DESIGNING AN INTERFACE TO OPTIMIZE READING WITH SMALL DISPLAY WINDOWS

TARJIN RAHMAN* and PAUL MUTER, University of Toronto, Toronto, Ontario, Canada

*tarjinr@microsoft.com

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

HUMAN FACTORS, 41
, No. 1, March 1999, pp. 106-117. Copyright (c) 1999, Human Factors and Ergonomics Society. All rights reserved.

ABSTRACT

The extent of electronic presentation of text in small display windows is mushrooming. In the present paper, 4 ways of presenting text in a small display window were examined and compared with a normal page condition: rapid serial visual presentation (RSVP), RSVP with a completion meter, sentence-by-sentence presentation, and sentence-by-sentence presentation with a completion meter. Dependent measures were reading efficiency (speed and comprehension) and preference. For designers of hardware or software with small display windows, the results suggest the following: (a) Though RSVP is disliked by readers, the present methods of allowing self-pacing and regressions in RSVP are efficient and feasible, unlike earlier tested methods; (b) slower reading in RSVP should be achieved by increasing pauses between sentences or by repeating sentences, not by decreasing the presentation rate within a sentence; (c) completion meters do not interfere with performance and are usually preferred; (d) the space-saving sentence-by-sentence format is as efficient and as preferred as the normal page format.

INTRODUCTION

A great amount of learning and working is done through reading. The majority of electronically presented text readily available today mimics the technology of the book by displaying "digital pages" on a screen. However, present technology is capable of more than just mimicking the technology of the book; new and improved ways of reading text with computers should be offered (Chiou, 1995; Muter, 1996). Thus, the question arises: Given the extraordinary power of computers and given what we know about the reading process, what is the best way of presenting electronic text?

New telecommunications technologies that can be used to access and send electronic text are likely to become extremely popular in the near future. Examples of this technology include wireless and desktop phones with displays. These devices typically have small display windows. Even on systems with large displays, text is often presented in small windows. In general, the extent of presentation of text in small display windows is mushrooming. The present paper focuses on a more specific question: What is the best way of presenting electronic text when display space is limited and when the goal is to maximize efficiency (reading speed and comprehension) and user preference? The experiments described in this paper explored methods of presenting text in a small display window. More specifically, the experiments had three main purposes: (a) to test sentence-oriented presentation in both spatial and temporal formats with small windows, (b) to assess the utility of completion indicators with these methods, and (c) to attempt to find a good method of allowing self-paced reading with small display windows.

Rapid Serial Visual Presentation

One way of presenting text in a small window is known as rapid serial visual presentation (RSVP; Juola, 1988; Juola, Tiritoglu, & Pleunis, 1995; Juola, Ward, & McNamara, 1982; Masson, 1983; Potter, 1984). With RSVP, text is presented one or more words at a time at a fixed location on the screen. (The general technique of presenting information temporally instead of spatially is often called RAP COM, for rapid communication; e.g., Konrad, Kramer, Watson, & Weber, 1996.)

RSVP was first introduced by Forster (1970) as a method for studying language processing and comprehension, although Gilbert (1959) studied tachistoscopic presentation of text in a similar sequential fashion. RSVP promised a different way to read text that would eliminate the need to make eye movements and would perhaps reduce the cognitive load in reading. This introduced the potential for reading at a speed that is faster than normal. However, most of the research showed the success of RSVP to be limited to reading rates and comprehension scores that were similar to normal page formats at best (Juola, 1988).

Sentence-Oriented Presentation

A second method of presenting text in a small window is sentence-by-sentence presentation. Several pieces of evidence suggest that sentence boundaries are important in reading and that sentence-by-sentence presentation may enhance efficiency. Stine (1990) studied on-line reading of single sentences by younger and older adults. She showed that younger adults spent more time at sentence boundaries and major and minor clause boundaries than at other places in the text. Just and Carpenter (1980) studied the reading of scientific passages by college students, and found that readers made longer pauses at points at which processing loads were the greatest, such as when accessing infrequent words, integrating information from important clauses, and making inferences after sentences.

Moore and Zabrucky (1995) showed that presenting complete sentences one at a time on-line and prompting the reader to click a button to advance or regress one sentence resulted in better comprehension than reading the complete text on paper. Unfortunately, this result is somewhat difficult to interpret, because participants spent more time reading text on the computer screens than on paper. However, a series of hierarchical regression analyses suggested that the time difference did not account for the greater comprehension in the computer presentations.

