Download Simulating Reading Struggles of Dyslexia: An Empirical Study on the Daniel Britton Font and more Study notes Psychology in PDF only on Docsity!
Running head: DYLEXIA SIMULATION FONT 1 Insights from a dyslexia simulation font: Can we simulate reading struggles of individuals with dyslexia? Z. Stark1, L. Franzen 1,2, & A. P. Johnson1,* ¹ Department of Psychology, Concordia University, Montréal, Canada (^2) Institute of Neuroscience and Psychology, University of Glasgow, Glasgow, United Kingdom (^3) CRIR/Centre de Réadaptation MAB-Mackay du CIUSSS du Centre-Ouest-de-l’Île-de-Montréal, Montréal, Canada Author Note Zoey Stark Orcid ID: 0000- 0002 - 0484 - 7722 Léon Franzen Orcid ID: 0000- 0003 - 2277 - 5408
- (^) Correspondence concerning this article should be addressed to Zoey Stark, Department of Psychology, SP244, Concordia University, 7141 Sherbrooke St. West, H4B 1R6, Canada. E-Mail: zoey.stark@mail.concordia.ca
Abstract Individuals with dyslexia struggle at explaining what it is like to have dyslexia and how they perceive letters and words differently. This led the designer Daniel Britton to create a font that aims to simulate the perceptual experience of how effortful reading can be for individuals with dyslexia (http://danielbritton.info/dyslexia). This font removes forty percent of each character stroke with the aim of increasing reading effort, and in turn empathy and understanding for individuals with dyslexia. However, its efficacy has not yet been empirically tested. In the present study, we compared participants without dyslexia reading texts in the dyslexia simulation font to a group of individuals with dyslexia reading the same texts in Times New Roman font. Results suggest that the simulation font amplifies the struggle of reading, surpassing that experienced by adults with dyslexia—as reflected in increased reading time and overall number of eye movements in the majority of typical readers reading in the simulation font. Future research could compare the performance of the Daniel Britton simulation font against a sample of beginning readers with dyslexia as well as seek to design and empirically test an adapted simulation font with an increased preserved percentage of letter strokes. Key words: Dyslexia, Type Font, Eye movements, Dyslexia Simulation, Reading
have demonstrated that visual symptoms associated with dyslexia are more prevalent among children with dyslexia (Raghuram, Gowrisankaran, & Swanson, 2018; Raghuram, Hunter, Gowrisankaran, & Waber, 2019), and can occur independent of the phonological component of reading (Bosse, Tainturier, & Valdois, 2007). One method that is well-suited for examining visual symptoms during reading is eye-tracking. Reports of different eye movements in dyslexia emerged in the 1980s (Rayner, 1983, 1985a), and have since received support from several studies with children (Bucci, Brémond-Gignac, & Kapoula, 2008; De Luca, Di Pace, Judica, Spinelli, & Zoccolotti, 1999a; Hutzler & Wimmer, 2004; Jainta & Kapoula, 2011; Razuk, Barela, Peyre, Gerard, & Bucci, 2018) and adults with dyslexia (Hawelka et al., 2010; Rayner, 1983, 1985 a; see Quercia et al., 2013, for a review). Recently, these differences were consolidated by a comprehensive eye movement profile of dyslexic adults during natural reading of standardised paragraphs showing a fundamental difference on all but one metric (Franzen, Stark, & Johnson, 202 1 ). Although this profile, and the on-going debate, suggests that visual differences may not constitute the main cause of dyslexia, eye movements provide researchers with the possibility to investigate differences in oculomotor control, visual attention span, and lexical and phonological processing alike. Based on the existing evidence eye movement differences in dyslexia, either as cause or consequence, can be considered established. In recent years, there has been a broadened awareness for individuals with dyslexia within mainstream society and media, as popular television shows have introduced characters with dyslexia such as Tiffany Doggett on Orange is the New Black and Evan Chapin from Atypical for instance. However, the idea of struggling to read can be hard for an average reader to
