Cognitive Psychology discussion posts

PAPER

Modeling a cascade of effects: the role of speed and executive functioning in preterm ⁄full-term differences in academic achievement

Susan A. Rose,1 Judith F. Feldman2 and Jeffery J. Jankowski1,2

1. Department of Pediatrics, Kennedy Center, Albert Einstein College of Medicine ⁄ Children’s Hospital at Montefiore, USA

2. Department of Social Sciences, Queensborough Community College ⁄ CUNY, USA

Abstract

This study identified deficits in executive functioning in pre-adolescent preterms and modeled their role, along with processing speed, in explaining preterm ⁄ full-term differences in reading and mathematics. Preterms (< 1750 g) showed deficits at 11 years on a battery of tasks tapping the three basic executive functions identified by Miyake – updating ⁄ working memory, inhibition, and shifting. Confirmatory factor analysis showed that these executive functions, though correlated, were distinct from one another and from processing speed, which later proved to account for much of the intercorrelation among executive functions. In the best-fitting structural equation model, the negative effects of prematurity on achievement were completely mediated by the three executive functions and speed in a cascade of effects: prematurity fi slower processing speed fi poorer executive functioning (working memory) fi lower achievement in math and reading.

Introduction

There is widespread agreement that children born pre- maturely are at risk not only for neurological and physical handicaps, but also for more subtle intellectual and behavioral problems (Aylward, 2002). Academic achieve- ment is particularly problematic, with preterms commonly reported to lag in math and reading, and to have a rela- tively high incidence of learning disabilities and school failure (Allen, 2008; Bhutta, Cleves, Casey, Cradock & Anand, 2002; Botting, Powls, Cooke & Marlow, 1998; Espy, Fang, Charak, Minich & Taylor, 2009; McCormick, 1989; Pinto-Martin, Whitaker, Feldman, Cnaan, Zhao, Rosen-Bloch, McCulloch & Paneth, 2004; Saigal, Hoult & Streiner, 2000). A population-based study of 11-year-olds from the United Kingdom found that up to 44% of extremely low gestational-aged preterms (£ 25 weeks) had a serious impairment in math and reading (Johnson, Hennessy, Smith, Trikic, Wolke & Marlow, 2010), and a recent meta-analysis found that very preterm and ⁄ or very low birth weight (VLBW) children scored around a half SD lower than full-term controls in these core subject areas (Aarnoudse-Moens, Weisglas-Kuperus & van Goudoever, 2009). In this article we present evidence that the relative difficulties found among preterms in academic achieve- ment can be explained by deficits in executive function (EF) and processing speed.

Executive function

EF is an umbrella term used to describe a number of higher-order cognitive processes needed for self-regula- tion, planning, problem solving, and flexible, goal- directed, thought and action. While EF is sometimes characterized as a single entity (Duncan, Emslie, Wil- liams, Johnson & Freer, 1996), most investigators have found that discrete sub-functions can be differentiated. One popular theoretical approach in this vein is that of Miyake and colleagues (Miyake, Friedman, Emerson, Witzki & Howerter, 2000) who, using confirmatory fac- tor analysis (CFA), isolated three EFs that are at the core of more complex EF tasks. These are: (1) Inhibition, the ability to deliberately suppress dominant, automatic, or prepotent responses when necessary; (2) Updating ⁄ working memory, the ability to revise and monitor rep- resentations in working memory, and (3) Shifting, the ability to switch attention flexibly between tasks, strate- gies, or mental sets. These three functions related dif- ferentially to more complex EF tasks that involve planning, abstract reasoning, and problem solving.

Though developed for adults, Miyake’s theoretical structure of EF has received some support in middle childhood. In one study of 8–13-year-olds, all three of Miyake’s factors were distinguishable (Lehto, Juujarvi, Kooistra & Pulkkinen, 2003). Other studies in this age

Address for correspondence: Susan A. Rose, Departments of Pediatrics and Psychiatry & Behavioral Sciences, Kennedy Center, Albert Einstein College of Medicine ⁄ Children’s Hospital at Montefiore, 1300 Morris Park Avenue, Bronx, NY 10461, USA; e-mail: susan.rose@einstein.yu.edu

� 2011 Blackwell Publishing Ltd, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA.

Developmental Science 14:5 (2011), pp 1161–1175 DOI: 10.1111/j.1467-7687.2011.01068.x

range found only two of Miyake’s factors (Huizinga, Dolan & van der Molen, 2006; van der Sluis, Jong & van der Leij, 2007). Van der Sluis distinguished working memory and inhibition, but not shifting, in 9–12-year- olds; Huizinga and colleagues found that a model having working memory and shifting, but not inhibition, as separable factors, fit multiple age groups (7-, 11-, 15-, and 21-year-olds). Some studies using other age groups found that only one factor was needed to summarize executive functions. For example, McCabe and col- leagues found a single factor structure using a lifespan sample of 18- to 90-year-olds tested on tasks of working memory and complex executive functions (McCabe, Roediger, McDaniel, Balota & Hambrick, 2010). Simi- larly, Wiebe and colleagues found that a structure with only one factor fit as well as a multi-factor structure in preschoolers tested on tasks of working memory and inhibition (Wiebe, Espy & Charak, 2008). In addition, progressive fractionation (differentiation) of EF con- structs across childhood has also been reported (Tsu- jimoto, Kuwajima & Sawaguchi, 2007). Although the majority of studies suggest that more than one factor is present by late childhood, it is not yet clear if all three of Miyake’s factors can be discerned by pre-adolescence.

Executive functions and academic achievement

Although all three of Miyake’s EFs have been linked to academic performance in the general population (Blair & Razza, 2007), studies are most extensive and the results most compelling for working memory, which has been linked to reading achievement in typically developing children (Alloway, Gathercole, Kirkwood & Elliott, 2009; Daneman & Carpenter, 1980; van der Sluis et al., 2007) and children with reading disabilities (Gathercole, Alloway, Willis & Adams, 2006; Siegel & Ryan, 1989). Working memory is also important for achievement in arithmetic and mathematical problem solving (Bull & Scerif, 2001; Passolunghi & Siegel, 2004; Siegel & Ryan, 1989; van der Sluis et al., 2007). One recent study found that over 80% of those who scored poorly in a test screening for working memory had problems in either reading or math, or most commonly, in both (Alloway et al., 2009).

There is evidence that inhibition and shifting may also be related to achievement, although studies of this issue are sparse and the results less clear cut. While some studies have found inhibition related to aspects of mathematical ability (Blair & Razza, 2007; Passolunghi & Siegel, 2001; St Clair-Thompson & Gathercole, 2006), others have failed to find relations between the two (Agostino, Johnson & Pascual-Leone, 2010; Swanson, 2006). The few studies examining the relation between shifting and math ability have had similarly mixed results (Agostino et al., 2010; Blair & Razza, 2007; Bull & Scerif, 2001; Lee, Ng & Ng, 2009), with these relations sometimes disappearing entirely when working memory and inhibition are controlled (Espy, McDiarmid, Cwik,

Stalets, Hamby & Senn, 2004). Inconsistencies across studies may be linked to the different tasks used to measure both the EF constructs and math ability, and with the tendency to use a single task to instantiate a construct.

Applicability to prematurity

Most EF studies in preterms have focused on the pre- school and early school years (for a listing of studies, see Mulder, Pitchford, Hagger & Marlow, 2009; Woodward, Edgin, Thompson & Inder, 2005); there are few data to indicate how pre-adolescent and older preterms might perform in this area. Given the continued rapid devel- opment through later childhood of the frontal cortical areas that underlie EF (Giedd, Blumenfeld, Jeffries, Castellanos, Liu, Zijdenbos, Paus, Evans & Rapoport, 1999; Lenroot & Giedd, 2006), and the different nature of the tasks used to assess early and later manifestations of EFs, more data are clearly needed.

The few studies targeting EFs in older preterms have been primarily concerned with complex problem solving, such as the Tower of Hanoi ⁄ Stockings of Cambridge (Anderson, Doyle & Group, 2004; Taylor, Minich, Bangert, Filipek & Hack, 2004), and verbal fluency (Nosarti, Shergill, Allin, Walshe, Rifkin, Murray & McGuire, 2009; Rushe, Rifkin, Stewart, Townsend, Roth, Wyatt & Murray, 2001). Only a handful have tar- geted the three EFs identified by Miyake – working memory, inhibition, shifting – and the findings have been mixed. For example, while deficits in some aspects of working memory have been found (Anderson et al., 2004; Curtis, Lindeke, Georgieff & Nelson, 2002; Luciana, Lindeke, Georgieff, Mills & Nelson, 1999; Taylor et al., 2004), they appear to abate over time (Curtis et al., 2002). Even fewer studies have examined inhibition or shifting in preterm pre-adolescents (Bayless & Stevenson, 2007; Elgen, Lundervold & Sommerfelt, 2004; Kulseng, Jennekens-Schinkel, Naess, Romundstad, Indredavik, Vik & Brubakk, 2006; Taylor, Klein, Drotar, Schluchter & Hack, 2006). Here too, the findings are mixed (Bayless & Stevenson, 2007; Luciana et al., 1999), and inconsis- tencies across studies may be linked to differences in the tasks and measurement strategy used.

