Not All Executive Functions Are Related to Intelligence

Transcript

1 PSYCHOLOGICAL SCIENCE Research Article Not All Executive Functions Are Related to Intelligence 1 1 2 1 1 Robin P. Corley, John C. DeFries, Akira Miyake, Naomi P. Friedman, Susan E. Young, and 1 John K. Hewitt 2 1 Department of Psychology, University Institute for Behavioral Genetics, University of Colorado at Boulder, and of Colorado at Boulder as the Wechsler Adult Intelligence Scale (WAIS), which pro- Accumulating evidence suggests that executive ABSTRACT— vides a composite intelligence quotient (IQ) derived from mul- functions (EFs) are related to intelligence, despite neuro- tiple subtests. These paradoxical neuropsychological findings psychological results initially considered evidence of no have long been interpreted as evidence that EFs are unrelated to such relation. However, findings that EFs are not unitary intelligence. raise the issue of how intelligence relates to different EFs. Duncan, Burgess, and Emslie (1995) proposed that this par- This study examined the relations of fluid and crystallized adox can be resolved by considering Cattell’s distinction be- intelligence and Wechsler Adult Intelligence Scale IQ to tween fluid intelligence (Gf)––which reflects higher mental three separable EFs––inhibiting prepotent responses (in- abilities, including reasoning––and crystallized intelligence hibiting), shifting mental sets (shifting), and updating working memory (updating)—in young adults. Updating (Gc)––which reflects knowledge acquired, partly through Gf, was highly correlated with the intelligence measures, but from culture, education, and other experiences (Carroll, 1993). inhibiting and shifting were not. Furthermore, in struc- Because already acquired knowledge may be more robust to tural equation models controlling for the inter-EF corre- frontal damage than fluid reasoning is, frontal lobe patients may lations, updating remained strongly related to intelligence, show deficits on tests of Gf, but not Gc. Hence, standard intel- but the relations of inhibiting and shifting to intelligence ligence tests like the WAIS may be relatively insensitive to were small and not significant. The results indicate that frontal damage because of their partial dependence on Gc intelligence measures differentially relate to these three measures. Duncan et al. found that frontal patients did show EFs, suggesting that current intelligence measures do not impaired intelligence on measures of Gf, such as Raven’s Pro- equally assess a wide range of executive control abilities gressive Matrices Test. Moreover, Duncan, Emslie, Williams, likely required for many ‘‘intelligent’’ behaviors. Johnson, and Freer (1996) found that executive problems such as neglecting to carry out goals were related to Gf in both normal adults and frontal lobe patients. Executive functions (EFs) are processes that control and regu- This neuropsychological evidence that Gf may be particularly late thought and action (e.g., suppressing habitual responses). sensitive to frontal lobe damage has influenced research with They are frequently associated with the brain’s frontal lobes; normal populations: Many nonneuropsychological studies ex- individuals with frontal lobe damage, in addition to performing amining the relation between EFs and intelligence have focused poorly on neuropsychological EF tasks such as the Wisconsin on Gf and largely ignored Gc. However, the distinction between Card Sorting Test, often exhibit deficits in planning, decision Gf and Gc may be less important for understanding the relations making, and generally regulating everyday behavior (Damasio, between EFs and intelligence in populations with no frontal 1994), which are considered hallmarks of intelligence (e.g., degradation. Because knowledge acquisition, the result of which Sternberg, 1988). Despite such deficits, some individuals with is Gc, may depend partly on Gf (Carroll, 1993), and because frontal lobe damage and corresponding EF deficits show normal there is no brain damage to selectively impair one type of in- intelligence, as measured by traditional psychometric tests such telligence, Gf and Gc may both be related to EFs in normal young adults. The current study evaluated this hypothesis by examining the relations of EFs to both Gf and Gc. Address correspondence to Naomi Friedman, Institute for Behavioral Previous studies of nonclinical populations have found some Genetics, 447 UCB, University of Colorado, Boulder, CO 80309, evidence that EFs are related to performance on tasks closely e-mail: [email protected] 172 Volume 17—Number 2 Copyright r 2006 Association for Psychological Science

