The influence of colour in the context of sport: a meta-analysis

ABSTRACT

This study conducted a meta-analysis to explore the influence of colour on human behaviour in the context of sport and to scrutinise potential moderators affecting this impact. P-curve analysis was used to test whether a set of findings in the published literature contains an evidential value. To analyse the methodological standard of included studies two quality assessment tools were applied: the QATSDD and guidelines recommended by Elliot [2019. A historically based review of empirical work on color and psychological functioning: Content, methods, and recommendations for future research. Review of General Psychology, 23(2), 177–200]. The PRISMA protocol was employed for data identification and selection. Sixty-nine articles met the inclusion criteria and were deemed eligible for analysis. The results of the p-curve analysis suggested that cumulatively the studies contained evidential value (p < .001). The meta-analysis with random effects models indicated a medium significant effect ($\hat{\theta }$ = 0.56, 95%-CI [0.40, 0.72], p < .001) with substantial heterogeneity (I² = 96.97%, t² = 0.34, p_Q < .001). Several moderators (i.e., age, study design, sport type, physical outcome measures, perceiving coloured environment, perceiving others in coloured equipment, red colour, blue colour) showed a significant impact on the relationship between colour and human behaviour in sports. Closer inspection of the quality of included studies, however, implied that carefully controlled empirical work on colour in the context of sport is scarce.

KEYWORDS:

• Colour
• sport
• meta-analysis
• review
• p-curve

Introduction

In preparation for the FIFA World Cup 2006, jersey colour played a significant role in German national soccer coach Jürgen Klinsmann’s meticulous planning (Zelustek & Niklaus, 2006). Klinsmann personally ensured that his national team was given a new look when he advocated replacing the traditional black alternate uniform with red jerseys. The famous German coach explained his decision by discussing the players’ personal preference for this colour, as well as the psychological influence of red as a symbol of aggression (Fritsch, 2011). The story of the German national coach changing the jersey colour of his team in order to increase his chance of winning emphasises the necessity of research to clarify this assumption and to question the impact of colour in sports in general (Meuren, 2006).

The current body of research on the impact of colours in sport does not share a clear perspective. Researchers’ interest in the field emerged in the 1980s as they began investigating the impact of colours on grip strength (O’Connell et al., 1985; Pellegrini et al., 1981; Pellegrini & Schauss, 1980), ball catching, and reaction time tests (Don Morris, 1976; Koslow, 1983). Based on this foundation, scientific research extended the analysis of colour effects to other parameters in sports, such as yards penalised and penalty minutes in team sports (Frank & Gilovich, 1988). Recent studies from the last two decades have incorporated more comprehensive methodologies and were able to broaden the scientific discourse surrounding the omnipresent role of colours in human society (Elliot, 2015; Elliot & Maier, 2007; Kaya & Epps, 2004). Most of this research has focused on the colour red and has shown a slight impact of coloured uniforms on sporting outcomes. While some research suggests a link between red and aggression/dominance (Attrill et al., 2008; Dreiskaemper et al., 2013; Hill & Barton, 2005; Krenn, 2015; Piatti et al., 2012; Sorokowski et al., 2014), other studies have revealed null results or even contradictory findings (Allen & Jones, 2014; Carazo-Vargas & Moncada-Jiménez, 2014; Furley et al., 2012; García-Rubio et al., 2011; Krenn, 2014; Pollet & Peperkoorn, 2013).

Closer inspection of the methodology used in these studies, however, implies that carefully controlled empirical work on colour and human behaviour in sports is scarce. Also, effect sizes of most results appeared to be small and may even overestimate the actual influence of colours in sport contexts (Elliot, 2015; Krenn et al., 2017). As a consequence, doubts have been raised about the consistency and the validity of colour impact (Lehmann et al., 2018). Goldschmied and Lucena (2018) also highlighted the low generalisability of the findings, as various studies did not differentiate between the self-perception of wearing coloured equipment, perceiving a coloured environment, or perceiving opponents in coloured equipment. These limitations render the research conducted essentially impossible to interpret. In addition to the aforementioned flaws, Elliot (2019) addressed various other consistent methodological issues (e.g., failing to control for colour-deficient participants, the target sample size via power analyses, background colour, or illumination, as well as the fact that colour varies on the three attributes of lightness, chroma and hue). Hence, from a scientific perspective, the impact of colour in sport and exercise settings seems rather vague, leaving an unclear picture of heterogeneous results. Thus, there is a need for research incorporating moderating variables when examining the impact of colour on sporting outcomes. One additional reason for the inconsistent and heterogeneous results might be that most empirically driven studies grounded their analysis only fragmentarily on theoretical assumptions and often did not provide or directly contribute to an overarching theoretical framework, which impedes the validity and possible generalisability of the published findings (Elliot & Maier, 2014; Whitfield & Whiltshire, 1990).

The most prominent theory dealing with the effects of colour on human beings was proposed by Elliot and Maier (2012). Their so-called Colour-in-Context Theory (CIC) presumes six specific propositions about the psychological impact of colour: (1) each colour carries a psychological meaning; (2) viewing colour influences psychological functioning and may foster motivational and behavioural processes; (3) responses to colour are automatic and are usually processed without conscious awareness; (4) colour associations can be rooted in learning and/or innate predispositions; (5) relations between colour perception and affect, cognition, and behaviour are reciprocal; (6) the effects and meanings of colour are context-specific (Meier et al., 2015). In this regard, the framework has been described as the “physical and psychological context in which colour is perceived to influence its meaning and, accordingly, responses to it” (Elliot, 2015, p. 2). These meanings and responses seem to emerge from social learning processes (Chantal & Bernache-Assollant, 2015; Chebat & Morrin, 2007; Mentzel et al., 2017) and/or evolutionarily ingrained predispositions (Changizi, 2009; Cuthill et al., 1997; Ilie et al., 2008; Pryke, 2009; Setchell & Wickings, 2005). Although CIC theory provides a robust attempt to explain colour effects, the framework is broad and non-sport specific and does not specify how to differentiate between different contexts (e.g., whether sport is a context in its own right or should it be divided into sub-contexts). On the basis of the previously described mixed findings and the lack of a clear sport-related theoretical foundation, it is about time to explore whether a colour effect in sports even exists.

An important consideration when investigating potential moderators is to take up the aforementioned limitation of CIC by making contextual sport-specific distinctions and breaking up the general idea of “sport” into different sub-contexts (types of sports). To date, the majority of research has focused on investigating the presence of colour influences in team sports (e.g., soccer: Allen & Jones, 2014; García-Rubio et al., 2011; basketball: Goldschmied & Lucena, 2018; Goldschmied & Spitznagel, 2020; ice hockey: Webster et al., 2012), combat sports (e.g., judo: Dijkstra & Preenen, 2008; boxing: Gülle et al., 2016; wrestling: Curby, 2016; taekwondo: Falco et al., 2016) and individual sports that do not involve a combat situation (e.g., running: Mentzel et al., 2019; cycling: Fujii et al., 2017; jumping: Lam et al., 2017). In this regard, the differing context of various sport types might act as a potential moderator, thus affecting the impact of colour in sports.

Another important potential moderator within the field of colour research is whether colour effects arise from wearing coloured equipment (e.g., Lam et al., 2017), perceiving a coloured environment (e.g., Payen et al., 2011), perceiving opponents in coloured equipment (e.g., Krenn, 2014), or some combination of these factors. One study examining the combination of wearing and perceiving coloured equipment is that of Feltman and Elliot (2011). Their study showed that participants imagining themselves wearing red in a taekwondo bout had enhanced self-perception of their own dominance and threat, whereas perceiving an opponent in red enhanced the perception of these qualities in the opponent. Further research found evidence that perceiving a coloured environment may also influence human behaviour in sporting contexts (Akers et al., 2012). Based on the discussion about wearing and perceiving effects of colours, it has been assumed that the aforementioned research on colour influences in sports might have been biased by not taking into account the impact on referees’ judgments (Carazo-Vargas & Moncada-Jiménez, 2014; Hagemann et al., 2008). This hypothesis was confirmed by Krenn (2014), who illustrated that referees judged tackles from behind more harshly for soccer players wearing red when compared to blue. Thus, wearer effects, perceiver effects of a coloured environment and perceiver effects of people in coloured jerseys may all operate and serve to influence sporting outcomes (Meier et al., 2015).

In addition to the moderators of sport context and colour location, colours might affect different outcome variables in different ways (Elliot, 2015). So far, colour research in sports was mainly focused on psychological outcome measures (for instance, aggression, dominance and threat) by the help of defined indicators like perceived number of blows (Sorokowski et al., 2014) or perceived harshness of tackles (Krenn, 2014). Other psychological variables that have been examined include measurements of mood states (e.g., anxiety: Profusek & Rainey, 1987; arousal: Payen et al., 2011; self-confidence: Recours & Briki, 2015; and positive affect: Williams et al., 2011). Various other studies have examined the influence of colours on performance metrics. One of the most well-known examples of this is Hill and Barton (2005), who investigated the number of competitions won in combat sports and found a significant advantage for red uniforms in comparison to blue. On the contrary, other studies (Caldwell & Burger, 2011; García-Rubio et al., 2011) reported no significant effect of jersey colour on performance outcomes. Both positive correlations and null results have been shown between colour and other performance metrics, such as winning probability (Curby, 2016; Dijkstra & Preenen, 2008; Julio et al., 2015), successful passes (Iwase, 2002), and covered distance (Briki et al., 2015; Londe et al., 2018). Finally, in addition to psychological behaviour and performance outcomes, researchers have examined the impact of colours on physical parameters in the context of sports (e.g., testosterone levels: Hackney, 2006; force production: Elliot & Aarts, 2011; Payen et al., 2011; heart rate: Akers et al., 2012; blood pressure: Etnier & Hardy, 1997; and VO2max: Fujii et al., 2017). While some experiments found support for the assumption that viewing colour is linked to enhanced force production (Elliot & Aarts, 2011; O’Connell et al., 1985) and heart rate (Dreiskaemper et al., 2013), other studies reported no significant differences (Crane et al., 2008; Ingram & Lieberman, 1985; Rezaeian et al., 2015). Overall, the studies indicate that a large variety of psychological, physical, and performance variables have been examined, yielding mixed results as to the actual influence that colours have.

Finally, colour research differs based on the methodology applied (Krenn et al., 2017). In retrospective studies incorporating archival data, uniform colour has been demonstrated to affect sporting outcomes in diverse sports such as soccer (Allen & Jones, 2014; Attrill et al., 2008), combat sports (Hill & Barton, 2005; Rowe et al., 2005), rugby (Piatti et al., 2012) and virtual competition (Ilie et al., 2008). Contrarily, experimental studies carried out in laboratory-based investigations most-commonly analysed colour effects in the contexts of physical strength tests (Dunwoody, 1998; Elliot & Aarts, 2011), cycling endurance tests on an ergometer (Briki et al., 2015; Etnier & Hardy, 1997) or ball catching and reaction time tests (Don Morris, 1976; Koslow, 1983).

Overall, the heterogeneity of methodologies and the mixed results described in the previous sections do not generate a clear picture of a potential colour impact in sports. A meta-analysis may resolve the ambiguity by enabling scrutinisation of potential colour impact moderators on sporting outcomes and providing a quantitative synthesis across diverse research results. Thus, meta-analyses might help to clarify whether and under which circumstances colour effects appear in the context of sport. As a consequence, this meta-analytic approach may also contribute to building a robust theoretical model predicting colour effects in sports. Since meta-analyses can be biased by type-I error and p-curve analyses test whether a set of findings in the published literature contains an evidential value, the p-curve becomes an ideal complement to the meta-analytic approach (Boggero et al., 2017). Therefore, the present research was conceptualised to complement previous scientific studies by systematically mapping and synthesising the research done in this area, as well as by identifying any existing gaps. Specifically, the following research question was formulated: what is the influence of colours in the context of sport? This question will be examined on the whole, but also while adjusting for the aforementioned differentiation within the context of sports research. Therefore, we will investigate the moderating effects of the study design (experimental, retrospective), context of sport type/activity (combat, team, individual sports), domain (psychological, physical, performance), colour location (wearing, perceiving environment, perceiving persons) and colour (red, blue, green, pink, yellow, orange, purple, brown, white, black, grey, not specified/control). Thus, the results of the meta-analysis might help to understand – from a theoretical but also applied perspective – in which conditions and circumstances colour effects appear, respectively disappear, and thus also contribute to our understanding of how colour exerts its potential influence. In addition, a meta-analysis also might help to identify what is mostly needed in this research area to profoundly enrich our knowledge about how colours affect human behaviour in the context of sport.

Methods

Protocol and registration

The Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines (Moher et al., 2009) were used to conduct this meta-analysis. The final protocol was registered prospectively with the Open Science Framework (https://osf.io/enwfh) on 6 October 2020.

Search strategy

A systematic search for identification of relevant literature was conducted using the following electronic databases of peer-reviewed journal articles and online research registers: Web of Science, PubMed, Google Scholar and Scopus. Search terms were agreed upon a priori by the first author and second author in order to reduce the risk of relevant literature being removed in initial e-journal searches (Gledhill et al., 2017). Databases were initially searched on 1 February 2020 and the final search was repeated on 23 July 2020 to ensure that no studies published in the intervening period were omitted. The terms used in conducting the search included: (1) “Color OR colour AND sports” OR (2) “Red AND sports” OR (3) “Black AND sports” OR (4) “Blue AND sports” OR (5) “Green AND sports” OR (6) “Pink AND sports” OR (7) “Yellow AND sports” OR (8) “Color OR colour AND performance” OR (9) “Color OR colour AND soccer” OR (10) “Color OR colour AND football” OR (11) “Color OR colour AND team” OR (12) “Color OR colour AND matches” OR (13) “Color OR colour AND competition” OR (14) “Color OR colour AND combat” OR (15) “Color OR colour AND hockey” OR (16) “Color OR colour AND cycling” OR (17) “Color OR colour AND running” OR (18) “Color OR colour AND endurance” OR (19) “Color OR colour AND win” OR (20) “Color OR colour AND loss” OR (21) “Color OR colour AND mood” OR (22) “Color OR colour AND exertion” OR (23) “Color OR colour AND heart rate” OR (24) “Color OR colour AND motor performance” OR (25) “Color OR colour AND force production” OR (26) “Color OR colour AND power” OR (27) “Color OR colour AND motor ability” OR (28) “Color OR colour AND psychology” OR (29) “Color OR colour AND physiology”. The search string was adapted appropriately for each database and searched, at minimum, in the fields of titles, abstracts and keywords. Truncated terms were used, where possible, to ensure that as many variant spellings of a term as possible were captured (Jackman et al., 2019). All retrieved titles and abstracts were exported into Rayyan (https://rayyan.qcri.org/welcome) reference management software. Duplicates were identified through the automatic duplication feature.

