The Link Between Vision and Reading: A Language-agnostic Window into Heterogeneity in Early Reading Development

The Link Between Vision and Reading: A Language-agnostic Window into Heterogeneity in Early Reading Development | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article The Link Between Vision and Reading: A Language-agnostic Window into Heterogeneity in Early Reading Development Mahalakshmi Ramamurthy, Klint Kanopka, Julian Siebert, Lucy Yan, and 7 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6804845/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Despite decades of research, the role of visual processing in learning to read remains contested—largely due to small, homogenous samples and the lack of reliable tools to capture the true heterogeneity of reading development. In this study, we administered theory-driven, carefully validated measures of rapid visual processing to a large, socioeconomically and linguistically diverse cohort of kindergarten and first-grade children in California public schools (N ~ 1200). These visual measures proved to be equitable —performance did not vary by home language or socioeconomic status— and independently accounted for 12–18% of the variance in reading outcomes. They were also important predictors of reading risk at year-end and a year- later. Latent profile analysis revealed subgroups invisible to traditional screeners: children with strong language but poor visual skills who later struggled to read, and children with visual strengths who outperformed expectations despite phonological weaknesses. Integrating measures of rapid visual processing into early screening offers a promising path forward, towards more equitable, personalized interventions and a deeper understanding of early reading development. Biological sciences/Psychology/Human behaviour Biological sciences/Neuroscience/Visual system/Pattern vision Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 Figure 7 Main Theories linking deficits in visual processing in individuals with dyslexia date back to the 1850s, when dyslexia was first conceived as a visual problem and was referred to as ‘congenital word blindness’ in the first case study published 1 . Ever since, myriad theories have been proposed linking deficits in visual processing to reading development 2–6 . Work by Lovegrove and colleagues in the 1980s showed that people with dyslexia have challenges in visual processing of transient stimuli 7 . Subsequent studies revealed deficits in visual sensitivity to transient and moving stimuli across a wide range of experimental conditions 8–12 . In the 1990s, physiological evidence came from Livingstone and colleagues 13 who studied the postmortem brains of five individuals with dyslexia and compared it to five Controls and reported that cell bodies in the magnocellular layers of the lateral geniculate nucleus (LGN) were fewer in terms of density and were approximately 27% smaller in dyslexic brains compared to control brains. Together, these observations led to the magnocellular theory of dyslexia 4,13 , which posits that difficulties in learning to read are the consequence of a low-level deficit in the magnocellular visual pathway, and that this deficit can be detected with either physiological responses 12–15 or psychophysical thresholds 9–11,16,17 to rapid, transient or moving stimuli. A number of studies that followed have failed to replicate the relationship between reading skills and motion processing 18–22 . Importantly, intervention studies revealed that intense training in reading neither improved visual motion sensitivity 10 nor did it hinder one’s ability to show improvements in reading 23 , leaving this as the most contested hypothesis in the field, till date. Later researchers reformulated the magnocellular hypothesis in more general terms, since the M-pathway reaches the dorsal striate cortex, implicating a more general issue with attention 24,25 . In the last decade, there has been growing interest in the visual-spatial attention hypothesis, which posits that people with dyslexia shift spatial attention more slowly than skilled readers. This hypothesis fits the “sluggish attention shift hypothesis,” which is purported to explain various other sensory deficits reported in children with dyslexia across visual and auditory domains when processing rapidly presented information 26–29 , 30 . Despite five decades of research, the field lacks clear, empirically validated visual tasks that can be reliably administered to younger children (ages 5–6) 2–6 at scale. On one hand, much of the existing literature is based on small- and homogenous- sample studies where samples are not sufficiently large and diverse to discern the contribution of different risk factors, with only a couple of studies exceeding N > 100 17,31,32 . On the other hand, there is growing consensus that reading development is heterogeneous shaped by multiple, interacting risk factors. Differences in temporal visual attention, visual crowding, motion sensitivity, and oculomotor control have all been linked to early reading outcomes, independent of language ability 33–35 . Similarly, variability in auditory processing—such as rapid auditory discrimination and temporal sensitivity—has been shown to contribute to individual differences in phonological development and reading acquisition 36,37 . These findings suggest that early reading trajectories are shaped by more than just phonological processing abilities. But capturing this heterogeneity in practice, specifically early in development, requires robust, scalable tools that can be reliably administered across diverse populations early in their reading development. A critical step towards understanding heterogeneity in early reading development is the systematic implementation of reliable and equitable screening tasks administered in school settings. Such an approach is particularly powerful because it not only reaches a large, diverse, and representative population but also enables longitudinal tracking of how early performance in sensory and cognitive tasks relates to later reading outcomes. The growing recognition that early deficits in visual processing might confer risk for dyslexia is reflected in recent dyslexia screening legislation across different states in the United States of America. Developmental dyslexia, with a global prevalence of 7–17% 2,38 substantially impacts children's educational, socioemotional, and subsequent socioeconomic outcomes 6 . This widespread occurrence across languages and cultures underscores the necessity for universal early screening tools. Reading acquisition not only represents a complex neurobiological developmental process that depends on cognitive, linguistic, and sensory neural systems⁹, but is significantly shaped by environmental and instructional context that the child experiences 39,40 . Consequently, building effective screening tools require substantial investment in developing age-, culture-, and language-appropriate 41–43 , assessment items that demonstrate no socioeconomic or linguistic bias upon implementation. Such equitable screening instruments provide significant value in elucidating the heterogeneity in reading development by facilitating the identification and characterization of distinct cognitive profiles underlying reading difficulties. Despite well-documented measurement biases in various cognitive and linguistic processes¹⁰⁻¹², the relationship between visual processing and reading acquisition remains inadequately characterized at the population level. Over the past five decades, two candidate visual tasks have emerged as central to this debate. The first, global motion coherence task (Motion), stems from the magnocellular theory of dyslexia, this theory was initially supported by studies demonstrating impaired motion sensitivity in individuals with dyslexia and was tested using both physiological 12–15 and psychophysical approaches 9–11,16,17 . However, replication failures 18–22 and theoretical reformulations 24,25 have since shifted the focus toward broader attentional mechanisms 26–29 , 30 . The second candidate task, multi-element processing (MEP), has gained traction more recently and centers on the ability to rapidly extract information from a string of visual symbols—a skill ecologically linked to reading. In this study we administered the MEP task that was optimized for reliability in kindergarten and first-grade children (see 44 ) and the Motion task adapted from a previous study 31 and tailored for administration to kindergarten and first grade children (see Methods). While both tasks are widely studied, their utility in identifying reading-related risk remains unclear, especially in large, diverse, representative samples. To address this gap in understanding the role of visual processing abilities in early reading development, our goal is to investigate whether developmentally appropriate, scalable visual processing measures—used alongside state-of-the-art language-based tasks—can predict reading difficulties independent of a child's home or instructional language¹⁵⁻¹⁹. We aim to explore the predictive validity of these visual measures, assess their equity across diverse linguistic backgrounds, and uncover heterogeneity in reading outcomes. To address these aims, in this study, we leveraged the state of California-funded Multitudes initiative to take some theory driven, carefully validated 44 , visual measures to a large and diverse sample of kindergarten and first graders in California public schools. Alongside the visual measures, children were administered conventional language-based measures followed by standardized reading assessments at the end of the school year. Results Visual measures are agnostic to a child’s primary language and socioeconomic status. We first asked if children in kindergarten and first-grade exhibit performance differences in the visual-processing measures across a) their eligibility for free or reduced-price meals (an indicator of socioeconomic status) and b) reported primary home language. Our motivation to investigate this question stems from the fact that socioeconomic status impacts educational opportunity and academic outcomes 39,45 . A child’s financial and social resources are among the strongest predictors of performance on reading assessments. Children growing up in lower socioeconomic environments have lower scores on common measures of reading, language, and executive function compared to children growing up in higher-resourced environments 46–49 . The socioeconomic disparity in elementary school