1. Introduction
Over the past decade, the popular, commercial, and scientific interest in physical exercise has grown. According to the International Health, Racquet and Sportsclub Association [
1], the health club industry revenue was estimated at
$87.2 billion in 2017 [
1]. Over 201,000 clubs served 174 million members around the world. In the US alone, the number of health club members in 2017 (
n = 60.9 million) has increased by 33.6%, compared to 2008 (
n = 45.6 million) [
1]. What is more, a recent ISI Web of Science search on the term “physical exercise” (PE) revealed a 269% increase of scientific papers in 2018 (
n = 9574), contrasted to those of a decade earlier (
n = 3557). Mass media coverage has gone as far as to compare the health consequences of sitting to those of smoking. A recent research analysis of news articles found nearly 300 articles claiming that “sitting is the new smoking” [
2]. Such claims, however, were found to be inaccurate since “absolute risk differences for smoking far outweigh those for sitting, except for type 2 diabetes” [
3]. Additionally, the fact that physical exercise has gained much attention over the past decade has also led to various studies regarding exercise’s effects on the human body. Interestingly, next to positive effects on health, physical exercise can also affect the human brain and cognition. However, there are competing views on the cognitive effects of physical activity and it is not clear what level of consensus exists among researchers in the field. In this study we tried to quantify the scientific consensus on the relationship between physical activity (PA) and cognitive function. In addition, we investigated if there is a gap between the public’s and scientists’ interpretations of scientific texts on this topic. We start by defining the concept of physical activity and provide scientific evidence on its effect on. Subsequently, we discuss the competing views on the effects of physical activity and its potential physiological and the psychological mechanisms.
To begin with, according to the World Health Organization (2010) [
4], physical activity is “any bodily movement produced by skeletal muscles that requires energy expenditure.” As such, PA includes any motor behavior in daily and leisure activities. Housework activities classify as a form of PA, and as such, it has been shown that they may have a greater beneficial effect on executive function compared to other physical activities, through the activation of the right ventrolateral prefrontal cortex (R-VLPFC) [
5]. PE, however, is “a sub-classification of PA that is planned, structured, repetitive, and has as a final or an intermediate objective the improvement or maintenance of one or more components of physical fitness” [
4]. Aerobic and anaerobic activity, characterized by a certain frequency, duration, and intensity, are examples of PE. Conversely, physical fitness (PF) is one’s ability to perform aspects of daily activities and sports with optimal performance, strength, and endurance [
6]. Another term that is used throughout the manuscript is “cognition” which is a set of mental processes that contribute to cognitive measures such as action, perception, intellect, and memory [
7].
Research has shown that PE might exert rather small benefits on cognitive capacities when compared to other enhancers such as caffeine, sugar, or modafinil; however, it has additional benefits, such as enhanced mental or physical health, without side effects [
8]. According to the U.S. Department of Health and Human Services (2018) [
5], PA can improve cognition and reduce the risk of depression in youth. For older adults, it can reduce symptoms of anxiety and depression, and improve cognition for those with dementia, multiple sclerosis, ADHD, and Parkinson’s disease [
5]. Lack of physical activity and physical exercise has also been associated with a higher risk of dementia among older populations [
9]. In school-age children, it was determined that physical exercise benefits academic achievement, perceptual skills, verbal and mathematical ability, and intelligence [
10]. Physical activity in the form of moderate cycling exercise can also enhance neurocognitive processing in adolescents with intellectual and developmental disabilities [
11]. Furthermore, a meta-analysis of randomized controlled trials revealed that aerobic exercise training could improve processing speed, executive function, and attention [
12]. However, the effects on working memory were less consistent [
12]. Mandolesi et al. (2018) [
13] pointed out that both chronic and aerobic PE can achieve similar benefits and that they play a role in counteracting normal and pathological aging.
