The effects of light physical activity on learning in adolescents: a systematic review

ABSTRACT

The effects of light physical activity (LPA) interventions on adolescents’ learning were reviewed. Learning was operationalized as academic performance and academic-performance-related cognitive functions. Using PubMed, Web of Science, ERIC, PsycINFO, PsycARTICLES, Psychology and Behavioral Sciences Collection, and Cochrane Library, only English-language studies published until 19-07-2021 were included. Of the resulting 49, 34 studies posed a low risk of bias, covering: executive functions (inhibition, updating, and shifting), memory, and academic performance. The quality of evidence of each outcome was very low, mainly due to inconsistencies in results between studies. Nonetheless, most studies found positive (n = 15) or null (n = 14) effects of LPA interventions on adolescents’ learning. Negative effects of LPA interventions were only found from studies with on-task procedures (n = 4). One study reported positive and negative effects. Only two studies utilized longitudinal LPA interventions, of which one reported positive effects, and the other reported null effects. Future studies should explore the role of potential moderators, e.g. timing of LPA interventions with regard to the learning and testing phases. Overall, while introducing LPA in the classroom may be effective at breaking up prolonged sedentary behavior, more high-quality research is needed to establish its positive effect on learning.

KEYWORDS:

• Academic performance
• executive function
• exercise
• youth
• memory

1. Introduction

Students tend to spend long durations being sedentary (Buckworth & Nigg, 2010; Guthold et al., 2010), which can be detrimental for their health and wellbeing (Rezende et al., 2014; Tremblay et al., 2011). Light physical activity (LPA) has been introduced into classrooms to interrupt sedentary behavior (SB), for example by using standing and cycling desks (Freeze, 2016; Wallace, 2015). Although favorable associations of LPA interventions have been reported for physical health (Butte et al., 2007; Carson et al., 2013), no consensus exists regarding whether LPA may affect learning, especially since the classroom’s function is to facilitate learning.

In this review, learning is operationalized to include academic performance and academic-performance-related cognitive functions. Executive functions (also known as cognitive control) are subdivided into three related, yet distinct categories: updating, the constant altering of working memory content as required by the task at hand; shifting, the ability to switch swiftly between mental sets or tasks; and inhibition, the ability to resist irrelevant stimuli, memory intrusions, and inappropriate prepotent responses (Friedman & Miyake, 2004; Miyake et al., 2000). Although executive functions and memory contribute to learning, learning is a more multifaceted and complex concept. In literature, learning is described as changes in an organism (including observable behavior, neural processes, and conscious thought) caused by regular or multiple environmental stimuli (de Houwer et al., 2013). In this review, we focus on learning in formal educational settings, operationalized to include academic performance and academic-performance-related cognitive functions, the latter including but not limited to executive functions (Best et al., 2011; Bull & Scerif, 2001; Diamond, 2013; St Clair-Thompson & Gathercole, 2006), memory (Hassevoort et al., 2016), processing speed (Rohde & Thompson, 2007), and attention (Stevens & Bavelier, 2012), all positively correlating to academic performance. Hereafter, academic performance and academic-performance-related cognitive functions are categorized under the broader category of learning.

This review focuses on adolescence (age 10–25) to understand the effects of introducing LPA interventions on learning in educational institutions. Adolescence is a distinct period of brain plasticity, where neurocognitive maturation of brain regions, such as the prefrontal cortex (Choudhury et al., 2006; Crone & Dahl, 2012), can be influenced by external stimuli (Crews et al., 2007). There has been growing recognition of the wide adolescence age range, with earlier puberty bringing the onset of adolescence to around the age of 10 in many populations, and an increased understanding that maturation of brain regions such as the prefrontal cortex, which is important for learning-related cognitive functions, completes around age 25 (Arain et al., 2013; Gogtay et al., 2004; Sawyer et al., 2018). Furthermore, brain maturation rates are susceptible to individual differences, with some reaching maturity around 18-19, while others reaching maturity in their 20’s (Foulkes & Blakemore, 2018), necessitating a study of the entire, wide adolescence age range.

Several explanatory mechanisms have been proposed for LPA’s potential effects on learning. For example, acute LPA (LPA that is performed in an immediate, one-off manner) is thought to improve cognitive performance due to its resulting changes in cerebral blood flow, enhancing oxygenation of brain regions, such as the prefrontal cortex (Giles et al., 2014; Holtzer et al., 2011; Timinkul et al., 2008), which are believed to be important for cognitive functions (Basso & Suzuki, 2017; Koechlin et al., 2003; Royall et al., 2002). Acute LPA has also been reported to increase arousal levels (Byun et al., 2014), affecting the storage and retrieval of information (Eysenck, 1976; Kaplan et al., 1968; Kleinsmith & Kaplan, 1963; Quas & Lench, 2007). Increased arousal has also been shown to lead to more elaborate information-sharing in groups (Knight & Baer, 2014), which may affect socially constructed learning (Vygotsky, 2012). The effects of LPA may be different for chronic LPA, which is performed repeatedly over a longer term, such as a few weeks, months, or even years. Chronic LPA has been suggested to affect learning by stimulating neuroplasticity, modulated by neurotransmitters (Barenberg et al., 2011).

Other than acute and chronic durations, the effects of LPA on learning may differ depending on whether LPA is performed concurrently with learning (i.e. on-task), or before learning (i.e. off-task). Several theories, such as the transient hypofrontality hypothesis (Dietrich, 2006), the capacity sharing model (Kahneman, 1973), and the central capacity sharing model (Tombu & Jolicoeur, 2003), argue that the brain has limited metabolic resources, thus when shared between LPA and learning (i.e. on-task), a deterioration in performance may be seen.

With LPA already being introduced into classrooms, this review studies whether LPA interventions, compared to SB, have different effects on learning in adolescents. Additionally, we make several distinctions regarding the LPA interventions: acute or chronic, and on-task or off-task.

2. Methods

2.1. Protocol

This study’s protocol was registered in PROSPERO (Chim et al., 2018).

2.2. Inclusion and exclusion criteria

Studies that reported the effects of LPA interventions on learning were included. LPA encompasses all activities between SB and moderate-to-vigorous physical activity (Ainsworth et al., 1993; Tremblay et al., 2017), defined as the range between 1.6 and 2.9 metabolic equivalents (METs). For this review, standing is considered LPA, as it does not fall within the definition of SB that includes only activities done while sitting, lying, or reclining (Tremblay et al., 2017). Other than standing, the Compendium of Physical Activities (Ainsworth et al., 1993, 2000, 2011) was used if activities were narratively described. When relative intensities were provided, we referred to Table 1.

Table 1. LPA classification.