reading scores has grown by over 40% in the second half of the 20th century 50 . Screening measures often reflect various biases, resulting in unequal performance in diverse school settings 51,52 . Oftentimes these measures reflect disparities in a child’s experience prior to formal schooling and are not informative in terms of potential neurobiological intervention targets 53,54 . So our first goal was to investigate if these visual measures exhibited bias in : a. Free and reduced-price meal (FRPM): The National School Lunch Program (NSLP) provides free meals for children whose household income is less than 130% of the poverty line and reduced-price meals for income between 130%–185% of the poverty line. FRPM metrics are highly correlated with county poverty rates (0.91 - 0.95) 55 . FRPM eligibility data is a student-level indicator of home income. We compared performance across students who were eligible or ineligible for FRPM. Ineligibility indexes higher socioeconomic status and was coded as 0 in our dataset. b. Student’s primary language: All students whose parents had indicated to the school that English is the only language spoken at home were categorized as English-only (coded as EO). Those students labeled as English learners (coded as EL) whose reported primary language was Spanish were classified as Spanish speakers. These two groups combined comprised ~90% of students in the dataset. The correlation between FRPM eligibility and a student’s primary language was 0.267 (phi coefficient for binary predictors). We examined how students across different grades performed in the visual tasks and found no group differences in terms of FRPM (Figure 1) or primary language (Figure 2). We performed a t -test to compare across groups and adjusted the significance threshold for multiple comparisons [Bonferroni adjusted p threshold: 0.004 (0.05 (p-cutoff) / 12 comparisons)]. Figure 1a shows how children from the FRPM Eligible and Ineligible groups perform across all visual tasks (first column). This can be visually compared to group differences in reading outcome measure (presented second column). We observed no significant difference in group performance between the two FRPM groups for the visual measures but a large effect size (Cohen's d) is observed for the reading outcome measure (WJ-LWI—see Methods for details). Figure 1b further shows how the effect size of the difference between the Eligible and Ineligible groups compares across visual, reading, and language-based measures. The number of participants with student-level data was about half the sample size for each measure because new regulation makes it more difficult for schools to share student-level data, so we also used the openly available school-level data on percentage of students with FRPM eligibility. The school-level FRPM eligibility data was available for the school year 2022-2023 from the California Department of Education 56 . We fit a linear model to the school-level median scores regressed on these measures with percentage of FRPM eligibility reported for each school and grades as predictors. As shown in Figure 1c, the coefficients of FRPM as predictor was close to zero for the visual measures compared to other measures, most evidently for the reading outcome measures as substantiated in Figure 1a. Since English learners in the United States (and California specifically) are predominantly Spanish speakers 57 , we next compared children whose reported primary spoken-language at home was English with those who reported speaking Spanish. Figure 2a shows how English and Spanish-speaking children perform across all visual tasks (first column) and reading outcome measure (second column). Children as young as kindergarten exhibited no group differences in task performance across all visual-processing measures. This is in stark contrast to the Woodcock-Johnson reading outcome measures in our dataset, which showed group differences as early as kindergarten. First-grade children showed a significant group difference only in the letter version of the MEP task. Since the MEP-L task involves letter knowledge, it is possible that factors like pre-school attendance, home environment and parental education could mediate the group difference we notice between the English and Spanish proficient groups by first grade. No such group difference was noticed in the pseudo-letter version of the task. Figure 2b, shows how the effect size of the difference between the English and Spanish groups compare across visual, reading, and language measures. Visual-processing measures are unbiased to differences in the primary language of the participants for kindergarten and the non-alphabetic visual tasks are unbiased to differences in the primary language of the participants in first grade. Note that all the visual tasks were administered in English but children were screened for task understandability (see Methods). Similar to Figure 1c, we fit a linear model to the school-level median scores regressed on these measures with groups based on the primary language of participants from each school and grade as predictors. As shown in Figure 2c, the coefficients of primary language as predictor was close to zero for the model predicting visual-processing measures compared to other measures, most evidently for the reading outcome measure (WJ-LWI) as substantiated in Figure 2a as boxplots. Together, we show that visual-processing measures, notably, the pseudo-letter version of the MEP task (MEP-P) and the global motion coherence tasks (Motion), which are non-alphanumeric visual tasks, show no group differences based on FRPM and primary language across kindergarten and first graders. This is crucial to establish, as we next aim to understand the correlation between these visual-processing measures and reading outcomes without the influence of environmental factors that typically affect reading assessments. By confirming the absence of group differences in these visual tasks, we can more confidently examine their relationship to reading performance. The relationship between visual measures and reading outcome is equivalent across groups. We then asked how performance in the visual measures relates to the end-of-year reading outcome (WJ-LWI—see Methods for details) for each grade. The development of visual processing abilities precedes formal reading instruction, so we hypothesize that individual differences in visual processing ability could account for individual differences in reading outcome as early as kindergarten, independent of SES and home language. To address this question, we first filtered those students who took all the visual measures and had end-of-the-year reading outcome measure (n= 539) and then fit two linear regression models. The first model regresses the reading outcome on all three visual measures and their interactions with FRPM eligibility, Do children with different FRPM eligibility show different associations between visual measures and reading outcomes? We investigated the relationship between visual-processing measures and reading outcome in kindergarten students ( n = 108) and first grade ( n = 240), for whom FRPM data was available. Across both grades, we fit two linear regression models, i) a visual-only model (we included a model with just the non-alphabetic visual tasks presented in Table 1) and ii) a visual with FRPM eligibility interactions as summarized in Table 1 and correlations are presented in Figure 3. First, the model including MEP-L, MEP-P, and Motion as predictors explained 18.22% of the variance in reading outcome in kindergarten and 9.42% in first grade. Only MEP-L performance was a significant predictor for kindergarten ( β = 0.218, p = 0.0253) and first grade ( β = 0.307, p = 0.00063). This is because performance in the MEP-L and MEP-P tasks are highly correlated ( r = 0.7) and correlated with the motion task ( r = 0.3); however, each measure is a significant univariate predictor of reading outcome across both grades (see Supplementary Table S1). An ANOVA comparison between the visual-only model and the model with visual and FRPM interactions demonstrated that including the FRPM eligibility improved model fit for both kindergarten [ F (4,100)= 3.930, p = 0.00526] and first graders [ F (4, 232) = 5.396, p = 0.00035]. The lack of significant interactions indicates that the relationship between visual measures and reading outcomes does not vary based on FRPM eligibility. While children with different FRPM eligibility did not differ significantly in their visual processing abilities, FRPM eligibility emerged as a strong predictor of reading outcome. This highlights the importance of socioeconomic factors in early reading development, independent of visual processing skills and its ineluctable role in reading development. Table 1. Regression models with visual measures and FRPM as predictors of end-of-year reading outcome. For each grade i. a visual-only model, ii. a model with just the non-alphabetic visual tasks and iii. a model with visual tasks and FRPM eligibility as predictors is presented. Kindergarten (n =108) First Grade (n=240) Variable Reading outcome ~ MEP-L + MEP-P + Motion Reading outcome ~ MEP-P + Motion Reading outcome ~ MEP-L * FRPM + MEP-P * FRPM + Motion* FRPM Reading outcome ~ MEP-L + MEP-P + Motion Reading outcome ~ MEP-P + Motion Reading outcome ~ MEP-L * FRPM + MEP-P * FRPM + Motion* FRPM (Intercept) -0.084 (0.069) -0.105 (0.070) -0.279 (0.083)** 0.141 (0.068)* 0.167 (0.069)* -0.122 (0.087) MEP-L 0.218 (0.096)* - 0.138 (0.107) 0.307 (0.089)*** - 0.341 (0.111)** MEP-P 0.032 (0.093) 0.174 (0.070)* 0.041 (0.101) 0.041 (0.085) 0.211 (0.071)** -0.043 (0.102) Motion 0.112 (0.068) 0.149 (0.068)* 0.121 (0.080) 0.040 (0.074) 0.064 (0.076) 0.002 (0.089) FRPM - - 0.530 (0.136)*** - - 0.589 (0.134)*** MEP-L * FRPM - - 0.242 (0.211) - - -0.083 (0.175) MEP-P * FRPM - - 0.008 (0.211) - - 0.164 (0.173) Motion * FRPM - - -0.055 (0.139) - - 0.052 (0.152) Multiple R² 0.182 0.142 0.293 0.094 0.048 0.171 F-statistic 7.73 (3,104)*** 8.67 (2,105)*** 5.93 (7,100)*** 8.19 (3,236)*** 6.01 (2,237)** 6.86 (7,232)*** Standard errors are in parentheses. Significance levels: *** p<0.001, ** p<0.01, * p<0.05, . p<0.1 Do children with different primary languages show different associations with reading outcome? We next investigated the relationship between visual-processing measures and reading outcome across different primary language groups in kindergarten (n=157) and first-grade (n=381). We fit two linear regression models, i) a visual-only model (we included a model with just the non-alphabetic