The Language, Working Memory, and Other Cognitive Demands of Verbal Tasks

Top Lang Disorders Vol. 33, No. 3, pp. 190–207 Copyright c 2013 Wolters Kluwer Health | Lippincott Williams & Wilkins The Language, Working Memory, and Other Cognitive Demands of Verbal Tasks

Lisa M. D. Archibald

Purpose: To gain a better understanding of the cognitive processes supporting verbal abilities, the underlying structure and interrelationships between common verbal measures were investigated. Methods: An epidemiological sample (n = 374) of school-aged children completed standardized tests of language, intelligence, and short-term and working memory, as well as nonstandardized measures of grammaticality judgment, rapid naming, and sentence recall. Results: Results of a principal component analysis revealed 4 factors corresponding to domain-general working memory, language processing, phonological short-term memory, and fluid reasoning. In corresponding analyses based on younger and older halves of the data, more variables loaded on the fluid reasoning factor for the younger group, and more task variance was explained by the language or phonological storage factors for the older group. The language processing factor correlated with all of the nonstandardized measures, whereas rapid naming was additionally correlated with working memory. Discussion/conclusions: Separable cognitive processes influence performance on common verbal measures, which has implications for assessment and intervention of children with developmental language impairments. Key words: intelligence, short-term memory, specific language impairment, working memory

ERBAL ABILITIES are measured in a logical information, and identify similarities V variety of ways, such as by asking people between verbally presented items. The to follow verbal directions, remember phono- cognitive processes proposed to underlie thesetasksareamatterofsomedebate.Some would argue that these verbal tasks tap a unitary language processing factor (MacDonald & Christiansen, 2002), whereas others have Author Affiliation: School of Communication Sciences and Disorders, Western University, London, suggested that these measures place demands Ontario, Canada. on various cognitive processes such as work- This work was supported by a Natural Sciences and ing memory (Alloway, Gathercole, Willis, & Engineering Research Council of Canada Discovery Adams, 2004) and crystallized knowledge Grant and an Academic Development Fund Grant from (McGrew, 2005). This study aimed to con- Western University. The valuable assistance of par- ticipating school personnel and families is gratefully tribute to understanding of the cognitive acknowledged. processes supporting verbal abilities by inves- The author has indicated that she has no financial and tigating the underlying structure and interre- no nonfinancial relationships to disclose. lationships among common verbal measures. Supplemental digital content is available for this article. Direct URL citations appear in the printed MEASURES OF LINGUISTIC ABILITY text and are provided in the HTML and PDF ver- sions of this article on the journal’s Web site (www. topicsinlanguagedisorders.com). One set of verbal tasks comes from those who are studying language and language dis- Corresponding Author: Lisa M. D. Archibald, PhD, School of Communication Sciences and Disorders, orders. The measures are aimed at assess- Elborn College, Western University, London, Ontario, ing core language ability, including the rules Canada, N6G 1H1 ([email protected]). for governing the content of language rep- DOI: 10.1097/TLD.0b013e31829dd8af resented in the semantic or meaning-based 190

system, and rules for governing the form tense, Pinker (1999) suggested that irregu- of language, including the phonological sys- lar words (e.g., drank) are stored directly tem for representing language sounds and the with their associated meaning whereas regu- morphosyntactic system for combining part lar words (e.g., jumped) must be constructed words, words, and sentences. The rules per- using morphological rules (e.g., add “ed”). taining to language use also may be targeted Consistent with this “words and rules” ac- in some tasks. Traditional tests of core lan- count is evidence of processing differences guage abilities include measures of the abil- in word retrieval tasks requiring or not re- ity to understand language known as recep- quiring the application of a linguistic rule tive language and the ability to produce lan- (Ullman et al., 1997). A preponderance of ev- guage known as expressive language. Recep- idence from connectionist models, however, tive language measures include tasks such as suggests that a single mechanism based on sta- following directions (e.g., “touch the third cir- tistical learning alone is sufficient to account cle and the first black square”) and choosing for the data (Joanisse & Seidenberg, 1999; one of several pictures corresponding to a Rumelhart & McClelland, 1986). Indeed, Bates given word or sentence. Expressive language and Goodman (1997) suggested that gram- tasks include creating a sentence using a given mar and the lexicon were inseparable; they word and completing a sentence using a word proposed a unified lexical account to explain with obligatory morphological markers to cor- a wide body of evidence pertaining to chil- respond to a picture. dren’s emergence of grammar and language The extent to which language tasks de- disorders in older children and adults. Simi- signed to assess semantic, phonological, and larly, Tomblin and Zhang (2006) reported that morphosyntactic knowledge tap a single or a one-dimensional model was sufficient to ac- multiple language processing factor(s) has count for the variance in omnibus language been assessed in only a few studies. Factor test performance, especially for younger chil- analytic studies of standardized tests of lan- dren. These latter findings, then, would pre- guage abilities have reported multiple fac- dict that a single language processing factor tors to explain the variance underlying per- would be sufficient to explain the variance formance (McKay & Golden, 1981; Skarakis- in performance on verbal tasks tapping mor- Doyle, Miller, & Reichheld, 2000). The Clin- phosyntactic or semantic knowledge. It must ical Evaluation of Language Fundamentals be noted, however, that the semantic system (CELF-4; Semel, Wiig, & Secord, 2003) is a encompasses more than just word meaning. standardized test commonly used in clinical It may be that, with a sufficiently compre- practice to identify language and communi- hensive assessment, semantic knowledge it- cation disorders in children. Studies based self may be shown to consist of separable on the normative sample for this test (Semel components. et al., 2003) identified a core language composite with four subfactors: expressive lan- MEASURES OF IMMEDIATE MEMORY guage, receptive language, language content, and language structure (or language memory Another group of verbal tasks have been in older groups). Importantly, the CELF-4 is designed by researchers interested in work- an omnibus language test with multiple sub- ing memory, which can be defined as the tests aimed at sampling not only syntactic and ability to hold information in the current semantic knowledge but also all aspects of focus of attention. According to the tripar- language functioning. tite working memory model of Baddeley and In other work, considerable debate sur- Hitch (1974), domain-specific phonological rounds the question of the separability of mor- and visuospatial short-term memory stores phosyntactic and semantic knowledge specif- hold relevant material for brief periods of ically. Using the example of English past time and have a rehearsal mechanism to

