Minnesota Symposia on Child Psychology, Volume 37: Developing Cognitive Control Processes: Mechanisms, Implications, and Interventions

By Philip David Zelazo

The collected papers from the most prestigious symposia in the field of child development provide scholars, students, and practitioners with access to the work of key researchers in human development. This volume focuses on changes in our understanding of cogisnitive control processes—constructs important to the field since Wundt and Freud. Our understanding of these constructs has advanced dramatically in recent years—both empirically and conceptually. This collection brings generalists and specialists alike up-to-date on this central process of human development and the implications for this new knowledge on school success and other areas.

• Language: English
• Publisher: Wiley
• Release date: Sep 24, 2013
• ISBN: 9781118415177

Book preview

Chapter 1

What Is Cognitive Control?

Philip David Zelazo and Jacob E. Anderson

WHAT IS COGNITIVE CONTROL?

Imagine yourself at age 3 years—let us say, 3 and a half. It is Saturday morning in late autumn, and instead of being at home in your pajamas watching Scooby-Doo, you find yourself in a university laboratory sitting across from a grown-up, a graduate student named Kay, who would also perhaps prefer to be somewhere else. You think of the Happy Meal you were promised if you got in the car in the first place. Kay places a cookie in front of you, and tells you that if you wait a few minutes, and do not eat the cookie, you will get two more cookies. Then she leaves the room.

What do you do? Do you wait? For how long? As it happens, your decision to wait, and especially how long you wait, will likely predict important developmental outcomes later in life. Pioneering work by Mischel and colleagues has revealed that children’s behavior in this experimental situation, the delay of gratification task, predicts future academic ability, including higher SAT scores, better social, cognitive, and emotional coping in adolescence, and better performance on measures of cognitive control in adulthood (e.g., Casey et al., 2011; Eigsti et al., 2006; Mischel, Shoda, & Peake, 1988; Mischel, Shoda, & Rodriguez, 1989; Shoda, Mischel, & Peake, 1990). Other researchers (e.g., Moffitt et al., 2011) have provided corroborating evidence that higher cognitive control measured in childhood is associated with a variety of salubrious outcomes in adulthood, including better physical health, higher socioeconomic status, and lower criminality. Findings like these have led to widespread recognition of the importance of cognitive control, and together with research on neuroplasticity—the way in which the brain adapts to the environment and changes as a function of behavior—they encourage the hope that efforts to improve children’s cognitive control may have lasting consequences for their well-being and adjustment.

The primary aim of this chapter is to provide a brief introduction to research on cognitive control and the way in which it develops during childhood and adolescence. In addition, however, we consider the implications of this research for the creation of interventions designed to promote the healthy development of cognitive control. These interventions have the notable appeal not only of helping children who are at risk for a wide range of problems, but also of allowing researchers to take an experimental approach to the study of behavioral and environmental influences on a key aspect of human development. Despite long-standing interest in these influences, most research on the development of cognitive control, including most longitudinal research, has been correlational. It is only through experimental research, with random assignment and proper controls, that it is possible to provide unambiguous evidence of causal influence.

DEFINING COGNITIVE CONTROL

Cognitive control is one of a number of overlapping constructs, such as executive function, self-regulation, self-control, and effortful control that refer to the neurocognitive processes involved in the top-down, goal-directed modulation of behavior, broadly defined to include attention, thought, emotion, motivation, and action. These are processes that are related to, but different from, what we normally mean by intelligence, and in contrast to the intellectual knowledge that is implied by intelligence, cognitive control concerns the use of that knowledge in the service of one’s goals. For example, it concerns the processes whereby the best laid plans are translated into successful action, as well as those self-reflective processes that allow for the adaptive revision of those plans.

Although much of what we do is habitual, the need for cognitive control is pervasive. We need it, for example, to resist impulses, to avoid distractions, to think flexibly, and to break habits. More generally, we can say that cognitive control is involved in deliberate problem solving—or at least those instances of problem solving that involve effort. Indeed, the feeling of effort is the subjective correlative of cognitive control (a point noted by James Mark Baldwin more than a century ago; Baldwin, 1892).