Eligibility criteria

The following inclusion and exclusion criteria were established to ensure that the parameters of this meta-analysis were clearly defined and that all literature relevant to the present objectives was identified: papers were included (a) if they were published in English (to ensure consistency in appraising articles, Tod et al., 2011); (b) if they were original, peer-reviewed journal publications (excluding dissertations, theses, abstracts, magazine articles, non-peer-reviewed sources); (c) if they were available as full-text articles (Gledhill et al., 2017); (d) if they contained empirical quantitative data (excluding qualitative or mixed methods data: reviews, meta-analyses, commentaries, letters, non-empirical articles); (e) if they involved human participants; (f) if they involved (1) participants in an actual sport/exercise setting, (2) participants judging pictures/videos/real-life actions depicting a sport/exercise setting, (3) participants imagining to be in a sport/exercise setting, (4) participants in virtual competition. Papers were excluded (g) if they focused on instrument development and/or validation (Jackman et al., 2019); (h) if they involved mentally disturbed participants.

Screening process

The author and a second screener sifted through the studies in three phases, both working independently to enhance the quality of the screening process. In the first screening phase, all reference titles were reviewed. In the second phase, abstracts of the remaining articles were screened, and in the third phase, full-text papers were obtained and eligibility reviewed in detail (Jones, 2004). Additionally, in order to identify articles that may have been missed due to the inconsistent use of search terms, manual searching was performed by the author to identify further potentially relevant articles through: (a) backward search by the help of reference lists of each identified study and (b) titles of forward citations of identified studies on Google Scholar (Jackman et al., 2019). This process was repeated for all articles identified through manual searches until no further potentially relevant studies were located. After concluding the manual search, all full texts of the relevant articles were obtained and searched in detail for compliance with the eligibility criteria by the author and the second screener independently. Both researchers met to discuss discrepancies in the outcomes of the screening process until a final consensus was reached (Wertli et al., 2014).

Data coding of moderators

Following the identification of all relevant studies, the authors read each included study several times to become immersed with the research context and to enhance familiarity with the data before extracting and synthesising all relevant findings. A data-charting form was jointly developed in Microsoft Excel 2020 by the first author to determine which variables and/or potential moderators to extract (Lenzen et al., 2017). The first author independently charted the data in an iterative process, with randomly chosen articles checked for coding accuracy by the second author. For this meta-analysis, the following moderators were extracted from each study: (a) year of publication; (b) number of sub-experiments; (c) study design (experimental, retrospective); (d) sample size; (e) mean age in years; (f) gender (all male, all female, mixed, not specified); (g) context of sport type/activity (individual [without combat], team, combat sports); (h) psychological domain; (i) physical domain; (j) performance domain; (k) wearing colour; (l) perceiving people in coloured equipment, (m) perceiving coloured environment, (n) mix between wearing and perceiving; (o–x) red, blue, green, pink, yellow orange, white, black, grey, not specified/control colours.

Effect size measures

A data-charting form was jointly developed in Microsoft Excel 2020 by the first author to determine which statistical values to extract. The first author independently charted the data (sample size, degrees of freedom, t-value, F-value, chi-square, Cohen’s d, Pearson correlation, odds ratio, z-value, eta square, standard deviation, mean) for each study in an iterative process, with randomly chosen studies being checked for charting accuracy by the second author. Some missing effect sizes were calculated manually based on recommendations by Lakens (2013). In total, 459 colour effects were reported; 196 of these (42.70%) had effect sizes that were copied, 119 (25.93%) had effect sizes that were calculated manually, and 144 (31.37%) had effect sizes that were not available, as not enough information was provided for a calculation.

Considering the primary studies’ designs, a standardised mean difference appeared to be most appropriate for the analysis (Morris, 2008). Thus, for each available effect size, Cohen’s d values were computed from the coded raw data. For less biased analysis, the effect sizes were transformed via Hedges transform (Hedges, 1981). Some primary studies reported multiple analyses using independent samples. In order to compare the effect sizes on the study level, the sub-studies’ effect sizes were averaged. Therefore, the subsamples’ effect sizes were first transformed to Hedges effect sizes. Mean values were computed to avoid the statistical dependence of effect sizes (Belanger & Vanderploeg, 2005; Cheung, 2019; Mann et al., 2007). As a consequence, 60 averaged effect sizes were used for the final computation.

Quality of studies

The quality assessment tool (QATSDD; Sirriyeh et al., 2012) was employed to assess the quality of each selected study. The QATSDD allows researchers to concomitantly appraise the methodological quality of included studies to ensure an acceptable scientific standard (Jackman et al., 2019). The tool consists of a list of 16 items that need to be rated on a four-point scale, ranging from zero (not at all) to three (complete). All studies were evaluated for relevant criteria that applied to quantitative designs, except criterions 11 (fit between stated research question and format and content of data collection tool, e.g., interview schedule) and 14 (assessment of reliability of analytical process), which were omitted as they applied to qualitative designs only. In addition, criterion 8 (detailed recruitment data) was evaluated for experimental studies only and was omitted for retrospective studies. The total score for each study was divided by the maximum possible score (42 for experimental studies and 39 for retrospective studies) and was reported as a percentage for standardisation purposes (Jackman et al., 2019).

In addition, the quality of all included studies was assessed by the help of the “Guidelines for a High-Quality Study on Color and Psychological Functioning” by Elliot (2019). This quality assessment tool was applied to specifically control for the complexity inherent in colour science methodology, as, for example, colour consists of multiple components, termed lightness, chroma and hue (this study solely focuses on the reported hue as the other two values were mainly not reported). The guidelines consist of five items (participant naiveté and confidentiality, colour vision deficiency, sample size, colour specification and control, background colour and illumination) that were evaluated as controlled (1) or not controlled/not mentioned (0). The total number of controlled criteria for each study was divided by a maximum of five and was also reported as a percentage for standardisation purposes. Studies using retrospective data could not be scored according to the given criteria set. In line with previous recommendations (Milner, 2015), two authors independently assessed the quality of all studies that met the inclusion criteria for both the QATSDD and the criteria by Elliot.

P-curve analysis

Within all included studies, 459 distinct relevant statistical results were reported. Of these results, 190 were statistically significant in the predicted direction and were recomputed from the provided information (e.g., degrees of freedom, t-statistic, χ²-statistic, F-statistic, Z-statistic). According to Simonsohn et al. (2014a), recomputation is necessary due to the fact that p-values are often reported as smaller than a particular benchmark and because they are sometimes reported incorrectly. In some cases, one study reported multiple significant p-values. Because p-values must be statistically independent for inference from a p-curve to be valid, there were strict selection rules for multiple p-values from one study (Boggero et al., 2017; Smyth & Ansari, 2019). Multiple p-values were extracted only if an article included multiple sub-experiments reporting p-values for independent samples. When studies reported multiple, significant p-values, Simonsohn et al. (2014b) recommended performing a robustness analysis by rerunning the p-curve analysis with alternate, dependent p-values. To be consistent within the selection rules, the first significant p-value that occurred in the studies/sub-experiments was used in the original main p-curve analysis. The robustness analyses were run by replacing the original p-values from any study/sub-experiment that had multiple qualified p-values with the final significant p-value reported. The data for both the main p-curve and the robustness test were input to the p-curve analytic software (http://p-curve.com) version 4.06 and were reported in a p-curve disclosure table, which can be found at the Open Science Framework (https://osf.io/enwfh/).

Meta-analytical procedures

Meta-analysis and moderator analysis

All reported analyses were conducted with the System for Statistical Computation and Graphics R (R Core Team, 2016), applying the R package metaphor (Viechtbauer, 2010). Overall effect sizes were computed using a random-effects model. Random-effects models are preferable to fixed-effects models in most analyses as they account for between-study variance in the true effect sizes and thus do not assume the same true effect size across studies (Borenstein et al., 2010). The mean effect size was calculated via a random-effects model. Statistical analyses were conducted at α = .05 significance level with 95% confidence intervals. The estimated effect sizes presented as Hedges g were interpreted as small, medium or large effect sizes guided by Cohen’s (1988) recommendations (0.20, 0.50, 0.80; Fritz et al., 2012). For assessment of between-study heterogeneity, I², τ² and Q-statistic were calculated. A significant result of the Q-statistic is an indicator for between-study heterogeneity and implies the necessity to use a random-effects model (Borenstein et al., 2010). Investigating the effect of specific study characteristics on the total effect size, the moderator variables were entered for the meta-regression.

Sensitivity analysis

To assess the sensitivity of overall estimates, analyses of both publication bias and the impact of estimates of single studies were conducted. Publication bias is a common problem in meta-analyses that can lead to an overestimation of effect sizes (Wilson & Lipsey, 2001). It is theorised that studies reporting significant effect sizes might be more likely to be published, compared to studies reporting no significant effect sizes. A funnel plot of the effect sizes was created and the trim-and-fill method (Duval & Tweedie, 2000) was applied to estimate the number of missing studies.

The impact of a single study on the overall estimate was tested by the leave-one-out method. A series of 60 meta-analyses was conducted, each leaving out one study. A change of the total estimate compared to the other analyses implies a larger impact of the excluded single study’s estimate.

Results

Literature identification

The electronic database search generated a total of 1267 results with a further 30 being returned from forward and backward manual citation searching. After the removal of 430 duplicates, the remaining 867 results were screened for titles and abstracts, which resulted in 766 articles being excluded from further analysis. Full texts for the remaining 101 records were reviewed for eligibility. Thirty-two full-text articles that did not meet the eligibility criteria were excluded for five main reasons: (a) articles were reviews, meta-analyses or commentaries that did not contain empirical quantitative data (n = 11) (e.g., Elkan, 2009; Meier et al., 2015; Rogerson, 2017); (b) articles did not involve participants in an actual sport/exercise setting; or participants judging pictures/videos/real-life actions depicting a sport/exercise setting; or participants imagining themselves in a sport/exercise setting; or participants in virtual competition (n = 11) (e.g., Briki & Hue, 2016; Little & Hill, 2007; Wiedemann et al., 2015); (c) articles were not related to the topic of the influence of colours in the context of sport (n = 7) (e.g., Donnelly et al., 2016; Spencer, 2017; Xu, 2019); (d) articles were non-peer-reviewed journal publications, such as dissertations or master’s theses (n = 2) (e.g., Wiedemann, 2016); (e) articles involved mentally disturbed participants (n = 1) (e.g., Gilliam, 1991). Following this process, 69 articles met the inclusion criteria and were deemed eligible for analysis according to the criteria for this meta-analysis (Figure 1).

Figure 1. PRISMA (preferred reporting items for systematic reviews and meta-analysis) chart of search strategy.

Characteristics of studies

All studies that were eligible for this review were published between 1976 and 2020. Twelve out of the total 69 studies consisted of two or more sub-studies (e.g., Chantal & Bernache-Assollant, 2015), yielding 84 experiments in total. Thirty-nine studies involved male participants only (e.g., Sorokowski & Szmajke, 2007), 33 studies involved both sexes (e.g., Koslow, 1983), one study involved female participants only (Goldschmied & Spitznagel, 2020), and sex was not specified in eleven of the studies (e.g., Keller & Vautin, 1998). The average sample size for all 58 experimental studies was 49.25 (SD = 45.60), ranging from six to 290 participants with a mean age of 22.50 years (SD = 3.45). The average sample size for all 26 retrospective studies was 2754.25 (SD = 9266.46), ranging from 10 to 45,874 (for more details on study characteristics, see Tables 1 and 2).

Table 1. Summary of all studies that were included (n = 69).

