Elsevier

Acta Psychologica

Volume 115, Issues 2–3, February–March 2004, Pages 167-183
Acta Psychologica

Executive function across the life span

https://doi.org/10.1016/j.actpsy.2003.12.005Get rights and content

Abstract

The development and determinants of executive function (EF) were studied in children (mean age=8.8 years), young adults (M=22.3 years), and elderly adults (M=71.1 years). EF was indexed by perseverative responding on two bidimensional sorting tasks (Visually Cued Color-Shape task and Auditorily Cued Number-Numeral task), and age-related changes in EF were considered in relation to estimates of conscious vs. unconscious memory that were obtained using the process dissociation procedure (PDP). Results revealed the rise and fall of EF across the life span, with significant quadratic trends found for performance on both sorting tasks and for the conscious recollection component (C) of the PDP task. Regression analyses indicated that PDP estimates of conscious memory accounted for variation in performance on the visual sorting task, but not on the auditory sorting task. The findings are discussed in terms of their implications for hierarchical models of EF and its development.

Introduction

Young children are often aptly characterized as stimulus bound, concrete, present-oriented, and impulsive (e.g., Inhelder & Piaget, 1964). As they develop, however, they are able increasingly to represent multiple aspects of a problem, plan a future course of action, keep that plan in mind and act on it, and detect and use information about errors. This growing ability to engage in deliberate, goal-directed thought and action, which depends on the increasing effectiveness of such processes as selective attention, working memory, and inhibitory control, is often studied under the rubric of executive function (EF). By now, a considerable body of research shows convincingly that there are systematic, age-related improvements in EF during childhood and into adolescence (for review, see Zelazo & Müller, 2002). It is also clear that EF declines during aging (see Mayr, Spieler, & Kliegl, 2001; McDowd & Shaw, 2000), suggesting that the development of EF follows an inverted U-shaped curve when considered across the life span (Dempster, 1992).

Inverted U-shaped developmental curves (or U-shaped curves, depending on whether the dependent variable is positively or negatively valued) have been documented for a variety of basic cognitive processes, such as processing speed and short-term memory (e.g., see Kail & Salthouse, 1994). However, relatively few studies have measured EF across a wide range of ages. An early study by Comalli, Wapner, and Werner (1962) used the Stroop Color-Word Task––a classic measure of EF––with participants ranging in age from 7 to 80 years. These authors found the largest Stroop interference effect among 7-year-olds, with the magnitude of the effect declining until late adolescence, remaining constant through young adulthood, and then increasing again for the oldest group of adults (65–80 years). More recently, Cepeda, Kramer, and Gonzalez de Sather (2001) examined task switching in individuals from 7 to 82 years. Task switching arguably provides a measure of participants’ ability to adopt and change a problem-solving set––a key aspect of EF. In a series of trials, participants were shown either one or three numerical ones or threes (i.e., 1, 111, 3, or 333) and required to classify these stimuli differently depending on a cue (i.e., they were required either to indicate which numeral was displayed or to indicate how many numerals were displayed). A U-shaped function was obtained for switch costs––the increase in reaction time (RT) on switch trials compared to non-switch trials. Cepeda et al. (2001) also found evidence that life span changes in switch costs could be attributed primarily to changes in the time needed to prepare for a new task, as opposed to changes in the decay rate of a previous task (i.e., task set inertia; Allport, Styles, & Hsieh, 1994).

In contrast to these studies, Williams, Ponesse, Schachar, Logan, and Tannock (1999) failed to find evidence of U-shaped age-related changes on another well-established measure of EF: stop-signal reaction time. In the stop-signal procedure, participants are presented with a series of stimuli and told to press one of two keys depending on whether an X or an O appears, unless they hear a tone (the stop signal), in which case they are to refrain from responding. These authors tested individuals ranging from 6 to 81 years of age, and while they found improvement between the youngest group (6–8 years) and the middle childhood group (9–12 years), there was no evidence of an age-related increase in stop-signal RT during adulthood (although there was evidence of age-related slowing on go-signal RT). Subsequent work, however, did reveal U-shaped changes in stop-signal RT on a modified stop-signal task in which participants were required to stop when they heard one tone but not when they heard another (Bedard et al., 2002).