2 N.P. Friedman et al. adults (e.g., Rockstroh & Schweizer, 2001). However, Salthouse associated with intelligence (Carpenter, Just, & Shell, 1990; et al. (1998) found a high correlation between shifting and Gf in Engle, Tuholski, Laughlin, & Conway, 1999; Miyake, Friedman, aging adults. Dempster (1991) referred to a study finding a small Rettinger, Shah, & Hegarty, 2001; Salthouse, Atkinson, & Be- but weak correlation in normal adults between WAIS IQ and the rish, 2003; Salthouse, Fristoe, McGuthry, & Hambrick, 1998). Wisconsin Card Sorting Test, which taps shifting (Miyake et al., In addition, Luciano et al. (2001) found that intelligence and 2000). In a study of adolescents, Ardila, Pineda, and Rosselli working memory capacity, a concept closely related to executive (2000) also found that performance on the Wisconsin Card functioning, share common genetic variance. Hence, both neu- Sorting Test correlated with WAIS IQ. ropsychological and nonclinical studies converge on the con- Unfortunately, much of the data on how intelligence relates to clusion that intelligence is related to EFs. different EFs comes from studies using popular neuropsycho- This conclusion, however, is complicated by emerging evi- logical tests as EF measures. Such complex EF-frontal tasks are dence that EFs, though correlated, are separable. Miyake et al. problematic EF measures for several reasons (Miyake et al., (2000) found that three EF latent variables––inhibiting prepo- 2000). First, they often have poor reliability. Second, because tent responses (inhibiting), updating working memory repre- they require controlling lower-level processes, they necessarily sentations (updating), and shifting between tasks or mental sets contain a good deal of variance unrelated to the EF of interest (shifting)––were moderately correlated but separable in college (i.e., task impurity). Finally, the particular EFs they tap are often students. Further, these EFs differentially predicted perfor- unclear. mance on four complex clinical and cognitive frontal lobe tasks In this study, we used multiple measures to construct EF la- (Wisconsin Card Sorting Test, Tower of Hanoi, random-number tent variables. Because latent variables extract the common generation, and a working memory span test). These results led variance from multiple measures, task-specific variance and Miyake et al. to conclude that EFs show both ‘‘unity and di- measurement error are largely eliminated, resulting in relatively versity.’’ pure EF measures and increased statistical power. Moreover, the These findings raise the question of how distinguishable EFs measures we used were simpler EF tasks (taken from Miyake relate to intelligence. Given that general intelligence is most et al., 2000) that are more tractable and well analyzed in the closely associated with complex reasoning and problem-solving literature, making their cognitive requirements better under- tasks (Carroll, 1993), and hence is ‘‘often taken to concern the stood than those of complex neuropsychological tasks. Such highest-level ‘executive’ or ‘supervisory’ functions of cognition’’ theoretically based cognitive constructs are an important tool in (Duncan et al., 1996, p. 258), one might posit that it would re- advancing understanding of intelligence, a still poorly under- late, possibly equally, to all EFs. However, this possibility re- stood psychometric construct (Kyllonen, 1996). mains untested. Although other EFs have been proposed (e.g., In addition to the issue of how closely individual EFs relate to time sharing), those examined by Miyake et al. (2000) have intelligence, a second question important to understanding the largely dominated the literature on the relations among EFs. relations between EFs and intelligence has not been answered: Available evidence suggests that at least some of these three EFs Do separable EFs show differential relations to intelligence, as might be related to intelligence. they do to various frontal lobe and complex cognitive tasks Several researchers have argued that intelligence is related to (Miyake et al., 2000)? Although there have been some proposals inhibiting. Salthouse et al. (2003) found that inhibiting strongly that all EF (and non-EF) abilities should be strongly associated correlated with Gf in aging adults. Dempster (1991) reviewed a with general intelligence in normal adults (Rabbitt, Lowe, & number of findings relating measures of intelligence, such as Shilling, 2001), we know of no study systematically targeting the WAIS IQ and aptitude-test performance, to several types of in- issue of how closely each of several EFs is related to intelligence hibition in both developmental populations (i.e., populations of and whether these relations are differential. Hence, we exam- children of several ages) and adult populations. These findings ined the relations of several intelligence measures (Gf, Gc, and led him to state that ‘‘intelligence cannot be understood without WAIS IQ) to inhibiting, updating, and shifting. reference to inhibitory processes’’ (p. 157). With respect to updating, Salthouse et al. (2003) found that it correlated with Gf in aging adults. Moreover, numerous re- METHOD searchers examining young adults have found moderate to strong relations between intelligence and working memory capacity Participants (see Ackerman, Beier, & Boyle, 2005). Because working memory Participants were 234 twins from the Colorado Longitudinal tests involve updating to maintain relevant information in the Twin Study. They were representative of the general population presence of interference, the finding that intelligence is related in their cognitive abilities (e.g., they showed a normally dis- to working memory capacity makes it likely that intelligence is tributed WAIS IQ distribution; see Table 1). All human research related to updating. guidelines were followed, and anonymity and confidentiality As for shifting, evidence is mixed. Some studies have found were maintained. The model estimation procedures corrected little relation between shifting tasks and intelligence in normal for the nonindependence of the twin pairs. Volume 17—Number 2 173

