The effect of exercise-induced arousal on cognitive task performance: A meta-regression analysis

Abstract

The effects of acute exercise on cognitive performance were examined using meta-analytic techniques. The overall mean effect size was dependent on the timing of cognitive assessment. During exercise, cognitive task performance was impaired by a mean effect of -0.14. However, impairments were only observed during the first 20 min of exercise. Otherwise, exercise-induced arousal enhanced performance on tasks that involved rapid decisions and automatized behaviors. Following exercise, cognitive task performance improved by a mean effect of 0.20. Arousal continued to facilitate speeded mental processes and also enhanced memory storage and retrieval. Positive effects were observed following exercise regardless of whether the study protocol was designed to measure the effects of steady-state exercise, fatiguing exercise, or the inverted-U hypothesis. Finally, cognitive performance was affected differentially by exercise mode. Cycling was associated with enhanced performance during and after exercise, whereas treadmill running led to impaired performance during exercise and a small improvement in performance following exercise. These results are indicative of the complex relation between exercise and cognition. Cognitive performance may be enhanced or impaired depending on when it is measured, the type of cognitive task selected, and the type of exercise performed.

Introduction

Individuals often describe changes in their ability to perform mental tasks during and after exercise. For some, exercise leads to reports of increased mental acuity and clarity of thought. Others report feelings of mental disorientation and difficulty making decisions following exercise. The relation between individual bouts of exercise and cognitive function has been examined in more than 150 empirical studies over the past 50 years and the results summarized in several narrative reviews (Brisswalter et al., 2002, McMorris and Graydon, 2000, Tomporowski, 2003, Tomporowski and Ellis, 1986). Only one quantitative review of the acute exercise literature has been published (Etnier et al., 1997), which provided tentative support for a causal relation between exercise and cognitive function. A number of studies have been conducted in the past decade that provide a larger database from which theory-based hypotheses concerning the effects of exercise on brain and cognitive function can be evaluated.

Much of the research on the relation between exercise and cognition has tested predictions drawn from “arousal” theories (e.g., Yerkes and Dodson, 1908, Kahnemann, 1973, Humphreys and Revelle, 1984, Sanders, 1986, Hockey et al., 1986). Common to these theories is the assumption that cognitive performance is dependent on the allocation of energetical resources to meet task demands. Acute exercise has been hypothesized to alter brain systems that influence how mental resources are dedicated to cognitive task performance (Audiffren, 2009). While the evaluation of exercise-induced arousal has been at the center of the majority of published studies, the methods used to manipulate participants' arousal levels have not been uniform. Three different approaches have been taken to test predictions drawn from the arousal hypothesis. One approach has been to use experimental protocols that mimic the exercise regimens prescribed to recreational runners or cyclists. These protocols involve steady-state cardiorespiratory exercise designed to improve mood and increase feelings of well-being. The aerobic-running movement that emerged in the 1970s was grounded on purported physical and psychological health benefits of steady-state exercise (Folkins and Sime, 1981). The existence of these benefits has been supported by empirical evidence, in which individuals report positive changes in affect following moderate levels of steady-state exercise lasting at least 20 min. Predictions for experiments that have examined cognition following exercise have been driven by this finding. Maintenance of exercise-induced arousal was expected for a short period following exercise and cognitive performance was also expected to be facilitated during this time.

A second approach researchers have taken to examine the exercise–cognition relation has been to model experimental protocols on predictions generated from the inverted-U hypothesis (Yerkes and Dodson, 1908) and other arousal theories (e.g. Humphreys and Revelle, 1984). Typically, cognitive performance was measured at multiple points during exercise that systemically altered participants' level of physiological arousal as measured by heart rate, oxygen uptake, RPE, or other biological indices. Cognitive performance was predicted to improve and peak as physiological arousal increased and then deteriorate as arousal levels approached maximal levels (see McMorris and Graydon, 2000).

Yet another approach to examine the influence of exercise on cognition has been to focus on the fatigue producing aspects of physical activity. Human factor researchers have long had an interest on the debilitating effects of physical fatigue on operational performance and exercise has been used to induce fatigue in many studies. Experimental protocols employed in these studies typically require participants to complete incremental-load exercise to voluntary exhaustion or to maintain a physically demanding steady-state exercise protocol for an extended period of time. In such studies, it has been predicted that participants' cognitive performance would be impaired both during and immediately following the termination of exercise.

Thus, depending on the approach taken, researchers have expected exercise to either facilitate or impair cognitive function. Compounding attempts to elucidate the relation between exercise and cognitive function have been the wide range of mental tasks that have been employed in these studies. Tasks range from those that measure basic processes such as perceptual organization, information-processing speed, and simple- and choice-response time, to tasks that measure memory and high-level executive control processes. The different approaches taken by researchers, coupled with the wide variation in outcome measures, have made efforts to summarize and synthesize research findings difficult. Several authors have reviewed the results of these studies in a narrative fashion. In an early review, Tomporowski and Ellis (1986) classified studies according to the intensity and duration of the acute exercise intervention. The authors concluded that the empirical evidence did not provide support for the notion that exercise influenced cognition. They pointed to individual difference factors as a potential explanation, as well as a lack of a cohesive theory driving the research. In a later review, Tomporowski (2003) classified studies based on the intensity and duration of exercise and interpreted the results in terms of information-processing theory. Based on the available evidence, it was concluded that acute exercise had selective effects on cognitive processing. Exercise appeared to facilitate certain aspects of processing such as response speed and accuracy and enhanced the processes involved in problem-solving and goal-oriented actions. This was particularly true when a task involved inhibition of a response, one component of executive function. In contrast, exercise appeared to have no effect on tasks that measured perceptual processing, sensory processing, or memory retrieval.