Processing speed

It has been proposed that performance on complex EF tasks depends on more fundamental abilities. One pop- ular theory holds that processing speed is a key cognitive resource underlying performance in a wide range of cognitive domains, including EF (Kail & Salthouse, 1994). Support for such a ‘common cause’ model comes from work with both adults and children that has focused largely on working memory. Age-related increases in processing speed during childhood (Kail, 1991) have been found to explain age-related improvement in working memory (Fry & Hale, 1996; Hale, 1990; Kail &

1162 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

Salthouse, 1994), and age-related decreases in processing speed during old age have explained the age-related declines seen later in life (Salthouse, 2005). Comparable relations between processing speed, shifting and inhibi- tion have been reported, but these data are limited (Christ, White, Brunstrom & Abrams, 2003; Salthouse, Fristoe, McGuthry & Hambrick, 1998).

There is behavioral and neuropsychological evidence indicating that preterms have deficits in processing speed. In infancy and toddlerhood, preterms have been found to take longer than full-term controls to encode stimuli (Rose, Feldman & Jankowski, 2002, 2009); similar defi- cits have been found among preterms in pre-adolescence (Rose & Feldman, 1996, 1997). Neurophysiological evi- dence indicates that processing speed is dependent on the integrity of white matter tracts (Kennedy & Raz, 2009; Turken, Whitfield-Gabrieli, Bammer, Baldo, Dronkers & Gabrieli, 2008). These tracts are often compromised in preterms, with recent imaging studies showing reductions in white matter concentration and volume, and wide- spread fiber tract disorganization (Constable, Ment, Vohr, Kesler, Fulbright, Lacadie, Delancy, Katz, Schneider, Schafer, Makuch & Reiss, 2008; Gim�nez, Junqu�, Narberhaus, Bargall�, Botet & Mercader, 2006). Moreover, a direct link has recently been found in a preterm sample between reductions in white matter integrity and slower processing speed (Soria-Pastor, Gim�nez, Narberhaus, Falcon, Botet, Bargall�, Mercader & Junqu�, 2008).

Present study

The present study examines EF and processing speed in pre-adolescent preterms and models their relation to academic achievement. It addresses four major ques- tions. First, do preterms of this age have deficits, relative to full-term controls, in processing speed and in the three EFs identified by Miyake (working memory, inhibition, shifting)? Second, are the three EFs identified by Miyake distinguishable from one another, and separable from processing speed? Confirmatory factor analysis (CFA) and latent variable analysis are used to examine this issue and to test the model’s applicability to a risk population. In effect, we are testing an extension of Miyake’s model (including speed in addition to EFs) and examining its applicability to a risk population. Third, is processing speed a driving force for the three EFs? Fourth, do the three EFs, together with processing speed, mediate pre- term ⁄ full-term differences in academic achievement? To examine these last two questions, we test a model based on Salthouse’s theory (Salthouse, 1996, 2000, 2005), in which a common cause – processing speed – drives individual differences in EFs. Our model posits that processing speed and EFs will mediate preterm ⁄ full-term differences in academic achievement in a cascade of ef- fects as follows: prematurity fi slower processing speed fi poorer executive functioning fi lower academic achievement.

Methods

Participants

Participants were preterm and full-term 11-year-olds who were enrolled in a prospective, longitudinal study of cognitive development. At testing, the average age of the preterms was 11.18 years (SD = .44; range = 10.4–12.1); the average age of the full-terms was 11.14 years (SD = .35; range = 10.5–12.5).

Original sample

The original sample included 203 children (59 preterm infants and 144 term controls), born between February 1995 and July 1997. While the original goal was to have twice as many full-term controls as preterms, full-terms were over-recruited in the expectation that they would be more difficult to retain over time, since the preterms received their medical care in an associated program. Preterm infants were recruited from consecutive births admitted to the neonatal intensive care units of two hos- pitals affiliated with Albert Einstein College of Medicine. Criteria for study intake were: singleton birth, birth weight < 1750 g, gestational age < 37 weeks, and the absence of any obvious congenital, physical, or neuro- logical abnormalities. Term infants were recruited from consecutive births from the same hospitals; criteria for study intake were birth weight > 2500 g, gestational age of 38–42 weeks, 5-minute Apgar scores of 9 or 10, and uneventful pre- and perinatal circumstances. The present study focuses on the 11-year follow-up of this sample. (Previous follow-ups were at 5, 7, 12, 24, and 36 months.)

Attrition

Of the original 203 children, 134 returned at 11 years, for a follow-up rate of 74.9% for preterms (N = 44) and 62.5% for full-terms (N = 90). The attrition would not appear unreasonable in view of the 8-year hiatus since the last visit. Reasons for loss to follow-up included (1) families now living out-of-state (five preterms; 19 full- terms); (2) inability to locate families that had moved (eight preterms, 27 full-terms), and refusal to participate (two preterms, five full-terms). Data for children who developed serious neurological conditions since their last follow-up at age 3 were excluded (N = 3 full-terms).

Background and medical characteristics

Background factors for the preterms and full-terms returning at 11 years (shown in Table 1) are similar for the two groups. Overall, the sample was evenly divided between males and females, about one-third were first born, and the majority were Black or Hispanic. Maternal education averaged 13.6 years (SD = 2.1) and SES, as assessed with the Hollingshead Four-Factor Index

Executive function in preterm adolescents 1163

� 2011 Blackwell Publishing Ltd.

(Hollingshead, 1975), averaged 36.6 (SD = 13.0). The distribution of these factors within groups is nearly identical to that of the original sample (Rose, Feldman & Jankowski, 2001). The medical risk factors for preterms returning at 11 years (shown in Table 2) are also similar to those of the original cohort (Rose et al., 2001).

Procedure

The children were seen at 11 years for a full day of testing. In general, the morning session ran from approximately 9:00am to 12:30pm, with one 20-minute break; lunch followed. The afternoon session ran from 1:15pm to 4:00pm, with one 20-minute break. Additional breaks were given if needed. The tasks presented here, which are part of a larger battery, were designed to assess (a) the three components of executive functioning – Working Memory, Inhibition, and Shifting, (b) processing speed, and (c) reading and mathematical achievement. We selected tasks that were well established and had been successfully used in prior research with children in this age range. Most were non-verbal, though two verbal tasks of working memory – listening span and counting span – were included because of previous literature suggesting that they might have differential relations to reading and math (Jarvis & Gathercole, 2003). With the exception of trail-making (see below), the measures from tasks assessing EF emphasized accuracy over speed (see Mulder et al., 2009, for a discussion of this issue).

The battery included both computerized and paper- and-pencil tasks.1 The computerized tasks were drawn

from the Cambridge Neuropsychological Testing Auto- mated Battery (CANTAB; Cambridge Cognition, 2005), a well-standardized battery of non-verbal tasks where responses are recorded via a touch screen; from the Cognitive Abilities Test, which also uses a touch screen (CAT; Detterman, 1988); or created in-house, using E-Prime, with responses recorded via computer key presses. The reliabilities of tasks from the CANTAB and CAT are good, with internal consistency coefficients on the CANTAB tasks ranging from .73 to .95 for 4- to 12-year-olds (Luciana & Nelson, 2002) and internal consistency and split-half reliabilities on the CAT gen- erally in the .80s or above (DeFries & Plomin, 1985; Detterman, 1988). The tasks created in E-Prime (counting span, go ⁄ no-go) followed well-established protocols; internal consistency and split-half correlations for these tasks are in the range of .70–.90 (Conway, Kane, Bunting, Hambrick, Wilhelm & Engle, 2005; Engle, Tuholski, Laughlin & Conway, 1999; Waters & Caplan, 2003). Many of the tasks used in the present study are graded in difficulty, thus minimizing floor and ceiling effects. The paper-and-pencil tasks included trail-mak- ing, a measure of shifting, and tests of math and reading from the Woodcock-Johnson III (WJ III) Tests of Achievement (Woodcock, McGrew & Mather, 1999). The WJ III is a standardized battery with high reliability and validity; standard scores are available for all subtests.

The number of children completing particular tasks varied from 125 to 131; data loss was due to equipment failure, parental time constraints, or failure to complete the task. All computerized tasks began with practice trials. The tasks, and measures included in the present study, are described below.

