Conceived and designed the experiments: HM SW YM-K KU TS. Performed the experiments: HM. Analyzed the data: HM. Wrote the paper: HM YM-K KU TS. Recruited the trainees: TS. Conceived and supervised the project: SW.
The authors have declared that no competing interests exist.
A growing body of evidence suggests that meditative training enhances perception and cognition. In Japan, the Park-Sasaki method of speed-reading involves organized visual training while forming both a relaxed and concentrated state of mind, as in meditation. The present study examined relationships between reading speed, sentence comprehension, and eye movements while reading short Japanese novels. In addition to normal untrained readers, three middle-level trainees and one high-level expert on this method were included for the two case studies.
In Study 1, three of 17 participants were middle-level trainees on the speed-reading method. Immediately after reading each story once on a computer monitor, participants answered true or false questions regarding the content of the novel. Eye movements while reading were recorded using an eye-tracking system. Results revealed higher reading speed and lower comprehension scores in the trainees than in the untrained participants. Furthermore, eye-tracking data by untrained participants revealed multiple correlations between reading speed, accuracy and eye-movement measures, with faster readers showing shorter fixation durations and larger saccades in X than slower readers. In Study 2, participants included a high-level expert and 14 untrained students. The expert showed higher reading speed and statistically comparable, although numerically lower, comprehension scores compared with the untrained participants. During test sessions this expert moved her eyes along a nearly straight horizontal line as a first pass, without moving her eyes over the whole sentence display as did the untrained students.
In addition to revealing correlations between speed, comprehension and eye movements in reading Japanese contemporary novels by untrained readers, we describe cases of speed-reading trainees regarding relationships between these variables. The trainees overall tended to show poor performance influenced by the speed-accuracy trade-off, although this trade-off may be reduced in the case of at least one high-level expert.
The relationship between reading speed and sentence comprehension has been under investigation for more than 50 years (for an overview, see
At present, little is known about how patterns of eye movements may be associated with potential relationships between reading speed and comprehension. Also, there is relatively little evidence about the relationships between eye movements, reading speed and comprehension in Japanese. Given such paucity of basic data, it is important to investigate these variables and how they relate to each other in reading Japanese text. The present study was designed to examine how eye movements and speed-accuracy relationships may be interconnected under conditions in which participants read Japanese novels as fast as possible while also comprehending the content, followed by content-related questions. The studies were conducted in a well-controlled setting using a computer-assisted setup customized for different participants to collect basic data from normal adults. An eye-tracking system which provides a natural testing environment without any glasses was used to record eye movements while reading.
Besides collecting such basic data, the present research was also focused on issues of speed-reading. The Park-Sasaki method of speed-reading originated in Korea and was later further developed in Japan
There have been a number of publications on this topic over decades of time, some of which reported seemingly positive effects of speed-reading training
Accordingly, in addition to investigating relationships between reading speed, comprehension, and eye movement while reading Japanese novels, the present study also examined sentence comprehension skills by speed-reading trainees. As a beginning step regarding the second point, we included three middle-level trainees and one high-level expert within a case-study approach. After reading Japanese novels on a computer screen, participants answered content-related questions, which aimed to examine the extent of comprehension of the novel. Potential differences in eye movements between the trainees and the untrained participants were also examined using an eye-tracking system.
The primary goal of the first study was to obtain basic data on eye movement measures in Japanese adults while reading novels with comprehension. Participants read each of the four stories, around 9,000 characters long, presented on a LCD screen while switching page displays by pressing a response button. They then answered 16 true or false content-related questions. Eye-tracking was done while reading, which enabled examination of the correlations between reading speed and various eye-movement measures including numbers and durations of fixations and saccadic sizes. Given the scores on the comprehension test, it is also possible to examine correlations between levels of comprehension and eye-movement measures. The second purpose was to preliminarily examine the efficacy of this setup to examine abilities and mechanisms associated with speed-reading. In addition to the normal untrained adults, three middle-level trainees of the Park-Sasaki method participated. If these trainees perform well, i.e., show high comprehension scores with short reading time, that would support the efficacy of this testing as well as the reading skills of the Park-Sasaki experts. In contrast, if they fail to perform well, this would either suggest poor performance of the trainees or potential points for improvement associated with this testing. We expected to find that multiple eye-movement measures correlate not only with reading speed but also with comprehension scores, with potential differences between groups.