The importance of sentence boundaries has not been completely neglected in RSVP research. Most RSVP studies have included a 250-500 ms blank window to enhance post-processing of sentences and the integration of ideas between sentences (Chen, 1986; Juola, 1988; Juola, Haugh, Trast, Ferraro, & Liebhaber, 1987; Masson, 1983; Potter, 1984).

In the present experiments, emphasis was placed on sentence boundaries in three ways. First, two conditions were included in which text was presented spatially one sentence at a time, similar to the experiment by Moore and Zabrucky (1995). Second, in the RSVP conditions, text was presented temporally one sentence at a time; participants pressed a key to advance to the next sentence. Third, in the RSVP conditions, the only type of regression permitted was repetition of the current sentence, and likewise, in the sentence conditions, participants were permitted to reread only the current sentence.

Completion Indicators

Another purpose of the present experiments was to examine passage completion indicators, which apparently have been overlooked in previous research. When people read text from a book or a typical computer screen, they are exposed to peripheral cues that allow them to judge how far they have read and how much more they need to read before they are finished. These cues are absent in the RSVP formats studied to date. The effects of a meter that graphically displayed both the number of sentences remaining in the passage and the number of sentences read were tested in the present experiments.

Self-Paced Reading

Past attempts to permit self-paced reading of computerized text with small windows have typically resulted in poor efficiency (Chen & Chan, 1990; Chen, Chan, & Tsoi, 1988; Muter, Kruk, Buttigieg, and Kang, 1988). Muter et al. (1988) tested RSVP with regressions in several experiments. The authors pointed out that a key difference between the printed page and a typical RSVP task is that the reader has the ability to reread words, phrases, or sentences when reading the printed page. Muter et al. permitted various kinds of regressions in RSVP and found that regressions back to the beginning of a sentence were more frequent than regressions two words back, and that reading time was substantially slower when the option of making regressions was given during the RSVP task. There was no compensating improvement in comprehension. The authors concluded that although permitting reader control in RSVP is feasible, it may result in slow reading. Similarly, in a task in which text was scrolled horizontally in a small window, Chen and Chan (1990) found that permitting readers to control the pace of text presentation by turning a knob resulted in poorer comprehension than did computer-controlled pacing.

Regressions for single-sentence computer presentations were examined by Zabrucky and Moore (1994). Participants could go back and reread any sentence they wished. The present study, however, did not allow for multiple-sentence regressions, following the recommendation of Muter et al. (1988) that permitting regressions may result in slow reading. By not permitting unlimited multiple-sentence regressions, we hoped that the present study would optimize reading efficiency.

EXPERIMENT 1

Experiment 1 tested four methods of interactively reading text on a computer screen: RSVP, RSVP with visual feedback in the form of a completion meter showing how much of the passage has been read, sentence-by-sentence presentation, and sentence-by-sentence presentation with a completion meter. These conditions were compared with a normal page format. In all four experimental conditions, a key-press initiated presentation of the next sentence. RSVP within-sentence rates for both practice and experimental trials were set at the participant's own average reading speed for a normal page format in a practice session.

Dependent measures were reading speed, comprehension, and user preference. The combined dependent measure of reading efficiency (Speed x Percentage Correct on the Comprehension Test; Jackson & McClelland, 1979; Juola, 1988) was computed as an overall measure of efficiency.

Method

Participants. In response to campus advertisements, 20 university students (6 men and 14 women between the ages of 18 and 26 years) were recruited as volunteers. The mean age was 21.1 years. Some students participated for credit for an introductory psychology course; others were paid an honorarium.

Materials and apparatus. Participants read 20 passages - 15 in the practice session and 5 in the experimental session - from the comprehension portion of the Graduate Record Examination study guide for the General Test (Brownstein, Weiner, & Green, 1994). The 15 practice passages had a mean of 400.0 words with a standard deviation of 126.3 words, and the range was from 197 to 571 words. The 5 experimental passages had a mean of 414.0 words with a standard deviation of 115.1 words, and the range was from 282 to 553 words. Passages were displayed in black on a white background using a PowerPC Macintosh with a 15-inch (38.1 cm) display with a resolution of 640 x 480 pixels. The text was presented in a 12-point sans serif variable-width font called Geneva. The normal page and sentence-by-sentence formats were 6 inches (15.24 cm) in length from left to right, the usual setting for 1-inch (2.54 cm) margins in Microsoft Word 5.1. In the normal page format, text was single-spaced with a maximum of 23 lines per page, and the indentations at the beginning of paragraphs were removed to be consistent with the RSVP formats. Similarly, in the sentence conditions, text was single-spaced with no indentations, but started on the middle line of the screen. In the RSVP conditions, text was centered horizontally and vertically.

