The Red Effect,Anxiety,and Exam Performance

Literature Review

The color perception literature is filled with numerous correlational findings between specific colors and affect, cognition, and behavior. For example, Rutchick, Slepian, and Ferris (2010) found that the use of a red pen significantly increased the number of corrections made when grading a standardized paragraph; Elliot et al. (2010) found that women rated men as more sexually desirable and attractive when they were dressed in red clothing or in a red background. Such studies indicate that color perception can have a significant effect on internal states and behaviors.

Nevertheless, the association between color presence and outcomes is complex and context specific. Research suggests that an eclectic array of associations can be made for any given color based on context environments in which colors are presented. Research examining color perception specific to context found that when primed by the color red, participants knocked fewer times on a laboratory door and moved their body further away from an IQ test, compared with participants primed with green (Elliot et al., 2009). Similarly, Maier, Barchfeld, Elliot, and Pekrun (2009) found that infants were less likely to choose red toys. These results suggest that red may prime avoidance orientations in some settings. Alternatively, in romantic contexts, red seems to increase attraction (Elliot et al., 2010) and, in competitive contexts (e.g., playing sports), red seems to increase perceived dominance and threat (Feltman & Elliot, 2011). Thus, the effects of red seem to be context specific, and researchers should specify the context of interest. The present study, like several before, focuses on the effects of the presence of the color red on cognitive task performance in achievement settings.

Several past studies have established significant negative relationships between the color red and assessment performance (Elliot et al., 2007, 2009; Elliot & Maier, 2007; Lichtenfeld et al., 2009). Mediating processes seem to reflect increased anxiety and/or the inducement of an avoidance orientation. Participants in the presence of the color red placed their chairs farther away from the color stimulus—suggesting that an avoidance orientation was activated (Elliot et al., 2009). Research suggests that avoidance is associated with the affective experience of anxiety (Kwallek, Lewis, & Robbins, 1988; Spielberg, Heller, Silton, Stewart, & Miller, 2011). Emotions of anxiety are thought to result from the perception of a threat for which blame cannot be assigned to an obvious source, and these emotions produce action tendencies toward avoidance (Lazarus, 1991). Thus, it seems that perceiving the color red in an achievement setting may trigger anxiety, which in turn, may adversely affect exam performance.

The majority of studies investigating the red effect were conducted in laboratory settings. In the lab, researchers have found statistically significant relationships between color name words (e.g., red, gray, and green), and cognitive task performance on IQ tests (Lichtenfeld et al., 2009). Others have used red, green, and white exam cover sheets as manipulations in laboratory-based IQ testing (Elliot et al., 2007). These studies have often found that participants in the red condition performed significantly worse than those in the contrast conditions. However, characteristics of the setting in which performance occurs may moderate these results, as discussed below.

Several aspects of the laboratory setting are worthy of note. First, these studies used measures of performance that participants were unable to prepare for, thus potentially increasing their performance anxiety relative to natural settings. Second, the results of the experimental task had few if any implications for the participants; whereas, in nonlaboratory settings, important outcomes may be associated with successful performance. Because of this, laboratory participants may have had little incentive to perform well on tasks presented to them. Third, laboratory settings may be less visually stimulating, exacerbating the effects of color perception due to lack of natural setting stimuli (e.g., colors of clothing, walls, floors, etc.). Finally, as discussed previously, color effects are context specific (Elliot et al., 2009). Thus, changing the context in which red is presented from laboratory to field may affect the psychological processes primed and, therefore, may affect performance.

Significantly fewer studies have examined the red effect in a field setting, and among those, results have been mixed (Sinclair, Soldat, & Mark, 1998; Tal, Akers, & Hodge, 2008). Sinclair, Soldat, and Mark (1998) printed exams, which were part of the regular curriculum, on red or blue paper and found that students performed better in the blue condition, especially on difficult questions. However, a similar study printed exams on red, green, blue, yellow, and white paper and found that students performed best in the white condition and worst in the blue condition, with red, green, and yellow falling in between (Tal et al., 2008).

The purpose of this study is to add to the existing literature on color perception and exam performance in achievement contexts by examining whether the red effect as well as the potential mediating variable of state anxiety, produce significant decreases in cognitive performance in a field setting. Our field setting includes students taking an in-class exam with real implications for their course grade. Thus, unlike many laboratory settings, this task provided students with motivation to perform, and many external variables were uncontrolled (e.g., amount of study time). That is, this study allows the researchers to examine whether the color red produces a sufficiently strong effect on task performance to be practically meaningful in combination with other external factors. In line with the results of past laboratory studies, we hypothesize that students in the red condition will perform significantly worse on an in-class exam than those in all other conditions (Hypothesis 1) and that anxiety will mediate this effect, with those in the red condition reporting higher anxiety (Hypothesis 2).