A total of 17 Japanese adults participated in the reading sessions. Among them, 14 (10 females and 4 males; age, 25–48 years; mean age = 33.9 years,
The study took place in a dim, sound-attenuated room, where the participants were seated in a comfortable height-adjustable chair. A 58-cm (23.0 inches) TFT LCD monitor (FlexScan EV2333W, EIZO, Ishikawa, Japan) was located on a table in front of the participant. The distance between the monitor and the participant’s eyes was set at 57 cm. One centimeter on the screen corresponded to a visual angle of approximately one degree. The heights of both the monitor and the chair were adjusted for each participant so that he/she could sit in a good posture and gaze at the fixation point in the center of the monitor right in front of him/her. A response device was used that was modified by removing unnecessary keys from a numerical keypad (TK-UFHBK, ELECOM, Osaka, Japan), which the participants put on their thigh. Participants could press a button on the device to switch pages on the display, in a way comparable to a mouse click. These height adjustments and device modifications were made to make the settings similar to that used for speed-reading training, by which the trainees could express their daily abilities. The program for presenting sentence stimuli was written in Microsoft VisualBasic 6.0.
An eye-tracking system (Tobii X120 Eye Tracker, Tobii Technology AB, Danderyd, Sweden) located beneath the monitor in front of the participant recorded the participants’ eye movements on the monitor throughout the reading period. Use of Tobii eye-trackers in studies of reading has become widespread
The texts used for the test sessions were four stories selected from the same series of Japanese novels (
Japanese sentences shown here, i.e., part of the well-known novel
Instruction on the general procedure using
Participants included in the additional two control conditions, i.e., no-reading control and gist/detail control, were non-speed-readers and were ignorant of the purpose of the study. No-reading control participants were simply presented with the question sheets and filled in the answers to the questions for all the test stories used in this study without reading the novels. Gist/detail control participants read the text, questions, referred to the correct answers to the questions and rated each question on a scale of one to ten, based on the extent to which the question reflected the (1) gist and (2) detail of the stories.
Eye-movement data post-processed by Tobii Studio software showing timestamps in milliseconds and coordinates of each fixation were used for the subsequent analyses. A fixation was scored if the gaze remained stationary within a radius of 35 pixels (about 0.9 degrees in visual angle); otherwise, the recorded sample was defined as part of a saccade. The following variables were used to indicate characteristics of eye movements: numbers of leftward and rightward saccades in the X axis; numbers of downward and upward saccades in the Y axis; mean size of saccades (absolute value integrating X and Y components) and its X and Y components separately with their variances; size of saccades in X and Y shown separately for the first-pass reading; mean duration of fixation and its variance; and first-pass reading time (i.e., sum of all fixation times during the first pass) and look-back time within each session. First-pass in this context was determined as all the fixations made from first entering each line of the text until leaving it. First-pass measures should help to consider to what extent comprehension was made during first-pass, or during re-reading to understand further details of the text. Saccades larger than 200 pixels (about 5.3 degrees in visual angle) in the rightward direction for the X axis or in the upper direction for the Y axis were analyzed separately, because these were regarded as saccades made to proceed to the beginning of the next page display and to the beginning of the next line (i.e., “return sweeps”), respectively. Regressive saccades were defined as those moving in the rightward direction for the X axis, or in the upward direction for the Y axis within each line. Correlations were examined for untrained participants between these measures and three independent variables: reading speed (characters read per minute), proportions of correct answers to the questions (comprehension scores), and residual values in Y from the regression line in the speed-accuracy plot, which was assumed to reflect the excellence of performance in the relationships between reading speed and comprehension scores.