Authors (date of publication)	Study design	Sample characteristics (Mean age in years)	Sport type (context)	Colours	Object in colour	Colour location	Hypotheses/Research questions	Domains	Measurements (grouped into domains)	(1) Key Findings: Psychological domain	(2) Key findings: Physical domain	(3) Key findings: Performance domain
Akers et al. (2012)	Experimental	14 Male participants (Mage = 20.7)	Cycling test(Individual)	Red, green, grey	Environment (video)	Perceivingenvironment	Cycling while being exposed to video footage of a green natural environment results in a more positive mood and lower perception of effort compared to achromatic or red-filtered video footage	(1) psychological, (2) physical	(1) Mood states (POMS), perceived exertion scale (Borg scale); (2) rate of oxygen uptake (VO2), respiratory exchange ratio, heart rate	Decreased mood disturbance and perceived exertion in green condition compared to red and grey, red is perceived as more aggressive than green and grey, no significant difference for feelings of tension, depression, fatigue, vigour, confusion	No difference for heart rate, VO2, respiratory exchange ratio between colour conditions	x
Allen and Jones (2014)	Retrospective	7720 Premier League matches between 1992 and 2012	Soccer (Team)	Red, blue, white, others	Jersey	x	Teams contesting their home fixtures in red show a greater home advantage than teams contesting their home fixtures in other colours	(3) performance	(3) Average finishing league position, home advantage	x	x	Teams in red finished in higher league positions than teams in blue, white and other colours; After controlling for team ability, teams in red did not show a greater home advantage than teams in other colour
Attrill et al. (2008)	Retrospective	68 English soccer clubs between 1946 and 2003	Soccer (Team)	Red, blue, yellow-orange, white	Jersey	x	English Soccer home league teams wearing red will, on average, be more successful than teams wearing other colours	(3) performance	(3) Home league rank, percentage wins (home and away), points per game, average league placing	x	x	Teams in red more often champion than other colours, red teams with higher percentage of maximum points achieved at home games than other colours, no difference in away games, teams in red have a better league placing than teams in other colours
Briki et al. (2015)	Experimental	10 Male cyclists (Mage = 17.4)	Cycling test (Individual)	Red, green, grey	Lens glasses, computer screen	Wearing, perceiving environment	Green exercise improves cycling performance in terms of covered distance, heart rate, and feelings	(1) psychological, (2) physical, (3) performance	(1) Perceived effort (Stamford and Noble’s scale), anxiety (Competitive State Anxiety Inventory-2 Revised), and enjoyment (Enjoyment Inventory); (2) heart rate; (3) Covered distance	Enjoyment higher in the green than red condition, no colour influence on perceived effort and anxiety	Heart rate lower in red environment than in grey and green environments	Covered distance lower in the red environment than in the grey and green environments
Caldwell and Burger (2011)	Retrospective	102 NHL matches between 2008 and 2010	Ice hockey (Team)	Red, black, others	Jersey	x	Analyse whether several measures of aggression levels are higher when teams wear black or red jerseys in comparison to differently coloured outfits	(1) psychological, (3) performance	(1) Total number of penalty minutes per game, (3) win percentage, scored points	No colour effects on levels of aggression/e.g., penalty minutes per game	x	No effect of jersey colour on win percentage and scored points
Carazo-Vargas and Moncada-Jiménez (2014)	Retrospective	718 Fights for taekwondo athletes* from World Taekwondo Championship 2013	Taekwondo (Combat)	Red, blue	Body protector	x	Red colour influences the outcome in taekwondo fights when participants wear a vest linked to an electronic scoring system	(3) performance	(3) Winning probability	x	x	No difference of winning a combat when wearing red or blue uniforms
Chantal and Bernache-Assollant (2015)	Experimental (E1, E2, E3)	67 Participants* (Mage = 20.5) (E1); 74 participants* (Mage 19.8) (E2); 102 participants* (Mage = 19.7) (E3)	Cycling (E1, E2, E3) (Individual)	Yellow, grey (E1, E2); yellow, grey, white (E3)	Image of syringe (E1), width frame (E2), cycler (E3)	perceiving environment (E1, E2, E3)	The image of a syringe on a yellow background, by contrast with grey, would more strongly evoke competitive cycling and doping (H1), State mood was also assessed but we expected null effects in that regard (H2) (E1); we anticipated stronger negative perceptions when yellow was put into play as compared with grey (H3) (E2); we anticipated that the racer in yellow would be perceived as a more deterrent model (H4), in comparison with grey and white, yellow would yield weaker impressions that sport could be attractive (H5) significant differences were not expected between the perceptions generated by the grey and white jerseys (H6) (E3)	(1) psychological (E1, E2, E3)	(1) Sports evocations and suspicions of doping (Performance Enhancement Attitude Scale), mood (nine-point scale) (E1); Social perceptions (Sport-Oriented Tall Poppy Scale), suspicions of doping (Performance Enhancement Attitude Scale), mood (nine-point scale) (E2); Racer’s suitability as role model, attractiveness of sport participation (Affective, Cognitive and Attitude Semantic Differential Scales) (E3)	Stronger association of a syringe to cycling than to track and field when displayed on yellow instead of grey background, Stronger suspicions of doping linked to yellow background, no colour effects for state mood (E1); Racer more criticised and perceived as less suitable model in yellow condition compared to grey, yellow condition yielded stronger suspicions of doping, no colour effect for state mood (E2); Participants in the yellow condition evaluated racer as less suitable role model than in the grey or white conditions, participants’ scores were equivalent across the grey and white conditions, participants in yellow condition viewed sport participation as less attractive than those in the grey or white conditions, similar scores in the grey and white conditions (E3)	x	x
Crane et al. (2008)	Experimental	18 Male participants (Mage = 20.3)	Cycling test, grip strength test (Individual)	Red, blue, white	Light stimulus	Perceiving environment	Analyse effects of red, blue, and white (neutral) light on muscular strength and average power	(2) Physical	(2) Muscular power measured on Wingate cycling ergometer; grip strength measured with dynamometer	x	Anaerobic power higher in room with red light compared to blue or white; Grip strength higher in room lit with white light compared to blue	x
Curby (2016)	Retrospective	952 Fights* during Senior World Wrestling Championships 2015	Wrestling (Combat)	Red, blue	Uniform	x	Analyse the presence of an unfair competitive bias in a random assignment of uniform colour	(3) Performance	(3) Winning probability/match outcome	x	x	No relation between the colour of the uniform and the match outcome (win/loose)
Dijkstra and Preenen (2008)	Retrospective	301 Male judo fights in the Athens Olympics 2004, 72 international male judo tournaments between 1996 and 2005	Judo (Combat)	Blue, white	Judogi	x	Winning bias for blue as reported by Rowe et al. (2005) should disappear when controlling for confounding factors (H1); winning bias for blue declines over the tournament rounds when including seeded athletes (H2), No colour-associated winning bias in the gold-medal finals, when controlling for seeding system, asymmetries in prior experience and recovery time	(3) Performance	(3) Winning probability	x	x	When excluding matches involving a seeded athlete, no difference between blue and white conditions for winning probability, in gold-medal finals was no difference between blue and white conditions; association between colour and athlete ability/winning probability declined over the tournament rounds, proportion of tournaments with a winning bias for blue was much higher after 2003
Dijkstra et al. (2018)	Retrospective	45,874 Fights* from international judo tournaments between 2008 and 2014	Judo (Combat)	Blue, white	Judogi	x	Analyse win bias for the first called athlete (blue before WC 2011, white after WC 2011) (H1); No significant effect of blue or white judogi colour on winning probability (H2)	(3) Performance	(3) Winning probability	x	x	Win bias for the first called athlete, but this effect was not significantly related to the colour of the judogi; No win effect of judogi colour
Don Morris (1976)	Experimental	90 Elementary school children (Mage = not reported)	Catching test (Individual)	Blue, yellow, white, black	Ball,background	Perceiving environment	Analyse effects colour of balls and colour of backgrounds have upon the catching performance of second, fourth, and sixth graders	(3) Performance	(3) Catching scores	x	x	Higher catching performance scores with yellow and blue balls than with white; catching performance higher with white background than with black
Dreiskaemper et al. (2013)	Experimental	28 Male handball players (Mage = 27.7)	Combat, leg strength test (Individual, Combat)	Red, blue, no colour (control)	Jersey	Wearing, perceiving people	Wearing a red jersey leads to an increase in heart rate, strength during a combat situation	(1) Psychological, (2) physical, (3) performance	(1) Subjective strain (RPE), (2) heart rate, force production, (3) combat performance (number of hits on opponents)	No colour difference for subjective strain (RPE)	Wearing red increased heart rate, no effects of colour on heart rate before and after the fight and compared to control group; Red increased force production to a greater extend than blue, no difference between blue and control for force production	No colour difference for combat performance (number of hits on opponents)
Dunwoody (1998)	experimental	30 male participants (Mage = not reported)	hand grip strength test (Individual)	red, blue	light stimulus	perceiving environment	Red and blue colour influence grip strength (H1); Assess potential effects of brightness on grip strength (H2)	2) physical	2) Grip strength measured with dynamometer	x	Higher grip strength in red condition, wherein red was not equated for brightness, no difference in the condition wherein brightness of the red and blue was equated	x
Elliot and Aarts (2011)	Experimental (E1, E2)	30 Participants* (Mage = 13.0) (E1); 46 participants* (Mage = 21.5) (E2)	Pinchgrip force test (E1) (Individual), handgrip strength test (E2) (Individual)	Red, grey (E1); red, blue, grey (E2)	Participant number (E1), instruction with coloured background (E2)	Perceiving environment (E1, E2)	Perceiving red, relative to other achromatic and chromatic colours, would enhance both the force and velocity of strength output in humans (E1, E2)	(2) Physical (E1, E2)	(2) Maximum pinchgrip force (E1), Maximum handgrip force (in Newtons [N]), mean force over time (in Newtons), and slope toward maximum force (in Newtons/second) (E2)	x	Participants in the red condition pinched the clasp open with greater force than those in grey condition (E1); Participants in red condition squeezed in the handgrip test with greater maximum, mean force and slope toward maximum force than those in the blue and grey conditions, no difference for mean, maximum force and slope toward maximum force between blue and grey conditions (E2)	x
Etnier and Hardy (1997)	Experimental	42 Participants* (Mage =18.6)	Cycling test, fine motor skill test (Individual)	Orange, green, white	Wall	Perceiving environment	Warm colour (orange) would improve performance on the gross motor task and would hinder performance on the fine motor task (H1); the cool colour (blue) would improve performance on the fine motor task and would hinder performance on the gross motor task (H2); Colour influences affect and arousal (H3)	(1) psychological, (2) physical, (3) performance	(1) Zuckerman Inventory of Personal Reactions (ZIPERS) measured affective states; (2) heart rate, blood pressure (systolic, diastolic); Wingate Anaerobic cycling Test (WAT), (3) performance capabilities measured using the McCloy Blocks Test	Prior to performance of the MBT, positive affect decreased in the orange room more than in green and white rooms, positive affect decreased the least in the orange room prior to performance of the WAT, from pre-WAT to post-WAT desire to cope increased in the orange room, remained the same in the white room, and decreased in the green room, no effects on arousal, sadness, anger	No effects for colour for heart rate, systolic and diastolic blood pressure, No difference in performance on the WAT cycling test measuring anaerobic power as a function of colour	No differences in performance on the MBT (motor test) as a function of colour
Falco et al. (2016)	Retrospective	462 Fights* during three pre-Olympic tournaments and the 2012 Olympic Games	Taekwondo (Combat)	Red, blue	Body protector	X	Analyse the relationship between the colour protector and success in taekwondo combats of the qualification championships when electronic body protectors are used (H1); confounding effect of being a top-ranked athlete in the 2012 London Olympic Games: Number of wins in the qualification tournaments by players in blue should be similar to the number of wins scored by athletes in red uniforms (H2)	(3) Performance	(3) Winning probability	x	x	For the whole sample, no relationship between the combat outcome and the colour of the winner’s protector; For individual tournaments, positive relationship between wearing red and winning combat in Asian Qualification Tournament and in European Qualification Tournament; For gender and weight categories no clear colour effects
Feltman and Elliot (2011)	Experimental (E1, E2)	32Participants* (Mage = 20.2) (E1); 49 participants* (Mage = 19.8) (E2)	Taekwondo (E1, E2) (Combat)	Red, blue (E1, E2)	Jersey (E1, E2)	Wearing (E1), perceiving people (E2)	Analyse whether wearing red influences perceptions of relative dominance and threat in an imagined same-sex competitive context (H1) (E1); analyse whether viewing red on an opponent influences perceptions of relative dominance and threat in an imagined same-sex competitive context (H2) (E2)	(1) Psychological (E1, E2)	(1) Self-perceived dominance and threat (nine-point scale) (E1), perceived dominance and threat of the opponent (nine-point scale) (E2)	Wearing red relative to blue resulted in more dominant and threatening self-perceptions (E1); viewing red relative to blue on an opponent resulted in more dominant and threatening perceptions of the opponent (E2)	x	x
Frank and Gilovich (1988)	Experimental (E1, E3, E4),retrospective (E2)	25 Participants* (Mage not reported) (E1); 28 NFL and 23 NHL teams between 1970 and 1986 (E2); 40 Soccer fans* (Mage not reported) and 20 Soccer referees* (Mage not reported) (E3); 72 male students (Mage not reported) (E4)	Soccer and ice hockey (E1,E2) (Team); Soccer (E3,E4) (Team)	Black, other (E1,E2); black, white (E3, E4)	Uniform (E1, E2, E3, E4)	Wearing (E4), perceiving people (E3), perceiving environment (E1)	Do the uniforms of the black-uniformed teams in the NFL and NHL look more evil, mean, and aggressive than the uniforms of the nonblack-uniformed teams (H1) (E1); teams with black uniforms in the NFL and the NHL should be penalised more than other teams (H2) (E2); test whether the results obtained in our analysis of penalty records were due to the uniforms’ effect on the referees’ judgments (H3) (E3); test whether the results obtained in our analysis of penalty records were due to the players’ actual behaviour (H4) (E4)	(1) Psychological (E1, E2, E3, E4)	(1) Rating uniform on malevolence, activeness, and strongness (E1); Number of penalty minutes (E2), penalty ratings (Likert scale), aggressiveness ratings (put an X on a dotted line) for coloured and non-coloured video conditions for play 1 and 2 (E3); Rated level of aggressiveness involved in the games (aggressiveness scale) (E4)	Black uniforms look more malevolent (as well as more active and strong) than the non- black uniforms (E1); Teams with black uniforms are penalised more than their rivals with nonblack uniforms (E2), Referees/participants were more inclined to penalise the defensive team if they saw the black versions of the two plays than if they saw the white versions (E3); Participants in black showed an increase in intended aggression relative to those in white (E4)	x	x
Fujii et al. (2017)	Experimental	13 Male runners (Mage = not reported)	Cycling test (Individual)	Red, blue, no colour (control)	Goggles	Wearing	Analyse effects of colour on anaerobic power and the cardiopulmonary function or a sprint runner in track and field	(2) Physical	(2) Anaerobic maximal, maximum oxygen uptake (VO2max) on bicycle ergometer	x	No difference in the colour effect on anaerobic maximal or VO2max	x
Furley et al. (2012)	Experimental (E1)	22 Male soccer goalkeepers (Mage = 24.3) (E1)	Soccer (E1) (Individual)	Red, white, no colour (control) (E1)	Uniform (video) (E1)	Perceiving people (E1)	Red jerseys would increase the effect of dominant nonverbal behaviour on impression formation compared with white jerseys (E1)	(1) Psychological (E1)	(1) Perception of target player (seven perception of target scales), power and accuracy of penalty, expectancy of success (five-point scale) (E1)	Penalty takers in red shown in videos were not perceived more positively than players dressed in white or the neutral (control) condition (E1)	x	x
García-Rubio et al. (2011)	Retrospective	160 Observations of 20 teams in the First Division of the Spanish Football League between 2001/02 and 2008/09	Soccer (Team)	Red, others	Jersey	x	Analyse effects of a red shirt on the sporting performance (ability of a Soccer team to achieve the maximum sporting result from a given endowment of resources) of professional Spanish Soccer teams	(3) Performance	(3) Model to calculate average programme efficiency scores (number of league points, number of matches, number of spectators, etc.)	x	x	No relationship between colour red and better sporting performance/average programme efficiency scores by using a model/formula