___________________________________________________

Figure 1. (a) The normal page condition with a portion of text from one of the test passages; (b) the RSVP condition with a completion meter showing that the current position is the ninth sentence of the same text as in (a); (c) the sentence-by-sentence condition with a completion meter showing that the current position is the ninth sentence of the same text as in (a). Figure is not actual size of display.

___________________________________________________

Figure 1a shows an example of the normal page condition. Figure 1b shows a portion from the same passage in the RSVP condition with a completion meter. The completion meter displays eight leading periods, indicating that the participant is currently reading the 9th sentence. There are 12 vertical bars to the right of the periods, indicating that there are 12 sentences left to read, including the present sentence. Each time the participant pressed a key to advance to the next sentence, a vertical bar turned into a period. Figure 1c shows the same portion of text in the sentence-by-sentence condition with a completion meter.

Design. There were five conditions: normal page, RSVP, RSVP with a completion meter, sentence-by-sentence, and sentence-by-sentence with a completion meter. Every participant performed in all five conditions.

Procedure. Two 1-h sessions were held on consecutive days. Day 1 served as a practice session in which participants read 15 passages, 3 in each of the five presentation formats, without being tested for comprehension. A 5 x 5 Latin square (four participants per row) determined the order in which the conditions were encountered by the participant, with one constraint: In Experiment 1, the normal page format was always the first condition that each participant experienced on Day 1. The within-sentence presentation rate for the RSVP conditions that followed was calculated for a particular participant from that participant's mean reading speed in the normal page format. Participants read 3 passages in each format before experiencing the next format. The 15 practice passages were randomly distributed among the five conditions for each participant. After the participant read each passage, the experimenter returned to the room to set up the next passage. This practice session was meant to provide the participant with some experience with each of the conditions before an experimental session on Day 2.

Each participant was tested in isolation. Each condition was thoroughly explained to the participant at the beginning of Day 1. Participants were instructed to read each passage as quickly and as accurately as possible. In every condition, the following message appeared centered on the screen before a passage was presented: "PRESS THE RED KEY TO BEGIN." The apostrophe key on the keyboard, directly to the left of the return key, was labeled with a red star. The semicolon key, immediately to the left of the apostrophe key, was labeled with a gold star.

In the normal page condition, participants were instructed to press the red key to advance to the next page. Pressing the gold key resulted in the display of the previous page. (It was not possible to back up two pages.) The current page number appeared in the top right-hand corner of each screen. After the participants had finished reading the last page, they were instructed to press the red key to arrive at a finish screen, at which time the experimenter would come back to set up the next passage.

Similarly, the RSVP conditions required the participant to press the red key to advance to the next sentence and to press the gold key to repeat the current sentence. The last word of each sentence remained on the screen until a red or gold key was pressed.

The sentence-by-sentence conditions required the participant to press the red key to advance to the next sentence; the gold key did nothing. Each sentence remained on the screen until the red key was pressed. Thus, participants were permitted to reread the current sentence as many times as they wished, as in the RSVP conditions. In all conditions, the final key-press of a passage stopped the clock for the determination of reading speed in words/min.

On Day 2, the participants read five new passages, one in each of the five presentation formats. Participants experienced the same order of the conditions as they did on Day 1, but without the constraint of always encountering the normal page condition first. In the RSVP conditions, the within-sentence presentation rates were the same as on Day 1. The Latin square from Day 1 was augmented to a 5 x 5 Graeco-Latin square (two superimposed orthogonal squares) in order to determine the counterbalancing of the five passages on Day 2. Thus, across the experiment on Day 2, every condition appeared equally often in each position, and every passage appeared equally often in each position.

Participants on Day 2 answered either four or six 5-alternative multiple choice questions immediately after reading each passage. The comprehension test was a pen-and-paper task. Comprehension was measured as percentage correct. As with Day 1, reading speed was measured in words/min. After the participants finished answering the questions for the last condition, they were asked to rank the five conditions in order of preference. Finally, the participants were fully debriefed.

Results

Throughout this paper, when the average of the Greenhouse-Geisser and Huynh-Feldt epsilon estimates of deviation from sphericity was greater than .70, Tukey post-hoc tests with specific mean-squared error terms were used (following Stevens, 1992, 448-454); when the average epsilon estimate was less than .70, Bonferroni tests were used. In both cases, the alpha for post-hoc tests was .05 overall; the alpha for each individual post-hoc test was therefore much less than .05. All of the results reported for Experiment 1 are for Day 2 (the experimental session).