visual tasks presented in Table 2) and ii) a visual with primary language interactions as summarized in Table 2, and correlations are shown in Figure 4. The visual-only model explained 16.3% of the variance in reading outcome measure in kindergarten and 11.9% in first grade. Only MEP-L performance was a significant predictor for kindergarten (β = 0.212, p = 0.0107) and first grade (β = 0.316, p =2.51x10 -6 ). An ANOVA comparison between the visual-only model and the model with visual and primary language interactions demonstrated that including the Primary language improved model fit for both kindergarten [ F (4,149)= 2.993, p =0.021] and first graders [ F (4, 373)= 7.896, p =4.058x10 -06 ]. The lack of significant interactions indicates that the relationship between visual-processing measures and reading outcomes does not vary between English and Spanish speakers. Primary language, however, emerged as a strong predictor of reading outcome measure. Table 2.Regression models with visual measures and child’s primary language as predictors of end-of-year reading outcome. For each grade i. a visual-only model, ii. a model with just the non-alphabetic visual tasks and iii. a model with visual tasks and child’s home language as predictors is presented. Kindergarten (n = 157) First Grade (n=381) Variable Reading outcome ~ MEP-L + MEP-P + Motion Reading outcome ~ MEP-P + Motion Reading outcome ~ MEP-L * Primary Language + MEP-P * Primary Lang + Motion* Primary Language Reading outcome ~ MEP-L + MEP-P + Motion Reading outcome ~ MEP-P + Motion Reading outcome ~ MEP-L * Primary Language + MEP-P * Primary Language + Motion* Primary Language (Intercept) -0.102 (0.060). -0.121 (0.060)* -0.333 (0.092)*** 0.037 (0.052) 0.049 (0.053) -0.303 (0.080)*** MEP-L 0.212 (0.082)* - 0.173 (0.130) 0.317 (0.066)*** - 0.293 (0.100)** MEP-P 0.063 (0.079) 0.194 (0.061)** -0.017 (0.126) 0.062 (0.068) 0.252 (0.057)*** 0.011 (0.107) Motion 0.109 (0.061). 0.137 (0.062)* 0.167 (0.086). 0.043 (0.055) 0.065 (0.057) 0.098 (0.079) Primary Language - - 0.386 (0.119)** - - 0.571 (0.104)*** MEP-L * Primary Language - - 0.048 (0.165) - - -0.044 (0.131) MEP-P * Primary Language - - 0.142 (0.159) - - 0.084 (0.136) Motion * Primary Language - - -0.087 (0.120) - - -0.107 (0.107) Multiple R² 0.163 0.126 0.225 0.119 0.066 0.188 F-statistic 9.894 (3,153)*** 11.09 (2,154)*** 6.165 (7,149)*** 17.00 (3,377)*** 13.31 (2,378)*** 12.33 (7,373)*** Standard errors are in parentheses. Significance levels: *** p<0.001, ** p<0.01, * p<0.05, . p<0.1 Overall, visual-processing measures accounted for ~16–18% of variance in reading outcome in kindergarten and ~10–12% in first grade. Visual measures are important predictors of reading difficulties. We next examined the potential application of the observed relationship between visual processing and reading outcomes, focusing on its utility for early screening and its ability to illuminate the heterogeneity in reading development. Two important considerations guide this analysis: (i) reading difficulties are not primarily visual processing disorders, although visual deficits may co-occur and hold predictive value; and (ii) robust evaluation of early visual markers requires a multi-year longitudinal design and a sufficiently large sample. We first built a risk prediction model with visual processing measures and language-based measures administered during the winter quarter of the school year. Gold-standard reading outcomes (Woodcock Johnson Letter Word Identification, WJ-LWI) were assessed at the end of the same school year (Spring 2023), with a subset of children retested one year later (Spring 2024). Risk is defined as children with scores on outcome measures below the 20th percentile. The sample sizes of children with all measures varied across years. Concurrent reading outcome measures (Spring 2023) were available for 108 kindergarteners and 216 first grade children. Longitudinal follow-up (Spring 2024), were available for K: 53, G1: 112. We performed multiple imputations to address missing data within language, visual and reading outcome domains before merging imputed datasets. The final sample size, after imputations, was K: 157 and G1: 241 (see Methods). In our sample, only ~20% of children were identified as at risk (Kindergarten: 32/157; Grade 1: 51/241), resulting in a notable class imbalance. In such cases, traditional classification metrics like sensitivity and specificity become less informative, as they are highly sensitive to class distribution and can yield unstable or misleading results in small samples. Importantly, class imbalance is an expected feature when the condition of interest—here, reading risk—has a relatively low base rate, and it does not compromise the validity of the predictors themselves. Given our current sample size, we intentionally chose not to emphasize on sensitivity and specificity metrics. As we continue to add cohorts and our sample grows, the absolute number of at-risk children will increase, allowing for more robust estimation of these metrics. Instead, our primary aim was to evaluate the predictive contribution of visual measures relative to conventional language-based predictors. Using a random forest classification model with leave one out cross validation (LOO-CV), we evaluated the combined predictive utility of the visual and language-based measures for identifying children at risk of word reading difficulties (WCJ-LWI < 20th percentile were classified as at risk, based on a previous study 58 ). For concurrent-year risk classification the model achieved accuracies of 87.96% (kindergarten) and 85.19% (first grade). Longitudinal prediction accuracies were 79.73% (kindergarten) and 80.32% (first grade). Given the imbalance in risk prevalence, we focus primarily on feature importance rankings to determine the unique predictive value of visual processing measures. To interpret how individual measures contributed to the predictions, we computed feature importance using the Boruta package in R 59 . Feature importance plots (Figures 4a-d) quantify the relative importance of each predictor - such as MEP, deletion, blending etc., - on the model’s ability to classify children at risk. Higher median importance values indicate that the measure was frequently used to split the decision nodes across the different decision trees in the random forest and thereby consistently improves prediction accuracy when included in the prediction model. These findings offer mechanistic insights into the cognitive processes that are most relevant to the reading outcome at developmentally different grade levels. Our results revealed visual measures, particularly MEP, as consistently important predictors across both time points and grades. It should be noted that since we have fewer children at-risk, models may prioritize variables that simply distinguish the majority-class patterns thereby skewing feature importance estimates. To verify that our findings reflected true patterns rather than imbalance artifacts, we replicated our analysis with class-balanced datasets using the ROSE package in R 60 that simulated a balanced dataset by synthetically oversampling minority-class instances (at-risk) and undersampling the majority class. Supplementary Figure S2 is identical to Figure 4 but, plotted with the simulated balanced data set. We found that the visual measures, especially MEP, emerged as an important and stable predictor of risk, in relation to the conventional language-based measures across both grades. This pattern held true for risk predicted at the end of the school year as well as at the one-year follow-up. To further validate these findings, we conducted a complementary analysis identifying predictors that maximized area under the curve (AUC) and accuracy (Table 3). We prioritized AUC over sensitivity and specificity metrics, as AUC provides a threshold-independent measure of model performance that is particularly well-suited to imbalanced datasets like ours. Risk status was computed as below the 20th percentile for each reading outcome measure (see Methods for a detailed description of composite scores here). The AUC-accuracy profile presented in Table 3, reinforces the role of MEP-L as one of the key predictors consistent with the ranking in the feature importance plots in Figure 4. This aligns with their prominence in feature importance analyses, reinforcing the role of visual processing measures in predicting reading risk. MEP-P, however, became a stable predictor in the absence of MEP-L (Supplementary Table S2), likely due to their high disattenuated correlation (r = 0.91 44 ). The stability of visual measures’ importance, especially MEPs, in both imbalanced and balanced scenarios (Figures 4 vs. Supplementary Figure S2) confirms their robustness as predictors, independent of class distribution. Latent profile analysis reveals subgroups of children with specific visual strength and specific visual challenge. To investigate heterogeneity in reading outcomes, we first looked at performance profiles across language-based and visual measures, to identify subgroups of children, guided by latent profile analysis (LPA). Bayesian Information Criterion (BIC)-optimized LPA, was implemented via the Mclust package in R 61 , on the kindergarten data (n=157). This revealed five latent profiles as the maximum number of subgroups based on the BIC criterion and such that each cluster has a sample size that is more than 10% of the total number of kindergarteners (Supplementary Figure S3). Subgroup criteria derived from the LPA solution on Kindergarten children were subsequently applied to define profiles consistently across children from both grades (see Methods). Figure 5a shows the subgroup performance profiles. As is common with LPA, three of the subgroups represented typical performance differences: (1) a high performing group that performed ~0.25 standard deviations above the mean on all tasks, (2) an average performing group that performed around the mean, and (3) a low performing group that performed ~0.25 standard deviations below the mean. Beyond these groups that differed in performance across all tasks, two interesting subgroups emerged: children with high performance in MEP tasks while demonstrating lower abilities in other domains—the Specific Visual Strength group (10% of total kindergarten children)—and children with low performance on MEP tasks who exhibited high performance across other domains—the Specific Visual Challenge group (15% of total kindergarten children). The Specific Visual Strength group showed longitudinal reading outcomes