Gilliam and Unruh (1988)	Experimental	54 Participants* (Mage = 31.0)	Hand grip strength test (Individual)	Pink, white	Carrel	Perceiving environment	Analyse effects of coloured stimuli, Baker–Miller pink and white on blood pressure, pulse rate, grip strength	(2) Physical, (3) performance	(2) Blood pressure (systolic, diastolic), pulse rate; grip strength measured by dynamometer for group 1 and 2, 3) Digit Symbol test score/completion time for group 1 and 2	x	No differences in blood pressure (systolic, diastolic) and pulse rate under white and pink visual stimuli, no differences for grip strength scores in group 1 and 2 between colour conditions	Digit Symbol Test scores under pink condition higher than under white condition, participants took less time to complete the test under the pink condition
Goldschmied and Lucena (2018)	Retrospective	441 Matches during NCAA National Basketball Tournament between 2012 and 2018	Basketball (Team)	Red, blue, black, green, white	Uniform	x	Red would win more than blue and the other colour uniformed lower-seeded teams vs. white uniform higher-seeded teams (H1); the effect would manifest itself especially when there were small differences in ability (through the ranking system) between the competing teams (H2); black clad teams would commit more fouls than other colours (H3)	(1) Psychological, (3) performance	(1) Aggression measured by the number of fouls; (3) Winning probability (for different degrees of relative ability)	No significant difference in the fouls committed based on the colour of the uniform	x	Lower ranked teams wearing either red, blue, black or green colours had similar winning success against higher ranked teams wearing white; Neither large nor small differences in ranking between the teams produced a significant effect on game outcome
Goldschmied and Spitznagel (2020)	Retrospective	503 Matches during women’s NCAA National Basketball Tournament between 2012 and 2019	Basketball (Team)	Red, blue, black, white	Uniform	x	Red would win more than blue and the other colour uniformed lower-seeded teams vs. white uniform higher-seeded teams (H1); the effect would manifest itself especially when there were small differences in ability (through the ranking system) between the competing teams (H2)	(3) Performance	(3) Winning probability (for different degrees of relative ability)	x	x	Lower ranked teams in women’s tournament wearing either red, blue or black colour had similar winning success against higher ranked teams wearing white; Neither large nor small differences in ranking between the teams produced a significant effect on game outcome
Green et al. (1982)	Experimental	30 Participants (Mage = not reported)	Hand grip strength, vertical jump, rotary pursuit tests (Individual)	Red, blue, pink	Light stimulus	Perceiving environment	Analyse effect of viewing colours on motor performance	(2) Physical, (3) performance	(2) Strength measured by the grip dynamometer, (3) jump height as measured by the vertical jump and precision as measured by the rotary pursuit apparatus	x	Mean grip strength was higher in red condition compared to blue and pink conditions, no difference for grip strength between blue and pink	No differences in performance while viewing different colour conditions for the vertical jump and rotary pursuit tasks
Greenlees et al. (2013)	Experimental	40 Male, collegiate soccer players (Mage = 21.9)	Soccer (Individual)	Red, blue, yellow, green	Jersey	Perceiving people	Participants who competed against an opponent wearing red would perform worse (number of penalties they scored) than participants who competed against an opponent wearing either a yellow-, green-, or blue-coloured uniform (H1); participants competing against red-clad opponents would have lower expectancies of success than participants competing against opponents dressed in other colours	(1) Psychological, (3) performance	(1) Expectancy of success (10-point scale), (3) penalty kick outcome	No effect of uniform colour on expectancy of success	x	Fewer goals scored against the goalkeeper in red compared to goalkeepers in blue or green, no difference between red and yellow conditions
Greenlees et al. (2008)	Experimental	12 Semi-professional male soccer goalkeepers (Mage = 21.4)	Soccer (Individual)	Red, white	Uniform (video)	Perceiving people	Participants would report lower expectancies of success in a penalty shootout against individuals displaying 90% gaze than for shootouts against individuals displaying 10% gaze (H1); participants would report lower expectancies of success for a penalty shootout scenario against individuals wearing red than for a shootout against individuals wearing white (H2)	(1) Psychological	(1) Perception of target player (nine-point scale), perception of target player’s penalty abilities (accuracy and power) (nine-point scale), expectancy of success (10-point scale) for different gaze durations (90% and 10%)	Penalty takers in red perceived more positive than those wearing white, no impact of colour on expectancies of success when models used 90% gaze, goalkeepers had lower expectancies of success when they viewed players using 10% gaze and wearing red compared to white	x	x
Gülle et al. (2016)	Retrospective	650 Competitions during Sakarya City Young Men Boxing Championship between 2005 and 2006	Boxing (Combat)	Red, blue	Uniform	x	Colours have an impact on sports performances of the youth	(3) Performance	(3) Winning probability	x	x	Sportsmen in blue more successful against the ones in red for the general total result, blue and red colour uniforms not relevant to competition results for the semi-final and final competitions
Hackney (2006)	Experimental	10 Male participants (Mage = not reported)	Cycling test (Individual)	Red, black	Uniform	Wearing, perceiving people	Analyse athletic men in a single exercise bout simulating a competitive effort to determine if red- coloured apparel influenced the testosterone responses	(1) Psychological, (2) physical	(1) Rate of perceived exertion (RPE); (2) VO2max, max power output and respiratory exchange ratio, pre and post-exercise hormonal testosterone concentration	Rate of perceived exertion did not differ across groups	VO2max, maximal power output, respiratory exchange ratio did not differ between colour groups; increase in the testosterone was observed in both treatment groups post-exercise at exhaustion measurement and at recovery measurement as compared to before exercise; no differences between colour groups in the before or post-exercise hormonal concentrations	x
Hagemann et al. (2008)	Experimental	42 Referees* (Mage = 29.3)	Taekwondo (Combat)	Red, blue	Shorts and head body protector (video)	Perceiving people	Analyse effect of the colour of the protective gear in tae-kwon do on the decisions of referees: changing the colour of the protective gear from blue to red would lead to an increase in points awarded, whereas changing the colour from red to blue would have the opposite effect	(3) Performance	(3) Mean number of points awarded	x	x	Competitors in red awarded more points than competitors in blue: number of points awarded increased for a blue competitor who was digitally transformed into a red competitor and decreased for a red competitor who was digitally transformed into a blue competitor
Hamid and Newport (1989)	Experimental	Six pre-school children* (Mage = not reported)	Hand grip strength test (Individual)	Blue, pink, grey	Light stimulus	Perceiving environment	Pink will enhance children’s physical strength (H1); the opposite effects would be observed when the children works in a blue (cool) environment (H2)	(1) Psychological, (2) physical	(1) Mood scores (seven-point scale), (2) Physical strength measured via ergometer	Higher mood scores in the pink room than in the blue room	Greater physical strength in the pink room than in the blue and grey room	x
Hasson et al. (1989)	Experimental	14 Participants* (Mage = 24.2)	Hand grip strength test (Individual)	Red, blue,unfiltered, no light (control)	Light stimulus	Perceiving environment	Viewing low wavelength light (blue) would result in a lowered EMG activity and force production than control (H1); viewing high wavelength light (red) would result in an elevated EMG activity and force production than control (H2)	(2) Physical	2) Electrical myographic activity and maximal force production measured with EMG device and dynamometer	x	Maximal force production was higher in red light condition compared to blue, unfiltered and control conditions; no differences between control, unfiltered and blue colour conditions for max. force production; Viewing red light resulted in higher EMG activity than blue, unfiltered and control colour conditions; no differences between control, unfiltered and blue colour conditions for EMG activity	x
Hill and Barton (2005)	Retrospective (E1, E2)	Male fights (sample size not reported) during Olympic Games 2004 (E1); Matches (sample size not reported) during Euro international soccer tournament 2004 (E2)	Boxing, Tae-kwondo, Greco roman wrestling, Freestyle wrestling (E1) (Combat); soccer (E2) (Team)	Red, blue (E1); red, blue, white (E2)	Body protector (E1), jersey E2)	x	Difference in number of matches won between combatants wearing red versus blue for Boxing, Tae-kwondo, GR, Free wrestling (H1); Difference in number of matches won between combatants wearing red versus blue for matches with small, medium, large, and similar asymmetries in competitive ability (H2) (E1); Difference in number of matches won and goals scored between the teams wearing red versus teams wearing blue or white at Euro 2004 (H3) (E2)	(3) Performance (E1, E2)	(3) Winning probability (for small, medium, large, similar degree of relative ability) (E1), Winning probability, amount of scored goals (E2)	x	x	Contestants in red win higher number of fights than contestants in blue, higher number of matches (with similar degree of competitive ability) won when combatants wearing red versus blue, no differences in win rate for matches with small, medium, large degree of relative ability (E1); Higher number of matches won and goals scored for teams in red vs blue and white (E2)
Ilie et al. (2008)	Retrospective	1347 Matches during Unreal Tournament Game 2004	Virtual sports (Team)	Red, blue	Uniform (video)	x	Analyse whether the red advantage observed in the Olympics extends to the world of virtual competition by comparing the performance of red and blue teams in a FPS game	(3) Performance	(3) Winning probability	x	x	Teams in red won higher amount of matches than teams in blue
Ingram and Lieberman (1985)	Experimental	48 Participants (Mage = not reported)	Hand grip strength test (Individual)	Red, blue, pink	Light stimulus	Perceiving environment	Participants in Group I (was told the study was to determine the effects of viewing colours on gross motor tasks) would have a stronger grip after viewing red (H1); the effect would not appear in Group 2 (was told that the study was to assess the effect of colour and physical exertion on memory) (H2)	(2) Physical	(2) Grip strength measured with dynamometer	x	No greater strength of grip after viewing red compared to blue or pink in group 1 and 2	x
Iwase (2002)	Experimental	University basketball club members and university students (sample size not reported)	Basketball (Team)	Red, blue, orange, green, white	Jersey	Wearing, perceiving people	Number of successful shots and passes will increase in order of the four colour conditions from 1 to 4 (1: red, blue, green, orange, white vs. Red, blue, green, orange, white; 2: white vs. white; 3: red vs. orange; 4: green vs. blue)	(3) Performance	(3) Successful shots, unsuccessful shots, successful passes, unsuccessful passes	x	x	No effects of uniform colour on successful passes, successful shots, unsuccessful passes and unsuccessful shots
Julio et al. (2015)	Retrospective	1233 Judo official combats (years not reported)	Judo (Combat)	Blue, white	Judogi	x	Analyse effect of judogi colours on competitive performance considering sex of athlete, age category, and competition level	(3) Performance	(3) Winning probability	x	x	Athletes of both sexes in blue won more frequently competing in regional, state level, junior and senior age categories than in white, no colour bias for pre-juvenile and juvenile athletes
Keller and Vautin (1998)	Experimental	83 Participants (Mage = not reported)	Hand grip strength test (Individual)	Red, purplish red, greyish purplish red, pink	Light stimulus	Perceiving environment	Analyse effect of colour viewing on hand-grip strength	(2) Physical	(2) Grip strength measured with dynamometer	x	No effect of viewed colour on hand-grip strength	x
Koslow (1983)	Experimental	32 Participants* (Mage = not reported)	Catching/reaction time test (Individual)	Red, blue, black, green, purple	Ball	Perceiving environment	Measure the stimulus background contrasts for five coloured balls presented against a woodgrain background on reaction time	(3) Performance	(3) Reaction times	x	x	No colour effect on reaction times
Krenn (2014)	Experimental	42 Male referees (Mage = 31.7), 81 participants* with high understanding of the rules of Soccer (Mage = 22.9), 82 participants* with minor understanding of the rules (Mage = 22.9)	Soccer (Team)	Red, blue, yellow, green	Uniform (jersey, stockings, shorts in video)	Perceiving people	Analyse effect of colour on tackle judgments in association Soccer (H1); analysis of the impact of uniform colour and level of expertise (participants with different degrees of understanding of the rules of Soccer) (H2)	(1) Psychological	(1) Judging tackles from behind, side and front on harshness on a scale from 1 to 9	Judgments of harshness of diverse tackles not affected by colours blue, green, red and yellow; neither when the tackle was committed by a player wearing one of those colours, nor when an player tackled, while wearing blue, green, red or yellow; referees and participants with a high level of understanding of the rules judged tackles from behind more harshly for players wearing red than blue	x	x
Krenn (2015)	Experimental (E1, E2, E3)	66 Participants* (Mage = 23.1) (E1); 66 participants* (Mage = 23.3) (E2); 62 participants* (Mage = 21.9) (E3)	Boxing, Tae-kwondo, wrestling (E1, E2, E3) (Combat)	Red, blue, green (E1, E2, E3)	Uniform (photo) (E1, E2, E3)	Perceiving people (E1, E2, E3)	Athletes wearing red uniforms should be judged more aggressive in combat sports than athletes wearing blue and green uniforms (H1) (E1); athletes wearing green uniforms would be judged fairer than athletes wearing blue or red (H2) (E2); athletes wearing red uniforms judged more likely to win than athletes wearing green or blue uniforms (H3) (E3)	(1) Psychological (E1, E2, E3)	(1) Perceived aggressiveness (E1), fairness (E2), chance of winning the fight (E3)	Boxing and wrestling athletes in red judged more aggressive than athletes in blue or green uniforms (E1); higher perceived chance of winning of boxing (E3), wrestling fighters in red compared to green; boxing and wrestling athletes judged fairer in green than in red (E2); In taekwondo no impact of uniform colour (E1, E2, E3)	x	x
Krenn (2018)	Retrospective	1530 Matches from seasons 2009/2010 to 2011/2012 of the first German Bundesliga and from seasons 2010/2011 and 2011/2012 of the second German Bundesliga	Soccer (Team)	Red, blue, black, green, yellow, white	Uniform (jersey, stockings, shorts)	x	For defending teams wearing green or black, more offside judgments were expected against their opponents’ attacks (H1); for attacking teams wearing green or black uniforms fewer offside judgments were assumed (H2)	(1) Psychological	(1) Number of offside judgments for attacking/defending teams	Lower number of offside judgments when attacking team wore black; Defending teams in black and green attended by the highest number of offside judgments given against opposing teams	x	x
Krenn et al. (2020)	Experimental (E1, E2)	58 Participants* (Mage = 24.43) (E1); 32 participants* (Mage = 23.93) (E2)	Knee strength test (E1, E2) (Individual)	Red, blue, grey (E1, E2)	Instructions on computer screen (E1, E2)	Perceiving environment (E1, E2)	Scrutinising the impact of perceiving red on peak torque and rate of torque production with colour manipulation implemented 5 s prior until the end of maximal torque measurement (H1) (E1); and with colour manipulation implemented 50 seconds prior the onset of maximal torque measurement (H2) (E2)	(1) psychological, (2) physical (E1, E2)	(1) Quality rating of own attempts via four bipolar items (bad-good, lethargic-energetic, lazy-explosive, powerless-powerful); (2) Peak torque (PT) and rate of torque development (RTD) measured with dynamometer (E1, E2)	Participants rated their own attempts better after viewing red than after viewing blue or grey (E1); self-reported quality of participants’ attempts did not turn out be affected by colour condition (E2)	No significant impact of colour on PT, red was found to show a significantly higher RTD than blue, whereas the difference to grey did not prove significance (E1); no significant impact of colour condition on PT or RTD (E2)	x
Krenn et al. (2017)	Experimental (E1, E2)	43 Nonprofessional male Soccer players (Mage = 25.6) (E1); 118 students* (Mage = 22.2) (E2)	Soccer (E1) (Team); soccer and handball (E2) (Team)	Red, blue (E1, E2)	Uniform (jersey, shorts in video) (E1, E2)	Perceiving people (E1, E2)	Any potential benefit of red uniforms in soccer penalty kicks was examined (H1) (E1); any potential benefit of red uniforms in handball penalties was examined (H2) (E2)	(1) Psychological (E1, E2)	(1) Ratings of players’ confidences to score, expected lateral position/height of ball (E1), ratings of players’ confidences to score, expected lateral position/height of ball, probability of penalty being scored (9-point scale) (E2)	No impact of colour on ratings of confidence and expected lateral position of the ball; Penalty takers in red expected to kick the ball lower than those in blue (E1); Neither the ratings of player’s/goalkeeper’s confidence level, probability of scoring the penalty, nor expected position of the ball differed significantly for blue and red conditions (E2)	x	x
Lam et al. (2017)	Experimental	36 Male university basketball players (Mage = not reported)	Vertical jump test (Individual)	Red, blue, black	Shoes	Wearing	Shoe colour influence perceptual and biomechanical variables in vertical countermovement jump	(1) Psychological, (3) performance	(1) Perceived body height, perceived jump height; (3) maximum jump height, propulsion peak, flight time, impulse measured via GRF	No shoe colour effect on perceived max jump height, perceived body height in total; red-preference participants perceived themselves taller when wearing red compared to black shoes	x	No colour effect on max. jump height, flight time, max propulsion GRF, propulsion impulse; Black-preference participants displayed longer flight time in red compared to blue condition