understand, since reading is typically a skill learned at a young age and is often taken for granted in today’s technology centred society. In an attempt to illustrate the reading difficulties associated with dyslexia, we have seen an increase in dyslexia simulations, often circulated on popular media platforms (e.g.; Refinergy29) and news cites (e.g.; CNN, Daily Mail and BBC). One well-circulated dyslexia simulation uses a webpage with a text that explains and defines dyslexia, however, the letters making up each word flip around within each word every second or so, making reading this passage strenuous (available at http://geon.github.io/programming/2016/03/03/dsxyliea). A second simulation that tries to help the general population understand how reading feels to an individual with dyslexia, and the stimulus for this study, was created by the UK graphic designer Daniel Britton. Here, Daniel Britton created a font type which removes forty percent of each character stroke with the aim of slowing down reading speed and increasing frustration among typical readers to similar levels experienced by individuals with dyslexia (n.d.; examples and description available at https://danielbritton.info/dyslexia). This font is a sans serif, roman, non- italicised font, every letter of which is capitalised. For simplicity reasons, we will refer to the Daniel Britton font henceforth as the simulation font. Effects of type fonts in dyslexia is a well-researched topic, although findings vary (Bernard, Chaparro, Mills, & Halcomb, 2003; Galliussi, Perondi, Chia, Gerbino, & Bernardis, 2020; Marinus et al., 2016; O’Brien, Mansfield, & Legge, 2005; Rello & Baeza-Yates, 2016). One well-documented phenomenon affected by changes in type fonts as well as inter/intra letter spacing is known as the visual crowding effect (Pelli & Tillman, 2008). Generally, when letter characters are closely spaced, visual crowding can occur, making reading more difficult for
e.g., De Luca, Di Pace, Judica, Spinelli, & Zoccolotti, 1999; Rayner, 1985). Based on these reports, should the simulation font adequately simulate the reading difficulty experienced by individuals with dyslexia, then we would expect to see comparable visual reading strategies as well as behavioural patterns between groups. Methods Participants****. Data was collected from 35 individuals with dyslexia who read texts in Times New Roman font (femaleDyslexia = 23, Mean age Dyslexia = 23.70, SD Dyslexia = 2.58, range Dyslexia = 18- 46 ). As well, we recruited 82 participants without dyslexia forming the simulation font group, of which 4 5 were included in the final analysis (femaleSim = 34, Mean age Sim = 2 4. 04 , SD Sim = 5.3 4 , range Sim = 18-42). A total of 3 7 participants were excluded from the simulation group for various reasons. Nine participants were excluded as their dominant language was not English. Eight scored above 40 points on the dyslexia screening measure that was used to ensure that participants in the simulation font group had no self-reported symptoms associated with dyslexia (see Measures in the Methods section for details). Twenty participants were removed from all analyses due to low-quality eye tracking data during calibration and or other technology related issues (no eye with average error < 0.5° and max error < 1.3°). As well, due to the difficulty of the task, our offline analysis showed that some of the latter 20 participants did not attempt to read trials in full by skipping to the end of the trial early on or skipping over several lines of text. The dyslexia group was exclusively composed of individuals who had received an official diagnosis of dyslexia by a specialist prior to participation. We did not diagnose these participants again in concordance with research policies for non-clinical
research in Quebec and at Concordia University. A score on the Adult Dyslexia Checklist was obtained solely as an indication of symptoms of dyslexia at the time of participation, but not to form the dyslexia group. All participants were either current or former college or university students pursuing a higher education or university degree in Canada at the time of participation or had pursued such a degree in the past. Hence, both groups were matched on chronological age and minimum education level. All participants received either course credits or $10 as compensation for their time. See Table 1 for a description of group characteristics. The present study was conducted in Montréal, Canada. The data of the dyslexia group was collected as part of a multi-study effort, where it has previously been used to demonstrate differences in eye movements between readers with and without dyslexia reading the IReST in Times New Roman and OpenDyslexic font (Franzen, Stark, & Johnson, 202 1 ). Though, only trials presented in Times New Roman, the more frequently used font, were included in this study. An a priori power analysis for a linear multiple regression random model was conducted in G*Power (version 3.1.9.4; Faul, Erdfelder, Lang, & Buchner, 2009) with a power of 0. 95 , and estimated population parameter rho effect size of 0.5 for the alternative hypothesis and 0.2 for the null hypothesis. This power analysis yielded a total sample size of 83 participants needed for this study to detect the presumed effects. The present study has received approval by the Concordia University Human Ethics Research Committee (UHREC Certificate: 30003975).