Measure	LPA category
Maximum aerobic power, or maximal oxygen uptake (VO₂ max/ VO₂peak)^a	< 40-51% depending on individual’s VO₂ max (Day et al., 2003; Howley, 2001)
Maximum power output (Wmax), Peak power output (PPO or Wpeak), Peak work rate, or maximal workload (EXmax)^b	< 32-45% (Arts & Kuipers, 1994)
Oxygen uptake reserve (VO₂R)^c	< 40% (American College of Sports Medicine, 2018; Howley, 2001)
Maximal heart rate (HRmax)^d	< 64% (Howley, 2001)
Heart rate reserve (HRR)e	< 40% (American College of Sports Medicine, 2018; Howley, 2001)
One repetition maximum (1RM)^f	< 50% (Howley, 2001)
Borg’s 15-partition Rate of Perceived Exertion scale (RPE)^g	< 12 (Borg, 1982; Howley, 2001)
Borg’s 10-partition Category-Ratio (CR-10) scale^h	< 3 (Borg, 1982)

Note.

^a Maximal oxygen uptake (VO₂max), also known as maximal aerobic power, is the highest rate of oxygen consumption achieved during heavy, dynamic exercise (American College of Sports Medicine, 2018; Howley, 2001). This occurs when an increase in work rate does not result in a further increase in oxygen consumption, with a plateau observed. When a plateau-like response cannot be achieved but the subject has reached their limit, the peak oxygen consumption (VO₂peak) is recorded and commonly assumed to be equivalent to VO₂max (Day et al., 2003).

^b Maximal power output (Wmax), also known as peak power output (PPO or Wpeak) and peak work rate, is the highest output achieved during a graded exercise to volitional exhaustion on a cycle ergometer. LPA was calculated based on the following equation: % VO₂max = 12.1 + 0.866%Wmax (Arts & Kuipers, 1994).

^c Oxygen uptake reserve (VO₂R) is the difference between resting oxygen consumption and VO₂ max (Howley, 2001).

^d Maximal heart rate (HRmax) is defined as the highest heart rate achieved during maximal incremental exercise to volitional exhaustion, with age-predicted HRmax calculated by subtracting one’s current age from 220 bpm (bpm; American College of Sports Medicine, 2018).

^e Heart rate reserve (HRR) is the difference between resting heart rate and maximal heart rate (American College of Sports Medicine, 2018; Howley, 2001).

^f One repetition maximum (1RM) is the heaviest weight that can be lifted once with good form (Howley, 2001).

^g Borg’s 15-partition Rate of Perceived Exertion scale (RPE) is a self-reported scale from 6 to 20, narratively ranging from ‘very, very light’ to ‘very, very hard’ (Borg, 1982).

^h Borg’s 10-partition Category-Ratio (CR-10) scale is a revised version of the RPE scale. The CR-10 scale is used for self-reporting perceived workload, ranging from 0, being ‘nothing at all’, to 10, denoting ‘very, very strong’ (Borg, 1982).

LPA interventions that were contaminated with other intensity levels or entailed other intervention components (e.g. diet, social exclusion, emotional frustration) were excluded. Furthermore, LPA interventions that were cognitively engaging or demanding (e.g. embodied learning, detailed coordination, exergaming, explicit instructions to prioritize certain activities, and learning novel movements such as balancing) were excluded. Referring to Mavilidi et al.’s (2018) matrix of (temporal) integration and task relevance, LPA interventions included in our review have low task relevance and include both low and high temporal integration. Hereafter, we refer to LPA interventions that are performed concurrently with learning (high temporal integration) as on-task, and LPA interventions that are performed separately from learning (low temporal integration) as off-task.

LPA interventions were compared with a rested, seated control condition. No distinction was made between active and passive control conditions, as a previous meta-analysis found no meaningful differences between them (Au et al., 2020).

Learning outcomes included academic performance and academic-performance-related cognitive functions (e.g. memory, processing speed, attention, and executive functions). Studies that recruited physically and mentally healthy adolescents (including athletes, but excluding individuals with a clinical diagnosis) were included in this review. We define the adolescence range to span from 10 to 25 years. As there are several studies that include samples with a mean age of around 10 or 25 years, we made the decision to include studies where at least 84% of their sample (mean age plus one standard deviation) was within the age range of 10–25 years.

2.3. Search strategy and study selection

A systematic literature search was carried out using the electronic databases PubMed, Web of Science, ERIC, PsycINFO, PsycARTICLES, Psychology and Behavioral Sciences Collection, and Cochrane Library. For time- and resource-saving purposes, only English language articles published until 19-07-2021 were included. Using the PubMed Medical Subject Headings database, the search string, reported in Supplementary Material A, was used to search articles by title. Additionally, references in relevant studies were manually screened. Initially, we placed no restrictions on the study design, as reported in our protocol. However, to provide a more focused review on the causal effects of LPA interventions, we decided to include only intervention studies.

2.4. Data extraction

In total, 8,121 non-duplicated records were found (see Figure 1). Two authors independently screened 500 studies – first screening the title and abstract, then evaluating the full paper – achieving more than 80% agreement on the included studies. The remaining studies were divided between the two authors. The type of data to be extracted (i.e. sample size, sex, age, type of LPA intervention, intensity classification, duration of LPA, study design, outcome measure, type of test, measurement timing, and result direction) from the included studies was determined a priori with consensus from all authors. Considering potential speed-accuracy trade-off that may result from prioritizing perceptual, cognitive, and motor processes (Rinkenauer et al., 2004), reaction time and accuracy were considered as separate outcomes where possible. Studies were also split up between those that assessed learning outcomes during the LPA intervention (on-task) and those that assessed learning outcomes after (off-task), because on-task studies may be affected by the sharing of metabolic resources between learning and LPA (Dietrich, 2006; Kahneman, 1973; Tombu & Jolicoeur, 2003). Finally, studies that included longer-term/chronic LPA interventions were annonated to differentiate them from acute effects of LPA interventions.

Figure 1. PRISMA 2009 flow diagram (Moher et al., 2009).

2.5. Quality assessment

Critical assessment of the included studies was done independently by two authors using the revised Cochrane risk of bias tool for randomized trials (RoB2) (Sterne et al., 2019). The RoB2 assesses biases that may affect the intervention outcomes: the randomization process, deviations from original protocol, missing data, measurement of outcome, and selective reporting. The overall judgement for the outcome can be rated as high risk of bias, some concerns, or low risk of bias, with the proposed ratings allowed to be overridden by the authors (Sterne et al., 2019). We overrode the RoB2 ratings for three situations. Firstly, for crossover studies that used counterbalancing, the randomization of allocation sequence would not influence the studies’ outcome, because we do not expect for participants or the experimenters to favor one condition over the other (i.e. the intervention or control). Secondly, concealment of allocation sequence is unnecessary because it is impossible to conceal the physical experience of LPA. Thirdly, although we recognize the importance of a published pre-specified plan, most papers in the current field have yet to adhere to this relatively new requirement. Therefore, we did not flag the studies as risky based solely on this criterion. Any discrepancies between the two authors were discussed among all authors and resolved through consensus. Only studies with a low risk of bias were extracted for further analysis to strengthen the argumentation of evidence.