Larionescu and Pantelimona (2012)	Experimental	10 Participants (mage = not reported)	Basketball (individual)	Red, blue	Light stimulus	Perceiving environment	Red, by comparison to blue, should improve performance of basketball throws	(3) Performance	(3) Success rate/efficiency of free throws	x	x	Efficiency/success rate of basketball throws were not different by colour, Efficiency/success rate of basketball throws when colours alternated were higher in red condition compared to blue
Londe et al. (2018)	Experimental	10 Male amateur soccer players (Mage = 21.0)	Running test (Individual)	Red, blue, no colour (control)	Lens glasses	Wearing	Glasses with red lenses increase both the testosterone concentration and exercise performance	(1) psychological, (2) physical, (3) performance	(1) Rate of perceived exertion/recovery; (2) testosterone concentration (pre and post measurement), blood lactate, relative heart rate (percentage of maximum), delta HR values (difference between resting HR and during the analysed time), (3) running distance	Perceived exertion, perceived recovery did not differ among conditions	Relative HR, blood lactate and absolute testosterone concentration did not differ among conditions, faster HR recovery in blue than red condition, testosterone (pre-post measurement) concentration increased more in red condition than in blue and control	Running distance did not differ among colour conditions
Matsumoto et al. (2007)	Retrospective	2794 Fights* during 2001, 2003, 2005 Judo World Championships and Olympic Games 2004	Judo (Combat)	Blue, white	Judogi	x	Blue judogis are associated with a winning bias in international judo competitions	(3) Performance	(3) Winning probability	x	x	Male judo athletes more likely to win a fight in blue than in white, colour bias got larger as matches proceeded in tournament, win ratio increased over four years of analysis, no differences for females
Mentzel et al. (2019)	Experimental	32 Participants* (Mage = 23.81)	Running (Individual)	Red, blue, grey	JERSEY (video)	Perceiving people	Runners depicted in red will be perceived as running at faster speeds than runners in blue (H1); beneficial red effect for the other performance-related parameters (power, posture, dynamics, economics, and fitness) (H2); Association between red and dominance and blue and rest in a Stroop task (H3); Colour effects for the individual colour association task will covary with individual colour effects when judging running speed (H4)	(3) Performance	(3) Perceived running speed, power, posture, dynamics, economics, and fitness; response time on a modified Stroop task	x	x	Runners shown in red being rated as running faster compared to runners in blue; results for other sport-related performance parameters (power, posture, dynamics, economics, and fitness) showed no significant differences between red and blue; rest-related words shown in blue or grey were classified faster compared to when shown in red and vice versa for the dominance-related words; no covariate effect of the modified Stroop task for the speed perception colour effect
Morris et al. (1994)	Experimental	12 College softball players (Mage = not reported)	Softball (Individual)	Yellow, white	Ball	Perceiving environment	Analyse effects of softball speed and colour on batting performance of collegiate players	(3) Performance	(3) Number of hits and rating of hits quality with production rating analysis chart	x	x	No difference in number of hits between colour groups, the mean quality of hit ratings for the yellow ball were lower at medium and fast speeds than for the white ball, At the slow speed, the quality ratings were higher for the yellow ball than for the white ball
O’Connell et al. (1985)	Experimental	40 Male undergraduate students (Mage = not reported)	Hand grip strength test (Individual)	Red, green	Wall	Perceiving environment	Arousing quality of colour is primarily responsible for the difference in the strength performance of participants	(2) Physical	(2) Grip strength measured with dynamometer	x	Greater grip strength scores when exposed to the red compared to green	x
Olde Rikkert et al. (2015)	Experimental (E1), retrospective (E2)	18 Students (Mage not reported) (E1), 10 soccer clubs from Premier League and Eredevise between 1995 and 2013 (E2)	Soccer (E1, E2) (Team)	Red, green, white (E1), not reported (E2)	Jersey (video) (E1), uniform (E2)	Perceiving people (E1)	Performance of a Soccer team may be improved by choosing kit colours with high visibility in the Soccer field environment (E1, E2)	(3) Performance (E1, E2)	(3) Tracking/assessment of player positions (E1); Conceded goals, total amount of points won (E2)	x	x	More difficult evaluation of players’ positions on the field for green than for white tricots (E1); Manchester City conceded in proportion to their home matches less goals in away games in high-contrast kits, Newcastle United in comparison to the home record conceded less goals and won more points while playing away matches in low-contrast kits, For the other eight teams, no effects of away jersey colour on match results (E2)
Payen et al. (2011)	Experimental	39 Male undergraduates (Mage = 23.2)	Leg strength test (Individual)	Red, blue, grey	Rectangle	Perceiving environment	Red stimulus viewed prior to the onset of motor action would produce slower and/or less forceful responding (H1); any observed effects would take place independent of general arousal and mood (H2)	(1) Psychological, (2) physical	(1) Self-reported arousal and mood states (nine-point scales); (2) grip strength (maximal voluntary contraction, rate of force development) measured with dynamometer	No differences between colour conditions for arousal and mood scores	No differences between colour conditions for baseline maximal voluntary contraction and baseline rate of force development, viewing red reduced peak rate of change of force development, but not in blue or grey conditions	x
Pellegrini and Schauss (1980)	Experimental	72 Right-handed participants* (Mage = not reported)	Hand grip strength test (Individual)	Blue, pink	Cardboard plate	Perceiving environment	Exposure to visual stimuli differing in hue can differentially affect muscle strength	(2) Physical	(2) Grip strength measured with dynamometer	x	Greater strength response to the blue than to the pink stimulus	x
Pellegrini et al. (1980)	Experimental	60 Male students (Mage = not reported)	Leg strength test (Individual)	Blue, pink	Cardboard plate	Perceiving environment	Analyse strength in a large skeletal muscle group across different colour conditions	(2) Physical	(2) Physical strength of quadriceps femoris measured via goniometer	x	Greater strength response while focused on a blue than pink stimulus	x
Pellegrini et al. (1981)	Experimental	36 Male right-handed school students (Mage = not reported)	Hand grip strength test (Individual)	Blue, pink	Cardboard plate	Perceiving environment	Investigate magnitude of colour effects on the grip strength of participants tested while standing, rather than sitting	(2) Physical	(2) Grip strength measured with dynamometer	x	No difference between hand dynamometer scores during pink and blue conditions	x
Piatti et al. (2012)	Retrospective	5604 Matches from Australian Rugby League between 1979 and 2008	Rugby (Team)	Red, others	Jersey	x	Analyse the presence of any red effect within the Australian National Rugby League	(3) Performance	(3) Winning probability	x	x	Red home teams win more often than teams in other colours
Pollet and Peperkoorn (2013)	Retrospective	210 Fights during Ultimate Fighting Championship (years not reported)	Mixed martial arts (Combat)	Red, blue, black, white, other	Shorts	x	Analyse a winning red effect based on the colour of shorts in the ultimate fighting championship	(3) Performance	(3) Winning probability	x	x	No effects of colour on fight outcome
Profusek and Rainey (1987)	Experimental	46 Undergraduate students* (Mage = not reported)	Hand grip strength test (Individual)	Red, pink	Light stimulus	Perceiving environment	Participants exposed to a room painted Baker-Miller Pink have lower state-anxiety scores, poorer grip strength, and better motor precision than participants exposed to a room painted in red	(1) Psychological, (2) physical	(1) State anxiety (State-Trait Anxiety Inventory), (2) grip strength measured via dynamometer, motor precision (Milton Bradley Game)	State-Anxiety scores were higher for participants in the red room than those in the Baker-Miller Pink room	No difference in grip strength and number of errors made in operation game between red and pink conditions	x
Recours and Briki (2015)	Experimental	60 Male undergraduate students (Mage = 22.0)	Virtual sports (Individual)	Red, blue	Gloves, shorts	Perceiving people	Competing with an opponent dressed in blue should generate more self-confidence than when competing with an opponent dressed in red (H1); competing with an opponent dressed in red should generate more competitive anxiety than when competing with an opponent dressed in blue (H2)	(1) Psychological	(1) Self-confidence, somatic anxiety, cognitive anxiety (Competitive State Anxiety Inventory 2 Revised)	No significant colour effect on somatic anxiety; self-confidence was higher in the blue condition than in the red condition; cognitive anxiety was higher in the red condition than in the blue condition	x	x
Rezaeian et al. (2015)	Experimental	30 Undergraduate students* (Mage = not reported)	Hand grip strength test (Individual)	Red, blue, no colour (control)	Computer screen	Perceiving environment	Analyse effects of colour of red and blue on grip strength and fatigue	(2) Physical	(2) Grip strength (Maximal grip strength, grip velocity, fatigue velocity and fatigue percentage) measured with dynamometer	x	No differences between colour conditions for maximal grip strength, grip velocity, fatigue velocity and fatigue percentage	x
Rowe et al. (2005)	Retrospective	301 Male judo fights in the Athens Olympics 2004	Boxing, tae-kwon do, Greco–Roman, freestyle wrestling (Combat)	Blue, white	Judogi	x	Analyse whether a winning bias occurs in contests in which neither contestant wears red, but instead blue and white	(3) Performance	(3) Winning probability for different degrees of relative ability (no asymmetry, small, medium and large asymmetries)	x	x	When considering only contests in the first round of competition higher winning probability for blue than white condition, higher number of matches (no asymmetry, with small, medium and large asymmetries in competitive ability) won when combatants wearing blue versus white
Smith et al. (1986)	Experimental (E1, E2)	59 Undergraduate students* (Mage = not reported) (E1); 54 students* (Mage = not reported) (E2)	Hand grip strength test (E1, E2) (Individual)	Red, blue, light blue, green, orange, pink, yellow, brown (E1); blue, pink, green, brown (E2)	Panels (E1, E2)	Perceiving environment (E1, E2)	Analyse whether demand characteristics play a role in the pink-weaken effect on muscle strength, as well as on the blue-strengthen effect (H1) (E1), Replicate Experiment 1 but with a control group added (H2) (E2)	(1) Psychological, 2) physical (E1, E2)	(1) Affective scales (Russell and Pratt's questionnaire), (2) Grip strength measured with dynamometer (E1, E2)	Blue and light blue were less distressing and more pleasant than red, light blue was more relaxing than was pink or red, and blue and light blue were rated as more sleepy and less exciting than red, pink and blue were more arousing under the pink-weaken instructions than under the pink-strengthen instructions (E1); brown was more distressing, less pleasant, less arousing, gloomier, and less exciting than either blue or pink, brown was less sleepier than pink and more unpleasant than blue, pink was less gloomy than blue and green, blue was more relaxing than green, women in the control condition viewing blue colour rated it as less unpleasant than women in either the pink-weaken or pink-strengthen condition viewing the colour brown, or the women in the pink- weaken condition viewing the colour green, they also rated the colour as less unpleasant than men in the pink-weaken condition viewing the colour brown (E2)	Men viewing pink and orange panels had higher grip strength under pink-strengthen than under pink-weaken instructions, for colour green the pink- weaken instruction resulted in greater strength than the pink-strengthen instruction; For women, the pink-weaken instructions resulted in a higher grip strength than did the pink-strengthen instructions, regardless of actual colour present (E1); No differences between colour conditions (E2)	x
Sorokowski and Szmajke (2007)	Experimental	45 Male karate and amateur boxers (Mage = not reported)	Boxing (Combat)	Red, blue, black/white (control)	Jersey (photo)	Perceiving people	Having a red decoration has a stimulating influence on its possessor, increasing his will to fight and thus facilitating his victory (H1); Red colour causes arbiter’s tendentious perception (“red” competitor seems more brave, aggressive etc. which increases his chances for being chosen a winner when the fight is fairly equal) (H2); Red colour causes the opponent’s tendentious perception (“Red” competitor seems more redoubtable and dangerous to the opponent, which might weaken the will to fight of the competitor, who is dressed in some other colour, and make it more difficult for him to win) (H3)	(1) Psychological	(1) Perceived physical preparation, technical abilities and bravery (seven-point scale) of Boxer A and B	No differences between colour conditions in perception of Boxer A on scales of physical preparation, technical abilities, bravery as well as for Boxer B on scales of physical preparation, technical abilities; perception of boxer B as more brave when depicted in red compared to blue clothing or a black-and-white photograph	x	x
Sorokowski et al. (2014)	Experimental	147 Students* from Poland (Mage = 22.1) and 143 students* from China (Mage = 23.6)	Boxing (Combat)	Red, blue, black	Shorts (video)	Perceiving people	Confirming the existence of a “black wins” effect in an experimental study (H1); Analyse blue, red and black simultaneously (H2); Analyse whether the “red wins” and “black wins” effects are related to the culture of the stereotypical Western perception (H3)	(1) Psychological	(1) Perceived number of blows (supremacy) for video boxers	Black and red trunks result in higher number of perceived blows when fought against a boxer in blue; no differences in perceived number of blows between red and black; in the case of the winning boxer, colour did not influence his perceived supremacy; for the losing boxer, blue-trunked boxer’s loss was more clearly perceived than the red and black-trunked boxer’s loss; cross-cultural differences (between Polish and Chinese participants) of the influence of boxing trunk colours on the number of blows was not confirmed	x	x
Tiryaki (2005)	Retrospective	2142 Turkish premier soccer league games played between 1995/1996 and 2002/2003	Soccer (Team)	Black, other	Uniform	x	Analyse whether Frank and Gilovich’s study (1988) approach that black uniforms are associated with a greater number of penalties is valid for the Turkish culture	(1) Psychological	(1) Number of yellow and red card penalties, total amount of penalty kicks given	No difference for teams with or without black uniforms on how often they were penalised, how many penalty kicks were given to their advantage, for the total number of penalty kicks	x	x
Webster et al. (2012)	Retrospective (E1, E2)	30 NHL teams across 52,098 games between 1984/1985 and 2009/2010 (E1); 30 NHL teams across 22,140 games between 2000/2001 and 2009/2010 (E2)	Ice hockey (E1, E2) (Team)	Black, other (E1); white, black, other (E2)	Jersey (E1,E2)	x	Black uniforms would be related to more penalty minutes than nonblack uniforms for the average team (H1) (E1); the switch from white to coloured jerseys would be related to increased penalty minutes at home for the average team (H2) (E2)	(1) Psychological (E1, E2), (3) performance (E2)	(1) Total assessed penalty minutes (aggressiveness) measured in penalty infraction minutes (PIM) (E1); (2) Bench minor infraction penalty minutes for aggressive (BMI) and non-aggressive (non-BMI) penalties; (3) analysis of total home goals (E2)	Teams were more aggressive (i.e., assessed more penalty minutes) in black versus other coloured jerseys; black- versus-coloured uniform effect generalised to aggressive penalties but not non-aggressive ones (E1); Teams were significantly more aggressive (i.e., assessed more penalty minutes) at home games after they wore coloured jerseys during the 2003–2004 season vs. wearing white jerseys before 2003–2004 (E2)	x	No difference between white and coloured home jersey colour for total home goals (E2)
Williams et al. (2011)	Experimental	14 Right-handed participants* (Mage = 26.0)	Goal-directed aiming test (Individual)	Red, blue, green, yellow, no colour (control)	Lens glasses	Wearing	If different colours have unique effects on action in accordance with unique effects on affective state, then a pattern of kinematic results related to the colours would emerge (H1); If different colours have unique effects on motivation, then we expected to observe a pattern of kinematic results related to the colours known to induce avoidance or approach motivations (H2); If only the colour red can modulate the emotion-action link, the affective report results indicate no differences between the colours (H3)	(1) psychological, (3) performance	(1) Perceived positive affect (PANAS questionnaire); 3) reaction time, movement time, peak acceleration (PA), time to PA, displacement at PA, peak velocity (PV), time to PV, displacement at PV, peak deceleration (PD), time to PD, displacement at PD, constant error and variable error	Perceived positive affect smaller in a red environment than in a yellow, green or clear visual environment, no effects for blue colour	x	Reaction time, movement time, peak acceleration (PA) , time to PA, displacement at PA, peak velocity (PV), time to PV, displacement at PV, peak deceleration (PD), time to PD, displacement at PD, constant error revealed no colour effects; variable error was greater in a red environment compared to green or control environments