comparable to the High-Performing group, indicating that strengths with MEP can be a protective factor for reading development. The Specific Visual Challenge group, on the other hand, showed longitudinal reading outcomes close to the average-performing group, despite the fact that they had strong language skills (see Figure 5b). Figure 5c shows the reading outcome (WCJ-LWI) trajectories across time — from concurrent-year to the one year follow up. A linear mixed effects model [specified as: reading outcome ~ Time * Subgroups + (1 | student_id)] revealed a significant main effect of time, F (1, 152) = 23.720, p = 2.769x10 -6 , indicating overall improvement in reading outcome across the year. There was also a significant main effect of subgroups, F (4, 152) = 6.80, p = 4.608x10 -05 , with high-performers and specific visual strength groups showing significantly higher baseline scores than low-performers ( p = 0.000489 and p = 0.0418, respectively). Although the time × group interaction was not significant ( F (4, 152)= 1.9403; p = 0.107), estimated marginal means revealed that all subgroups—except low-performers—exhibited statistically significant reading gains over time. Importantly, initial disparities in reading performance persisted, with low-performers showing no measurable growth. In the first grade cohort (n= 241), similar trends were observed (see Figure 6a and 6b). Figure 6c shows the reading outcome (WCJ-LWI) trajectories across time—from concurrent-year to the year after. Reading outcome improved significantly over the one-year period, F (1, 236) = 39.937, p =1.296x10 -09 and there were marked baseline differences among subgroups, F (4, 236) = 15.027, p = 5.880x10 -11 : low-performers < specific visual strength group < high-performers. Similar to Kindergarten, the interaction between time and group was not significant, F (4, 236) = 1.071, p = 0.372, suggesting that although all groups improved, their rates of growth were comparable and the initial group differences persisted over time. Notably, the Specific Visual Challenge group continued to demonstrate significantly lower reading outcomes over time despite improving in the WCJ-LWI standard score by ~7 ( p = .0102), whereas the Specific Visual Strength group maintained reading levels on par with Average Performers, reinforcing their more resilient developmental trajectory. These results show that children with strong language but poor visual processing abilities exhibited a persistent risk for lower reading development that would have been completely missed if the screening battery did not include visual processing measures. Similarly, children with strong visual processing abilities demonstrated comparable reading performance to High Performers suggesting that there might be compensatory mechanisms early in development that those with Specific Visual Strengths can leverage to build reading skills. The subgroup profiles in our dataset and their developmental trajectories suggest that these patterns emerge early and remain stable across the foundational years of reading acquisition, underscoring the need for early identification and targeted support tailored to distinct profiles. Discussion Despite 40 years of controversy, the debate over whether individual differences in visual processing can explain individual differences in reading ability remains active 3–5,7,34 . This might seem surprising in light of hundreds of studies that report visual processing deficits in children with dyslexia 31,33,34,37,62,63 . Progress has been hindered by small, homogenous samples. Despite best efforts to recruit representative samples, the families who respond to university recruitment efforts are rarely representative of regional or national diversity, yet it is precisely this heterogeneity that must be understood to develop appropriate screening tools and intervention strategies. This dual constraint—methodological and conceptual—has kept visual processing at the periphery of reading science, despite long-standing debate of its relevance in dyslexia and reading difficulties. Here, we addressed these limitations: the visual-processing measures were selected based on prior evidence of their ability to identify struggling readers missed by conventional language-based assessments and were rigorously validated through item response theory for use with young children in a school setting 44 . We then administered these measures to a diverse and representative sample of kindergarteners and first-graders in California public schools and showed that visual processing abilities predict variation in early reading development. We first examined whether the visual-processing measures exhibit bias related to students' home language or eligibility for free and reduced-price meals (an indicator of family income). This investigation addresses a fundamental challenge in early screening: many measures are biased by socioeconomic disparities. Consequently, even when measures demonstrate good predictive power for reading challenges, it often remains unclear whether this predictive ability stems from the measure assessing a construct fundamental to reading development or simply from capturing the same differences in educational opportunity that are known to influence reading outcomes 45 . We observed that visual-processing measures are invariant to socioeconomic status. We filtered for task understandability—blinded from the demographics of the participating students, we first cleaned all the data from the MEP and Motion tasks based on a predetermined rubric for performance (see Methods). We then investigated if our visual measures were biased to the child's home language. Our results show that in kindergarten children grouped by home language did not show differences in task performance for both the MEP and the Motion tasks. However, in first grade children whose home language was English performed significantly higher than the Spanish speakers only in the MEP-L (letter version of the task). These group differences, by first grade, may be mediated by factors such as preschool attendance, home environment, and parental education, given that the MEP-L task involves letter knowledge. We did not observe group differences in the pseudoletter version of the MEP task (MEP-P). Thus, our visual-processing measures, especially the non-alphanumeric based, are both scalable and equitable— free from biases linked to language exposure at home or socioeconomic background. These visual-processing measures explained unique variance in reading outcomes and consistently emerged as important predictors across both feature extraction and AUC-based approaches. Our initial aim was to directly compare prediction models built exclusively with visual measures against those using state-of-the-art language-based tasks, to determine which children were commonly or uniquely identified as at-risk. However, such comparisons require substantially larger samples—especially ensuring sufficient representation of children classified as at-risk. Until these sample sizes are achieved, a more practical and informative approach is to assess whether visual-processing measures remain consistently important predictors within models combining both language-based and visual predictors. Both our random forest classification model and our AUC-based analysis of predictor performance showed that MEPs are a consistent predictor of reading risk—both at the end of the same school year and at a one-year follow-up. This pattern aligns with longstanding debates about the visual pathways that underpin reading. Global motion discrimination tasks, often invoked to probe magnocellular (M‑pathway) integrity, have yielded mixed evidence for a direct mechanistic role in reading acquisition 64,65 . By contrast, MEP tasks are ecologically closer to the demands of word reading: they index how efficiently a child can encode multiple elements within a single fixation—precisely the perceptual bottleneck 66,67 that fluent reading must overcome. It is therefore unsurprising that MEP outperforms the Motion task in forecasting later reading outcomes, providing convergent support for the view that rapid, fixation‑bound visual encoding—not low‑level motion processing—is the more critical constraint on early reading development. We then examined the performance profiles across language-based and visual measures, to identify subgroups of children with distinct profiles. Five distinct profiles emerged based on task performance, three of the five profiles are common to most clustering analysis. The high-, average- and low performers were unique sets of children who in general had scores across language-based and visual tasks that were high-, average- and low. A year later, their median reading outcomes mirrored these tiers, with high performers achieving the highest median scores, low performers the lowest, and average performers falling in between. Interestingly our analysis revealed two novel subgroups of children: a Specific Visual Challenge group, whose reading outcomes lagged despite average to high language skills, and a Specific Visual Strength group, whose reading outcomes matched those of High-Performers despite low to average language skills. Without our visual measures, the Specific Visual Challenge group would have mimicked the High Performers in terms of their scores in the language-based tasks, making it hard to reconcile their poor reading outcomes a year later. In contrast, we found a subgroup of children with Specific Visual Strengths who mimicked the Low Performers in terms of their scores in the language-based tasks but exhibited reading outcomes comparable to the High Performers. We were curious to examine the FRPM eligibility and home language composition of these subgroups (see Supplementary Figure S4). First, based on socio-economic status, we did not observe marked distinctions across most subgroups. While we initially hypothesized the High Performers would predominantly come from higher SES backgrounds, all groups had substantial proportions of FRPM-eligible children, with the highest percentages appearing in the Low Performer group by Grade 1. In contrast, subgroup composition based on home language revealed notable differences. As expected, High Performers were predominantly English-speaking, whereas Low Performers were overwhelmingly Spanish-speaking, reinforcing the cumulative disadvantage faced by linguistically marginalized students and the fact that the reading outcome measures were all in English. Average Performers maintained a balanced language composition. Most intriguing, however, were