The quality of evidence for each outcome, from only low risk of bias studies was assessed independently by two authors using the Grades of Recommendation, Assessment, Development, and Evaluation (GRADE; Guyatt et al., 2011). GRADE provides guidelines to rate the quality of evidence per outcome level, ranging from high to moderate, low, and very low quality. The assessment rating was based on five criteria: inconsistency, indirectness, imprecision, publication bias, and risk of bias. Inconsistency describes the heterogeneity of results. Indirectness indicates how well the evidence answers the research question. Imprecision informs on whether the 95% confidence interval around the absolute effect is sufficiently narrow to avoid random errors. Publication bias indicates whether selective publication has been detected.

3. Results

This review included 49 studies (see Supplementary Material B for study characteristics). The PRISMA flow diagram in Figure 1 provides more information on the number of records identified, screened, and full-text assessed.

3.1. Critical assessment of individual studies

Using the RoB2, several included studies posed risks to the current review’s aim of studying the effects of LPA interventions on learning, detailed in Supplementary Material C. Twelve studies were rated as high-risk (Akazawa et al., 2019; Brisswalter et al., 1997; Del Giorno et al., 2010; Frodsham et al., 2020; Haynes et al., 2019; Lindheimer et al., 2017; Lo Bue-Estes et al., 2008; Lowe et al., 2016; Martins et al., 2013; Tsukamoto et al., 2017; Williams et al., 1985; Yamazaki et al., 2017), due to deviations from the intended intervention or missing outcome data. Three studies were rated with some concerns (Hasegawa et al., 2001; Schramke & Bauer, 1997; Wick et al., 2018), because of potential bias from the randomization process or deviations from the intended intervention. Notably, although some studies were well-designed, their research aim was not congruent with the current review’s aim (e.g. the lack of counterbalancing may not be a risk within the individual studies, but poses a risk to the current review). The learning outcomes were analyzed solely from the 34 remaining studies with a low risk of bias.

3.2. Study characteristics

Of the 34 studies with low risk of bias, 19 focused purely on LPA interventions. Five studies aimed to measure moderate physical activity but their resulting intensities fall within the current review’s definition of LPA (Cuttler et al., 2018; Harveson et al., 2019; Joyce et al., 2009; Loprinzi, Koehler, et al., 2019; Soga et al., 2015). The remaining 10 studies studied the effects of various physical activity intensities, from which we only extracted the effects of LPA (Brush et al., 2016; Budde et al., 2010; Endo et al., 2013; Hötting et al., 2016; Kamijo et al., 2009; Loprinzi & Kane, 2015; Loprinzi, Day, et al., 2019; Olson et al., 2016; Wang et al., 2013; Zuniga et al., 2019).

Within-groups design was utilized in 19 studies, between-groups design in 10 studies, and mixed design in five studies. The included studies comprised of 1,643 participants, with sample sizes ranging between 10 and 117. The participants’ mean age per study ranged between 11 and 23 years. No individual study sampled the full age range of 10–25 years. Only one study explicitly focused on early-adolescence (mean age 11). All other studies focused on mid- to late–adolescence (mean age 14–23 years). All studies utilized acute interventions, except for two studies that used chronic interventions: using ergometers for approximately 11 min/week across eight weeks (Liu et al., 2017) and ergometers with desks for at least 120 min/week across 10 weeks (Pilcher et al., 2017).

3.3. Evidence synthesis

The learning outcomes were categorized into inhibition, updating, shifting, memory, and academic performance. Based on the GRADE assessment, the quality of evidence for each outcome was rated as ‘very low’, because of three main reasons. First, the effects found were inconsistent: with positive, null, and negative effects found for all outcomes (except for shifting, with positive and null effects only). Second, the aggregated sample size for updating, shifting, and memory was small, leading to issues of imprecision, according to GRADE. Third, the studies did not directly measure the full age range targeted in this review. The quality of evidence is summarized in Table 2, with detailed assessment for each outcome listed in Tables 3–7. Of these studies, 15 studies reported positive effects, four studies reported negative effects, one study reported positive and negative effects, with the remaining 14 studies finding no significant effects. Nonetheless, the ‘very low’ quality of evidence indicates no clear effects of LPA interventions on learning among adolescents.

Table 2. Summary of quality of evidence according to GRADE.

Learning outcome	No. of studies	Quality of evidence
Inhibition	15	Very low
Updating	9	Very low
Shifting	5	Very low
Memory	6	Very low
Academic performance	10	Very low

Table 3. Effects of LPA interventions on inhibition.

15 studies (n = 591) [1] Alderman et al. (2014), n = 66 [2] Brush et al. (2016), n = 28 [3] Byun et al. (2014), n = 25 [4] Ellemberg and St-Louis-Deschênes (2010), n = 36 [5] Endo et al. (2013), n = 13 [6] Harveson et al. (2019), n = 63 [7] Harveson et al. (2016), n = 94 [8] Joyce et al. (2009), n = 10 [9] Kamijo et al. (2009), n = 12 [10] Larson, LeCheminant, Carbine, et al. (2015), n = 69 [11] Loprinzi and Kane (2015), n = 43 [12] O’Leary et al. (2011), n = 36 [13] Olson et al. (2016), n = 27 [14] Pontifex and Hillman (2007), n = 41 [15] Soga et al. (2015), n = 28		Condition	Accuracy		Reaction time
			Results	Studies	Results	Studies
		Flanker task
		Congruent	?	1, 9, 10, 12, 13, 14, 15	?	1, 9, 10, 12, 13, 14, 15
		Incongruent	?	1, 9, 10, 12, 15	?	1, 9, 10, 12, 13, 14, 15
			-	13, 14
		Interference	n/a	n/a	+	12
		Stroop test
		Congruent	?	1,2,3	+	7
					?	1,2,3
		Incongruent	?	1, 2, 3, 5	+	5^a, 6
Quality Assessment					?	1, 2, 3, 5^a
Risk of bias	No	Interference	n/a	n/a	+	3, 6^b
Inconsistency	Very serious (−2)^1				?	2, 6^b
Indirectness	Serious (−1)^2	Simon task
Imprecision	No	Congruent	?	2	?	2
Publication bias	Undetected	Incongruent	?	2	?	2
Other	Undetected	Choice response
Quality of evidence	Very Low	General	?	4	+	4
		Stop-signal task
		Go	?	8, 10	+	8^c
					?	8^c
		Stop	?	8, 10	+	8
		Condition	General performance
			Results		Studies
		Feature Match
		General	?		11
		Polygon
		General	?		11

Note.