Factor-analytic work with adults is consistent with the suggestion that cognitive control is a hierarchical construct that is characterized by both unity and diversity (Miyake et al., 2000), such that performance on measures of cognitive control (or executive function) in adults can be captured by three partially independent latent variables, reflecting cognitive flexibility, inhibitory control, and working memory, as well as a higher-order factor that captures shared variance common to each measure. Cognitive flexibility involves thinking about something in multiple ways—for example, considering someone else’s perspective on a situation. Inhibitory control is the process of deliberately suppressing a response to something (e.g., ignoring a distraction or stopping an impulsive utterance). Working memory involves both keeping information in mind and, usually, manipulating it in some way.

These three aspects of cognitive control may be thought of as neurocognitive functions that contribute in important ways to the goal-directed modulation of behavior. Research on cognitive control examines their role in goal-directed behavior, but efforts to understand these functions are typically directed at another level of analysis—namely, the brain. That is, theories of cognitive control generally aim to understand these neurocognitive functions, and their development, in terms of the neural circuits that underlie them (Figure 1.1).

Figure 1.1 Theories of cognitive control address the neural circuits that underlie the neurocognitive functions that contribute to the goal-directed modulation of behavior, broadly defined. These are the functions whereby the brain modulates its own activity in the service of a goal.

COGNITIVE CONTROL AND THE BRAIN

It has long been known that a part of the brain, prefrontal cortex (PFC), plays a key role in cognitive control (Figure 1.2). An early piece of evidence was the case of Phineas Gage, the railroad foreman who suffered damage to his prefrontal cortex when a tamping iron was blown through the front part of his head (Harlow, 1848, 1868). Gage survived the accident, and most of his basic cognitive functions were preserved. At the same time, however, he underwent what was described as a transformation of his personality. Previously an upstanding citizen, Gage was now fitful, irreverent . . . devising many plans of future operations, which are no sooner arranged than they are abandoned in turn for others appearing more feasible (Harlow, 1868). These difficulties, along with other neuropsychological studies of head injuries (see Levine & Craik, 2012), have contributed greatly to our understanding of the construct of cognitive control, and they illustrate well the differences between (a) basic cognitive functions (intact) and their modulation (impaired), and (b) having goals or plans (intact) and being able to act on them (impaired).

Figure 1.2 A hierarchical model of rule representation in PFC.

A lateral view of the human brain is depicted at the top of the figure, with regions of PFC identified by the Brodmann areas (BA) that comprise them: Orbitofrontal cortex (BA 11), ventrolateral PFC (BA 44, 45, 47), dorsolateral PFC (BA 9, 46), and rostrolateral PFC (BA 10). The formulation and maintenance in working memory of more complex rules depends on the recruitment of additional regions of PFC into an increasingly complex hierarchy of PFC activation. Note: S = stimulus; check = reward; cross = nonreward; R = response; C = context, or task set. Brackets indicate a bivalent rule that is currently being ignored.

From A Brain-Based Account of the Development of Rule Use in Childhood, by S. Bunge and P. D. Zelazo, 2006, Current Directions in Psychological Science, 15, pp. 118–121. Reprinted with permission.

One of the most iconic neuropsychological observations of impaired cognitive control associated with prefrontal cortical damage comes from work by Milner (1963). Patients were given the Wisconsin Card Sorting Test (WCST; Grant & Berg, 1948), in which they were presented with target cards that differ on various dimensions (e.g., one red triangle, two green stars), and then shown test cards (e.g., three green triangles) that match different target cards on different dimensions. The patient’s task was to determine the rule according to which cards must be sorted (e.g., match by color), and the examiner informed patients after each sort whether they were correct or incorrect. After a certain number of consecutive correct responses, the target dimension was shifted, and the patient needed to discover the new sorting rule. Milner observed that patients tended to persist in sorting by the initial dimension despite being told it was wrong—they perseverated. More remarkably, they did so despite apparently knowing that they were wrong (e.g., they sometimes said something like, I know this is going to be wrong.). Teuber (1964, in the recorded discussion following Brenda Milner’s presentation) referred to this as a curious dissociation between knowing and doing (p. 333).