Reading efficiency was defined as Speed (reading speed) x Comprehension (percentage correct on the comprehension test). Efficiency results are presented in Figure 2a. The means in Figure 2a suggest that efficiency was at least as high in the sentence conditions as in the page condition, the completion indicator did not detract from performance, and efficiency was poorer in the RSVP conditions than in the other conditions.

___________________________________________________

Figure 2. Results of the five conditions on Day 2 in Experiment 1 with standard error bars (w indicates "with completion meter"): (a) reading efficiency (Speed x Comprehension); (b) reading speed (words/min); (c) comprehension (percentage correct on the comprehension test); (d) preference rankings.

___________________________________________________

A one-way within-subject analysis of variance (ANOVA) was performed on reading efficiency and revealed significant differences among the five conditions, F(4, 76) = 3.24, p < .017. The Greenhouse-Geisser epsilon correction for deviation from sphericity (Stevens, 1992) resulted in p < .022. The Tukey tests yielded significant differences between the sentence condition and the RSVP condition and between the sentence condition and the RSVP condition with the completion meter. The differences between the sentence condition with completion meter and the RSVP conditions were not quite significant. Also, the differences between the page condition and the RSVP conditions were not quite significant. The effect of the completion meter on RSVP and the effect of the completion meter on the sentence condition did not approach significance. A 2 x 2 (RSVP vs. Sentence x Presence vs. Absence of Completion Meter) ANOVA on the four experimental conditions (excluding the normal page condition) revealed that there was a significant effect of presentation mode (Sentence > RSVP), F(1, 19) = 16.54, p < .001; that completion meter had no effect, F(1, 19) < 1; and that there was no interaction, F(1, 19) < 1.

Reading speed in Experiment 1 is plotted in Figure 2b. The pattern appears to be similar to the pattern for efficiency. The ANOVA yielded a significant effect of reading speed, F(4, 76) = 5.36, p < .001 (with the Greenhouse-Geisser correction, p < .005). Bonferroni tests yielded significant differences between the RSVP condition and the sentence condition and between the RSVP condition with the completion meter and the sentence condition. The differences between the sentence condition with completion meter and the RSVP conditions approached significance, as did the differences between the page condition and the RSVP conditions. A 2 x 2 ANOVA on the four experimental conditions revealed a significant effect of presentation mode (RSVP vs. sentence), F(1, 19) = 11.29, p < .003; did not quite reveal a significant effect of completion meter, F(1, 19) = 3.305, p < .085; and revealed no interaction, F(1, 19) < 1.

Comprehension (percentage correct on the comprehension test) in Experiment 1 is plotted in Figure 2c. The overall ANOVA on comprehension failed to show any differences among the conditions, F(4, 76) = 1.391. However, the 2 x 2 ANOVA (excluding the page condition) revealed a significant effect of presentation mode, F(1, 19) = 6.80, p < .017. Inspection of Figure 2c suggests that comprehension was better in the sentence conditions than in the RSVP conditions. In this 2 x 2 ANOVA, the main effect of completion meter was not significant, F(1, 19) < 1, nor was the interaction between presentation mode and presence versus absence of completion meter, F(1, 19) < 1.

Participants' preference was quantified (Figure 2d) by awarding 4 points for first choice, 3 points for second choice, 2 points for third choice, 1 point for fourth choice, and 0 points for last choice rankings. It is clear from Figure 2d that RSVP was preferred less than the other conditions, and that in RSVP, a completion meter was preferred. A one-way within-subject ANOVA on preference revealed a significant effect, F(4, 76) = 26.84, p < .001 (with the Greenhouse-Geisser correction, p < .001). Tukey tests indicated that all pairwise comparisons were significant except sentence versus sentence with completion meter, sentence versus page, and sentence with completion meter versus page. According to the 2 x 2 ANOVA (excluding the page condition), the effect of presentation mode (RSVP vs. sentence) was significant, F(1, 19) = 77.90, p < .001, the effect of completion meter was significant, F(1, 19) = 10.33, p < .005, and the interaction was not significant, F(1, 19) < 1.

Discussion

In Experiment 1, sentence-by-sentence reading was just as efficient as normal page reading. RSVP was less efficient than sentence-by-sentence presentation and was less preferred. Completion meters did not interfere with efficiency, and in the RSVP conditions, users reliably preferred a completion meter.

One reason that readers performed poorly in both RSVP conditions might be that the RSVP within-sentence rates in this experiment, like those in Muter et al. (1988) but unlike those reported and reviewed by Juola (1988), were not necessarily rapid. The theoretical advantage of RSVP is based on the reduction of unnecessary eye-movements, and participants may have been making multiple eye-fixations on each word during RSVP in the present experiment. The diversion of cognitive resources from the task of comprehension to the task of programming eye movements may have affected performance.