Note. E = Experiment, * = participants of both genders, H = Hypothesis, Colour location = only applicable to experimental studies.

Table 2. Frequency distribution of moderators and variables in all included studies (and sub-studies).

Moderator	Variable	N
Study design	Experimental	58
	Retrospective	26
Sport type/activity	Strength test	22
	Combat sport (judo, boxing, wrestling, taekwondo)	20
	Soccer	14
	Cycling (test)	9
	Ice hockey	5
	Rugby/Soccer	4
	Basketball	4
	Virtual sport	2
	Running	2
	Jumping	2
	Catching/reaction time test	2
	Aiming test	1
	Handball	1
	Softball	1
	Rotary pursuit test	1
	Fine motor skill test	1
Context	Individual (without combat)	41
	Combat	20
	Team	23
Colour	Red	58
	Blue	57
	White	26
	Black	18
	Green	17
	Grey	12
	Pink	12
	Yellow	11
	Not specified/other	10
	No colour/control	8
	Orange	4
	Brown	2
	Purple	1
Object in colour	Uniform	15
	Jersey	14
	Other (panel, gloves, rectangle, carrel, goggles, etc.)	11
	Light stimulus	9
	Judogi	5
	Body protector	4
	SHORTS	4
	Lens glasses	3
	Carboard plate	3
	Ball	3
	Computer screen	3
Colour location	Perceiving environment	32
	Perceiving people	20
	Wearing	10
Domain	Psychological	42
	Performance	39
	Physical	28

Quality assessment

For all eligible studies, the percentual average score on the QATSDD ranged from 14% to 82% (M = 51%, SD = 17%) and the percentual average score in terms of the criteria by Elliot ranged from 0% to 100% (M = 34%, SD = 26%). The inter-rater reliability was calculated by the weighted Cohen’s kappa in terms of the percentual average score and indicated a perfect level of agreement for the criteria by Elliot (κ = 1.00 (95% CI, 1.00–1.00), p < .001) and an almost perfect level of agreement for the QATSDD (κ = .89 (95% CI, 0.86–0.92), p < .001). An overview of the quality assessment is depicted in Table 3.

Table 3. Quality assessment of included studies.

Authors (date of publication)	Reviewer	A1	A2	A3	A4	A5	A6	A7	A8	A9	A10	A12	A13	A15	A16	Total score A	% A	B1	B2	B3	B4	B5	Total score B	% B
Akers et al. (2012)	JW	1	3	2	0	1	3	2	0	0	3	3	3	0	1	22	52%	0	1	0	1	0	2	40%
	SM	1	3	2	0	1	3	2	0	0	3	2	3	0	1	21	50%	0	1	0	1	0	2	40%
Allen and Jones (2014)	JW	3	2	2	1	2	2	3	x	0	3	3	0	0	2	23	59%	x	x	x	x	x	x	x
	SM	3	1	2	1	2	1	3	x	0	3	3	0	0	2	21	54%	x	x	x	x	x	x	x
Attrill et al. (2008)	JW	2	3	2	1	2	2	1	x	1	3	3	3	0	2	25	64%	x	x	x	x	x	x	x
	SM	1	3	2	1	2	3	1	x	1	3	3	3	0	2	25	64%	x	x	x	x	x	x	x
Briki et al. (2015)	JW	3	3	3	1	1	3	2	2	0	3	3	3	1	3	31	74%	1	1	0	1	0	3	60%
	SM	3	3	3	1	1	3	2	2	0	2	3	2	1	3	29	69%	1	1	0	1	0	3	60%
Caldwell and Burger (2011)	JW	2	1	2	1	2	2	1	x	0	2	2	0	0	1	16	41%	x	x	x	x	x	x	x
	SM	2	1	2	1	2	1	1	x	0	2	2	0	0	1	15	38%	x	x	x	x	x	x	x
Carazo-Vargas and Moncada-Jiménez (2014)	JW	2	2	2	1	3	1	1	x	0	2	1	3	0	0	18	46%	x	x	x	x	x	x	x
	SM	2	2	2	1	3	1	1	x	0	2	2	3	0	0	19	49%	x	x	x	x	x	x	x
Chantal and Bernache-Assollant (2015)	JW	3	3	3	1	1	2	2	1	0	3	3	0	3	3	28	67%	1	1	0	1	0	3	60%
	SM	3	3	3	1	1	2	2	1	0	3	2	0	3	3	27	64%	1	1	0	1	0	3	60%
Crane et al. (2008)	JW	2	3	3	1	1	3	2	0	0	3	3	0	0	1	22	52%	1	0	0	0	1	2	40%
	SM	2	3	3	1	1	3	2	0	0	3	2	0	0	1	21	50%	1	0	0	0	1	2	40%
Curby (2016)	JW	1	1	1	1	2	1	0	x	0	1	1	0	0	0	9	23%	x	x	x	x	x	x	x
	SM	1	1	2	1	2	1	0	x	0	2	1	0	0	0	11	28%	x	x	x	x	x	x	x
Dijkstra and Preenen (2008)	JW	2	3	3	1	3	3	2	x	0	3	3	2	0	3	28	72%	x	x	x	x	x	x	x
	SM	3	3	3	1	3	3	2	x	0	3	3	2	0	3	29	74%	x	x	x	x	x	x	x
Dijkstra et al. (2018)	JW	2	3	3	1	3	2	1	x	0	2	3	3	0	3	26	67%	x	x	x	x	x	x	x
	SM	2	3	3	1	3	2	2	x	0	2	3	3	0	3	27	69%	x	x	x	x	x	x	x
Don Morris (1976)	JW	2	2	2	1	2	2	2	0	2	2	2	0	1	1	21	50%	0	1	0	0	1	2	40%
	SM	2	3	2	1	2	2	2	0	2	2	2	1	1	1	23	55%	0	1	0	0	1	2	40%
Dreiskaemper et al. (2013)	JW	2	3	3	1	1	3	1	1	2	3	3	2	0	3	28	67%	1	0	0	0	1	2	40%
	SM	2	3	3	1	1	2	1	1	2	2	2	1	0	3	24	57%	1	0	0	0	1	2	40%
Dunwoody (1998	JW	1	3	2	1	1	2	1	0	0	2	1	0	0	1	15	36%	0	1	0	1	0	2	40%
	SM	1	3	1	1	1	2	2	0	0	2	2	0	0	1	16	38%	0	1	0	1	0	2	40%
Elliot and Aarts (2011)	JW	3	3	3	1	1	2	1	0	0	3	3	0	0	2	22	52%	1	1	0	1	1	4	80%
	SM	3	3	3	1	1	2	2	0	0	3	3	1	0	2	24	57%	1	1	0	1	1	4	80%
Etnier and Hardy (1997)	JW	3	3	2	0	1	3	2	0	3	3	3	0	0	1	24	57%	0	1	0	1	1	3	60%
	SM	2	3	2	0	1	3	2	0	3	3	3	0	0	1	23	55%	0	1	0	1	1	3	60%
Falco et al. (2016)	JW	3	2	3	1	3	2	1	x	0	1	3	3	0	3	25	64%	x	x	x	x	x	x	x
	SM	3	2	3	1	3	2	1	x	0	1	3	2	0	3	24	62%	x	x	x	x	x	x	x
Feltman and Elliot (2011)	JW	2	1	3	1	1	3	1	0	0	3	3	0	0	3	21	50%	1	1	0	1	0	3	60%
	SM	2	2	3	2	1	3	1	0	0	2	2	0	0	3	21	50%	1	1	0	1	0	3	60%
Frank and Gilovich (1988)	JW	2	2	3	1	1	1	1	x	0	2	2	1	0	2	18	43%	x	x	x	x	x	x	x
	SM	2	3	3	1	1	1	1	x	0	2	2	1	0	2	19	45%	x	x	x	x	x	x	x
Fujii et al. (2017)	JW	2	3	2	1	1	2	1	0	0	3	1	1	0	1	18	43%	0	1	0	0	0	1	20%
	SM	2	3	2	1	2	2	1	0	0	3	2	1	0	1	20	48%	0	1	0	0	0	1	20%
Furley et al. (2012)	JW	3	2	2	1	1	3	3	0	0	3	2	3	0	2	25	60%	1	0	0	0	0	1	20%
	SM	3	3	2	1	2	3	3	0	0	3	2	3	0	2	27	64%	1	0	0	0	0	1	20%
García-Rubio et al. (2011)	JW	1	2	2	1	2	3	2	x	0	3	1	2	0	1	20	51%	x	x	x	x	x	x	x
	SM	1	2	2	1	2	3	3	x	0	3	2	2	0	1	22	56%	x	x	x	x	x	x	x
Gilliam and Unruh (1988)	JW	2	3	2	1	1	3	1	0	0	3	1	0	0	1	18	43%	0	0	0	0	0	0	0%
	SM	2	3	2	1	1	3	1	0	0	3	1	0	0	1	18	43%	0	0	0	0	0	0	0%
Goldschmied and Lucena (2018	JW	2	3	3	2	3	3	3	x	2	3	3	1	0	3	31	79%	x	x	x	x	x	x	x
	SM	2	3	3	2	3	3	3	x	2	3	3	1	0	3	31	79%	x	x	x	x	x	x	x
Goldschmied and Spitznagel (2020)	JW	3	3	3	2	3	3	2	x	0	3	3	0	0	3	28	72%	x	x	x	x	x	x	x
	SM	3	3	3	2	3	3	2	x	0	3	3	0	0	3	28	72%	x	x	x	x	x	x	x
Green et al. (1982)	JW	0	2	1	0	1	1	0	0	0	1	2	1	0	0	9	21%	0	1	0	0	1	2	40%
	SM	0	2	1	0	1	1	0	0	0	1	2	1	0	0	9	21%	0	1	0	0	1	2	40%
Greenlees et al. (2013)	JW	2	3	3	0	1	3	3	0	0	3	3	3	2	3	29	69%	1	1	0	0	0	2	40%
	SM	2	3	3	0	1	3	2	0	0	3	3	3	2	3	28	67%	1	1	0	0	0	2	40%
Greenlees et al. (2008)	JW	2	3	3	1	1	3	3	0	0	3	3	3	0	3	28	67%	0	1	0	0	0	1	20%
	SM	2	3	3	1	1	3	3	0	0	3	3	2	0	3	27	64%	0	1	0	0	0	1	20%
Gülle et al. (2016)	JW	1	1	2	1	3	2	0	x	0	1	2	1	0	1	15	38%	x	x	x	x	x	x	x
	SM	1	1	2	1	3	2	1	x	0	1	2	1	0	1	16	41%	x	x	x	x	x	x	x
Hackney (2006)	JW	1	2	1	1	1	3	2	0	0	3	3	3	0	2	22	52%	1	0	0	0	0	1	20%
	SM	1	2	1	1	1	3	2	0	0	3	3	3	0	2	22	52%	1	0	0	0	0	1	20%
Hagemann et al. (2008)	JW	2	1	2	1	2	3	1	0	0	2	3	2	0	1	20	48%	0	0	0	0	0	0	0%
	SM	1	1	2	1	2	2	1	0	0	2	3	2	0	1	18	43%	0	0	0	0	0	0	0%
Hamid and Newport (1989)	JW	2	3	2	0	1	2	1	0	0	2	1	0	0	1	15	36%	0	1	0	0	1	0	40%
	SM	2	3	2	0	1	2	1	0	0	2	1	0	0	1	15	36%	0	1	0	0	1	0	40%
Hasson et al. (1989)	JW	1	3	2	0	1	2	1	0	1	3	2	2	0	2	20	48%	0	1	0	0	1	0	40%
	SM	1	3	2	0	1	2	1	0	1	3	2	2	0	2	20	48%	0	1	0	0	1	0	40%
Hill and Barton (2005)	JW	1	2	2	0	1	0	0	x	0	1	1	0	0	1	9	23%	x	x	x	x	x	x	x
	SM	1	1	2	0	1	0	0	x	0	1	1	0	0	1	8	21%	x	x	x	x	x	x	x
Ilie et al. (2008)	JW	1	2	1	1	1	2	0	x	0	1	2	1	0	1	13	33%	x	x	x	x	x	x	x
	SM	0	2	1	1	1	2	0	x	0	1	2	1	0	1	12	31%	x	x	x	x	x	x	x
Ingram and Lieberman (1985)	JW	1	1	1	0	1	1	0	0	0	1	1	0	0	0	7	17%	0	1	0	0	0	1	20%
	SM	0	1	1	0	1	1	0	0	0	1	1	0	0	0	6	14%	0	1	0	0	0	1	20%
Iwase (2002)	JW	0	1	0	0	1	1	0	0	0	1	1	0	0	1	6	14%	0	0	0	1	0	1	20%
	SM	0	1	0	0	1	1	0	0	0	1	1	0	0	1	6	14%	0	0	0	1	0	1	20%
Julio et al. (2015	JW	2	2	3	1	3	3	1	x	0	3	3	3	0	2	26	67%	x	x	x	x	x	x	x
	SM	2	2	3	1	3	3	1	x	0	3	3	6	0	2	29	74%	x	x	x	x	x	x	x
Keller and Vautin (1998)	JW	2	2	2	1	2	3	1	1	0	3	2	0	0	1	20	48%	0	1	0	1	1	3	60%
	SM	2	2	2	1	1	3	1	1	0	3	2	0	0	1	19	45%	0	1	0	1	1	3	60%
Koslow (1983)	JW	2	2	3	0	1	3	1	0	0	3	2	0	0	1	18	43%	0	1	0	0	1	2	40%
	SM	2	2	2	0	1	3	1	0	0	3	2	0	0	1	17	40%	0	1	0	0	1	2	40%
Krenn (2014)	JW	2	2	3	1	2	3	2	0	0	3	3	2	0	3	26	62%	1	0	0	1	0	2	40%
	SM	2	2	3	1	3	3	2	0	0	3	3	2	0	3	27	64%	1	0	0	1	0	2	40%
Krenn (2015)	JW	3	3	3	1	2	3	2	0	0	3	3	2	0	3	28	67%	0	1	0	1	0	2	40%
	SM	3	3	3	1	2	3	2	0	0	3	3	1	0	3	27	64%	0	1	0	1	0	2	40%
Krenn (2018)	JW	2	3	3	1	3	3	3	x	2	3	3	3	0	3	32	82%	x	x	x	x	x	x	x
	SM	2	3	3	1	3	3	3	x	2	3	3	3	0	3	32	82%	x	x	x	x	x	x	x
Krenn et al. (2020)	JW	2	3	3	3	2	3	3	0	0	3	2	2	0	3	29	69%	1	1	1	1	1	5	100%
	SM	2	3	3	3	2	3	3	0	0	3	2	3	0	3	30	71%	1	1	1	1	1	5	100%
Krenn et al. (2017)	JW	3	3	3	1	1	3	2	1	0	3	3	3	0	3	29	69%	1	1	0	1	1	4	80%
	SM	3	3	3	1	1	3	2	1	0	3	3	3	0	3	29	69%	1	1	0	1	1	4	80%
Lam et al. (2017)	JW	1	2	1	1	1	1	0	0	0	1	2	1	0	0	11	26%	0	1	0	0	0	1	20%
	SM	1	2	1	1	1	1	0	0	0	1	2	1	0	0	11	26%	0	1	0	0	0	1	20%
Larionescu and Pantelimona (2012)	JW	1	1	1	1	1	2	1	0	0	1	1	0	0	0	10	24%	1	1	0	0	0	2	40%
	SM	2	1	1	1	1	2	1	0	0	1	1	0	0	0	11	26%	1	1	0	0	0	2	40%
Londe et al. (2018	JW	2	2	3	1	1	3	3	0	0	3	3	3	0	3	27	64%	0	0	0	0	1	1	20%
	SM	2	2	2	1	1	3	3	0	0	3	3	3	0	3	26	62%	0	0	0	0	1	1	20%
Matsumoto et al. (2007)	JW	2	3	3	1	3	2	1	x	0	2	1	2	0	2	22	56%	x	x	x	x	x	x	x
	SM	2	3	3	1	3	2	1	x	0	2	1	2	0	2	22	56%	x	x	x	x	x	x	x
Mentzel et al. (2019)	JW	3	3	3	3	2	3	2	1	0	3	3	3	0	3	32	76%	1	1	1	1	1	5	100%
	SM	3	3	3	3	2	3	2	1	0	3	3	3	0	3	32	76%	1	1	1	1	1	5	100%
Morris et al. (1994)	JW	0	2	1	0	1	1	0	0	0	2	2	0	0	0	9	21%	0	0	0	0	0	0	0%
	SM	0	2	1	0	1	1	0	0	0	2	2	0	0	0	9	21%	0	0	0	0	0	0	0%
O’Connell et al. (1985)	JW	2	1	2	0	1	1	0	0	0	1	2	0	0	0	10	24%	0	0	0	0	0	0	0%
	SM	2	1	2	0	1	1	0	0	0	1	2	0	0	0	10	24%	0	0	0	0	0	0	0%
Olde Rikkert et al. (2015)	JW	2	3	3	0	1	3	2	0	0	3	2	3	0	3	25	60%	0	1	0	0	0	1	20%
	SM	2	3	3	0	1	3	2	0	0	2	2	3	0	3	24	57%	0	1	0	0	0	1	20%
Payen et al. (2011	JW	2	3	3	1	1	3	3	0	0	3	3	3	0	2	27	64%	1	0	0	1	0	2	40%
	SM	2	3	3	1	1	3	3	0	0	3	3	3	0	2	27	64%	1	0	0	1	0	2	40%
Pellegrini and Schauss (1980)	JW	2	2	2	1	1	3	1	1	0	3	1	0	0	1	18	43%	0	0	0	0	0	0	0%
	SM	2	1	2	1	1	3	1	1	0	3	1	0	0	1	17	40%	0	0	0	0	0	0	0%
Pellegrini et al. (1980)	JW	2	1	2	1	1	3	1	1	0	2	3	0	0	1	18	43%	0	0	0	0	0	0	0%
	SM	2	1	2	1	1	3	1	1	0	2	3	0	0	1	18	43%	0	0	0	0	0	0	0%
Pellegrini et al. (1981)	JW	2	2	2	1	1	2	1	1	0	2	2	0	0	0	16	38%	0	0	0	0	0	0	0%
	SM	2	2	2	1	1	2	1	1	0	2	2	0	0	0	16	38%	0	0	0	0	0	0	0%
Piatti et al. (2012)	JW	2	2	2	1	3	3	2	x	0	3	3	2	0	2	25	64%	x	x	x	x	x	x	x
	SM	3	2	2	1	3	3	2	x	0	3	3	2	0	2	26	67%	x	x	x	x	x	x	x
Pollet and Peperkoorn (2013)	JW	2	3	3	1	1	1	0	x	0	1	3	2	0	1	18	46%	x	x	x	x	x	x	x
	SM	2	3	3	1	1	1	0	x	0	1	3	2	0	1	18	46%	x	x	x	x	x	x	x
Profusek and Rainey (1987)	JW	1	3	1	0	1	1	0	0	0	2	1	0	0	1	11	26%	0	0	0	0	1	1	20%
	SM	1	3	1	0	1	1	0	0	0	2	1	0	0	1	11	26%	0	0	0	0	1	1	20%
Recours and Briki (2015)	JW	2	3	1	1	1	2	1	0	0	2	3	0	0	1	17	40%	0	1	0	0	1	2	40%
	SM	2	3	1	1	1	2	1	0	0	2	3	0	0	2	18	43%	0	1	0	0	1	2	40%
Rezaeian et al. (2015)	JW	2	2	2	1	1	3	1	1	0	3	1	3	0	3	23	55%	0	1	0	1	1	3	60%
	SM	1	2	2	1	1	3	1	1	0	3	1	3	0	3	22	52%	0	1	0	1	1	3	60%
Rowe et al. (2005)	JW	1	2	1	1	1	1	0	x	0	1	3	0	0	1	12	31%	x	x	x	x	x	x	x
	SM	1	2	1	1	1	1	1	x	0	1	2	0	0	1	12	31%	x	x	x	x	x	x	x
Smith et al. (1986)	JW	2	3	3	1	1	2	1	1	0	2	3	0	0	1	20	48%	0	1	0	0	0	1	20%
	SM	3	3	3	1	1	2	1	1	0	2	3	0	0	1	21	50%	0	1	0	0	0	1	20%
Sorokowski and Szmajke (2007)	JW	2	3	3	0	2	2	1	0	0	2	3	0	0	3	21	50%	0	0	0	0	0	0	0%
	SM	2	3	3	0	2	2	1	0	0	2	3	0	0	3	21	50%	0	0	0	0	0	0	0%
Sorokowski et al. (2014	JW	2	3	3	1	3	3	1	0	0	3	3	1	2	3	28	67%	0	0	0	0	0	0	0%
	SM	2	3	3	1	3	3	1	1	0	3	3	1	2	3	29	69%	0	0	0	0	0	0	0%
Tiryaki (2005	JW	2	2	3	1	1	2	1	x	0	3	3	0	0	2	20	51%	x	x	x	x	x	x	x
	SM	1	2	3	1	1	2	1	x	0	3	3	0	0	2	19	49%	x	x	x	x	x	x	x
Webster et al. (2012)	JW	2	3	3	1	3	2	2	x	0	3	3	2	0	3	27	69%	x	x	x	x	x	x	x
	SM	2	3	3	1	3	2	2	x	0	3	3	3	0	3	28	72%	x	x	x	x	x	x	x
Williams et al. (2011)	JW	2	3	3	1	1	3	2	0	0	3	3	3	0	3	27	64%	0	1	0	0	0	1	20%
	SM	3	3	3	1	1	3	2	0	0	3	3	3	0	3	28	67%	0	1	0	0	0	1	20%

