HUMAN FACTORS, 1982, 24(5), 501-508

Extended Reading of Continuous Text on Television Screens

PAUL MUTER, SUSANE A LATREMOUILLE, and WILLIAM C. TREURNIET, Communications Research Center, Ottawa, Canada, and PAUL BEAM, University of Waterloo, Waterloo, Canada

Address correspondence to Paul Muter, Department of Psychology, University of Toronto, Toronto, Ont., M5S 3G3, Canada, muter@psych.utoronto.ca.

(c) 1982, The Human Factors Society. All rights reserved.

Thirty-two subjects read continuous text for 2 hours. Half read from a television screen (video condition) and half read from a book. Subjects experienced little nausea or headache in either condition. A small amount of dizziness, fatigue, and eyestrain was produced by reading, but there were no significant differences between the book condition and the video condition on these measures, nor was there a difference in comprehension scores. Video subjects read 28.5% more slowly than book subjects. Possible reasons for this difference are discussed. In the video condition, the presence of proportional horizontal spacing had no effect on reading speed or comprehension. The results suggest that extended reading of continuous text on television screens is certainly feasible.

INTRODUCTION

Within the next few decades, people may be able to access much of the published information in the world from their living rooms by using videotex. (Videotex, sometimes called "interactive television" or "two-way television," is a technology permitting inexpensive, public, remote access to information by means of modified television sets, computers, and telecommunications links.) In the distant future, much of this information may be presented in the form of a pictographic language, but, in the relatively near future at least, a large amount of information will be presented on videotex in the form of phonetic text, such as written English.

For example, it may soon be possible to display any major novel on home television sets. An obvious question that arises is whether it is feasible to read text on television screens for extended periods of time, in the way that people often read books for such periods. There have been many studies of various issues regarding legibility and readability of text displayed on CRTs (e.g., Scanlan and Carel, 1976; Shurtleff, 1967; Snyder and Maddox, 1978; Treurniet, 1980), but there is apparently no published evidence indicating whether it is feasible to read continuous text on videotex, or on CRTS in general, for two hours or longer. Would there be intolerable fatigue, eyestrain, headaches, or other problems? How would reading speed and comprehension compare with reading speed and comprehension for material in book form?

The present study provides answers to these questions. A second purpose of this study is to assess the effects of proportional horizontal spacing of characters on reading speed and comprehension. On computer displays, text is typically presented nonproportionally: That is, the complete character matrix containing descriptions of the desired letters is displayed. In some cases, such as the letters i and l, not all columns of the matrix are used. These empty columns appear as extra space beside the displayed letters. Removal of the empty columns prior to displaying the letter is referred to as proportional spacing. Proportional spacing helps to maintain the unitary appearance of words. In addition, it destroys the vertical alignment of letters on the page, which often results in the appearance of potentially distracting vertically oriented rivers of background color.

In the present experiment, one group of subjects (book condition) read short stories from a book for 2 hours, and a second group of subjects (video condition) read the same material displayed on a videotex terminal for 2 hours. The video group read text with proportional spacing for 1 hour and text with nonproportional spacing for the same period. All subjects underwent tests of reading speed and comprehension, and gave subjective estimates of dizziness, headache, nausea, fatigue, eyestrain, and desire to read more.

METHOD

Subjects
Thirty-two people who had responded to advertisements posted on the campus of the University of Ottawa served as subjects. They ranged in age from 18 to 40, with a median age of 22. The advertisements specified that sujects were required to be able to read English; however, English was not the first language of some subjects. All subjects had at least 20/20 vision, corrected or uncorrected.

Materials and Apparatus

The reading material was 47 short stories of H.H. Munro (also know as "Saki") taken from The Complete Works of Saki (Munro, 1976). This material was split arbitrarily into two sets approximately equal in size. For the most part, the stories were read in order in the book condition, but in three places subjects encountered messages in the book, such as "Go to page 220." The order of the stories was the same in the video condition as in the book condition. More generally, the material was as similar as possible in all respects in the two conditions. In the book condition, there was a maximum of 40 rows of text per page, with a maximum of approximately 60 characters per row. The number of words per page was approximately 400. In the video condition, there was a maximum of 18 rows of text, with a maximum of 39 characters per row. The number of words per page was approximately 120.

In the video condition, the text was displayed on a 48-cm color television receiver, modified for RGB (Red-Green-Blue) video input. The television was controlled by a Norpak Telidon videotex terminal. (Telidon is a form of videotex developed by the Department of Communications of Canada. See Bown, O'Brien, Sawchuck, Storey, and Treurniet, 1980.) The terminal was equipped with the Bronsard 3 character set described in Treurniet (1981). This character set is designed on a 5 X 8 matrix, with the bottom row of the matrix being used only for descender characters. The displayed matrix height measured 1 cm. Examples of the two kinds of text, proportional and nonproportional, are presented in Figures 1 and 2, respectively. Text was displayed in white on a blue background. The luminance of the white source was 181.6 cd/M2, and the luminance of the blue source was 15.4 cd/M2. (Luminance was determined by measuring from a screen filled with each color.)