Completion meters did not detract from efficiency in the RSVP and sentence-by-sentence conditions. Preference scores confirm the hypothesis that this kind of feedback could be important to readers using these technologies.

EXPERIMENT 2

Experiment 2 was conducted to extend and confirm the findings discussed in Experiment 1. The RSVP within-sentence rate was fixed at a relatively high rate (approximately 260 words/min) to reduce the probability of participants making eye-movements during the presentation of a word. With a fixed rate in RSVP, there was no need to run the normal page condition first in Day 1; thus, the practice trials were completely counterbalanced in Experiment 2.

Method

Participants. In response to campus advertisements, 20 university students - 6 men and 14 women between the ages of 19 and 38 years - were recruited as volunteers. The mean age of participants was 22.1 years. Some students participated for credit for an introductory psychology course, and others were paid an honorarium.

Materials, Apparatus, and Design. These were the same as in Experiment 1.

Procedure. The procedure in Experiment 2 was the same as in Experiment 1, with the following exceptions: (a) The RSVP within-sentence rate was fixed at approximately 260 words/min for each participant; (b) the constraint that the normal page condition was always the first to be experienced was removed; and (c) the order of the Graeco-Latin square was reversed, right-to-left, to extend the robustness of the findings.

Results

Efficiency results for the experimental session in Experiment 2 are presented in Figure 3a. A one-way within-subject ANOVA on reading efficiency (Speed x Percentage Correct on the Comprehension Test) revealed no significant differences among the conditions, F(4, 76) = 1.68. Similarly, the 2 x 2 ANOVA (excluding the page condition) indicated that the effect of completion meter was not quite significant, F(1, 19) = 3.317; the effect of presentation mode (RSVP vs. sentence) was not significant, F(1, 19) = 2.02; and the interaction was not significant, F(1, 19) = 1.18.

___________________________________________________

Figure 3. Results of the five conditions on Day 2 in Experiment 2 with standard error bars (w indicates "with completion meter"): (a) reading efficiency (Speed x Comprehension); (b) reading speed (words/min); (c) comprehension (percentage correct on the comprehension test); (d) preference rankings.

___________________________________________________

With regard to reading speed (Figure 3b), the one-way ANOVA indicated no significant effect, F(4, 76) = 1.73. Similarly, the 2 x 2 ANOVA (excluding the page condition) yielded no significant effects of presentation mode, F(1, 19) = 2.42, completion meter, F(1, 19) < 1, or their interaction, F(1, 19) < 1. However, there was a reliable overall effect on comprehension (Figure 3c), F(4, 76) = 2.70, p < .037 (with the Greenhouse-Geisser correction, p < .043). Visual inspection of Figure 3c suggests that comprehension in the sentence condition with the completion meter was superior to comprehension in all of the other conditions. According to the Tukey tests, comprehension in the sentence condition with the completion meter was significantly better than in the RSVP condition and the sentence condition without the completion meter, and almost significantly superior to the other two conditions. The 2 x 2 ANOVA indicated that the effect of completion meter on comprehension was significant, F(1, 19) = 5.41 p < .031; the effect of presentation mode (RSVP vs. sentence) was not quite significant, F(1, 19) = 3.06; and the interaction was not quite significant, F(1, 19) = 2.87.

Preference results are displayed in Figure 3d. Among the four experimental conditions, the rank order of preference was sentence with completion meter, sentence, RSVP with completion meter, and RSVP. The ANOVA on preference revealed significant differences among the conditions, F(4, 76) = 15.55, p < .001 (with the Greenhouse-Geisser correction, p < .001). According to the Tukey tests, the four experimental conditions were all significantly different from each other. In addition, the page condition was significantly preferred over the RSVP condition. The 2 x 2 ANOVA confirmed a significant effect of presentation mode, F(1, 19) = 38.80, p < .001; a significant effect of completion meter, F(1, 19) = 9.62, p < .006; and no interaction, F(1, 19) < 1.

