Age-related change in executive function: Developmental trends and a latent variable analysis
Introduction
Across development, children become increasingly more able to control their thoughts and actions (for a review see: Diamond, 2002). This change has been associated with the development of executive function (EF), which is an umbrella term for various cognitive processes that subserve goal-directed behavior (Miller & Cohen, 2001; see also Luria, 1966, Shallice, 1982). EF is especially important in novel or demanding situations (Stuss, 1992), which require a rapid and flexible adjustment of behavior to the changing demands of the environment (Zelazo, Muller, Frye, & Marcovitch, 2003). EF is thought to rely strongly on prefrontal cortex (PFC), as indicated by studies showing that patients with lesions to PFC perform poorly on tasks such as the Wisconsin Card Sorting Task (WCST; Grant & Berg, 1948) and the Tower of London (ToL; Shallice, 1982; for a review see: Stuss & Levine, 2002). On the WCST, which requires flexible switching between sorting rules, PFC patients typically perseverate, i.e., they persist in sorting according to the rule that was previously correct (e.g., Anderson, Damasio, Jones, & Tranel, 1991; Milner, 1963; Nagahama, Okina, Suzuki, Nabatame, & Matsuda, 2005; Stuss et al., 2000). On the ToL, which requires spatial problem solving by moving balls in order to reach a pre-specified goal, PFC patients require more moves to solve the problem (e.g., Andres & Van der Linden, 2001; Carlin et al., 2000; Morris, Ahmed, Syed, & Toone, 1993; Owen, Downes, Sahakian, Polkey, & Robbins, 1990).
Children show a similar pattern as patients with PFC damage; that is, they also perseverate on the WCST and require more moves to solve ToL problems (Anderson, Anderson, & Lajoie, 1996; Baker, Segalowitz, & Ferlisi, 2001; Chelune & Baer, 1986; Chelune & Thompson, 1987; Heaton, Chelune, Talley, Kay, & Curtis, 1993; Kirk & Kelly, 1986; Lehto, 2004; Lehto, Juujaervi, Kooistra, & Pulkkinen, 2003; Paniak, Miller, Murphy, & Patterson, 1996; Welsh, Pennington, & Groisser, 1991). The slow development of EF has been attributed to the protracted maturation of PFC (e.g., Diamond, 2002). Conclusive evidence about the developmental trajectories of the different EF components in relation to the performance on standard neuropsychological EF tasks has yet to be established. In this study, we examined the development of EF component processes by using a multi-group confirmatory factor analysis. Where we have at our disposal multiple indicators of a given latent variable, this approach has the advantage that it allows us to study performance at the level of the latent variables, according to a pre-specified model of EF.
A major theoretical issue concerns the organization of EF. It has been suggested that EF is unitary, i.e., it does not include distinct sub-functions or sub-components. This means that the cognitive and behavioral impairments seen after PFC damage can be explained entirely in terms of one dysfunctional system (e.g., Cohen & Servan-Schreiber, 1992; Duncan, Emslie, Williams, Johnson, & Freer, 1996; Kimberg, D’Esposito, & Farah, 1997). For example, Kimberg et al. (1997) posited that all deficits in PFC function can be attributed to deficits in working memory. In contrast, others view EF as multi-faceted (non-unitary). These authors argued that EF involves several discrete cognitive processes that have a relatively focal neural representation (e.g., Baddeley, 1986; Stuss, Shallice, Alexander, & Picton, 1995; see also Teuber, 1972). The multi-faceted nature of EF is suggested by behavioral studies incorporating batteries of widely used EF tasks. These studies yielded low or non-significant correlations between tasks and exploratory factor analysis yielded multiple factors (Brocki & Bohlin, 2004; Culbertson & Zillmer, 1998; Lehto, 1996, Levin et al., 1996, Pennington, 1997, Robbins et al., 1994, Welsh et al., 1991).