In retrospect, Gage’s injuries are now known to have involved primarily a particular part of prefrontal cortex—ventromedial prefrontal cortex—whereas Milner’s patients had lesions to dorsolateral prefrontal cortex. Prefrontal cortex is a large, heterogeneous expanse of neural tissue that comprises roughly the front third of the human brain, and several distinct regions can be distinguished on the basis of neuroanatomical connections, the relative importance of different neurotransmitter systems, and behavioral correlations (see Johnson, Munro, & Bunge, this volume). In contrast to functional circuits involving more lateral regions of prefrontal cortex (e.g., MacDonald, Cohen, Stenger, & Carter, 2000), those involving ventromedial regions, including parts of orbitofrontal cortex, play a prominent role in what has been called hot executive function (Zelazo & Müller, 2002).

Traditionally, conceptualizations of cognitive control have focused on its relatively cool, cognitive aspects, often associated with lateral prefrontal cortex and elicited by relatively abstract, decontextualized problems, such as the WCST and other well-established measures, including the classic Color-Word Stroop task (Stroop, 1935), versions of the Eriksen flanker task (Rueda et al., 2004), and the Dimensional Change Card Sort (DCCS; Zelazo, 2006). It is now clear, however, that there is an important distinction to be made between these more cool forms of cognitive control and the more hot, emotional forms of executive function that play a key role in motivationally significant situations (e.g., Brock, Rimm-Kaufman, Nathanson, & Grimm, 2009; Hongwanishkul, Happaney, Lee, & Zelazo, 2005; Willoughby, Kupersmidt, Voegler-Lee, & Bryant, 2011). Hot executive function relies more heavily on networks involving ventral and medial regions of PFC (e.g., orbitofrontal cortex, which is involved in the flexible reappraisal of the affective or motivational significance of stimuli) (e.g., Happaney, Zelazo, & Stuss, 2004; Zelazo & Müller, 2002).

The distinction between hot and cool executive function is similar in some respects to the hot/cool systems distinction made by Metcalfe and Mischel (1999), but it is also fundamentally different: In the Metcalfe and Mischel (1999) framework, hot processes are bottom-up emotional influences on behavior (e.g., associated with the amygdala) that tend to undermine top-down processes. The construct of hot executive function captures the suggestion that different top-down processes are required in motivationally significant versus insignificant contexts.

The construct of hot executive function is supported by neuroscientific research on the functions of orbitofrontal cortex (e.g., Bechara, 2004; Rolls, 2004). The requirement that representations of specific stimulus-reward associations be modified is common to a wide range of measures shown to depend on orbitofrontal cortex (see Happaney et al., 2004, for a review), including measures of reversal learning (in which a learned approach-avoidance discrimination must be reversed), delay discounting (in which the value of an immediate reward must be reconsidered relative to larger delayed reward), extinction (when a previously rewarded stimulus is no longer rewarded and must now be avoided), and gambling (when what initially appears to be advantageous is revealed over time to be disadvantageous).

Lesion studies (both human and nonhuman) also show clearly that hot and cool aspects are dissociable. For example, considerable research with both adult and pediatric patients (e.g., Bechara, 2004; Eslinger, Flaherty-Craig, & Benton, 2004) has shown that patients with damage to orbitofrontal cortex are often unimpaired on classic measures of cognitive control (e.g., the WCST) but nonetheless have considerable problems in their daily lives and on measures such as the Iowa Gambling Task.