For Experiment 2, the order of conditions was completely counterbalanced on Day 1, and some results are presented here. Reading speeds during the practice trials are shown in Figure 4. (There were no comprehension tests on Day 1.) The ANOVA on the mean reading speed for each condition revealed significant differences, F(4, 76) = 3.94, p < .006 (with the Greenhouse-Geisser correction, p < .015). A 4 (conditions) x 3 (trials) ANOVA on the four small-window conditions alone was also performed. The effects of both condition, F(3, 57) = 3.06, p < .05, and trial, F(2, 38) = 12.39, p < .001, were significant. The interaction was not significant, F(6, 114) = 0.95. Reading speed increased with practice in the unfamiliar small-window conditions: The means for Trials 1, 2, and 3 were 184.6, 194.4, and 201.0 words/min, respectively. In contrast, for the familiar normal page condition, there was no such increase. The means for Trials 1, 2, and 3 were 207.5, 213.2, and 205.8 words/min.

___________________________________________________

Figure 4. Reading speed (words/min) for each trial on Day 1 (practice) in Experiment 2 with standard error bars (w indicates with completion meter; 1, 2, and 3 indicate trial number).

___________________________________________________

Discussion

The different method of RSVP presentation in Experiment 2 appears to have eliminated the poor performance in RSVP. In the RSVP conditions, perhaps the higher within-sentence rate of presentation in Experiment 2 prevented or reduced eye movements and thereby facilitated performance. To our knowledge, Experiment 2 is the first experiment in which an RSVP format with self-pacing and regressions has been demonstrated to be as efficient as normal page reading. As in Experiment 1, sentence-by-sentence reading was also just as efficient as normal page reading. Within the sentence-by-sentence conditions, the completion meter reliably facilitated comprehension.

Though performance in the RSVP conditions was better in Experiment 2 than in Experiment 1, RSVP was still less preferred than the other conditions. In addition, results indicated a preference for the completion meter.

In the practice session, performance improved over the three trials for the unfamiliar conditions, but not for the familiar normal page condition. Perhaps with extended practice, performance with these novel formats would continue to improve.

GENERAL DISCUSSION

Electronic text is becoming increasingly common, and it is important to develop and test technologies and interfaces for electronic text that are both appealing and efficient for users. The present experiments explored several ways of exploiting the power of a computer to present continuous text in a small display window. The methods of Experiment 2 - sentence-by-sentence presentation, either spatial or temporal (RSVP), with capability of rereading only one sentence, and with rapid within-sentence presentation rates in RSVP - seem to be viable means of presenting continuous text and allowing self-pacing. Reading efficiency was just as high in these conditions as in a conventional normal page condition.

In Experiment 1, individual differences in reading speed were accommodated partly by varying within-sentence presentation rates, and poor efficiency resulted. In Experiment 2, speed was varied only by permitting participants to increase pauses between sentences or to repeat sentences, and efficiency was not reliably different from efficiency in the other conditions. The present experiments suggest that slower reading in RSVP should be achieved by increasing pauses between sentences or by repeating sentences, not by decreasing the within-sentence presentation rate. RSVP within-sentence rate should be sufficiently high to reduce or eliminate eye movements. This conclusion is consistent with the theoretical basis for the conception of RSVP (Potter, 1984).

Experiment 2 is the first experiment we know of in which an RSVP format with self-pacing and regressions has been demonstrated to be as efficient as page format reading. The present study is encouraging for the development of RSVP software, not only for skilled readers but also for special populations such as dyslexics (Potter, 1984) and people who have retinitis pigmentosa or other problems that cause impaired peripheral vision (Williamson, Muter, and Kruk, 1986).

Of course, RSVP and (to a lesser extent) spatial sentence-by-sentence presentation are relatively unfamiliar to readers. Extended practice may result in even better performance with these formats. In a monitoring task, Lang, Utesch, and Fly (1993, Experiment 2) found that extended practice with temporally presented information resulted in more improvement in error rate than did extended practice with the more familiar spatial format. Some evidence for practice effects in the unfamiliar formats was found in Day 1 of Experiment 2 in the present study, even though each participant read for approximately only 10 min in each of the conditions. Extended practice may also reduce the readers' dislike of RSVP, which was clear in the present experiments.

The present research introduced a reading completion meter and showed that it did not interfere with performance. In Experiment 2 in the sentence conditions, the completion meter actually facilitated comprehension. In most cases, participants preferred to have this feedback. The use of completion indicators demands further research, perhaps with different modalities. One reason that completion meters may have been important in the present experiment is that the lengths of the passages varied considerably. The practice passages ranged from 197 to 571 words in length, and the experimental passages ranged from 282 to 553 words in length. A completion meter may not be necessary when passages are short and homogeneous in length, because participants may be able to judge how much material is still to be read.