Whereas the first study revealed correlations between reading speed, accuracy, and eye movement, it also revealed poor performance in the trainees. Potential problems and points for improvement are: (1) participants should be those at the high stage of expertise, who can both comprehend and remember the content while reading at a high speed; and (2) the present instruction to read as fast as possible while also comprehending may have been misleading, because it could have urged participants to accelerate reading speed while sacrificing comprehension. It would be better to more strongly encourage participants to focus on comprehension. Thus, in the second study, we included a high-level expert as well as new untrained participants and modified the instructions so that participants would prioritize comprehension. We expected that the expert would show performance exceeding that of the control participants, while potentially showing different eye-movement patterns as well.
An advanced expert of the Park-Sasaki method (FT; female, 59 years old) participated. Prior to participation FT had participated in the training course for 7.4 years and attained the high stage of expertise. According to the specific program of the method, she was reportedly capable of reading more than 120,000 characters per minute using the familiar sentences in the training textbook of the method
The same setup as in Study 1 was used for Study 2. The materials were the same as in Study 1 except that one additional story from the same series of novels was used for the test, which now involved five sessions in total. This new story,
Eye movement measures by the eye-tracker were analyzed and examined generally in the same manner as in Study 1. Examination of correlations between eye-movement measures, reading speed and comprehension scores, already reported in Study 1, were not duplicated in this study. In order to identify the parts of the sentence display where the participants’ gaze was frequently fixated, the sentence area of 20×14 cm was divided into subareas of every 2 cm for both the vertical and horizontal axes. These subareas are referred to as X1 to X10 (from left to right) for the horizontal axis and as Y1 to Y7 (from top to bottom) for the vertical axis. Durations of fixations at each of these subareas were examined. All fixations, including those during first-pass reading, re-reading, and return sweeps, were included in this analysis.
Data from all the test sessions were included in the analysis, but one untrained participant (female, 30 years old) quit after finishing two sessions. The average score of the no-reading control participants across the four stories was 55.5%, which was considered to reflect the baseline “zero-comprehension”. The mean rating values regarding the questions by the gist/detail control participants were 5.3 for gist and 7.4 for detail, showing that the questions reflected both the gist and the detail of the stories. Reading speed and percentages of correct responses to the questions for both the trainees and the untrained participants are shown in
Each dot shows data obtained from each reading session. The dashed line indicates the averaged comprehension score by no-reading control participants who gave answers to questions without reading the novels.
speed, accuracy, and eye-movementmeasures | Study 1 | Study 2 | ||
untrained participants | trainees | untrained participants | trained individual(FT) | |
mean (variance) | mean (variance) | mean (variance) | mean (variance) | |
reading speed (character/min) | 1236.6 | 2600.2 | 1193.4 | 5644.0 |
% accuracy on questions | 83.8 | 68.1 | 84.9 | 77.5 |
N leftward saccades | 870.9 | 602.1 | 924.7 | 258.0 |
N rightward saccades | 575.3 | 384.9 | 635.3 | 49.0 |
N downward saccades | 1184.8 | 643.8 | 1278.1 | 220.6 |
N upward saccades | 261.4 | 343.2 | 281.9 | 86.4 |
N return sweeps | 294.8 | 205.3 | 297.5 | 60.2 |
saccade size: total (degree) | 3.6 (7.2) | 3.7 (4.8) | 3.6 (7.9) | 3.0 (2.9) |
saccade size in X: total (degree) | 0.3 (0.7) | 0.4 (1.4) | 0.2 (0.7) | 1.1 (1.3) |
saccade size in Y: total (degree) | 1.7 (5.1) | 1.2 (9.5) | 1.6 (5.3) | 0.7 (4.8) |
saccade size in X: first pass (degree) | 0.3 (0.7) | 0.5 (1.4) | 0.3 (0.7) | 1.1 (1.3) |
saccade size in Y: first pass (degree) | 1.8 (4.7) | 1.4 (9.0) | 1.8 (5.0) | 0.7 (4.8) |
fixation time (msec) | 268 (15525) | 224 (8876) | 278 (24098) | 274 (10527) |
first-pass reading time (sec) | 441.2 | 240.0 | 498.0 | 99.3 |
look-back time (sec) | 26.4 | 22.8 | 27.3 | 0.1 |
Saccade sizes in X and Y are shown both for all saccades and for saccades during the first pass. Saccade sizes are presented in terms of degrees of visual angle. Depicted are the representative values that were averaged across sessions, and then across participants. For saccade size and fixation time measures, variances are also shown in parenthesis.