3 Executive Functions and Intelligence TABLE 1 Descriptive Statistics Twin correlation SD Mean N Task Minimum Maximum Skewness Kurtosis Reliability MZ DZ b n .83 .11 .45 .99 230 Antisaccade 0.50 .90 0.91  .62 .03 b n 1.49 220 Stop-signal .80 280 ms 1.06 503 151 60 .62 .09 b n 81 52 Stroop 0.52 0.04 .80 221 222 ms 469 .18 .47 c n n .10 .44 1.00  0.73 1.09 .65 Keep-track 231 .79 .30 .47 c n 232 Letter-memory .57 .01 .51 0.31 0.71  1.00 .53 .10 .86 a c n Spatial 2-back 0.19 0.50 1.57  1.20 1.86 .91 232 .29 1.17 .08 b n 0.58 0.85 985 23 .76 200 361 ms 230 Number-letter  .51 .16 b n 317 ms 186  Color-shape 974 0.90 1.12 .97 224 82 .48 .19 b n 230 360 ms 196 Category-switch 998 0.71 0.15 .78 –83 .18 .53 d n n 4 Raven 12 29 232  21 .84 0.65  0.06 .56 .37 e n n 0.13 19 4 3 .43 11 .60 234 .88 Block Design (WAIS) 0.19 b n n 10 1 61 0.66 Vocabulary .76 0.67 .89 231 .55 26 e n n 11 234 19 5 3 0.26 –0.08 Information (WAIS) .63 .89 .77 e n n 234 WAIS IQ .41 73 103 142 .84 11 .97 0.63 0.41 5 dizygotic (fraternal); WAIS 5 monozygotic (identical); DZ 5 MZ Note. Wechsler Adult Intelligence Scale. b a Scores for this task were arcsine transformed because of a ceiling effect. Internal reliability was calculated by adjusting split-half cor- c relations (Part 1–Part 2 for the vocabulary task and odd-even for all others) with the Spearman-Brown prophecy formula. Reliability was d e calculated using Cronbach’s alpha. Internal reliability from Internal reliability from DeFries, Plomin, Vandenberg, and Kuse (1981). Wechsler (1997). n .05. p < of two to four target categories and a series of 15 words, then Materials, Design, and Procedure recalled the last word presented that belonged to each of the The majority of the EF tasks were taken from Miyake et al. target categories. In each trial of the letter-memory task (after (2000), with some of the tasks from that study modified or Morris & Jones, 1990), participants continuously said the last replaced to make them more appropriate for the general popu- three letters presented in a running series of unpredictable lation or improve their validity, their reliability, or both. Par- length (five, seven, or nine letters), then recalled the final three ticipants completed the EF tasks (computerized) at ages 16 to 18 letters presented once the series stopped. In the spatial 2-back 5 M ( 5 SD 0.27) and the intelligence measures (paper and 17, task, participants saw a screen with small boxes darkened in a 0.27). SD 16, 5 M pencil) at ages 16 to 17 ( 5 random order and indicated whether each darkened box was the same as the one darkened two trials before. Inhibiting The inhibiting tasks required suppressing dominant or auto- Shifting matic responses. In the antisaccade task (Roberts, Hager, & The shifting tasks required participants to switch between Heron, 1994), participants suppressed the reflexive tendency to 5 subtasks (DV reaction time on switch trials  reaction time on look at a cue and instead looked in the opposite direction to no-switch trials). On each trial, a random cue indicating which identify a briefly appearing target (dependent variable, or DV 5 subtask to perform was presented just before the stimulus and proportion correct). The stop-signal task (Logan, 1994) allowed remained on the screen until the participant responded. In the participants to build a prepotent word-categorization response, number-letter task (Rogers & Monsell, 1995), participants then asked them to withhold that response on trials with beeps switched between classifying numbers and classifying letters. In 5 estimated stop-signal reaction time; see Logan). In the (DV the color-shape task (Miyake, Emerson, Padilla, & Ahn, 2004), Stroop task (after Stroop, 1935), participants resisted the dom- participants switched between classifying shapes and classify- inant tendency to read color words, instead naming the incon- ing colors. In the category-switch task (Mayr & Kliegl, 2000), gruent font color; on neutral trials, participants named the color participants switched between classifying the animacy of words 5 of a string of asterisks (DV naming time on incongruent trials and classifying the size of words. – naming time on neutral trials). Intelligence Updating Gf was measured with two tasks tapping problem-solving and The updating tasks required adding and deleting information in reasoning abilities. In Raven’s Progressive Matrices Test (Ra- proportion correct). In each trial of 5 working memory (all DVs ven, 1960), participants chose which pieces completed complex the keep-track task (after Yntema, 1963), participants saw a set 174 Volume 17—Number 2

4 N.P. Friedman et al. TABLE 2 Correlation Matrix 5 4 3 2 1 Task 7 6 8 9 10 11 12 13 14 –– 1. Antisaccade n .43 2. Stop-signal –– n n 3. Stroop .24 .22 –– n n n 4. Keep-track .21 –– .23 .20 n n n n .32 5. Letter-memory .25 .42 –– .25 n n n n 6. Spatial 2-back .17 .28 .29 .19 –– .10 n n n n n n .23 7. Number-letter .28 .22 .24 .13 .21 –– n n n n n n n 8. Color-shape .20 .15 .34 .20 .17 .47 –– .27 n n n n n n n n 9. Category-switch .26 .30 .16 .13 .48 –– .52 .20 .13 n n n n .23 10. Raven .00 .03 .03 .24 .28 .33 .09 –– .07 n n n n n .19 11. Block Design (WAIS) .01 .26 .17 .22 .12 .06 .11 .43 –– .05 n n n n n n n n n 12. Vocabulary .12 .16 .11 .36 –– .16 .32 .29 .51 .26 .25 .15 n n n n n n n n n n .17 13. Information (WAIS) .37 .24 .68 .04 .15 –– .15 .16 .41 .38 .19 .10 n n n n n n n n n n n n 14. WAIS IQ .25 .62 .06 .25 .69 .37 .80 .52 –– .23 .16 .19 .21 .47 Note. The correlation matrix was estimated with maximum likelihood and was adjusted for nonindependence and missing data by Mplus. Wechsler Adult 5 For reaction time measures, scores were reversed so that higher numbers indicate better (faster) performance. WAIS Intelligence Scale. n < p N 5 234. .05 for certained by testing whether dropping them resulted in a sig- patterns. In the WAIS Block Design subtest (Wechsler, 1997), nificant chi-square difference, appropriately scaled (Satorra & participants used blocks to reconstruct patterns. This subtest Bentler, 2001); this method is more reliable than test statistics has been used previously to measure Gf because of its sensitivity based on standard errors (Gonzalez & Griffin, 2001). to frontal damage (Lezak, 1995) and close relation to other Gf tasks (Carpenter et al., 1990). Gc was measured with two tests assessing acquired knowledge. A multiple-choice vocabulary test (described in DeFries, Plomin, Vandenberg, & Kuse, 1981) RESULTS AND DISCUSSION required participants to identify the synonyms or definitions of words. The WAIS Information subtest asked questions tapping Descriptive information and zero-order correlations for the in- general factual knowledge. WAIS IQ, a composite general in- 1 dividual tasks are presented in Tables 1 and 2. To examine the telligence measure based on 11 subtests, was also examined. relations between the EFs and intelligence, we used confirma- Except for WAIS IQ, the DV for each measure of intelligence tory factor analysis (CFA), a procedure that allows one to extract was the number correct. latent variables based on a priori factor patterns and estimate the correlations between these latent variables. Specific hypotheses can be evaluated by imposing model constraints (e.g., con- Statistical Procedures straining a correlation to zero) and testing whether the con- ́ ́ We used Mplus 3.1 (Muthe n & Muthe n, 2004) to estimate latent strained model provides a significantly worse fit to the data variable models with missing data (34 participants were missing 2 w (indicated by a significant difference test). We estimated a data for one or more tasks because of experimenter, equipment, CFA with the three EFs, Gf, and Gc. The top portion of Table 3 or participant error; see Table 1). Because the participants were provides the factor loadings of each task on its latent variable, twin pairs, we used an Mplus option that calculates parameters, and the bottom portion provides the interfactor correlations. standard errors, and a scaled chi-square robust to noninde- The tasks all loaded significantly on their intended constructs, pendence. Because models with sample sizes necessary for la- 2 2 and model fit was good, w 1.53; (55) df 5 / 5 84.10, p 5 .007; w tent variable analyses often show a significant chi-square 5 CFI 5 SRMR .952; .044. (suggesting poor fit to the data) despite only trivial differences between the model’s predictions and the observed data, we used 1 Table 1 presents raw twin correlations based on approximately 60 monozy- the common criterion of the chi-square divided by the degrees of gotic and 50 dizygotic twin pairs. Although this sample is too small to provide accurate estimates of heritabilities, the generally higher correlations for mono- freedom being less than 2 as an indication of adequate fit. We zygotic twins (who share 100% of their genes) than for dizygotic twins (who share also used two other fit criteria recommended by Hu and Bentler on average 50% of their genes) suggest that the abilities tapped by these tasks, and likely the latent variables, are heritable. Note that the lower intertwin cor- (1998): Bentler’s comparative fit index (CFI) greater than .95 relations for dizygotic twins do not mean that dizygotic individuals had lower and standardized root mean squared residual (SRMR) less than intertask correlations than monozygotic individuals, so the correlations between .08. Significance of correlation and path coefficients was as- measures pooled over zygosity (Table 2) are not driven by the monozygotic data. Volume 17—Number 2 175