Physical activity has also been identified as a protective factor against age-related cognitive decline. This notion is supported by a neuroimaging study assessing PA in individuals in their early seventies [
14]. They determined that PA preserves the structural volume in the prefrontal and temporal cortices after nine years of follow-up. Additionally, preserved grey matter volume was observed in several cortical and subcortical regions, such as the hippocampus [
14]. Another study found that aerobic training can even significantly increase hippocampal volume in older women with mild cognitive impairment [
15]. Physical activity has also been associated with changes in grey and white matter structures, metabolite concentration, and corticospinal/intracortical excitability [
16]. These changes were observed in young, adult, and elderly populations but were reversed in athletes with concussions [
16]. As in almost every other cognition-enhancing method, some people seem to benefit more than others [
8].
Scientists do not yet know the precise mechanisms of the way exercise changes the structure and function of the brain. For instance, a study by Brisswalter, Collardeau, and René (2002) [
17] points at an increase in arousal level related to physical exertion as a potential mechanism for improvement in cognitive performance during exercise. Although there is no clear functional hypothesis that explains the relationship between arousal and exercise, motivation and attention have been mentioned as important psychological mediators in this relationship [
17].
Another mechanism was uncovered by Mata, Thompson, and Gotlib (2010) [
18]. They demonstrated that a
brain-derived neurotrophic factor (
BDNF) genotype moderates the protective effect of PA on depressive symptoms in adolescent girls. The researchers tested 82 girls by using a psychological and biological test. They found that physical activity served as a protective factor for girls in the high-risk group who carry a
BDNF gene variation, called met allele. On the contrary, the girls with a different, homozygous variant, called val allele, did not benefit as much from physical activity. The interaction between the physical activity and the
BDNF gene has not manifested with respect to depression. The majority of the studies in both adolescents and adults have failed to replicate the effect when it comes to depression outcomes [
19].
A literature review by Marmeleira (2012) [
20] draws attention to multiple physiological and psychological mechanisms that support a positive relationship between PA and cognition, emphasizing the potential effects of different types of exercise. For instance, cardiovascular (aerobic) exercise has been considered to mediate the positive association between physical activity and cognition. This is also known as “the cardiovascular fitness hypothesis.” The cognitive benefits from aerobic exercise are thought to be due to BDNF, glucose availability, cerebral oxygen, changes in cerebral structure, and neurotransmitters’ levels, which themselves have been associated with improved cognitive performance [
20,
21].
Another line of research line builds on the human movement effect [
22,
23]. Using the theoretical frameworks of cognitive load theory and embodied cognition, Sweller and colleagues (2019) [
23] provided an explanation for positive effects of fine movements on cognition and learning. It is assumed that fine movements, such as making gestures during problem solving (e.g., finger counting, pointing, tracing), or gross motor movements, such as enacting to learned words, can improve learning by sharing the load between the cognitive and motor systems (i.e., cognitive offloading) [
24,
25] and by providing additional cues that can be incorporated in cognitive schemas and used in subsequent knowledge retrieval [
22]. Mavilidi et al. (2018) [
26] have suggested that for movements to be effective for learning they need to be integrated into and relevant for the learning task.
1.1. Scientific Consensus
The “scientific consensus” approach in this field is relatively new, and to date there is only one study which has employed it. An expert panel approach was proposed by Singh et al. (2019) [
27]. The authors conducted a systematic review with an international expert panel to evaluate the evidence on the effects of PA interventions on cognitive and academic performance in children. They determined that the current state of scientific literature is inconclusive, regarding the beneficial effects of physical activity interventions on cognitive and academic performance in children [
27].
A longitudinal experiment on the potential role of exercise in preventing cognitive decline found that exercise helped play a protective role in cognitive functioning in elders over time [
28]. Their subjects were rural elders, 65 years of age or older of low socioeconomic status and education. The results showed that a higher exercise level was associated with an absence of substantial cognitive decline two years later, even after adjusting for variables such as age, sex, education, previous level of cognitive function, self-rated health, and exercise frequency [
28]. This could be explained by the fact that increasing energy output from a variety of physical activities is related to larger gray matter volumes in the elderly, regardless of cognitive status [
21]. Another follow-up study in China found that people with limited physical activity had a higher risk of developing dementia [
29].
