It is well known that people who read print or braille sometimes make eye or finger movements against the reading direction. The way these regressions are elicited has been studied in detail by manipulating linguistic aspects of the reading material. Actually, it has been shown that reducing the physical intensity or clarity of the visual input signal can also lead to increased regressions during reading. We asked whether the same might be true in the haptic realm while reading braille. We set the height of braille dots at three different levels (high, medium, and low) and asked adult blind, practiced braille readers to read standardized texts without any repetition of content. The results show that setting the braille dot height near the tactile threshold significantly increased the frequency of regressive finger movements. Additionally, at the lowest braille dot height, braille reading speed significantly diminished. These effects did not occur at braille dot heights that were closer to the height of standard braille (medium and high). We tentatively conclude that this effect may be due to a heightened sense of uncertainty elicited by perception near the threshold that seems to be common to the reading process, independent of the sensory input modality. Furthermore, the described effect may be a feature of a brain area that contributes to the reading process mediated by vision as well as touch.
Citation: Lei D, Stepien-Bernabe NN, Morash VS, MacKeben M (2019) Effect of modulating braille dot height on reading regressions. PLoS ONE 14(4): e0214799. https://doi.org/10.1371/journal.pone.0214799
Editor: Koji Miwa, Nagoya University, JAPAN
Received: July 27, 2018; Accepted: March 20, 2019; Published: April 17, 2019
Copyright: © 2019 Lei et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Data Availability: All relevant data are within the paper and its Supporting Information files.
Funding: This work was supported by the National Institute on Disability, Independent Living, and Rehabilitation Research (www.acl.gov/about-acl/about-national-institute-disability-independent-living-and-rehabilitation-research) under grant numbers 90RE5008-01-00 and 90RE5024-01-00, awarded to John Brabyn; the National Institutes of Health (www.nih.gov) under award number K99EY026997 and a gift from Facebook (www.facebook.com), awarded to VM; and the National Science Foundation Graduate Research Fellowship (www.nsfgrfp.org) under award number DGE 1752814 and the Minnie F. Turner Fund for Impaired Vision Research, awarded to NSB. The content is solely the responsibility of the authors and does not necessarily represent the official views of the funding agencies. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Reading constitutes the sequential acquisition of information that can be imparted through a variety of media. In visual reading, the eyes wander across visible text by a series of fast (saccadic) eye movements mainly along the lines of text [1–4]. The saccades are interspersed by fixations, with mean duration of 275–325 ms for oral reading and 225–350 ms for silent reading, during which the actual process of acquiring information occurs [5–7].
The rich body of knowledge of the visual reading process has led to the formulation of several theoretical models of reading and eye movements. They all addressed the task of recognizing letters in the presence of noise, identifying words and making the appropriate eye movements to the next word.
Most prominent was the creation of “Mr. Chips”, based on an ideal observer strategy [8–9], while others have used Bayesian observer models that derive the inference from letter data: “E-Z Reader” [10,11], “SWIFT” [12,13], a “rational model”  (which particularly focused on the occurrence of between-word regressions). Others have also hypothesized that regressions are mainly determined by low-level visuomotor processes (e.g., word skipping due to oculomotor aiming errors) . An additional influence of different reading strategies cannot be ruled out, but it is beyond the purview of this study. For a detailed discussion of different reading and eye movement models see .
Regarding eye movements, the models consider different functional states to choose between three actions: (a) continuing to fixate a currently-fixated position, (b) initiating a saccade to a new position, or (c) stopping to read a sentence. One parameter to differentiate between these options is word and sentence comprehension. Option (b) can also include the decision whether to move the eye forward or backward to a previously read word .
In the tactile reading of braille the process is similar, only that the fingers wander continuously over a series of raised dots that are grouped to denote letters or small groups of letters; but, in this case, acquiring information happens during movement .
There is an objective aspect of the reading process that depends on the quality of the physical input signal. The aim of this study was to investigate the role of physical input quality on braille reading. Our approach addresses the fact that the signal can be degraded by physical properties, in the visual as well as the tactile realms. Visual input can be degraded by very low contrast; tactile input can be degraded by reduced braille dot heights.
There is also a subjective and individual aspect that determines the quality of the sensory input by a change of sensitivity. Visual sensitivity can be diminished by an eye disease, tactile sensitivity by age. Both these components affect the ability of the individual sensory organ to derive an unequivocal message from the signal received from the input medium. We focus here on the objective aspect of reading, because it is fundamental for solving the task and has not received enough attention.
In reading, most eye and finger movements are progressive; that is, they travel in the reading direction and, thus with the flow of grammatical and semantic content. It has long been known, however, that not all movements are progressive [18–20]. According to Rayner , 10–15% of saccades by readers of English print are regressions that lead either just a few letters back, or to a previously read word.
Additionally, it has long been known that there are several features that can provoke regressions, such as lack of reading experience (especially in children) and unfamiliarity with the vocabulary or text layout [18, 21]. Visuomotor factors have also been shown to influence regression frequency [22–24]. In general, Bayle  conducted an early eye movement analysis and recognized the importance of the quality of the input signal and postulated that regressions appear “when perception is inadequate”. Later research has confirmed these early notions [26,27] and added many linguistic aspects of the text as possible causes [28, 23, 29–30]. Indeed, linguistic manipulations of the text can have the same effect on reading braille and increase the likelihood of regressive hand movements [31, 17, 32–35]. The kinematic properties of reading hand movements and regressive hand movements are not qualitatively different .