5 Executive Functions and Intelligence 2 .001. < p 42, : 13 Þ¼ 2 ð TABLE 3 correlations to be equal worsened fit, w diff Confirmatory Factor Analysis with Gf and Gc Hence, the three EFs differentially relate to both Gf and Gc. The CFA model calculates the EF-intelligence relations Latent factor without controlling for EF intercorrelations. We used structural Variable Shifting Updating Gf Gc Inhibiting equation models (SEMs) to examine how the three EFs relate to Factor loadings intelligence, considering their intercorrelations. SEMs allow for n Antisaccade .61 examination of more complex hypotheses by specifying direct n Stop-signal .64 paths (i.e., structural relations) from predictors to predicted n Stroop .43 latent variables in place of correlations. Further, SEMs allow n Keep-track .72 one to estimate how much variance in each predicted variable is n Letter-memory .60 n explained by the predictors considered together. In the models Spatial 2-back .41 n here, the difference between CFA models and SEMs is analo- .65 Number-letter n .73 Color-shape gous to that between correlations and multiple regressions. We n .72 Category-switch estimated an SEM with the correlated EFs predicting Gf and Gc n Raven .67 (see Fig. 1; the tasks used to construct the latent variables are n Block Design (WAIS) .65 not included for simplicity; task loadings remained the same as n .86 Vocabulary those in Table 3). Note that the arrow directionality is not meant n .79 Information (WAIS) to imply causality; it merely allowed a test of how the EFs relate 2 Interfactor correlations to intelligence, controlling for their intercorrelations. Model fit Inhibiting –– 2 2 was good, w 84.10, / .952; 5 (55) 5 5 p 5 .007; w CFI df 1.53; n Updating .62 –– 3 SRMR 5 .044. Controlling for the other EFs, updating still n n Shifting .64 .39 –– < .001, but inhibiting ( p s 5 .680 predicted Gf and Gc, both p s n Gf .29 .64 –– .13 n n n n s 5 and .145, respectively) and shifting ( p .632 and .295, Gc .31 .68 –– .62 .31 respectively) did not. Hence, model fit was unharmed when all For reaction time measures, scores were reversed so that higher numbers Note. 2 paths from inhibiting and shifting were dropped, w ð 4 Þ¼ diff crystallized indicate better (faster) performance. Gf 5 fluid intelligence; Gc 5 : .382, and the resulting path coefficients for updating 5 p 18, 4 Wechsler Adult Intelligence Scale. 5 intelligence; WAIS n < p .05. predicting Gf and Gc were .61 and .67, respectively. As with the CFA, constraining the three EF path coefficients to either Gf or 2 w Gc to be equal significantly reduced fit, 91, ð 2 Þ¼ 18 : Before examining the correlations of the EFs with Gf and Gc, diff 2 w .001, and < p < .001, respectively. Hence, 2 Þ¼ 17 : 53, p ð we wanted to determine whether the structure of EFs found by diff the SEM results support the CFA finding that the three EFs Miyake et al. (2000) with a selected sample (college students) differentially relate to both Gf and Gc. was replicated in this sample, which was more cognitively In the CFA model, Gf and Gc correlated significantly (.62), representative of the general population. Miyake et al. found p < .001, as expected in this sample of normal young adults. In that inhibiting, shifting, and updating were moderately corre- the SEM, when the variance due to the three EFs was removed s lated ( 5 .42 to .63) but separable (none of the correlations r from Gf and Gc (the resulting residual variances are depicted in could be set to 1.0 without harming model fit). As shown in the the small circles on the far right of Fig. 1), the correlation top three rows of the correlation matrix in Table 3, the EF cor- p 5 .002). This between them dropped to .17 (still significant, relations ranged from .39 to .64, all within the 95% confidence decrease was significant; constraining the .62 correlation in the intervals of the respective correlations found by Miyake et al. 2 w CFA model to .17 significantly reduced model fit, 1 ð Þ¼ Model comparisons including only the EFs, to make the anal- diff .001. Hence, the EFs, particularly updating, con- 98, < 13 : p yses comparable to those of Miyake et al., indicated that con- tribute substantially to the Gf-Gc correlation. straining any EF correlation to one or zero worsened model We used similar models to examine WAIS IQ, a frequently fit (all .001). Hence, these data replicate the unity and < s p used composite measure of general intelligence that includes diversity of EFs found by Miyake et al. subtests tapping Gf and Gc (two of which were used to construct Turning now to the EF-intelligence correlations, shown in the the Gf and Gc latent variables), as well as other abilities such as last two rows of the correlation matrix in Table 3, Gf showed 2 w short-term memory. Model fits were good, 5 (30) 37.37, 5 p nonsignificant correlations with inhibiting (.29), p 5 .081, and .277, but a larger correlation with updating 5 p shifting (.13), 2 (.64), .001. Constraining these correlations to be equal p < Note that the moderate EF intercorrelations resulted in some multicolline- 2 arity, which tends to increase the standard errors (decreasing the precision) of the w worsened fit, 17 : 41, p ð .001. The EF-Gc correla- 2 Þ¼ < diff path coefficient estimates. tions were slightly higher and all significant, but updating still 3 The CFA and SEM fit indices were identical because these models are showed a higher correlation (.68), .001, than inhibiting < p ‘‘equivalent’’ in that they mathematically predict the same covariances, but with p .002, and shifting (.31), 5 p < .001. Constraining these (.31), different theoretical decompositions of these covariances. 176 Volume 17—Number 2