However, other longitudinal studies failed to find such an association. For instance, a 7-year prospective study in Japan discovered that neither work nor leisure PA was protective against Alzheimer’s disease for Japanese participants (
n = 828) [
30]. Additionally, a systematic review study revealed largely insufficient evidence for the effectiveness of any exercise intervention on promoting cognitive function and preventing cognitive decline in older adults [
31]. What is more, they also found that most of the trial studies were small, underpowered, and unable to assess the clinical significance of cognitive test outcomes [
31].
The above-mentioned contradictions in research findings have been addressed by mass media, such as Time magazine [
32] and STAT, produced by Boston Globe Media [
33]. Currently, there is no complete consensus regarding the effects of PE on the human brain. When evidence is uncertain or not quantified, people tend to erroneously reach conclusions about the gravity of evidence. This is a result of a well-established human information processing mechanism, called the “availability heuristic” [
34]. In such instances, a scientific consensus is a tool that serves as a form of social proof, easily comprehended by laymen and experts alike [
35]. The major stakeholders in creating and communicating consensus are scientists and non-scientists. Scientists are the ones who determine and quantify the scientific consensus, which is important, as it provides a novel methodology for assessing the scientific weight of evidence. They do that by implementing structured communication techniques and/or methods such as the Delphi method [
36,
37]. Consensus is crucial as it safeguards the public against influential misinformation. This is where the non-scientists play a role as communicating the scientific consensus has a powerful effect on realigning public views of the issue with expert opinions [
35]. For example, there have been conflicting views on the state of evidence whether brain games can improve cognitive function in daily life [
38,
39]. In an attempt to determine the state of evidence, Simons et al. (2016) [
40] reviewed literature cited by brain-training proponents and leading companies and concluded that there was not sufficient evidence to justify the claim that brain training is an effective tool for enhancing cognition in the real world. They found that such studies lacked consistency and had methodological shortcomings, such as lack of preregistration, incomplete reporting, small sample sizes, and no controls for placebo effects [
40]. These types of reviews are important because they resolve ambiguities in the current state of knowledge and explore the present state of understanding on a topic [
41]. For this reason, the present study sought to clarify the state of the scientific consensus on the effect of PA on cognition.
To date, there are two scientific consensus statements on the effect of physical activity and aging, and cognitive and academic performance [
26,
42]. Regarding aging, twenty-six researchers from nine different countries and a variety of academic disciplines met in Denmark to reach an evidence-based consensus about physical activity and its effect on older adults. According to the consensus statement, physical activity slows down age-associated cognitive decline and neurodegeneration in physically active adults. Additionally, acute moderate-intensity PA could produce short-term benefits in cognitive performance [
42]. This consensus, however, was based on a face-to-face meeting. In such meetings, there is a conformity pressure to adjust one’s own opinion to that of the group, especially in homogeneous groups [
43]. Because of this social pressure to conform with group norms, there is a risk that scientists may agree with the group consensus even though they may have different personal views.
1.2. Current Study
The goal of this consensus study was twofold: Firstly, to quantify the scientific consensus on whether physical activity has cognitive benefits. The present study distinguishes between physical activity and physical exercise, even though both terms are often used interchangeably. We further distinguish between acute and chronic effects of physical activity. Acute effects develop during short-term exposure or a single bout of physical activity, whereas repeated bouts of physical activity are needed for the chronic effects to take place [
26]. Both effects are significant, with acute effects, on one hand, expressed as improved attention and cognitive function [
13,
26]. Chronic effects, on the other hand, are associated with brain structure changes, and improved learning and memory [
13,
26].