Experiments on reading print under extreme conditions, like very low contrast, have shown that, as the quality of the physical input diminished near the sensory threshold, there is an increased uncertainty about the input signal. For instance, it has been found that reducing brightness contrast of printed text to near the contrast sensitivity threshold leads to more frequent regressive saccades . The same effect occurs if the sensory apparatus is weakened by a pathological condition, as is the case with an eye disease  or with abnormal signal processing in the brain due to amblyopia .
Considering that the effects of linguistic manipulations occur independent of medium, one might assume that the common reason underlying regressions could be that the reader feels unsure about the meaning of the recently read text . We ask, therefore, whether the role of input quality is a domain-specific characteristic of the visual system, or if this influence also exists during haptic reading. Particularly, we investigated whether diminishing the intensity of tactile input can increase the likelihood of regressive hand movements while reading braille. Consequently, the aim of this project was to perform experiments with tactile input of different physical levels of quality on adult, experienced braille readers.
There are several methods of measuring braille reading via tracking finger movements. For example, researchers have recorded finger movements using a video recorder and then coding the observed finger movements offline. More recent methods include using an infrared camera to track finger movements (for a review, see ). In this study, however, we used a relatively novel alternative method developed in our lab to record finger movements while participants read braille. Details on this finger tracking method are provided in the Procedures sections below.
Based on pilot data and evidence from previous vision research, we hypothesized that braille reading regressions could occur due to simply diminishing the physical quality of the tactile input without manipulating any linguistic variables. If true, this finding would imply that reading regressions can result from uncertainty about recently-read text without linguistic manipulations and independent of the sensory modality. This concept is explained in detail in the Discussion.
We report data from 12 blind adults (age M = 41 years, SD = 13.94, range = 22 to 70 years, eight female). Only fluent braille readers were included in this study. An additional six adults were excluded from further analysis: two because they read braille in the typical braille dot height too slowly (their reading times were more than 1.5 times the interquartile range above the third quartile (Q3), which are outliers according to the IQR outlier rule), one because of excessive reading mistakes, one because of a hand injury that impacted braille reading, and two because of experiment error. All participants had begun receiving braille instruction between the ages of 4–8 years. Participants responded to the Edinburgh Handedness Inventory , a questionnaire that assesses hand preference for everyday items or tasks. Their handedness scores and their preferred hand during one-handed braille reading were unrelated (Kendall’s Tau-b = 0.04, p = .936).
There are many ways one can read braille text; individuals may use one hand or two hands. Additionally, in individuals who habitually read braille with two hands, there are a variety of movement patterns they may use, e.g., reading with both index fingers side by side across the entire page, or reading part of the line with the left hand, and part of the line of text with the right hand . All but one of our participants preferred to read braille using two hands. For these experiments, however, we asked participants to use only one hand, although this may have reduced the reading speed of habitual two-handed braille readers. In the discussion of the experimental results, it will be inevitable to make comparisons between reading braille and reading print. Any such comparison would be unfair if we juxtaposed reading print by two-handed braille reading, since the latter allows independent hand movements, while reading print does not include independent movements of the eyes. Hence, this report only discusses braille reading movements observed during one-handed reading of braille.
All participants came from the San Francisco Bay Area and were recruited through an email list and word-of-mouth. They received monetary compensation for their time. The Smith-Kettlewell Eye Research Institute Institutional Review Board approved the experimental procedures, and all participants provided informed written consent prior to their enrollment.
To control for linguistic complexity as a potential factor for causing regressions, we ensured that texts were of the same reading level and of similar length and syntactic complexity by using the International Reading Speed Texts (IReSTs), a set of standardized texts designed for assessing reading performance in teenagers and adults . These texts allowed us to control for linguistic complexity as a potential factor for causing regressions, since they were intentionally designed to have the same reading ability level (sixth grade), similar length (830 ± 2 characters), high frequency words, and varied but comparable linguistic complexity [43, 44].
Six English IReST paragraphs were used as experimental stimuli (titled “mice”, “beaver”, “prey”, “trees”, “desert”, and “island”), and one was selected for a practice trial (titled “colors”). The texts received minor edits to conform to American English spelling and word usage, and they were translated into American English contracted braille. Contracted braille is not a direct letter-to-symbol mapping with printed text. The most important difference is that braille symbols and symbol groups are used to replace a mix of single letters and punctuation, parts of words (e.g., “ea” and “ount”), and whole words (e.g., “enough” and “were”). Spaces between some words are removed in contracted braille. Due to the larger footprint of braille, an IReST paragraph occupied a full 11.5 x 11 inch (29.21 x 27.94 cm) standard braille page. Each sheet was used only once, so that repeated use would not lower the height of the dots.