All analyzed eye-movement measures together with reading speed and comprehension score measures for the trainees and the untrained participants from both studies are summarized in
eye-movementmeasure | reading speed | % correct answers | residual value in y | ||||
|
|
|
|
|
|
||
saccade size: total (pixels) | mean | 0.124 | 0.185 | 0.036 | 0.398 | 0.071 | 0.306 |
variance | −0.468 | <0.001 |
0.046 | 0.369 | −0.084 | 0.273 | |
saccade size in X: total (pixels) | mean | 0.897 | <0.001 |
−0.190 | 0.085 | 0.060 | 0.333 |
variance | 0.362 | 0.004 |
0.245 | 0.037 |
0.346 | 0.005 |
|
saccade size in Y: total (pixels) | mean | −0.175 | 0.103 | −0.038 | 0.394 | −0.086 | 0.267 |
variance | 0.505 | <0.001 |
0.084 | 0.272 | 0.226 | 0.051 | |
saccade size in X: first pass (pixels) | mean | 0.893 | <0.001 |
−0.186 | 0.089 | 0.063 | 0.325 |
variance | 0.355 | 0.004 |
0.247 | 0.036 |
0.347 | 0.005 |
|
saccade size in Y: first pass (pixels) | mean | −0.168 | 0.112 | −0.005 | 0.486 | −0.052 | 0.355 |
variance | 0.541 | <0.001 |
0.089 | 0.260 | 0.241 | 0.040 |
|
fixation time (msec) | mean | −0.577 | <0.001 |
−0.219 | 0.056 | −0.381 | 0.002 |
variance | −0.571 | <0.001 |
−0.113 | 0.207 | −0.273 | 0.023 |
|
% rightward regression | −0.234 | 0.044 |
0.260 | 0.029 |
0.195 | 0.079 | |
% upward regression | 0.454 | <0.001 |
0.058 | 0.338 | 0.185 | 0.090 |
Shown is the summary of correlations between independent variables (reading speed, comprehension scores, and residual values in the speed-accuracy linear regression) and eye-movement measures (saccade size, fixation time, and proportion of regression) across all the sessions from untrained participants.
:
:
:
Correlations between reading speed and eye-movement measures for untrained participants revealed a number of significant tendencies. First, faster reading corresponded to shorter duration of each fixation and smaller variance in fixation time. Second, faster reading corresponded to smaller variances in total saccade sizes. Analyzed separately for X and Y components, faster reading corresponded to larger saccade size in X (in the leftward direction), but not in Y (in the downward direction). Variances in saccade sizes both in X and Y were larger in fast readers than in slow readers. Comparable trends were found for saccade sizes in X and Y during the first pass. Finally, faster reading corresponded to smaller proportions of rightward regressive saccades, i.e., moving rightwards back to the sentences where the participant had already read, but to larger proportions of upward regressive saccades within each line.
Correlations between comprehension scores and eye-movement measures generally failed to reveal strong statistical significance, compared with the former analyses regarding reading speed. One statistical significance was found for the higher comprehension scores corresponding to larger proportions of rightward regressive saccades, as opposed to the case for reading speed.Variances in saccade sizes in X were larger in readers with higher comprehension scores than in those with lower scores, both for all saccades and saccades during the first pass. Otherwise no statistically significant correlations were revealed.