It should be noted that although hot and cool aspects of cognitive control can be dissociated in lesioned brains, they typically work together, and there is considerable overlap among the neural systems underlying hot and cool cognitive control. Right ventrolateral PFC, for example, appears to play a role in a wide range of situations, including what might be considered both hot and cool contexts (Aron, Robbins, & Poldrack, 2004). Bunge and Zelazo (2006) presented a neural model of rule use that captures the relation between relatively hot and relatively cool processes. According to this model, different regions of prefrontal cortex are involved in the use of rules at different levels of complexity (Figure 1.2). Orbitofrontal cortex furnishes simple approach-avoidance (stimulus-reward) rules and is also involved in learning to reverse these rules. The formulation and use of more complex rules that control the application of simpler rules (e.g., if color game, then if red, then it goes here) involves the recruitment of increasingly anterior regions of lateral prefrontal cortex into an increasingly complex, hierarchically arranged network of PFC regions. Higher levels in the hierarchy operate on the products of lower levels (see also Badre & D’Esposito, 2007; Botvinick, 2008; Christoff & Gabrieli, 2000; Goldberg & Bilder, 1987; Koechlin, Ody, & Kouneiher, 2003).

As rules become more complex, they also become more abstract (i.e., abstracted away from the exigencies of a situation), and this can be viewed as a shift from hotter to cooler aspects of cognitive control (see also Munakata, Snyder, & Chatham, this volume). In general, on this view, the development of prefrontal cortical circuitry proceeds in a bottom-up fashion that parallels well-documented age-related changes in the complexity of the rules that children can formulate, maintain in working memory, and use when solving problems (e.g., Zelazo, Müller, Frye, & Marcovitch, 2003).

DEVELOPMENT OF COGNITIVE CONTROL

A major impediment in the study of cognitive control and its development has been the lack measures that are suitable across a wide range of ages. Most measures of cognitive control used with young children, such as the standard version of the Dimensional Change Card Sort (DCCS; Zelazo, 2006) or the Less is More task (Carlson, Davis, & Leach, 2005), are too easy for most older children, whereas most classic neuropsychological measures of cognitive control, such as the WCST, are either too difficult for young children or inappropriate for other reasons. The creation of the Cognition Battery from the NIH Toolbox for the Assessment of Neurological and Behavioral Function (NIH Toolbox; Zelazo & Bauer, in press) is an important methodological advance that has made it easier to compare cognitive control across ages and ability levels. The NIH Toolbox includes measures of cognitive flexibility, inhibitory control, and working memory that are brief (< 5 minutes each), suitable for use in repeated trials (with minimal practice effects), and appropriate for participants age 3 to 85 years. These measures include, respectively, a version of the Dimensional Change Card Sort (Zelazo, 2006), a version of the Eriksen flanker task derived from the Attention Network Task (Rueda et al., 2004), and a List Sorting task derived from the Spanish and English Neuropsychological Assessment Scales (Mungas, Reed, Marshall, & Gonzalez, 2000). (One thing the NIH Toolbox currently lacks, however, is a measure of hot executive function.)

Results from a validation study of the NIH Toolbox (N = 476) not only confirmed that the measures are reliable and valid but also yielded valuable information about the developmental course of cognitive control. As shown in Figure 1.3, which depicts performance on the NIH Toolbox DCCS, there were two periods between the ages of 3 and 15 years during which relatively rapid age-related improvements were observed (a cubic model provided the best fit of the data, R² = .76). Although performance improved most rapidly during the preschool period, the rate of improvement was also relatively high during the transition to adolescence.

Figure 1.3 Performance on the NIH Toolbox DCCS Test across age-groups.

Pediatric data from a cross-sectional validation study of 476 individuals ages 3 to 85 years. Error bars are +/– 2 standard errors.

From NIH Toolbox Cognition Battery (CB): Measuring Executive Function and Attention, by P. D. Zelazo, J. E. Anderson, J. Richler, K. Wallner-Allen, J. L. Beaumont, & S. Weintraub, in press, Monographs of the Society for Research in Child Development. Reproduced with permission.