Each text was printed on a standard braille sheet using a VP Max embosser (ViewPlus Technologies, Inc.). While the Library of Congress  recommends that braille dots on embossed paper be 0.48 mm tall, the maximum and default braille dot height printed by the VP Max embosser was 0.38 mm, which was used as the “high” condition. Because the high braille dot height is the default height of the braille embosser we used, this can also be considered the “standard” height for braille embossed on paper. Three braille dot heights were printed by the VP Max embosser by manipulating the “opacity” setting in Adobe Photoshop (Adobe Systems Inc., San Jose, CA). Braille dots were printed in one of three heights: high M = 0.38 mm (SE = 0.01 mm), medium M = 0.18 mm (SE = 0.01 mm), or low M = 0.04 mm (SE = 0.01 mm). These values were determined by 18 measurements of each height, taken from used and unused stimuli on different positions on the braille page (neither variable statistically affected heights), using a 10x magnifying reticle.
For the duration of the experimental procedures, participants were seated at a table on which braille sheets were presented in a fixed location (Fig 1). A microphone above the table captured their verbal responses. Participants had electromagnetic sensors taped to the tops of their index and middle fingernails, which tracked finger positions at 240 Hz (3DG trakSTAR system by Northern Digital Inc.; 2.0 mm sensors with 1.4 mm root mean square position error). They were asked to read the braille texts aloud and were informed that the texts would vary in braille dot height. The decision to have the participants read the texts aloud was made to help with detecting and recognizing reading errors. Indeed, reading visual print aloud invokes more frequent regressions than silent reading , which may also be true for reading braille.
Participants completed one practice trial with high braille dots to familiarize them with the procedures. They were allowed to read using the hand and fingers they preferred. Following the practice trial, they read three texts with one hand. Assignment of text to braille dot height condition was pseudorandom, with the constraint that each text appeared an equal number of times in each condition across the 12 participants. Specifically, each participant read three different texts, each at a different dot height (low, medium, and high). Across the 12 participants, each text was read once at each dot height condition. The reading time for a page of text ranged from 66 to 206 seconds (equivalent to 136 and 44 words per minute).
Index finger movements of the reading hand were labeled either as forward movements, backward movements, or line returns. Forward movements are left to right movements, indicating reading movements. Backward movements are right to left movements to a previously-read word, or to a word on the previous line. Unlike visual reading of text characterized by a series of fixations and saccades, braille reading is characterized by continuous hand movements and is typically exhaustive reading of all characters on the page . Hence, backward movements of at least a quarter of an inch, or the width of a braille cell, were labeled as regressions. A regression from line n to a previous line (n-1) was labeled as a regression from line n.
All analyses were conducted in R (The R Foundation for Statistical Computing, Vienna, Austria, version 3.3.3). Specifically, the lme4 package  was used for the general and generalized mixed effects modeling, and the emmeans package  was used for the post-hoc pairwise comparisons of the estimated marginal means. All post-hoc tests were corrected for multiple comparisons using the Holm-Bonferroni procedure for the 95% confidence level adjustment and p-value adjustment .
Average reading rates in words per minute were calculated as the number of words in each text divided by the participants’ reading times in minutes. Because of the wide age span of our cohort (22 to 70 years), we had to consider the possibility that there may be differences in reading speed based on the documented decline of tactile sensitivity with age . In our sample, however, age and reading speed were not significantly correlated (Pearson’s correlation r = .27, p = .398).
Otherwise, reading rates were analyzed using a general mixed effects model with random intercept effect of participant and a fixed effect of braille dot height. Likelihood ratio tests were used to compare the different models’ goodness of fit, followed by post-hoc pairwise comparisons of the estimated marginal means on braille dot height.
Unlike with reading print, where information is acquired during fixations, braille reading requires individuals to move their fingers across the raised dots to acquire information. It is possible that there will be multiple left to right movements to reread the same word. To better interpret regressions in braille reading, we differentiated between primary regressions, the first regressive movement to a particular word or a series of words, and repeat regressions, all other regressive movements to the same word or a series of words. To include the repeat regressions with the primary regressions would superficially inflate the unique number of regressions.
Accordingly, regressions were investigated using four separate metrics. The first two metrics were the number of primary regressions and the number of repeat regressions. The third was the number of times each word was read using a distinct forward (left to right) movement with a length of at least 0.25 in (6.35 mm). Forward movements were separated from one another by regressive movements of at least 0.25 in (6.35 mm). If no reading regressions were made, the number of times each word was read would be one. The fourth metric was the percentage of time spent in regressive movements during each trial. The analyses for the number of primary regressions that occurred, the number of repeat regressions that occurred, and the number of times each word was read were conducted as a general (linear) mixed effects model with a random intercept effect of participant and a fixed effect of braille dot height. Likelihood ratio tests were used to compare the different models’ goodness of fit, followed by post-hoc pairwise comparisons of the estimated marginal means on braille dot height. Analyses of the percentage of time in regressive movements were conducted as generalized mixed effects models using a Gamma distribution with a random intercept effect of participant, due to violations of normality, agreement with Gamma model assumptions, and better fit based on the Akaike Information Criterion (AIC).