The differential sensitivity of different measures of EF to developmental changes may provide useful information about which aspects of EF change across the life span, and about the nature of EF itself. EF is a functional construct, and as such, it is defined solely in terms of its behavioral outcome: deliberate, goal-directed thought and action. The actual mechanisms that make EF possible are likely varied, and they remain a matter of considerable debate.

One approach to understanding EF and its development during childhood is the Cognitive Complexity and Control (CCC) theory (Zelazo & Frye, 1998), according to which age-related changes in EF can be attributed to changes in the maximum hierarchical complexity of the rules that children can formulate and use when solving problems. Age-related changes in maximum rule complexity are, in turn, made possible by biologically determined developmental increases in the degree to which children can consciously reflect on the rules they represent (i.e., age-related increases in highest level of consciousness that children can muster in response to situational demands; Zelazo, 2004). According to this approach, 3-year-olds can easily integrate two rules (e.g., “If red then here; if blue then there”) into a single rule system (Zelazo & Reznick, 1991). However, 3-year-olds have difficulty reflecting on these rules and consequently cannot switch flexibly between incompatible pairs of rules (e.g., “If sorting by color, then red goes here and blue there. If sorting by shape, then car goes here and flower goes there”; Frye, Zelazo, & Palfai, 1995). Reflection on lower-order rules is required in order to consider them in contradistinction to other, incompatible rules and embed them under higher-order rules. Higher-order rules are needed in order to select the appropriate lower-order rules. Characteristic failures of EF, such as perseveration and knowledge-action dissociations, are likely to occur until incompatible rule systems are integrated into a single, more complex rule system via a higher level of reflection or re-entrant processing.

This approach to EF can be extended to account for the impairments in EF associated with aging. Although elderly adults are capable of high levels of conscious reflection and capable of formulating and using high-order rules, doing so is likely to be resource-demanding and effortful, as is maintaining rules in working memory so that they can be used to constrain inferences and guide behavior (Braver, Barch, Keys, et al., 2001). Extending CCC theory in this way is compatible with recent proposals by Craik, 2002a, Craik, 2002b. According to Craik, knowledge may be represented as a hierarchy of levels of representation, with higher levels corresponding to more abstract representations and lower levels corresponding to more specific representations (e.g., specific details of an event). Consideration of the overlap between this approach, formulated to understand the effects of aging, and CCC theory, formulated to understand child development, prompts the following set of suggestions.

Children and older adults both show poorer performance relative to young adults on “working memory” and “executive function” tasks. Very young children simply cannot reflect on lower-order rules and cannot construct superordinate rules that govern the appropriate selection of a lower-order rule when different lower-order rules result in different responses (Zelazo & Frye, 1998). In contrast, older children and young adults can construct increasingly higher-order rules, but they may have difficulty doing so on the fly, and even when successful, they may have difficulty holding the higher-order rule in working memory, resulting in perseveration on a prepotent lower-order rule. We assume that such complex cognitive processing requires considerable expenditure of attentional resources, whose availability depends on the integrity of the frontal lobes (Craik & Grady, 2002) and the dopamine system (Braver et al., 2001). These biological systems decline in efficiency in the course of normal aging, with the result that older adults may need extra time to access and reflect on higher-order representations in the ‘levels of consciousness’ hierarchy (Zelazo, 2004). Moreover, because many situations will be familiar to older adults (in contrast to young children), older adults may be more likely to access higher-order representations that are pre-formed, context-bound, and hence, relatively inflexible. For all these reasons, the necessity to switch rapidly between sets of rules will be difficult for older adults, resulting again in perseverative errors. In fact, age-related decrements in some but not all aspects of task-switching have been reported (Meiran, Gotler, & Perlman, 2001), and older adults do make more perseverative errors on the Wisconsin Card Sorting Task (WCST; Craik, Morris, Morris, & Loewen, 1990; Grant & Berg, 1948; Heaton, 1981). More generally, older adults will exhibit difficulty in the intentional modulation of levels of consciousness and in the ability to navigate knowledge hierarchies flexibly and effectively. Some levels of representation may be difficult to access at all; for example Craik, 2002a, Craik, 2002b has suggested that in memory tasks older adults often fail to retrieve information from lower levels associated with specific knowledge (e.g., names) and with specific contextual details. The present suggestion implies that they may also fail to intentionally access and utilize higher-level representations––needed, for example, to understand analogies, generalize old knowledge to new situations, and switch flexibly between sets.