6 N.P. Friedman et al. Fig. 1. Structural equation model predicting fluid intelligence (Gf) and crystallized intelligence (Gc) with the three executive functions (EFs). Ellipses represent latent variables. The model is conceptually similar to multiple regressions with Gf and Gc as the dependent variables and the three EFs as correlated independent variables. The numbers on the straight, single-headed arrows are the standardized path coefficients (interpretable as stan- dardized regression coefficients). These numbers reflect the contribution of each EF to Gf and Gc, controlling for the other EFs. Boldface type and solid lines indicate coefficients significant at the .05 level, whereas dotted lines and normal type indicate nonsignificant coefficients. The circled numbers next to the small arrows pointing to Gf and Gc are their residual variances. Subtracting each of these numbers from 1.0 provides the amount of variance in each type of intelligence that is predicted by the three EFs, considered together. Squaring each path coefficient provides the amount of variance in each type of intelligence that is predicted by each EF. The number on the curved arrow connecting the Gf and Gc residual variances indicates the extent to which the two types of intelligence correlate after the variance due to the EFs is removed from each one. 2 df 5 1.25; CFI 5 .981; SRMR 5 .036. In the CFA, .167; / w relations between updating and intelligence measures were WAIS IQ significantly correlated with inhibiting (.38), p < .001; undiminished, but the relations between inhibiting and intelli- updating (.69), p < .001; and shifting (.23), .003. These 5 p gence and between shifting and intelligence were no longer 2 correlations could not be equated, w significant. These results suggest that the small correlations of Þ¼ ð .001. < p 44, : 33 2 diff 4 inhibiting and shifting with intelligence measures in the CFAs In the SEM predicting WAIS IQ with the correlated EFs, the were due to the variance they shared with updating. Even con- .07), .01), .639, and shifting ( paths from inhibiting (  p p 5 5  trolling for its correlations with the other EFs, updating alone .841, were not significant, whereas the path from updating (.74), accounted for 37% to 45% of the intelligence measures’ vari- < p .001, was. Dropping the paths from inhibiting and shifting 2 ances. Moreover, from the CFA to the SEM with Gf and Gc, the w did not harm model fit, 2 .763, and the path Þ¼ 5 0 ð p 54, : diff Gf-Gc correlation dropped 73%, indicating that the EFs, par- coefficient from updating alone was .67. As with the CFA, ticularly updating, accounted for a significant portion (though equating the paths between the EFs and WAIS IQ worsened fit, 2 not all) of the Gf-Gc correlation. Note that 49% to 57% of the w 2 .001. Hence, the EFs showed the same ð p Þ¼ : < 76, 42 diff variances in the intelligence measures were unexplained by the differential relations to WAIS IQ as to Gf and Gc, with updating EFs, reflecting the fact that EFs, though important correlates of relating similarly to all three intelligence measures. intelligence, are not the only ones. In this population of young adults, the three different mea- GENERAL DISCUSSION sures of intelligence showed virtually identical patterns of rela- tionships with the three EFs examined. Although the literature The goal of this study was to examine, in normal young adults, to on the relations between EFs and intelligence has focused what extent inhibiting, updating, and shifting relate to intelli- largely on Gf, the similarity of the results with Gf to the results gence measures, and whether these relations are differential. with Gc and WAIS IQ suggests that Gf may not necessarily be The conclusions based on Gf, Gc, and WAIS IQ were identical: more strongly associated with EFs than are other measures of These three EFs differentially relate to intelligence in normal intelligence in young adults, for whom Gf likely strongly influ- young adults, with updating being the EF most closely related to ences knowledge acquisition (the result of which is Gc). How- intelligence. In the CFAs, the three intelligence measures ever, in populations with reduced frontal integrity, such as older shared 41% to 48% of their variances with updating, but only adults and frontal lobe patients, one might expect Gf to show 2% to 14% of their variances with inhibiting and shifting. SEMs more EF involvement than Gc or WAIS IQ, because Gc may be revealed that when inter-EF correlations were considered, the relatively unaffected by frontally related EF dysfunction (Duncan 4 et al., 1995). The three EFs together explained 52% of WAIS IQ variance. Volume 17—Number 2 177