The display was controlled remotely by a minicomputer communicating via a full duplex, 1200-baud modem. There was no scrolling of text. Subjects initiated requests for new pages by pressing the lower-right key of the hand-held keypad. The time between requesting a new page and initiating display of that page was approximately 0.5 s. The time to fill the screen was 9 s, substantially slower than for most Telidon videotex terminals. In other respects, the system provided a typical videotex display.

Subjects in the video condition sat on a couch approximately 2.5 m from the screen. The approximate distance from a typical subject's eyes to the floor was 94 cm, and the center of the screen was 63.5 cm from the floor, so the approximate viewing angle was 7 deg below the horizontal. Subjects in the book condition sat at the other end of the same couch and were permitted to hold the book in whatever manner they preferred. The room was well illuminated by an overhead light source.

Design
Condition (book or video) was a between-subjects variable; 16 subjects read in each condition. Spacing (proportional or nonproportional) was a within-subject variable for video subjects only. Eight of the video subjects read text with proportional spacing for the first hour (Set 1) and text with nonproportional spacing for the second hour (Set 2), and eight did the reverse. For subjects in the book condition, of course, the text was proportionally spaced in both sets. Set 1 material was always presented in the first hour, and Set 2 material was always presented in the second hour.

Procedure
Twenty-eight of the 32 subjects were run in 14 pairs, one subject in each condition. The remaining 4 subjects were run individually.

Subjects were given the following instructions:

In this experiment, you will be asked to read some short stories by H. H. Munro (also known as "Saki"). You will read for one hour, then there will be a short break, then you will read for a second hour. Your comprehension and reading speed will be measured, but please try to read the stories as you would if you were reading them for pleasure. You may pause for a few seconds once in a while.

In this experiment, you are not permitted to turn back to a previous page at any time.


The subjects in pairs were also instructed not to talk during the experiment. Subjects in the video condition were told to press only the lower-right button on the hand-held keypad. After 60 min of reading, subjects were asked to stop by the experimenter, who then noted the extent of their progress. Subjects were then asked to answer 25 multiple-choice questions on the material in Set 1. Twenty of these questions concerned the first eight stories of the set. Each question included the title of the relevant story. Subjects were asked to answer all questions, even those on stories that they had not read. The mean duration of the break between Set I reading and Set 2 reading was approximately 10 min.

Subjects then read Set 2 for 60 min, and were given a comprehension test as in Set 1. Subjects then rated, on a seven-point scale (from "none" to "extreme") how much dizziness, headache, nausea, fatigue, and eyestrain they felt at present, and how much they had felt before the experiment began. Finally, they were asked to rate, on the same seven-point scale, their desire to read more. Subjects in the video condition were also asked to indicate, on a seven-point scale, whether they would prefer to read a novel on videotex or in book form, and why.

RESULTS

Subjective Measures of Discomfort
There were no significant differences in subjective measures of discomfort between the video condition and the book condition. Table 1 shows that the mean discomfort produced by reading on video tended to be slightly higher than that produced by reading a book, but none of these differences approached statistical significance. (The t-tests in Table 1 reflect the difference between the two conditions in the amount of discomfort produced by reading. The amount of discomfort produced is determined by the difference between the "after" and "before" scores). Table 1 suggests that negligible headache and nausea occurred in either condition. Reading for two hours did produce subjective reports of small amounts of dizziness, fatigue, and eyestrain, but these problems were approximately equal in the two conditions.

Subjects in the video condition expressed slightly more desire to continue reading (mean = 2.75) than subjects in the book condition (mean = 2.25), but this difference did not approach statistical significance t(30) = 0.90. (A response of 2 indicated "a trace" of desire to read more; a response of 3 indicated "mild" desire to read more.)

Subjects in the video condition were asked whether they preferred reading a novel in video or book form. The mean response was 5.25 (5 = mild preference for a book, 6 = moderate preference for a book), but this result is misleading because some subjects incorrectly assumed that one cannot normally turn back to previous pages when reading videotex. The most common reason subjects gave for preferring the book form was that it allowed them to return to previous pages.

Reading Speed
There was no effect of spacing, proportional or nonproportional, on reading speed in the present experiment. A Latin square analysis of variance (Winer, 1962, pp. 541-542) was performed on the reading speed of subjects in the video condition only. This analysis revealed that subjects read faster in Set 2 than in Set 1, F(1,14) = 7.31, p < 0.05, but there was no effect of spacing, F(l, 14) < 1, or order, F(l, 14) < 1. The relevant means and standard deviations are presented in Table 2.

In the subsequent overall analysis of variance, the variable of spacing was ignored. This overall analysis revealed a reliable effect of condition (video or book), F(1,30) = 10.3, p < 0.005. Subjects in the book condition read at an overall mean rate of 222.3 words/min; subjects in the video condition read at an overall mean rate of 158.9 words/min, 28.5% more slowly. Overall, subjects read faster in Set 2 than in Set 1, F(1,30) = 16.0, p < 0.001. There was no reliable interaction between set and condition, F(1,30) < 1. The means and standard deviations are given in Table 3.

Comprehension
In order to assess performance on material that had been read by all subjects, only the first eight comprehension questions for each set were used in the analysis. (The slowest reader was slower than expected.) Because only eight questions were used, the comprehension tests may have been relatively insensitive.