Measures of executive function

Working memory

Spatial working memory (CANTAB), modeled after Petrides and Milner’s self-ordered pointing task (1982),

Table 1 Demographic characteristics of the sample at 11 years

Variable

Full-terms (n = 87)

Preterms (n = 44)

X2% %

Gender –% male 48.3 56.8 .85 Birth order – % first born 35.6 43.2 .71 Ethnic composition

Black 41.4 52.3 Hispanic 42.5 36.4 White 16.1 11.4 1.50

M SD M SD t

Maternal age 31.1 6.2 30.6 6.3 .46 Mother’s education (years) 13.7 2.0 13.3 2.5 .88 Father’s education (years)a 13.5 2.2 13.1 2.6 .92 SES (Hollingshead Four-Factor Index)

37.8 12.2 34.4 14.4 1.45

Note:. SES = socioeconomic status. None of the preterm ⁄ full-term differences were significant.

Table 2 Medical characteristics of the preterms (N = 44)

Mean SD Range %

Birth weight (g) 1165.2 268.4 751–1747 Gestational age at birth (weeks) 29.7 2.8 25–36 Apgar

1 minute 5.7 2.5 1–9 5 minute 7.5 1.5 4–9

Time on respirator (days) 6.1 8.0 0–31 Time on oxygen (days) 18.9 21.3 0–82 Time in hospital (days) 49.9 22.5 7–106 Neonatal Medical Index (NMI)a 3.6 1.4 1–5 Small for Gestational Age (SGA)b 27.3 Respiratory Distress Syndrome (RDS) 50.0 Ultrasound diagnosis

No abnormality 56.8 Grade I or II Hemorrhage 43.2 Grade III or IV Hemorrhage 0.0

a Higher scores indicate more severe complications. b Defined as having a birth weight below 10% of the population for a given gestational age.

1 Data for two additional tasks included in the battery are not pre- sented. One, a non-verbal task of pattern working memory, modeled after the change-detection task (Luck & Vogel, 1997), was excluded because there is some question as to whether it requires the same level of concurrent processing as do the other working memory tasks and it did not correlate with them. Data from the other, a Stroop task (assessing Inhibition), was not saved due to software problems.

1164 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

assesses the child’s ability to retain spatial information and manipulate remembered items in working memory. The test begins with three colored boxes displayed on the computer screen. By touching them and using a process of elimination, the child finds one blue ‘token’ under each, which is used to fill an empty column on the right- hand side of the screen. On each trial, once a token has been removed from a box, that box is empty; a return search to it constitutes an error. The number of boxes is gradually increased, with the child receiving four trials each with 4, 6, and 8 boxes. Measure: number of errors (searching the same box more than once).

Listening span (E-Prime), modeled after Daneman and Carpenter (1980), requires retaining in memory, and updating, a list consisting of the last words of a series of sentences. After listening to each sentence (e.g. ‘Turtles run fast’) the child judges its veracity and then repeats the final word. When all sentences in a set have been presented, the child must recall, in order, the final words of each sentence. The test begins with a set of two sen- tences; the number of sentences in a set is increased by one every three trials, to a maximum of six. Testing continues until the child errs on all three trials of a set. Measure: A partial-credit unit weighted score, reflecting the average proportion of items correctly recalled per set size (following Conway et al., 2005).

Counting span (E-Prime), modeled after Case, Kurland and Goldberg (1982), requires retaining in memory counts of the red circles found in a series of arrays containing three to nine randomly placed red circles (along with 2 to 14 distracters – blue circles and blue squares). Squares are 2.5 mm on a side, circles 2.5 mm in diameter. The child must count aloud the number of red circles in each array and repeat the total. When all the arrays in a set have been presented, the child must recall, in order, the final tallies. The test begins with sets con- taining two arrays; the number of arrays in each set is increased by one every three trials, to a maximum of five. Measure: A partial-credit unit weighted score, reflecting the average proportion of items correctly recalled per set size (following Conway et al., 2005).

Inhibition2

Go ⁄ No-go (E-Prime), adapted from Casey, Trainor, Orendi, Schubert, Nystrom and Giedd (1997), assesses the child’s ability to withhold or suppress a prepotent response. There are two types of trials: go-trials, which require the child to press a computer key as quickly as possible when any eight of nine different Looney Tune

cartoon characters are displayed on the computer screen, and no-go trials, which require withholding a response when the remaining cartoon character (‘Bugs Bunny’) appears. Stimuli were presented sequentially, for 500 ms each (ISI = 1500 ms); the degree of inhibitory control required is increased by increasing the number of ‘go’ trials (1, 3, or 5) preceding ‘no-go’ trials. Foil trials (no-go trials after 2 or 4 go trials) were included to prevent learning of the pattern, but these trials were not included in the analysis. There were 228 trials in all (172 ‘go’ trials; 56 ‘no-go’ trials), presented in four blocks of 57 trials. Each block contained 48 valid trials (16 each with 1, 3, or 5 preceding go-trials), and eight foil trials. Measure: number of correct no-go trials.

Rapid visual information processing (CANTAB) is a continuous performance task in which the child moni- tors a stream of digits for a particular target sequence (3-5-7). The target sequence is displayed on the screen, in the upper right hand corner, serving as a reminder throughout the task. Single digits, ranging from 2 to 9, appear in a pseudo-random order, at the rate of 100 ⁄ min. The child registers a response on a press pad after the last digit of the target sequence appears (withholding responses to all other numbers). There are 300 trials, containing 24 target sequences. False alarms (errors of commission), one of several measures avail- able from this task, are considered a reliable index of inhibition (Mulder et al., 2009). Measure: number of false alarms.

Shifting

Trail-making, a paper-and-pencil test (Reitan & Wolfson, 1992), requires the child to connect the dots of 25 con- secutive targets. It has two parts. In Part A, a baseline assessment of motor speed, the targets are all numbers (1, 2, 3, etc.). In Part B, a measure of cognitive flexibility, the targets alternate between numbers and letters (1, A, 2, B, 3, C, etc.). The goal of the subject is to finish the test as quickly as possible. Measure: difference in time to complete Parts A and B.

Intra-dimensional ⁄ extra-dimensional shift (CANTAB), derived from the Wisconsin Card Sorting Task, assesses rule acquisition and reversal. The test uses two artificial dimensions: color-filled shapes and white lines. Simple stimuli are made up of just one of these dimensions, whereas compound stimuli are made up of both, namely white lines overlying color-filled shapes. The child starts by seeing two simple color-filled shapes, and must learn which one is correct by touching it. Feed- back teaches the child which stimulus is correct, and after six correct responses, the stimuli and ⁄ or rules are changed. These shifts are initially intra-dimensional (ID) shifts (e.g. color filled shapes remain the only relevant dimension), then later extra-dimensional (ED) shifts (white lines become the only relevant dimension). For both types of shifts, some blocks of trials (or ‘stages’) involve reversing the previously ‘correct’ rule.

2 While several subtypes of inhibition have been distinguished (Fried- man & Miyake, 2004; Nigg, 2000), the present study focuses on the one that requires deliberate suppression of a prepotent behavioral response, often considered the most straightforwardly associated with executive function (Friedman & Miyake, 2004). Other subtypes, such as resis- tance to proactive interference and resistance to distracters, were not assessed here.

Executive function in preterm adolescents 1165

� 2011 Blackwell Publishing Ltd.

These are the trials of interest here. Progress through the test entails satisfying a set criterion of learning at each stage (six consecutive correct responses). If the subject fails to reach this criterion after 50 trials in any stage, the test terminates. Measure: total number of errors on all reversal blocks.

Measures of processing speed

Tachistoscopic threshold (CAT), a measure of inspection time, assesses the child’s rapidity at encoding simple information. Two patterns are presented briefly and the child must indicate, by a key press, whether the two are the same or different. Trials begin with a 1 ⁄ 60 s exposure and, using the ascending method of limits, exposure time is incremented by 1 ⁄ 60 s after an error, and reduced by 1 ⁄ 60 s after a correct response. Presentations continue until the child makes five consecutive correct responses. The exposure time for these five trials is taken as the threshold for the block. There are 20 such blocks of trials. Measure: Median threshold (in msec) over the 20 blocks.

Reaction time (CAT) is a task of simple and choice RT. The test begins with one window displayed on the screen and the child is to touch it as quickly as possible when it lights up. The number of boxes is gradually increased, so that 2, 4, 6, or 8 are displayed at one time, only one of which lights up. Variable delays are used (200, 300, 400 ms) and there are 24 trials at each set size (1, 2, 4, 6, 8), for a total of 120 trials. RT (decision time) is mea- sured by the interval between stimulus onset and response initiation (lifting finger from a ‘home bar’ on the touch screen). Measure: RT.

Measures of academic achievement

Reading

Letter-word identification (WJ III) assesses reading decoding by requiring the child to read aloud a list of words of increasing difficulty. Measure: number correct (standardized score).

Reading fluency (WJ III) assesses the ability to quickly read simple sentences and decide if the statement is true or false; the difficulty of the sentences increases gradu- ally. This test has a 3-min time limit. Measure: number correct (standardized score).

Mathematics

Math fluency (WJ III) assesses the ability to solve simple addition, subtraction, and multiplication problems. This test has a 3-min time limit. Measure: number correct (standardized score).