Finger movement speeds were compared for forward and regressive movements using a generalized mixed effects model using a Gamma distribution (due to violations of normality) with a random intercept effect of participant and fixed effects of movement direction (forward and regressive) and braille dot height. Likelihood ratio tests were used to examine which fixed effects should be included in the model and whether there was an interaction between the fixed effects, followed by a post-hoc pairwise comparison of the estimated marginal means on movement direction and braille dot height.
Reading text embossed in medium and high braille dots was faster than reading low height braille dots. The participants’ average reading rates (Fig 2) were normally distributed and were significantly affected by braille dot height, χ2(2) = 85.86, p < .001, (see Table 1 for a statistics table of the fixed effects of the model). Post-hoc pairwise estimated marginal means analysis revealed significant differences in reading rates for low and medium braille dot heights (p < .0001, 95% CI [-23.97, -11.69]) and low and high braille dot heights (p < .0001, 95% CI [-25.12, -12.85]), but no significant difference between medium and high braille dot heights (p = .635, 95% CI [-7.29, 4.99]).
We found that 24.54% of all regressive movements occurred in high braille dot height, 29.04% in medium height, and 46.42% in low height. Across all regressive movements, 57.20% were within a word, 35.88% were within a line beyond a word, and 1.53% were to a previous line. On average, 1–4% of regressions were to a previously-read line (low: M = 2.09%, SEM = 0.98%; medium: M = 1.66%, SEM = 0.62%; high: M = 3.29%, SEM = 1.15%). The fixed effects statistics table for each model related to regressive movements are reported in Table 1.
Primary and repeat regressions.
Some regressions (5.63%) began at an empty space in the text (rather than a braille cell), so that it was difficult to determine the start and end word of these regressions. Thus, the number of primary regressions was determined after excluding those regressions that began at an empty space. Of the remaining regressions, 71.61% were primary regressions and 28.49% were repeat regressions. Primary regressions and repeat regressions had a similar proportion of within-word regressions (61.55% and 58.25%, respectively). The average number of primary regressions were M = 30.67, SEM = 3.69; M = 21.17, SEM = 2.81; and M = 18.17, SEM = 3 for low, medium, and high braille dot heights respectively. The average number of repeat regressions for each braille dot height condition were low: M = 14.17, SEM = 3.2; medium: M = 7.5, SEM = 1.94; and high: M = 6.08, SEM = 1.37.
The number of primary regressions (Fig 3) was significantly affected by braille dot height, χ2(2) = 35.02, p < .0001. Post-hoc pairwise estimated marginal means comparisons indicated that the number of primary regressions was significantly different between low and medium braille dot heights (p = .001, 95% CI [3.78, 15.21]) and between low and high braille dot heights (p < .0001, 95% CI [6.79, 18.21]), but not significantly different between medium and high braille dot heights (p = .188, 95% CI [-2.71, 8.71]).
The number of repeat regressions (Fig 4) was significantly affected by braille dot height, χ2(2) = 16.76, p < .001. Post-hoc pairwise estimated marginal means comparisons indicated that the number of repeat regressions was significantly different between low and medium braille dot heights (p = .009, 95% CI [1.20, 12.13]) and between low and high braille dot heights (p = .003, 95% CI [2.62, 13.55]), but not significantly different between medium and high braille dot heights (p = .509, 95% CI [-4.05, 6.88]).
Number of times each word was read.
The average number of times each word was read (Fig 5) with a distinct forward movement was significantly associated with braille dot height χ2(2) = 18.089, p < .001. Post-hoc pairwise estimated marginal means comparisons revealed significant differences between low and medium braille (p = .010, 95% CI [0.25, 0.21]) and low and high braille (p = .002, 95% CI [0.05, 0.23]), but no significant difference between medium and high braille (p = .807, 95% CI = [-0.068, 0.11]).
Percentage of time rereading.
The average percentage of time spent in regressive movements were 11.91% (1.11% SEM), 9.06% (0.73% SEM), and 9.46% (1.43% SEM) for low, medium, and high braille dots respectively. Although there was an increased percentage of time for low braille dot heights compared to the medium and low braille dot heights, the average percent of time spent in regressive movements were not statistically associated with braille dot height χ2(2) = 5.05, p = .080.
The average finger movement speeds (Fig 6) were significantly related to movement direction χ2(1) = 172.21, p < .001; and were significantly associated with braille dot height χ2(2) = 19.52, p < .001. However, they were not associated with the interaction of movement direction and braille dot height χ2(2) = 0.28, p = .871. Post-hoc pairwise comparisons indicated that for each braille dot height condition, backward movement speeds were significantly faster than forward movement speeds (all dot heights: p < .0001, 95% CI [0.111, 0.151]). Furthermore, when controlling for movement direction, post-hoc pairwise comparisons indicated that finger movement speeds were significantly different between low and medium braille dot heights (p = .0002, 95% CI [0.019, 0.074]) and between low and high braille dot heights (p = .0002, 95% CI [0.017, 0.073]), but not between medium and high braille dot heights (p = .880, 95% CI [-0.027, 0.024]). On average, backwards movements were 2.77 cm/s faster than forward movements.
In sum, for the low braille dot height condition, we found significantly slower reading rates, significantly more primary regressions, significantly more repeat regressions, and a significant increase in the number of times each word was read compared to medium and high braille dot height. However, there was no significant difference between medium and high braille dot height conditions for those four measures (reading rates, number of primary regressions, number of repeat regressions, and the number of times each word was read).