Again, the variable of spacing was ignored for the overall analysis of variance, since spacing clearly had no effect on comprehension scores in the present experiment.

There was no effect of condition (video or book) on comprehension scores, F(1,30) < 1. Nor was there a significant effect of set, F(1,30) = 2.98, or of the interaction between condition and set, F(1,30) < 1. The relevant means are given in Table 4.

DISCUSSION

The present experiment suggests that it is certainly feasible to read continuous text on videotex for at least 2 hours. There was no significant difference between the video condition and the book condition in reported desire to read more, or in reported buildup of dizziness, headache, nausea, fatigue, or eyestrain. Similarly, there was no significant difference in comprehension scores.

There was a statistically significant difference in reading speed. Subjects in the video condition read 28.5% more slowly than subjects in the book condition. There are several possible reasons for this difference.

First, the difference may be related to familiarity with the medium. The subjects presumably had had extensive experience in reading continuous text from books, and little or no experience in reading continuous text from television screens. The lack of familiarity may have hindered reading performance; alternatively, the novelty may have produced a mild fascination that caused them to read slowly.

Second, the effect of medium on reading speed may have been due to the difference between the mean number of words on a book page (approximately 400) and the mean number of words on a video page (approximately 120). Perhaps subjects have a tendency to consume pages at fixed rate, as opposed to a tendency only to consume words at a fixed rate.

Third, a related possibility is that the number of characters per line was a factor. Kolers, Duchnicky, and Ferguson (1981) found that reading speed on a CRT was 17% faster with 80 characters per line than with 40 larger characters per line. Starting a new line may waste time. In the present experiment, the maximum number of characters per line was 39 in the video condition and approximately 60 in the book condition.

Fourth, subjects were free to adjust their posture at will in both conditions, but only in the book condition were subjects able to adjust the distance from their eyes to the reading material, as well as the height and the angle of the reading material. Ideally, these variables should be under the user's control on videotex also, and the absence of such control may have retarded reading speed.

Fifth, the 9 seconds that were required to fill the screen under the present procedure may have distracted subjects in the video condition.

Subjects read faster in Set 2 than in Set 1. For the comprehension test on Set 1, subjects may have noted that the questions covered far more stories than most subjects had read, and this may have induced subjects to attempt to increase their reading speed for Set 2.

There was no discernible effect of proportional spacing, in the present experiment. Of course, it does not follow that proportional spacing is worthless. The extra expense of providing proportional spacing may be justifiable on aesthetic grounds alone. Furthermore, though the number of words per line was equal in the proportional and nonproportional conditions in the present experiment, proportional spacing permits more compact display of text: slightly more words per line and hence more words per page.

CONCLUSION

In the present experiment, people read continuous text on a television screen for 2 hours without undue discomfort, though they read more slowly than people reading the same material in book form. In the video condition, presence of proportional horizontal spacing had no effect on reading speed or comprehension.

Further research is needed to determine why reading was slower in the video condition and how this problem can be remedied. However, the present results suggest that videotex system designers and database creators should bear in mind that reading extensive passages of prose, for example novels, on a television screen is indeed feasible.

ACKNOWLEDGEMENTS

This work was partially supported by Research Grant UO149 from the Natural Sciences and Engineering Research Council of Canada to the first author. We are grateful for the help of Candace Mayson, Alan Orr, and William Henneker. We also thank Robert Duchnicky, Paul Hearty, Eric Lee, Dorothy Phillips, and Thomas Whalen for useful comments.

REFERENCES

Bown, H. G., O'Brien, C. D., Sawchuck, W., Storey, J. R., and Treurniet, W. C. Telidon Videotex and user-related issues. In P. A. Kolers, M. E. Wrolstad, and H. E. Bouma (Eds.) Processing of visible language 2. New York: Plenum, 1980.

Kolers, P. A., Duchnicky, R. L., and Ferguson, D. C. Eye movement measurement of readability of CRT displays. Human Factors, 1981, 23, 517-527.

Munro, H. H. The complete works of Saki. Garden City, NJ: Doubleday, 1976.

Scanlan, C. A., and Carel, W. L. Human performance evaluation of matrix displays: Literature and technology review. Wright-Patterson AFB, OH: Aerospace Medical Research Laboratory, Report No. AMRL-TR76-39, 1976.

Shurtleff, D. A. Studies in television legibility. Information Display, Jan/Feb 1967, 40-45.

Snyder, H. L., and Maddox, M. E. Information transfer from computer-generated dot-matrix displays. Research Triangle Park, NC: U.S. Army Research Office, Report No. HFL-78-3/ARO-78-1, 1978.

Treurniet, W. C. Spacing of characters on a television display. In P. A. Kolers, M. E. Wrolstad, and H. Bouma (Eds.) Processing of visible language 2. New York: Plenum, 1980.

Treurniet, W. C. Display of text on television. Ottawa: Dept. of Communications, C.R.C. Technical Note No. 705-E, 1981.

Winer, B. J. Statistical procedures in experimental design. New York: McGraw-Hill, 1962.