to allow for natural reading behaviour (Schotter & Payne, 2019). In accordance with the British Dyslexia Association guidelines for dyslexia friendly written material, text was presented in roman format (i.e., not italicised or bolded). As in the IReST validation study, we made use of attention questions. These questions aim to incentivise participants to read for comprehension (Morrice et al., 2020). Language and organisation, both elements of literacy, were screened using the Adult Dyslexia Checklist (Smythe & Everatt, 2001). Items on this screening questionnaire required the respondent to rate “symptoms” of dyslexia on a 4-point Likert scale, such as problems with literacy skills, word finding and organisation (e.g., “Do you find it difficult to read words you haven't seen before?”). The scoring sheet outlines a score of 45 or more as indicative of mild to severe dyslexia symptoms (Smythe & Everatt, 2001). As stated above, to avoid including participants with even mild symptoms of dyslexia, we used a score of 40 as a cut-off for inclusion in the simulation font group. This checklist was not used to form the dyslexia group. The Wechsler Adult Scale of Intelligence Symbol Search and Coding subtests were administered to assess processing speed abilities of all participants, both of which are non- linguistic measures (WAIS; Wechsler, 2008). Reports of links between slower cognitive processing speed and reading speed in adults with dyslexia motivated their inclusion (Breznitz & Misra, 2003). The WAIS has an interscorer agreement coefficient that ranges from 0.98–0.99, and an intraclass correlation ranging from 0.91–0.97. Its internal consistency coefficient on processing speed index tasks ranges from 0.87–0.98 (Canivez, 2010). Correlations between scores on tests that measure similar constructs were in the 0.80 range on criterion-related validity measures (Schraw, 2010).
The Minnesota Read sentences (MN read) were used as the stimulus on the first day during training of the simulation font. Each sentence is comprised of 60 characters including spaces and are written at the second and third grade level (Mansfield, Atilgan, Lewis, & Legge, 2019). These sentences were chosen for the training day of the simulation font group to avoid introducing the IReST during this training, which would have removed their novelty for the testing day. Further, MN read sentences are made up of 10-14 words per sentence. For our purposes, 10 of the original 24 sentences where used (Mansfield et al., 2019). Procedure****. Participants in the simulation groups underwent slightly different procedures since data of the dyslexia group was collected as part of a multi-study dyslexia research initiative. Participants in the dyslexia group visited the lab once for testing. Due to the novelty and complexity of the simulation font, individuals in the simulation font group visited the lab on two consecutive days; once for training and once for testing. On the first day, the training day, participants in the simulation group completed the Adult Dyslexia Checklist (Smythe & Everatt, 2001) as well as the two processing speed tasks (WAIS; Wechsler, 2008). Subsequently, these participants completed a self-paced learning paradigm. This training paradigm used the MN read sentences (Mansfield et al., 2019) in which one MN read sentence was displayed in Times New Roman font on the right-hand side of the screen while the same sentence was displayed in the simulation font on the left-hand side, both in point size 24. This procedure was repeated four more times with different sentences. For the proceeding five trails, the simulation font appeared on its own. Once participants pressed the space bar, the corresponding Times New Roman text appeared adjacently, giving participants the ability to verify their reading. This was repeated another four times with different
Fig. 1. Experimental paradigm. a) Pictorial depiction of the self-paced testing paradigm. Participants controlled the paradigm by pressing the space bar. A drift correct fixation point appeared first, followed by the text and one multiple-choice question. This process was repeated nine times, with both text order and font presented at random. b) Example of text stimulus presented in Times New Roman font. Please note this is an example text comparable to the ones from the commercial reading assessment (IReST), as these texts are protected. c) Identical text to panel b presented in the Daniel Britton simulation font. The attention question was still presented in Times New Roman font to ensure unobstructed reading. d) Attention question accompanying the texts in panels b and c. Attention questions were always presented in Times New Roman font.