Note: A1–A16 = QATSDD items, B1–B5 = Elliot (2019) items.

P-curve analysis

Main p-curve

In total, 459 tests were conducted and 452 p-values were reported, of which 190 (41.39%) were significant. After the exclusion of eight incomplete tests (tests with missing df, t-value, F-value) and the application of selection rules (e.g., using the first significant p-value), the main p-curve analysis consisted of 44 statistically significant (p < .05) results. Results indicated a statistically significant right skew of p-values for both the binomial test (p < .001), and the continuous half p-curve test (Z = −10.58, p < .001), as well as the full p-curve test (Z = −10.91, p < .001). The test of flatness (flatter than 33% power) was nonsignificant for the binomial test (p = .89), the continuous half p-curve test (Z = 11.48, p > .999), and full p-curve test (Z = 5.93, p > .999). The post hoc test of statistical power indicated that the average estimated power of tests included in the p-curve (correcting for selective reporting) was 87% (CI = 77% to 93%). Figure 2 depicts the distribution of p-values in the main p-curve analysis.

Figure 2. Main p-curve distribution of 44 statistically significant results (blue solid lines), compared to the expected distribution when the null hypothesis is true (red dotted lines) or the alternative hypothesis is true and studies were powered at 33% (green striped lines).

Robustness test

After the application of selection rules (e.g., using the last significant p-value), the robustness test consisted of 32 statistically significant (p < .05) results. Results indicated a statistically significant right skew of p-values for the binomial test (p = .01), the continuous half p-curve test (Z = − 6.59, p < .001), and the full p-curve test (Z = − 7.06, p < .001). The test of flatness (flatter than 33% power) was nonsignificant for the binomial test (p = .65), the continuous half p-curve test (Z = 8.53, p > .999), and the full p-curve test (Z = 3.17, p > .999). The post hoc test of statistical power indicated that the average estimated power of tests included in the p-curve (correcting for selective reporting) was 73% (CI = 55%–86%). Figure 3 depicts the distribution of p-values in the robustness test.

Figure 3. Robustness test of 32 statistically significant results (blue solid1 lines), compared to the expected distribution when the null hypothesis is true (red dotted lines) or the alternative hypothesis is true and studies were powered at 33% (green striped lines).

Meta-analysis and moderator analysis

The results of the random effects model and the moderator analysis are presented in Tables 4 and 5. The Q-statistic was significant for the model (p < .001), indicating high heterogeneity and therefore justifying a random effects model. Overall, a significant medium-size colour effect was observed.

Table 4. Overall random effects model.

Parameter	Estimator	Standard error/variance	Test Statistic	p-value	Confidence Interval
Effect size	$\hat{\theta }$ = 0.56	SE = 0.08	t(59) = 6.90	p < .001***	[0.40, 0.72]
Heterogeneity	τ^2 = 0.34	I^2 = 96.97%	Qe (59) = 506.51	PQ < .001***

Note: ***p < .001; was calculated via Hedges g; p-values refer to the test statistics.

Table 5. Moderator analysis of Hedge’s g for the effect of colour.

Moderator	Levels	(No. of studies)	Estimate	95%-CI	p-Value
1. Year of publication	b-weight	−60	−0.02	[−0.04, −0.01]	<.001***
2. Number of sub-experiments	b-weight	−60	0.11	[−0.07, 0.29]	.242
3. Sample size	b-weight	−60	0	[−0.01, 0.01]	.678
4. Mean age	b-weight	−30	−0.09	[−0.13, −0.06]	<.001***
5. Study design	Experimental retrospective	−60	−0.37	[−0.71, −0.03]	.033*
6. Gender	All female all male mixed not specified	−60	0.46	[−0.77, 1.68]	.466
		−60	0.48	[−0.74, 1.71]	.438
		−60	0.73	[−0.64, 2.09]	.296
7. Context of sport type	Combat team individual	−60	0.08	[−0.33, 0.49]	.713
		−60	0.45	[0.08, 0.81]	.017*
8. Psychological	No vs. yes	−60	−0.11	[−0.44, 0.22]	.506
9. Physical	No vs. yes	−60	0.43	[0.10, 0.77]	.011*
10. Performance	No vs. yes	−60	−0.31	[−0.63, 0.08]	.055
11. Wearing	No vs. yes	−44	0.04	[−0.61, 0.70]	.897
12. Perceiving person	No vs. yes	−44	−0.5	[−0.92, −0.09]	.017*
13. Perceiving Environment	No vs. yes	−44	0.43	[0.01, 0.84]	.043*
14. Mixed (wearing/perceiving)	No vs. yes	−44	−0.24	[−1.34, 0.86]	.673
15. Colour red	No vs. yes	−60	−0.43	[−0.75, −0.10]	.009**
16. Colour blue	No vs. yes	−60	−0.35	[−0.68, −0.02]	.035*
17. Colour green	No vs. yes	−60	0.04	[−0.34, 0.42]	.839
18. Colour yellow	No vs. yes	−60	−0.8	[−0.49, 0.33]	.706
19. Colour orange	No vs. yes	−60	0.3	[−0.43, 1.03]	.426
20. Colour pink	No vs. yes	−60	0.26	[−0.24, 0.76]	.304
21. Colour white	No vs. yes	−60	−0.07	[−0.41, 0.26]	.667
22. Colour grey	No vs. yes	−60	0.14	[−0.27, 0.56]	.501
23. Colour black	No vs. yes	−60	−0.1	[−0.50, 0.29]	.641
24. Colour control/not specified	No vs. yes	−60	−0.16	[−0.78, 0.45]	.599

Note: Estimates are presented as Hedges g transformed standardised effect sizes, CI = credible interval, moderators except 1–4 are (1) effect coded: 6. Gender & 7. Context of sport type (reference category highlighted in italics); or (2) dummy-coded: 5. Study Design & 8–24 (reference category highlighted in italics); *p < .05, **p < .01, ***p < .001.