7 Executive Functions and Intelligence cognitive flexibility) are often used to denote the ability to The strong relation between updating and intelligence is flexibly switch mental sets or avoid being stuck on ineffective consistent with numerous findings of an association between strategies or mundane viewpoints (e.g., Lohman, 2000; Stern- intelligence and working memory capacity (e.g., Carpenter et berg, 1988). Thus, to the extent that the simple EFs in the al., 1990; Engle et al., 1999). These results highlight the im- current study actually tap these hypothesized roles (see Fried- portance of updating abilities in current conceptions of intelli- man & Miyake, 2004, for evidence suggesting a link between gence. Updating and working memory capacity have been inhibiting and everyday cognitive functioning), the current described as abilities that involve attentional control to maintain finding that not all EFs are related to psychometric intelligence relevant information (including task goals) in the face of inter- suggests that traditional measures of intelligence are missing ference, delete this information when it becomes irrelevant, and some fundamental supervisory functions. replace it with new information (Engle et al., 1999; Miyake et al., 2000). This description overlaps considerably with a definition of intelligence articulated by Binet: ‘‘[It] consists of two chief Acknowledgments— This research was supported by grants processes: First to perceive the external world, and then to re- from the National Institute of Mental Health (MH63207 and instate the perceptions in memory, to rework them, and to think MH01865) and National Institute of Child Health and Human about them’’ (translation by Carroll, 1993, p. 35). Development (HD007289 and HD010333). We thank Sally Ann The weak to nonexistent relations between intelligence and Rhea and Scott Sabella for coordinating the project and col- the other two EFs, particularly inhibiting, may initially seem surprising; however, much of the evidence for strong relations lecting data. between these EFs and intelligence comes from studies of populations with compromised frontal lobe integrity, such as REFERENCES clinical and aging populations (e.g., Salthouse et al., 1998, 2003). Although the current data do not speak directly to the Ackerman, P.L., Beier, M.E., & Boyle, M.O. (2005). Working memory relations between EFs and intelligence in these populations, one Psychological and intelligence: The same or different constructs? possibility suggested by Rabbitt et al. (2001) is that when frontal Bulletin , , 30–60. 131 Ardila, A., Pineda, D., & Rosselli, M. (2000). Correlation between in- lobe functioning is generally compromised, multiple EFs may be Archives of telligence test scores and executive function measures. affected, leading to higher inter-EF correlations. These higher 15 , Clinical Neuropsychology , 31–36. correlations could then result in generally higher EF-intelli- Carpenter, P.A., Just, M.A., & Shell, P. (1990). What one intelligence gence correlations. Indeed, Salthouse et al. (2003), examining test measures: A theoretical account of the processing in the an aging sample, found substantially higher inhibiting-updating 97 Raven Progressive Matrices Test. , Psychological Review , 404– 431. (.71), inhibiting-Gf (.73), and updating-Gf correlations (.93) Human cognitive abilities: A survey of factor-an- Carroll, J.B. (1993). than those found here. . New York: Cambridge University Press. alytic studies The current study has interesting implications for intelligence Damasio, A.R. (1994). Descartes’ error: Emotion, reason, and the human research. Specifically, the weak to nonexistent relations be- brain . New York: Avon Books. tween the intelligence measures and both inhibiting and shifting Current Directions in Das, J.P. (2002). A better look at intelligence. highlights some discrepancies between psychometric intelli- 11 Psychological Science , , 28–33. DeFries, J.C., Plomin, R., Vandenberg, S.G., & Kuse, A.R. (1981). gence measures and many theoretical conceptions of intelli- Parent-offspring resemblance for cognitive abilities in the Colo- gence. Although intelligence theorists have repeatedly noted rado Adoption Project: Biological, adoptive, and control parents the importance of inhibiting and shifting abilities to intelligent , 245–277. 5 , and one-year-old children. Intelligence behavior, tests traditionally used to assess intelligence in normal Dempster, F.N. (1991). Inhibitory processes: A neglected dimension of adults do not seem to tap these two EFs much (if at all). intelligence. , 15 , 157–173. Intelligence For example, Thurstone (1924) defined intelligence as self- Duncan, J., Burgess, P., & Emslie, H. (1995). Fluid intelligence after , 261–268. 33 , Neuropsychologia frontal lobe lesions. control over reflexive or instinctive impulses, arguing that in- Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). hibiting such impulses allows one to rationally consider options Intelligence and the frontal lobe: The organization of goal-directed and intelligently adjust to situations. More recently, Sternberg 30 , Cognitive Psychology behavior. , 257–303. (1988) defined intelligence as ‘‘mental self-management’’ (p. 72) Engle, R.W., Tuholski, S.W., Laughlin, J.E., & Conway, A.R.A. (1999). needed to adapt, select, and shape the environment, citing lack Working memory, short-term memory, and general fluid intelli- of impulse control as one cause of self-management failures. gence: A latent-variable approach. Journal of Experimental Psy- 128 chology: General , , 309–331. In addition, Dempster (1991) and Das (2002) have argued that Friedman, N.P., & Miyake, A. (2004). The relations among inhibition the ability to resist interference from distracting, irrelevant in- and interference control processes: A latent variable analysis. formation (an ability closely related to inhibiting, as demonstra- , 101–135. Journal of Experimental Psychology: General , 133 ted by Friedman & Miyake, 2004) is an important component Gonzalez, R., & Griffin, D. (2001). Testing parameters in structural of intelligent behavior. Similar theoretical claims can also equation modeling: Every ‘‘one’’ matters. Psychological Methods , be found for shifting, although different terms (e.g., mental or 6 , 258–269. 178 Volume 17—Number 2