Mean residual value was 2.30 for the untrained participants and −5.86 for the trainees. These numerical data show lower performance by the trainees than by the untrained participants within the speed-accuracy plot. For the untrained participants, shorter duration of each fixation and smaller variance in fixation time corresponded to larger residual values. Also, larger variances in saccade sizes both in X and Y corresponded to larger residual values, more apparently for saccades during the first pass than for all saccades. These show that untrained readers with higher performance within the speed-accuracy plot tended to make fixations with shorter and less variable duration, and to make saccades more variable in size for both the vertical and horizontal axes, compared with those with lower performance. The other correlations involving total saccade sizes and regressive saccades failed to reach statistical significance.
Data from all the test sessions were included in the analysis, but one untrained participant (male, 19 years old) quit after completing four test sessions other than Story 2. The average scores of the participants from no-reading control and FT’s-speed control groups across the five stories were 52.2% and 58.9%, respectively. The average rating values regarding the questions by the gist/detail control participants were 5.0 for gist and 7.1 for detail, showing that the questions reflected both the gist and the detail of the stories. Mean reading speed was numerically 4.7 times faster and mean proportion of correct answers to the questions was numerically 7.4% lower, in the trained individual (FT) than in the untrained participants.
Each dot shows data obtained from each reading session. The dashed line indicates the averaged comprehension score by no-reading control participants, who gave answers to questions without reading the novels. The long dashed dotted line indicates the averaged comprehension score by FT’s-speed control participants, who answered questions after having page displays switched in the same presentation times as FT.
For both reading speed and comprehension score measures, we conducted Mann-Whitney’s
To visualize the differences in looking patterns between the trained in individual (FT) and the untrained participants,
Fixation times for each area of the sentence display are shown in different colors. Each figure shows data for the representative test session using Story 1, (
Shown are the data averaged across stories and then across participants for the untrained. The sentence area was divided into subareas of every 2 cm for the horizontal (X1–X10; [A]) and vertical (Y1–Y7; [B]) axes, respectively. Dashed lines show the proportions expected by the chance level (i.e., 1/10 for [A] and 1/7 for [B]). Error bars indicate standard errors of the mean.
The results revealed various relationships between reading speed, comprehension scores, and eye-movement measures in untrained participants and the trainees. In particular, eye-tracking data for untrained participants revealed multiple significant correlations between reading speed and comprehension scores and eye-movement measures. Specifically, faster reading was associated not only with shorter fixation times and fewer/more rightward/upward regressive saccades, but also with larger horizontal saccadic movements. Saccade sizes during the first pass generally produced trends similar to those for all saccades. This suggests that eye-movement strategies, such as larger saccade sizes in X in fast readers, largely reflect those during first-pass reading of lines of text. In addition, larger variances in saccade sizes in faster reading suggest the use of irregular or unsteady eye-movement strategies as reading rate accelerated. On the other hand, comprehension scores generally revealed less apparent correlations with the eye-movement measures, compared with reading speed. However, higher scores on the comprehension test were associated with more rightward regressive saccades. This could be interpreted as moving back to the previous lines in order to obtain more accurate understanding of the content. Finally, residual value from linear regression in the speed-accuracy plot showed that shorter fixation times with smaller variances and larger variances in saccade sizes corresponded to larger residual values. These support the notion that relatively excellent reading performance was achieved by using more variable eye-movement strategies. These correlations were found in a situation in which untrained participants read as fast as possible while also comprehending the content, that is, while performing at their best. Thus, it seems possible that these data may reflect patterns generally observed when normal native participants read comparable Japanese text, while using reading strategies to read faster than their usual daily speed.