+: Significant improvement;

?: no significant difference;

–: significant deterioration;

n/a: not available;

Alphabetic superscripts denote that results were found in multiple directions due to multiple measures presented in one study (see Supplementary Material B);

¹ Effects pointed at all directions (positive, no significant differences, and negative);

² The studies covered mean ages of 10 and 14 - 23, with gaps showing a poor representation of the full 10–25 age range.

Table 4. Effects of LPA interventions on updating.

9 studies (n = 341) [1] Brush et al. (2016), n = 28 [2] Budde et al. (2010), n = 39 [3] Burgess and Hokanson (1964), n = 40 [4] Delli Paoli et al. (2017), n = 48 [5] Grubaugh and Rhea (2014), n = 16 [6] Larson, LeCheminant, Hill, et al. (2015), n = 75 [7] Loprinzi, Day, et al. (2019), n = 24 [8] Loprinzi & Kane (2015), n = 43 [9] Soga et al. (2015), n = 28		Condition	Accuracy		Reaction time
			Results	Studies	Results	Studies
		Verbal Running Span task
		Short list	?	1
		Long list	?	1
		n-back task
		General	+	4^b	?	9^a
			?	1, 4^b, 9	-	9^a
		Aospan task
		General	?	5	?	5
Quality Assessment		modified-Paced Auditory Serial Addition Test
Risk of bias	No	General	–	6	n/a	n/a
Inconsistency	Very serious (−2)^1	Brown-Peterson task
Indirectness	Serious (−1)^2	General	?	7
Imprecision	Serious (−1)^3	Condition	General performance
Publication bias	Undetected		Results		Studies
Other	Undetected	Digit Span
		General	+		2^c, 3^d
Quality of evidence	Very low		?		2^c, 3^d
		Spatial Search
		General	?		8
		Spatial Slider
		General	?		8

Note.

+: Significant improvement;

?: no significant difference;

−: significant deterioration;

n/a: not available;

Alphabetic superscripts denote that results were found in multiple directions due to multiple measures presented in one study (see Supplementary Material B);

¹ Effects pointed at all directions (positive, no significant differences, and negative);

² Total sample size is smaller than 400;

³ The studies covered mean ages of 14–16 and 18-24, with gaps showing a poor representation of the full 10–25 age range.

Table 5. Effects of LPA interventions on shifting.

5 studies (n = 289) [1] Brush et al. (2016), n = 28 [2] Harveson et al. (2016), n = 94 [3] Jaffery et al. (2018), n = 44 [4] Loprinzi and Kane (2015), n = 43 [5] Wang et al. (2013), n = 80		Condition	Accuracy		Reaction time
			Results	Studies	Results	Studies
		Dimension-Switching task
		Simple blocks	?	1	?	1
		Mixed blocks	?	1	?	1
		Plus-Minus task
		Simple blocks	?	1	?	1
		Mixed blocks	?	1	?	1
Quality Assessment		Trail Making test
Inconsistency	Serious (−1)^1	A	n/a	n/a	?	2, 4
Indirectness	Serious (−1)^2	B	n/a	n/a	+	2^a, 3 ^b
Imprecision	Serious (−1)^3				?	2^a, 3^b, 4
Publication bias	Undetected	B-A	n/a	n/a	?	2
Other	Undetected	Condition	General performance
Quality of evidence	Very low		Results		Studies
		Wisconsin Card Sorting test
		Conceptual level response	?		5
		Categories completed	?		5
		Perseverative responses	?		5
		Failure to maintain set	?		5

Note.

+: Significant improvement;

?: no significant difference;

−: significant deterioration;

n/a: not available;

Alphabetic superscripts denote that results were found in multiple directions due to multiple measures presented in one study (see Supplementary Material B);

¹ Effects pointed at two directions (positive and null);

² Total sample size is smaller than 400;

³ The studies covered mean ages of 16 and 21-22, with gaps showing a poor representation of the full 10–25 age range.

Table 6. Effects of LPA interventions on memory.

6 studies (n = 341) [1] Abou Khalil et al. (2020), n = 18 [2] Cuttler et al. (2018), n = 80 [3] Larson, LeCheminant, Hill, et al. (2015), n = 75 [4] Loprinzi and Kane (2015), n = 43 [5] Loprinzi, Koehler, et al. (2019), n = 38 [6] Zuniga et al. (2019), n = 87		Condition	General performance
			Results	Studies
		Free Recall
		Rate	+	1^a
			?	1^a
		Number	+	1^a
			?	1^a
		Delayed Recall
		Rate	+	1^a
Quality Assessment			?	1^a
Risk of bias	No	Number	+	1^a
Inconsistency	Very serious (−2)^1		?	1^a
Indirectness	Serious (−1)^2	Prospective Memory test
Imprecision	Serious (−1)^3	Viral Video	?	2
Publication bias	Undetected	Reminder	+	2
Other	Undetected	Red Pen	?	2
Quality of evidence	Very low	California Verbal Learning test II
		Number recalled	?	2
		Number retained	?	2
		Brief Visuospatial Memory test-Revised
		Number recalled	?	2
		Number retained	?	2
		Rey Auditory Verbal Learning test
		Short-delay recall	?	3
		Long-delay recall	?	3
		Total learning	–	3
		Spatial Span
		General	?	4
		Paired Associates
		General	?	4
		Logical Memory task
		General	?	5
		Free recall
		General	+	6
		Judgement of Learning Magnitude
		General	?	6
		Metacognitive Accuracy
		Calibration	?	6
		Resolution	?	6

Note.

+: Significant improvement;

?: no significant difference;

−: significant deterioration;

Alphabetic superscripts denote that results were found in multiple directions due to multiple measures presented in one study (see Supplementary Material B);

¹ Effects pointed at all directions (positive, no significant differences, and negative);

² Total sample size is smaller than 400;

³ The studies covered mean ages of 20 - 21, with gaps showing a poor representation of the full 10–25 age range.

Table 7. Effects of LPA interventions on academic performance.