In the current review, physical activity is the field of interest. The present work aims to report on the actual consensus in the field, regardless of the form of physical activity and of the acute/chronic effects on cognition.
This goal was explored in the first stage of the study. During the first stage, a sample of scientific literature, published over a 15-year period, was examined to determine the level of scientific consensus in the field. Each abstract was categorized per author based on the level of endorsement (explicit endorsement, implicit endorsement, neutral, implicit rejection, explicit rejection, or partial endorsement/partial rejection). To prevent conformity biases, our study was anonymized so that scientists did not feel social pressure to conform to certain expectations, and thus, not refrain from expressing their own views.
The second goal was to explore if there is a gap between the general public’s and scientists’ interpretations of scientific texts. It was hypothesized that there would be no association between the scientists’ and laymen’s interpretations of scientific abstracts. This hypothesis was based on the discrepancy between the public’s and scientists’ views on key issues [
44,
45]. In the second phase, the same scientific literature was distributed to participants with a non-academic background. Each abstract was then classified by two independent raters, and if any disagreements arose, they were resolved by a third party; namely, an arbitrator. Upon completion of the final ratings, both scientists’ and non-scientists’ ratings were compared to see if there is an association between interpretations.
Nevertheless, our study differs from the approaches adopted by Singh et al. [
27] and Bangsbo et al. [
42] in three ways: First, we adopted a theory-based consensus approach, which differs from the expert panel approach by Singh et al. (2019) [
27]. Instead of authors recommending international experts in the field of physical exercise, we contacted authors who published manuscripts related to the effects of PE on cognitive performance. This means that we contacted authors from a wide range of expertise. Secondly, our approach further investigated the degree of match-mismatch between understanding and beliefs of experts versus non-experts. Third, both experts and non-experts remained anonymous; hence, alleviating social conformity pressure.
Through analysis of physical activity-related manuscripts published from 2004 to 2019 by scientific and non-scientific participants, this study provides the first consensus analysis of its kind to quantify and evaluate the level of consensus in nearly two decades of research.
4. Discussion
In the present study we tried to determine the scientific consensus regarding the effects of physical activity on cognition. Scientific consensus is a tool that serves as a form of social proof and hence, determining and quantifying the consensus is important for evidence-based policy. Science methodology and vocabulary are repeatedly deemed inaccessible to the general public, and they often do not specify what actions can be taken by individuals and organizations alike. The present study was an attempt to help make the first step to addressing these challenges by communicating the scientific consensus and helping to realign public views with experts’ opinions.
Overall, the first stage of the study corresponded to a previous consensus statement, released by Bangsbo et al. (2019) [
42]. In particular, scientists (80%) reached a consensus that physical activity benefited cognitive processes and brain health based on the surveyed literature (76.1%), similar to a previous consensus on aerobic physical activity and cognitive function in older adults [
42]. The scientists’ ratings were similar to those of the non-scientists’ (70.4%). Despite this, however, there was a methodological difference between the study of Bangsbo and colleagues and the current study. Unlike Bangsbo et al. (2019) [
42], where scientists had a face-to-face meeting, the present study required the respondents to remain anonymous. This was done to eliminate bias and peer pressure that could occur in face-to-face meetings. When considering the second stage of the review, non-scientists were able to interpret scientific abstracts in the way scientists intended (standardized C = 0.52,
p < 0.05). The results were surprising, since it was expected that there would be no association between the scientists’ and laymen’s interpretations of scientific abstracts. A disadvantage of the contingency coefficient, however, is that its maximum possible value depends on the number of cells in the table. Therefore, for future replication studies, it is advisable to adhere to the same number of cells in the table to allow for comparability across studies.