8 N.P. Friedman et al. Rabbitt, P., Lowe, C., & Shilling, V. (2001). Frontal tests and models for Hu, L., & Bentler, P.M. (1998). Fit indices in covariance structure , cognitive ageing. , 13 European Journal of Cognitive Psychology modeling: Sensitivity to underparameterized model misspecifi- 5–28. 3 , 424–453. , Psychological Methods cation. . Standard Progressive Matrices: Sets A, B, C, D, & E Raven, J.C. (1960). g ?InI. Kyllonen, P.C. (1996). Is working memory capacity Spearman’s London: H.K. Lewis & Co. Dennis & P. Tapsfield (Eds.), Human abilities: Their nature and Roberts, R.J., Hager, L.D., & Heron, C. (1994). Prefrontal cognitive measurement (pp. 49–75). Mahwah, NJ: Erlbaum. processes: Working memory and inhibition in the antisaccade Neuropsychological assessment (3rd ed.). New York: Lezak, M.D. (1995). , 374–393. task. Journal of Experimental Psychology: General , 123 Oxford University Press. Rockstroh, S., & Schweizer, K. (2001). The contributions of memory Logan, G.D. (1994). On the ability to inhibit thought and action: A user’s and attention processes to cognitive abilities. Journal of General guide to the stop signal paradigm. In D. Dagenbach & T.H. Carr , 30–42. Psychology , 128 Inhibitory processes in attention, memory, and language (Eds.), Rogers, R.D., & Monsell, S. (1995). Costs of a predictable switch be- (pp. 189–239). San Diego, CA: Academic Press. Lohman, D.F. (2000). Complex information processing and intelli- Journal of Experimental Psychology: tween simple cognitive tasks. (pp. 285– Handbook of intelligence gence. In R.J. Sternberg (Ed.), , 207–231. General , 124 340). New York: Cambridge University Press. Salthouse, T.A., Atkinson, T.M., & Berish, D.E. (2003). Executive Luciano, M., Wright, M.J., Smith, G.A., Geffen, G.M., Geffen, L.B., & functioning as a potential mediator of age-related cognitive de- Martin, N.G. (2001). Genetic covariance among measures of in- Journal of Experimental Psychology: cline in normal adults. Behavior formation processing speed, working memory, and IQ. 132 General , , 566–594. , 581–592. 31 , Genetics Salthouse, T.A., Fristoe, N., McGuthry, K.E., & Hambrick, D.Z. (1998). Mayr, U., & Kliegl, R. (2000). Task-set switching and long-term memory Relation of task switching to speed, age, and fluid intelligence. Journal of Experimental Psychology: Learning, Memory, retrieval. Psychology and Aging , 13 , 445–461. and Cognition , 1124–1140. 26 , Satorra, A., & Bentler, P.M. (2001). A scaled difference chi-square test Miyake, A., Emerson, M.J., Padilla, F., & Ahn, J. (2004). Inner speech , 66 Psychometrika statistic for moment structure analysis. , 507– as a retrieval aid for task goals: The effects of cue type and ar- 514. ticulatory suppression in the random task cuing paradigm. Acta The triarchic mind: A new theory of human Sternberg, R.J. (1988). , Psychologica , 123–142. 115 intelligence . New York: Viking Penguin. Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., Howerter, A., Stroop, J.R. (1935). Studies of interference in serial verbal reactions. & Wager, T.D. (2000). The unity and diversity of executive func- 18 , 643–662. Journal of Experimental Psychology , tions and their contributions to complex ‘‘frontal lobe’’ tasks: A . London: Kegan Paul, The nature of intelligence Thurstone, L.L. (1924). , 49–100. , latent variable analysis. Cognitive Psychology 41 Trench, Trubner & Co. Miyake, A., Friedman, N.P., Rettinger, D.A., Shah, P., & Hegarty, M. Wechsler Adult Intelligence Scale (3rd ed.). San Wechsler, D. (1997). (2001). How are visuospatial working memory, executive func- Antonio, TX: Psychological Corp. tioning, and spatial abilities related? A latent variable analysis. Yntema, D.B. (1963). Keeping track of several things at once. Human Journal of Experimental Psychology: General , , 621–640. 130 , 7–17. Factors , 5 Morris, N., & Jones, D.M. (1990). Memory updating in working memory: , British Journal of Psychology The role of the central executive. 81 , 111–121. ́ ́ n, B.O. (2004). n, L.K., & Muthe (3rd ed.). Los Mplus user’s guide Muthe (R EVISION ACCEPTED ECEIVED 12/7/04; R 2/2/05; ́ ́ Angeles: Muthe n & Muthe n. F INAL MATERIALS RECEIVED 2/14/05) Volume 17—Number 2 179

Related documents

PostScript Language Reference Manual

PostScript Language Reference Manual

See also cross-references: abcdefghijklmnopqrltuvw PLRM 2nd Edition July 27, 1992 ® PostScript Language Reference Manual SECOND EDITION Adobe Systems Incorporated Addison-Wesley Publishing Company, In...