By using the same measures at different ages, from ages 3 to 85 years, it was also possible to examine whether there are age-related changes in the relations among measures (Mungas et al., in press). In general, there was good evidence of the increasing differentiation of cognitive control from other aspects of cognitive function, consistent with a characterization of neurocognitive development as interactive functional specialization (Johnson, 2011). The pattern has also been seen within cognitive control, in research using different measures at different ages. For example, whereas a three-factor model fits data from adults (e.g., Miyake et al., 2000), data from preschool-age children are more consistent with a one-factor model (e.g., Wiebe et al., 2011; Wiebe, Espy, & Charak, 2008).

THEORETICAL APPROACHES TO THE DEVELOPMENT OF COGNITIVE CONTROL

In recent years, research on cognitive control in childhood has emphasized a number of different underlying neurocognitive processes (Carlson, Zelazo, & Faja, 2013), with some researchers addressing the role of working memory (e.g., Morton & Munakata, 2002), others focusing on inhibitory control (e.g., Kirkham, Cruess, & Diamond, 2003), and still others emphasizing different processes altogether. Kloo and Perner (2003), for example, suggested that young children’s cognitive control depends crucially on the conceptual understanding that an object can be seen from two different perspectives or that the object can be redescribed (e.g., that a red boat can be described either by its shape or by its color).

As noted, there is also interest in the importance of rule use in cognitive control (e.g., Bunge & Zelazo, 2006). According to the cognitive complexity and control theory-revised (CCC-r), developmental changes in cognitive control result from an increase in the hierarchical complexity of rules that children can formulate, keep in working memory, and use (Zelazo et al., 2003). On this view, cognitive flexibility, inhibitory control, and working memory all depend on the iterative reprocessing of information, which permits the formulation of more complex rules that can then be used to control behavior (e.g., Zelazo & Cunningham, 2007).

Research in developmental cognitive neuroscience has provided support for all of these approaches, and indeed, it now seems likely that various processes, including working memory, inhibitory control, reflection, rule use, and conceptual changes all play a role within the context of a complex, dynamic (developing) neurocognitive system. Key changes in such a system have been modeled in terms of changes in the long and short range functional interactions among neuronal populations tuned to represent higher-order rules as well as information from the external world (shape, color, location) (see Spencer & Buss, this volume).

NEUROPLASTICITY AND COGNITIVE CONTROL

Although it is clear that cognitive control is highly heritable (Friedman et al., 2008; Lee et al., 2012), there are also many naturally occurring environmental correlates of cognitive control. Among those correlates likely to have a causal influence are socioeconomic status (SES; e.g., Noble, Norman, & Farah, 2004) and parenting style (Bernier, Carlson, & Whipple, 2010). Genetic and environmental influences interact dynamically (over time) to yield cognitive control phenotypes, and it will be of considerable interest to examine the bidirectional causal pathways, including epigenetic changes, underlying this interaction.

In contrast to the old-fashioned idea that development unfolds as a function of genetically programmed maturation (e.g., Gesell, 1933), research in developmental neuroscience suggests that neurocognitive development can be seen as a dynamic process of adaptation wherein neural systems are constructed in a largely use-dependent fashion. When we use our brains in particular ways, the neural circuits upon which we rely become more efficient. Fibers connecting regions within a network (and between networks) are myelinated when used, and unused synapses are pruned. The human brain is an inherently plastic organ, continually adapting to its environment, but there are periods of relatively high plasticity (often called sensitive periods) when particular regions of the brain and their corresponding functions are especially susceptible to environmental influences. These periods typically correspond to times of rapid growth in those regions and functions, when relevant neural regions are adapting especially rapidly to structure inherent in the environment (Huttenlocher, 2002).