Completion indicators enable readers to know what proportion of the text has been read. A further potential advantage of completion meters relates to location cues. Rothkopf (1971; see also Piolat, Roussey, & Thunin, 1997) found that with page format presentation (assuming text is advanced by paging and not scrolling), incidental memory for location within a page may facilitate processing. Completion indicators may at least partially compensate for the fact that there are no incidental location cues with small-window displays; that is, incidental memory for the state of the completion indicator may play a role similar to incidental memory for location.

In addition to RSVP and sentence-by-sentence spatial presentation, another means of delivering text in a small window is the Times Square method, also known as leading. With this method, text scrolls from right to left in a limited area of the screen in a fashion similar to advertisements on electronic billboards. Most researchers have found that the Times Square method results in inferior performance compared with the RSVP and page format (Granaas, McKay, Laham, Hurt, & Juola, 1984; Juola et al., 1995; Sekey & Tietz, 1982). However, Kang and Muter (1989) found that pixel-by-pixel scrolling with a Times Square format resulted in performance at least as good as RSVP. (The other Times Square studies used more coarse-grained scrolling.) Kang and Muter also found that readers preferred Times Square over RSVP. It remains to be seen how a similar self-paced version of Times Square that allows only single-sentence regressions would compare with the present findings.

The results from the present study have implications for the effective management of larger displays as well. Internet web sites can potentially benefit from restructuring text-intensive pages into user interfaces that display text interactively in small windows like those used in the present study. For example, in a task in which participants were required to recall digits, Matin and Boff (1988) found that RSVP was particularly effective when the display was cluttered, as web pages often are. The savings in screen space could be used to offer other resources and more functionality for the user. Similarly, if small display windows are used, computers (which have been unpopular for the purpose of reading entire books) may allow the reader to work with the text in ways that have been enjoyed only through printed media (see O'Hara & Sellen, 1997). It is feasible to design an interface in which text is interactively displayed one sentence at a time while a separate region on the screen is devoted to note taking. By designing interfaces to work with text rather than to only display it, the paperless office may take another step toward an efficient reality.

With earlier computer systems, reading was often found to be substantially slower with computers than with books (Gould & Grischkowsky, 1984; Kruk & Muter, 1984; Muter, Latremouille, Treurniet, & Beam, 1982). This deficit was apparently eliminated with better screens (Muter & Maurutto, 1991). The question remains as to whether a means of presenting text electronically can be found that surpasses the technology of the book. It is possible that the limit on the efficiency of reading is in the human brain and is not set by the format of presentation, but we are optimistic that better presentation methods can produce better results. The present experiments may provide clues for future directions and improvements.

ACKNOWLEDGMENTS

This work was supported by the Natural Sciences and Engineering Research Council of Canada. We thank Monica Castelhano, Joan M. Cherry, David Olson, and Nancy L. Williamson for their valuable comments. This work is based on Tarjin Rahman's master's thesis with Paul Muter.

REFERENCES

Brownstein, S. C., Weiner, M., & Green, S. W. (1994). Barron's Students' #1 Choice: How to Prepare for the GRE Graduate Record Examination General Test, Eleventh Edition. New York: BarronŐs Educational Series Inc.

Chen, H.-C. (1986). Effects of reading span and textual coherence on rapid-sequential reading. Memory & Cognition, 14, 202-208.

Chen, H.-C, & Chan, K.-T. (1990). Reading computer-displayed moving text with and without self-control over the display rate. Behaviour & Information Technology, 9, 467-477.

Chen, H.-C, Chan, K.-T, and Tsoi, K.-C. (1988). Reading self-paced moving text on a computer display. Human Factors, 30, 285-291.

Chiou, G.-F. (1995). Reader interface of computer-based reading environment. International Journal of Instructional Media, 22, 121-133.

Forster, K. L. (1970). Visual perception of rapidly presented word sequences of varying complexity. Perception & Psychophysics, 8, 215-221.

Gilbert, L. C. (1959). Speed of processing verbal stimuli and its relation to reading. Journal of Educational Psychology, 55, 8-14.

Gould, J. D., & Grischkowsky, N. (1984). Doing the same work with hard copy and with cathode-ray tube (CRT) computer terminals. Human Factors, 26, 323-337.

Granaas, M. M., McKay, T. D., Laham, R. D., Hurt, L. D., & Juola, J. F. (1984). Reading moving text on a CRT screen. Human Factors, 26, 97-104.

Jackson, M. D. & McClelland, J. L. (1979). Processing determinants of reading speed. Journal of Verbal Learning and Verbal Behavior, 108, 151-191.

Juola, J. F. (1988). The use of computer displays to improve reading comprehension. Applied Cognitive Psychology, 3, 87-95.