More info »
CDIR 2018 07 27

CDIR 2018 07 27

S. Pub. 115-7 2017-2018 Official Congressional Directory 115th Congress Convened January 3, 2017 JOINT COMMITTEE ON PRINTING UNITED STATES CONGRESS UNITED STATES GOVERNMENT PUBLISHING OFFICE WASHINGTO...

More info »
AT Commands Reference Guide

AT Commands Reference Guide

AT Commands Reference Guide GE863-QUAD, GE863-PY, GE863-GPS, GM862-QUAD, GM862-QUAD-PY, GE862-GPS, GE864-QUAD, GE864-PY, GC864-QUAD and GC864-PY 80000ST10025a Rev. 0 - 04/08/06

More info »
DoD7045.7H

DoD7045.7H

DoD 7045.7-H EPARTMENT OF D EFENSE D F UTURE Y EARS D EFENSE P ROGRAM (FYDP) S TRUCTURE Codes and Definitions for All DoD Components Office of the Director, Program Analysis and Evaluation A pril 2004

More info »
10749

10749

Cash transfers: what does the evidence say? A rigorous review of programme impact and of the role of design and implementation features Francesca Bastagli, Jessica Hagen-Zanker, Luke Harman, Valentina...

More info »
A 01.00

A 01.00

M.P.S.C. No. 13 - Electric Second Revised Sheet No. A 1.00 Consumers Energy Company - Cancels First Revised Sheet No. A 1.00 (To revise link to website) - CONSUMERS ENERGY COMPANY BOOK RATE FOR ELECTR...

More info »
Investors Exchange Rule Book

Investors Exchange Rule Book

Updated April 8 * , 2019 through ubsequent changes 222 (June 1 - 78101; File No. 10 - Investors Exchange rules were adopted by Securities Exchange Act Release No. 34 * 7, 2016). Dates of s are indicat...

More info »
javanotes7 linked

javanotes7 linked

Introduction to Programming Using Java Version 7.0, August 2014 ( Version 7.0.2, with just a few corrections, December 2016 ) David J. Eck Hobart and William Smith Colleges This is a PDF version of a ...

More info »
SR288.PS

SR288.PS

113th Congress S. Report ! " SENATE 2d Session 113–288 REPORT of the SENATE SELECT COMMITTEE ON INTELLIGENCE COMMITTEE STUDY of the CENTRAL INTELLIGENCE AGENCY’S DETENTION AND INTERROGATION PROGRAM to...

More info »
ref frame matter.book

ref frame matter.book

Universal Verification Methodology (UVM) 1.1 Class Reference June 2011

More info »
MPI: A Message Passing Interface Standard

MPI: A Message Passing Interface Standard

MPI : A Message-Passing Interface Standard Version 3.0 Message Passing Interface Forum September 21, 2012

More info »
GPO PLUMBOOK 2016

GPO PLUMBOOK 2016

[COMMITTEE PRINT] RT . 114–26 S. P UNITED STATES GOVERNMENT Policy and Supporting Positions f Committee on Homeland Security and Governmental Affairs United States Senate 114th Congress, 2d Session DE...

More info »
Gutmans Frontmatter

Gutmans Frontmatter

Gutmans_Frontmatter Page i Thursday, September 23, 2004 9:05 AM PHP 5 Power Programming

More info »
Justification Book

Justification Book

UNCLASSIFIED Department of Defense Fiscal Year (FY) 2019 Budget Estimates February 2018 Office of the Secretary Of Defense Defense-Wide Justification Book Volume 3B of 5 Research, Development, Test & ...

More info »
94755 II

94755 II

94TH CONGRESS SENATE NoREPORT 2d Session ](No. 91,-755 INTELLIGENCE ACTIVITIES AND THE RIGHTS OF AMERICANS II BOOK FINAL REPORT THE OF SELECT COMMITTEE TO STUDY GOVERNMENTAL OPERATIONS WITH RESPECT TO...

More info »
C:\JIM\PLUMBOOK\2012PL~1\LIVE\76304.002

C:\JIM\PLUMBOOK\2012PL~1\LIVE\76304.002

[COMMITTEE PRINT] UNITED STATES GOVERNMENT Policy and Supporting Positions f Committee on Oversight and Government Reform U.S. House of Representatives 112th Congress, 2d Session DECEMBER 1, 2012 Avai...

More info »
WJP ROLI 2017 18 Online Edition 0

WJP ROLI 2017 18 Online Edition 0

l i a a o a i z a n r a s n i a i i a a n a e r d v v z F o i g a w i B o l l o l o m s d J a I o b e r t u g e s a t a u l a n i e i m e o B u B B n i l m z a r y g h s a a B r l k a m a n i o r J l ...

More info »
SAP S/4HANA 1809   Feature Scope Description

SAP S/4HANA 1809 Feature Scope Description

PUBLIC Document Version: 1.2 – 2019-01-21 SAP S/4HANA 1809 - Feature Scope Description company. All rights reserved. affiliate THE BEST RUN 2019 SAP SE or an SAP ©

More info »
Security and Privacy Controls for Federal Information Systems and Organizations

Security and Privacy Controls for Federal Information Systems and Organizations

-53 NIST Special Publication 800 Revision 4 Security and Privacy Controls for Federal Information Systems and Organizations JOINT TASK FORCE TRANSFORMATION INITIATIVE This publication is available fre...

More info »