10 studies (n = 566) [1] Abou Khalil et al. (2020), n = 18 [2] Alderman et al. (2014), n = 66 [3] Finch et al. (2017), n = 96 [4] Harveson et al. (2019), n = 63 [5] Hötting et al. (2016), n = 55 [6] Isip (2014), n = 24 [7] Liu et al. (2017), n = 40* [8] Pilcher and Baker (2016), n = 38 [9] Pilcher et al. (2017), n = 117* [10] Schmidt-Kassow et al. (2014), n = 49	Condition	Accuracy		Reaction time
		Results	Studies	Results	Studies
	Word-Picture Verification
	General	+	7^a*	+	7*
		?	7^a*
	Semantic Judgment task
	General	+	7*	+	7*
	Condition	General performance
		Results	Studies
	Reading
	General	?	2, 3, 6
	Mathematics/Mental Arithmetic
Quality Assessment	General		+	4^b, 6
Risk of bias		No	?	1^c, 4^b
Inconsistency		Very serious (−2)^1	−	1^c
Imprecision	No	Reasoning
Indirectness	Serious (−1)^2	General	?	8
Publication bias	Undetected	Introductory Psychology exam
Other	Undetected	General	?	9*
Quality of evidence	Very low	Polish-German vocabulary
		General	+	10
			?	5

Note.

*: chronic/longitudinal LPA intervention;

+: Significant improvement;

?: no significant difference;

−: significant deterioration;

Alphabetic superscripts denote that results were found in multiple directions due to multiple measures presented in one study (see Supplementary Material B);

¹ Effects pointed at all directions (positive, no significant differences, and negative);

² The studies covered mean ages of 14 and 18–23, with gaps showing a poor representation of the full 10–25 age range.

3.4. Timing of LPA intervention

A comparison was made for studies that used an on-task (Table 8) or off-task protocol (Table 9). From separating the studies into on- or off-task, the only observable pattern was that negative effects from the LPA intervention were only present from on-task protocols (Abou Khalil et al., 2020; Larson, LeCheminant, Hill, et al., 2015; Olson et al., 2016; Pontifex & Hillman, 2007; Soga et al., 2015).

Table 8. Studies where the LPA intervention and the outcome measure are conducted concurrently (i.e. on-task).

Study	Results
Executive function: Inhibition
Alderman et al. (2014)	?
Joyce et al. (2009)	+
Larson, LeCheminant, Carbine, et al. (2015)	?
Olson et al. (2016)	−
Pontifex and Hillman (2007)	−
Soga et al. (2015)	?
Executive function: Updating
Grubaugh and Rhea (2014)	?
Loprinzi, Day, et al. (2019)	?
Soga et al. (2015)	−
Executive function: Shifting
Wang et al. (2013)	?
Memory
Abou Khalil et al. (2020)	+
Cuttler et al. (2018)	?
Larson, LeCheminant, Hill, et al. (2015)	−
Academic performance
Abou Khalil et al. (2020)	−
Alderman et al. (2014)	?
Finch et al. (2017)	?
Isip (2014)	+
Pilcher and Baker (2016)	?

Note.

+: Significant improvement;

?: no significant difference;

−: significant deterioration.

Table 9. Studies where the LPA intervention are conducted before assessing the outcome measure (i.e. off-task).

Study	Timing since LPA intervention	Results
Executive function: Inhibition
Brush et al. (2016)	Post-15 min	?
	Post-180 min	?
Byun et al. (2014)	Post	+
Ellemberg and St-Louis-Deschênes (2010)	Post	+
Endo et al. (2013)	Post-5 min	+
	Post-20 min	?
Harveson et al. (2016)	Post (5-40 min)	+
Harveson et al. (2019)	Post (5-20 min)	+
Joyce et al. (2009)	Post	+
	Post-30 min	+
Kamijo et al. (2009)	Post	?
Loprinzi and Kane (2015)	Post	?
O’Leary et al. (2011)	Post (22.2 ± 0.6 min)	?
Soga et al. (2015)		?
Executive function: Updating
Brush et al. (2016)	Post-15 min	?
	Post-180 min	?
Budde et al. (2010)	Post	+
Burgess and Hokanson (1964)	Post	+
Delli Paoli et al. (2017)	Post	?
Loprinzi and Kane (2015)	Post	?
Soga et al. (2015)	Post	?
Executive function: Shifting
Brush et al. (2016)	Post-15 min	?
	Post-180 min	?
Harveson et al. (2016)	Post (5-40 min)	+
Jaffery et al. (2018)	Post-5 min	+
Loprinzi and Kane (2015)	Post	?
Memory
Cuttler et al. (2018)	Post	+
Hötting et al. (2016)	Post-20 min	?
	Post-24 h	?
	Post-24 h	?
Loprinzi and Kane (2015)	Post	?
Loprinzi, Koehler, et al. (2019)	Post	?
	Post-25 min	?
Zuniga et al. (2019)	Post	+
Academic performance
Harveson et al. (2019)	Post (5-20 min)	+
Liu et al. (2017)*	Post (8 tests)	+
	Post-one month	+
Loprinzi and Kane (2015)	Post	?
Pilcher et al. (2017)*	Post	?
Schmidt-Kassow et al. (2014)	Post-24 h	+

Note.

*: chronic/longitudinal LPA intervention;

+: Significant improvement;

?: no significant difference;

−: significant deterioration.

A detailed look into the on-task studies reveals two further types of protocols, whether the LPA was administered before or during a learning phase, and whether the LPA was administered before or during the measurement of learning outcomes (see Table 10). A learning phase was incorporated within seven studies, where the participants were given material to learn before a test. The timing of LPA differed between studies, with some administering LPA before or after the learning phase. Comparing the studies by Hötting et al. (2016) and Schmidt-Kassow et al. (2014), the participants had to translate newly learnt Polish vocabulary into the German language. Both groups of researchers administered LPA interventions that elicited similar heart rates (around 100 bpm) and measured similar outcomes 24 h later. Hötting and colleagues did not find a significant effect, whereas Schmidt-Kassow’s group reported that their LPA intervention boosted the learning of Polish-German words. The main difference is that, in Hötting et al.’s study, the participants learnt the Polish-German words while seated, then participated in the LPA intervention; whereas Schmidt-Kassow et al.’s participants learnt the Polish-German words during the LPA intervention. This finding that LPA during learning may lead to better performance is supported by Liu et al. (2017), whose participants learnt a second-language while carrying out the LPA intervention, also finding improved performance.

Table 10. Timing of LPA intervention with respect to the learning phase.