Neuroimaging studies provide evidence in support of the multi-faceted nature of EF, as different components of EF are seen to rely on different parts of PFC. For example, the ability to maintain information in working memory has been found to recruit mostly lateral PFC (Narayanan et al., 2005; Smith & Jonides, 1999). In contrast, switching between tasks is thought to rely on medial PFC (Crone, Wendelken, Donohue, & Bunge, 2005; Rushworth, Walton, Kennerley, & Bannerman, 2004). Finally, the ability to inhibit responses was found to rely on orbitofrontal cortex (e.g., Aron, Robbins, & Poldrack, 2004; Roberts & Wallis, 2000). Thus, different regions within PFC subserve different components of goal-directed behavior.
At this point, it should be noted that the problem of “task impurity” hinders the interpretation of results reported in behavioral and neuroimaging studies using multiple EF tasks. Task impurity refers to the fact that a single indicator (operationalization) of a given construct (e.g., Working Memory) can rarely, if ever, be viewed as a pure measure of that construct. Most measures are contaminated by random error and systematic error (see Kline, 1998, p. 55). The task impurity problem is highly relevant to EF research, as the manifestation of EF components invariably involves other (non-EF) cognitive processes (e.g., Miyake et al., 2000).
Miyake et al. (2000) presented one way to address the task impurity problem. They proposed using multiple tasks to measure each EF component and adopting a latent variables approach to extract the variance common to those tasks. Latent variables (as incorporated in structural equation models; SEM) refer to what is shared among tasks that are assumed to tap a given EF. The latent variable approach minimizes the task impurity problem, and is therefore especially informative in developmental studies (e.g., Nunally & Bernstein, 1994, p. 85). Using confirmatory factor analysis, Miyake et al. (2000) examined the separability of three frequently postulated EF components: “Working Memory”, “Shifting”, and “Response Inhibition” (henceforth: Inhibition). Miyake et al. (2000) focused on these three EF components because: (1) they are well-circumscribed, lower-level functions that can be operationalized in a fairly precise manner; (2) they can be studied using commonly used tasks; and (3) they have been implicated in the performance of more complex EF tasks, such as the WCST and ToL. Miyake et al. (2000) tested healthy young-adults on multiple tasks tapping Working Memory, Shifting, and Inhibition, and several standard, but complex, neuropsychological tasks, including the WCST and the Tower of Hanoi (similar to the ToL). The results showed that, although moderately correlated, Working Memory, Shifting, and Inhibition were separable constructs (see also Fisk & Sharp, 2004). Moreover, the EF component processes differentially predicted performance on the complex neuropsychological tasks. For example, Shifting predicted WCST performance, whereas Inhibition predicted ToH performance.
Developmental studies using standard neuropsychological tasks have shown that EF has a protracted course of development, beginning in early childhood and continuing into adolescence. However, these EF tasks are subject to distinct developmental trajectories. For example, on the WCST, analysis of perseverative errors indicates that the performance of children is comparable to that of young-adults by 12 years of age; however, analysis of failure-to-maintain set indicates that children do not reach adult levels of performance until 13–15 years of age (e.g., Chelune & Baer, 1986; Chelune & Thompson, 1987; Levin et al., 1991, Welsh et al., 1991). Similarly, on the ToL task, performance based on errors appears to continuously improve from middle childhood into young-adulthood; however, when performance is based on both errors and time, adult levels of performance may be attained as early as 13 years of age (Baker et al., 2001; see also Levin et al., 1996).