the home language dynamics within the two novel visually-defined subgroups. In Kindergarten, the Specific Visual Strength group largely consisted of Spanish-speaking children, suggesting visual strengths might offset early language-related disadvantages. Indeed, these Kindergarten Spanish-speaking children with visual strengths went on to achieve strong reading outcomes a year later (as shown in Figure 5b). Conversely, the Specific Visual Challenge group at kindergarten primarily consisted of English-speaking children—a reversal of the typical demographic seen in struggling readers, suggesting that the visual challenge they face is independent of linguistic advantage. These children with visual challenges went on to show considerably poor reading outcomes a year later (Figure 5b). These patterns raise critical questions about the nature of the specific visual processing difficulties faced by these children and the kinds of early interventions that might mitigate their impact on future reading outcomes. Although addressing these questions lies beyond the scope of our current work, future studies that characterize the underlying visual deficits in these subgroups and evaluate tailored, early-stage interventions could yield critical insights—both theoretically and practically—into optimizing reading outcomes for all learners. Notably by Grade 1, the Specific Visual Strength group was predominantly English-speaking children, and the Specific Visual Challenge group flipped to being predominantly Spanish-speaking similar in composition to Low-Performers, reversing the pattern observed in Kindergarten. This cross-sectional shift suggests that the protective influence of visual strengths may be most salient early in development for multilingual learners but may erode over time as reading demands escalate and mismatches between home and instructional language accumulate. Critically, children facing dual vulnerabilities—limited proficiency in the language of instruction and reduced visual processing abilities—appear disproportionately represented among those struggling with reading, underscoring the need for early, multidimensional screening approaches that go beyond language-focused metrics alone. Conclusion By identifying distinct subgroups of children within a general population, our study provides compelling empirical evidence that visual processing abilities contribute meaningfully to heterogeneity in early reading development. This has profound implications for how we design, implement, and interpret early screening programs. As these visual measures have now been adopted into two of the four universal screeners recommended by the state of California, forthcoming cohorts will offer valuable opportunities to evaluate the predictive utility—particularly the sensitivity and specificity—of various combinations of predictors. A central limitation of the current study is that the language-based measures administered showed considerable variability in empirical reliability (0.61–0.97), as they were not originally designed with the linguistic and demographic diversity of California public schools in mind. This may have attenuated estimates of language skill contributions and limits the generalizability of findings based on these measures alone. However, as part of the state-funded Multitudes initiative, language assessments have since been redesigned and calibrated for California’s diverse learner population, complementing the equitable visual measures introduced here. Together, these results challenge long-standing assumptions about the role of vision in reading and lay the foundation for a reconceptualization of variability in reading development. By establishing a population-level framework for longitudinal follow-up, this work enables future studies to trace how visual and language-based risk factors interact over time. In doing so, this research lays the groundwork for a new generation of early identification tools—ones that are inclusive, developmentally appropriate, and empirically grounded—opening the door to precision interventions that can transform outcomes for young readers worldwide. Methods Participating school recruitment Participants were recruited by the University of California, San Francisco (UCSF) Dyslexia Center’s Multitudes project, an initiative supported by the State of California to develop a universal screener for reading challenges and dyslexia in California public schools. The sample was drawn from schools in locations and communities intended to be representative of the state in terms of race, ethnicity, socioeconomic status, and home language. Care was taken to reach remote rural, urban, and suburban communities to ensure that the sample was truly representative of the diversity of lived experiences in California. The study aimed to reduce selection bias with a passive consent process. The Institutional Review Board of UCSF determined that a process of informing parents about the study and how their children would be participating, with clear instructions on how to opt-out through communications with their school administrators, was appropriate for the research going into a universal screener. Parents of eligible kindergarten and first-grade students received an information sheet describing the project, which involved normal classroom activities and presented no more than minimal risk to participants. Children whose parents wished to opt out did not participate in study activities. General Procedure. A team of 6-10 UCSF-trained proctors administered these assessments to kindergarten and first-grade students in each of the participating schools based on the school’s convenience. Reading outcome measures were individually administered to each student. In each school, kindergarten and first-grade children were brought in batches to complete a battery of language, reading, and visual measures. All children were assigned a unique student tracking ID and were randomly grouped in batches. Thus, missing data means that either a child was absent on the day or did not wish to participate. Demographics reported by the participating school consisted of reported primary language of the student, age, English proficiency designation, grade, as well as race and ethnicity information. The student-level data on FRPM eligibility reported in this study was discontinued after the 2022-2023 school year. This data was available only for a selected set of children from the schools that provided FRPM eligibility information. Supplementary Figure S5 shows the percentage of children reported as eligible compared to the eligibility percentage reported by the California Department of Education. Battery of assessments Standardized Reading measures. The Woodcock-Johnson IV Tests of Achievement were administered to all children at the end of the school year to assess key literacy skills through Letter-Word Identification, Word Attack, and Spelling tests. For this study, we used the letter word identification (WCJ-LWI) subtest as the reading outcome for both English and Spanish-speaking children. It should be noted these reading outcome measures in English are a measure of children’s classroom reading demands. The Letter-Word Identification test evaluates basic reading skills by measuring, untimed, word recognition and decoding abilities, it requires them to read and correctly identify letters and words that range from high-frequency to those of increasing difficulty. The LWI tests are integral components of the broader WCJ IV battery, designed to comprehensively evaluate single-word reading abilities and written language capabilities across various developmental stages 68 , 69,70 , 71 . To capture more comprehensive profiles of reading ability, we also examined composite scores from combining subtests derived from the WCJ IV battery: i. Basic Reading Skills: Composed of the letter-word identification and word attack subtests, this composite evaluates fundamental decoding skills, including sight word recognition, phonemic decoding, and structural analysis. It reflects a child’s capacity to accurately read both familiar and novel words using a combination of visual and phonological strategies. ii. Broad Reading composite: This composite integrates performance across letter-word identification, passage comprehension, and sentence reading fluency subtests. It provides a multidimensional index of reading achievement, encompassing decoding accuracy, reading fluency (rate), and comprehension. The composite score offers insight into how effectively students read connected text and extract meaning under time constraints. iii. Reading composite: Comprising letter-word identification and passage comprehension, this composite encompasses decoding and reading comprehension (only administered to first graders). It captures the ability to both recognize individual words and understand meaning in sentences and brief passages, thereby offering a reliable estimate of core reading competence. iv. Reading Fluency: This reflects performance in the sentence reading fluency subtest for first graders. Across all these reading outcome measures, it is important to acknowledge that for Spanish-speaking children, performance may partially reflect English knowledge or ongoing language acquisition rather than pure decoding or automaticity. This conflation of linguistic and reading skills is a known limitation when using English-based assessments in emergent bilingual populations and underscores the need for interpreting reading scores in light of children's instructional language context and exposure. Reading Risk. We chose the 20th percentile of the within-sample distribution of the WJ-LWI as the cut-off for risk classification (this was based on a previous study on the same dataset 58 ). Visual-processing measures. Included rapid visual processing of multiple elements of letters and pseudo letters and global motion processing ability. All visual measures were administered using a touchscreen Chromebook and all instructions were through an engaging audio-visual illustration of the task. Multi-element Processing letters (MEP-L) & Multi-element Processing pseudo-letters (MEP-P) : For task description, trial structure, design, and validation please see 44 . The assessment was gamified to make it more engaging and fun for younger children. The task simply involved a string of elements that were briefly presented for 240ms. After the elements disappeared a blue line appears below one of the element positions and participants were required to identify the element that was presented above the blue