An interesting question that follows this study is what the tactile threshold value is for humans reading braille. Unfortunately, we could not find a systematic psychophysical determination of the tactile threshold for braille-like dots. For a single dot, it is in the range of only a few microns ; however; sensitivity to a single dot is not what is required for reading braille. Instead, we need to consider that lateral interactions with the sensation of adjacent dots are likely to occur while reading braille, especially as the sensing finger is moved.
The tactile height of the dots in the high condition was 0.38 mm, which is near the recommended standard braille dot height and can be assumed to be above the threshold. This assumption is confirmed by the fact that the high dots could be easily read by all of our participants. Our medium dots of 0.18 mm were obviously not near the threshold, since our results were hardly different than the ones obtained in the high dot condition. This height has also been shown to be easily read by many braille readers (71%) .
Dots in the low height condition, however, were 0.04 mm, which is likely to be closer to the tactile threshold. This produced an increased frequency of regressive finger movements and diminished reading rate. This finding is confirmed by the fact that braille dots of 0.037 mm height printed with a UV ink-jet layering technique can still be read by blind and visually-impaired individuals .
It is important for us to relate our stimuli to the threshold height in the presence of adjacent dots. Note that we need to know the critical tactile height of the dots, not the “tactile acuity” as described by , which determines the minimum separable distance between two adjacent dots for them to be perceptually distinct. “Tactile acuity” is not relevant in the context of this study, since braille dots are always an equal distance away from each other in one cell (2.3 mm, a standard set by the Library of Congress ). This is well-adjusted to the finding that the maximum response of mechano-receptive afferents to dot arrays is highest between 2 and 3 mm spacing for the two most important tactile receptor types .
Lateral interactions also play a role in experiments testing tactile “hyperacuity”, where the lateral displacement of the middle one in a row of three dots must be detected. If no finger movements are allowed, the threshold is very high, at a mean of 0.31 mm (± 0.05 SEM) . If the finger pad is moved over a stationary dot, however, the threshold drops considerably . This condition is relevant here, because we can assume that there will always be finger pad movements while reading braille as in our experiments.
In summary, we conclude that our low braille dot height was close enough to the tactile threshold to have an effect on reading rate and frequency of regressions, but not below the threshold, since our participants could still accurately read the braille text. Our results showed that regressive finger movements during braille reading can be elicited by low levels of stimulus intensity, the same way that has been shown for visual input and reading print.
It is a well-documented fact that regressions occur in the realm of reading print as well as reading braille. Due to the incompatibilities in the physical and physiological mechanisms, the steps involved require a detailed examination to pinpoint commonalities and differences.
Reading is a dynamic process that depends on the intention to acquire information by comprehending text. This requires movement from beginning to end of the text in both media. In the case of reading print, the eye progresses through the text with saccadic eye movements, while the information is acquired during the holding positions between the fast, step-like movements. In the case of reading braille, the information is continuously acquired while the reading finger slides relatively slowly and continuously over raised dots that are arranged in braille cells.
Reading braille with one’s fingertips is an involved task that requires movement of one’s entire hand across the page. Braille readers are aware of their regressive movements, since they can be admonished to avoid them . However, that might not be the case for the fast regressive eye movements during reading print. Given the decision to continue reading, the process is perpetuated until the goal is reached. Nevertheless, a fundamental question that must be addressed repeatedly in both media, which is determining which direction the next movement is to go. If the reader is sure enough that the information is sufficiently clear to allow an unequivocal interpretation of the content, the movement will progress in the reading direction. Conversely, if the reader is unsure about the content, the next movement can go in the other direction to re-examine input that has been read before.
The functional role of this point in time has been recognized as fundamental and has received much attention in the published models of the reading process. This strategic point has been described as a moment of “uncertainty” [8, 40, 14, 59–60] or as “ambiguity” [61, 34, 8–9, 62, 59]. This point in time can be seen as an implementation of an informational threshold that determines whether the ultimate tool of resolving it should be used to reduce uncertainty (or entropy) to zero .
The literature shows that regressive eye movements that occur when reading print can be influenced by linguistic elements of the reading material (word frequency, word length, language complexity; see [63, 64], visuomotor processes , as well as the physical quality of the input signal in reading print [8, 37–38]). What the current study adds to the existing knowledge is the fact that the physical quality of the haptic input in reading braille also influences the occurrence of regressions. In summary, this means that the special state of “uncertainty” can be evoked by diminished input quality alone and, thus, trigger the generation of regressive movements during reading in the visual and tactile realms.
As uncertainty obviously plays an important role in reading, it is tempting to speculate whether there is a region in the human brain that could support this kind of decision-making. The critical feature would have to be that it is involved in reading and independent of the input stimulus modality. Such an area has been described as the VWFA (visual word form area), based on imaging experiments on the human brain [65, 66]. This area in the ventral visual pathway is multimodal, receives input from the visual and tactile systems, and is activated when reading in either input modality. Experiments using transcranial magnetic stimulation (TMS) have shown that disruption of VWFA activity can decrease tactile reading accuracy in sighted braille readers  and lesions in that area cause pure alexia .