With regard to speed-reading, on the other hand, the results revealed rather poor performance in the Park-Sasaki trainees. That is, the negative speed-accuracy correlation involving both untrained participants and the trainees suggested a speed-accuracy trade-off. This trade-off appears consistent with previous studies: even though a small increase in reading speed is not known to impair comprehension
The reading speed of the trained individual (FT) was numerically 4.7 times that of the untrained participants, and her comprehension scores were significantly higher than the zero-comprehension level for all the five test sessions. This was statistically no different from the comprehension scores of the untrained participants, albeit numerically 7.4% lower on average. FT’s comprehension scores were also higher than those of the untrained participants having comparable exposure to the text (FT’s-speed control). This suggests that untrained participants might not produce results comparable to FT by simply adopting different strategies to read as fast as FT. After finishing all the sessions, FT reported that she understood the stories as a whole and gave details about what types of questions she found difficult to answer; for example, those on the specific characteristics of a person in one story. Thus, with reading speed much faster than those of all the untrained participants the trained individual (FT) maintained levels of sentence comprehension close to those of untrained participants, while the questions reflected both the gist and the detail of the stories. FT’s performance thus appears to show reduced speed-accuracy trade-off as observed in Study 1.
Analyses of eye movements during these sessions revealed that fixation time and saccade sizes of FT were numerically not distinctively different from those of the untrained participants. However, FT tended to make saccades larger in horizontal, leftward directions rather than in vertical, downward directions typical of the untrained participants. In addition, rightward regressive saccades and return sweeps were less frequent in FT compared with the untrained participants. These fewer return sweeps, larger saccade sizes in X than one degree, and more frequent upward saccades in FT than in the untrained participants, which appears consistent with but more apparent than in the trainees of Study 1 (
Given the comprehension scores, it appears that even though FT did not direct her gaze over all characters or lines while speed-reading, she appeared to understand the content of the stories to a certain extent. It might not be, for example, that FT obtained information from only a few words on which her eyes were fixated during the first pass: such strategies might not have produced her comprehension scores, as the lower scores by FT’s-speed control participants suggest. The Park-Sasaki method teaches looking strategies in which the trainee shifts one’s attention ahead while never feeling the movement of one’s own eyes. Consequently, the trainee learns to move the eyes horizontally while seemingly viewing all the characters in order. The present results thus may support the possibility that at least one expert might have developed such looking strategies through training to a level at which she can grasp the content of sentences to a certain extent.
The present study examined relationships between reading speed, scores on comprehension tests and eye-movement measures while native participants read Japanese contemporary novels. In addition to normal untrained participants, four trainees of the Park-Sasaki speed-reading method were included in the two case studies. In Study 1, the Park-Sasaki trainees showed poor sentence comprehension performance influenced by the speed-accuracy trade-off. In addition, for normal untrained readers, multiple statistically significant correlations were revealed between reading speed, comprehension scores, residual value from linear regression in the speed-accuracy plot, and eye movement measures. In Study 2, one trained speed reader (FT) showed comprehension scores approaching those of the untrained participants while reading faster. This single case suggested the possibility of reduced speed-accuracy trade-off. FT’s eye-movements appeared different from those of the untrained participants in that she mainly made saccades in horizontal directions while looking at limited parts of the sentence display with little re-reading, scarcely moving her eyes in the vertical direction as did the untrained readers.
Some of the the conclusions from the present studies can be discussed in relation to previous work in English. The speed-accuracy trade-off as found in Study 1 is consistent with other previous works
Regarding Study 2, numerically similar comprehension scores between the trained individual (FT) the and untrained participants in Study 2 might be somewhat better than Just and Carpenter
There appear to remain a series of further issues that need to be carefully considered or addressed in the future, before making better interpretation of the data and conclusive remarks regarding efficacy of the Park-Sasaki method. First, small sample sizes in the present study clearly impose limitations on the conclusions drawn about speed-reading. That is, conclusions related to speed-reading based on the present study require particularly cautious considerations given that they mainly derive from numerical differences or non-parametric tests. Even though the speed-accuracy trader-off appeared to be reduced for the trained individual in Study 2, this single case could be insufficient for researchers to conclude that the trade-off is overcome in high-level or top-level Park-Sasaki experts. More trained participants who did not participate need to be included in the future when they have achieved sufficiently high levels of expertise, so that the present results could be generalized to a larger population. There were also age differences between the trainees and the untrained participants, in particular in Study 2. It would thus be more ideal in the future to control the ages of these participants between groups.