Study 1

Method

Results and Brief Discussion

Descriptive statistics for the red and green color conditions are provided in Table 1. The α internal consistency reliability was acceptable for the anxiety scale (α = .92).

OPEN IN VIEWER

Table 1.

Study 1—Mean Performance and Anxiety Scores by Group.

Color Condition	n	M exam (SD exam)	M anxiety (SD anxiety)
Green	69	37.22 (5.61)	2.45 (.61)
Red	68	36.34 (5.63)	2.37 (.79)

Note. Exams were worth 50 points.

Hypothesis 1 stated that participants in the red condition would have significantly lower performance than participants in the green condition. This hypothesis was tested using an independent samples t-test to compare the means of the two groups. Results suggested there was no significant main effect of color condition, t(135) = −.92, p = .36, with participants in the green condition scoring nonsignificantly better than those in the red condition. Thus, Hypothesis 1 was not supported.

Hypothesis 2 suggested that participants in the red condition would have significantly greater anxiety than those in the green condition, and that anxiety would mediate the effect of the color red on performance. Using an independent samples t-test, we found no significant relationship between color condition and anxiety, t(130) = −.62, p = .16. Due to the lack of significant results between the predictor and mediator, we did not proceed with mediation tests. Thus, Hypothesis 2 was not supported.

Our findings revealed no significant main effect of color condition with the green condition scoring nonsignificantly better than the red condition. Further analysis showed no significant relationship between color condition and anxiety. The results of our study, therefore, indicate that the color red may not produce significant effects on exam performance or anxiety in field settings.

Although we used cover sheet manipulations similar to those in past laboratory studies that found a red effect, we found no significant results. This suggests that cognitive task performance may not be as susceptible to the biasing effects of color presence as previously thought. Other factors, such as ability and motivation, may dominate the variance in task performance outside of the laboratory.

Nevertheless, Study 1 had several limitations. These limitations were addressed in a second study. First, in Study 1, state anxiety was assessed retrospectively, after the exam had concluded. Because retrospective reports of experienced anxiety may be biased, we measured state anxiety prospectively in Study 2. Furthermore, we included additional contrasting colors, specifically blue and white. Finally, we increased the strength of the manipulation by placing the assigned color on every page of the students’ exam, rather than on only the cover page.

The addition of blue as a contrast color allows us to replicate the contrasts performed by Sinclair et al. (1998) who found significantly higher performance for those in the blue condition relative to red. Furthermore, they found that the effect was moderated by difficulty, such that individuals in the blue condition received more performance advantage on difficult as opposed to simple questions. Thus, we examine difficulty at the item level and also perceived difficulty at the exam level as potential moderators.

Finally, Sinclair et al. (1998) proposed that the performance advantage from the color blue on difficult items may result from affective changes. Relative to red, blue is thought to invoke more negative affect (Sinclair et al., 1998). Further, negative affect induces a more detailed processing strategy (e.g., Forgas, 1995), which may provide performance advantages on difficult exam items. Therefore, we measured negative affect as a potential mediating variable for color effects.

Study 2

Method

Results and Brief Discussion

Descriptive statistics are provided by color condition and course in Table 2. None of the variables differed significantly by color condition in class 1 (Fs = .04–.66, η² = .00–.04, ns), class 2 (Fs = .24–1.49, η² = .02–.11, ns), or class 3 (Fs = .19–3.28, η²= .07–.55, ns). However, we note that the sample sizes were small, particularly for Class 3, so power to detect significant effects was low. The effect sizes for color were rather large in Class 3 for exam score, anxiety, and negative affect. For exam score, individuals in the red condition had the highest scores and very little variance in scores—the standard deviation was only 1.18. In contrast, individuals in the white condition had the lowest scores and more variation in scores. Those in the blue and green conditions seemed to fare similarly and fell in between the red and white conditions. However, it appears that individuals in the blue condition in Class 3 may have experienced greater levels of anxiety and negative affect during the exam than their classmates in the other color conditions. Although this result did not reach statistical significance, it is consistent with the hypothesis that blue may convey more negative affect than red (e.g., Soldat, Sinclair, & Mark, 1997).

OPEN IN VIEWER

Table 2.

Descriptive Statistics by Class and Color Condition.