It is important to note that three measures of tactile sensitivity (tactile acuity [gap width], line length, and orientation) show a steady decline with age . This opens the possibility that braille reading in our participants might be affected accordingly, especially given their considerable age range (22 to 70 years). The fact that we did not see an age effect in our cohort can be explained by our use of standardized braille cells that are configured at a size that is far above the thresholds measured in the study by Stevens et al. . A further point to consider is the fact that blind braille readers, like all of our participants, have been found to show higher tactile sensitivity than sighted control subjects . Since we did not conduct any comparisons with sighted subjects, this is a concern we can disregard.
To conclude, the results of our experiment demonstrate that the physical characteristics of tactile input can influence behavioral aspects of reading braille. Since this parallels previous findings on reading print, it is surprising that the role of input strength in the generation of reading regressions has not found a place in unified theories of reading [69, 70, 71]. This is relevant, because a role of regressions during reading was hypothesized early on by Bayle , who concluded from an eye movement analysis that they “serve a definite purpose in reading”. A recent series of advanced experiments has demonstrated that regressive movements made while reading print are not only normal and acceptable, but actually augment text comprehension [27, 72]. The present study has increased the likelihood that regressive hand movements while reading braille are likely to serve the same function.
Future studies should investigate the influence of systematic decreases of dot height near the braille reading threshold to further assess regressive movements and their influence on reading comprehension. Taken together with an equivalent study of reading print, this could help to further elucidate those aspects of the reading process that are independent of input modality.
S1 Data. Participants’ demographics (sheet titled “demographics”) and participants’ reading rates, regression counts, reading times, and movement speeds at the three dot heights (sheet titled “data”).
The authors would like to thank Anita Tung for help with the illustration (Fig 1). They are also indebted to two reviewers, for their careful reading and several suggestions how to improve the quality of this publication.
- 1. Javal E. Essai sur la physiologie de la lecture. Annales d'Ocilistique. 1878;80:61–73.
Hering E. Der Raumsinn und die Bewegungen des Auges. In: Hermann L, editor. Handbuch der Physiologie, Vol. 3, Part 1 Physiologie des Gesichtssinnes. Leipzig: Vogel; 1879. p. 343–601.
- 3. Lamare M. Des mouvements des yeux dans la lecture. Bulletins et Mémoires de la Société Française d’Ophthalmologie 1892;10:354–364.
Huey EB. The psychology and pedagogy of reading. The Macmillan Company; 1908.
- 5. O’Regan JK. The control of saccade size and fixation duration in reading: The limits of linguistic control. Percept Psychophys. 1980 Mar 1;28(2):112–7. pmid:7432983
- 6. Rayner K. Eye movements in reading and information processing: 20 years of research. Psychol Bull. 1998 Nov;124(3):372. pmid:9849112
- 7. Rayner K. Eye movements and attention in reading, scene perception, and visual search. Q J Exp Psychol. 2009 Aug 1;62(8):1457–506.
- 8. Legge GE, Klitz TS, Tjan BS. Mr. Chips: an ideal-observer model of reading. Psychol Rev. 1997 Jul;104(3):524–53. Review. pmid:9243963
- 9. Legge GE, Hooven TA, Klitz TS, Mansfield JS, Tjan BS. Mr. Chips 2002: new insights from an ideal-observer model of reading. Vision Res 2002. 42: 2219–2234. pmid:12207981
- 10. Reichle ED, Pollatsek A, Fisher DL, & Rayner K. Toward a model of eye movement control in reading. Psychol. Rev. 1998, 105,125–157. pmid:9450374
- 11. Reichle ED, Pollatsek A, Rayner K. E-Z Reader: A cognitive-control, serial-attention model of eye-movement behavior during reading. Cogn. Syst. Res. 2006;7(1):4–22.
- 12. Engbert R, Longtin A, Kliegl R. A dynamical model of saccade generation in reading based on spatially distributed lexical processing. Vision Res. 2002 Mar 1;42(5):621–36. pmid:11853779
- 13. Engbert R, Nuthmann A, Richter EM, Kliegl R. SWIFT: a dynamical model of saccade generation during reading. Psychol. Rev. 2005 Oct;112(4):777. pmid:16262468
Bicknell K, Levy R. A rational model of eye movement control in reading. In Proceedings of the 48th annual meeting of the Association for Computational Linguistics 2010 Jul 11 (pp. 1168–1178). Association for Computational Linguistics.
Vitu F. Visual extraction processes and regressive saccades in reading. In Cognitive processes in eye guidance. 2005 Jul 28:1–32. Oxford University Press.
- 16. Rayner K. Eye movements in reading: Models and data. J Eye Mov Res. 2009 Apr 3;2(5):1. pmid:20664810
- 17. Bertelson P, Mousty P, D'alimonte G. A study of braille reading: 2. Patterns of hand activity in one-handed and two-handed reading. Q J Exp Psychol A. 1985 May;37(2):235–56. pmid:4023274
- 18. Dearborn WF. The psychology of reading. 1906.
- 19. Dodge R. Visual motor functions. Psychol Bull. 1916 Nov;13(11):421.
