(Summary of the Class Experimental Project: PSY 440, 1998)
A growing number of people in our society, as well as others around the world, use computers everyday. Whether it is for work or home use, there is a great need to find out how to make working with computers easier. One thing that is important in dealing with computers is the ease and ability to read the monitor, or video display terminal (VDT). The color of the background, the color of the text, and the contrast between the two may play a crucial role in the VDT's readability. A large body of existing research examining the effects of foreground/background colors and contrasts on readability of video display terminals has found similar as well as contradictory results. Some studies, even before examining the use of color in VDTs, examined whether or not work done at VDTs may be easier than work done through the use of pen and paper. One such study (Burns,1991) says yes.
Opposing evidence suggests that reading from a computer screen is more difficult than reading from paper ("Planning," 1998). But for the majority of people, regardless of whether it is simple or difficult, working with computers remains a part of everyday life, so the simpler it can be made, the better.
It has been found that several display aspects, including color and contrast of the characters and background, affect readability (Mills & Weldon, 1987). A study conducted by Legge, Parish, Luebker, and Wurm (1990) compared speed of reading between color contrasts and luminance contrasts, and found that when color contrast is high, subjects read as rapidly as with high luminance contrasts. Luminance and color are both factors when making comparisons on a computer VDT.
But what colors are best for readability? This is where the contradictions begin. Some recommend simply "experimenting" with background and foreground colors and choosing the ones that seem to work best ("Text," 1998), such recommendations suggest that preference is the best way to decide. Others, such as Downes (1996), claim that red type catches our attention more than blue, although it is not effective on a computer screen because people have "grown up" using blue and magenta links. Some studies recommended a colorless background for VDT viewing. Both Young and Miller's study (1991) as well as that of Zhu and Cao (1994) provided evidence that a black screen background with brighter, more luminous symbol colors provide much better visual discrimination than less luminous symbol colors ona white background. Another source (Zawitz, 1998) states in her guidelines for legibility that "dark text on light background is easiest to read." Despite their differences, several (but not all) of these sources have one thing in common: They exhibit a lack of empirical evidence. Some are even based solely on personal preference.
The current experiment hopes to investigate individuals' abilities on a readability task by taking their performance (reaction time) scores and individual preferences to find out if "liking" correlates with response time. It is thought that participants will do the best when contrast is highest. This is predicted due to the fact that text is a fairly high spatial frequency, and when working with high spatial frequencies, ability to resolve detail is lowered. It is also predicted that blue text will not be processed very well due to thelack of short-wavelength cones in the fovea. The next prediction is that participants will perform poorly on such text/background combinations as blue on yellow and red on green, and vice versa. This color prediction is based on Herring's opponent-process theory of color vision. This theory affirms that humans have different types of center-surround receptive fields which are composed of red-excitatory/green-inhibitory and green-excitatory / red-inhibitory, and the same for blue/yellow. Thus, it should be difficult to focus on these colors at the same time.
Participants included 20 individuals, with the majority being between the ages of 18 and 25. Nine of those participating were naive to the purpose of the experiment and 11 were not.
The B/C Power Laboratory experimental package presented the stimuli and recorded response time on twenty 7200/120 Power Macintosh computers. Statistica was used to analyze the data.
A within-subjects design consisting of three independent variables including background color, text color, and contrast of background color was utilized. Background color was either red, yellow, or achromatic. Text color was black, blue, or green. Light or dark versions of the three background colors were used to lead to two contrast conditions, high and low, respectively. The dependent variable was reaction time, which was used to infer readability. The preference task consisted of a rating of each of the color trials presented again in their original form.
Text paragraphs were approximately 100 words long, 4" wide, and had side margins of 1". There were continuing paragraphs one space above and below the target paragraph which ran off the screen and were not utilized for any other purpose but to suggest reading flow. Six text excerpts were presented for each of the 18 stimulus conditions. Paragraphs were presented in 12 point font, Times NewRoman style on the six background colors (light and dark yellow, grey, and red).
All participants entered the classroom and were given a brief synopsis of the experiment. All signed and returned consent forms before the experiment took place. Instructions were presented on the computer screen of the individual. The task was to read the target (middle) paragraph and search for one of the three target words(circle, square, triangle) and then click the corresponding shapeicon at the bottom of the screen as soon as the target word was found. There were 18 conditions with 6 trials per condition. This produced a total of 108 randomized trials. Placement of target words was completely counterbalanced within each condition. Four practice trials were completed before the trials began. The trials were self-paced with a pause in between each trial. There were 18 example pages at the end of the trials cataloging each color combination, and the participants were asked to rate each one on a 5-point Likert scale according to their preference.