As a further factor of consideration, the database search was set to include all published manuscripts from 1913 until 2018, covering 105 years of scientific inquiry. Due to the selection protocol, the 1060 manuscripts selected were published between 1991 and 2018; thus, covering 27 years of research. They were from peer-reviewed journals only. Unfortunately, the current review covered a much smaller span, representing 15 years of research due to the low response rate (10.3%). For future studies that manage to include earlier research in their study, it could be interesting to conduct a time series analysis of each level of endorsement, analyzed in terms of the number and the percentage of abstracts to see if consensus changed over the years, and if so, how.
Overall, we found a 76.1% consensus that physical activity has cognitive benefits. In fact, experts’ and laymen’s abstract ratings were similar, which means that non-scientists could correctly understand and interpret the scientific language used in abstracts. This is important, since as mentioned earlier, communicating the scientific consensus is a powerful tool for assessing the weight of scientific evidence and for realigning public views of the issue with expert opinions [
35]. However, it should be noted that different disciplines produce different answers. To illustrate, 75% (
n = 9) of the rejection endorsing studies were in the field of neuroscience. The other three rejection endorsement manuscripts are in the fields of sport sciences (17%,
n = 2) and rehabilitation (8%,
n = 1). Therefore, in our sample, there were no rejection endorsing manuscripts from research in psychology, clinical neurology, or cognitive science. This could be attributed to the differences in methodology and cognitive performance measurement methods. A general problem in the methodology of exercise research is the relative lack of single standardized criteria for defining “exercise” and determining its effectiveness, so more standardized definitions would be useful in future research [
51]. It is also interesting to note that the criteria for establishing cognitive impairment were not consistent across studies, making it difficult to compare the differential effects of exercise across diagnostic categories (such as neurodevelopmental disorders and anxiety disorders), concluding that more standardized criteria are needed. Adhering to official guidelines, such as World Health Organization’s guidelines, on definitions of PE and PA and on the recommended intensity, duration, and volume of PA per age group, could be effective for comparative research. Further research is required from a range of disciplines to advance the understanding of the theoretical models of cognitive enhancement. For instance, forms of physical activity, such as balance and resistance training, have been insufficiently explored. More evidence is needed on the cognitive effects of such forms of PA, specifically in older adults [
42].
Such interventions could also be designed to be executed in a straightforward manner, and preferably, in real-life tasks. Likewise, cognitive research has expressed a disproportionate interest in the effects of PA on aspects, such as executive function and working memory. Research within the child and developmental studies have pointed out the importance of learning processes, induced by motor movements [
25]. Nonetheless, there is a scarceness of research on the effects of learning outside of the field of child and developmental studies. More systematic incorporation of various movements such as gesturing, head-tilting, and tracing movements could contribute to the ability to design and implement ecologically valid interventions that combine complex and diverse environments with efficacy for both cognitive and neural health. Concomitantly, a comprehensive battery of neuropsychological assessments and neuroimaging tools would complement behavioral findings [
52]. Future studies would benefit from such measures since they would significantly contribute to the understanding of how PA-induced changes in neural circuitry might be associated with accompanying behavioral changes [
52].
In the age of globalization, social and life sciences have remained largely American and European. An empirical review study has noted that volunteers from Western, educated, industrialized, rich, and democratic countries are the usual spokesmen of humanity who think differently than those from other parts of the world [
53]. Therefore, there is a need for replication studies on the effects of a physical activity intervention on cognition with participants from different nationalities. Additionally, another study has indicated the difficulty of estimating the true robustness and effect size of the cognitive enhancement effect by surveying the published literature [
54]. Publication bias has been implicated in decreasing the efficiency of the science and bringing the credibility of published research to lower standards [
54]. Hence, replication studies are needed to combat the publication bias and low diversity in the scientific literature. To conclude, remarkable progress has been made in the scientific understanding of the PA–cognition correlation during normal development, as well as in psychiatric and neurological populations. Studies have provided promising support for PA to be associated with both early and late-life cognitive functioning.
4.1. Study Limitations
There are several limitations in the present literature analysis, such as the representativeness of the literature sample, lack of clarity in the abstracts, the scholars’ sample representativeness and risk of evaluation bias.