- 20. Vernon MD. The movements of the eyes in reading. Br J Ophthalmol. 1928 Mar;12(3):130. pmid:18168705
Schmidt WA. An experimental study in the psychology of reading [dissertation]. Chicago (IL): The University of Chicago Libraries Chicago; 1917.
- 22. Andriessen JJ, De Voogd AH. Analysis of eye movement patterns in silent reading. IPO Annual Progress Report. 1973;8:29–34.
- 23. Vitu F, McConkie GW. Regressive saccades and word perception in adult reading. In: Reading as a perceptual process. 2000. p. 301–326.
Vitu F, McConkie GW, Zola D. About regressive saccades in reading and their relation to word identification. In Eye guidance in reading and scene perception. 1998 Jan 1 (pp. 101–124). Elsevier Science Ltd.
- 25. Bayle E. The nature and causes of regressive movements in reading. The Journal of Experimental Education. 1942 Sep 1;11(1):16–36.
- 26. Bicknell K, Levy R. Why readers regress to previous words: A statistical analysis. In Proceedings of the Annual Meeting of the Cognitive Science Society; 2011 Jan 1; Vol. 33.
- 27. Booth RW, Weger UW. The function of regressions in reading: Backward eye movements allow rereading. Mem Cognit. 2013 Jan 1;41(1):82–97. pmid:22886737
- 28. Frazier L, Rayner K. Making and correcting errors during sentence comprehension: Eye movements in the analysis of structurally ambiguous sentences. Cogn Psychol. 1982 Apr 1;14(2):178–210.
- 29. Meseguer E, Carreiras M, Clifton C. Overt reanalysis strategies and eye movements during the reading of mild garden path sentences. Mem Cognit. 2002 Jun 1;30(4):551–61. pmid:12184556
- 30. Staub A. Eye movements and processing difficulty in object relative clauses. Cognition. 2010 Jul 1;116(1):71–86. pmid:20427040
- 31. Davidson PW, Appelle S, Haber RN. Haptic scanning of braille cells by low-and high-proficiency blind readers. Res Dev Disabil. 1992 Jan 1;13(2):99–111. pmid:1574626
- 32. Hislop DW, Zuber BL, Trimble JL. Text-scannning patterns of blind readers using Optacon and braille. J Rehabil Res Dev. 1985 Jul:54–65. pmid:3835265
- 33. Davidson PW, Appelle S, Haber RN. Haptic scanning of braille cells by low-and high-proficiency blind readers. Res Dev Disabil. 1992 Jan 1;13(2):99–111. pmid:1574626
- 34. Mousty P, Bertelson P. Finger movements in braille reading: The effect of local ambiguity. Cognition. 1992 Jan 1;43(1):67–84. pmid:1591903
- 35. Wright T, Wormsley DP, Kamei-Hannan C. Hand movements and braille reading efficiency: Data from the Alphabetic Braille and Contracted Braille Study. J Vis Impair Blind. 2009 Oct 1;103(10):649–61.
- 36. Hughes B, McClelland A, Henare D. On the nonsmooth, nonconstant velocity of braille reading and reversals. Scientific Studies of Reading. 2014 Mar 4;18(2):94–113.
- 37. Legge GE, Ahn SJ, Klitz TS, Luebker A. Psychophysics of reading—XVI. The visual span in normal and low vision. Vision Res. 1997 Jul 1;37(14):1999–2010. pmid:9274784
- 38. Bullimore MA, Bailey IL. Reading and eye movements in age-related maculopathy. Optom Vis Sci. 1995 Feb;72(2):125–38. pmid:7753526
- 39. Kanonidou E, Proudlock FA, Gottlob I. Reading strategies in mild to moderate strabismic amblyopia: an eye movement investigation. Invest Ophthalmol Vis Sci. 2010 Jul 1;51(7):3502–8. pmid:20207968
- 40. Levy R, Bicknell K, Slattery T, Rayner K. Eye movement evidence that readers maintain and act on uncertainty about past linguistic input. Proceedings of the National Academy of Sciences. 2009 Dec 15;106(50):21086–90.
- 41. Marcet A, Jiménez M, Perea M. Why braille reading is important and how to study it/Por qué es importante la lectura en braille y cómo estudiarla. Cultura y Educación. 2016 Oct 1;28(4):811–25.
- 42. Oldfield RC. The assessment and analysis of handedness: the Edinburgh inventory. Neuropsychologia. 1971 Mar 1;9(1):97–113. pmid:5146491
- 43. Hahn GA, Penka D, Gehrlich C, Messias A, Weismann M, Hyvärinen L, et al. New standardised texts for assessing reading performance in four European languages. Br J Ophthalmol. 2006 Apr 1;90(4):480–4. pmid:16547331
- 44. Gibson E. Linguistic complexity: Locality of syntactic dependencies. Cognition. 1998 Aug 1;68(1):1–76. pmid:9775516
- 45. The Library of Congress. National Library Service for the Blind and Physically Handicapped. 2008. Available from: http://www.loc.gov/nlsold/specs/800_march5_2008.pdf
- 46. Rayner K, Pollatsek A, Ashby J, Clifton Jr C. Psychology of reading. Psychology Press; 2012 May 22.