There is a growing body of research indicating differential trends in the development of EF component processes.1 These studies, although not entirely unequivocal, show that adult-level performance on different EF tasks is attained at different ages during childhood and adolescence (for reviews see: Diamond, 2002, Welsh, 2002). Working memory capacity has been found to gradually develop throughout childhood and into adolescence (e.g., Beveridge, Jarrold, & Pettit, 2002; Brocki & Bohlin, 2004; DeLuca et al., 2003; Gathercole, Pickering, Ambridge, & Wearing, 2004; Hitch, Halliday, Dodd, & Littler, 1989; Luciana, Conklin, Hooper, & Yarger, 2005; Luciana & Nelson, 1998; Luna, Garver, Urban, Lazar, & Sweeney, 2004). In addition, recent studies on the development of task shifting abilities all show that the cost of switching between tasks decreases as children grow older, with adult levels of performance being attained around the age of 12 (Cepeda, Kramer, & Gonzalez de Sather, 2001; Crone, Bunge, Van der Molen, & Ridderinkhof, in press; Huizinga & Van der Molen, in preparation-a; Kray, Eber, & Lindenberger, 2004). Finally, inhibitory control was found to increase throughout childhood (e.g., Klenberg, Korkman, & Lahti Nuuttila, 2001), and to reach adult level of performance in late childhood, around the age of 12 (Bédard et al., 2002; Bunge, Dudukovic, Thomason, Vaidya, & Gabrieli, 2002; Durston et al., 2002; Ridderinkhof & Van der Molen, 1995; Van den Wildenberg & Van der Molen, 2004), or early adolescence (Williams, Ponesse, Schachar, Logan, & Tannock, 1999).
A straightforward interpretation of the developmental trends of the EF component processes is hampered by a number of factors. First, different tasks are used across studies to measure the same EF component. For example, the developmental trajectory of Inhibition has been assessed by using a Go/NoGo task (e.g., Durston et al., 2002), the Eriksen Flankers task (e.g., Bunge et al., 2002; Ridderinkhof & Van der Molen, 1995), and the Stop-signal task (e.g., Bédard et al., 2002; Van den Wildenberg & Van der Molen, 2004; Williams et al., 1999). Second, it is unclear whether children at various ages use the same strategy when performing on EF tasks. This issue concerns the question of measurement invariance, i.e., whether we are actually measuring the same construct across age (Meredith, 1993). Third, many developmental studies focused on a single EF component process. This precludes a reliable assessment of developmental change across EF component processes, because differential rates might be due to different samples rather than components. Thus, a reliable assessment of the developmental patterning of EF component processes requires homogeneous age groups and the application of a latent variables approach to extract what the various tasks used to tap EF component processes have in common. This approach has been adopted by Lehto et al. (2003), who were the first to assess the patterning of EF component processes in children by using SEM. Importantly, they observed the same factor structure in a group of 8–13-year old children as previously found by Miyake et al. (2000) in adults. In the present study, we hope to contribute to these results by adopting a multi-group design, and thus providing a more graded assessment of developmental change in EF component processes.
In the present study, we adopted the conceptual framework of Miyake et al. (2000) to assess developmental change in EF. The main goal of this study was to examine age-related changes in the three EF components distinguished by Miyake et al. (2000), i.e., Working Memory, Shifting, and Inhibition. In order to shed some light on the development of these EF component processes, we tested children in three homogeneous age groups (i.e., 7-, 11-, 15-year olds), in addition to a group of young-adults (i.e., 21-year olds). Adult level on EF tasks is typically reached in late childhood or early adolescence (for reviews see: Diamond, 2002, Welsh, 2002). The decision to limit the youngest group to 7-year olds was based on the consideration that the present task battery was probably too difficult for children younger than 7 years of age. The EF components, Working Memory, Shifting, and Inhibition, were indexed by nine experimental tasks, three for each EF component.
Working Memory was defined as the collection of cognitive processes that temporarily retain information in an accessible state, suitable for carrying out any mental task (Cowan, 1998). The essence of this component is the monitoring and coding of incoming information with respect to relevance and replacement of information that is no longer relevant by newly relevant information. Shifting was interpreted as shifting back and forth between multiple tasks (Allport, Styles, & Hsieh, 1994; Monsell, 1996, Monsell, 2003). When different (usually choice RT) tasks are mixed within blocks, shifting between tasks typically results in an increase in RT and a decrease in accuracy (i.e., shift costs). Inhibition was conceptualized as the ability to deliberately inhibit dominant, automatic, or pre-potent responses (Logan & Cowan, 1984). The essence of this EF component lies in the suppression of a response or in the control of interfering stimuli or competing responses.