Applied problems (WJ III) assesses the child’s ability to solve problems that require increasingly difficult mathe- matical operations. Measure: number correct (standard- ized score).

Data analysis

Univariate and bivariate distributions were initially examined for all variables, separately by group, and outlying values, > 2.5 SD from the mean or regression line, were removed (four data points for preterms, all from rapid visual processing; three data points for full- terms, two from spatial working memory and one from trail-making). For each domain, a multivariate test (MANOVA) was used to determine whether the two groups differed, with follow-up ANOVAs to determine which tasks differed between the groups. In preliminary analyses, tasks having multiple levels of difficulty (e.g. go ⁄ no-go, RT task), were analyzed with repeated mea- sures ANOVAs to determine if there were interactions of difficulty level with group. There were none. Supple- mentary analyses were carried out (1) using gender as a second grouping variable, and (2) removing children with IQs < 70 (N = 5 preterms). There were no gender effects and none of the results were altered by excluding the five children with IQ < 70.

Confirmatory factor analysis (CFA) and path analyses were accomplished with structural equation modeling (SEM) using LISREL (Ver 8.54: Jçreskog & Sçrbom, 2003) with maximum likelihood estimation. Missing data (1.4% for the CFA, 1.5% for the path models) were imputed using the Expected Maximization (EM) algo- rithm in PRELIS. Model fit was assessed using the normal theory weighted least squares v2, the root mean square error of approximation (RMSEA), the compara- tive fit index (CFI), and the non-normed fit index (NNFI). Values indicative of good fit are a non-signifi- cant v2 at p > .05, a RMSEA < .05, and a CFI and NNFI > .90; lower values of Akaike’s consistent infor- mation criterion (CAIC) indicate better fit. Nested models were compared with one another using the chi- square difference test, which subtracts the chi-square for the full model from a nested, restricted model with fewer free parameters. A significant difference indicates that the fuller model provides a significantly better fit.

A four-factor CFA model was evaluated which posited three EFs (working memory, inhibition, shifting), and a separate speed factor. Its fit was compared with a series of alternative models positing fewer latent factors. In addition, a series of two-group CFA models were fit to the data to determine whether the basic four-factor model fit equally well for full-terms and preterms. The two-group models differed in the number of equality constraints that were imposed between preterms and full- terms. The first model constrained only factor loadings to be equal across groups; the second additionally con- strained to equality the variances of and covariances among latent variables; and the third model additionally constrained error variances to be equal.

The path model tested the hypothesis that pre- term ⁄ full-term differences in academic achievement could be explained by a cascade of effects involving speed and the three EFs. Specifically, the following was pro-

1166 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

posed: prematurity fi slower processing speed fi poorer executive functioning fi lower academic achievement.

Results

Preterm ⁄ full-term differences

The first set of analyses sought to determine which domains of functioning were compromised by preterm birth. Multivariate ANOVA revealed that there were significant or marginally significant group differences in all domains; working memory, Hotelling’s T = .08, F(3, 122) = 3.26, p = .024, g2 = .07; inhibition, Hotelling’s T = .044, F(2, 125) = 2.74, p = .07, g2 = .04; shifting, Hotelling’s T = .116, F(2, 125) = 7.27, p = .001, g2 = .10, processing speed, Hotelling’s T = .11, F(2, 127) = 7.05, p = .001, g2 = .10, and academic achievement, Hotell- ing’s T = .087, F(4, 118) = 2.56, p = .042, g2 = .08. Univariate tests, along with means and standard devia- tions, are shown in Table 3.

Working memory

Preterms made significantly more errors than full-terms on Spatial Working Memory, and on Counting Span; the groups did not differ on Listening span.3

Inhibition

The two groups performed similarly on Go ⁄ No-go, which was difficult for both groups, with children failing to

inhibit responses on about 40% of the trials. However, preterms did not perform as well as full-terms on Rapid visual information processing, having significantly more false alarms (errors of commission).

Shifting

Preterms needed significantly more time than full-terms to connect points when alternating between numbers and letters, on Part B of the Trail-making task, even when controlling for motor speed on Part A. (Similar results were found using an ANCOVA, with performance on Part B covaried for performance on Part A, F(1, 126) = 7.29, p < .01, g2 = .06.)4 Preterms also made significantly more errors on reversal trials on Intra- dimensional ⁄ extra-dimensional shift.

Processing speed

Tachistoscopic thresholds were significantly higher for preterms, who needed an exposure time about 40% longer than full-terms to determine whether two briefly presented stimuli were the same or different. Reaction times were marginally slower, with preterms taking around 50 ms longer than full-terms to respond.

Academic achievement

Reading. The preterms had significantly poorer perfor- mance on the test of Reading fluency, which requires

Table 3 Preterm ⁄ full-term differences on tasks of executive functioning, processing speed, and achievement

Domains ⁄ Measures

Preterm Full-term

F-tests for birth status g2N M SD N M SD

EXECUTIVE FUNCTION Working memory

Spatial working memory – # errors 43 44.12 15.40 85 38.20 17.59 F(1,126) = 3.68, p = .05 .03 Listening Span – partial credit unit scoring 44 .35 .15 87 .36 .13 F(1,129) = .45, p = .50 .00 Counting Span – partial credit unit scoring 44 .59 .23 85 .70 .27 F(1,127) = 6.22, p < .05 .05

Inhibition Go ⁄ No-go – % correct no-go trials 43 56.83 16.72 87 60.51 15.45 F(1,128) = 1.57, p = .22 .02 Rapid visual processing – # false alarms 43 6.02 11.09 86 3.30 4.22 F(1,127) = 4.11, p < .05 .03

Shifting Trail-making – time to compete Part B-A 43 73.58 65.16 85 42.80 25.58 F(1,126) = 14.61, p < .001 .10 ID ⁄ Ed Shift – # errors on reversal learning 43 6.74 6.01 87 4.46 1.93 F(1,128) = 10.46, p < .01 .08

PROCESSING SPEED Tachistoscopic threshold – Median (ms) 43 167.07 79.94 87 121.57 56.22 F(1,128) = 14.11, p < .001 .10 Reaction time – Mean decision time (ms) 43 580.65 139.48 87 540.32 101.97 F(1,128) = 3.50, p < .10 .03

ACHIEVEMENT Reading

Letter-word identification – standard score 44 97.95 14.54 86 100.60 9.76 F(1,128) = 1.52, p = .22 .01 Fluency – standard score 44 95.89 14.88 84 101.40 13.01 F(1,128) = 4.70, p < .05 .04

Math Fluency – standard score 43 93.09 14.91 84 96.39 14.51 F(1,125) = 1.44, p = .23 .01 Applied problems – standard score 42 97.50 13.00 84 103.69 10.00 F(1,128) = 8.75, p < .01 .07

3 Results for the two span tasks scored for span length (maximum number of words ⁄ numbers recalled in correct serial order) are as fol- lows: Listening span, M = 2.66 (SD = .96) for preterms and M = 2.60 (SD = .98) for full-terms, F(1, 129) < 1, ns; for Counting span the corresponding scores are M = 3.43 (SD = 1.21) and M = 4.05 (SD = 1.00), F(1, 127) = 9.51, p < .01.

4 Trail-making has a relatively large motoric requirement due to the distributed arrangement of the targets. Part A was subtracted from Part B to avoid confounding preterm ⁄ full-term differences in shifting with differences in motor speed. (Motor speed was not involved in the measures of processing speed.)

Executive function in preterm adolescents 1167

� 2011 Blackwell Publishing Ltd.

sentence comprehension, but not Letter-word identifica- tion, which requires only word decoding.

Mathematics. The preterms performed significantly worse than the full-terms on Applied problems, where they had to solve word problems involving mathematical reasoning, but not in Math fluency, which requires only simple addition, subtraction, and multiplication.

Confirmatory factor analysis: EF and processing speed

The CFA model tested whether the three EF constructs posited by Miyake (working memory, inhibition, shifting) were distinct from one another, and whether processing speed could be viewed as a separate construct. Correla- tions among the measures used in the CFA are shown in Table 4. Because fewer than 5% of the correlations dif- fered significantly between preterms and full-terms, the data are combined across groups and the correlations are partialed for birth status (to avoid their being inflated by group mean differences).5 Two points are worth noting about these results. First, the two measures of processing speed (tachistoscopic threshold and reaction time) were, as anticipated, related to most measures of EF. Second, processing speed, along with the three EF constructs, was related to reading fluency and applied math.

The results of the CFA are presented for the two groups combined (in view of the fact that the metric model did not differ across groups; see below). The basic four-factor model is shown in Table 5, with factor

loadings of each task in the top portion, and correlations among factors in the bottom. The standardized factor loadings ranged from .35 to .87; all were significant (p < .05). Correlations among factors ranged from .24 to .78; the three EFs themselves were moderately inter- correlated (.24 to .66). Fit statistics, shown in Table 6 (top line, four-factor model), indicate that the model provides an excellent fit to the data. These results sup- port Miyake’s conceptualization, with the three EFs forming distinct, though correlated, latent factors.