Study	Timing of LPA intervention	Results	Outcome: Type of test
Hötting et al. (2016)	Learning, seated → LPA → Testing, seated	?	Academic performance: Polish-German vocabulary
Schmidt-Kassow et al. (2014)	Learning during LPA → Testing, seated	+	Academic performance: Polish-German vocabulary
Liu et al. (2017)	Learning during LPA → Testing, seated	+	Academic performance: English (second language) vocabulary
Pilcher et al. (2017)	Learning during LPA → Testing, seated	?	Academic performance: Introductory Psychology exam
Larson, LeCheminant, Hill, et al. (2015)	Learning during LPA → Testing during LPA	−	Memory: Rey Auditory Verbal Learning test (native language)
Zuniga et al. (2019)	LPA → Learning and testing seated	+	Memory: Logical Memory task
Loprinzi, Koehler, et al. (2019)	LPA → Learning and testing seated	?	Memory: Free Recall

Note.

+: Significant improvement;

?: no significant difference;

−: significant deterioration.

Although there seems to be a pattern that LPA during learning improves language-learning performance, this finding was not replicated for an Introductory Psychology exam (Pilcher et al., 2017), where participants who cycled while studying did not do any better than their sedentary peers. This suggests that the effects of LPA interventions may be different depending on the context or subject. Furthermore, verbal memory appears to be affected differently, compared to language-learning. In a verbal memory study, participants who memorized lists of words in their native language while walking, hence having LPA during their learning, recalled less items than the group that was seated (Larson, LeCheminant, Hill, et al., 2015). Notably, Larson and colleagues’ intervention differed from the aforementioned studies (Schmidt-Kassow et al. (2014) and Liu et al. (2017) who found that LPA during learning enhanced performance) because both the learning and the testing phase were done while walking, suggesting that the timing of the LPA intervention relative to the learning measurement moderates its effects as well.

3.5. Timing of delayed testing

Related to the timing of the LPA intervention, nine studies measured learning after a delay. With 5-minute to 40-minute delays, positive effects were found for inhibition (Endo et al., 2013; Harveson et al., 2016; Joyce et al., 2009; O’Leary et al., 2011), shifting (Harveson et al., 2016; Jaffery et al., 2018), and mathematics (Harveson et al., 2019).

Although Endo et al. (2013) found positive effects for inhibition 5 min after the LPA intervention, this effect was not replicated just 20 min later. Harveson et al. (2016) reported positive effects for mathematics after a delay of 5-20 min, but no significant effect was found for inhibition. Likewise, no significant effects were found after a delay of 15-25 min for inhibition, updating, and shifting (Brush et al., 2016), Polish-German vocabulary (Hötting et al., 2016), and for memory (Loprinzi, Koehler, et al., 2019). Brush et al. (2016) also measured inhibition, updating, and shifting after a 3-hour delay, and similarly did not find any significant effect from the LPA intervention.

Longer delays include a 24-hour and one-month temporal gap. After a 24-hour delay, Schmidt-Kassow et al. (2014) found positive effects on learning Polish-German words, but Hötting et al. (2016) did not find any significant effect. Only one study measured learning after a one-month delay (Liu et al., 2017), and found positive effects despite the long delay.

3.6. Duration, type, intensity, and frequency of LPA intervention

Across the included studies, LPA interventions lasted from 30 s to 240 min. Inconsistency remains a problem, as studies that administered the same duration of LPA find conflicting results (detailed in Supplementary Material B). No discernerable pattern was observed with regard to duration.

Reported effects were different depending on the type of LPA intervention. Alderman et al. (2014) and Olson et al. (2016) tested inhibition during LPA. Alderman et al. did not find that their LPA intervention had an effect, whereas Olson et al. found a deterioration in accuracy when performing the same inhibition test. The only difference is that participants walked in Alderman et al.’s study, while Olson et al.’s participants cycled on an ergometer. Therefore, the type of LPA appears to influence the effects found. Another example is the comparison of two studies by Harveson et al. (2016, 2019). Shifting was found to improve from aerobic LPA (Harveson et al., 2016), whereas mathematics and inhibition improved only from resistance LPA (Harveson et al., 2019). One final example on how different types of LPA may lead to different results is provided by Abou Khalil et al. (2020). In their study, immediate and delayed free recall of words were better during a bout of walking compared to sitting, but no significant differences were found between standing and sitting. Also, for mental arithmetic, walking (but not standing) led to a lower performance on the rate of correct answers, while standing (but not walking) led to a lower performance on the reaction time of correct answers. These patterns suggest that the different types of LPA may differentially affect learning.

Only one included study manipulated the intensity of LPA. Within the same group of participants, Endo et al. (2013) found that light cycling at 40% maximal workload boosted performance on inhibition, whereas very light cycling at 20% maximal workload did not show any significant performance changes.

There were no studies that manipulated the frequency of LPA administration, for example, by changing how many times per week the participants were exposed to the LPA interventions. Of the two longitudinal studies that did administer frequent LPA, all three studies were very different in their outcome measure and overall design (detailed in Supplementary Material B). Once again, the LPA interventions were found to have inconsistent results, with no effects on a Psychology exam (Pilcher et al., 2017) and positive effects on language-learning (Liu et al., 2017).

3.7. Individual differences

Burgess and Hokanson (1964) found that performance on an updating task only improved post-LPA within the fitter group with a lower resting heart rate, but not in the less fit group with a higher resting heart rates. Yet, Zuniga et al. (2019) did not find differences between low-fit and high-fit participants on memory performance, with both groups benefiting from a bout of treadmill walking.

With regard to baseline performance, Budde et al. (2010) compared between those who did well and those who did less well during their baseline updating performance. Only those who performed worse at baseline showed improvements in their updating performance after LPA.

3.8. Type of test used

The observable effects of LPA interventions differed when different types of tests were used, even if the different tests were supposed to measure the same outcome. For example, when testing inhibition, positive effects were found after acute LPA using the Flanker task (O’Leary et al., 2011), Stroop test (Byun et al., 2014; Endo et al., 2013; Harveson et al., 2016, 2019), Simon task (Ellemberg & St-Louis-Deschênes, 2010), and Stop-signal task (Joyce et al., 2009), but no effects were found for the Feature Match and Polygon tests (Loprinzi & Kane, 2015). Also regarding updating, positive effects were found after acute LPA using the n-back task (Delli Paoli et al., 2017) and Digit Span task (Budde et al., 2010; Burgess & Hokanson, 1964), but not using the Verbal Running Span task (Brush et al., 2016), Aospan task (Grubaugh & Rhea, 2014), Spatial Search task, or the Spatial Slider task (Loprinzi & Kane, 2015). For shifting, positive results after an acute bout of LPA were only found for the Trail Making test (Harveson et al., 2016; Jaffery et al., 2018). Even within the same study, Cuttler et al. (2018) measured prospective memory after an acute bout of LPA and found that LPA interventions produced positive effects when tested with the Reminder Prospective Memory test, but null effects when tested with the Red Pen Prospective Memory test. Thus, it appears that different tests may differ in their sensitivity to the effects of LPA interventions.