From this perspective, age-related changes in EF across the life span can be understood in terms of corresponding changes in the ability to formulate higher-level hierarchical representations in childhood and to consciously select and maintain them in aging. One purpose of the present study was to document age-related changes in EF using the same measures of EF in participants ranging from children to elderly adults. A second purpose was to examine whether age-related changes can be accounted for by changes in conscious control.

To assess EF, we relied on two sorting tasks, based on existing measures of EF such as the WCST, the Dimensional Change Card Sort (DCCS; Frye et al., 1995), and various measures of task switching (e.g., Goschke, 2000; Rogers & Monsell, 1995). Performance on the WCST shows considerable improvements in performance over the school age years (e.g., Chelune & Baer, 1986) but older adults perform less well than young adults (Craik et al., 1990; Raz, 2000). However, the WCST is an inductive, hypothesis-testing task that taps numerous aspects of EF simultaneously, and, as a result, the origin of errors on this task is difficult to determine (e.g., see Delis, Squire, Bihrle, & Massman, 1992). For example, perseveration could occur after a rule change in the WCST either because a new rule was not hypothesized, was hypothesized but not selected, or was selected but not acted upon. In contrast, the DCCS, which has been used with preschool age children, is relatively simple, and errors are consequently relatively easy to interpret. In this deductive rule-use task, children are shown two bivalent, bidimensional target cards (e.g., depicting a blue rabbit and a red boat), and they are told to match a series of test cards (e.g., red rabbits and blue boats) to these target cards first according to one dimension (e.g., color) and then according to the other (e.g., shape). Regardless of which dimension is presented first, 3-year-olds typically perseverate by continuing to sort cards by the first dimension after the rule is changed.

The new sorting tasks were deductive rule-use tasks, like the DCCS, and they were designed to be suitable across a wide range of ages. One of the new sorting tasks was the Visually Cued Color-Shape task, in which participants were shown four target cards and asked to sort test cards either by color or shape, depending on a visually presented cue (an X or a Y). Forty out of 50 trials were cued as color trials; the remaining 10 trials were shape trials, and they occurred as single trials unpredictably throughout the sequence. The preponderance of color trials was intended to induce a strong set to sort by one dimension. The second new sorting task was the Auditorily Cued Number-Numeral task, in which participants viewed a 2 × 2 grid where each quadrant contained either one, two, three, or four small squares and also a numeral 1, 2, 3, or 4; the number of squares in a quadrant did not correspond to the numeral. Four keys in the same configuration as the grid were used for responding. The stimuli and cues for responding were presented as numbers by a male or female voice; if the number was spoken by one voice, the rule was “respond by pressing the key corresponding to the quadrant containing that numeral,” if the number was spoken by the other voice, the rule was “respond by pressing the key corresponding to the quadrant containing that number of small squares.” Again, 40 trials were cued to one dimension and the remaining 10 trials were cued to the other dimension. Fig. 1 depicts the displays for both sorting tasks.