We adopted two approaches in analyzing the data. First, we conducted a standard analysis of variance approach. Second, we took a latent variable approach, i.e., multi-group confirmatory factor analyses (Dolan, 2000, Meredith, 1993). We examined (i) the organization of executive function in children and young-adults by investigating whether the indicators of the Working Memory, Shifting, and Inhibition tasks measured the same constructs across age, (ii) whether this organization changes across development, and (iii) how EF component processes contribute to the performance on the WCST and the ToL across age groups, again following Miyake et al. (2000). We included the WCST and the ToL because these tasks have been used previously to study the development of EF (Anderson, Byrd, & Berg, 2005; Baker et al., 2001; Chelune & Baer, 1986; Lehto, 2004, Welsh et al., 1991).
Section snippets
Sample
The present study included four age groups: seventy-one 7-year olds (39 female, M age = 7.2 (age range: 6–8); M Raven-quartile = 3.6 (S.D. = 0.88); M number of years of education = 0.56 (S.D. = 0.13)), one hundred and eight 11-year olds (62 female, M age = 11.2 (age range: 10–12); M Raven-quartile = 3.2 (S.D. = 0.93); M number of years of education = 3.92 (S.D. = 0.13)); one hundred and eleven 15-year olds (58 female, M age: 15.3 (age range: 14–16); M Raven-quartile = 3.1 (S.D. = 0.99); M number of years of education:
Results
We performed three sets of analyses. The first set included analyses of variance to assess developmental trajectories for each task. The second set included multi-group confirmatory factor analysis, to assess when the latent components Working Memory, Shifting, and Inhibition reached adult levels. The third set included regression analyses to assess the contribution of the latent factors to the performance on the WSCT and ToL.
Discussion
In this study, we examined the developmental trajectories of three frequently postulated EF components, Working Memory, Shifting, and Inhibition, in relation to performance on standard neuropsychological EF tasks, the WCST and the ToL. In so doing, we adopted both standard analyses and multi-group latent variable modeling. The latter enabled us to model the structure of the underlying latent variables across age groups, thus going beyond the usual analysis of observed developmental trends in
Acknowledgements
The authors thank Ingmar Visser, Ellen Hamaker, and Verena Schmittmann for valuable comments on an earlier version of this manuscript. Bert van Beek is gratefully acknowledged for programming the tasks. Thanks to Susanne Stuurman, Manon de Beer, and Esther Kooymans for assistance with the data collection.
References (136)
- et al.
Inhibition and the right inferior frontal cortex
Trends in Cognitive Sciences
(2004) - et al.
Both random and perspective errors underlie WCST deficits in prefrontal patients
Neuropsychologia
(2002) - et al.
The resources construct in cognitive development: Diverse sources of evidence and a theory of inefficient inhibition
Developmental Review
(1990) - et al.
Development of frontal lobe contributions to cognitive control in children: Evidence from fMRI
Neuron
(2002) - et al.
Planning impairments in frontal lobe dementia and frontal lobe lesion patients
Neuropsychologia
(2000) - et al.
Operational efficiency and the growth of short-term memory span
Journal of Experimental Child Psychology
(1982) - et al.
Imaging the developing brain: What have we learned about cognitive development?
Trends in Cognitive Sciences
(2005) - et al.
The rise and fall in information-processing rates over the life-span
Acta Psychologica
(1994) - et al.
The development of serial short-term-memory and the articulatory loop hypothesis
Intelligence
(1990) - et al.
Working memory capacity and its relation to general intelligence
Trends in Cognitive Sciences
(2003)