The issue of sample representativeness when investigating extensive concepts, such as PA and cognition, was addressed by selecting a large sample for this type of literature analysis by using broad search terms (i.e., “brain” and “exercise”). Nonetheless, 1060 manuscripts are still a small fraction (9.2%) of the physical activity and cognition literature, since a Web of Science search yielded 11,519 manuscripts. Additionally, the search terms we used were a possible limitation. Therefore, the sourcing techniques employed in the current analysis could be expanded to include more manuscripts by using multiple databases, such as Scopus, Cochrane, PubMed, PsycINFO, and many others. Additionally, a combination of different search terms, including “physical activity, exercise, aerobic exercise,” and “cognition, cognitive function, executive function, learning, memory, academic, cognitive, and learning performance” could be more informative for future research.
Another area of uncertainty is the nature of the language used for writing the abstracts and abstract formats. In some cases, the ambiguous language made it difficult for the second stage laymen to determine the intended meaning of the authors. The implementation of the authors’ self-rating process allowed us to compare their abstract categorizations to those of laymen. The descriptive analysis revealed that, whereas non-scientists categorized nine papers as belonging to the uncertain category due to ambiguous language and dissimilar abstract formats, scientists categorized only one paper as belonging to the uncertain category. With regard to the “scholars sample representativeness,” it may be questionable whether someone with one published paper in the field should be considered a representative of the field. This, in turn, would have lowered the reliability of the survey and decreased response and completion rates [
55].
Possibly, the characteristics and the size of the non-expert group may not have been large enough to make a sensible comparison with the expert group. We deliberately opted for non-representativeness in our study design (“intentional” non-representativeness) for a couple of reasons. The first reason is practical; recruiting a large number of volunteers is challenging and not always feasible. Additionally, if we had matched the non-expert sample size to that of experts, the number of manuscripts per volunteer would have decreased to approximately two manuscripts per a non-expert. This means that the survey would not be representative of the different categories of endorsement, and it could have increased the content validity bias. Secondly, we aimed at the age group of 15–64, as it represents 65% of the current world population [
56]. By choosing this age group for the non-expert group in our study, the findings were more representative on a global scale. However, it should be noted that these findings should not be extrapolated to other non-expert age groups. Furthermore, the non-expert group cannot be considered as a representative sample due to its small size and the selection protocol employed. Future research should increase the size of the non-expert group relative to the one of experts’ if the manuscript sample is big enough to allow for sufficient representativeness of the different categories of endorsement.
What is more, participants’ subjectivity is inherent in the abstract rating process for both experts and non-experts. Regardless of defining the criteria for determining ratings prior to the rating period, it is possible that experts showed evaluation bias. Such a source of rating bias is that experts might have been more likely to classify papers as sharing the endorsement if they themselves were endorsing it, regardless of what the paper says. Another source of subjectivity bias could be scientific reticence which would mean that scientists would be more biased towards a “no position” categorization to avoid conflict. This bias was partially addressed by using multiple independent raters and comparing the abstract categorization results to those of author self-ratings. A comparison between both types of ratings revealed that this bias had minimal impact on the level of consensus. Authors’ “no position” categorizations (n = 26, 16.4%) were almost identical to those of volunteers’ (n = 25, 15.7%).
4.2. Government Policies
Scientists often aspire to see their findings being used to benefit society. Consecutively, decision-makers such as civic organizations, government officials, and citizens often seek out the best scientific evidence to make well-informed decisions for policies and regulations [
35]. The scientific evidence can be used to improve practice, especially when research findings are results of well-designed studies with methodological rigor that minimized the chances of bias. For instance, international public health agencies such as the World Health Organization often implement research findings into their policies, which serves to highlight the importance of research findings in daily life [
57]. The findings from the current study could contribute to efforts to translate research evidence into effective community programs for older adults with and without cognitive impairments, and among children and young people.