To assess conscious control––particularly as it contributes to memory––independently of EF, we relied on the process dissociation procedure (PDP) suggested and elaborated by Jacoby and his colleagues (Jacoby, 1991; Jacoby, Toth, & Yonelinas, 1993). The basic idea is that the relative contributions of consciously controlled and automatic influences on behavior may be estimated by comparing performance when the two processes work together, to performance when the processes are set in opposition to each other. For example, in one paradigm used by Jacoby and colleagues, participants first studied two lists of words, one presented visually (seen) and another presented auditorily (heard). Participants were then given a series of word stems to complete. Under one set of instructions (inclusion instructions), participants were told to try to complete the stems with any word, seen or heard, that they had encountered in the presentation phase. Under another set of instructions (exclusion instructions) they were told to complete only those word stems that could be completed to make a word they had heard; stems of previously seen words or completely new words were to be left uncompleted. When seen word stems are considered, they could be completed under inclusion instructions either because the participant consciously recollected that it had been presented visually, because it was on the original list (without recollecting its presentation modality), or simply because it felt familiar. In this case automatic and controlled processes work in concert. On the other hand, if a seen stem is completed under exclusion conditions, it must mean that the participant does not recollect that it was presented visually; that is, the automatic influence to complete it is unopposed by any consciously controlled influence.

By comparing the probability of completing seen stems when instructed to do so (inclusion) versus when instructed not to do so (exclusion), it is possible to derive estimates of the extent to which performance on the task is determined by controlled (C) versus automatic (A) processes. On the assumption that C and A are independent processes, the probability of using a previously seen word in the inclusion condition is taken to reflect the additive influences of C and A, minus the overlap, that is:p(Inclusion)=C+A−CAorC+A(1−C)If a previously seen word is used (in error) in the exclusion condition, the assumption is that controlled processes are not operating in that portion of the trials (thus represented by 1−C) and that behavior is controlled by automatic influences only. That is:p(Exclusion)=A(1−C).From these two equations, estimates of C and A can easily be derived (see Jacoby, 1991).C=InclusionExclusionA=Exclusion/(1−C)

Errors produced in the exclusion condition, referred to as action slips, indicate that automatic processes exert more influence on behavior than controlled processes. Moreover, as noted, it is assumed that action slips occur only given a failure of controlled processes. As Jacoby and Kelley (1992) state, “In the exclusion condition, a studied word will be produced only when there is a failure to consciously remember that it was on the list” (p. 176). While this assumption may be reasonable for healthy young adults (although even this is debatable), it seems unlikely to hold in children and in the elderly. Children and the elderly may have more difficulty than young adults in bringing consciously accessed information to bear on a situation when this information conflicts with response tendencies. For example, 3-year-olds can repeat both the pre- and the post-switch pairs of rules in the DCCS (and hence, represent them at one level of consciousness), but they nonetheless fail to use them (Zelazo, Frye, & Rapus, 1996). When given exclusion instructions, children may sometimes fail to exclude words that they could nonetheless correctly categorize as seen. By this analysis there is therefore an interesting category of genuine or dissociated action slips, which occur when a participant makes an action slip in the exclusion condition despite conscious recollection. Indeed, in contrast to the process-dissociation approach, which treats controlled processes as an all-or-none phenomenon, developmental data argue for a more nuanced notion of age-related levels of consciousness (Zelazo, 2004), and a distinction between levels may be revealed by the occurrence of genuine action slips.

In the current study, participants were required to study lists of words presented auditorily or visually, and then complete a list of word stems under both inclusion and exclusion conditions. Following each condition, participants were shown the study lists and asked to identify the modality in which words were originally presented. This measure was designed to provide a preliminary assessment of the frequency of genuine action slips.

In summary, the purpose of the present study was to examine age-related changes in EF across the life span using a common set of measures, and explore the extent to which these changes could be understood in terms of corresponding changes in conscious control. School-age children (between 8 and 9 years), young adults, and elderly adults were tested on two sorting tasks based on the DCCS, and estimates of conscious and automatic influences on memory were obtained using the PDP, which involved a word stem completion task administered under both inclusion and exclusion conditions. Eight- to nine-year-old children were selected because pilot testing raised questions about whether all of the tasks were suitable for younger children.