3.9. Speed and accuracy

For the outcome of inhibition, positive effects of LPA were seen only in reaction time (Byun et al., 2014; Ellemberg & St-Louis-Deschênes, 2010; Endo et al., 2013; Harveson et al., 2016, 2019; Joyce et al., 2009; O’Leary et al., 2011), while negative effects were found only for accuracy (Olson et al., 2016; Pontifex & Hillman, 2007). As for the outcome of updating, negative effects were found for both accuracy (Larson, LeCheminant, Hill, et al., 2015) and reaction time (Soga et al., 2015), while positive effects were only found for accuracy (Delli Paoli et al., 2017). This contrasted the patterns found for inhibition, where positive effects were found for reaction time and negative effects for accuracy. Finally, for the outcomes of shifting and language-related academic performance, positive effects were found for both accuracy (Liu et al., 2017) and reaction time (Harveson et al., 2016; Jaffery et al., 2018; Liu et al., 2017). Overall, the effects of LPA interventions on speed and accuracy did not produce consistent, discernerable patterns.

4. Discussion

The aim of this paper was to systematically review the effects of LPA interventions on learning in healthy adolescents aged 10–25 years. A total of 49 studies was found, of which only 34 were judged to have low risk of bias. From these 34 studies, positive effects were found in 16 studies, negative effects in four studies, one study reported both positive and negative effects, and the remaining 14 studies reported null effect. The learning outcomes were grouped into five categories: (1) inhibition, (2) updating, (3) shifting, (4) memory, and (5) academic performance. The quality of evidence was very low, indicating no clear effects of LPA interventions on learning.

To improve future quality of evidence, the issues of inconsistency, imprecision, and indirectness need to be addressed. To reduce inconsistencies between future studies, one solution would be to account for potentially moderating variables, identified in the current review – a recommendation that has been strongly advocated before (Brisswalter et al., 2002; Tomporowski & Ellis, 1986). Considering the numerous potential moderators that we have identified, the inclusion of too many covariates may result in overfitting (Zhang, 2014). Therefore, we advise to recruit sufficiently large samples (other solutions to overfitting are discussed by Babyak (2004) and Hawkins (2004)), which would also narrow the margin of error, reducing imprecision over the aggregated included studies. To reduce indirectness, we implore for more representation of 10–25 age range, as early adolescence (10–13 years) is currently understudied in literature, and the overall age range is an important period for brain plasticity (Arain et al., 2013; Gogtay et al., 2004; Sawyer et al., 2018).

To better understand why inconsistencies exist in literature, variables that may have potential confounding effects have to be teased apart and studied. We recommend that researchers make careful choices with regard to: timing of LPA intervention, timing of delayed testing, duration and type of LPA, individual differences, and type of test used. With the various protocols currently utilized, it is difficult to make a direct comparison of the results. Furthermore, several factors remain understudied, such as the different LPA intensities and frequency of LPA interventions. Therefore, we encourage for more dose–response studies to explain the inconsistencies in results currently found.

With regard to timing of the LPA intervention, it appears that administering LPA during the learning phase fosters performance, especially in language-learning (Liu et al., 2017; Schmidt-Kassow et al., 2014). The positive effects could be ascribed to LPA increasing arousal level (Byun et al., 2014; Knight & Baer, 2014) or brain oxygenation (Giles et al., 2014; Holtzer et al., 2011; Timinkul et al., 2008). Administering LPA before the testing phase of learning may lead to boosted performance, while administering LPA during the testing phase may lead to a drop in performance. This finding aligns with the theories that suggest a sharing of metabolic resources between learning and maintaining concurrent LPA, leading to lower learning performance (Dietrich, 2006; Kahneman, 1973; Tombu & Jolicoeur, 2003). Therefore, the timing of LPA, in relation to the learning and testing phase, is highly likely a moderator.

When it comes to delayed testing, learning outcomes measured after a very short (5-minute) delay tend to show positive effects (Endo et al., 2013; Harveson et al., 2016, 2019; Jaffery et al., 2018). Longer delays (20 min to 3 h) produce inconsistent findings. A previous meta-analysis reported similar findings, with larger positive effects observed from shorter delays (Chang et al., 2012). For positive effects to persist after longer delays, the LPA intervention would need to last for longer durations or be of higher intensities (Chang et al., 2012). With regard to language-related outcomes, beneficial effects appear to withstand longer delays of 24 h (Schmidt-Kassow et al., 2014) and one month (Liu et al., 2017). Inconsistencies still exist, with Hötting et al. (2016) not finding a significant effect on language-learning, but this could possibly be explained with the participants only carrying out LPA after learning the language (see Table 10 and section 3.4). Overall, positive effects on learning can be found after a very short delay, while positive effects on language-related outcomes tend to withstand longer delays.

Inconsistency in results was high when scrutinizing on the duration and type of LPA. For example, with duration, when inhibition was measured after 30 min of LPA, positive effects (Ellemberg & St-Louis-Deschênes, 2010; Harveson et al., 2016; Joyce et al., 2009), negative effects (Olson et al., 2016), and null effects (Alderman et al., 2014; Loprinzi & Kane, 2015) were reported. As for the type of LPA, different types of LPA recruit different musculoskeletal functions and bodily systems (e.g. aerobic vs. resistance movements (Harveson et al., 2019)), possibly explaining the inconsistencies found. Further research into the potential moderators is required to understand the nuances that are present in extant literature.

When controlling for individual differences, two important factors were found from literature – fitness levels and baseline performance on the learning measure. The potential moderating effects of these two factors have been highlighted elsewhere (Pesce, 2009). Fitness levels were expected to only affect cognitive outcomes that were measured concurrently with physical activity (Pesce, 2009). In our review, Burgess and Hokanson (1964) found that after an acute bout of LPA, physically fit individuals perform better in an updating task, compared to their less-fit peers. This hints at fitness levels affecting one’s capability to cope with the demands of a physical activity intervention even after the conclusion of the physical activity intervention. This result is supported by studies of higher intensity showing that fit individuals, compared to their unfit counterparts, show better cognitive performance during (Delignières et al., 1994) and after a bout of physical activity (Sjöberg, 1980). Nonetheless, one study from our review (Zuniga et al., 2019) did not find an effect of fitness levels, necessitating future studies to investigate the potential moderating effects of fitness on the relationship between LPA interventions and learning.