line from a set of six choices. The game sets children on a mission to help a lost whale in the sea, by playing this game of elements that is the secret map to the lost treasures and friends. Motion : This experiment was built using JS psych and launched online on the ROAR app platform 72 . The stimuli for the motion task were adapted from a previous study 31 . In this single-interval forced-choice experiment, on each trial random dots with varying proportions of signal and noise were presented. The assessment was gamified to make it more engaging and fun for younger children. The task simply involves participants judging the global direction of motion of the entire dot field. The game sets children on a mission to help Beatrice the bee, in finding where the colony of bees is headed in a dense forest. If the colony of bees is headed to the apple tree they have to click on the apple tree on the left or the lemon tree on the right. Task understandability: For the visual measures, we wanted to distinguish task understandability from task performance. Blinded from the demographics of the participating students, we first cleaned all the data from the MEP and Motion tasks based on a predetermined rubric for performance. MEP task filtering criterion : For the MEP tasks only children who attempted and engaged with at least 24 trials were considered for further analysis. In addition, those participants whose raw scores were below 4 correct out of the 24 attempted trials had ability estimates that were overestimated by the IRT model. These participants are highlighted in green in Supplementary Figures S1a and S1b and were excluded from further analysis. Supplementary Figure 1c shows the reading outcome distribution for the excluded children 115 (MEP-L) and 135 (MEP-P) confirming that we were not categorically excluding poor readers. Motion task filtering criterion : This task had a non-response contingent trial succession, meaning trials are a fixed length time and the game automatically will progress to the next trial. Therefore, it is critical to define a performance cut-off that would inform us about the child’s task understandability. All participants who surpassed a performance cut-off of > 57% in the highest signal coherence conditions of 96% were filtered and included. This cut-off in performance was to verify that children understood the task and performed above chance in at least the easiest trials. This criterion excluded around 220 participants. The excluded participants reading outcome distributions were normal and we weren’t categorically eliminating poor readers (see Supplementary Figure S1d). Language-based measures . A pilot version of the Reach Every Reader (RER) digital assessment battery 73 with individual tasks that were widely used in screening was administered 74 . These tasks tapped into constructs of phonological awareness (deletion, blending), decoding (letter naming, word reading, non-word reading), phonological working memory (digit span, non-word repetition), rapid object naming, and oral language (expressive vocabulary, sentence repetition). We selected the Woodcock-Johnson Letter-Word Identification (WCJ LWI) subtest as our reading outcome measure, as it assesses word reading skills—a core component of decoding, therefore the additional decoding measures were not included in the language model. Administration and data collection of RER language-based measures: All language measures were administered on iPads with headphones. The administration involved a proctor-facing iPad and a student-facing iPad. Students were asked to wear headphones for instructions and prompts. All instructions were very engaging and child-friendly audio-visual illustrations of the task. Data were collected by UCSF proctors working individually with each child, ensuring the accuracy and efficiency of administration and data collection. Most measures were designed as computer adaptive testing and the platform returns ability estimates after each child completes the task. For measures that are not computer-adaptive like rapid automatic naming of objects, the platform recorded raw scores. Procedure: Phonological awareness was measured using the blending (BLE) and deletion (DEL) tasks. For the blending task, students join two phonemes or words to form a single word and remove one phoneme or word from a given word for the deletion task. Decoding skills were assessed using Word Reading (WRE) and Non-Word Reading (NRE), where students read the word or nonsense word presented on screen. Language and vocabulary skills were assessed with expressive vocabulary (EVO) and sentence repetition (SRT) tasks. EVO involved naming the visually presented object correctly and in SRT students repeated a set of sentences. Letter Naming context (LNC) and rapid automatic naming (RAO) skills were both time-sensitive measures and were scored based on the total number of correct responses in 45 seconds. In the RAO task, children were presented with a grid of 5x5 objects and a practice to name all the objects. The task is to name the objects in the 5x5 grid as quickly as possible. Similarly, for the LNC task, children were presented with random upper and lowercase letters and were asked to read out the grid of 5x5 letters as quickly as possible. Phonological working memory was assessed using the digit span (DGS) task and non-word repetition (NWR) tasks. In the DGS task, students repeated two-digit number sequences that incrementally went up to six digits. The task was discontinued after three consecutive incorrect responses. In the NWR task, students repeated a list of nonsense words. Reliability of language-based measures: The empirical reliability of the language measures used in this study ranged from .61 to .97. Most of the criterion-referenced reliability for the language measures ranged from .54 to .78. These numbers indicate moderate to high consistency across the majority of measures (see Appendix B Table B1 in a previous study 58 ). Data Imputations. To address missing values and preserve statistical power across a wide range of predictors, we performed multiple imputation using the mice package in R 75 . Imputations were conducted separately for kindergarten and first grade cohorts before each analysis. Imputations were applied to distinct subtests of variables based on their domain: the domains were: i. language-based measures (deletion, bending, non-word repetition, sentence repetition, digit span, rapid automatic naming, evocative vocabulary); ii. rapid visual processing measures (MEP-L and MEP-P); iii. Motion (is its own domain); iv. reading outcome at the end of the same school year (WCJ: LWI, WA, ORF, RF, SRF) and v. reading outcome measures at the end of the next school year (WCJ: LWI, WA, ORF, RF, SRF, Spelling). Within each subset, missing values were imputed using predictive mean matching (PMM) and a fixed random seed to ensure reproducibility. After imputation, the subsets were merged using a common student identifier to generate a single, fully imputed dataset per grade. The imputed kindergarten and first grade datasets were then combined for all downstream analyses. Importantly, imputation was performed prior to the calculation of outcome-based classifications (e.g., risk status) to avoid introducing bias. This approach enabled the integration of diverse behavioral and cognitive data while preserving the natural variance structure and ensuring the reproducibility of the analysis pipeline Feature importance plots. To identify which language-based and visual predictors most reliably distinguished children at risk for reading difficulties, we employed the Boruta feature selection algorithm 59 , a wrapper-based method built on random forests that iteratively compares the importance of actual features against permuted “shadow” features. This approach allows for the identification of all relevant features without relying on arbitrary thresholds or assumptions about linearity. Data preparation: We selected a comprehensive set of predictor variables based on theoretical relevance to early reading development. These included measures of rapid visual processing (MEP–L; MEP–P; Motion), language skills (sentence repetition and expressive vocabulary), phonological awareness (deletion, blending), rapid automatized naming (rao), and phonological working memory (digit span and non-word repetition). Each variable was z-scored prior to analysis. The binary reading risk outcome was derived based on WCJ at the end of first grade. Two datasets were constructed: 1. An original (imbalanced) dataset with class imbalances of those at risk; 2. A synthetically balanced dataset generated using the ROSE (Random Over-Sampling Examples) package in R 60 , which synthesizes new data points by bootstrapping and interpolating between existing minority class observations to counteract the effects of class imbalance. Feature selection was conducted separately on both the imbalanced and balanced datasets using the Boruta function in R with 100 random forest iterations and 100 trees per forest. The outcome variable was reading risk status, and all language-based and visual predictors were included as candidate features. Feature importance was assessed using Z-score normalized mean decrease in accuracy values, and predictors were ranked based on their median importance scores across iterations. A custom post-processing function was applied to clean and extract the Boruta output. Importance values were visualized using boxplots ordered by median importance as shown in Figure 4 and Supplementary Figure S2. Latent profile analysis. To identify subgroups of children with distinct profiles based on performance, we conducted a model-based latent profile analysis (LPA) using the Mclust package in R 61 . This approach fits gaussian finite mixture models under various constraints to the covariance matrices. Preferred models were selected using the Bayesian Information Criterion (BIC). All continuous input variables were standardized (z-scored) prior to modeling. The LPA was first run on kindergarten children ( n = 157). We chose the EEI model—which assumes equal volume and shape for each group with diagonal covariance matrices across aligned to the coordinate axes (Supplementary Figure S3). The EEI parameterization assumes the grouping variables to be uncorrelated and offers a balanced compromise between model complexity and parsimony, particularly suitable for relatively small samples where overparameterization can result in unstable solutions. Among all models evaluated, the EEI model with five latent profiles achieved