Relative to young adults, both children and elderly adults were expected to make more perseverative errors on the sorting tasks. Children and elderly adults were also expected to exhibit lower estimates of conscious control on the word stem completion task, and to produce more genuine action slips in the exclusion condition of this task. We also expected that variations in EF as measured by sorting could be accounted for by variations in conscious contributions to memory. Together these results would support the suggestion that the development of EF across the life span follows an inverted U-shaped function, and encourage efforts to understand EF in terms of underlying changes in the ability to consciously set up, maintain, and access representations at the appropriate level of complexity.

Section snippets

Participants

There were 20 participants in each of three groups: 8- to 9-year-old children (mean age=8.8 years, range: 8.2–9.5 years), young adults (mean age=22.3 years, range: 19.5–26.6), and elderly adults (mean age=71.1 years, range: 65.8–74.2). Ten of the children were male, as were eight of the participants in each of the other groups. Children and elderly adults were recruited from databases containing names of individuals who had expressed an interest in participating in research. All young adults

Results

Analyses were conducted to address four questions: (1) Is EF, as measured by the two sorting tasks, an inverted U-shaped function of age across the lifespan? (2) Do conscious contributions to memory, estimated via the PDP, also show an inverted U-shaped pattern? (3) Can estimates of conscious memory help explain age-related changes in EF? And finally, (4) are children more likely than adults to commit genuine action slips in the exclusion condition of the PDP?

Discussion

The present study was designed to examine age-related changes in EF across the life span using two bidimensional sorting tasks, and to explore the extent to which these changes could be understood in terms of corresponding changes in conscious versus automatic influences on memory.

As expected, perseverative errors on both sorting tasks exhibited a U-shaped function when plotted against age. On the Visually Cued Color-Shape task, both children and older adults made more perseverative errors than

Acknowledgements

This research was supported by grants from the Natural Sciences and Engineering Research Council (NSERC) of Canada to P.D. Zelazo and F.I.M. Craik. The authors would like to thank Jennie Sawula and Dana Liebermann for their help in preparing this article, and Ellen Bialystok for helpful comments on a previous draft.

References (42)

  • L.J Anooshian et al.

    Conscious and unconscious retrieval in picture recognition: A framework for exploring gender differences

    Journal of Personality and Social Psychology

    (1996)
  • A.-C Bedard et al.

    The development of selective inhibitory control across the life span

    Developmental Neuropsychology

    (2002)
  • T.S Braver et al.

    Context processing in older adults: Evidence for a theory relating cognitive control to neurobiology in healthy aging

    Journal of Experimental Psychology: General

    (2001)
  • N.J Cepeda et al.

    Changes in executive control across the lifespan: Examination of task-switching performance

    Developmental Psychology

    (2001)
  • G.J Chelune et al.

    Developmental norms for the Wisconsin Card Sorting Test

    Journal of Clinical and Experimental Neuropsychology

    (1986)
  • P.E Comalli et al.

    Interference effects of Stroop Color-Word Test in childhood, adulthood, and aging

    The Journal of Genetic Psychology

    (1962)
  • F.I.M Craik

    Human memory and aging

  • F.I.M Craik

    Levels of processing: Past, present … and future?

    Memory

    (2002)
  • F.I.M Craik et al.

    Aging, memory, and frontal lobe functioning

  • F.I.M Craik et al.

    Relations between source amnesia and frontal lobe functioning in older adults

    Psychological Aging

    (1990)
  • T Goschke

    Intentional reconfiguration and involuntary persistence in task switching

  • Cited by (428)

    View all citing articles on Scopus
    1

    Present address: Rotman Research Institute, Baycrest Centre.

    2

    Present address: Department of Psychology, York University.

    View full text