		Class 1 3,000-Level Multiple-Choice Exam, Psychology, Mean Difficulty = 2.41				Class 2 3,000-Level Multiple-Choice Exam, Management, Mean Difficulty = 2.92				Class 3 1,000-Level Short-Answer Exam, Psychology, Mean Difficulty = 3.20
Variable	N	Red	Blue	Green	White	Red	Blue	Green	White	Red	Blue	Green	White
		18	13	14	12	14	11	10	6	4	3	3	3
Exam score	M	80.94	80.69	81.14	79.67	83.21	76.46	81.50	79.83	86.38	71.17	71.67	66.83
	SD	13.28	9.27	12.40	14.91	15.65	13.16	14.18	12.17	1.18	11.03	10.73	14.43
Expected score	M	1.61	1.64	1.50	1.75	1.86	1.90	1.60	1.83	2.25	2.67	2.50	2.33
	SD	0.85	0.63	0.65	0.87	1.10	0.88	0.52	0.75	0.50	0.58	0.71	1.16
Anxiety	M	2.40	1.93	2.26	2.16	3.01	2.88	2.69	3.31	2.81	4.13	3.56	2.54
	SD	1.05	0.80	1.24	0.95	0.90	1.05	1.07	0.94	0.99	0.43	0.44	0.40
Negative affect	M	1.51	1.42	1.74	1.41	1.98	2.19	1.61	2.29	1.50	2.64	2.25	1.94
	SD	0.73	0.34	0.84	0.70	0.89	0.69	0.55	0.62	0.59	0.39	1.34	0.36

Across all three courses and across all color conditions, correlation analysis revealed that expected grade and actual grade achieved were significantly correlated (r = −.38, p < .05). ¹ Furthermore, anxiety and negative affect were significantly correlated (r = .77, p < .05). Expected grade was significantly correlated with anxiety (r = .31, p < .05) and with negative affect (r = .24, p < .05). Actual grade was inversely associated with negative affect (r = −.19, p < .05) but not with anxiety (r= −.11, ns).

We next examined the hypothesis that color condition would be significantly associated with anxiety, negative affect, and exam scores. This hypothesis was tested using analysis of covariance (ANCOVA), controlling for perceived exam difficulty and expected grade. We also tested whether color condition might moderate the effects of anxiety on exam score. ²

Consistent with our findings in Study 1, no statistically significant color effects were found. Color condition was not significantly associated with exam score, anxiety, or negative affect. Effect sizes indicated that color condition accounted for 3% of the variance in anxiety, 2% of the variance in exam score, and 0% of the variance in negative affect (see Table 3). Although these effect sizes are small, the effects for anxiety and exam performance might be practically meaningful when considering the experiences of a large number of students across a large number of exams.

OPEN IN VIEWER

Table 3.

Analysis of Covariance (ANCOVA) Results for Color Effects on Performance, Anxiety, and Negative Affect.

	Color			Perceived Difficulty (Control)			Expected Grade (Control)
	F	p	Partial η2	F	p	Partial η2	F	p	Partial η2
Performance	.65	.58	.02	5.61	.02	.05	13.44	<.01	.11
Anxiety	.88	.46	.02	40.98	<.01	.28	10.67	<.01	.09
Negative affect	.06	.98	.00	40.65	<.01	.28	5.39	.02	.05

Furthermore, because Sinclair et al. (1998) found that difficulty may play a role in the red effect, we also performed the analysis again, selecting only students who self-reported perceptions that the exam was either “difficult” or “very difficult.” In this case, we removed perceived exam difficulty as a control variable, given that it was used to select cases. The results were similar. Color condition was nonsignificant (F = .26, p = .85), and color accounted for 4% of the variance in exam scores. Among participants who rated the exam as less difficult, color accounted for 2% of the variance in performance.

We also tested for effects of actual item difficulty in the 3,000-level psychology course (the only course for which we had access to the necessary data to do so). Items were considered easy if at least 75% of students got them correct and were otherwise considered difficult. This classification produced 24 easy and 12 difficult items. ANCOVA was performed, controlling for expected grade. Color condition was nonsignificant for percentage correct for both easy items (F = .53, p = .67; η² = .03) and difficult items (F = .56, p = .64, η² = .03).

Finally, we examined whether anxiety might moderate the effects of color condition on exam scores. In this case, color might have greater effects on exam performance for students who are more anxious than for those who are less anxious. For this analysis, a linear regression was performed in which the control variables (expected score and perceived difficulty) were entered in Block 1. Color conditions were dummy coded with red as the contrast category, such that significant β weights for any of the three dummy-coded color variables would suggest a significant contrast with the red condition. The dummy-coded color variables and the centered main effect of anxiety were entered in Block 2, then the three interaction terms (the interactions of each color with anxiety) were entered in Block 3. The results of this analysis can be found in Table 4. As shown, the interactions between color condition and anxiety had no significant effects on exam scores. Thus, the effect of color did not differ significantly based on students’ level of exam anxiety.

OPEN IN VIEWER

Table 4.