With regard to baseline performance of the learning measure, Budde et al. (2010) found that those who performed worse from the onset experienced a larger improvement in their learning. Previous studies that used moderate-to-vigorous physical activity interventions align well with Budde et al.’s study, with those who performed less well at baseline benefiting more during measures of executive functions (Drollette et al., 2014; Sibley & Beilock, 2007) and classroom on-task behavior (Mahar et al., 2006). Budde and colleagues ruled out ceiling effects for the higher performing group, as no participant was able to complete the entire Letter Digit Span task of updating. Further research is needed to uncover the explanation for why physical activity interventions appear to be most beneficial for those with a larger learning margin. Overall, there are hints that individual differences may influence the relationship between LPA interventions and learning outcomes. However, due to the low number of studies that have accounted for individual differences, we cannot yet conclude which individual differences influence the effect of LPA interventions on learning and how these differences influence the effects.

In this review, different results were found for different tests, despite assessing the same outcome (see section 3.8 Type of test used). Assuming the tests used are valid and reliable, the effect of any intervention comes into question when different results are found with different types of tests assuming to measure the same construct. For future studies, we recommend researchers to select outcome measures based on clear theoretical rationales. For example, executive functions have over 128 different tests available (Etnier & Chang, 2009). Therefore, the type of test should be selected based on empirical evidence showing that the test is sensitive to the effects of LPA interventions on adolescents.

As for speed and accuracy, no consistent effect was found from the LPA interventions. Although previous studies reported a trade-off between speed and accuracy in cognitive tests (Rinkenauer et al., 2004), the studies included in our review did not find support for this trade-off. Initially, it appeared that LPA interventions may have a positive effect on the reaction time of inhibition and a negative effect on accuracy. However, these findings were not sufficient to indicate that the effects of LPA interventions were affected by a trade-off in speed-accuracy, because no one study reported both a positive effect on reaction time and a negative effect on accuracy. Furthermore, the pattern found for inhibition was contrasted by updating, shifting, and language-related academic performance (see section 3.9 Speed and Accuracy). Overall, the effects of LPA interventions on speed and accuracy remain inconsistent across studies.

Two factors remain understudied in extant literature – intensity and frequency of LPA interventions. Despite already focusing on the lighter end of the intensity spectrum, different intensities still appear to have a different effect on learning outcomes (e.g. 20% vs. 40% maximal workload in, Endo et al. (2013)’s study). With only one study manipulating the intensity, more studies are needed to check whether intensity may moderate the effects of LPA interventions on learning. As for the frequency of LPA interventions, none of the included studies studied the frequency of LPA interventions (e.g. multiple LPA interventions in a week vs. once per week). Further research on manipulating the frequency of LPA interventions is needed because having more frequent LPA interventions may enlarge the effect size. Moreover, the LPA interventions introduced within classrooms, such as standing and cycling desks, are not meant for one-time use. These arguments can also be made for longitudinal designs. Yet, only two longitudinal studies were found in this review (Liu et al., 2017; Pilcher et al., 2017) with inconsistent results found. It is important to study the longitudinal effects of LPA interventions because the results may vary according to how experienced the participant is to the LPA intervention. In a qualitative study, participants explained that similar to starting a new exercise program, the novel experience of using standing desks is different from long-term use, requiring time needed for familiarization (Wilkerson et al., 2019). In sum, further research is needed on the frequency and long-term effects of LPA interventions, to explores its possible dose–response effect on learning.

Despite the inconsistencies between studies, we can still derive some practical implications from this review. Of the included studies, 15 studies reported a positive effect, while 14 studies reported null effects. Apart from the focus on learning, other researchers have found LPA to be beneficial for physical health (Chastin et al., 2019; Füzéki et al., 2017), counteracting the negative risks of prolonged SB (Rezende et al., 2014; Tremblay et al., 2011). Thus, adolescents should not be held back from the health benefits of LPA, considering the probable null or positive effects that LPA has on learning. However, we do not discount the negative effects found from four studies (Larson, LeCheminant, Hill, et al., 2015b; Olson et al., 2016; Pontifex & Hillman, 2007; Soga et al., 2015) and one study that reported both positive a and negative effects (Abou Khalil et al., 2020). Within these studies, learning outcomes were measured during the LPA intervention. Therefore, the practical recommendation would be to not administer LPA during exams (despite one study finding boosted performance on a mathematics exam when the students were standing (Isip, 2014)). Instead, we encourage education institutions to provide students with the option to perform LPA during their learning phase, especially during language-learning, or to provide LPA breaks in-between classes.

To our knowledge, this is the first systematic review to focus exclusively on LPA and its effects on learning in adolescents. Previous systematic reviews tend to focus on general physical activity, usually of moderate-to-vigorous intensities (Álvarez-Bueno et al., 2017; Biddle et al., 2019; Daly-Smith et al., 2018; Singh et al., 2018; Wassenaar et al., 2020). These moderate-to-vigorous physical activity reviews present similar findings as the current LPA review. Mainly, the majority of studies on physical activity interventions pointed towards either positive or null effects on cognitive and academic performance, yet the inconsistent effects lead to inconclusive evidence. Similarly, these reviews also highlight the high heterogeneity in the intervention design and measurement of outcomes, with a lack of studies attempting to clarify the dose–response effects of various potential moderators.

The limitation of this review is that during data extraction, only one author screened the majority of studies. For the first 500 studies that were screened by two authors, a high interrater agreement (> 80%) was achieved, therefore the remaining studies were screened by only one author for efficiency. Still, human error may have led to the overlooking of relevant articles that could have been included in this review. The strengths of this review include the thorough and independent assessment of each study by two authors. The RoB2 was used to assess the risk of bias from each study, then the quality of evidence at the outcome level was further evaluated using GRADE. Basing our analysis on only low-risk-of-bias studies also serves as a strength of this review, as the conclusions drawn would be less susceptible to methodological errors in trial design, conduct, or reporting.

Hitherto, there is a lack of high-quality evidence to substantiate the effects of LPA interventions on learning among adolescents. This review found 15 studies reporting positive effects, four studies reporting negative effects, one study reporting both positive and negative effects, while the remaining 14 studies reporting null effects. These studies were conducted in laboratory settings, with the exception of Pilcher et al.’s (2017) library setting. Further studies are therefore needed for three reasons. First, to understand the effects of LPA interventions in field settings. Second, to elucidate potential moderators. And third, to replicate current studies to ascertain the reliability of results. Apart from learning, LPA has various benefits for physical health (Chastin et al., 2019; Füzéki et al., 2017), with a high feasibility and acceptance in classrooms (Benzo et al., 2016; Sherry et al., 2016). Thus, there is no reason to refrain from introducing LPA interventions among adolescents. The majority of included studies in this review find positive or null effects of LPA interventions on adolescents’ learning. This suggests that introducing more LPA in the classrooms to break up prolonged SB fosters learning, but further research is warranted.

Disclosure statement

No potential conflict of interest was reported by the author(s).