This model fit better than all the nested alternatives shown in Table 6, each of which specified fewer latent factors. The three-factor alternatives were created by collapsing each possible pair of executive functions into a single latent variable; the two-factor alternative was created by collapsing all three EFs into one; the one- factor alternative was created by collapsing speed along with the three EFs into a single latent factor. As indi- cated by the significance of the v2 difference tests shown in Table 6, the fit of the four-factor model was superior to all these alternatives. In sum, these results support the four-factor model as a good characterization of the data.

Another series of model comparisons was used to determine whether the four-factor model fit equally well for preterms and full-terms. Here we constructed a series of two-group models, sequentially introducing equality constraints on factor loadings, variances and covariances among latent variables, and error variances of the man- ifest variables. Model 1, which constrained only factor loadings, fit both groups well: v2 (52) = 34.13, p = .97, RMSEA = .00, CFI = 1.00, NNFI = 1.08. Model 2, which placed equality constraints on both factor load- ings, and the variances of the latent variables and covariation among them, also fit the data well, v2

(58) = 35.59, p = .99, RMSEA = .00, CFI = 1.00, NNFI = 1.08, and fit as well as Model 1, D v2 (6) = 1.46, p > .05. Model 3, which provides the most stringent test

Table 4 Correlations among measures (partialed for prematurity)

Domains ⁄ Measures1 1 2 3 4 5 6 7 8 9 10 11 12 13

Working memory 1 Spatial working memory 2 Counting span .25 3 Listening span .37 .39

Inhibition 4 Go ⁄ No-go (no-go trials) .17 .10 .16 5 Rapid visual processing .24 .23 .11 .37

Shifting 6 Trail-making (Time B – A) .22 .26 .22 .04 .13 7 ID ⁄ ED shifts (reversal trials) .11 .17 .10 ).01 .10 .36

Processing speed 8 Tachistoscopic threshold .18 .17 .22 .31 .38 .21 ).01 9 Reaction time .17 .18 .24 .24 .34 .22 .08 .42

Academic achievement 10 Reading: Letter-word .16 .24 .23 .25 .30 .01 ).04 .27 .12 11 Reading: Fluency .31 .24 .38 .24 .35 .25 .06 .32 .33 .62 12 Math: Fluency .19 .28 .20 .35 .21 .26 .05 .24 .14 .45 .54 13 Math: Applied problems .32 .34 .41 .15 .21 .36 .12 .28 .24 .56 .47 .54

Note: N = 124–130. Correlations in bold are significant at p < .05. 1 All measures are scaled so that high scores indicate good performance.

5 Originally, we computed correlations among measures within each group separately, and tallied the number of correlations that differed significantly across groups (a = .05). By chance, then, one would have expected 5% of the correlations to differ significantly (n � 5 of 91 correlations [(142 – 14) ⁄ 2]). The actual number differing across groups was four (4%).

1168 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

of metric invariance, imposes additional equality con- straints, so that there were now constraints on factor loadings, variances of and covariances among factors, and on the error variances. This model also fit well, v2

(66) = 43.49, p = .98, RMSEA = .00, CFI = 1.00, NNFI = 1.08; and constraining these paths to be equal did not harm model fit, Dv2 (8) = 12.10, p > .05. Given that the imposition of constraints did not result in any degradation of the model, we can conclude that the basic four-factor measurement model fits well for both groups.

Path model: mediation of preterm ⁄ full-term differences in academic achievement

We next used SEM to test the hypothesis that these specific latent factors would account for preterm ⁄ full- term differences in academic achievement. The model, portrayed in Figure 1, assumes that one latent factor (processing speed) contributes to three other latent

factors (working memory, inhibition, and shifting), which, in turn, contribute to mathematics and reading.

Latent variables are defined as in the CFA. Birth status (coded ‘0’ for preterms and ‘1’ for full-terms) is a single- indicator variable, as are reading and math. The tests representing reading (fluency) and math (applied prob- lems) are those for which there were preterm ⁄ full-term differences.

The model yielded an excellent fit to the data, v2(36, N = 131) = 23.21, p = .95, RMSEA = .00, CFI = 1.00, NNFI = 1.04; the model CAIC = 269.97. Several points about the model are noteworthy. First, as shown by the path coefficient from birth status to processing speed (b = .38), preterms are considerably slower than their full-term peers. Second, consistent with the key assumption that speed of processing underlies perfor- mance on many measures of EF, there were strong paths between processing speed and the three EFs (b ranged from .41 to .84). Third, there was also a significant direct

Table 5 Confirmatory factor analysis

Measures1

Latent factors

Working memory Inhibition Shifting Processing speed

Factor loadings (standardized)

Spatial working memory .60* Counting span .57* Listening span .60* Go ⁄ No-go (no-go trials) .57* Rapid visual processing .68* Trail-making (Time B – A) .87* ID ⁄ ED shifts (reversal trials) .35* Tachistoscopic threshold .72* Reaction time .62*

Interfactor correlations

Working Memory – Inhibiting .48* – Shifting .66* .24� – Processing Speed .56* .78* .54* –

1 All measures are scaled so that high scores indicate good performance. � p < .10; *p < .001.

Table 6 Goodness of fit indices for CFA models and comparisons of the four-factor model with alternatives

Models

Fit indices

Comparisons of 4-factor model with alternatives

V2 df p RMSEA CFI NNFI CAIC Dv2 Ddf

Four factors: working memory, inhibition, shifting, speed 16.02 21 .77 .00 1.00 1.04 157.02 Alternatives

Three factors: collapsing working memory & inhibition 51.66 24 .001 .09 .91 .87 175.04 35.64** 3 Three factors: collapsing working memory & shifting 28.54 24 .24 .04 .98 .97 151.92 12.52** 3 Three factors: collapsing inhibition & shifting 44.39 24 .01 .08 .93 .90 167.77 28.37** 3 Two factors: collapsing working memory, inhibition & shifting 63.07 26 .00 .11 .87 .82 174.70 49.05** 5 One factor: collapsing all four latent variables 73.57 27 .00 .12 .85 .79 179.32 59.55** 6

Note: The endorsed model is in bold. *p < .05; **p < .01.

Executive function in preterm adolescents 1169

� 2011 Blackwell Publishing Ltd.

effect of birth status to shifting (b = .38), indicating that preterms’ problems in cognitive flexibility are not entirely due to processing speed. Fourth, of the three EFs, only working memory had an independent effect on achieve- ment, with similar effects on math (b = .38) and reading (b = .28). Fifth, processing speed and EF completely mediated the relation of birth status to reading (b = .03) and math (b = .01). Taken together, it is clear that the relation between prematurity and academic difficulties can be explained by the following cascade of effects: prematurity fi slower processing speed fi poorer exec- utive functioning (working memory) fi lower achieve- ment in math and reading.

We also examined a theoretically plausible alternative model in which speed and the three EFs were on the same level, with all four serving as direct predictors of academic achievement. In this model, as in the original, correlations were allowed among the three EFs. In essence, this model differs from that portrayed in Figure 1 in that the paths between processing speed and the three EFs are absent. This model did not fit, v2 (39, N = 131) = 55.88, p < .05, RMSEA = .06, CFI = .96, NNFI = .93; the model CAIC = 285.01. A chi-square difference test indicated

that the model depicted in Figure 1 was superior to this alternative, Dv2 (3) = 32.57, p < .001.

Discussion

This study examined processing speed and EFs in preterm and full-term children and their relation to academic achievement. There are three main findings. First, preterm deficits at 11 years were found in all three components of EF identified by Miyake as critical for more complex thought – updating ⁄ working memory, inhibition, and attentional shifting. Second, a CFA validated their struc- ture as three quasi-independent latent factors, with processing speed forming a fourth quasi-independent la- tent factor. The invariance of the CFA model across groups suggests that the underlying architecture of EF is the same in preterms and full-terms. Third, using SEM, we showed that the negative effects of prematurity on reading and math were completely mediated by processing speed and EF. The model posited a cascade of effects, in which prematurity influenced processing speed, which then influenced EF, which, in turn, influenced achievement. Of

Figure 1 Structural equation model of a cascade of effects from birth status to achievement (reading and math) at 11 years via processing speed and three executive functions. Ovals represent multiple-indicator latent variables; rectangles represent single- indicator latents. Birth status is coded in a binary manner, ‘1’ for full-terms and ‘0’ for preterms. Single-headed arrows represent paths by which latent variables influence each other; double-headed arrows represent correlations. Solid lines indicate significant path coefficients (p < .05), while dotted lines represent non-significant coecients. Parameter estimates are shown for the completely standardized solution.

1170 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

the three EFs, working memory influenced math and reading independently of inhibition and shifting. In effect then, the cascade was as follows: prematurity fi slower processing speed fi poorer executive functioning (working memory) fi lower achievement in math and reading.