Second, what types of looking and reading strategies the Park-Sasaki trainees used has to be cautiously considered. The pattern of eye movements shown by the trained individual in Study 2 (
Third, a major problem in comparing the trainees and the untrained participants is that there may well be underlying differences between these participants that are not simply associated with the training procedure: (1) Those who self-selected themselves for training may be generally more highly motivated to read efficiently. (2) Those who are successful in the training process and hence choose to continue may have individual differences that make them more suited to this training method. For example, they may have particularly good parafoveal vision or they may be particularly skilled readers from before training, so that they would be better able to accurately identify words further from the fixation point that are visually degraded. (3) Those who select themselves for training may simply benefit from having more reading practice. To clarify these issues, further to comparing advanced experts and untrained participants, it would be required to randomly assign naïve participants to training and no-training groups, and make comparisons between pre- and post-training. Also, another potential approach could be to employ within-participants designs with the same readers taking either a speed-reading or normal-reading strategies, hence controlling for other variables including reading practice, ability to use heuristics, age, and so on.
Fourth, the nature of comprehension being assessed in this study needs to be carefully considered. In the present task, the comprehension questions asked about the plot of the stories. This appears appropriate when examining comprehension of the context in reading short contemporary novels. Ratings on the questions by gist/detail control participants suggested that they reflected both the gist and the detail of the stories. The extent to which the questions reflected the gist/detail of the stories did not seem to have effect on comprehension. However, it should be noted that the degree of comprehension is specific to the tasks used, and thus much further investigations are required before concluding that comprehension levels in the trained individual in Study 2 were the same regardless of reading speed. Actually, comprehension scores of FT were somewhat lower, though not statistically significantly, than those of the untrained participants. In addition, as described in the Introduction, the present study used relatively easy text that does not require specialized knowledge/schema on specific areas. Thus, the present results may apply only to comparable types of text and reading strategies. Other types of reading (e.g. comprehension during student study) may require more complex comprehension, such as integration of the contexts with prior knowledge or understanding of complex concepts. It would surely be required to further modify comprehension questions by using more various types of text, so that different types of comprehension could be separately examined. In addition, there is so far no clear evidence that any participants including the trainees could comprehend the details of non-fixated text. This also needs to be examined using modified procedures.
Fifth, even though the instructed reading strategies changed across Studies 1 and 2, little difference in the reading speed, comprehension, or eye-movement measures was found for the untrained participants. It is thus possible that the change in instructions had little impact on the reading strategies. However, this point remains unclear for the trainees in Studies 1 and 2, whose levels of expertise were different. It would thus be necessary to further examine how trained participants may perform under different instructions.
The fact that FT appeared to understand the text without gazing over the whole sentence areas suggests the possibility that effective perceptual span may have been widened through the course of training. This seems consistent with the preliminary finding that performance on visual search tasks including conjunction search tasks was better in the trainees on this method than in the untrained participants
From a methodological perspective, the present research made use of Tobii eye-tracking system in a reading situation. The method can be safely used not only for adults but also for children
To summarize, in addition to finding correlations between reading speed and comprehension scores and eye-movement measures in reading Japanese contemporary novels by normal untrained readers, we also obtained cases by trainees of a speed-reading method regarding relationships between these variables. The trainees overall showed poor performance influenced by the speed-accuracy trade-off, although this trade-off appeared to be reduced in the case of one high-level expert, who used horizontal eye-movements. Effective use of eye-tracking systems in computer-assisted settings would allow for potential applications in reading studies as well as in studies of speed-reading. By accumulating both behavioral and neurocognitive evidence by using various experimental tasks and modified procedures, it would be possible to further elucidate the potentials of humans’ cognitive capabilities that may be achieved through meditation-based visual training.
We would like to thank Sayaka Ishii and Soichiro Matsuda for their assistance in conducting the studies.