increase retention (Baddeley, 1986, 2003; tion to a greater extent than verbal short-term Logie, 1995). The third component of work- memory tasks, especially in younger children ing memory is the central executive, a (Alloway, Gathercole, & Pickering, 2006). As capacity-limited domain-general resource as- well, executive functions have been found to sociated with attentional control, high-level support visuospatial tasks when the presen- processing activities, and the coordination of tations involve dynamic rather than static im- activities within working memory (Alloway ages (Logie, 1995). et al., 2004; Baddeley, 1998; Baddeley, Della Salla, Gray, Papagno, & Spinner, 1997). Verbal MEASURES OF INTELLIGENCE tasks such as immediate recall of words or digits in a list would be considered to tap phono- A final set of verbal tasks to be consid- logical short-term memory, whereas nonver- ered in this article comprises those that are bal tasks that involve recall of locations, such included in tests of intelligence (IQ). Tests as dots on a grid, would be considered to tap of intelligence traditionally have included in- visuospatial short-term memory. dicators of the ability to analyze information In addition to short-term retention of ma- and solve problems using language-based rea- terial, working memory tasks require addi- soning (verbal IQ) or nonverbal reasoning tional processing of information. Examples of (performance IQ). Alternatively, largely anal- verbal working memory tasks include judg- ogous indicators of crystallized and fluid in- ing the veracity of a sentence while retain- telligence have also been described (Horn & ing the last word or counting similar items Cattell, 1967; McGrew, 2005). Crystallized in- and recalling the total. Such tasks are believed telligence is the ability to use skills, knowl- to be supported by both phonological short- edge, and experience acquired in a lifetime. term memory (for item retention) and the Fluid intelligence or fluid reasoning is the ca- controlled attentional resources of the central pacity to think logically and solve problems executive. Corresponding visuospatial work- efficiently in novel situations, independent of ing memory tasks also tap both visuospatial acquired knowledge. Verbal tasks that involve short-term memory and the central executive application of acquired vocabulary and gen- because they involve recall of visuospatial in- eral knowledge are considered to tap crystal- formation such as locations or orientations lized intelligence, whereas largely nonverbal and visuospatial processing such as mental tasks requiring relational reasoning test fluid rotation. reasoning. Several investigations provide support for Theories of intelligence aim to describe the basic structure of the tripartite work- the full complement of cognitive abilities. ing memory model by demonstrating separa- These theories typically include a single, ble factors reflecting distinct domain-specific general factor (g). General intelligence, or short-term memory stores and a domain- general mental ability, has been suggested general attentional control resource (Bayliss, to account for 40%–50% of the variance Jarrold, Gunn, & Baddeley, 2003; Engle, in performance on all types of cognitive Tuholski, Laughlin, & Conway, 1999; Kane et tests (Kamphaus, Rowe, Winsor, & Kim, al., 2004). It should be noted that some studies 2005). Nevertheless, factor analytic studies have failed to distinguish factors correspond- of intelligence have led to the proposal of ing to visuospatial short-term memory and the a hierarchical three-stratum model known central executive (Gathercole & Pickering, as the Cattell–Horn–Carroll (CHC) theory of 2000; Miyake, Friedman, Rettinger, Shah, & cognitive development (Carroll, 1993; Cattell Hegarty, 2001). Findings from a developmen- & Horn, 1978; McGrew, 2005). The CHC tal study of working memory suggested that vi- theory specifies about 70 specific or narrow suospatial short-term memory tasks may draw abilities at Stratum 1, eight or nine second- on executive resources or controlled atten- order or broad abilities at Stratum 2, and an

overall general factor at Stratum 3. In addition for the separability of demands related to lan- to crystallized intelligence and fluid reason- guage, short-term memory, working memory, ing, the second-order broad abilities may and intelligence across this range of tasks. include quantitative reasoning, visuospatial The study was not designed to investigate the processing, auditory processing, short-term structure of the variation within each domain, memory, long-term storage and retrieval, and for example, within the domains of language processing speed (Flanagan, Ortiz, & Alfonso, (e.g., morphology, syntax). 2007). Although not specifically represented Although the interrelationships between in this model, working memory has been verbal tasks tapping this range of abilities found to be a significant predictor of g have not been systematically assessed pre- (Ackerman, Beier, & Boyle, 2002; Conway, viously, findings of studies assessing links Cowan, Bunting, Therriault, & Minkoff, between some of these measures provide 2002; Engle, Kane, & Tuholski, 1999). It has relevant information regarding the poten- been suggested that working memory is an tial separability of these processes. For ex- executive attentional control mechanism, ample, close and specific associations have variably tapping different cognitive processes been found between a phonological short- represented in the model depending on task term memory measure known as nonword demands. In this way, working memory sup- repetition involving the immediate repeti- ports the active maintenance of goal-relevant tion of novel, multisyllabic words, and one information in the face of interference (Kane aspect of language, vocabulary knowledge & Engle, 2000; Lustig, May, & Hasher, 2001). (Gathercole, Service, Hitch, Adams, & Martin, 1999). RELATIONSHIPS BETWEEN The relationship between nonword repeti- VERBAL TASKS tion and vocabulary is strongest during the early stages of vocabulary development in It is clear that different motivations have led both native (Gathercole, Willis, Emslie, & Bad- to the design of different verbal tasks to assess deley, 1992; Masoura & Gathercole, 2005) different cognitive abilities. However, there and foreign language acquisition (Masoura & has been no systematic assessment of the joint Gathercole, 2005). It has been suggested that variations in performance across verbal tasks phonological short-term memory is particu- commonly included in standardized tests. A larly important to new word learning early in better understanding of the interrelationships vocabulary development when there is little between these tasks is important because available support from existing lexical knowl- such findings will contribute to understand- edge (Gathercole, 2006). This notion would ing of the cognitive processes that support lead to the prediction that phonological short- them. This study explored the principal com- term memory and amassed lexical knowledge ponents necessary to explain the variation in should be separable factors exerting unique performance across tasks from standardized influences on performance depending on the tests of language, working memory, and in- demands of the task. telligence. Because of practical constraints, it Strong positive links have been demon- was not possible to include an adequate sam- strated between the domain-general central pling of all of the tasks considered to tap, for executive component of working memory example, language or crystallized intelligence. and aspects of sentence-level language pro- Nevertheless, at least two measures were in- cessing, including measures of comprehen- cluded that were considered to tap language, sion (Gathercole, Durling, Evans, Jeffcock, & phonological short-term memory, visuospatial Stone, 2008) and production (Adams & Gath- short-term memory, working memory, crys- ercole, 1995). Working memory, rather than tallized intelligence, and fluid reasoning. Ac- short-term memory, has been implicated in cordingly, the main question addressed by this these studies because the memory tasks used study concerned whether there was evidence have imposed verbal processing demands in

addition to the brief retention of phonological working memory, and intelligence in a sample information. of nearly 400 school-aged children. All verbal Working memory may be expected to sup- tasks involving language processing such as port language functioning, given the multi- understanding a sentence or defining a word ple demands of verbal communication, such were expected to load on a single-language as the need to translate thoughts and ideas processing factor, given their overlapping ver- rapidly to verbal codes, produce well-formed bal demands. Nevertheless, a separate compo- discourse, and negotiate social nuances. The nent corresponding to those tasks tapping ap- relationship between working memory and plication of general knowledge (e.g., defining language would not be expected to be spe- a word) would distinguish a crystallized intel- cific, however. Indeed, working memory has ligence factor from the language processing been found to be linked to a number of other factor. It was further hypothesized that tasks complex cognitive activities including mental involving verbal retention would pose differ- arithmetic (DeStefano & LeFevre, 2004) and ent demands than those involving language scholastic achievement (Bayliss et al., 2003). processing, leading to the identification of a These findings would suggest the hypothesis phonological short-term memory factor dis- that domain-general working memory and lan- tinct from the language processing factor. A guage processing may be separable but re- single domain-general working memory factor lated factors influencing language function- was predicted to reflect performance on tasks ing. It should be noted that no association has involving immediate information processing been found between measures of visuospa- and retention across both verbal and visu- tial short-term memory and language (Adams, ospatial domains. Tasks imposing both work- Bourke, & Willis, 1999) except when spatial ing memory and language demands were ex- language (e.g., above, below) is tested specif- pected to have a complex loading pattern, ically (Phillips, Jarrold, Baddeley, Grant, & with variance explained by both the language Karmiloff-Smith, 2004). processing and domain-general working mem- The extent to which the influence of one ory factors. A fluid reasoning factor was pre- aspect of intelligence, crystallized intelli- dicted to account for performance on tradi- gence, may be separable from more specific tional nonverbal intelligence tasks requiring language processing involving morphosyn- solving of novel problems, and a visuospatial tactic or semantic knowledge is unclear. short-term memory factor was expected to ex- Arguably, crystallized knowledge may sub- plain variability on the tasks requiring recall sume both morphosyntactic abilities and of visuospatial information. semantic knowledge. Indeed, descriptions of A second aim of the study was to examine crystallized intelligence include knowledge the relationship between latent variables and application of the grammatical features identified in the factor analysis and other of the language, as well as lexical and general nonstandardized verbal tasks including gram- knowledge (McGrew, 2009). As such, any matical judgment, sentence recall, and rapid distinction between language processing and automatic naming (RAN). Grammaticality crystallized intelligence is subtle, and the judgment, a task requiring the judgment of the design of specific tasks to measure one but grammatical accuracy of a sentence, clearly not the other is challenging. As a result, it taps language processing. However, the task may be very difficult to separate these factors also requires retention of the sentence while empirically except in extremely large studies assessing or comparing different aspects of with multiple measures. the sentence. As such, grammatical judgment tasks may also impose working memory THIS STUDY demands. Sentence recall chiefly requires short-term retention of the phonological This study explored the interrelationships form of the sentence. Nevertheless, language between standardized measures of language, processing can be expected to be recruited as

an obligatory process in sentence recall tasks. dardized tests of language, short-term and Rapid automatic naming, a measure of how working memory, and verbal and nonverbal quickly individuals can name familiar objects intelligence, as well as nonstandardized mea- or letters, imposes language processing and sures of rapid picture naming, rapid letter other additional demands. Critical aspects of naming, grammaticality judgment, and other RAN include the need for focused sustained tasks not reported here. The sentence recall attention over time (Wolf, Bowers, & Biddle, task was completed during the screening visit. 2000) and the dynamic cognitive suppression All tasks were administered individually in a of previous and upcoming responses (Arnell, quiet room in the child’s school by a trained Joanisse, Klein, Busseri, & Tannock, 2009). research assistant. It has been suggested that efficient coding in working memory may account for unique Standardized test battery variance in RAN performance (Arnell et al., Language 2009), as would retrieval from long-term memory and processing speed. Each child completed the four core subtests appropriate for the child’s age for the Composite Language Score from the CELF-IV METHODS (Semel et al., 2003). In the concepts and following directions subtest, the child pointed Participants to aspects of a picture following a spoken in- Data from 374 children (199 boys) aged struction. For recalling sentences, the child 5 years 0 months to 9 years 11 months repeated sentences immediately after hear- were drawn from our existing database in ing them, and for formulated sentences, cre- this study (see Supplemental Digital Con- ated a sentence using a given word. Children tent Appendix A, available at: http://links.lww younger than 9 years completed the word .com/TLD/A21). The database was developed structure subtest, which involves completing in two parts described in detail elsewhere a sentence with the grammatically correct (Archibald, Oram Cardy, Joanisse, & Ansari, word form, and those 9 years and older com- 2013). Briefly, all children in senior kinder- pleted the word classes 2 subtest, which in- garten to Grade 4 were invited to participate volves identifying which two of four words from 34 area schools in the southwest re- have a related meaning. Scaled and standard gion of Ontario, Canada. A total of 1,387 chil- scores were based on published test norms. dren completed three screening tasks mea- suring language (sentence recall, see below), Verbal and visuospatial short-term and reading (Test of Word Reading Efficiency; working memory Torgensen, Wagner, & Raschotte, 1999), and Eight subtests from the Automated Work- math (math fluency from the Woodcock- ing Memory Assessment (Alloway, 2007) were Johnson III Test of Achievement; Woodcock, administered. Measures tapping phonological McGrew, & Mather, 2001). A subset of this short-term memory involved immediate rep- group (n = 374), representing a range of abil- etition of numbers or nonword forms (digit ities on the screening measures, completed recall, nonword recall), and those tapping vi- the measures reported here. suospatial short-term memory required recall of locations (dot matrix, block design). Verbal Procedures working memory measures involved recall of Within 6 months of the initial 10-min counts or final words after counting or pro- screening visit, all participants completed a cessing a sentence (counting recall, listening comprehensive test battery over three 30- to recall), respectively, whereas those involving 40-min study visits occurring approximately visuospatial working memory required the re- 1 week apart. The test battery included stan- call of location or orientation after identifying

a different shape or mentally rotating an im- row to the bottom right. Any naming errors age, respectively (odd one out, spatial recall). were recorded by the administrator. The time Standard scores were based on published test required in seconds to name all the items in norms. the grid was recorded for each task (pictures; letters). Verbal and nonverbal intelligence Children 6 years and older completed the Grammaticality judgment four subtests of the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999), In the grammaticality judgment task, the and those younger than 6 years completed child was asked to decide whether an au- six subtests from the Wechsler Preschool and ditorily presented sentence sounded correct Primary Scale of Intelligence, Third Edition (“sounds like something a person would really (WPPSI-3; Wechsler, 2002). The nonverbal in- say”) or incorrect (“sounds funny or wrong”). telligence subtests included block design, in The task was derived from the grammaticality which children arranged blocks to match a judgment task used by Miller, Leonard, and model, and matrix reasoning, which involved Finneran (2008) and has been described in choosing a picture to complete a pattern; for detail in Noonan, Redmond, and Archibald children younger than 6 years, picture con- (2013). Sentences consisted of 24 sentences = cepts, in which children picked pictures that with a mean length of 10.95 words (SD go together. The verbal intelligence subtests 0.91). Twelve sentences were grammatically included vocabulary, in which children pro- correct (e.g., “You must stir your gravy so it vided definitions, and similarities, which in- doesn’t become too lumpy”; participant re- volved identifying related pictures or describ- sponse: “yes”), and 12 sentences contained ing similarities between words; for children a grammatical error (e.g., “Joan bikes and younger than 6 years, information, in which *skate in the park every day after school”; children provided general world knowledge participant response: “no”). Sentences were such as the color of grass. T scores from the presented via a digital audio recording of WASI and scaled scores from the WPPSI pub- an adult female speaker in fixed random or- lished test norms were converted to standard der on a laptop computer. Each child’s “yes” scores (M = 100; SD = 15). or “no” response was recorded online and scored offline as the total number correct (out Nonstandardized measures of 24). Rapid naming Two RAN tasks were administered, the first Sentence recall involved naming of four letters (g, k, m,and The sentences were taken from Redmond r), and the second, naming of common pic- (2005) and consisted of 16 sentences, each tures (book, dog, chair,andhand). To famil- composed of 10 words (10–14 syllables), iarize the child with the items, an 11 in. × with an equal number of active and passive 8.5 in. paper with a single row of stimuli in- sentences. Although not standardized, this cluding one copy of each of the items in the task has been found to have good sensitivity task was presented for the child to name. The and specificity for identifying children with RAN task for the respective stimuli set was language impairment (Archibald & Joanisse, presented immediately after this. Items were 2009). The sentences were presented in fixed presented in a five-row × 10-column grid on order via a digital audio recording of an adult astandard11× 8.5 in. paper. The children female speaker, using headphones. Sentences were instructed to name each stimulus item were scored online by the research assistant accurately as quickly as possible, beginning at as 2 (correct), 1 (three or fewer errors), or 0 the top left corner and proceeding along each (more than four errors or no response).

RESULTS was performed to enhance the interpretation of the factors. The rotated component matrix Interrelationships between language, factor loadings for each of the 15 measures are memory, and intelligence measures given in Table 1, together with the percentage of variance explained in the rotated solution Appendix A (see Supplemental Digital Con- by each factor. Factors 1 and 2 each account tent, available at: http://links.lww.com/TLD/ for approximately 20% of the variance, respec- A21) presents descriptive statistics for all par- tively, and factors 3 and 4, 10%–13%. ticipants grouped by chronological age. There Factor 1 shows high loadings (>.5) for all of was a significant developmental increase in the verbal and visuospatial working memory > < η2 > raw scores, F 7.1, p .001, p .07 (all tasks and the visuospatial short-term memory = cases, except nonword recall: F(4, 369), p measures. The concepts and following direc- η2 > .031, p .03). Given the wide age range tions subtest also show a considerable sec- of the sample, all analyses involving the stan- ondary loading (>.3) on this factor. Taken dardized tests were completed on the age- together, this pattern reflects domain-general adjusted scores (scaled for the language sub- processing in the context of immediate mem- tests; standard scores for all remaining tests). ory demands, and as such, this factor may be Correlations examining the relationships be- considered a working memory factor. Factor tween the language, intelligence, and short- 2 is associated with high loadings for all of term and working memory measures com- the language subtests and the verbal intel- pleted by all participants revealed significant ligence subtests. The listening recall work- ≥ correlations in all (but one) cases, r .12, ing memory task and the matrix reasoning < p .05 (see Supplemental Digital Content Ap- subtests show minimal secondary loadings pendix B, available at: http://links.lww.com/ TLD/A20). Although the large sample size may Table 1. Factor loadings (>.25) for the ex- have increased the chance of reaching sig- ploratory factor analysis for the entire sample nificance, 68% of the correlations were at or (n = 374) with varimax rotation above the .3 level conventionally considered to have practical significance (Cohen, Cohen, West, & Aiken, 2002). Component To examine the underlying structure in 1234 these data, an exploratory factor analysis was performed. A principal components analysis Digit recall .77 was performed on the data from the lan- Nonword recall .75 guage, intelligence, and short-term and work- Dot matrix .76 ing memory tests completed by all partici- Block recall .74 pants (i.e., excluding word structure, word Listening recall .51 .28 .50 classes, information, and picture concepts, Counting recall .63 .38 which were each completed by a subset of Odd one out .68 .26 participants depending on age). Five factors Spatial recall .73 Concepts and FD .38 .66 met Jolliffe’s (1972) criterion that factors with Recalling sentences .68 .51 eigenvalues greater than 0.7 be retained. Nev- Formulating sentences .78 ertheless, the five-factor solution resulted in Vocabulary .75 one trivial factor defined as a factor with only Similarities .69 .26 one variable loading above .3 (Comrey & Lee, Block design .81 1992). As a result, four factors were retained Mazes reasoning .26 .80 with eigenvalues of 5.8, 1.6, 1.3, and 0.8, % variance explained 21 19 10 13 respectively, which, together accounted for 63.5% of the total variance. A varimax rotation Note.FD= following directions.

(.25–.3) on this factor. This suggests that Fac- Table 2. Factor loadings (>.25) for the ex- tor 2 is a language factor involving domain- ploratory factor analysis for the younger sam- specific verbal processing. Factor 3 shows ple (n = 183) with varimax rotation high loadings for digit recall, nonword recall, recalling sentences, and listening recall. Component Counting recall shows a considerable secondary loading whereas odd-one-out loads 1234 minimally on factor 3. All of these tasks, with the exception of odd-one-out task, explicitly Digit recall .33 .69 Nonword recall .78 involve verbal storage of digits or words, sug- Dot matrix .76 gesting that Factor 3 is a phonological stor- Block recall .70 .26 age factor. Finally, Factor 4 is associated with Listening recall .32 .51 .51 the block design and matrix reasoning sub- Counting recall .48 .53 .31 tests with a minimal loading of the similarities Odd one out .59 .36 .29 subtest. Thus, Factor 4 may be considered a Spatial recall .75 nonverbal or fluid reasoning factor. Concepts and FD .64 .44 Recalling sentences .74 .44 Developmental patterns in the factor Formulating sentences .76 structure Vocabulary .65 .33 To examine the development of these fac- Similarities .43 .61 Block design .63 tors, the sample was split into a younger ver- Mazes reasoning .79 sus older groups on the basis of the median Word structure .70 .39 age of 91 months (7 years 7 months). Children % variance explained 19 19 13 12 younger than 91 months (7 years 6 months or younger) constituted the younger group (n = 183), whereas those 91 months or older (older Note.FD= following directions. than 7 years 6 months) formed the older group (n = 191). For both of these groups, the sam- digit recall, listening recall, and similarities ple size exceeded the recommended mini- test show considerable secondary loadings. mum for factor analysis of 10 observations per This suggests that Factor 1 is a language fac- variable (Nunnally, 1978). The principal com- tor involving domain-specific verbal process- ponents analysis completed on the younger ing, corresponding to the second factor in group included all of the 15 variables present the analysis of the entire sample. Factor 2 in the previous analysis, as well as the scores shows high loadings for three of the four for the word structure subtest completed by working memory tasks, with the remaining all participants in this group. The four-factor verbal working memory task, counting recall, solution resulted in factors with eigenvalues showing considerable secondary loadings. As of 5.9, 1.7, 1.4, and 1.0, respectively, which well, the concepts and following directions together accounted for 63% of the total vari- and word structure subtests have consider- ance. The rotated component matrix factor able secondary loadings on this factor. This loadings for each of the 16 measures are given pattern is consistent with that seen for Fac- in Table 2, together with the percentage of tor 1 in the entire sample and suggests that variance explained in the rotated solution by Factor 2 in the younger sample is a working each factor. Factors 1 and 2 each account memory factor. It should be noted that the for 19% of the variance, and Factors 3 and 4, variance explained in the rotated solution for 12%–13%. Factors 1 and 2 was 19% for each and thus the Factor 1 is associated with high loadings of reversal of the working memory and language all of the language subtests (including word factors relative to the entire sample may not structure) and the vocabulary subtest. The be meaningful. Factor 3 shows high loadings

for digit recall, listening recall, and counting Table 3. Factor loadings (>.25) for the ex- recall and considerable secondary loadings for ploratory factor analysis for the older sample recalling sentences and odd one out. As in the (n = 191) with varimax rotation previous analysis, all of these tasks with the exception of the odd-one-out task explicitly Component involve verbal storage of digits or words, suggesting that Factor 3 is a phonological storage 1234 factor. Factor 4 is associated with the block design, matrix reasoning, and similarities sub- Digit recall .81 Nonword recall .63 tests, with considerable secondary loadings Dot matrix .75 of the vocabulary and counting recall subtests Block recall .75 and minimal loadings of the odd-one-out sub- Listening recall .50 .27 .46 test. As before, Factor 4 may be considered a Counting recall .74 .29 fluid reasoning factor; however, the factor is Odd one out .71 .31 not strictly nonverbal in this younger group, Spatial recall .70 as both verbal and nonverbal tasks load on the Concepts and FD .33 .69 factor. Recalling sentences .63 .59 The corresponding four-factor solution for Formulating sentences .76 the older group resulted in factors with eigen- Vocabulary .75 values of 6.1, 1.6, 1.1, and 0.8, respectively, Similarities .75 Block design .26 .83 which together accounted for 64% of the total Mazes reasoning .28 .79 variance. The rotated component matrix fac- % variance explained 22 20 13 10 tor loadings for each of the 15 measures completed by all participants are given in Table 3, together with the percentage of variance ex- Note.FD= following directions. plained in the rotated solution by each factor. Factors 1 and 2 individually account for ing subtests. The similarities subtest does not approximately 20% of the variance, respec- show a minimal loading on this factor. Thus, tively, and Factors 3 and 4, 10%–13%. The fac- Factor 4 may be considered a fluid reasoning tor structure and pattern of loadings exactly factor, with only nonverbal tasks loading on mirrored those for the entire sample. Factor the factor. 1, the working memory factor, shows high To summarize the factor analyses, factors loadings for all of the verbal and visuospatial corresponding to working memory, language working memory tasks and the visuospatial processing, phonological storage, and nonver- short-term memory measures, with a consid- bal fluid reasoning explained 63% of the vari- erable secondary loading of the concepts and ance in language tasks and verbal and non- following directions subtest. Factor 2, the lan- verbal short-term memory, working memory, guage factor, is associated with high loadings and intelligence measures in a sample of 374 of all of the language subtests and the ver- children aged 5 years 0 months to 9 years bal intelligence subtests. The listening recall 11 months. There were some notable differ- working memory task and the matrix reason- ences in the loadings for each variable across ing subtests show minimal secondary loadings these four factors for the younger versus on this factor. Factor 3, the phonological stor- older groups. The greatest differences were age factor, shows high loadings for digit recall, for the fluid reasoning factor, which included nonword recall, and recalling sentences, with loadings for the two nonverbal intelligence considerable secondary loadings for listening measures only for the older group. For the recall and odd-one-out and minimal loading younger group, however, the fluid reasoning of counting recall. Finally, Factor 4 is associ- factor also included some loading from sev- ated with the block design and matrix reason- eral other variables (both verbal intelligence

measures, counting recall, odd one out, block the language processing (55%) than phono- recall). The older and younger groups also dif- logical storage factors (20%) for the younger fered in the proportion of variance in perfor- group. In addition to loading on the working mance (as reflected by squared factor load- memory factor for both groups (25%–26% of ings) explained for some tasks. For the older variance explained), listening recall was asso- group, the majority of the variance (54%) in ciated with somewhat higher loadings on the counting recall was explained by the work- language processing and phonological storage ing memory factor (compared with 8% by the factors for the younger (36% total for both fac- phonological storage factor), whereas count- tors) than older group (28%). ing recall performance was explained relatively equally by both the working memory Relationships between identified factors (23%) and phonological storage factors (27%) and related cognitive measures for the younger group. Digit recall loaded Both by way of validating the identified on the phonological storage factor for both factors and by examining their relationship groups (47% and 65% of the variance ex- to related measures, correlations were com- plained for the younger and older groups, puted between the factor scores and the respectively), but a proportion of the vari- rapid object naming, rapid letter naming, ance was also explained by the language pro- grammatical judgment, and sentence recall cessing factor for the younger group (11%). tasks separately for the entire sample and Variance in recalling sentences was explained for the younger and older groups. Table 4 relatively equally by the language processing presents the three correlation matrices. For (40%) and phonological storage factors (35%) the entire sample, the working memory factor for the older group but had higher loadings on was moderately correlated with the rapid

Table 4. Correlations between the identiﬁed factors and four nonstandardized measures for the entire sample and for the younger and older groups

Nonstandardized Measures

Grammaticality Sentence Factors RAN Objects RAN Letters Judgment Recall

Entire sample (n = 374) Working memory − .15** − .13* − .04 .02 Language − .24** − .22** .33** .47** Phonological storage .07 .08 .05 .12* Fluid reasoning − .08 − .11* .21** .18** Younger group (n = 183) Working memory − .08 − .07 − .10 − .02 Language − .20** − .23** .25** .42** Phonological storage − .04 − .05 .06 .17* Fluid reasoning − .22** − .28** .19* .34** Older group (n = 191) Working memory − .23** − .27** .04 .02 Language − .26** − .27** .43** .51** Phonological storage − .08 − .08 .16* .37** Fluid reasoning − .11 − .01 .28** .18*

Note.RAN= rapid automatic naming. *p < .05; **p < .01.

naming but not the grammaticality judgment older group by either the language process- or sentence recall tasks. The relationship with ing factor (digit recall, recalling sentences, the rapid naming tasks was stronger in the listening recall) or the phonological storage older group and not significant in the younger factor (counting recall, listening recall). Of the group. In contrast, the fluid reasoning factor nonstandardized rapid naming, grammatical- was strongly correlated with the rapid naming ity judgment, and sentence recall measures, tasks for the younger age group but not for the language factor was associated with all, the older age group, resulting in mixed results regardless of age. Rapid naming additionally for the entire sample (moderate correlation correlated with working memory in the older for one rapid naming task but not the other). group and with fluid reasoning in the younger The language factor was correlated with all group. Phonological storage was consistently measures in all of the analyses, whereas the linked with sentence recall regardless of age phonological storage factor was correlated and with grammaticality judgment in the older with the sentence recall task in all cases and group. with the grammaticality judgment task only This study provides evidence that signifi- for the older group. It must be acknowledged cant variability in children’s performance on that the correlations between the nonstan- common verbal tasks included in standard- dardized sentence recall task and the language ized tests is explained by separable factors and phonological short-term memory factors related to domain-general working memory, are not surprising, given that one of the language processing, phonological short- subtests contributing to both of the latter two term memory, and, depending on age, fluid factors was the recalling sentences subtest of reasoning. The current results are consistent the CELF-4 (Semel et al., 2003). Nevertheless, with the tripartite working memory model inclusion of the correlations with the non- (Baddeley & Hitch, 1974) distinguishing standardized sentence recall task contributes a domain-general processing component to the overall pattern of the results. corresponding to the central executive and a domain-specific phonological short-term DISCUSSION memory component corresponding to the phonological loop (Bayliss et al., 2003; Engle This study examined the factors and inter- et al., 1999; Kane et al., 2004). relationships underlying traditional measures As predicted, verbal tasks requiring only re- of language, short-term and working mem- tention without additional information pro- ory, and verbal and nonverbal intelligence in cessing loaded on the phonological short- children. Four factors were found to best ex- term memory factor. Verbal recall tasks with plain variance in performance on these mea- added processing demands also loaded on sures: a working memory factor characterized the domain-general working memory factor by domain-general processing in the context together with corresponding nonverbal vi- of immediate memory demands, a language suospatial tasks. The distinction between a factor involving domain-specific verbal pro- phonological short-term storage mechanism cessing, a phonological storage factor, and supporting brief verbal recall and an addi- a fluid reasoning factor. Corresponding anal- tional resource supporting immediate pro- yses based on younger and older halves of cessing of verbal information is indicated by the data revealed some interesting develop- the pattern of loadings on these factors for the mental patterns in these factors. Generally, other language tasks included in the analyses. more variables loaded on the fluid reasoning For example, recalling sentences places high factor for the younger group than for the demands on storage and loaded on the phono- older group. As well, a greater proportion logical short-term memory factor, whereas fol- of the variance on several verbal tasks was lowing directions requires some immediate explained in the younger group than in the processing of verbal information and loaded

on the domain-general working memory fac- ing memory factor. All tasks involving some tor. Similarly, the nonstandardized sentence processing of linguistic relationships loaded repetition task correlated with the phonolog- on this factor including understanding a sim- ical short-term memory factor whereas the ple sentence, following directions, identifying rapid naming tasks requiring dynamic cog- commonalities between words, formulating a nitive activation and suppression correlated sentence given a word, and giving the defini- with the working memory factor. Of less inter- tion of a word. The latter two tasks required est to this study was the lack of a factor corre- the expression of complex language and were sponding to visuospatial short-term memory the only tasks to load solely on the language specifically. As has been suggested in previ- processing factor. ous studies (Alloway et al., 2006; Gathercole Interestingly, the recalling sentences test & Pickering, 2000; Logie, 1995; Miyake et al., loaded on the language processing factor, 2001), it may be that the act of briefly main- suggesting that sentence repetition involves taining visuospatial information in dynamic sentence processing and retention. Converg- tasks such as those in this study imposed pro- ing evidence that semantic processing sup- cessing demands, at least for our young par- ports recall comes from findings of more ac- ticipants. curate recall of words than nonwords (Hulme, Interestingly, this study is one of the first Maughan, & Brown, 1991), related than unre- to provide evidence of the separability of lated word lists (Poirier & Saint-Aubin, 1995), domain-general working memory and lan- and story-based than unrelated sentence lists guage processing factors. Contrary to sugges- (Jeffries, Lambon Ralph, & Baddeley, 2004). tions that verbal working memory and lin- Additional evidence of the separability of the guistic measures tap a unitary language pro- language processing and working memory cessing construct (MacDonald & Christiansen, factors was the results of the correlations 2002), tests and nonstandardized measures in- with the nonstandardized tasks. In particular, volving verbal information processing loaded grammaticality judgment was correlated with on, or correlated with, one or both fac- the language but not working memory factor, tors in this study. The tasks loading on the showing a distinct relationship to language domain-general working memory component processing but not domain-general working required the immediate processing of rela- memory. It may be that this pattern of findings tively simple verbal information while hold- informs MacDonald and Christiansen’s (2002) ing some material in mind. For example, the notion that language processing is not sup- listening recall task requires true or false judg- ported by a capacity-limited working mem- ments about short sentences made up of fre- ory resource. In particular, the present results quent words and tapping common knowl- indicate that nontrivial language processing edge as in “Lions have four legs.” The con- tasks without specific storage demands are cepts and following directions language test supported by a language processing factor but also loaded on the domain-general working not a working memory factor. memory factor to some extent. At least some The language processing factor was found of the items in this subtest rely on the pro- to be distinct from the phonological short- cessing of familiar concepts and vocabulary term memory factor in this study as well. and require retention of several items such as Tasks requiring the retention of verbal infor- “Point to the second black shoe and the third mation processing but no additional informa- white ball.” Finally, the rapid naming tasks tion processing such as digit recall and non- also correlated with the working memory word recall loaded only on the phonological factor. short-term memory factor. Tasks with addi- Tasks requiring the processing of linguistic tional processing demands also loaded on an- relationships loaded on a language processing other corresponding factor, the language pro- factor separate from the domain-general work- cessing factor for tasks requiring processing

of linguistic relationships such as listening re- requires further investigation. The language call and recalling sentences and the working processing factor was distinguished from a memory factor for those requiring immediate fluid reasoning factor in this study. Only the domain-general processing such as odd one two nonverbal intelligence measures consis- out and counting recall. tently loaded highly on this factor. The corre- Phonological short-term memory has been lations between the fluid reasoning factor and closely linked to language learning. Consis- the nonstandardized grammaticality judgment tently, children with an unexplained language and sentence recall tasks may have arisen due learning deficit known as specific language to shared variance with the other factors. impairment (SLI) have been found to have The current findings provide some prelimi- a phonological short-term memory deficit nary indications of developmental changes in (Archibald & Gathercole, 2006; Archibald & the cognitive processes supporting the verbal Joanisse, 2009). Nevertheless, a phonologi- tasks in this study. For example, digit recall cal short-term memory deficit alone has not loaded on the language processing factor for been found to lead to a lasting language learn- the younger group but not the older group. As ing impairment (Gathercole, Tiffany, Briscoe, well, more variance was explained for the lis- Thorn, & ALSPAC Team, 2005). Gathercole tening recall and sentence recall tasks by the (2006) suggested distinct roles in the learn- language processing factor for the younger ing of new words forms for both phonologi- group than the older group. These findings cal short-term memory and long-term memory may reflect the diminishing language process- for language content, with the former being ing demands imposed by these tasks as linguis- important to the brief retention of unfamiliar tic competency improves. The counting recall phonological forms and the latter supporting task, and to some extent, the listening recall learning when new information activates ex- task loaded more strongly on the phonologi- isting representations. Interestingly, the non- cal short-term memory factor for the younger standardized grammatical judgment task cor- group than the older group. It has been sug- related with the language processing factor as gested that developmental increases in work- expected, but it showed only a weak relation- ing memory tasks such as these occur due to ship with phonological short-term memory in the more efficient encoding and rehearsal of the older group. It may be that the task did not items enabled by growing improvements in place sufficient demands on short-term mem- linguistic abilities (Flavell, Beach, & Chinsky, ory skills or only taxed immediate memory in 1966; Ottem, Lian, & Karlsen, 2007). some participants (see Noonan et al., 2013). It follows from these results that the stor- This study failed to distinguish a language age demands of working memory tasks con- processing from a crystallized knowledge fac- strain performance to a greater extent in tor, with both language and verbal intelli- younger children than in older children. Nev- gence tasks loading on the same factor. Al- ertheless, the primary factor identified for the though this result might suggest a unitary younger group was the language processing long-term language knowledge and process- factor whereas the primary factor for the older ing factor, the finding may be related to the group was the working memory factor. This limited range of language processing tasks in- reversal may reflect the developing nature of cluded in this study. For example, no specific the linguistic system for the younger group grammatical tests administered to all partic- and the importance of working memory once ipants were included in the analysis. Given language abilities are largely intact for the that a grammatical learning deficit has been older group. suggested as a hallmark of SLI (Rice, 2003), Finally, more variables loaded on the fluid the extent to which grammatical learning can reasoning factor for the younger group than be distinguished from other types of language for the older group, including both verbal processing in a large developmental group (counting recall; vocabulary) and nonverbal

tasks (odd one out; block recall). Similarly, ing investigated. In addition, the sample size the nonstandardized rapid naming tasks were was relatively small for an investigation of correlated with the fluid reasoning factor in this nature. As well, the factor structures may the younger group and the working memory have been influenced by the tests and subtests factor in the older group. To our knowledge, used. However, any relationship between fac- these findings are the first to suggest that tors and tests was not direct in that some performance on working memory tasks may measures did span a number of factors and be supported by different cognitive processes some factors included loadings of subtests inyoungerchildrenthaninolderchildren. from more than one test. Overall, the findings Specifically, such tasks may be particularly provide a useful exploration of the underlying novel to young children, leading to the processes supporting verbal tasks commonly recruitment of fluid reasoning for younger used in various standardized tests purporting children than for older children. to tap different cognitive mechanisms.

Clinical implications CONCLUSION The current results have important clinical implications for assessing the verbal abilities A large sample of school-aged children of children with language learning difficul- completed standardized verbal tasks designed ties. Measures commonly used in standard to measure language abilities, phonological language batteries include following direc- short-term memory, visuospatial short-term tions, repeating sentences, and formulating memory, working memory, crystallized intel- a sentence from a given word, and questions ligence, and fluid intelligence. They also com- often arise regarding the extent to which pleted nonstandardized grammatical, naming, these tasks tap cognitive processes additional and recall tasks. A four-factor structure ex- to language. The findings from this study plained 63.5% of the variance on the stan- suggest that these language tasks primarily dardized tests, with separable factors corre- tap a language processing factor separable sponding to domain-general working mem- from both working memory and phonolog- ory, language processing, phonological short- ical short-term memory. It is important to term memory, and fluid reasoning. The per- remember, however, that deficits in the latter formance on some measures was explained two processes could affect performance on to a greater extent by the language pro- these language tasks. cessing, phonological short-term memory, and fluid reasoning factors for younger chil- Study limitations dren than for older children. These results One of the main limitations of this study were suggested to reflect a developmental is the small number of tasks corresponding change in the processes tapped by these to each of the hypothesized constructs be- measures.

REFERENCES

Ackerman, P. L., Beier, M. E., & Boyle, M. B. (2002). Indi- children. Journal of Speech and Hearing Research, vidual differences in working memory within a nomo- 38, 403–414. logical network of cognitive and perceptual speed Alloway, T. P. (2007). The automated working memory abilities. Journal of Experimental Psychology: Gen- assessment. London: Harcourt Assessment. eral, 131, 567–605. Alloway, T. P., Gathercole, S. E., & Pickering, S. J. (2006). Adams, A.-M., Bourke, L., & Willis, C. S. (1999). Work- Verbal and visuo-spatial short-term and working mem- ing memory and spoken language comprehension in ory in children: Are they separable? Child Develop- young children. International Journal of Psychology, ment, 77, 1698–1716. 34, 364–373. Alloway, T. P., Gathercole, S. E., Willis, C., & Adams, Adams, A.-M., & Gathercole, S. E. (1995). Phonological A.-M. (2004). A structural analysis of working mem- working memory and speech production in preschool ory and related cognitive skills in early childhood.

Journal of Experimental Child Psychology, 87, Conway,A.R.A.,Cowan,N.,Bunting,M.F.,Therriault, 85–106. D. J., & Minkoff, S. R. B. (2002). A latent variable anal- Archibald, L. M. D., & Gathercole, S. E. (2006). Short- ysis of working memory capacity, short-term memory term and working memory in specific language im- capacity, processing speed, and general fluid intelli- pairment. International Journal of Language and gence. Intelligence, 30, 163–183. Communication Disorders, 41, 675–693. DeStefano, D., & LeFevre, J. (2004). The role of working Archibald, L. M. D., & Joanisse, M. F. (2009). On the sen- memory in mental arithmetic. European Journal of sitivity and specificity of nonword repetition and sen- Cognitive Psychology, 16, 353–386. tence recall to language and memory impairments in Engle, R. W., Kane, M. J., & Tuholski, S. W. (1999). Individ- children. Journal of Speech, Language, and Hearing ual differences in working memory capacity and what Research, 52, 899–914. they tell us about controlled attention, general fluid in- Archibald, L. M. D., Oram Cardy, J., Joanisse, M. F., telligence, and functions of the prefrontal cortex. In A. & Ansari, D. (2013). Language, reading and math Miyake & P. Shah (Eds.), Models of working memory: learning profiles in a school age epidemiological Mechanisms of active maintenance and executive sample. Manuscript submitted for publication. control (pp. 102–134). New York: Cambridge Univer- Arnell, K. M., Joanisse, M. F., Klein, R. S., Busseri, M., sity Press. & Tannock, R. (2009). Decomposing the relation be- Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, tween Rapid Automatized Naming (RAN) and reading A. R. A. (1999). Working memory, short-term mem- ability. Canadian Journal of Experimental Psychol- ory, and general fluid intelligence: A latent variable ogy, 63, 173–184. approach. Journal of Experimental Psychology Gen- Baddeley, A. D. (1986). Working memory. Oxford, Eng- eral, 128, 309–331. land: Oxford University Press. Flanagan, D. P., Ortiz, S. O., & Alfonso, V. C. (2007). Baddeley, A. D. (1998). Recent developments in working Essentials of cross-battery assessment with C/D ROM memory. Current Opinion in Neurobiology, 8, 234– (2nd ed.). New York: Wiley. 238. Flavell, J. H., Beach, D. R., & Chinsky, J. M. (1966). Sponta- Baddeley, A. D. (2003) Working memory: Looking back neous verbal rehearsal in a memory task as a function and looking forward. Nature Reviews Neuroscience, of age. Child Development, 37, 283–299. 4(10), 829–839. Gathercole, S. E. (2006). Complexities and constraints in Baddeley, A. D., Della Salla, S., Papagno, C., & Spinnler, nonword repetition and word learning. Applied Psy- H. (1997). Dual task performance in dysexecutive and cholinguistics, 27, 599–613. non-dysexecutive patients with a frontal lesion. Neu- Gathercole, S. E., Durling, M., Evans, S., Jeffcock, E., & ropsychology, 11(2), 187–194. Stone, S. (2008). Working memory abilities and chil- Baddeley, A. D., & Hitch, G. (1974). Working memory. dren’s performance in laboratory analogues of class- In G. Bower (Ed.), The psychology of learning and room activities. Applied Cognitive Psychology, 22, motivation (Vol. 8, pp. 47–90). New York: Academic 1019–1037. Press. Gathercole, S. E., & Pickering, S. J. (2000). Assessment of Bates, E., & Goodman, J. C. (1997). On the inseparability working memory in six- and seven-year old children. of grammar and the lexicon: Evidence from acquisi- Journal of Education Psychology, 92, 377–390. tion, aphasia, and real-time processing. Language and Gathercole, S. E., Service, E., Hitch, G. J., Adams, A., & Cognitive Processes, 12, 507–584. Martin, A. J. (1999). Phonological short-term memory Bayliss, D. M., Jarrold, C., Gunn, D. M., & Baddeley, A. D. and vocabulary development: Further evidence on the (2003). The complexities of complex span: Explaining nature of the relationship. Applied Cognitive Psychol- individual differences in working memory in children ogy, 13, 65–77. and adults. Journal of Experimental Psychology— Gathercole, S. E., Tiffany, C., Briscoe, J., & Thorn, A. S. C. General, 132, 71–92. ALSPAC Team. (2005). Developmental consequences Carroll, J. B. (1993). Human cognitive abilities: A sur- of poor phonological short-term memory function in vey of factor-analytic studies. Cambridge, England: childhood: A longitudinal study. Journal of Child Psy- Cambridge University Press. chology and Psychiatry and Allied Disciplines, 46, Cattell, R. B., & Horn, J. L. (1978). A check on the theory 598–611. of fluid and crystallized intelligence with description Gathercole, S. E., Willis, C., Emslie, H., Baddeley, A. D. of new subtest designs. Journal of Educational Mea- (1992). Phonological memory and vocabulary devel- surement, 15, 139–164. opment during the early school years: A longitudinal Cohen, P., Cohen, J., West, S. G., & Aiken, L. S. (2002). study. Developmental Psychology, 28, 887–898. Applied multiple regression/correlation analysis for Horn, J. L., & Cattell, R. B. (1967). Age differences in fluid the behavioral sciences (3rd ed.). Mahwah, NJ: Erl- and crystallized intelligence. Acta Psychologica, 26, baum. 107–129. Comrey, A. L., & Lee, H. B. (1992). A first course in factor Hulme, C., Maughan, S., & Brown, G. D. A. (1991). Mem- analysis. Hillsdale, NJ: Erlbaum. ory for familiar and unfamiliar words: Evidence for a

long-term memory contribution to short-term memory Miller, C. A., Leonard, L. B., & Finneran, D. (2008). Gram- span. Journal of Memory and Language, 30, 685– maticality judgements in adolescents with and with- 701. out language impairment. International Journal of Jefferies, E., Lambon Ralph, M. A., & Baddeley, A. D. Language and Communication Disorders, 43, 346– (2004). Automatic and controlled processing in sen- 360. tence recall: The role of long-term and working mem- Miyake, A., Friedman, N. P., Rettinger, D. A., Shah, P., ory. Journal of Memory and Language, 51, 623–643. & Hegarty, M. (2001). How are visuo-spatial working Joanisse, M. F., & Seidenberg, M. S. (1999). Impairments memory, executive functioning and spatial abilities in verb morphology after brain injury: A connection- related? A latent-variable analysis. Journal of Experi- ist model. Proceedings of the National Academy of mental Psychology; General, 130, 621–640. Sciences U S A, 96, 7592–7597. Noonan, N. B., Redmond, S. M., & Archibald, L. M. D. Jolliffe, I. T. (1972). Discarding variables in a principal (2013). Differentiating linguistic and working mem- component analysis, I. Artificial data. Applied Statis- ory demands on children’s grammaticality judg- tics, 21, 160–173. ments. Manuscript submitted for publication. Kamphaus,R.W.,Rowe,E.W.,Winsor,A.P.,&Kim, Nunnally, J. C. (1978). Psychometric theory (2nd ed.). S. (2005). A history of intelligence test interpretation. New York: McGraw Hill. In D. Flanagan & P. Harrison (Eds.), Contemporary Ottem, E. J., Lian, A., & Karlsen, P. J. (2007). Reasons Intellectual Assessment (3rd ed., pp. 23–38). New for the growth of traditional memory span across age. York: Guilford Press. European Journal of Cognitive Psychology, 19, 233– Kane, M. J., & Engle, R. W. (2000). Working memory ca- 270. pacity, proactive interference, and divided attention: Phillips, C. E., Jarrold, C., Baddeley, A. D., Grant, J., & Limits on long-term memory retrieval. Journal of Ex- Karmiloff-Smith, A. (2004). Comprehension of spatial perimental Psychology: Learning, Memory, and Cog- language terms in Williams syndrome: Evidence for an nition, 26, 336–358. interaction between domains of strength and weak- Kane,M.J.,Hambrick,D.Z.,Tuholski,S.W.,Wilhelm,O., ness. Cortex, 40, 85–101. Payne, T. W., & Engle, R. W. (2004). The generality of Pinker, S. (1999). Words and rules. New York: Harper working-memory capacity: A latent-variable approach Perennial. to verbal and visuo-spatial memory span and reason- Poirier, M., & Saint-Aubin, J. (1995). Memory for related ing. Journal of Experimental Psychology: General, and unrelated words: Further evidence on the influ- 133, 189–217. ence of semantic factors in immediate serial recall. Logie, R. H. (1995). Visuo-spatial working memory. Quarterly Journal of Experimental Psychology, 48A, Hove, England: Erlbaum. 384–404. Lustig, C., May, C. P., & Hasher, L. (2001). Working mem- Redmond, S. M. (2005). Differentiating SLI from ADHD us- ory span and the role of proactive interference. Jour- ing children’s sentence recall and production of past nal of Experimental Psychology: General, 130(2), tense morphology. Clinical Linguistics and Phonet- 199–207. ics, 19, 109–127. MacDonald, M. C., & Christiansen, M. H. (2002). Reassess- Rice, M. L. (2003). A unified model of specific and gen- ing working memory: A comment on Just & Carpenter eral language delay: Grammatical tense as a clinical (1992) and Waters & Caplan (1996). Psychological Re- marker of unexpected variation. In Y. Levy & J. Schaef- view, 109, 35–54. fer (Eds.), Language competence across populations: Masoura, E. V., & Gathercole, S. E. (2005). Phonological Toward a definition of specific language impair- short-term memory skills and new word learning in ment (pp. 63–94). Mahwah, NJ: Erlbaum. young Greek children. Memory, 13, 422–429. Rumelhart, D. E., & McClelland, J. L. (1986). On learn- McGrew, K. (2009). Editorial. CHC theory and the human ing the past tenses of English verbs. In J. L. McClel- cognitive abilities project. Standing on the shoulders land, & D. E. Rumelhart & the PDP Research Group. of the giants of psychometric intelligence research. Parallel distributed processing: Explorations in the Intelligence, 37, 1–10. microstructure of cognition. Volume 2: Psychologi- McGrew, K. S. (2005). The Cattell-Horn-Carroll (CHC) cal and biological models (pp. 216–271). Cambridge, theory of cognitive abilities. Past, present and future. MA: Bradford Books/MIT Press. In D. P. Flanagan & P. L. Harrison (Eds.), Contem- Semel, E., Wiig, E. H., & Secord, W. A. (2003). Clinical porary intellectual assessment. Theories, tests, and evaluation of language fundamentals, Fourth Edi- issues (pp. 136–202). New York: Guilford Press. tion (CELF-4). Toronto, Ontario, Canada: The Psycho- McKay, S. E., & Golden, C. J. (1981). The assessment of logical Corporation/A Harcourt Assessment Company. specific neuropsychological skills using scales derived Skarakis-Doyle, E., Miller, L. T., & Reichheld, M. (2000). from neuropsychological skills using scales derived Construct validity as a foundation of evidence-based from factor analysis of the Luria-Nebraska Neuropsy- practice: The case of the preschool language as- chological Battery. International Journal of Neuro- sessment instruction. Journal of Speech-Language science, 14, 189–204. Pathology and Audiology, 24(4), 180–192.

Tomblin, J. B., & Zhang, X. (2006). The dimensionality Wechsler, D. (1999). Wechsler Abbreviated Scale of In- of language ability in school-age children. Journal of telligence. San Antonio, TX: The Psychological Corpo- Speech, Language, and Hearing Research, 49, 1193– ration. 1208. Wechsler, D. (2002). Wechsler Preschool and Primary Torgensen, J. K., Wagner, R. K., & Raschotte, C. A. (1999). Scale of Intelligence–III. San Antonio, TX: The Psy- Test of word reading efﬁciency.Austin,TX:AGS chological Corporation. Publishing. Wolf, M., Bowers, P., & Biddle, K. (2000). Naming- Ullman, M. T., Corkin, S., Coppola, M., Hickok, G., Grow- speed processes, timing, and reading: A concep- don, J. H., Korosheta, J. H., et al. (1997). A neural tual review. Journal of Learning Disabilities, 33, dissociation within language: Evidence that the men- 387–407. tal dictionary is part of declarative memory, and that Woodcock, R. W., McGrew, K. S., & Mather, N. (2001). grammatical rules are processed by the procedural sys- Woodcock-Johnson Tests of Achievement. Itasca, IL: tem. Journal of Cognitive Neuroscience, 9, 266–276. Riverside Publishing.