Juola, J. F., Haugh, D., Trast, S., Ferraro, R., & Liebhaber, M. (1987). Reading with and without eye movements. In J. K. OŐRegan and A. Levy-Schoen (Eds), Eye Movements: From Physiology to Cognition. Amsterdam: Elsevier.

Juola, J. F., Tiritoglu, A., & Pleunis, J. (1995). Reading text presented on a small display. Applied Ergonomics, 26, 227-229.

Juola, J. F., Ward, N. J., & McNamara, T. (1982). Visual search and reading of rapid serial presentations of letter strings, words, and text. Journal of Experimental Psychology: General, 111, 208-227.

Just, M. A., & Carpenter, P. A. (1980). A theory of reading: From eye fixations to comprehension. Psychological Review, 87, 329-354.

Kang, T. J., & Muter, P. (1989). Reading dynamically displayed text. Behaviour & Information Technology, 8, 33-42.

Konrad, C.M., Kramer, A.F., Watson, S.E., & Weber, T.A. (1996). A comparison of sequential and spatial displays in a complex monitoring task. Human Factors, 38, 464-483.

Kruk, R. S., & Muter, P. (1984). Reading of continuous text on video screens. Human Factors, 26, 339-345.

Lang, V.A., Utesch, B.S., & Fly, P.M. (1993). Effects of display format and colour coding schemes on visual monitoring performance using spatially and temporally distributed visual displays. Displays, 14, 195-206.

Masson, M. E. J. (1983). Conceptual processing of text during skimming and rapid sequential reading. Memory & Cognition, 11, 262-274.

Matin, E. & Boff, K.R. (1988). Information transfer rate with serial and simultaneous visual display formats. Human Factors, 30, 171-180.

Moore, D., & Zabrucky, K. (1995). Adult age differences in comprehension and memory for computer-displayed and printed text. Educational Gerontology, 21, 139-150.

Muter, P. (1996). Interface design and optimization of reading of continuoustext. In van Oostendorp, H., & de Mul, S. (Eds.), Cognitive Aspects of Electronic Text Processing. Norwood, N.J.: Ablex.

Muter, P., Kruk, R. S., Buttigieg, M. A., & Kang, T. J. (1988). Reader-controlled computerized presentation of text. Human Factors, 30, 473-486.

Muter, P., Latremouille, S. A., Treurniet, W. C., & Beam, P. (1982). Extended reading of continuous text on television screens. Human Factors, 24, 501-508.

Muter, P., & Maurutto, P. (1991). Reading and skimming from computer screens and books: The paperless office revisited? Behaviour & Information Technology, 10, 257-266.

O'Hara, K. & Sellen, A. A comparison of reading paper and on-line documents (1997). In Proceedings of CHI '97 (Atlanta, Georgia), New York: ACM Press.

Piolat, A., Roussey, J.-Y., & Thunin, O. (1997). Effects of screen presentation on text reading and revising. International Journal of Human-Computer Studies, 47, 513-607.

Potter, M. C. (1984). Rapid serial visual presentation (RSVP): A method for studying language processing. In Kieras, D. E., & Just, M. A. (Eds.), New Methods in Reading Comprehension Research. Hillsdale, N.J.: Erlbaum.

Rothkopf, E.Z. (1971). Incidental memory for location of information in text. Journal of Verbal Learning and Verbal Behavior, 10, 608-613.

Sekey, A. & Tietz, J. (1982). Text display by 'saccadic scrolling'. Visible Language, 16, 62-76.

Stevens, J. (1992). Applied Multivariate Statistics for the Social Sciences, Second Edition. Hillsdale, N.J.: Erlbaum.

Stine, E. A. L. (1990). On-line processing of written text by younger and older adults. Psychology and Aging, 5, 68-78.

Williamson, N. L., Muter, P., & Kruk, R. S. (1986). Computerized presentation of text for the visually handicapped. In E. Hjelmquist & L. G. Nilsson (Eds.), Communication and Handicap: Aspects of Psychological Compensation and Technical Aids. Amsterdam: North Holland.

Zabrucky, K., & Moore, D. (1994). Contributions of working memory and evaluation and regulation of understanding to adultsŐ recall of text. Journal of Gerontology: Psychological Sciences, 49, 201-212.


Tarjin Rahman received his M.A. in applied cognitive science from the Ontario Institute for Studies in Education, University of Toronto, in 1997. He is a graduate student at McGill University in Montreal, Quebec, Canada.

Paul Muter received his Ph.D. in experimental psychology in 1979 from the University of Toronto. He is an assistant professor in the Department of Psychology at the University of Toronto.

Date received: January 20, 1998
Date accepted: August 13, 1998