one of the best BIC values while maintaining the constraint that each cluster contain at least 10% of the sample, ensuring sufficient subgroup stability for interpretation. The five subgroups reflected a range of performance patterns. These included a high-performing group with scores ≥ +0.25 SD across all tasks, an average-performing group with scores near the sample mean, and a low-performing group with scores ≤ – 0.25 SD. The visual strength group exhibited scores in the MEPs ≥ +0.25 SD and scores on deletion < -0.10 and the visual challenge group exhibited scores in deletion ≥ +0.10 SD and scores on MEPs < -0.25 SD. These empirically derived profiles were subsequently applied to the cohort of first graders to assess the generalizability of the classification framework across samples. Declarations Competing interests. The authors declare no competing interests. Data Availability. De-identified data is available in the analysis repository. Code Availability. The experiment code is available at https://github.com/yeatmanlab/roar-mep The analysis code is available at https://github.com/yeatmanlab/Visual-Tasks-as-Language-agnostic-Early-Identification-Measures-of-Reading-Challenges References Snowling, M. J. Dyslexia: a hundred years on. BMJ 313 , 1096–1097 (1996). Pennington, B. F. & Edition, W. Controversial Therapies for Dyslexia. Perspect. Lang. Lit. Winter , 7–8 (2011). Vellutino, F. R., Fletcher, J. M., Snowling, M. J. & Scanlon, D. M. Specific reading disability (dyslexia): What have we learned in the past four decades? J. Child Psychol. Psychiatry 45 , 2–40 (2004). Stein, J. F. & Walsh, V. To see but not to read; the magnocellular theory of dyslexia. Trends Neurosci. 20 , 147–152 (1997). Hulme, C. The implausibility of low-level visual deficits as a cause of children’s reading difficulties. Cogn. Neuropsychol. 5 , 369–374 (1988). Wolf, M. et al. Towards a dynamic, comprehensive conceptualization of dyslexia. Ann. Dyslexia (2024) doi:10.1007/s11881-023-00297-1. Lovegrove, W. J., Bowling, A., Badcock, D. & Blackwood, M. Specific reading disability: Differences in Contrast Sensitivity as a Function of Spatial Frequency. Science 210 , 439–440 (1980). Demb, J. B., Boynton, G. M., Best, M. & Heeger, D. J. Psychophysical evidence for a magnocellular pathway deficit in dyslexia. Vision Res. 38 , 1555–1559 (1998). Talcott, J. B. et al. Dynamic sensory sensitivity and children’s word decoding skills. Proc. Natl. Acad. Sci. U. S. A. 97 , 2952–2957 (2000). Talcott, J. B., Hansen, P. C., Assoku, E. L. & Stein, J. F. Visual motion sensitivity in dyslexia: Evidence for temporal and energy integration deficits. Neuropsychologia 38 , 935–943 (2000). Boets, B., Vandermosten, M., Cornelissen, P., Wouters, J. & Ghesquière, P. Coherent Motion Sensitivity and Reading Development in the Transition From Prereading to Reading Stage. Child Dev. 82 , 854–869 (2011). Eden, G. F. et al. Abnormal processing of visual motion in dyslexia revealed by functional brain imaging. Nature 382 , 66–69 (1996). Livingstone, M. S., Rosen, G. D., Drislane, F. W. & Galaburda, a. M. Physiological and anatomical evidence for a magnocellular defect in developmental dyslexia. Proc. Natl. Acad. Sci. U. S. A. 88 , 7943–7947 (1991). Demb, J. B., Boynton, G. M. & Heeger, D. J. Brain activity in visual cortex predicts individual differences in reading performance. Proc. Natl. Acad. Sci. U. S. A. 94 , 13363–13366 (1997). Ben-Shachar, M., Dougherty, R. F., Deutsch, G. K. & Wandell, B. A. Contrast responsivity in MT+ correlates with phonological awareness and reading measures in children. Neuroimage 37 , 1396–1406 (2007). Cornelissen, P., Richardson, A., Mason, A., Fowler, S. & Stein, J. Contrast sensitivity and coherent motion detection measured at photopic luminance levels in dyslexics and controls. Vision Res. 35 , 1483–1494 (1995). Talcott, J. B. et al. On the relationship between dynamic visual and auditory processing and literacy skills; results from a large primary-school study. Dyslexia 8 , 204–225 (2002). Paulesu, E., Danelli, L. & Berlingeri, M. Reading the dyslexic brain: multiple dysfunctional routes revealed by a new meta-analysis of PET and fMRI activation studies. Front. Hum. Neurosci. 8 , 830 (2014). Amitay, S., Ben-Yehudah, G., Banai, K. & Ahissar, M. Disabled readers suffer from visual and auditory impairments but not from a specific magnocellular deficit. Brain 125 , 2272–2285 (2002). Amitay, S., Ben-Yehudah, G., Banai, K. & Ahissar, M. Reply to: Visual magnocellular deficits in dyslexia. Brain 126 , 3E–3 (2003). Skottun, B. C. The magnocellular deficit theory of dyslexia: The evidence from contrast sensitivity. Vision Res. 40 , 111–127 (2000). Skottun, B. C. The need to differentiate the magnocellular system from the dorsal stream in connection with dyslexia. Brain Cogn. 95 , (2015). Joo, S. J., Donnelly, P. M. & Yeatman, J. D. The causal relationship between dyslexia and motion perception reconsidered. Sci. Rep. 7 , 4185 (2017). Flint, S. & Pammer, K. It is the egg, not the chicken; dorsal visual deficits present in dyslexia are not present in illiterate adults. Dyslexia 0 , 1–15 (2018). Gori, S., Seitz, A. R., Ronconi, L., Franceschini, S. & Facoetti, A. Multiple Causal Links between Magnocellular-Dorsal Pathway Deficit and Developmental Dyslexia. Cereb. Cortex 26 , 4356–4369 (2016). Facoetti, A., Paganoni, P., Turatto, M., Marzola, V. & Mascetti, G. G. Visual-spatial attention in developmental dyslexia. Cortex 36 , 109–123 (2000). Facoetti, A., Lorusso, M. L., Paganoni, P., Umiltà, C. & Mascetti, G. G. The role of visuospatial attention in developmental dyslexia: evidence from a rehabilitation study. Brain Res. Cogn. Brain Res. 15 , 154–164 (2003). Facoetti, A. et al. The relationship between visuo-spatial attention and nonword reading in developmental dyslexia. Cogn. Neuropsychol. 23 , 841–855 (2006). Facoetti, A., Ruffino, M., Peru, A., Paganoni, P. & Chelazzi, L. Sluggish engagement and disengagement of non-spatial attention in dyslexic children. Cortex 44 , 1221–1233 (2008). Hari, R. & Renvall, H. Impaired processing of rapid stimulus sequences in dyslexia. Trends Cogn. Sci. 5 , 525–532 (2001). O’Brien, G. & Yeatman, J. D. Bridging sensory and language theories of dyslexia: Toward a multifactorial model. Dev. Sci. 24 , e13039 (2021). Valdois, S., Reilhac, C., Ginestet, E. & Line Bosse, M. Varieties of Cognitive Profiles in Poor Readers: Evidence for a VAS-Impaired Subtype. J. Learn. Disabil. 22219420961332 (2020) doi:10.1177/0022219420961332. Franceschini, S., Gori, S., Ruffino, M., Pedrolli, K. & Facoetti, A. A causal link between visual spatial attention and reading acquisition. Curr. Biol. 22 , 814–819 (2012). Vidyasagar, T. R. & Pammer, K. Dyslexia: a deficit in visuo-spatial attention, not in phonological processing. Trends Cogn. Sci. 14 , 57–63 (2010). Gori, S. & Facoetti, A. How the visual aspects can be crucial in reading acquisition? The intriguing case of crowding and developmental dyslexia. J. Vis. 15 , 15.1.8 (2015). Tallal, P. Improving neural response to sound improves reading. Proc. Natl. Acad. Sci. U. S. A. 109 , 16406–16407 (2012). Goswami, U. Sensory theories of developmental dyslexia: three challenges for research. Nat. Rev. Neurosci. 16 , 43–54 (2015). Peterson, R. L. & Pennington, B. F. Developmental dyslexia. Annu. Rev. Clin. Psychol. 11 , 283–307 (2015). Bradley, R. H. & Corwyn, R. F. Socioeconomic status and child development. Annu. Rev. Psychol. 53 , 371–399 (2002). Pugh, K. & McCardle, P. How Children Learn to Read: Current Issues and New Directions in the Integration of Cognition, Neurobiology and Genetics of Reading and Dyslexia Research and Practice . (Psychology Press, 2011). Solano-Flores, G. & Li, M. Generalizability of test scores across languages: Using multiple data sources to investigate the translation of achievement tests for multilingual populations. International Journal of Testing 9 , 98–123 (2009). Solano-Flores, G. & Trumbull, E. Examining language in context: The need for new research and practice paradigms in the testing of English-language learners. Educ. Res. 32 , 3–13 (2003). Solano-Flores, G. This paper examines how language, dialect, and register affect the assessment of multilingual students and highlights the challenges of creating assessments that are valid across linguistic and cultural groups. Review of Research in Education 30 , 117–150 (2006). Ramamurthy, M. et al. Design and validation of a rapid visual processing measure for screening reading difficulties in early childhood. https://osf.io/x6ysc/download. Reardon, S. F. The widening academic achievement gap between the rich and the poor. in Inequality in the 21st Century 177–189 (Routledge, 2018). doi:10.4324/9780429499821-33. Sarsour, K. et al. Family socioeconomic status and child executive functions: the roles of language, home environment, and single parenthood. J. Int. Neuropsychol. Soc. 17 , 120–132 (2011). Lawson, G. M., Hook, C. J. & Farah, M. J. A meta-analysis of the relationship between socioeconomic status and executive function performance among children. Dev. Sci. 21 , (2018). Christensen, J., Aarøe, L., Baekgaard, M., Herd, P. & Moynihan, D. P. Human Capital and Administrative Burden: The Role of Cognitive Resources in Citizen‐State Interactions. Public Administration Review 80 , 127–136 (2020). Zelazo, P. D. & Carlson, S. M. The neurodevelopment of executive function skills: Implications for academic achievement gaps. Psychol. Neurosci. 13 , 273–298 (2020). Reardon, S. F. The widening academic achievement gap between the rich and the poor. in Social stratification 536–550 (Routledge, 2018). Valencia Sheila W & Riddle Buly Marsha. Behind Test Scores: What Struggling Readers REALLY Need. The Australian Journal of Language and Literacy 27 , 217–233 (2004). Rathbun, A. & West, J. From Kindergarten Through Third Grade: Children’s Beginning School Experiences. Mathematica Policy Research Reports . Norton, E. S., Beach, S. D. & Gabrieli, J. D. E. Neurobiology of dyslexia. Curr. Opin. Neurobiol. 30 , 73–78 (2015). Raschle, N. M., Chang, M. & Gaab, N. Structural brain alterations associated with dyslexia predate reading onset. Neuroimage 57 , 742–749 (2011). Croninger, R., Rice, J. K. & Checovich, L. Evaluation of the Use of Free and Reduced-Price Meal Eligibility as a Proxy for Identifying Economically Disadvantaged Students. Alternative Measures and Recommendations . (Maryland State Department of Education. Retrieved from http://www …, 2015). DataQuest (CA Dept of Education). https://dq.cde.ca.gov/dataquest/. Facts about English Learners in California. https://www.cde.ca.gov/ds/ad/cefelfacts.asp#:~:text=Basic%20Facts. Siebert, J. M. et al. Differences in reading screening accuracy by percentile cutoff and English proficiency: Feature selection and group-wise prediction evaluation. EdArXiv (2024) doi:10.35542/osf.io/kswau. Kursa, M. B. & Rudnicki, W. R. Feature Selection with theBorutaPackage. J. Stat. Softw. 