- 47. Bates D, Maechler M, Bolker B, Walker S. lme4: Linear mixed-effects models using Eigen and S4. R package version. 2014 Jan;1(7):1–23.
- 48. Lenth R. V. (2018) emmeans: Estimated Marginal Means, aka Least-Squares Means. R package version 1.1.3. Available from: https://CRAN.R-project.org/package=emmeans
- 49. Holm S. A simple sequentially rejective multiple test procedure. Scandinavian journal of statistics. 1979 Jan 1:65–70.
- 50. Stevens JC, Foulke E, Patterson MQ. Tactile acuity, aging, and braille reading in long-term blindness. J Exp Psychol Appl. 1996 Jun;2(2):91.
- 51. LaMotte RH, Whitehouse J. Tactile detection of a dot on a smooth surface: peripheral neural events. J Neurophysiol. 1986 Oct 1;56(4):1109–28. pmid:3097273
Douglas G, Weston A, Whittaker J, Wilkins SM, Robinson D. Braille dot height research: investigation of Braille dot elevation on pharmaceutical products. Final Report, University of Birmingham and RNIB. 2008.
- 53. Urbas R, Rotar B, Hajdu P, Stankovič Elesini U. Evaluation of the modified braille dots printed with the UV ink-jet technique. Journal of Graphic Engineering and Design. 2016 Jan 1;7(2):15–24.
- 54. Stevens JC. Aging and spatial acuity of touch. J Gerontol. 1992 Jan 1;47(1):35–40.
- 55. Phillips JR, Johansson RS, Johnson KO. Responses of human mechanoreceptive afferents to embossed dot arrays scanned across fingerpad skin. J Neurosci. 1992 Mar 1;12(3):827–39. pmid:1545242
- 56. Grant AC, Thiagarajah MC, Sathian K. Tactile perception in blind Braille readers: a psychophysical study of acuity and hyperacuity using gratings and dot patterns. Percept Psychophys. 2000 Jan 1;62(2):301–12. pmid:10723209
- 57. Johansson RS, LaMotte RH. Tactile detection thresholds for a single asperity on an otherwise smooth surface. Somatosens Res. 1983 Jan 1;1(1):21–31. pmid:6679912
Olson MR, Mangold SS. Guidelines and games for teaching efficient braille reading. American Foundation for the Blind; 1981.
- 59. Norris D, Kinoshita S. Reading through a noisy channel: Why there's nothing special about the perception of orthography. Psychol Rev. 2012 Jul;119(3):517. pmid:22663560
- 60. Linzen T, Jaeger TF. Uncertainty and expectation in sentence processing: Evidence from subcategorization distributions. Cogn Sci. 2016 Aug;40(6):1382–411. pmid:26286681
- 61. Rayner K, Duffy SA. Lexical complexity and fixation times in reading: Effects of word frequency, verb complexity, and lexical ambiguity. Mem Cognit. 1986 May 1;14(3):191–201. pmid:3736392
Levy R. A noisy-channel model of rational human sentence comprehension under uncertain input. In Proceedings of the conference on empirical methods in natural language processing 2008 Oct 25 (pp. 234–243). Association for Computational Linguistics.
- 63. Kliegl R, Grabner E, Rolfs M, Engbert R. Length, frequency, and predictability effects of words on eye movements in reading. Eur J Cogn Psychol. 2004 Jan 1;16(1–2):262–84.
- 64. Barton JJ, Hanif HM, Eklinder Björnström L, Hills C. The word-length effect in reading: a review. Cogn Neuropsychol. 2014 Aug 18;31(5–6):378–412. pmid:24665973
- 65. Büchel C, Price C, Friston K. A multimodal language region in the ventral visual pathway. Nature. 1998 Jul;394(6690):274. pmid:9685156
- 66. Reich L, Szwed M, Cohen L, Amedi A. A ventral visual stream reading center independent of visual experience. Curr Biol. 2011 Mar 8;21(5):363–8. pmid:21333539
- 67. Siuda-Krzywicka K, Bola Ł, Paplińska M, Sumera E, Jednoróg K, Marchewka A, et al. Massive cortical reorganization in sighted Braille readers. Elife. 2016 Mar 15;5:e10762. pmid:26976813
- 68. Dehaene S, Cohen L. The unique role of the visual word form area in reading. Trends in cognitive sciences. 2011 Jun 1;15(6):254–62. pmid:21592844
- 69. Just MA, Carpenter PA. A theory of reading: From eye fixations to comprehension. Psychol Rev. 1980 Jul;87(4):329. pmid:7413885
- 70. Sadoski M, Paivio A. Toward a unified theory of reading. Sci Stud Read. 2007 Sep 26;11(4):337–56.
- 71. Frost R. A universal approach to modeling visual word recognition and reading: Not only possible, but also inevitable. Behav Brain Sci. 2012 Oct;35(5):310–29. pmid:23251930
- 72. Schotter ER, Tran R, Rayner K. Don’t believe what you read (only once) comprehension is supported by regressions during reading. Psychol Sci. 2014 Jun;25(6):1218–26. pmid:24747167