Because cognitive control undergoes a particularly rapid transformation during early childhood, the preschool period may be a window of opportunity for the cultivation of fundamental cognitive control skills via well-timed, targeted scaffolding and support. Indeed, research on interventions in the preschool period has now shown clearly that even relatively brief interventions targeting cognitive control not only change children’s behavior but also change children’s neural structure and function (Diamond, this volume). For example, Rueda, Rothbart, McCandliss, Saccomanno, and Posner (2005) improved 4- and 6-year-olds’ performance on a computerized attention task with five training sessions using computerized games. Children in the training condition showed improvement on an attention task and a measure of general intelligence, as well as more adult-like patterns in the N2 component of the event-related potential (ERP), located over frontoparietal and prefrontal areas. The N2 (e.g., Botvinick, 2007; Botvinick, Braver, Barch, Carter, & Cohen, 2001; Lahat, Todd, Mahy, & Zelazo, 2010; Nieuwenhuis, Yeung, Van Den Wildenberg, & Ridderinkhof, 2003; Rueda et al., 2004; Waxer & Morton, 2011; Yeung & Nieuwenhuis, 2009) has consistently been associated with anterior cingulate cortex (ACC)-mediated detection of conflict in a variety of cognitive control paradigms, including Go-No-Go tasks, flanker tasks, and the DCCS. Lamm, Zelazo, and Lewis (2006) found that the reductions in N2 amplitude typically seen as children get older were better predicted by performance on independent measures of cognitive control than by age per se, and children who perform well on the DCCS show smaller N2 amplitudes than same-aged children who perform poorly (Espinet, Anderson, & Zelazo, 2012).

Espinet, Anderson, and Zelazo (2013) provided evidence that cognitive control can be modified through even briefer exercises that encouraged children to reflect on more aspects of the context in which they were responding. These authors assigned children who failed the DCCS to one of three conditions: an experimental condition that consisted of reflection training, and two control conditions consisting of minimal feedback training or mere practice. Children who received reflection training showed significant improvement in performance on a subsequent administration of the DCCS, unlike children in the two control conditions, and they also showed reductions in N2 amplitude. Children who pass the DCCS may resolve the conflict inherent in the task more efficiently than children who fail, resulting in smaller N2 amplitudes. One possibility is that for these children, the detection of conflict initiated reflection and higher-order rule use (mediated by lateral prefrontal cortical networks) effectively resolved the conflict inherent in the stimuli and down-regulated ACC activation (cf. Botvinick et al., 2001).

This example also illustrates another important characteristic of cognitive control: There is a dynamic interaction between top-down cognitive processes and a wide range of more bottom-up influences on behavior. Relatively rapid, automatic, bottom-up neurocognitive responses (e.g., the N2-indexed ACC response to conflict) appear to influence relatively slow, voluntary, top-down cognitive control processes (e.g., by triggering the PFC activation underlying reflection), and these processes, in turn, appear reciprocally to influence the more bottom-up influences (e.g., reduction in N2 amplitude). Blair’s longitudinal research on executive function and stress/stress reactivity (e.g., Blair, this volume) addresses another aspect of this dynamic interaction.

Although the preschool years may be an especially sensitive period for EF, there is also considerable reorganization of prefrontal systems during the transition to adolescence, when gray matter volume in prefrontal cortex reaches a peak (Giedd et al., 1999). Prefrontal cortical plasticity is clearly not limited to the preschool period, and an example of a successful intervention with older children and adults is CogMed, designed to train working memory. Following 5 weeks of training, Klingberg et al. (2005) found improved working memory and reduced attention deficit/hyperactivity disorder (ADHD) symptomatology in a group of 7- to 12-year-olds with ADHD. In a study of CogMed with adults, Olesen, Westerberg, and Klingberg (2003) found training-related changes in activity in cortical regions known to be involved in working memory (i.e., increases in activity in frontal and parietal areas, as well as decreases in activity in cingulate cortex).

TOWARD THE PERSONALIZATION OF INTERVENTIONS

Interventions targeting cognitive control have the potential to help children who are at risk for a wide range of deleterious developmental outcomes, and there is a growing interest in how to make these interventions more effective. In the three experiments described by Espinet et al. (in press, 2013), for example, approximately half of the children who received reflection training responded positively (i.e.,