Other narrative reviews have targeted studies that employed specific exercise protocols. McMorris and Graydon (2000), for example, evaluated only those studies that were designed to test hypotheses derived from the inverted-U relation (Yerkes and Dodson, 1908). As mentioned, these researchers have predicted that cognitive test performance would rise to an optimal level as exercise-induced arousal increased and then decline with higher levels of physiological arousal. A similar approach to selecting research studies for evaluation was taken by Brisswalter et al. (2002). The conclusion drawn from these two narrative reviews was that certain aspects of cognitive performance improved during and after acute exercise. All three of these narrative reviews echoed a similar overall theme—the results of studies were “equivocal.”

Meta-analytic procedures provide an alternative method to examine the impact of acute exercise on cognition in terms of the magnitude and direction of its effect. While narrative reviews of acute exercise are informative, they do not permit a detailed examination of methodological differences that exist among studies and how these differences influence the size of the overall effect. To estimate the overall effect, individual effect sizes are calculated for the relevant studies and are aggregated in a systematic manner. An effect size is a statistic that captures the vital quantitative information from each relevant study finding. Effect sizes are standardized such that the values from different studies can be interpreted in a consistent way (Lipsey and Wilson, 2001, pp. 3–4). Meta-analytic procedures also permit examination of study and sample characteristics that moderate the size of the effect. An advantage of meta-analytic techniques compared to traditional narrative reviews is that the results are independent of probability values and levels of statistical significance, which are largely dependent on sample size (Lipsey and Wilson, 2001, p.6).

Meta-analytic methods have been used to examine the relation between acute bouts of exercise and cognition. Etnier et al. (1997) derived 852 effects from 134 studies that examined the influence of both acute bouts of exercise and chronic exercise training on cognitive functioning. Of these effects, 371 represented the impact of acute exercise on cognition. The authors found that acute exercise had an overall small but significant positive (ES = 0.16) effect on cognition. Several moderator variables affected the magnitude of effect sizes, and those most relevant to the present work include: the type of cognitive task, the sampling method, and the number of threats to internal validity. Post hoc analyses were performed to evaluate the role of cognitive task type and revealed larger effects for studies that used measures of simple reaction time than studies using choice- or discriminant-reaction time tasks. Study effect sizes were found to increase as the rigor of the experimental methods decreased, suggesting that poorly designed studies yielded larger effects.

However, the interpretation of the analyses conducted by Etnier et al. (1997) is limited by several methodological factors. First, effect sizes were derived from different study designs (correlational, between-subject, within-subject, etc.) and combined for analysis. The merits of limiting meta-analytic procedures to effects sizes derived from comparable research designs have been described (Lipsey and Wilson, 2001 p. 45). Second, methods to adjust for the small sample sizes of the studies were not employed. Weighting each effect based on sample size is important because larger studies provide more reliable parameter estimates of experimental effects than do studies with fewer participants. Each study's contribution to the meta-analysis should be proportionate to its reliability (Lipsey and Wilson, 2001, p. 106). Third, the analysis was limited to a fixed-effects model, which considers only within-study variability. Contemporary researchers suggest that a random-effects model is preferable to a fixed-effect model because it accounts for between-studies heterogeneity. This heterogeneity is associated with the sampling error within studies as well as the shared random effects variance (Lipsey and Wilson, 2001, pp. 124–125, 140–142, 216–220).

The wide differences that exist in exercise interventions, populations studied, and types of cognitive tasks employed by researchers in this area warrant the use of a random effects model to aggregate the studies. This method may provide a more sensitive analytic approach than the methods used in previous studies. The present review focused exclusively on acute-exercise studies in which young adults' cognition was measured during or after single bouts of exercise using a within-subjects, repeated-measures design. The dual-task conditions of research protocols used to measure cognition during exercise (e.g., concurrent running and cognitive testing) differ theoretically from single-task protocols used to measure cognition following exercise (e.g., cognitive test only). Given these fundamental differences in attentional demands, studies that measured cognitive function during exercise were analyzed separately from those that measured cognitive function following exercise.

Selection of potentially moderating variables was guided by previous narrative and quantitative reviews and by contemporary cognitive theories. First, the relation between acute exercise and cognition was expected to be dependent on the exercise intensity and duration requirements placed on participants. As described previously, researchers have employed exercise protocols designed with a priori assumptions that interventions would either facilitate or degrade cognitive test performance. Second, the time at which cognitive tests were administered was expected to influence effect size, with larger effects expected during exercise as the length of the exercise bout increased and the effects dissipating gradually following termination of exercise. The third hypothesis focused on exercise mode. Ergometer cycling and treadmill running exercise protocols have been used most frequently but relatively little distinction has been made between the two modalities. However, the attentional demands required to maintain a desired treadmill running pace might be greater than the demands when seated on a cycle ergometer. As such, fewer attentional resources would be available to runners to perform cognitive tasks than cyclists under dual-task conditions. Effect sizes were predicted to be smaller during exercise in studies that utilized running protocols when compared to effect sizes obtained from studies utilizing ergometer cycling protocols. The fourth hypothesis tested the supposition that acute exercise has selective effects on cognitive test performance. Effect sizes were predicted to be larger for tasks that emphasize processing speed, decision-making, and executive processing and smaller for tasks that involve memory encoding and retrieval processes. The final hypothesis addressed study-design factors. Studies with greater experimental rigor typically result in smaller effect sizes than studies with fewer controls. Thus, studies in which a resting control condition was included in the design were hypothesized to exhibit smaller effect sizes than studies that employed a pre- and post-exercise measurement design.