particularly low in letter number and sentence memory. The latter is a global test of language and memory competence involving working memory, 30 which has been found to be particularly impaired in language impairment. 12 Memory for stories was impaired but less severely than other verbal memory skills, suggesting that access to meaning supports story memory (incidentally supporting a principal component analysis of the WRAML subtests, 27 which included it in verbal learning rather than memory). Speed of processing measured in the WISC is a combination of visuoperceptual and visuomotor tasks. It is a timed task similar to some components of tasks used in other studies. 4,5 Our finding of slowness of processing, greatest in the C-LI group but also still present in the P-LI group, agrees with other suggestions that this could be an additional impairment in language impairment. 4,5 The finding of a significant level of impairment in verbal memory index, NWR (but not reverse digit span), and processing speed in the P-LI group, even though the language competence had improved on CELF scores, is important educationally and confirms others findings of NWR as a persistent sign of some underlying impairment even in apparently recovered language impairment. 13 Non-word repetition is frequently used as a measure of phonological working memory capacity and is said to be a more sensitive measure of the phonological subcomponent than digit recall because of an absence of any stored lexical specification. However, NWR taps a number of processes: speech perception, motor planning, and articulation, as well as lexical and phonological knowledge, and memory. 31 The letter number sequence task of the WRAML is a serial recall task. In the discrepancy analysis, the letter number and Memory and Language Impairment in Children Gillian Baird et al. 539

6 sentence memory subtests of the WRAML, the reverse digit task of the WISC (a working memory task), and NWR were all impaired and do not differ in effect from each other (which may be a matter of power). Limitations of the study stem from the sample selection, self-referral for a genetic study of families who may not be representative of language impairment, the wide age range and the small size of group with P-LI. The use of a comparison group who are family members may not provide an unbiased comparison if subthreshold language problems and visual processing skills are familial, even though any bias is towards lower sibling scores, and would make the comparisons more conservative rather than increasing the likelihood of finding differences. Further studies using an independent control group are recommended. The use of standardized memory and learning assessment tools plus two well-known verbal memory tests is a strength of the study; the sample size was large enough to identify moderate effect-size differences between the C-LI and N-LI groups. It remains unclear whether the verbal memory impairment that co-occurs with language impairment is a cause of, an association with, or an integral part of speech language processing. The clinical implication is a reminder of the persistent significant problems that children with both early and continuing language impairment are likely to experience in listening to, remembering, and understanding complex language, which is highly relevant in the classroom. Educators will need to take into account the impact on the child and make use of the comparative strengths in visual memory and learning. ACKNOWLEDGMENTS All authors contributed to drafting and revising the paper. GB and VS carried out the study, and ES and KD made substantial contributions to the interpretation and are responsible for the statistical analysis. This study was funded by a grant to GB by the Wellcome Trust, reference B 95 B, in conjunction with a grant to Professor Monaco, Wellcome Trust Centre for Human Genetics Oxford. The funder had no involvement in study design, data collection, or analysis. We thank members of the SLI Consortium and Dr Leila Jannoun for help with data collection. REFERENCES 1. Tomblin JB, Records NL, Buckwalter P, Zhang X, Smith E, O Brien M. Prevalence of specific language impairment in kindergarten children. J Speech Lang Hear Res 1997; 40: Miller CA, Kail R, Leonard LB, Tomblin JB. Speed of processing in children with specific language impairment. J Speech Lang Hear Res 2001; 44: Archibald LMD, Gathercole SE Short term working memory