Medians were computed for each subject for each condition. These were put into a within-participants 3-way ANOVA. A significant main effect was found for background [F(2, 38)=8.51, p<.01], with the means for red, achromatic, and yellow being 12.78, 11.05, and 11.24 seconds, respectively. Another significant main effect was found for text color [F(2, 38)=3.63, p<.05], with the means for black, green, and navy being 11.83, 12.14, and 11.09 seconds, respectively. There was no significant main effect for contrast. There were 2 significant 2-way interactions. These existed between background color and text color [F(4,76)=6.50, p<.01], and contrast of background color and text color [F(2,38)=3.54, p< .05]. Finally, there was a significant 3-way interaction between text, background, and contrast [F(4,76)=2.77, p<.05]. The means for this are plotted in Figure 1. Notice how contrast affects response times. For example, the response times for green text on light yellow background are much faster than response times for the green text on dark yellow background.
Data was further analyzed to indicate whether or not knowledge affected the outcome (students enrolled in the class knew the hypotheses, while the naive observers did not). A 4-way mixed ANOVAwith knowledge as a between-participants variable found no significant results. In this ANOVA, N=18 was used because data from the teacher and teaching assistant were not used in order to equalize the number of participants with and without knowledge. Since there was no significant effect, the 3-way ANOVA will be used to discuss the effects of text color, background, and contrast on the readability of VDTs.
Finally, a correlation was performed to test the relationship between preference and performance. No significant correlation was found between the two. See Figure 2 for the response time means in relation to the preference means. Notice that often subjects performed slowly on the conditions they preferred and more quickly on the conditions they did not prefer. For example, subjects generally liked red text on a green background, though they performed poorly. On the other hand, they disliked black text on a red background but performed rather well. It can also be seen that, sometimes, subjects do prefer the conditions that they performed well on, such as the case with green on yellow. Overall, this lack of significant correlation suggests that the colors a subject prefers does not necessarily determine the conditions in which they will perform the best.
Overall, these findings provide partial support for the original hypothesis. Participants performed the most quickly when contrast was highest (i.e. lighter background). This is seen in the response times of light yellow background with green text as opposed to slower response times to a dark yellow background with green text. Secondly, participants did respond more slowly to such opposing colors as red background with green text. The prediction of slow processing when using blue text was not supported because subjects performed fairly well, in fact, second fastest, when processing blue text, especially on a white background. This is quite possibly due to the high contrast (when the background was white) as well as color cues. Had the blue text not been so dark and the background not been so light,r esponse times may have been slower.
Knowledge of the experiment showed no significant effect on performance. This is not surprising due to the fact that reading speed is based on sensory experience, something that is virtually the same for everyone with normal vision, regardless of prior knowledge.
Another interesting fact is the impressive performance of subjects on the gray background/black text condition (10.48 seconds). This is important for Internet designers because this is the default currently used by many Internet browsers.
The lack of relationship between performance and preference highlights an important point. Subjects did not necessarily perform more quickly on the conditions that they rated high in liking. This is contradictory to sources that suggest otherwise, such as the 'Text on a Background Pattern" web article.
The evidence uncovered in the current experiment concerning contrast supports Zawitz's argument that dark text on a lighter background is easiest to read.
Future research might be aimed toward further investigatingtext/background combinations not examined in this experiment. Perhaps when other colors are researched for readability, Internet pages will not only be aesthetically pleasant to look at, but will benefit from their ease of readability as well.
Burns, D.H. (1991). Characteristics of visual display units that may cause visual difficulties. Ophthalmic and Physiological Optics,15, 99-104.
Downes,S. (March,1996). International University Consortium. Effective interaction and communication with web courses. [On-line].
Legge, G., Parish, D., Luebker, A., & Wurm, L. (1990). Psychophysics of reading: XI. Comparing color contrast and luminance contrast. Journal of the Optical Society of America A, 7, 2002-2010.
Mills, C. & Weldon, L. (1987). Reading text from computer screens. ACM Computing Surveys, 19, 329-358.
Planning, design, and production- G. readability. (March, 1998). [On-line].
Sweet home on-line- Sweet home Oregon, USA. (March, 1998). Color legibility. [On-line].
Text on a background pattern. (March, 1998). [On-line].
Young, H. & Miller, J. (1991). Visual discrimination on colour VDTs at two viewing distances. Behaviour and Information Technology, 10 (3), 191-205.
Zawitz,M. (March, 1998) How do you improve legibility? [On-line].
Zhu, Z. & Cao, L. (1994). Effects of target-background color matching of visual performance on color cathode-ray tubes (CRTs). Acta Psychologica Sinica, 26, 128-135.