Sensitivity analysis

Assessing a potential publication bias, funnel plots were inspected and trim-and-fill analyses were conducted (see Appendix). The trim-and-fill analysis indicated no missing studies. The adjusted effect size was estimated ${\hat{\theta }}_{\mathrm{adj}}$ = 0.56 (p < .001). The funnel plot supported these findings, indicating a minor imbalance on the left side.

Applying the leave one out method, the impact of a single study’s effect size on the total effect size was estimated. The estimated effect sizes varied from $\hat{\theta }$ = 0.48 to 0.57 when leaving out one study, indicating no extreme impact of a single study’s effect size. The most extreme outliers were Pellegrini et al. (1980), indicating a total effect size of $\hat{\theta }$ = 0.48 and O’Connell et al. (1985), indicating a total effect size of $\hat{\theta }$ = 0.52 when leaving out the study. All other studies ranged between $\hat{\theta }$ = 0.56 and 0.57 when leaving out one study. Overall, no significant changes in effect sizes were observed applying the leave one out method.

Discussion

Quality assessment

This study aimed to systematically analyse the growing literature on the influence of colours in the context of sports. Two quality assessment tools were applied: the QATSDD for assessing the overall methodological quality and the guidelines by Elliot for assessing five colour-specific criteria. The results of the quality assessment with the QATSDD revealed a low average quality (overall average score of the evaluated criteria was 51%, with an SD of 17%). Especially the items A4 (evidence of sample size considered in terms of analysis), A8 (detailed recruitment data), A9 (statistical assessment of reliability and validity of measurement tools), A15 (evidence of user involvement in design) consistently received low ratings, with only four, one, five, and three studies, respectively, graded as moderately or completely fulfilled. Additionally, the quality assessment of the colour-specific criteria by Elliot was even lower, as the overall score revealed that on average only 34% (SD = 26%) of the examined criteria were controlled. Specifically, the item B3 (sample size acquired via power analysis) was only controlled for in two studies.

The results of the QATSDD indicate that most of the included studies did not control for general methodological and scientific standards of quantitative research. Additionally, in terms of the assessed colour criteria (e.g., control for the multidimensionality of colour, background colours, colour blindness), results imply that the overall colour effect in this study must be interpreted with caution, as it was based on less-than-high quality studies. The two quality assessment tools must also be viewed critically, since Elliot’s criteria are not an official evaluation scheme and were deduced from his recommendations to specifically increase the quality of studies in colour research. Additionally, the QATSDD has been criticised by Fenton et al. (2015), as the definitions of the evaluation criteria would be unspecific and the tool would be partly subjective in nature. Nevertheless, we do highly encourage the reuse of both quality assessment tools in future work, as the evaluation criteria were based on a strong scientific foundation, the inter-rater reliability appears to be high, and applying the tools as an integral part of colour research would improve the overall quality of scientific work in this field (Elliot & Maier, 2014).

P-curve analysis

In total, 41.39% of the tests conducted to analyse the impact of colours in sport proved significant. The results of the main p-curve indicated that the analysed studies cumulatively contained evidential value and did not appear to exhibit evidence of intense p-hacking (see Simonsohn et al., 2014a). The robustness analysis was in line with the results obtained by the initial p-curve analysis. In sum, based on the distribution of significant p-values, the published studies do provide support for the hypothesis that colours influence human behaviour in sports, as the observed distribution of p-values mirrors the expected distribution of p-values for a true effect. Although the effect can still turn out to be false, at this moment, there is quantifiable evidence to differentiate between the likelihood that this specific hypothesis is true. Nevertheless, it is important to note that several significant results had to be excluded for reporting issues and missing data (see results section for details). Thus, future research should conscientiously determine how they report characteristics of their studies in order to ensure replicability and to clarify the evidential value of the data (Simonsmeier et al., 2020).

Meta-analysis and moderator analysis

The main purpose of this study was to meta-analyse the presence of colour effects in the context of sport by identifying relevant moderating variables in order to provide an estimate of the effect size. Results indicated a medium average effect size in the predicted direction that is statistically significant. In addition, results based on a 95% confidence interval indicated that the true effect size is between 0.40 (small to medium) and 0.72 (medium to large). This effect topped several past findings, which mainly reported a general small impact of colour in different contexts (Elliot, 2015; Elliot & Maier, 2012). However, the generally low quality of studies included in this meta-analysis lowers the validity of this finding and must be considered when interpreting these results. An investigation of the funnel plot (see Appendix E) identifies the studies by Pellegrini et al. (1980) and O’Connell et al. (1985) as being potential outliers. This needs to be taken into account when interpreting the results, as both studies were conducted before 1986 and involved both physical outcome variables and an individual sport context, which may have affected the significance level of the investigated moderators. Nevertheless, the leave one out method indicated no extreme impact of a single study’s effect size on the total effect size estimate. Additionally, an inspection of the forest plot (see Appendix D) shows that the effect sizes appear to vary across studies. In fact, this is not surprising given that the meta-analysis examines multiple studies differing in terms of participant characteristics, colour stimuli, and study design. It points to the possibility of heterogeneity in the results, specifically, a reliably strong effect under some conditions and a reliably weak effect under other conditions (Lehmann et al., 2018). In line with this assumption, the Q-statistic was calculated to examine the homogeneity of effect sizes across studies. A highly significant Q-statistic was observed indicating heterogeneity of results and highlighting potential moderator effects (Belanger & Vanderploeg, 2005).

Nine statistically significant moderators were identified including year of publication, mean age, study design, individual context of sport type, physical domain, perceiving persons in coloured outfit, perceiving environment in colour, and red and blue colours. The results provide support for a multi-layered relationship, as various moderators appear to exert a significant influence on the association between colours and sports.

A significant moderation of publication year emphasised that studies published more recently tended to show smaller effects (e.g., Goldschmied & Lucena, 2018; Goldschmied & Spitznagel, 2020), suggesting that older studies may have overestimated the colour effect (Lehmann et al., 2018). A higher methodological research standard in more recent studies might have caused this effect, as the conducted quality assessments (QATSDD and Elliot’s criteria) revealed higher ratings for more recently published studies.

When interpreting the results in terms of participant characteristics, the moderator mean age indicates that studies involving younger participants tended to have larger effects whereas studies involving older participants tended to have smaller effects. This could imply that colour effects are robust within certain populations (Mentzel et al., 2017). However, more research on this phenomenon is required to draw precise conclusions and make recommendations.

The moderator study design seems to justify a distinction between different methodologies since the colour effect differed significantly between retrospective studies and experimental studies. Experimental studies were indeed accompanied by larger effects sizes. This finding might be interpreted as a consequence of higher methodological standards and a more precise control of confounding variables in experimental as opposed to retrospective study designs. This result clarifies that study design appears to be an important factor when investigating the relationship between colours and sport and should be considered critically in future studies.

When considering the context of sport type within the field of colour research, results indicated a significant moderation by individual sports not involving a combat situation (e.g., running, cycling, jumping) in comparison to team sports and combat sports. One potential explanation is that sports in an individual context can be investigated and controlled straightforwardly, as they are more compelling than, for example, team sports or combat sports (García-Rubio et al., 2011; Kocher & Sutter, 2008). This potential issue should be considered when discussing practical implementations for follow-up research, as most of the findings in the context of sport type apply almost exclusively to an unstandardised competition involving participants with different degrees of skill and strength. Subsequent research would be required to examine the association between colour and performance outcomes by controlling for an athlete’s degree of relative competitive ability (Allen & Jones, 2014; Goldschmied & Lucena, 2018; Goldschmied & Spitznagel, 2020). However, this finding is surprising from a theoretical perspective, as most studies hypothesised an association between a colour, mainly red and black, with aggression/dominance (e.g., Elliot & Aarts, 2011). Thus, the expectation arose that this association should be most easily detectable in combat sports, as aggression and dominance seem to play the most predictive role for performance outcomes. Therefore, this result also questions this assumed association derived from evolutionary biology, as the major part of studies dealt with the assumed link between the colour red and aggressiveness and/or dominance. In this regard, future studies are needed to better understand this significant effect.

Concerning the moderation effects of the investigated outcome variables, the results showed that physical domain was a significant predictor for the effect size, whereas psychological and performance domains were not significant predictors. These results indicate that physical outcome variables might be more strongly influenced by colour than psychological or performance variables. In this regard several authors argued that colours – especially red and black – might act as a cue of threat, danger and/or aggression, thus causing an immediate and intense flight or fight reaction (Elliot & Aarts, 2011; Krenn et al., 2020). This might present an explanation for this finding and emphasises the evolution-based assumption of colour effects (e.g., Hill & Barton, 2005). However, one critical flaw is that the categorisation of outcome measurements into the three domains was challenging and somewhat arbitrary, as clear-cut groupings were not always possible. An additional explanation for these mixed findings is based on the fact that too many different outcome variables were assessed per study, as each statistical analysis came with a potential type-I error (Mentzel et al., 2019). Thus, it can be concluded that there is a need to further differentiate, specify, and categorise the investigated outcome measures in order to better understand the influence that colours have upon various sport-specific parameters.

Another important differentiation within the field of colour research is whether colour effects arise from wearing coloured equipment, perceiving a coloured environment, perceiving opponents in coloured equipment, or some combination of all these factors. The findings from this study show only significant effects for the variables relating to perceiving a coloured environment and perceiving people in a coloured outfit. From both theoretical and applied perspectives, this finding seems to be important, as it suggests that colour effects in sports are mainly caused by the perception of opponents uniform colour or the colour of the environment, but not by wearing a specific colour, as was discussed in several studies (Feltman & Elliot, 2011; Krenn et al., 2017). Further research will be required to investigate whether the perception of coloured equipment grants an unfair advantage in sport contexts and raises the question of whether a regulation will be necessary to guarantee equal and fair conditions for all athletes (Feltman & Elliot, 2011).

Having considered the role of both wearing and perceiving effects, it is reasonable to look at manipulated colour stimuli. In the present research, the effects of more than 10 different colours were examined, but only the colours red and blue appeared to significantly moderate the investigated relationship. As can be seen in Table 2, the colours blue and red were by far most frequently analysed in the different studies, while colours such as brown or purple, for example, were not even part of the moderator analysis, as they were too infrequently investigated. Based on these results, it can be argued that the present results justify the extensive analyses of the colours red and blue, as they seem to have a significant impact on sporting outcomes (Dreiskaemper et al., 2013; Hagemann et al., 2008). However, considering that colours have their own meanings and associations that affect human performance and behaviour (Elliot & Maier, 2012; Krenn et al., 2020), it might be beneficial for future research to examine red and blue relative to other chromatic contrast colours, such as pink, purple, orange or brown before permanently ruling out an impact of other, less popular colours on sports. In addition, empirical work that carefully attends to the complexity inherent in colour science methodology is needed (Elliot, 2015), for example by controlling for both lightness and chroma, in addition to hue (Little & Hill, 2007; Stokes et al., 1992). For all the other moderators, no significant effects were found.

Limitations

This study attempted to systematically review and analyse growing literature on the influence of colours on human behaviour in sports. There are several methodological strengths in the current meta-analysis, including an inclusive search strategy, specific and well-defined exclusion criteria, as well as the implementation of two independent quality assessment tools. In addition, this study provides a concrete example of how p-curve and meta-analysis can complement each other to present the most complete analysis of the literature on the effects of colours in sports to date.

Although every effort was taken in ensuring uniformity within this scientific research, it is important to highlight several limitations employed within the present approach. As a meta-analysis is only as strong as the methodological quality of the studies contributing to it (Lehmann et al., 2018), the overall low quality of included studies should be considered when examining this meta-analysis. One potential weakness is the fact that some studies calculated effect sizes using parametric methods without reporting whether all assumptions were met in order to properly conduct the analysis (e.g., test of normality, homogeneity of variance). In addition, most studies did not report as many variables as they have seemed to have tested. Based on the tables and statistics provided, 25.93% of the effect sizes needed to be manually calculated by the authors of this paper. In addition, 31.37% of the reported results in the included studies did not provide sufficient information for effect size calculation. To control for these issues, a standardised average for the overall effect size calculation was used, as this has been shown to help avoid the dependence of multiple effect sizes (Cheung, 2019). According to Cheung (2019) and Maynard et al. (2014), however, this approach may lead to missed opportunities, as it does not utilise all available data and removes valuable within-study variations stemming from potential moderators. In addition, this meta-analysis solely included published studies, increasing the probability of detecting false positive findings due to publication bias and Tower of Babel bias (cf. Cumming, 2012; Ellis, 2010). However, considering this shortcoming in the current meta-analysis, the combination with a p-curve analysis might help provide a clearer picture about a potential publication bias in colour research in sports. As we did not detect suspicious data in our p-curve analysis, the reported findings of our meta-analysis are actually strengthened. Nevertheless, future scientific research should be aware of these limitations and may investigate the colour-sports relationship on the basis of multivariate and three-level meta-analysis designs.

In summation, this meta-analysis provides support for a multi-layered relationship between colours and sports. Although slight trends were identified, the investigated research question is still something of a puzzle. To achieve substantial progress in future scientific work and to get a more precise picture of the relationship between colours and human behaviour in sports, awareness of the numerous moderators influencing each other and the various factors that need to be controlled and standardised will be crucial. Subsequent work would also do well to identify and manipulate potential additional moderators, including some not considered herein such as the duration of colour stimuli presentation. Further practical implications of colour effects are still to be determined.

Conclusion

In conclusion, the present meta-analysis adds to the growing body of scientific literature that has scrutinised the role of coloured stimuli in sporting activities. Both p-curve and meta-analysis provided support for the hypothesis that colours influence human behaviour in sports. Findings did not appear to lack evidential value or to exhibit evidence of intense p-hacking. In addition, several moderators (e.g., experimental and retrospective study design and mean age; individual, team, and combat context of sport type; psychological, physical, and performance outcome measures; wearing and perceiving colour; red, blue and other colour stimuli) appear to play a role in the relationship between colour and human behaviour in sports. Most notably, colour effects were revealed in the following situations: in older people; in experimental research designs; for the colours red and blue; when colours were perceived in the environment or on an opponent; when individual sports without a combat situation were analysed; and mainly at physical outcome measures. Closer inspection of the quality of these studies, however, implied that carefully controlled empirical work is scarce, which also impedes the validity of the reported findings. As such, the information provided by the current meta-analysis must be considered tentatively and with caution. As many questions in this research area still remain unanswered, future studies are challenged to follow the guidelines offered by Elliot (2019) and to contribute to the theoretical framework to better understand how colour affects human behaviour in the context of sport.

Disclosure statement

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

Supplemental material

The data that support the findings of this study are available from the corresponding author (BK) https://osf.io/enwfh/.