in children with specific language impairment. Int J Lang Commun Disord 2006; 41: Leonard LB, Ellis Weismer S, Miller CA, Francis DJ, Tomblin JB, Kail RV. Speed of processing, working memory, and language impairment in children. J Speech Lang Hear Res 2007; 50: Montgomery JW, Windsor J. Examining the language performances of children with and without specific language impairment: contributions of phonological short-term memory and speed of processing. J Speech Lang Hear Res 2007; 50: Alloway TP, Gathercole SE, Adams AM, Willis C. Working memory and other cognitive skills as predictors of progress towards early learning goals at school entry. Br J Dev Psychol 2005; 23: Gathercole SE, Pickering SJ, Knight C, Stegmann Z. Working memory skills and educational attainment: evidence from National Curriculum assessments at 7 and 15 years of age. Appl Cogn Psychol 2004; 18: Baddeley A. Working memory and language: an overview. J Commun Disord 2003; 36: Shah P, Miyake A. The separability of working memory resources for spatial thinking and language processing: an individual differences approach. J Exp Psychol Gen 1996; 125: Ellis Weismer S, Evans J, Hesketh LJ. An examination of verbal working memory capacity in children with specific language impairment. J Speech Lang Hear Res 1999; 42: Estes KG, Evans JL, Else-Quest NM. Differences in the nonword repetition performance of children with and without specific language impairment: a meta-analysis. J Speech Lang Hear Res 2007; 50: Montgomery JW. Sentence comprehension in children with specific language impairment: the role of phonological working memory. J Speech Lang Hear Res 1995; 38: Bishop DV, North T, Donlan C. Genetic basis of specific language impairment: evidence from a twin study. Dev Med Child Neurol 1995; 37: Newbury DF, Bishop DV, Monaco AP. Genetic influences on language impairment and phonological short-term memory. Trends Cogn Sci 2005; 9: Hoffman LM, Gillam RB. Verbal and spatial information processing constraints in children with specific language impairment. J Speech Lang Hear Res 2004; 47: Bavin EL, Wilson PH, Maruff P, Sleeman F. Spatio-visual memory of children with specific language impairment: evidence for generalized processing problems. Int J Lang Commun Disord 2005; 40: Hick R, Botting N, Conti-Ramsden G. Cognitive abilities in children with specific language impairment: consideration of visuo-spatial skills. Int J Lang Commun Disord 2005; 40: Archibald LMD, Gathercole SE. Visuospatial immediate memory in specific language impairment. J Speech Lang Hear Res 2006; 49: Ferguson AN, Bowey JA. Global processing speed as a mediator of developmental changes in children s auditory memory span. J Exp Child Psychol 2005; 91: Windsor JRL, Milbrath EJ, Carney EJ, Rakowski SE. General slowing in language impairment: methodological considerations in testing the hypothesis. J Speech Lang Hear Res 2001; 44: Bolton P, Macdonald H, Pickles A, et al. A case control family history study of autism. J Child Psychol Psychiatry 1994; 35: Semel E, Wiig EH, Secord W. Clinical Evaluation of Language Fundamentals, 3rd edn UK (CELF-3UK). San Antonio, TX: Psychological Corporation, Anthony A, Bogle D, Ingram TTS, McIsaac MW. E.A.T. The Edinburgh Articulation Test. Edinburgh: Livingstone, Wechsler D. The Manual of the Wechsler Intelligence Scale for Children, 3rd edition UK (WISC-III UK). London: The Psychological Corporation, Wechsler D, Rust J, Golombok S. The Manual of the Wechsler Preschool and Primary Scale of Intelligence, UK edn (WPPSI-III UK). London: The Psychological Corporation, Gathercole SE, Pickering S. Working Memory Test Battery for Children. London: Pearson Assessment, Sheslow D, Adams W. Wide Range Assessment of Memory and Learning. Wilmington, DE: Jastak Associates, The SLI Consortium. A genome-wide scan identifies two novel loci involved in specific language impairment. Am J Hum Genet 2002; 70: Huber PJ. The behavior of maximum likelihood estimates under non-standard conditions. Proceedings of the 5th Berkeley Symposium on Mathematical Statistics and Probability 1967; 1: Alloway TP, Gathercole SE, Willis C, Adams AM. A structural analysis of working memory and related cognitive skills in young children. J Exp Child Psychol 2004; 87: Archibald LM, Gathercole SE. Nonword repetition in specific language impairment: more than a phonological shortterm memory deficit. Psychon Bull Rev 2007; 14: Developmental Medicine & Child Neurology 2010, 52: