11
Predicting school achievement from cognitive and non-cognitive variables in a Chinese sample of elementary school children Liping Lu a,b , Heike S. Weber c , Frank M. Spinath c , Jiannong Shi a,b, a The Institute of Psychology, Chinese Academy of Sciences, China b Graduate University of Chinese Academy of Sciences, GUCAS, China c Department of Psychology, Saarland University, Germany article info abstract Article history: Received 14 August 2010 Received in revised form 9 December 2010 Accepted 10 February 2011 Available online 17 March 2011 The present study had two aims: First, to investigate the joint and specific roles of working memory (WM) and intelligence as predictors of school achievement. And second, to replicate and extend earlier findings (Spinath, Spinath, Harlaar, & Plomin, 2006) on the incremental validity of non-cognitive over cognitive abilities in the prediction of school achievement. The present sample consisted of N = 179 Chinese primary school children in the fourth grade. All measures including working memory (WM), intelligence and motivational items were assessed in class. Teachers provided test scores for the domains of Chinese and Math. We found that WM was a good predictor of school achievement and comparable in predictive power to intelligence. Together, cognitive ability including both WM and intelligence explained 17.8% and 36.4% of the variance in children's Chinese and Math scores, respectively. The relative importance of WM and intelligence varied with school domains with greater predictive power of WM for Math while intelligence explained a greater proportion of the variance in Chinese although the magnitude of this difference was only moderate. Domain-specic motivational constructs contributed only marginally to the prediction of school achievement for both Chinese and Math. Crown Copyright © 2011 Published by Elsevier Inc. All rights reserved. Keywords: School achievement Intelligence Working memory Self-perceived ability Intrinsic values Earlier research published in this journal indicated that non-cognitive variables can have predictive power on school achievement beyond the inuence of general intelligence (Spinath, Spinath, Harlaar, & Plomin, 2006). In that study, children's self-reports on domain-specic self-perceived ability (SPA) and intrinsic values (IV) from Eccles' expectan- cyvalue theory of motivation (Eccles (Parsons) et al., 1983; Wigeld & Eccles, 2000) explained a signicant proportion of the variance in teacher-rated school achievement unaccount- ed for by intelligence. It was one of the rst studies to emphasize the joint and specic effects of cognitive and non- cognitive variables on scholastic achievement in elementary school. However, that study was limited in several ways: First of all, it assessed intelligence with four short verbal and nonverbal tests adapted from the Wechsler Intelligence Scale for Children, third edition (WISC-III-PI; Kaplan, Fein, Kramer, Delis, & Morris, 1999), and the Cognitive Abilities Test 3 (CAT3;Smith, Fernandes, & Strand, 2001) modied for parental administration. The utilization of data from parent- administered intelligence tests is not uncommon in twin research (e.g., Oliver & Plomin, 2007), since participants live far apart and in-person testing is typically not feasible. However, in-person testing carried out by trained personnel is preferable in order to minimize variance due to differential administration. Second, Spinath et al. (2006) used teacher ratings of school achievement rather than achievement test scores. It has been argued that non-cognitive variables might contribute to the prediction of achievement over cognitive variables especially when achievement is assessed by teachers compared to more objective measures (Hansford & Hattie, 1982; Helmke, 1992). For example, Schicke and Fagan (1994) showed that pupils' academic self-concepts explained Intelligence 39 (2011) 130140 Corresponding author at: The Institute of Psychology, Chinese Academy of Sciences, China. E-mail addresses: [email protected], [email protected] (J. Shi). 0160-2896/$ see front matter. Crown Copyright © 2011 Published by Elsevier Inc. All rights reserved. doi:10.1016/j.intell.2011.02.002 Contents lists available at ScienceDirect Intelligence

Predicting school achievement from cognitive and non-cognitive variables in a Chinese sample of elementary school children

Embed Size (px)

Citation preview

Intelligence 39 (2011) 130–140

Contents lists available at ScienceDirect

Intelligence

Predicting school achievement from cognitive and non-cognitive variablesin a Chinese sample of elementary school children

Liping Lu a,b, Heike S. Weber c, Frank M. Spinath c, Jiannong Shi a,b,⁎a The Institute of Psychology, Chinese Academy of Sciences, Chinab Graduate University of Chinese Academy of Sciences, GUCAS, Chinac Department of Psychology, Saarland University, Germany

a r t i c l e i n f o

⁎ Corresponding author at: The Institute of Psycholof Sciences, China.

E-mail addresses: [email protected], luli_1119@12

0160-2896/$ – see front matter. Crown Copyright ©doi:10.1016/j.intell.2011.02.002

a b s t r a c t

Article history:Received 14 August 2010Received in revised form 9 December 2010Accepted 10 February 2011Available online 17 March 2011

The present study had two aims: First, to investigate the joint and specific roles of workingmemory (WM) and intelligence as predictors of school achievement. And second, to replicateand extend earlier findings (Spinath, Spinath, Harlaar, & Plomin, 2006) on the incrementalvalidity of non-cognitive over cognitive abilities in the prediction of school achievement. Thepresent sample consisted of N=179 Chinese primary school children in the fourth grade. Allmeasures including working memory (WM), intelligence and motivational items wereassessed in class. Teachers provided test scores for the domains of Chinese and Math. Wefound that WM was a good predictor of school achievement and comparable in predictivepower to intelligence. Together, cognitive ability including bothWM and intelligence explained17.8% and 36.4% of the variance in children's Chinese andMath scores, respectively. The relativeimportance of WM and intelligence varied with school domains with greater predictive powerof WM for Math while intelligence explained a greater proportion of the variance in Chinesealthough the magnitude of this difference was only moderate. Domain-specific motivationalconstructs contributed only marginally to the prediction of school achievement for bothChinese and Math.

Crown Copyright © 2011 Published by Elsevier Inc. All rights reserved.

Keywords:School achievementIntelligenceWorking memorySelf-perceived abilityIntrinsic values

Earlier research published in this journal indicated thatnon-cognitive variables can have predictive power on schoolachievement beyond the influence of general intelligence(Spinath, Spinath, Harlaar, & Plomin, 2006). In that study,children's self-reports on domain-specific self-perceivedability (SPA) and intrinsic values (IV) from Eccles' expectan-cy–value theory of motivation (Eccles (Parsons) et al., 1983;Wigfield & Eccles, 2000) explained a significant proportion ofthe variance in teacher-rated school achievement unaccount-ed for by intelligence. It was one of the first studies toemphasize the joint and specific effects of cognitive and non-cognitive variables on scholastic achievement in elementaryschool. However, that study was limited in several ways: Firstof all, it assessed intelligence with four short verbal and

ogy, Chinese Academy

6.com (J. Shi).

2011 Published by Els

evier In

nonverbal tests adapted from the Wechsler Intelligence Scalefor Children, third edition (WISC-III-PI; Kaplan, Fein, Kramer,Delis, & Morris, 1999), and the Cognitive Abilities Test 3(CAT3;Smith, Fernandes, & Strand, 2001) modified forparental administration. The utilization of data from parent-administered intelligence tests is not uncommon in twinresearch (e.g., Oliver & Plomin, 2007), since participants livefar apart and in-person testing is typically not feasible.However, in-person testing carried out by trained personnelis preferable in order to minimize variance due to differentialadministration. Second, Spinath et al. (2006) used teacherratings of school achievement rather than achievement testscores. It has been argued that non-cognitive variables mightcontribute to the prediction of achievement over cognitivevariables especially when achievement is assessed byteachers compared to more objective measures (Hansford &Hattie, 1982; Helmke, 1992). For example, Schicke and Fagan(1994) showed that pupils' academic self-concepts explained

c. All rights reserved.

131L. Lu et al. / Intelligence 39 (2011) 130–140

a greater proportion of variance in grades than in achieve-ment test scores after intelligence was controlled. In a similarvein, Gottfried (1990) found that intrinsic values incremen-tally predicted teacher-assessed achievement rather thantest-based achievement.

The sampling frame in the study by Spinath et al. (2006)was the Twins Early Development Study (TEDS; Oliver &Plomin, 2007), investigating children living in a Westernsociety whose primary language spoken at homewas English.From comparative intercultural studies, however, numerousfindings have been reported suggesting systematic differ-ences between Asian andWestern students from preschool tocollege. Some researchers found that Asian countries havebetter school practice and pedagogy (Kobayashi, 1994; Lewis,1995; Stevenson & Stigler, 1992). Furthermore, it has beendemonstrated that Asian children, especially Chinese chil-dren, display higher achievement motivation than theirWestern peers and believe in learning through effort ratherthan fixed ability (Dweck, 1999; Tweed & Lehman, 2002)while others reported that Asian parents have higherexpectations to their children and emphasize more onpromoting their children's learning (Au & Harackiewicz,1986; Li, 2003; Stevenson & Stigler, 1992; Yao, 1985). Forexample, Chao (1994), who also studied elementary schoolchildren to investigate cultural differences affecting children'sschool performance, found that Chinese parents emphasizedhard effort and discipline at an early age, whereas Americanparents focused more on the role of ability. Since thesecultural differences can affect bothmean level differences andcan also influence the relative importance of variablespredicting school achievement, we aimed at investigatingthe explanatory power of cognitive and non-cognitivevariables in a Chinese sample and extend the study designby adding a second cognitive construct, namely workingmemory (WM).

1. Working memory as an important predictor of schoolachievement

Working memory (WM) constitutes “mechanisms orprocesses that are involved in the control, regulation, andactive maintenance of task-relevant information in theservice of complex cognition” (Miyake & Shah, 1999, p.450). It is considered to be a relatively pure measure ofchildren's learning potential without being overly saturatedwith their prior experiences such as preschool education, orsocioeconomic background (Alloway, Gathercole, Willis, &Adams, 2004; Weismer et al., 2000).

There is strong empirical evidence that WM is substan-tially related to learning abilities and school achievement(Alloway et al., 2005; Gathercole, Brown, & Pickering, 2003;Gathercole & Pickering, 2000; Gathercole, Pickering, Knight, &Stegmann, 2004). For example, Daneman and Carpenter(1980) reported a correlation of r=.59 between WM testsand verbal SAT results. Gathercole and Pickering (2000)identified seven-year-old students with poor scholasticachievement by assessing their WM capacity (see alsoGathercole et al., 2003). St. Clair-Thompson and Gathercole(2006) argued that WM “plays a causal role in children'sdeveloping skills and knowledge, particularly in the domainof literacy” (p. 755).While Gathercole et al. (2004) found that

intellectual operations required in Math and Science areconstrained by the limited capacity of working memoryacross the childhood years. As a construct fundamentallyimportant to higher mental processes, WM involves theoperations and nature of higher mental processes such asdiscrimination, attention and intelligence (Peterson, 1925).

Researchers studying the relation between workingmemory and intelligence have typically reported substantialcorrelations (Ackerman, Beier, & Boyle, 2002, 2005; Colom,Rebollo, Palacios, Juan-Espinosa, & Kyllonen, 2004; Colom,Abad, Rebollo, & Shih, 2005; Colom & Shih, 2004; Conway,Cowan, Bunting, Therriault, & Minkoff, 2002; Kane et al.,2004; Kyllonen & Christal, 1990). As intelligence has a long-standing tradition as a predictor of school achievement, someresearchers have suggested that the key factor underlying therelationship between WM and school achievement might beintelligence (Alloway et al., 2004; Gathercole, Alloway, Willis,& Adams, 2006). In order to identify common and specificvariance components in school achievement associated withintelligence and WM, both have to be studied in the samesample. There is, however, a surprising lack of studies thathave actually done this. We found only four empirical studiesthat reported a joint assessment of intelligence and WM,three of which focused only on learning disabilities (e. g.,Alloway, 2009; Swanson, Jerman, & Zheng, 2008; Maehler &Schuchardt, 2009). The single remaining study used anundergraduate student sample and did not employ a standardWM model (Krumm, Ziegler, & Buehner, 2008). In this study,reasoning appeared to be the best predictor of schoolachievement. In addition, common variance shared by WMand reasoning as well as variance specific to WM alsosignificantly accounted for individual differences in languagecourses beyond reasoning.

2. Predicting school achievement by WM and intelligence

There are two lines of reasoning why intelligence andWMare studied together so rarely. The first is that it appears to bedifficult to find an appropriate level of comparison betweengeneral intelligence (g, or general fluid intelligence, Gf) andWM. WM tasks typically address circumscribed basic pro-cesses (Luo, Thompson, & Detterman, 2006) whereas g isconsidered to be conceptually opaque and defined as thecommon variance in a set of cognitive abilities (Oberauer,Schulze, Wilhelm, & Suess, 2005). The second explanation isthat some cognitive theorists regard working memory and gas isomorphic (Jensen, 1998). Kyllonen and Christal (1990),for example, argued that WM capacity and intelligence werelargely the same. Stauffer, Ree, and Carreta (1996) reported acorrelation of .99 between two latent factors representing gand WM. However, there is growing consensus that WMcannot simply be equated with intelligence, despite substan-tial correlations between the two. Furthermore, numerousstudies have reported much lower associations between WMand intelligence. For example, Engle, Tuholski, Laughlin, andConway (1999) reported correlations between WM tests andCattell's Culture Fair Test (Cattell, 1973) ranging from .24 to.29. Extending this study, Conway et al. (2002) found similarmoderate correlations from .28 to .37 between WM andintelligence. The estimated correlation between WM and g ina latent structural equation model was .60. In a review on the

1 We did not include “7” in our digit sorting task because the study wasdesigned to ultimately allow comparisons with assessments in a Germansample, and 7 is the only digit with two syllables in the German languagewhen all other digits are only one-syllable words both in German andChinese.

132 L. Lu et al. / Intelligence 39 (2011) 130–140

relation between working memory capacity and g, Conway,Kane, and Engle (2003) summarized that WM accounted forat least one third of the variance in g, which means despitebeing highly associated, g and WM are not identical. In asimilar vein, Ackerman et al. (2005) found that the averagecorrelation between WM and g was far from unity, lendingsupport to the notion that WM and intelligence are differentconstructs (see also Cain, Oakhill, & Bryant, 2004; Gathercoleet al., 2006) and should be assessed separately wheninvestigating the importance of cognitive variables in theprediction of school achievement.

Regarding WM, our objectives in this study are thefollowing: First, to investigate the explanatory power ofWM and compare it to intelligence. Second, to assess theextent to which WM and intelligence predict school achieve-ment through unique and common proportions of variance.And finally, to study the incremental validity of non-cognitivevariables after variance explained by WM and intelligencewas accounted for.

3. The contribution of motivation to school achievement

The study by Spinath et al. (2006) introduced the rationalefor our interest in studying cognitive and non-cognitivevariables in a joint design.

First, cognitive variables, albeit powerful predictors, leave asubstantial, yet reliable portion of the variance in scholasticachievement unaccounted for. Second, one primary interest inthe educational sciences concerns the potential improvementof learning and achievement outcomes through educators'efforts and intervention. Typically, motivational constructs areconsidered to be more malleable than cognitive variables. Inthe realm of theories and constructs that have been proposedto explain variation in students' commitment to academictasks, one long-standing perspective on motivation is expec-tancy–value theory. Its representatives argue that “individuals'choice, persistence, and performance can be explained by theirbeliefs about how well they will do on the activity and theextent to which they value the activity” (Wigfield & Eccles,2000, p. 68). As this is one of the best elaborated motivationaltheories and since constructs from this theoretical approachwere tested in the earlier study (Spinath et al., 2006), wechoose the expectancy–value model as our theoretical frame-work for the present study again.

In the light of this theoretical background, our hypothesesin this study were first, thatWM and intelligence predict bothshared and specific proportions of variance in schoolachievement, and second, that motivation contributes incre-mentally to this prediction, beyond cognitive abilities.

4. Method

4.1. Participants

The children in this study were recruited from one of thelargest primary schools in Beijing. Both native Beijingchildren and also children from other areas whose parentsworked in Beijing attend this school. All 203 children from thefourth grade in this school were invited to participate withparents' consent. The data collection was divided into threeparts: In the first session lasting 30 min, intelligence was

assessed. Within a week, the children completed a question-naire set including non-cognitive variables in a secondsession that lasted 45 min. A week later, the workingmemorydata was collected in a third session lasting 45 min. Teachersprovided test scores of 193 children. The scores of 10 childrenwere missing for uncontrollable reasons (e. g., if a child wastransferred to a different school). The final sample in theanalysis consisted of 171 children with complete data.

5. Instruments and procedures

5.1. Cognitive measures: working memory

To assess the components of Baddeley's working memorymodel, (Baddeley, Emslie, Kolodny, & Duncan, 1998), weselected three different tasks called Digit Sorting, AnimalSorting and Backward Matrix Span. The assessments with thematerial developed for group testing were pilot tested twicewith 15 fourth graders each. Procedural details such asanswering time were decided upon the pilots. Since childrenwere tested in the classroom, the tasks were presentedthrough a laptop with attached speakers and a projector. Ineach experiment, not more than 35 children were allowed toattend simultaneously to assure that sight and sound wereclear to all participants.

1. Digit Sorting. In this task, prerecorded lists comprisingdifferent digits ranging from 1 to 12 (except 71) werepresented at the rate of one digit per second. The taskrequired the children to remember the presented digits ineach trial and to reproduce them in ascending order. Forreproduction, answer sheets were provided. The taskincluded a total of 12 trials. Set size was mixed with arange from 2 to 6 stimuli. Depending on the set size, thechildren had up to 15 s to provide answers. Digits wererandomly grouped with the restriction that no digit wasincluded twice within the same trial. Across trials, all digitsappeared with roughly the same frequency. The totalnumber of correct reproductions was used as the partici-pant's Digit Sorting score.Before testing, example files were played to familiarize thechildren with the audio files. To make sure that each childunderstood the instruction, three practice trials wereperformed afterwards which were explained by theexperimenter.

2. Animal Sorting. In this task, prerecorded lists of animalnameswere presented via speakers at the rate of one animalname per second (e.g. dog and mouse). The task requiredthe children to remember the presented animals in eachtrial and to reproduce them sorted according to their bodysize starting with the smallest animal in the list. Forreproduction, answer sheets were provided. A total of 10trials were presented. Set sizes were mixed with a rangefrom 2 to 6 (out of 11) stimuli. Depending on the set size,children had up to 30 s to write down their answers.Animals were randomly grouped into trials with the

Table 1Means (M), standard deviations (SD), and inter-correlations of all measured variables.

Test items N M SD Non-cognitive abilities Cognitive abilities School achievement

SPA_m IV_c IV_m CFT WM Chinese Math

SPA_c 191 3.99 0.75 .49 ⁎⁎ .59 ⁎⁎ .38 ⁎⁎ .03 .06 .10 .08SPA_m 191 4.17 0.73 – .32 ⁎⁎ .69 ⁎⁎ .11 .11 −.01 .23 ⁎⁎

IV_c 191 4.07 0.73 – .46 ⁎⁎ .03 .04 −.01 .03IV_m 191 4.14 0.91 – .07 .18 ⁎⁎ −.10 ⁎ .19 ⁎⁎

CFT 188 27.05 6.07 – .36 ⁎⁎ .41 ⁎⁎ .49 ⁎⁎

WM 191 .00 1.00 – .13 .55 ⁎⁎

Chinese 193 88.40 9.20 – .62 ⁎⁎

Math 193 85.74 12.11 –

Note: SPA _c: children's report about their Chinese self-perceived ability; SPA _m: children's report about their Mathematics self-perceived ability; IV _c: children'sreport about their Chinese intrinsic values; IV _m: children's report about their Mathematics intrinsic values; CFT: CFT test scores; WM: factor score of workingmemory; Chinese: mean scores of two examinations; and Math: mean scores of two examinations. Minor differences in N result from missing values for childrenwho did not attend all three measurements.⁎⁎ pb .01.⁎ pb .05 (2-tailed).

133L. Lu et al. / Intelligence 39 (2011) 130–140

restrictions that no animal was included twice within thesame trial. Carewas taken that the animals in any given trialdiffered in size enough so that only one order was correct.Across trials, all animal names appeared with roughly thesame frequency. The total number of correct reproductionswas used as the participant's Animal Sorting score.Before the testing, pictures of all included animals werepresented to the class and the children had to name eachanimal to ensure that they were familiar with all includedstimuli. To make sure that each child understood theinstructions, the testing was also preceded by three practicetrials.

3. Backward Matrix Span. In this task, a 3⁎3 square matrixwas projected to a screen in front of the class. Eight out ofnine cells in this matrix were white and one was red. At arate of 2 s, a different cell turned red and the former redcell became white. The task required the children toremember the sequence of the red cells within the matrixin each trial. Upon the projection of three black questionmarks on the screen, the children were asked to reproducethe sequence in backward order. To do so, an answer sheetwas provided which included one 3⁎3 matrix for eachtrial. Using digits, the children marked the backwardsequence starting with the last red cell in trial andproceeding with the last but one red cell in the trial, andso on. The task included a total of 10 trials. Set sizes weremixed with a range from 2 to 6 stimuli. Depending on theset size, the children were allowed up to 11 s to provideanswers. The sequences of red cells were random with therestriction that no position was included twice within thesame trial and that no trial contained only adjacent cells.Across trials, all cells turned red with roughly the samefrequency. The total number of correct trials was used asthe participant's score.

Before testing, three practice trials were performed andexplained by the experimenter.

5.2. Cognitive measures: intelligence

Children were tested with a short version of Cattell'sCulture Fair Test (Cattell, 1973), a well-established measure

of Cattell's Gf–Gc Model with excellent predictive validity forschool achievement (Williams, McCallum, & Reed, 1996).

The CFT short version consists of four separate and timedpaper-and-pencil subscales with 8 to 12 items each and atotal of 46 items. All subscales only include non-verbalmaterial. For each subscale, practice items were administeredwhich were explained by the experimenter to make sure thateach child understood the task. Children were allowed 2.5 to4 min for each subtest. When time expired for a subtest,children were instructed to move on to begin the nextsubtests.

5.3. Non-cognitive measures

5.3.1. Self-perceived ability (SPA)Children's self-perceived ability was assessed by means of

three items for Chinese and three items for Math (Eccles(Parsons) et al., 1983). The items were closely related totypical curricular content for Chinese and Math, such asreading for Chinese and calculation for Math. Children wererequired to indicate on a 5-point Likert scale how good theythought they were in these essential activities. These shortscales show acceptable reliabilities (average Cronbach'salpha=.69).

5.3.2. Children's intrinsic values (IV)Children's intrinsic values were also assessed with three

items per domain, reflecting the same content as the SPAitems. Children were asked to indicate on a 5-point Likertscale how much they liked these activities. The averagereliability of children's intrinsic values was also acceptable(average Cronbach's alpha=.71).

5.4. School achievement

Children's school achievement was based on the averagetest scores of their midterm and endterm examinations forChinese andMath. Thesemidterm and endterm examinationsare the two most important tests for school children in China.Although children in the same grade in China work with thesame textbooks, they don't have uniform tests. So during thedays before midterm and endterm examinations, all teachersof the same subject in the same grade of a primary school

Table 2Regression analyses of domain-specific school achievement on cognitive abilities.

Beta T p R R2 ΔR2 ΔF(df) Δp

ChineseModel 1 CFT .374 5.352 .000 .374 .140 .140 28.649 (1, 176) .000Model 2 WM .338 4.757 .000 .338 .114 .114 22.631 (1, 176) .000Model 3 CFT .290 3.975 .000 .433 .187 .047 20.165 (2, 175) .000

WM .233 3.192 .002Math

Model 1 CFT .463 6.957 .000 .463 .215 .215 48.402 (1, 177) .000Model 2 WM .527 8.255 .000 .527 .278 .278 68.137 (1, 177) .000Model 3 CFT .314 4.879 .000 .603 .364 .149 50.360 (2, 176) .000

WM .414 6.426 .000

Note: CFT: CFT test scores; WM: factor score of working memory; Chinese: mean scores of two examinations; and Math: mean scores of two examinations. In oursample, there were 178 students with both Chinese test scores and cognitive abilities (bothWM and intelligence) and 179 students with bothMath test scores andcognitive abilities.

Table 3Commonality analyses of WM and intelligence predicting schoolachievement.

R2 (Chinese)=.225R2 (Math)=.416

Explainedvariancefor Chinese

Explainedpercent forChinese (%)

Explainedvariancefor Math

Explainedpercent forMath (%)

WM .051 22.7 .138 33.2CFT .092 40.8 .124 29.8Common variancefor Chinese

.082 36.4

Common variancefor Math

.154 37.0

Note: CFT: CFT test scores; WM: factor score of working memory; Chinese:mean scores of twoexaminations; andMath:mean scores of two examinations.

134 L. Lu et al. / Intelligence 39 (2011) 130–140

(normally 4–6 teachers) discuss and prepare the test itemsfor midterm examination according to what their studentslearned during the first half of the year. In preparation of theendterm examination, teachers need to consider whatchildren learned throughout the full year. The correlationbetween the midterm and endterm scores was r=.47 forChinese and r=.43 for Math, and we aggregated the twoscores within domains to reflect children's schoolachievement.

6. Results

Means, standard deviations, and zero order inter-correla-tions of our measures are depicted in Table 1.

The three WM tasks showed a positive manifold (averagecorrelation r=.32) and a principal component analysis wasperformed. This yielded a unifactorial solution explaining54.8% of the variance among WM scores. We used this factorscore (Table 1) in all subsequent regression and commonalityanalyses as our WMmeasure. Table 1 shows, intelligence andWM showed a significant positive association of r=.36,Domain-specific SPA and IV showed significant and substan-tial overlap of r=.59 for Chinese and r=.69 for Math. Acrossdomains SPA and IV correlated to a somewhat lesser extent(r=.49 for SPA and .38 for IV). Interestingly, SPA and IV werenot significantly related to intelligence. Significant associa-tions between non-cognitive and cognitive variables werefound only for IV Math and WM (r=.18 with the WM factorscore). Significant moderate to strong correlations werefound between cognitive variables and school achievementtest-scores, with a tendency for higher associations for Mathcompared to Chinese. The largest correlation with the Mathtest-score of r=.55 was found for the WM factor score, whilefor the Chinese test-score the largest correlation was foundwith intelligence (r=.41). Significant but small correlationsbetween test-scores and non-cognitive variables were foundonly for Math (r=.23 for SPA and r=.19 for IV).

To investigate the contributions of intelligence andWM tothe prediction of domain-specific school achievement, re-gression analyses were performed.

Table 2 includes the results of three regression analysesper domain. Models 1 and 2 used intelligence and WM as thesole predictors of school achievement, whereas Model 3 usedboth intelligence and WM. Intelligence alone explained 14%of the variance in school achievement for Chinese and 21.5%

for Math. Working memory alone explained 11.4% of thevariance in Chinese and 27.8% in Math. The multipleregression analysis (Model 3) indicated that intelligenceand WM together explained a significantly greater portion ofthe variance in school achievement than each cognitivepredictor alone (R2=.187 for Chinese and .364 for Math).

Results of a commonality analysis (Cooley & Lohnes, 1976)are displayed in Table 3.

Commonality analysis decomposes the explained variancein multiple regression analyses with two predictors into threecomponents: 1) explained variance specific to the firstpredictor, 2) explained variance specific to the secondpredictor, and 3) variance explained commonly by the twopredictors. As Table 3 indicates, a substantial proportion of theexplained variance in school achievement was shared betweenintelligence and WM (35.3% for Chinese and 35.6% for Math).For Chinese, intelligence explained the largest proportion ofthe variance (30.0%) whereas for Math the largest proportionof the variance was explained by WM (40.8%).

A final series of regression analyses was performedincluding SPA and IV into the prediction of domain-specificschool achievement along with cognitive abilities. Twomodels were compared in the domain-specific analyses:Model 1 included only cognitive variables whereas Model 2added SPA and IV.

As presented in Table 4, motivational variables predictedonly a small portion of the variance in school achievementbeyond cognitive variables. Taking SPA and IV together, theincremental variance explained was only 1.9% in Chinese and

C F T

E 2

C F T 2

E 3

C F T 3

E 4

C F T 4

W M

E 5 W M _ D S

E 6 W M _ M S

E 7 W M _ A S

S P A

E 8 S P A 1

E 9 S P A 2

E 1 0 S P A 3

I V

E 1 1 I V 1

E 1 2 I V 2

E 1 3 I V 3

E 1

C F T 1

S A

E 1 4t e s t 1

t e s t 2 E 1 5

E 1 6

Fig. 1. Cognitive and motivational predictors of school achievement: full latent variable model. CFT: intelligence; CFT1 to CFT4: CFT subtests; WM: workingmemory;WM_DS: Digit Sorting;WM_MS: BackwardMatrix Span;WM_AS: Animal Sorting; SPA: self-perceived ability; SPA1 to SPA3: children's report about theirdomain-specific self-perceived ability; IV: intrinsic value; IV1 to IV3: children's report about their domain-specific intrinsic value; SA: school achievement; test1:domain-specific midterm test score; test2: domain-specific endterm test score; and E1 to E16: error terms.

135L. Lu et al. / Intelligence 39 (2011) 130–140

2.2% in Math. At the level of single predictors, only SPAChinese yielded a significant regression weight.

Our main reason for performing regression analyses wasto provide results directly comparable to earlier studies onthe predictability of school achievement through cognitiveand non-cognitive variables. To overcome inherent limita-tions of this regression-based approach, however, additionalstructural equation modeling (SEM) was applied for domainspecific analyses with four latent predictors (WM, intelli-gence, SPA, and IV) and one latent dependent variable

Table 4Regression analyses of domain-specific school achievement on cognitive abilities, c

Beta T p

ChineseModel 1 CFT .266 3.539 .001

WM .252 3.352 .001Model 2 CFT .258 3.446 .001

WM .249 3.332 .001SPA_c .172 1.990 .048IV_c −.108 −1.252 .212

MathModel 1 CFT .312 4.732 .000

WM .418 6.329 .000Model 2 CFT .305 4.664 .000

WM .402 6.111 .000SPA_m .106 1.195 .234IV_m .055 .613 .541

Note: SPA _c: children's report about their Chinese self-perceived ability; SPA _m: chireport about their Chinese intrinsic values; IV _m: children's report about their Matmemory; Chinese: mean scores of two examinations; and Math: mean scores of two171 students for Chinese and 170 students for Math. In this table, Model 1 is conceptnumbers differ slightly.

(achievement). In the full model, no restrictions were appliedwith the exception that correlations between the error termsof corresponding IV and SPA itemswere allowed. See Fig. 1 fora depiction of the full model.

Results from the full model were compared to moreparsimonious models by successively fixing non-significantpaths to zero. The overall model fit was evaluated by the rootmean square error of approximation (RMSEAb .05 indicate agood and RMSEAb .08 indicate an acceptable fit) and the χ2

test. The significance of model parameters was evaluated by a

hildren's self-perceived ability, and intrinsic value.

R R2 ΔR2 ΔF(df) Δp

.428 .183 .1831 8.861 (2, 168) .000

.450 .202 .019 1.983 (2, 166) .141

.604 .365 .365 47.972 (2, 167) .000

.622 .387 .022 3.014 (2, 165) .052

ldren's report about their Mathematics self-perceived ability; IV _c: children'shematics intrinsic values; CFT: CFT test scores; WM: factor score of workingexaminations. With missing data in motivation items, our sample reduced toually the same as Model 3 in Table 2, but due to missing motivation scales, the

136 L. Lu et al. / Intelligence 39 (2011) 130–140

maximum likelihood based 95% confidence interval (CI) andthe χ2 difference test in which nested models with moreparameters were compared to models with fewer para-meters. SEM analyses were based on a slightly smaller sample(N=150) than the regression analyses due to the fact that fortwo classes separate subtest results for the CFT were notavailable. SEM analyses were carried out using AMOS 7.0.Variance–covariance matrices for SEM analyses were esti-mated by using expectation maximization procedures forhandling missing data (Little & Rubin, 2002).

For both domains, the full model provided a good fit to thedata (Chinese: χ2 (77)=102.86, p=.03; RMSEA=.047;Math: χ2 (78)=104.95, p=.02; RMSEA=.048). The mostparsimonious models that did not fit significantly worse thanthe full model are depicted in Figs. 2 (Chinese) and 3 (Math).For both subjects, the χ2 difference test (likelihood-ratio test,LRT) indicated no significant differences in fit between fulland reduced models (Chinese: LRT=4.87, df=6, p=.56,Math: LRT=8.99, df=6, p=.17).

From Figs. 2 and 3, it is apparent that across domainsfixing the paths from SPA and IV to the respective dependentvariable as well as their correlation with the cognitivevariables to zero did not result in a significant deteriorationof fit. The findings shown in Fig. 2 are consistent with ourregression results in that the strongest direct effect onChinese school achievement was observable for intelligence(path, or β=.43), but that WM also had a significant effect(β=.29). Taken together, the latent cognitive variablesexplained 40% of latent Chinese achievement. The findingsfor Math depicted in Fig. 3 also mirror our regression resultsin that the strongest direct effect on school achievement camefrom WM (path, or β=.59), but that intelligence also

C F T

E 2

C F T 2

E 3

C F T 3

E

C F T

W M

E 5 W M _ D S

E 6 W M _ M S

E 7 W M _ A S

S P A c

E 8 S P A 1 c

E 9 S P A 2 c

E 1 0 S P A 3 c

I V c

E 1 1 I V 1 c

E 1 2 I V 2 c

E 1 3 I V 3 c

E 1

C F T 1

. 3 2 . 4 2 . 4 6 . 1 8

. 5 7 . 6 5 . 6 8 . 4 3

. 6 0

. 2 7

. 3 2

. 7 7

. 5 2

. 5 7

. 4 0

. 2 7

. 1 7

. 2 2

. 5 6

. 2 5

. 6 5

. 4 0

. 4 4

. 5 2

. 6 4

. 4 2

. 4 6

. 7 5

. 5 0

. 5 2

. 0 0

. 5 8

. 0 0

. 0 0

. 0 0

.

. 2 9

. 0 0

Fig. 2. Reduced latent variable model for the prediction of domain-specific schoolworking memory; WM_DS: Digit Sorting; WM_MS: Backward Matrix Span; WM_children's report about their Chinese self-perceived ability; IVc: Chinese intrinsic vaschool achievement for Chinese; test1c: Chinese midterm test score; test2c: Chines

contributed substantially to the prediction, while intelligencealso had a powerful effect (β=.37). Taken together, latentcognitive variables explained 70% of latent Mathachievement.

7. Discussion

The twomain purposes of this paperwere to investigate thejoint and specific roles of working memory and intelligence aspredictors of school achievement in Chinese elementary schoolchildren and whether non-cognitive variables (self-perceivedabilities and intrinsic values) contributed to school achieve-ment beyond cognitive abilities. Our main findings indicatethat across domains, a solid portion of the variance in schoolachievement (roughly one third of the explained variance) waspredicted jointly by intelligence and WM. The relativeimportance of specific predictor variance was to some degreedomain-specific in that intelligence added more to theprediction of Chinese test scores whereas WM added more tothe prediction of Math test scores. The amount of absolutevariance explained in Math was substantially larger (36.4%)than in Chinese (18.7%). In our study, the incremental validityof motivational constructs was negligible. These results wereconsistent across different methods of analyses, that is, usingmanifest variables in a regression approach and latent variablesin a SEM approach.

8. Cognitive predictors of achievement

Consistent with previous research that children's workingmemory skills are closely associated with academic progress inboth Reading (Swanson, Ashbaker, & Lee, 1996; Gathercole

4

4

S A c

E 1 4t e s t 1 c

t e s t 2 c E 1 5

E 1 6

4 3

. 0 0

. 4 0

. 7 8

. 9 8

. 6 0

. 9 6

achievement (Chinese). CFT: intelligence; CFT1 to CFT4: CFT subtests; WMAS: Animal Sorting; SPAc: Chinese self-perceived ability; SPA1c to SPA3clues; IV1c to IV3c: children's report about their Chinese intrinsic value; SAce endterm test score; and E1 to E16: error terms.

:::

C F T

E 2

C F T 2

E 3

C F T 3

E 4

C F T 4

W M

E 5 W M _ D S

E 6 W M _ M S

E 7 W M _ A S

S P A m

E 8 S P A 1 m

E 9 S P A 2 m

E 1 0 S P A 3 m

I V m

E 1 1 I V 1 m

E 1 2 I V 2 m

E 1 3 I V 3 m

E 1

C F T 1

S A m

E 1 4t e s t 1 m

t e s t 2 m E 1 5

E 1 6

. 3 3 . 4 2 . 4 4 . 2 0

. 5 8 . 6 4 . 6 6 . 4 4

. 6 2

. 2 6

. 3 1

. 7 9

. 6 1

. 5 6

. 8 7

. 2 6

. 8 0

. 2 5

. 8 6

. 8 3

. 4 5

. 1 0

. 1 4

. 5 1

. 9 3

. 8 9

. 5 0

. 9 3

. 9 1

. 5 2

. 0 0

. 7 4

. 0 0

. 0 0

. 0 0

. 3 7

. 5 9

. 0 0

. 0 0

. 7 0

. 8 2

. 8 8

. 6 7

. 7 7

Fig. 3. Reduced latent variable model for the prediction of domain-specific school achievement (Math). CFT: intelligence; CFT1 to CFT4: CFT subtests; WM:working memory; WM_DS: Digit Sorting; WM_MS: Backward Matrix Span; WM_AS: Animal Sorting; SPAm: Math self-perceived ability; SPA1m to SPA3m:children's report about their Math self-perceived ability; IVm: Math intrinsic values; IV1m to IV3m: children's report about their Math intrinsic value; SAm: schoolachievement for Math; test1m: Math midterm test score; test2m: Math endterm test score; and E1 to E16: error terms.

137L. Lu et al. / Intelligence 39 (2011) 130–140

et al., 2004) and Math (Swanson, 2006; Geary, Hoard, Byrd-Craven, & DeSoto, 2004), our study showed that workingmemory was an excellent predictor of children's schoolachievement both across domains and beyond intelligence.This is also in line with previous studies reporting that thespecific associations between WM and achievement remainwhen controlling for intelligence (Alloway, 2009; Swanson et al.,2008; Maehler & Schuchardt, 2009). Baddeley (1986) claimedthat a critical feature of working memory is simultaneousstorage and processing of information, an aspect characteristic ofmany learning activities, such as text comprehension, oldknowledge storage and new knowledge manipulation.

Concerning the relationship between WM and intelli-gence, the moderate inter-correlation of .36 is in line with theview that the two constructs are not isomorphic. Even in ourSEM analyses and using latent variables, the inter-correlationbetween WM and intelligence did not exceed .52. However,since we assessed a rather small number of WM andintelligence tasks instead of full test batteries, this findingshould not be overstated. A further look at the commonalityanalyses also illustrates that WM and intelligence share anotable portion of predictor variance in school achievement(e.g., Engle et al., 1999; Kyllonen & Christal, 1990). Theprocesses driving the covariation amongWMand intelligenceare not yet fully understood (e.g., Fry & Hale, 1996), althoughfindings from a recent study suggest that a large proportion ofthis covariance is accounted for by short-term storage(Colom, Abad, Quiroga, Shih, & Flores-Mendoza, 2008).

On a domain-specific level, differences between intelligenceand WM concerning their relative predictive validity emerged.Unique predictor variance ofWMexplained a larger proportionof achievement in Math whereas the opposite was true for

Chinese. The subject of Chinese places heavy demands onanalogical reasoning, which is correlated with general intelli-gence (Chiappe & MacDonald, 2005). Chinese children need tolearn the underlying logical relationships among words andamong sentences. In grade 4, for example, children are requiredto learn synonyms, antonyms, personification, comparison, andanalogy, and to contrast learned words and sentences. ForMath, on the other hand, greater emphasis is placed on theprocessing of number information, calculation, application ofarithmetic rules, and retrieval of arithmetic facts from longterm memory (Geary, Hoard, & Hamson, 1999; LeBlanc &Weber-Russell, 1996; Swanson & Sachse-Lee, 2001).

Another interesting finding concerns the strong differencein explained absolute variance betweenMath and Chinesewithalmost twice as much of the variance explained in the former(36.4%) compared to the latter (18.7%). Krumm et al. (2008)who studied German undergraduate students also reportedthat the amount of variance explained by cognitive predictorswas greater in science compared to language course grades.However, the absolute variance explained for Chinese achieve-ment in the present is particularly low. One possible explana-tion lies in content differences between Chinese and Westerntextbooks. According to a comparative study of Chinese andAmerican fourth grade reading textbooks, Chinese textbooksfocus to a substantial degree on values such as respectfulness,preservation of traditions, cooperation, leadership and econo-my,while readings inAmerican textbooks focusmore onvaluessuch as self-development and personal independence, and alsoput greater emphasis on the issue of systematic problemsolving in real life scenarios (Zheng, 1997). Another differencemight lie in a greater preoccupation of Chinese educators inlanguage classes with aesthetic versus analytic aspects of the

138 L. Lu et al. / Intelligence 39 (2011) 130–140

course content. For example, a comparison of different readingpreferences suggested that Chinese educatorswere particularlyinterested in evoking admiration for elegancy and style in theirstudents and inspiring them to read the articles with sentiment(Wang, 2002). It can be argued that a stronger focus on readingas a tool to expand knowledge, develop thinking styles, andencourage students to question content based on theirunderstanding, all of which might involve reasoning to agreater extent, might partially explain the differences in theamount of variance explained which we encountered in ourstudy.

9. The role of non-cognitive variables

The second area of interest in the present study concernedthe relation among cognitive and non-cognitive predictors andthe incremental validity of motivational variables. Unlikeearlier reports from Western samples (Spinath & Spinath,2005; who studied German children; Spinath et al., 2006 whostudied children in the UK), we found no significant correlationbetween cognitive abilities and children's reports on self-perceived ability and intrinsic values, although our motiva-tional measures showed no indication of variance restrictionand also have acceptable reliabilities. Obviously, the children'sanswers in our study reflect something different than theanswers of Western children. Having a closer look at Eccles'model, its emphasis on the parents' role as expectancysocializers becomes evident (Frome & Eccles, 1998). Aschildren's self-perceived ability and intrinsic motivation arenot only affected by previous achievement-related experience,but also by their parents' perceptions, expectations andattitudes towards their children (Bandura, 1997; Pintrich &Schunk, 2002; Sarason, Pierce, Bannerman, & Sarason, 1993), itis plausible that cultural specifics in parental attitudes mayhave influenced the children's self-perceptions. It is well-documented that Chinese parents, adopting Confucian doc-trines which emphasize single-minded effort, consider theirchildren's behavior and ability to be malleable and encouragechildren to compensate limited abilities with conscientious-ness and hard work (Tweed & Lehman, 2002). Previous studieshave shown that Chinese grade-school students tend toattribute academic success to effort (Hau & Salili, 1991),whereas American children were more likely to attributeacademic success to less controllable factors such as possessinginherent ability (Stevenson, Chen, & Lee, 1993; Stevenson &Stigler, 1992). Further evidence indicates that individuals whobelieved that effort led to success were also inclined to hold animplicit incremental theory that one can change importantaspects of the self such as one's ability to perform intellectualtasks (Dweck, Chiu, & Hong, 1995; Levy & Dweck, 1998). To theextent that the children in our study interpreted the ability-related question as a question about their willingness to showeffort, the lack of association to ability is less surprising.

Domain-specific motivation in our study did not showconvergent correlations with Chinese achievement, yet it didcorrelate moderately withMath. This could be, at least to somedegree, due to the unreliability in themeasurement of SPA andIV. But even after a double correction for attenuation, thedomain-specific correlations between non-cognitive variablesand achievement in the present study (average r=.19between SPA and achievement) were considerably lower

than the uncorrected correlations reported by Spinath et al.(2006) with an average r=.40 between SPA and achievement.The finding that a correlation emerged for Math but not forChinese ties in with the fact that both Chinese parents andchildren put more emphasis and value on Math rather thanReading (Huntsinger, Jose, Larson, Krieg, & Shaligram, 2000).Chinese parents also engage in fewer reading activities withtheir children and provide less books at home, two possiblyrelevant influences on children's reading achievement (e.g.,Adams, 1990; Ko & Chan, 2009). In addition, it has beendemonstrated that Chinese students practice more in the fieldof Math compared to American students (Stigler, Lee, &Stevenson, 1986), corroborating the greater concern of Chinesestudents with Math. The fact intrinsic values for Math werepositively correlated with Math and negatively correlated withChinese also fits into this picture. Chinese children study hardto meet the curricular requirements and high expectationsboth from their parents and teachers. They receive largeamounts of assignments for various subjects, are required toattend cram school after finishing their regular classes, andoften stay up until late at night to complete their homework.Given that children have to allocate limited time resourcesaccording to their priorities, it is comprehensible that childrenwith higher intrinsic values for Math will spend more time onstudying Math and invest less time for their language studies,resulting in lower Chinese test scores.

The fact that motivation in our study explained only verysmall incremental portions of variance in Chinese (1.9%) andMath (2.2%) beyond cognitive abilities was unexpected. UsingSEM, the paths from motivation to achievement did not evenreach significance. We see three possible reasons for thedifferent findings between this study and the study by Spinathet al. (2006) who reported incremental validities of 8% forEnglish and 9% for Math in a sample of elementary schoolchildren in the UK. The first reason may involve the differentcriteria chosen in the two studies. In the present study, wemeasured children's school achievement on the basis of testscoreswhile Spinath et al. (2006) chose teacher ratings instead.According to Voyer (1996), tests reflect scholastic skills in amore direct and accurate waywhile grades and teacher ratingsare amalgamates of both achievement as well as behavioralinformation which do not influence test scores to the sameextent. In otherwords, grades not onlymeasure knowledge butalso reflect behavior in the classroom. Studies which haveconducted systematic comparisons between school grades andobjective test scores have shown that both self-perceivedability and intrinsic values explainedmore in grades comparedto test scores (Gottfried, 1990; Schicke & Fagan, 1994).

A second reason for the greater incremental validity ofmotivation in the Spinath et al. (2006) study may involve theparent-administered testing. Because the sample in theearlier study consisted of twins, traditional face-to-faceintelligence testing conducted by experimenters was notfeasible. Instead, test sets were sent into the family andparents were instructed to administer them to their children.It is possible that this procedure introduced some degree ofunreliability in the intelligence measurement, thus increasingthe probability for non-cognitive constructs to account forgreater portions of variance.

The third reason follows up a point made earlier in thisdiscussion, namely the notion of a cultural difference

139L. Lu et al. / Intelligence 39 (2011) 130–140

between Chinese and Western samples regarding theiremphasis on ability versus effort which could have influencedchildren's ability self-perceptions. That is, even for thoseChinese children with lower performance, their parentsmight still encourage them to ignore poor ability, workharder and maybe try new learning methods to increase theirperformance. If this was the case, the traditional item contentused tomeasure Eccles expectancy–valuemodel might not befully suitable for use in the Chinese culture.

In the case of intrinsic values, problems might arise since inChinese culture values are conceived more of an asset of thewhole family, whileWestern culture emphasizes individualismand propagates self-development. In other words, a Chinesechild's achievement is viewed as a joint result of the wholefamily's effort rather than simply a result of the child's ownefforts (Chin, 1988). In addition, Chinese learners tend to focusmore on extrinsic motivation, such as practical outcomesbrought by education, compared toWestern learners. Previousstudies found that Chinese students regarded education as ameans to an end rather than self interest (e.g., Salili, 1996;Winter, 1996). Moreover, Chinese parents and teachersconsider a successful education more as a prerequisite for agood job and economic prosperity (Llewellyn, Hancock, Kirst, &Roeloffs, 1982) than seeking intrinsic values as a goal (Dewey,1916).

10. Limitations of the current study

The present study provides a first glance at the relativeimportance of cognitive and non-cognitive variables in theexplanation of elementary school achievement in Chinesechildren. Due to time constraints we had to test children inclass and adapt standard WM measures to be usable in aclassroom setting. We cannot rule out the possibility thatthese adaptations influenced the validity of our WMmeasures. The correlational pattern of WM and intelligencetested with a standard measure as well as its explanatorypower, however, can be regarded as indication that the testadaptations did not introduce pronounced bias.

To our knowledge this is the first time that traditionalexpectancy–value constructs were assessed in a Chinesesample of primary school children and used to explain variancein achievement test scores. We have offered possible explana-tions for the lack of predictive validity of these measures,however, we are well aware that these interpretations arepost-hoc and require further testing. The fact that domain-specificity in the predictability of achievement appears to playamuch greater role in Chinese samples is a novel findingwhichmight stimulate future intercultural research.

References

Ackerman, P. L., Beier, M. E., & Boyle, M. O. (2002). Individual differences inworking memory within a nomological network of cognitive andperceptual speed abilities. Journal of Experimental Psychology: General,131, 567−589.

Ackerman, P. L., Beier, M. E., & Boyle, M. O. (2005). Working memory andintelligence: The same or different constructs? Psychological Bulletin,131, 30−60.

Adams, M. J. (1990). Beginning to read: Thinking and learning About print.Cambridge, MA: MIT Press.

Alloway, T. P. (2009). Working memory, but not IQ, predict subsequentlearning in children with learning difficulties. European Journal ofPsychological Assessment, 25, 92−98.

Alloway, T. P., Gathercole, S. E., Adams, A. M., Willis, C., Eaglen, R., & Lamont,E. (2005). Workingmemory and phonological awareness as predictors ofprogress towards early learning goals at school entry. British Journal ofDevelopmental Psychology, 23, 417−426.

Alloway, T. P., Gathercole, S. E., Willis, C., & Adams, A. M. (2004). A structuralanalysis of working memory and related cognitive skills in earlychildhood. Journal of Experimental Child Psychology, 87, 85−106.

Au, T. K. F., & Harackiewicz, J. M. (1986). The effects of perceived parentalexpectations on Chinese children's mathematics performance. Merrill-Palmer Quarterly, 32, 383−392.

Baddeley, A. D. (1986).Working memory. New York: Oxford University Press.Baddeley, A. D., Emslie, H., Kolodny, J., & Duncan, J. (1998). Random

generation and the executive control of memory. The Quarterly Journal ofExperimental Psychology, 51A, 819−852.

Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W.H. Freeman.Cain, K., Oakhill, J., & Bryant, P. (2004). Children's reading comprehension

ability: Concurrent prediction by working memory, verbal ability, andcomponent skills. Journal of Educational Psychology, 96, 31−42.

Cattell, R. (1973).Measuring intelligence with the culture-fair tests: Manual forscales 2 and 3. Institute for Personality and Ability Testing, U.S.A.

Chao, R. K. (1994). Beyond parental control and authoritarian parentingstyle: Understanding Chinese parenting through the cultural notion oftraining. Child Development, 65, 1111−1119.

Chiappe, D., & MacDonald, K. B. (2005). The evolution of domain-generalmechanisms in intelligence and learning. The Journal of GeneralPsychology, 132, 5−40.

Chin, A. P. (1988). Children of China. New York: Alfred A. Knopf.Colom, R., Abad, F. J., Quiroga, M. A., Shih, P. C., & Flores-Mendoza, C. (2008).

Working memory and intelligence are highly related constructs, butwhy? Intelligence, 36, 584−606.

Colom, R., Abad, F. J., Rebollo, I., & Shih, P. C. (2005). Memory span and generalintelligence: A latent-variable approach. Intelligence, 33, 623−642.

Colom, R., Rebollo, I., Palacios, A., Juan-Espinosa, M., & Kyllonen, P. C. (2004).Working memory is (almost) perfectly predicted by g. Intelligence, 32,277−296.

Colom, R., & Shih, P. C. (2004). Is working memory fractionated onto differentcomponents of intelligence? Intelligence, 32, 431−444.

Conway, A. R. A., Cowan, N., Bunting, M., Therriault, D., &Minkoff, S. (2002). Alatent variable analysis of working memory capacity, short-termmemory capacity, processing speed, and general fluid intelligence.Intelligence, 30, 163−183.

Conway,A. R. A., Kane,M. J., & Engle, R.W. (2003).Workingmemory capacity andits relation to general intelligence. Trends in Cognitive Sciences, 7, 547−552.

Cooley, W. W., & Lohnes, P. R. (1976). Evaluation research in education. NewYork, NY: Irvington.

Daneman, M., & Carpenter, P. A. (1980). Individual differences in workingmemory and reading. Journal of Verbal Learning and Verbal Behavior, 19,450−466.

Dewey, J. (1916). Democracy and education: An introduction to the philosophyof education. New York: Macmillan.

Dweck, C. S. (1999). Self-theories. Philadelphia: Psychology Press.Dweck, C. S., Chiu, C., & Hong, Y. (1995). Implicit theories and their role in

judgments and reactions: A world from two perspectives. PsychologicalInquiry, 6, 267−285.

Eccles (Parsons), J., Adler, T. F., Futterman, R., Goff, S. B., Kaczala, C. M., Meece,J. L., et al. (1983). Expectancies, values, and academic behaviors. In J. T.Spence (Ed.), Achievement and achievement motives (pp. 75−146). SanFrancisco: Freeman.

Engle, R. W., Tuholski, S. W., Laughlin, J. E., & Conway, A. R. A. (1999).Working memory, short-term memory, and general fluid intelligence: Alatent-variable approach. Journal of Experimental Psychology: General,128, 309−331.

Frome, P. M., & Eccles, J. S. (1998). Parents' influence on children'sachievement-related perceptions. Journal of Personality and SocialPsychology, 74, 435−452.

Fry, A. F., & Hale, S. (1996). Processing speed, working memory, and fluidintelligence: Evidence for a developmental cascade. PsychologicalScience, 7, 237−241.

Gathercole, S. E., Alloway, T. P., Willis, C. S., & Adams, A. M. (2006). Workingmemory in children with reading disabilities. Journal of ExperimentalChild Psychology, 93, 265−281.

Gathercole, S. E., Brown, L., & Pickering, S. J. (2003). Working memoryassessments at school entry as longitudinal predictors of national curriculumattainment levels. Educational and Child Psychology, 20, 109−122.

Gathercole, S. E., & Pickering, S. J. (2000). Working memory deficits inchildren with low achievements in the national curriculum at 7 years ofage. The British Journal of Educational Psychology, 70, 177−194.

140 L. Lu et al. / Intelligence 39 (2011) 130–140

Gathercole, S. E., Pickering, S. J., Knight, C., & Stegmann, Z. (2004). Workingmemory skills and educational attainment: Evidence from nationalcurriculum assessments at 7 and 14 years of age. Applied CognitivePsychology, 18, 1−16.

Geary, D. C., Hoard, M. K., Byrd-Craven, J., & DeSoto, M. C. (2004). Strategychoices in simple and complex addition: Contributions of workingmemory and counting knowledge for children with mathematicaldisability. Journal of Experimental Child Psychology, 88, 121−151.

Geary, D. C., Hoard, M. K., & Hamson, C. O. (1999). Numerical and arithmeticalcognition: Patterns of function and deficits in children at risk for amathematical disability. Journal of Experimental Child Psychology, 74,213−231.

Gottfried, A. E. (1990). Academic intrinsic motivation in young elementaryschool children. Journal of Educational Psychology, 82, 525−538.

Hansford, B. C., & Hattie, J. A. (1982). The relationship between self andachievement/performance measures. Review of Educational Research, 52,123−142.

Hau, K. T., & Salili, F. (1991). Structure and semantic differential placement ofspecific causes: Academic causal attributions by Chinese students inHong Kong. International Journal of Psychology, 26, 175−193.

Helmke, A. (1992). Selbstvertrauen und schulische Leistung [Self-confidenceand school achievement]. Göttingen: Hogrefe.

Huntsinger, C. S., Jose, P. E., Larson, S. L., Krieg, D. B., & Shaligram, C. (2000).Mathematics, vocabulary, and reading development in Chinese Ameri-can and European American children over the primary school years.Journal of Educational Psychology, 92, 745−760.

Jensen, A. R. (1998). The g factor: The science of mental ability. Westport, CT:Praeger.

Kane, M. J., Hambrick, D. Z., Tuholski, S. W., Wilhelm, O., Payne, T. W., & Engle,R. W. (2004). The generality of working memory capacity: A latent-variable approach to verbal and visuospatial memory span andreasoning. Journal of Experimental Psychology: General, 133, 189−217.

Kaplan, E., Fein, D., Kramer, J., Delis, D., & Morris, R. (1999).WISC-III as a processinstrument (WISC-III-PI). New York: The Psychological Corporation.

Ko, H. W., & Chan, Y. L. (2009). Family factors and primary students' readingattainment — A Chinese community perspective. Chinese Education andSociety, 42, 33−48.

Kobayashi, Y. (1994). Conceptual acquisition and change through socialinteraction. Human Development, 37, 232−241.

Krumm, S., Ziegler, M., & Buehner, M. (2008). Reasoning and workingmemory as predictors of school grades. Learning and IndividualDifference, 18, 248−257.

Kyllonen, P. C., & Christal, R. E. (1990). Reasoning ability is (little more than)working-memory capacity? Intelligence, 14, 389−433.

LeBlanc, M. D., & Weber-Russell, S. (1996). Text integration and mathemat-ical connections: A computer model of arithmetic word problem solving.Cognitive Science, 20, 357−407.

Levy, S. R., & Dweck, C. S. (1998). Trait- versus process-focused socialjudgment. Social Cognition, 16, 151−172.

Lewis, C. C. (1995). Educating hearts andminds: Reflections on Japanese preschooland elementary education. New York: Cambridge University Press.

Li, J. (2003). U.S. and Chinese cultural beliefs about learning. Journal ofEducational Psychology, 95, 258−267.

Little, R. J. A., & Rubin, D. B. (2002). Statistical analysis with missing data (2ndedition). Hoboken: Wiley.

Llewellyn, J., Hancock, G., Kirst, M., & Roeloffs, K. (1982). A perspective oneducation in Hong Kong. Hong Kong: Hong Kong Government Press.

Luo, D., Thompson, L. A., & Detterman, D. K. (2006). The criterion validity oftasks of basic cognitive processes. Intelligence, 34, 79−120.

Maehler, C., & Schuchardt, K. (2009). Working memory functioning inchildren with learning disabilities: Does intelligence make a difference?Journal of Intellectual Disability Research, 53, 3−10.

Miyake, A., & Shah, P. (1999). Models of working memory: Mechanisms ofactive maintenance and executive control. New York, NY: CambridgeUniversity Press.

Oberauer, K., Schulze, R., Wilhelm, O., & Suess, H. M. (2005). Working memoryand intelligence — Their correlation and their relation: Comment onAckerman, Beier, and Boyle (2005). Psychological Bulletin, 131, 61−65.

Oliver, B. R., & Plomin, R. P. (2007). Twins' early development study (TEDS):A multivariate, longitudinal genetic investigation of language, cognition,and behavior problems from childhood through adolescence. TwinResearch and Human Genetics, 10, 96−105.

Peterson, J. (1925). Early conceptions and tests of intelligence. Yonkerson-Hudson, New York: World Book.

Pintrich, P. R., & Schunk, D. H. (2002). Motivation in education. EnglewoodCliffs, NJ: Merrill.

Salili, F. (1996). Accepting personal responsibility for learning. In D. A.Watkins, & J. B. Biggs (Eds.), The Chinese learner: Cultural, psychological,and contextual influences (pp. 85−106). Hong Kong: ComparativeEducation Research Centre.

Sarason, B. R., Pierce, G. R., Bannerman, A., & Sarason, I. G. (1993).Investigating the antecedents of perceived social support: Parents'views of and behavioral toward their children. Journal of Personalityand Social Psychology, 66, 1071−1085.

Schicke, M., & Fagan, T. K. (1994). Contributions of self-concept andintelligence to the prediction of academic achievement among grade 4,6, and 8 students. Canadian Journal of School Psychology, 10, 62−69.

Smith, P., Fernandes, C., & Strand, S. (2001). Cognitive abilities test TechnicalManual (Third Edition). . London: nferNelson.

Spinath, F. M., & Spinath, B. (2005). Development of self-perceived ability inelementary school: The role of parents' perceptions, teacher evaluations,and intelligence. Cognitive Development, 20, 190−204.

Spinath, B., Spinath, F. M., Harlaar, N., & Plomin, R. (2006). Predicting schoolachievement from general cognitive ability, self-perceived ability, andintrinsic value. Intelligence, 34, 363−374.

St. Clair-Thompson, H. L., & Gathercole, S. E. (2006). Executive functions andachievements in school: Shifting, updating, inhibition, and workingmemory. The Quarterly Journal of Experimental Psychology, 59, 745−759.

Stauffer, J., Ree, M., & Carreta, T. (1996). Cognitive-components tests are notmuch more than g: An extension of Kyllonen's analyses. The Journal ofGeneral Psychology, 123, 193−205.

Stevenson, H. W., Chen, C., & Lee, S. -Y. (1993). Mathematics achievement ofChinese, Japanese, and American children: Ten years later. Science, 259,53−58.

Stevenson, H. W., & Stigler, J. W. (1992). The learning gap: Why our schools arefailing and what we can learn from Japanese and Chinese education. NewYork: Simon & Schuster.

Stigler, J. W., Lee, S., & Stevenson, H. W. (1986). Digit memory in Chinese andEnglish: Evidence for a temporally limited store. Cognition, 23, 1−20.

Swanson, H. L. (2006). Cross-sectional and incremental changes in workingmemory and mathematical problem solving. Journal of EducationalPsychology, 98, 265−281.

Swanson, H. L., Ashbaker, M. H., & Lee, C. (1996). Learning disabled readersworking memory as a function of processing demands. Journal ofExperimental Child Psychology, 61, 242−275.

Swanson, H. L., Jerman, O., & Zheng, X. (2008). Growth inworkingmemory andmathematical problem solving in children at risk and not at risk for seriousmath difficulties. Journal of Educational Psychology, 100, 343−379.

Swanson, H. L., & Sachse-Lee, C. (2001). Mathematic problem solving andworking memory in children with learning disabilities: Both executiveand phonological processes are important. Journal of Experimental ChildPsychology, 79, 294−321.

Tweed, R. G., & Lehman, D. R. (2002). Learning considered within a culturalcontext: Confucian and Socratic approaches. The American Psychologist,57, 89−99.

Voyer, D. (1996). The relation between mathematical achievement andgender differences in spatial abilities: A suppression effect. Journal ofEducational Psychology, 88, 563−571.

Wang, R. S. (2002). Comparison and discussion of different readingpreference between Chinese and Japanese teaching. Chinese abroadtransmission, 10, 40−42.

Weismer, S. E., Tomblin, J. B., Zhang, X., Buckwalter, P., Chynoweth, J. G., &Jones, M. (2000). Nonword repetition performance in school-agechildren with and without language impairment. Journal of Speech,Language, and Hearing Research, 43, 865−878.

Wigfield, A., & Eccles, J. S. (2000). Expectancy–value theory of achievementmotivation. Contemporary Educational Psychology, 25, 68−81.

Williams, P. C., McCallum, R. S., & Reed, M. T. (1996). Predictive validity of theCattell–Horn Gf–Gc constructs to achievement. Assessment, 3, 43−51.

Winter, S. (1996). Peer tutoring and learning outcomes. In D. A. Watkins, & J.B. Biggs (Eds.), The Chinese learner: Cultural, psychological, and contextualinfluences (pp. 221−242). Hong Kong: Comparative Education ResearchCentre.

Yao, E. (1985). A comparison of family characteristics of Asian–American andAnglo-American high achievers. International Journal of ComparativeSociology, 16, 198−208.

Zheng, B. Y. (1997). Values in primers: A comparative study of Chinese andAmerican fourth grade reading textbooks. Humanities and Social Sciences,57, 505.