Apparatus. Stimuli were presented and data collected using an iMac (2011 27” i7, 16GB RAM) with an external monitor. Participants viewed stimuli on a linearised video monitor (View Sonic G225fb 21” CRT, 1024 × 768 pixel resolution, 100-Hz refresh rate). A chin rest was used to stabilise head position at a distance of 70 cm from the screen. Eye position was acquired remotely using a video-based eye movement monitor (Eyelink 1000, SR Research, Ottawa, Ontario). Calibration used a series of 9 dots across the screen, with participants needing an accuracy of < 0.5 degrees on average, with no point exceeding 1.3 degrees. Analysis****. For the purpose of this study’s analysis, we included only trials of individuals with dyslexia reading in the Times New Roman font and the simulation group reading in the simulation font. The last three of the five trials presented in the simulation font were analysed in order to give participants extended practice under experimental testing conditions with this font. Conversely, only four out of five trails were analysed for the dyslexia group due to do technical difficulties during the presentation of the text. It is important to note that the used multiple-choice questions intended to incentivize readers to read for comprehension. Further, due to a difference in the number of analyzed trials per participant, the scales of the attention questions differ by group with regard to the percentage of accurate responses. Therefore, results of these attention questions should be interpreted with caution. Eye movement data processing. Fixations and saccades were recorded at a sample rate of 1000 Hz and stored for offline analysis with Data Viewer (version 3.1.97, SR Research, 2017, Ottawa, Ontario) and MATLAB (version 2019a, The MathWorks, 2019, Natick, Massachusetts). During offline analysis, we removed the first two trials presented in the simulation font on the testing day of each participant in the simulation group from all analyses. A total of 2 89 trials
simulation group was regarded as the group receiving the experimental manipulation. A negative effect size signifies a higher number or ratio of the respective eye movement by the simulation group and a positive effect size represents evidence for the alternative. Further, to be able to quantify evidence for the null hypothesis given our data, we also computed Bayes Factors for the above effect sizes in JASP (version 0.9.2; JASP Team, 2020 ). To visualise these data and statistical results without obfuscation, we used raincloud plots in MATLAB (Allen, Poggiali, Whitaker, Marshall, & Kievit, 2019). In addition to single-trial data points and well- known boxplots, these raincloud plots illustrate the differences between groups by means of probability density functions using kernel density estimation. Results This study aims to empirically evaluate whether behavioural reading performance (i.e., reading speed—measured in words per minute—and attention to text) as well as visual reading strategy (i.e., eye movements) seen in individuals with dyslexia when reading in Times New Roman font can be simulated in typical readers using a dyslexia simulation font. Specifically, we compared the mean differences between both groups/fonts for each eye movement metric using Glass’s delta effect sizes and their 95 % confidence intervals. No outliers were removed or other corrections performed on the data (see Methods section for details). Behavioural results. Standardised mean contrast for reading speed measured in words per minute (wpm) showed that the simulation font led the simulation group to read fewer words per minute compared to our sample of individuals with dyslexia ( Median Sim= 58 wpm, Median Dyslexia = 174 wpm; 2. 48 standard deviation units in magnitude; Tables 2 and 3 ; Fig. 2a).