Preterm deficits

The deficits found in EF are in agreement with a recent meta-analysis that concluded that this was an area of weakness in preterms (Mulder et al., 2009). On working memory, preterms had more difficulty than full-terms remembering where they had previously searched (spatial working memory), replicating earlier findings (Curtis et al., 2002; Luciana et al., 1999; Taylor et al., 2004). They also had more difficulty in keeping an increasingly long sequence of tallies in mind on counting span, a task which, to our knowledge, has not previously been used with preterms. On shifting, they took longer to complete trail-making, even controlling for motor speed (see also Nosarti, Giouroukou, Healy, Rifkin, Walshe, Reichen- berg, Chitnis, Williams & Murray, 2008), and made more reversal learning errors on the ID ⁄ ED task. Although previous studies have not found preterm difficulties in this latter task (Curtis et al., 2002; Taylor et al., 2004), none examined performance on reversal trials. These trials require shifting from a reinforced stimulus to a previously non-reinforced one, and may entail more cognitive flexibility than many of the other measures from the task. On inhibition, preterms had more false alarms than full-terms on rapid visual processing. These findings are consistent with those of Taylor et al. (2004). While there were no preterm ⁄ full-term differences on the Go ⁄ No-go task, both groups showed higher error rates (44.6% for preterms; 39.5% for full-terms) than is gen- erally found (10%–15%; Durston, Thomas, Yang, Ulug, Zimmerman & Casey, 2002). The relatively short inter- trial intervals used in the present study (1500 ms com- pared with 3500 ms in the Durston et al. study) may have increased the pace of the task, making it more difficult for children in both groups to withhold responses.

Preterm deficits were also found in speed and aca- demic performance. The preterms’ longer latencies on the RT task and higher tachistoscopic thresholds are consistent with earlier findings in this age group on these same tasks (Rose & Feldman, 1996), and with recent findings in preterms who have reached adulthood (Strang-Karlsson, Andersson, Paile-Hyvarinen, Darby, Hovi, Raikkonen, Heinon, Jarvenpaa, Erkisson & Kaj- antie, 2010). Preterm academic difficulties were evident here on the more challenging reading and math tasks. They had difficulty with reading fluency, which involved sentence comprehension, but not with letter-word iden- tification, which entailed simple word decoding; simi- larly, they had difficulty with applied problems, which involved reasoning, but not with math fluency, which entailed only simple arithmetic.

The CFA model: EF and speed

One of the strengths of the present study was the use of CFA to define the factor structure of EF. The use of CFA is a rarity in preterm studies, where constructs are often operationalized by a single task, and overall, a rarity with preadolescent children. Because latent variables extract the common variance that is shared among multiple tasks, minimizing task-specific variance and measurement error, they provide a more reliable and accurate measure of the intended construct than do the individual tasks (Friedman, Miyake, Corley, Young, Defries & Hewitt, 2006). Despite the possible limitation of some latent factors having only two indicators, not only did we confirm the existence of the same three EF factors Miyake found with this technique, but we did so using a different battery of tasks, a younger age group, and a risk group.

Our findings also shed light on the nature of the commonality that exists among Miyake’s three EFs. They suggest that speed is at the core of this common- ality. In the present study, as in Miyake’s study, the three EFs were moderately intercorrelated, r = .24 to .43 (see also Lehto et al., 2003). The inference that speed is responsible for much of this commonality is based on the sharp drop in these intercorrelations in the path model, when speed was partialed (see Figure 1). The correlation between working memory and shifting fell from .66 to .32; between working memory and inhibition from .48 to .04; and between inhibition and shifting from .24 to ).07. This finding deepens our understanding of the nature of EF constructs, and is consistent with suggestions that it may be useful to remove the variance due to speed when studying the unique properties of particular EFs (Huizinga et al., 2006; Mulder et al., 2009).

These results also indicate that the three EFs singled out by Miyake as separate but interrelated constructs in adults, are already similarly structured by 11 years. It is not, however, clear whether these EFs are separable at earlier ages. As indicated earlier, Wiebe et al. (2008) found that a one-factor model fit as well as a multi-factor model in a study that assessed working memory and inhibition in preschoolers. Moreover, Tsujimoto et al. (2007) found a developmental change in the structure of working memory and inhibition between 5–6 years and 8–9 years, such that the constructs correlated with one another in the younger group but not in the older group. Together, these results raise the possibility of changes in the structure of EF between preschool and pre- adolescence.

Modeling the effects of prematurity

To our knowledge, this report represents the first attempt to use a latent variable approach to model the relations of EF to academic achievement, and one of the few attempts to model preterm deficits. The model, which posited a developmental cascade, with processing speed

Executive function in preterm adolescents 1171

� 2011 Blackwell Publishing Ltd.

and EFs explaining preterm ⁄ full-term differences in reading and math, was supported. These findings are consistent with the theory of Salthouse (1996), which considers processing speed to be a driving force for higher cognition. They are also consistent with findings of Frye and Hale (1996), Kail (2007) and Nettelbeck and Burns (2010), who all describe a developmental cascade in childhood and adolescence in which age-related improvements in processing speed result in improved working memory which, in turn, is linked to improved reasoning ability. The results of the present study extend this model, showing that additional aspects of executive functioning (inhibition and shifting) are similarly influ- enced by processing speed, and that speed and all three EFs mediate preterm ⁄ full-term differences in achieve- ment. These results are supported by those of a recent study (Mulder, Pitchford & Marlow, 2010) which showed that processing speed and working memory partially mediated preterm ⁄ full-term differences in teacher- reported academic competence.

While processing speed affects all three EFs, it is not clear how it specifically drives executive deficits in chil- dren born preterm. One possibility is that the relevant cognitive operations are executed too slowly to be suc- cessfully completed in the available time. Another is that fewer rehearsals loops are possible within the same time frame, with the result that representations in working memory are weaker.

Despite the fact that processing speed accounted for considerable variance in all three EFs, there was also an independent effect of preterm birth on one particular EF, namely, shifting. Our data offer no leads as to what these variables might be. However, one possibility is that group differences may exist in the tendency to perseverate, and this may be independent of processing speed, which may principally affect the rule-learning aspects of shifting attention.

The concept of working memory has become central in many theories of cognition (Cowan, 1997; Fry & Hale, 1996). Our finding that working memory is the EF most pivotal to academic achievement is consistent with findings mentioned in the introduction, and with the large body of work by Gathercole and colleagues (Alloway et al., 2009; Gathercole et al., 2006). These investigators have linked working memory to academic achievement in several large-scale studies in the general population, as well as in children with reading disabili- ties. A screening study of over 3000 children, aged 5–11 years, found that 10% had deficits in working memory and about two-thirds of these performed poorly on tests of reading and math (Alloway et al., 2009). The findings here also show that working memory plays a much larger role in achievement than do the other two EFs (inhibition and shifting); indeed, relations of these two EFs to achievement disappeared once working memory was taken into account. This finding suggests that inhibition and shifting are related to achievement only through the variance they share with working

memory. Although we had thought that there might be domain-specific links between working memory and achievement, with listening span related to reading and counting span to math, this was not the case. Instead, all three types of working memory – spatial, listening span, and counting span – correlated similarly to reading and math. These findings, though at odds with some earlier data suggesting domain-specific links (Jarvis & Gather- cole, 2003), are consistent with more recent findings (St Clair-Thompson & Gathercole, 2006), and suggest that working memory reflects a domain-general ability important to various areas of academic achievement.

In summary, preterms do poorly, on average, on tasks of the three components of EF identified by Miyake – working memory, inhibition, and shifting. Moreover, these three components of EF, along with processing speed, mediate preterm ⁄ full-term differences in academic ability as part of a cascade of influence: prematurity fi slower processing speed fi poorer executive functioning (working memory) fi lower achievement in math and reading. The present results highlight the fundamental role of basic cognitive abilities in EF and have implica- tions for interventions to improve school-related abilities.

Acknowledgements

This research was funded in part by Grants HD 13810 and HD 049494 from the National Institutes of Health. The authors are grateful to all the participants and their parents, and to Keisha Phillips for her invaluable help in testing children and scoring data.

References

Aarnoudse-Moens, C.S., Weisglas-Kuperus, N., & van Gou- doever, J.B. (2009). Meta-analysis of neurobehavioral out- comes in very preterm and ⁄ or very low birth weight children. Pediatrics, 124, 717–728.

Agostino, A., Johnson, J., & Pascual-Leone, J. (2010). Execu- tive functions underlying multiplicative reasoning: problem type matters. Journal of Experimental Child Psychology, 105, 286–305.

Allen, M.C. (2008). Neurodevelopmental outcomes of preterm infants. Current Opinion in Neurology, 21, 123–128.

Alloway, T.P., Gathercole, S.E., Kirkwood, H.J., & Elliott, J.E. (2009). The cognitive and behavioural characteristics of children with low working memory. Child Development, 80, 606–621.