Moderation of Anxiety on the Effect of Color on Exam Performance.

		B	SE	β	t	p
Block 1, R 2 = .28	Intercept	100.97	4.54		22.23	<.01
	Perceived difficulty	−2.68	1.38	−.21	−1.94	.06
	Expected grade	−8.64	2.10	−.45	−4.10	<.01
Block 2, R 2 = .30	Blue	−0.08	1.84	−.01	−0.04	.97
	Green	−0.87	1.82	−.06	−0.48	.64
	White	0.27	2.19	.02	0.12	.90
	Anxiety	3.29	2.67	.15	1.23	.22
Block 3, R 2 = .31	Anxiety × Blue	−2.01	3.34	−.14	−0.60	.55
	Anxiety × Green	−0.50	3.15	−.04	−0.16	.87
	Anxiety × White	2.68	4.14	.17	0.65	.52

Note. Red is the contrast category.

In Study 2, we again found no significant effects by color condition. Participants assigned to different color conditions did not differ significantly on exam performance, expected grade, anxiety, or negative affect. Furthermore, no significant moderation effects were found for either perceived exam difficulty or anxiety level.

General Discussion

The two studies reported here sought to examine the extent to which the red effect might occur in classroom exam settings and further to explore the potential mechanisms by which color effects could occur. The results for color condition were nonsignificant in all cases, with color condition not significantly related to performance, expected performance, anxiety, or negative affect. Nevertheless, the effect sizes we found were consistent with those reported in previous field research in actual testing situations. We found that color accounted for 2% of the variance in exam scores, which matches the effect size of η² = .02 found by Sinclair et al. (1998) and is also similar to that of η² = .04 found by Tal et al. (2008). Thus, it seems that the effects of color on exam performance may be quite small, accounting for only 2–4% of the variance in exam performance.

Nevertheless, small effects may have large practical impacts. Agars (2004) describes how a small gender discrimination effect accounting for less than 1% of the variance in an individual hiring or promotion decision compounds across time and across individuals. While a small effect may not have large impact on any one outcome for any individual person, their impacts may be cumulative. In Agars’ example, after only four decision rounds with a discrimination effect size of less than 1%, the percentage of women selected had decreased from 50% to less than 42%. Thus, it seems that although a color effect size of η² = .02 may not have large implications for any individual student’s performance on a single exam, it could have implications for the performance of groups of students over time.

Unlike gender, the color exam to which students are assigned could easily be varied over time. Instructors who wish to use color-coded exam forms should randomly assign exam cover colors so that no students are systematically affected by repeated assignment to the same color. Nevertheless, color effects might have broader implications. For example, assigning students to teams where each team is represented by a different color (e.g., red team, blue team), and where each team works together throughout the semester, could have a significant impact on students’ grades. Furthermore, if color effects persist over time, then school colors might be associated with organization-level disparities in performance. School colors are often present in classroom decor and are frequently worn by students and teachers. Might schools with blue colors perform differently at the organizational level than those with red colors? If the small effects of color are cumulative across performance opportunities, then an effect size of η² = .04 could produce significant school level over time. Future research might further examine the cumulative effects of color in academic settings.

In addition, we found little evidence to suggest that changes in either anxiety or negative affect underlie any color effects that do occur. A trend did emerge, however, that might provide a potential avenue for future research. We noted that the color effects may have been larger (although still nonsignificant) in Class 3 than in the others. This particular exam differed from the others in that it was (a) a 1,000-level first-year course; (b) perceived as more difficult; and (c) an essay, as opposed to multiple choice, exam. However, our sample size from Class 3 was quite small. Any of these differences could be explored in future research as potential moderating factors. Future research might also explore potential differences of the color effect in mental processing of multiple-choice versus short-answer exams.

Although past research has found significant color effects in the laboratory, we were unable to replicate those significant effects in our studies. The generalizability of laboratory contexts to field contexts is not yet well understood in the color perception literature. The laboratory offers researchers a high degree of control of confounding variables; researchers are able to standardize protocol in a laboratory setting. It is also generally an unfamiliar context for students—especially when compared to a classroom in which students have attended many class sessions. When considering the generalizability of a finding, the more similar the experimental conditions are to the field setting, the more one can be sure about external validity. Identifying which factors are responsible for potential differences between lab-based and field-based results on color perception is an important topic for future consideration.

Conclusion

Extraneous influences (e.g., color effects) should be further explored to fully understand their effects on performance in order to ensure accurate measurement of achievement. Only with a clear understanding of potential nonability factors contributing to negative performance outcomes can we reduce biases affecting exam performance, thus creating a fair and equal environment in which student ability and preparedness are the only factors affecting performance outcomes.