Out of 45 typical readers in the simulation group, seven (15.5%) of the participants reading in the simulation font did not appear to struggle beyond that experienced by individuals with dyslexia, as their median reading speed was above that of the median reading speed of the dyslexia group. However, most reading speeds in the simulation group fell below the dyslexia group’s median (38/45, 84.4%), and even below their first quartile (32/45, 71%). These results illustrate longer processing times when reading texts in the simulation font. Attention to the texts was illustrated by high performance on post-hoc multiple-choice attention questions in both groups ( M Sim = 90. 37 %, M Dyslexia = 82. 05 % correct; Table 2; Fig. 2b). Both groups demonstrated high performance, with no conclusive evidence for a difference in performance (G lass’s d = - 0. 46 , 95% CI = [-0.91, - 0. 004 ]; BF 10 = 1.17; Tables 2 and 3 ). It is important to note the difference in scales between the simulation and dyslexia groups (three versus four multiple choice questions, respectively; Fig. 2 b). Since, this leads to one correct/erroneous answer being associated with a different change in percent accuracy in each group, these results should be interpreted with caution. Further, group differences were found on the coding subtest of the WAIS-IV, however, only with anecdotal evidence supporting alternative hypothesis according to the Bayes Factor ( Glass’s d = - 0.5887, 95% CI = [-1.0545, - 0.123]; BF 10 = 1.46; Table 3). On the contrary, no significant group differences were observed on the Symbol Search subtest of the WAIS-IV as illustrated by the confidence interval of the effect size including 0 and a Bayes Factor providing insufficient evidence for the null hypothesis of no difference ( Glass’s d = - 0.3049, 95% CI = [- 0.7582, 0.1485]; BF 10 = 0.48; Table 3).
Eye movement results. While reading text in the simulation font, typical readers exhibited a larger number of fixations ( Glass’s d = - 3. 50 , 95% CI = [- 4. 37 , - 2. 63 ]; Tables 2 and 3; Fig. 3a) that were longer in duration compared to individuals with dyslexia when reading the nominally same texts in Times New Roman font (Tables 2 and 3; Fig. 3b). The standardised mean difference in median fixation duration was 2. 12 standard deviation units in magnitude ( Glass’s d = - 2.12, 95% CI = [-2.76, - 1. 49 ]; Table 3; Fig. 3b). Bayes Factors corroborated these results by showing that our data provides substantial evidence for the alternative hypothesis of a difference between groups on these fixation metrics (BF 10 = 4. 86 E+ 8 and BF 10 = 1. 57 E+ 6 , respectively; Table 3). Group differences were likewise observed on saccade related eye movements. Specifically, individuals reading in the simulation font demonstrated a larger number of saccades ( Glass’s d = - 2. 78 , 95% CI = [- 3. 52 , - 2. 04 ]; Tables 2 and 3; Fig. 3c). However, their saccades were shorter in length compared to individuals with dyslexia reading in Times New Roman font ( Glass’s d = 1. 26 , 95% CI = [0. 74 ,1. 79 ]; Tables 2 and 3; Fig. 3d). Furthermore, results from the Bayes Factor analysis underline that the present data provides substantial evidence for the alternative hypothesis of group differences on the two measures number of saccades and saccade length (BF 10 = 1. 67 E+ 8 and BF 10 = 277 .18, respectively; Table 3). Taken together, these results demonstrate a similar pattern across most of the examined eye movement metrics. A higher number of fixations and saccades was observed alongside an increase in fixation duration among typical readers reading in the simulation font. Overall, this pattern suggests, that for the majority of typical readers, the simulation font amplifies the degree of effort needed to read the text above and beyond that observed in our
sample of individuals with dyslexia. An overview of the statistics supporting these results is displayed in Table 3. Fig. 3. Results of eye movement analysis. Data of the simulation font group is depicted in red across all panels, whereas data of the dyslexia group is depicted in light blue. Dots represent single-trial data. a) Overall number of fixations of each trial. b) Median fixation duration per trial in milliseconds. c) Overall number of saccades of each trial. d) Median saccade length per trial in degrees of visual angle.