Anderson, P.J., Doyle, L.W., & Victorian Infant Collaborative Study Group. (2004). Executive functioning in school-aged children who were born very preterm or with extremely low birth weight in the 1990s. Pediatrics, 114, 50–57.

Aylward, G.P. (2002). Cognitive and neuropsychological out- comes: more than IQ scores. Mental Retardation and Devel- opmental Disabilities Research Reviews, 8, 234–240.

Bayless, S., & Stevenson, J. (2007). Executive functions in school-age children born very prematurely. Early Human Development, 83, 247–254.

1172 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

Bhutta, A.T., Cleves, M.A., Casey, P.H., Cradock, M.M., & Anand, K.J.S. (2002). Cognitive and behavioral outcomes of school-aged children who were born preterm: a meta-analy- sis. Journal of the American Medical Association, 288, 728– 736.

Blair, C., & Razza, R.P. (2007). Relating effortful control, executive function, and false belief understanding to emerg- ing math and literacy ability in kindergarten. Child Devel- opment, 78, 647–663.

Botting, N., Powls, A., Cooke, R.W.I., & Marlow, N. (1998). Cognitive and educational outcome of very-low-birthweight children in early adolescence. Developmental Medicine and Child Neurology, 40, 652–660.

Bull, R., & Scerif, G. (2001). Executive functioning as a pre- dictor of children’s mathematics ability: inhibition, switching, and working memory. Developmental Neuropsychology, 19, 273–293.

Case, R., Kurland, M., & Goldberg, J. (1982). Operational efficiency and the growth of short-term memory span. Jour- nal of Experimental Child Psychology, 33, 386–404.

Casey, B.J., Trainor, R.J., Orendi, J.L., Schubert, A.B., Nystrom, L.E., & Giedd, J.N. (1997). A developmental functional MRI study of prefrontal activation during per- formance of a go-nogo task. Journal of Cognitive Neurosci- ence, 9, 835–847.

Christ, S.E., White, D.A., Brunstrom, J.E., & Abrams, R.A. (2003). Inhibitory control following perinatal brain injury. Neuropsychology, 17, 171–178.

Constable, R.T., Ment, L.R., Vohr, B.R., Kesler, S.R., Ful- bright, R.K., Lacadie, C., Delancy, S., Katz, K.H., Schnei- der, K.C., Schafer, R.J., Makuch, R.W., & Reiss, A.L. (2008). Prematurely born children demonstrate white matter microstructural differences at 12 years of age, relative to term control subjects: an investigation of group and gender effects. Pediatrics, 121, 306–316.

Conway, A.R.A., Kane, M.J., Bunting, M.F., Hambrick, D.Z., Wilhelm, O., & Engle, R.W. (2005). Working memory span tasks: a methodological review and user’s guide. Psychonomic Bulletin & Review, 12, 769–786.

Cowan, N. (1997). The development of memory in childhood. Hove, East Sussex: Psychology Press.

Curtis, W.J., Lindeke, L., Georgieff, M.K., & Nelson, C.A. (2002). Neurobehavioural functioning in neonatal intensive care unit graduates in late childhood and early adolescence. Brain, 125, 1646–1659.

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

DeFries, J.C., & Plomin, R. (1985). Origins of individual dif- ferences in infancy: The Colorado Adoption Project. Orlando, FL: Academic Press.

Detterman, D.K. (1988). CAT: Cognitive Abilities Test (unpublished test). Unpublished manuscript, Cleveland, OH.

Duncan, J., Emslie, H., Williams, P., Johnson, R., & Freer, C. (1996). Intelligence and the frontal lobe: the organization of goal-directed behavior. Cognitive Psychology, 30, 257–303.

Durston, S., Thomas, K.M., Yang, Y., Ulug, A.M., Zimmer- man, R.D., & Casey, B.J. (2002). A neural basis for the development of inhibitory control. Developmental Science, 5, F9–F16.

Elgen, I., Lundervold, A.J., & Sommerfelt, K. (2004). Aspects of inattention in low birth weight children. Pediatric Neu- rology, 30, 92–98.

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

Espy, K.A., Fang, H., Charak, D., Minich, N., & Taylor, G.H. (2009). Growth mixture modeling of academic achievement in children of varying birth weight risk. Neuropsychology, 23, 460–474.

Espy, K.A., McDiarmid, M.M., Cwik, M.F., Stalets, M.M., Hamby, A., & Senn, T.E. (2004). The contribution of exec- utive functions to emergent mathematic skills in preschool children. Developmental Neuropsychology, 26, 465–486.

Friedman, N.P., & Miyake, A. (2004). The relations among inhibition and interference control functions: a latent-vari- able analysis. Journal of Experimental Psychology: General, 133, 101–135.

Friedman, N.P., Miyake, A., Corley, R.P., Young, S.E., Defries, J.C., & Hewitt, J.K. (2006). Not all executive func- tions are related to intelligence. Psychological Science, 17, 172–179.

Fry, A.F., & Hale, S. (1996). Processing speed, working memory, and fluid intelligence: evidence for a developmental cascade. Psychological Science, 7, 237–241.

Gathercole, S.E., Alloway, T.P., Willis, C.S., & Adams, A.M. (2006). Working memory in children with reading disabilities. Journal of Experimental Child Psychology, 93, 265–281.

Giedd, J.N., Blumenfeld, R.S., Jeffries, N.O., Castellanos, F.X., Liu, H., Zijdenbos, A., Paus, T., Evans, A.C., & Rapoport, J.I. (1999). Brain development during childhood and adolescence: a longitudinal study. Nature Neuroscience, 2, 861–863.

Gim�nez, M., Junqu�, C., Narberhaus, A., Bargall�, N., Botet, F., & Mercader, J.M. (2006). White matter volume and concentration reductions in adolescents with history of very preterm birth: a voxel-based morphometry study. NeuroIm- age, 32, 1485–1498.

Hale, S. (1990). A global developmental trend in cognitive processing speed. Child Development, 61, 653–663.

Hollingshead, A.B. (1975). The four-factor index of social status. Unpublished manuscript, Yale University. (Available from A.B. Hollingshead, Department of Sociology, Yale University, New Haven, CT 06520.)

Huizinga, M., Dolan, C.V., & van der Molen, M.W. (2006). Age-related change in executive function: developmental trends and a latent variable analysis. Neuropsychologia, 44, 2017–2036.

Jarvis, H.L., & Gathercole, S.E. (2003). Verbal and non-verbal working memory and achievements on national curriculum tests at 11 and 14 years of age. Educational and Child Psy- chology, 20, 123–140.

Johnson, C., Hennessy, E., Smith, R., Trikic, R., Wolke, D., & Marlow, N. (2010). Academic attainment and special edu- cation needs in extremely preterm children at 11 years of age: the EPICure study. Archives of Disease in Childhood – Fetal Neonatal Edition, 94, F283–F289.

Jçreskog, K., & Sçrbom, D. (2003). LISREL (Version 8.54). Chicago, IL: Scientific Software International, Inc.

Kail, R. (1991). Developmental change in speed of processing during childhood and adolescence. Psychological Bulletin, 109, 490–501.

Kail, R. (2007). Longitudinal evidence that increases in pro- cessing speed and working memory enhance children’s rea- soning. Psychological Science, 18, 312–313.

Executive function in preterm adolescents 1173

� 2011 Blackwell Publishing Ltd.

Kail, R., & Salthouse, T.A. (1994). Processing speed as a mental capacity. Acta Psychologica, 86, 199–225.

Kennedy, K.M., & Raz, N. (2009). Aging white matter and cognition: differential effects of regional variations in diffu- sion properties on memory, executive functions, and speed. Neuropsychologia, 47, 916–927.

Kulseng, S., Jennekens-Schinkel, A., Naess, P., Romundstad, P., Indredavik, M., Vik, T., & Brubakk, A.-M. (2006). Very- low-birthweight and term small-for-gestational-age adoles- cents: attention revisited. Acta Paediatrica, 95, 224–230.

Lee, K., Ng, E.L., & Ng, S.F. (2009). The contributions of working memory and executive functioning to problem representation and solution generation in algebraic word problems. Journal of Educational Psychology, 101, 373–387.

Lehto, J.E., Juujarvi, P., Kooistra, L., & Pulkkinen, L. (2003). Dimensions of executive functioning: evidence from children. British Journal of Developmental Psychology, 21, 59–80.

Lenroot, R.K., & Giedd, J.N. (2006). Brain development in children and adolescents: insights from anatomical magnetic resonance imaging. Neuroscience and Biobehavioral Reviews, 30, 718–729.

Luciana, M., Lindeke, L., Georgieff, M., Mills, M., & Nelson, C.A. (1999). Neurobehavioral evidence for working-memory deficits in school-aged children with histories of prematurity. Developmental Medicine and Child Neurology, 41, 521–533.

Luciana, M., & Nelson, C.A. (2002). Assessment of neuro- psychological function through use of the Cambridge Neu- ropsychological Testing Automated Battery: performance in 4- to 12-year-old children. Developmental Neuropsychology, 22, 595–624.

Luck, S.J., & Vogel, E.K. (1997). The capacity of visual working memory for features and conjunctions. Nature, 390, 279–281.

McCabe, D.P., Roediger, H.L., McDaniel, M.A., Balota, D.A., & Hambrick, D.Z. (2010). The relationship between working memory capacity and executive functioning: evidence for a common executive attention construct. Neuropsychology, 24, 222–243.

McCormick, M.C. (1989). Long-term follow-up of infants discharged from neonatal intensive care units. Journal of the American Medical Association, 261, 1767–1772.

Miyake, A., Friedman, N.P., Emerson, M.J., Witzki, A.H., & Howerter, A. (2000). The unity and diversity of executive functions and their contribution to complex ‘frontal lobe’ tasks: a latent variable analysis. Cognitive Psychology, 41, 49– 100.

Mulder, H., Pitchford, N.J., Hagger, M.S., & Marlow, N. (2009). Development of executive function and attention in preterm children: a systematic review. Developmental Neuro- psychology, 34, 393–421.

Mulder, H., Pitchford, N.J., & Marlow, N. (2010). Processing speed and working memory underlie academic attainment in very preterm children. Archives of Disease in Childhood – Fetal Neonatal Edition, 95, F267–F272.

Nettlebeck, T., & Burns, N.R. (2010). Processing speed, working memory and reasoning ability from childhood to old age. Personality and Individual Differences, 48, 379–384.

Nigg, J.T. (2000). On inhibition ⁄ disinhibition in developmental psychopathology: views from cognitive and personality psy- chology and a working inhibition taxonomy. Psychological Bulletin, 126, 220–246.

Nosarti, C., Giouroukou, E., Healy, E., Rifkin, L., Walshe, M., Reichenberg, A., Chitnis, X., Williams, S.C.R., & Murray,

R.M. (2008). Grey and white matter distribution in very preterm adolescents mediates neurodevelopmental outcome. Brain, 131, 205–217.

Nosarti, C., Shergill, S.S., Allin, M.P., Walshe, M., Rifkin, L., Murray, R.M., & McGuire, P.K. (2009). Neural substrates of letter fluency processing in young adults who were born very preterm: alterations in frontal and striatal regions. Neuro- Image, 47, 1904–1913.

Passolunghi, M.C., & Siegel, L.S. (2001). Short-term memory, working memory, and inhibitory control in children with difficulties in arithmetic problem solving. Journal of Experi- mental Child Psychology, 80, 44–57.

Passolunghi, M.C., & Siegel, L.S. (2004). Working memory and access to numerical information in children with disability in mathematics. Journal of Experimental Child Psychology, 88, 348–367.

Pinto-Martin, J.A., Whitaker, A., Feldman, J., Cnaan, A., Zhao, H., Rosen-Bloch, J., McCulloch, D., & Paneth, N. (2004). Special education services and school performance in a regional cohort of low birthweight infants at age nine. Paediatric and Perinatal Epidemiology, 18, 120–129.

Petrides, M., & Milner, B. (1982). Deficits in subject-ordered tasks after frontal and temporal-lobe lesions in man. Neu- ropsychologia, 20, 249–262.

Reitan, R.M., & Wolfson, D. (1992). Neurospychological eval- uation of older children. Yuscon, AZ: Neuropsychology Press.

Rose, S.A., & Feldman, J.F. (1996). Memory and processing speed in preterm children at 11 years: a comparison with full- terms. Child Development, 67, 2005–2021.

Rose, S.A., & Feldman, J.F. (1997). Memory and speed: their role in the relation of infant information processing to later IQ. Child Development, 68, 630–641.

Rose, S.A., Feldman, J.F., & Jankowski, J.J. (2001). Attention and recognition memory in the first year of life: a longitu- dinal study of preterms and full-terms. Developmental Psy- chology, 37, 135–151.

Rose, S.A., Feldman, J.F., & Jankowski, J.J. (2002). Processing speed in the 1st year of life: a longitudinal study of preterm and full-term infants. Developmental Psychology, 38, 895– 902.

Rose, S.A., Feldman, J.F., & Jankowski, J.J. (2009). Informa- tion processing in toddlers: continuity from infancy and persistence of preterm deficits. Intelligence, 37, 311–320.

Rushe, T.M., Rifkin, L., Stewart, A.L., Townsend, J.P., Roth, S.C., Wyatt, J.S., & Murray, R.M. (2001). Neuropsycho- logical outcome at adolescence of very preterm birth and its relation to brain structure. Developmental Medicine and Child Neurology, 43, 226–233.

Saigal, S., Hoult, L.A., & Streiner, D.L. (2000). School difficulties in adolescence in a regional cohort of children who were extremely low birth weight. Pediatrics, 105, 325– 331.

Salthouse, T.A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103, 403– 428.

Salthouse, T.A. (2000). Aging and measures of processing speed. Biological Psychology, 54, 35–54.

Salthouse, T.A. (2005). Relations between cognitive abilities and measures of executive functioning. Neuropsychology, 19, 532–545.

Salthouse, T.A., Fristoe, N., McGuthry, K.E., & Hambrick, D.Z. (1998). Relation of task switching to speed, age, and fluid intelligence. Psychology and Aging, 13, 445–461.

1174 Susan A. Rose et al.

� 2011 Blackwell Publishing Ltd.

Siegel, L.S., & Ryan, E.B. (1989). The development of working memory in normally achieving and subtypes of learning disabled children. Child Development, 60, 973–980.

Soria-Pastor, S., Gim�nez, M., Narberhaus, A., Falcon, C., Botet, F., Bargall�, N., Mercader, J.M., & Junqu�, C. (2008). Patterns of cerebral white matter damage and cognitive impairment in adolescents born very preterm. International Journal of Developmental Neuroscience, 26, 647–654.

St Clair-Thompson, H.L., & Gathercole, S.E. (2006). Executive functions and achievements in school: shifting, updating, inhibition and working memory. Quarterly Journal of Experimental Psychology, 59, 745–759.

Strang-Karlsson, S., Andersson, S., Paile-Hyvarinen, M., Darby, D., Hovi, P., Raikkonen, K., Heinon, K., Jarvenpaa, A.-L., Erkisson, J.G., & Kajantie, E. (2010). Slower reaction times and impaired learning in young adults with birth weight <1500g. Pediatrics, 125, e74–e82.

Swanson, H.J. (2006). Cross-sectional and incremental changes in working memory and mathematical problem solving. Journal of Educational Psychology, 98, 265–281.

Taylor, G.H., Klein, N.,Drotar, D., Schluchter,M., & Hack, M. (2006). Consequences and risk of <1000-g birth weight for neuropsychological skills, achievement, and adaptive func- tioning. Developmental and Behavioral Pediatrics, 27, 459–469.

Taylor, H.G., Minich, N., Bangert, B., Filipek, P.A., & Hack, M.B. (2004). Long-term neuropsychological outcomes of very low birth weight: associations with early risks for peri- ventricular brain insults. Journal of the International Neuro- psychological Society, 10, 987–1004.

Tsujimoto, S., Kuwajima, M., & Sawaguchi, T. (2007). Developmental fractionation of working memory and response inhibition in childhood. Experimental Psychology, 54, 30–37.

Turken, A.U., Whitfield-Gabrieli, S., Bammer, R., Baldo, J.V., Dronkers, N.F., & Gabrieli, J.D.E. (2008). Cognitive pro- cessing speed and the structure of white matter pathways: convergent evidence from normal variation and lesion stud- ies. NeuroImage, 42, 1032–1044.

van der Sluis, S., Jong, P.F., & van der Leij, A. (2007). Exec- utive functioning in children, and its relation with reasoning, reading, and arithmetic. Intelligence, 35, 427–449.

Waters, G.S., & Caplan, D. (2003). The reliability and stability of verbal working memory measures. Behavior Research Methods, Instruments, & Computers, 35, 550–564.

Wiebe, S.A., Espy, K.A., & Charak, D. (2008). Using confir- matory factor analysis to understand executive control in preschool children: 1. Latent structure. Developmental Psy- chology, 44, 575–587.

Woodcock, R., McGrew, R., & Mather, N. (1999). Woodcock- Johnson III – Tests of Achievement. Itnoca, IL: Riverside Publishing.

Woodward, L.J., Edgin, J.O., Thompson, D., & Inder, T.E. (2005). Object working memory deficits predicted by early brain injury and development in the preterm infant. Brain, 128, 2578–2587.

Received: 25 October 2010 Accepted: 10 March 2011

Executive function in preterm adolescents 1175

� 2011 Blackwell Publishing Ltd.

Copyright of Developmental Science is the property of Wiley-Blackwell and its content may not be copied or

emailed to multiple sites or posted to a listserv without the copyright holder’s express written permission.

However, users may print, download, or email articles for individual use.