36 , 1–13 (2010). Lunardon, N., Menardi, G. & Torelli, N. ROSE: A package for binary imbalanced learning. R J. 6 , 79 (2014). Scrucca, L., Fraley, C., Murphy, T. B. & Raftery, A. E. Model-Based Clustering, Classification, and Density Estimation Using Mclust in R . (Chapman & Hall/CRC, Philadelphia, PA, 2023). Lobier, M., Zoubrinetzky, R. & Valdois, S. The visual attention span deficit in dyslexia is visual and not verbal. Cortex 48 , 768–773 (2012). Mascheretti, S. et al. Visual motion and rapid auditory processing are solid endophenotypes of developmental dyslexia. Genes Brain Behav. 17 , 70–81 (2018). Norcia, A. M. & Tyler, C. W. The role of the magnocellular pathway in dyslexia. Current Directions in Psychological Science 13 , 200–204 (2004). Vidyasagar, T. R. & Pammer, K. Impaired visual search in dyslexia relates to the role of the magnocellular pathway in attention. Neuroreport 10 , 1283–1287 (1999). Ronen, I. & Yeshurun, Y. Perceptual and cognitive mechanisms in reading. Psychological Science 24 , 2335–2344 (2013). Rey, A. M. & Valdois, S. The role of visual processing in reading development: Insights from the visual processing speed theory. Psychological Bulletin 139 , 1224–1252 (2013). Mather, N. & Wendling, B. J. Essentials of WJ IV Tests of Achievement . (John Wiley & Sons, 2015). Miciak, J., Ahmed, Y., Capin, P. & Francis, D. J. The reading profiles of late elementary English learners with and without risk for dyslexia. Ann. Dyslexia 72 , 276–300 (2022). Hajovsky, D. B., Villeneuve, E. F., Schneider, W. J. & Caemmerer, J. M. An Alternative Approach to Cognitive and Achievement Relations Research: An Introduction to Quantile Regression. Journal of Pediatric Neuropsychology 6 , 83–95 (2020). Schrank, F. A., Mcgrew, K. S., Mather, N. & Woodcock-Johnson, I. V. Woodcock-Johnson IV Tests of Achievement. Tests of Achievement (2014). Yeatman, J. D. et al. Rapid Online Assessment of Reading (ROAR). https://roar.stanford.edu/technical/ (2024). Catts, H. W. & Petscher, Y. A Cumulative Risk and Resilience Model of Dyslexia. J. Learn. Disabil. 55 , 171–184 (2022). January, S.-A. A. & Klingbeil, D. A. Universal screening in grades K-2: A systematic review and meta-analysis of early reading curriculum-based measures. J. Sch. Psychol. 82 , 103–122 (2020). Van Buuren, S. & Groothuis-Oudshoorn, K. “mice: Multivariate Imputation by Chained Equations in R. Journal of Statistical Software 45 , 1–67 (2011). Table 3 Table 3 is available in the Supplementary Files section. Additional Declarations There is NO Competing Interest. Supplementary Files SupplementaryMaterial.pdf Supplementary Material: The link between vision and reading: A language-agnostic window into heterogeneity in early reading development. Table3.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6804845","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":472395773,"identity":"80655135-038a-4913-baf7-cdca0a73e149","order_by":0,"name":"Mahalakshmi Ramamurthy","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA10lEQVRIiWNgGAWjYFACHgaGBAMbMPMDAwMzkEogQsuDijQQxTiDaC2MD84cJkGLbvvZgx8S287b27OfPdjwcY81Az97jgFeLWZn8pIlEttuJ/bw5CU2zniWziDZ84aAlgM5BiAtCTwMOeaPeQ4cZjC4QciW82+MfyS2nbPn4X9j2AzSYk9Qy40cM4mEMwcYeyRyIFoMJAhqeWNmkVCRnNhz441h44wD6TwSZ54VEHBYjvHNHwZ29uz9OYYNHw5Yy/G3J2/AqwUD8JCmfBSMglEwCkYBVgAAByFLW7K6dkcAAAAASUVORK5CYII=","orcid":"","institution":"Stanford University","correspondingAuthor":true,"prefix":"","firstName":"Mahalakshmi","middleName":"","lastName":"Ramamurthy","suffix":""},{"id":472395774,"identity":"e5ac025d-bbfe-4dfb-80ad-1ae636f95591","order_by":1,"name":"Klint Kanopka","email":"","orcid":"","institution":"New york University","correspondingAuthor":false,"prefix":"","firstName":"Klint","middleName":"","lastName":"Kanopka","suffix":""},{"id":472395775,"identity":"6bd89bab-e3aa-4533-bce0-8324cd70a4a3","order_by":2,"name":"Julian Siebert","email":"","orcid":"","institution":"Stanford University","correspondingAuthor":false,"prefix":"","firstName":"Julian","middleName":"","lastName":"Siebert","suffix":""},{"id":472395776,"identity":"27d02806-31ce-400e-90a8-6bb12f281a79","order_by":3,"name":"Lucy Yan","email":"","orcid":"","institution":"Weill Institute for Neurosciences, University of California, San Francisco","correspondingAuthor":false,"prefix":"","firstName":"Lucy","middleName":"","lastName":"Yan","suffix":""},{"id":472395777,"identity":"54ea5820-4b3f-4124-849d-79ce1b6eb19e","order_by":4,"name":"Carrie Townley-Flores","email":"","orcid":"","institution":"Stanford University","correspondingAuthor":false,"prefix":"","firstName":"Carrie","middleName":"","lastName":"Townley-Flores","suffix":""},{"id":472395778,"identity":"3b20d44d-34a6-4ccd-a008-3858726e0fce","order_by":5,"name":"Mónica Zegers","email":"","orcid":"https://orcid.org/0000-0002-9672-6417","institution":"Weill Institute for Neurosciences, University of California, San Francisco","correspondingAuthor":false,"prefix":"","firstName":"Mónica","middleName":"","lastName":"Zegers","suffix":""},{"id":472395779,"identity":"cd143d67-7127-4925-be3c-95c26b01a256","order_by":6,"name":"Francesca Pei","email":"","orcid":"","institution":"Weill Institute for Neurosciences, University of California, San Francisco","correspondingAuthor":false,"prefix":"","firstName":"Francesca","middleName":"","lastName":"Pei","suffix":""},{"id":472395780,"identity":"d94620ad-690b-4da2-80e5-17dd266faa22","order_by":7,"name":"Phaedra Bell","email":"","orcid":"","institution":"University of California San Francisco","correspondingAuthor":false,"prefix":"","firstName":"Phaedra","middleName":"","lastName":"Bell","suffix":""},{"id":472395781,"identity":"0b4f560c-eea3-4f9d-9090-b6fe8823022e","order_by":8,"name":"Hugh Catts","email":"","orcid":"","institution":"Florida State University","correspondingAuthor":false,"prefix":"","firstName":"Hugh","middleName":"","lastName":"Catts","suffix":""},{"id":472395782,"identity":"1b7657bb-a9f7-4e5f-b986-0ea0755b215f","order_by":9,"name":"Maria Luisa Gorno-Tempini","email":"","orcid":"","institution":"","correspondingAuthor":false,"prefix":"","firstName":"Maria","middleName":"Luisa","lastName":"Gorno-Tempini","suffix":""},{"id":472395783,"identity":"fcd0edf4-acc2-4707-938c-1eeccacd8549","order_by":10,"name":"Jason Yeatman","email":"","orcid":"https://orcid.org/0000-0002-2686-1293","institution":"Stanford University","correspondingAuthor":false,"prefix":"","firstName":"Jason","middleName":"","lastName":"Yeatman","suffix":""}],"badges":[],"createdAt":"2025-06-02 19:30:14","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6804845/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6804845/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":85180426,"identity":"dec913d9-d870-4f93-a37d-0209df163491","added_by":"auto","created_at":"2025-06-23 07:14:31","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":439342,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003ea\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e.\u003c/strong\u003e The first column shows differences in task performance in the MEP-L , MEP-P and Motion (MEP-L: Multi-element processing of letters; MEP-P: Multi-element processing of pseudo-letters; Motion: Global motion coherence discrimination) measures across the free and reduced-price meals eligible and ineligible groups for both grades and the second column shows differences in reading outcome measure (Woodcock-Johnson (WJ) LWI: Letter word identification) for the same participants. \u003cem\u003e\u003cstrong\u003eb\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e. \u003c/em\u003eThe magnitude of effect size between the eligible (E) and ineligible (InE) groups across visual, and language (Language measures included: LNC: Letter Naming Context; DEL: Deletion; RAO: Rapid Automatic naming; DGS: Digit Span; BLE: Blending; SRT: Sentence repetition; NWR: Non-word repetition; NRE: Non-word reading; WRE: Word reading; EVO: Expressive Vocabulary and other Woodcock Johnson subtests Spell: Spelling; WA: Word Attack) measures administered at the same time point across both grades. Sample sizes for each group for each measure are printed next to each measure; \u003cem\u003e\u003cstrong\u003ec. \u003c/strong\u003e\u003c/em\u003eThe beta coefficient from a linear model fit to schools’ median score on each measure with FRPM eligibility for each school and grade as predictors. Across both\u003cem\u003e\u003cstrong\u003e b\u003c/strong\u003e\u003c/em\u003e. and \u003cem\u003e\u003cstrong\u003ec.\u003c/strong\u003e\u003c/em\u003ewe see that visual measures are more unbiased than the other measures and most evidently the reading outcome measures. Error bars represent confidence intervals.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/ad622419de0a43e67ee6f33f.png"},{"id":85179033,"identity":"399cd70f-7699-4f79-9a0d-f1c5af7aea49","added_by":"auto","created_at":"2025-06-23 07:06:31","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":474713,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cem\u003e\u003cstrong\u003ea\u003c/strong\u003e\u003c/em\u003e\u003cstrong\u003e.\u003c/strong\u003e The first column shows differences in task performance in the MEP-L , MEP-P, and Motion (MEP-L: Multi-element processing of letters; MEP-P: Multi-element processing of pseudo-letters; Motion: Global motion coherence discrimination) measures across the two primary language groups for both grades. The second column shows the differences in reading outcome measure (Woodcock-Johnson (WJ) LWI: Letter word identification) for the same participants. \u003cem\u003e\u003cstrong\u003eb\u003c/strong\u003e\u003c/em\u003e\u003cem\u003e. \u003c/em\u003eshows the magnitude of effect size between the English and Spanish-speaking groups across visual, and language (Language measures included: LNC: Letter Naming Context; DEL: Deletion; RAO: Rapid Automatic naming; DGS: Digit Span; BLE: Blending; SRT: Sentence repetition; NWR: Non-word repetition; NRE: Non-word reading; WRE: Word reading; EVO: Expressive Vocabulary and other Woodcock Johnson subtests Spell: Spelling; WA: Word Attack) measures administered at the same point across both grades. \u003cem\u003e\u003cstrong\u003ec. \u003c/strong\u003e\u003c/em\u003eshows the coefficient from a linear model fit to schools’ median score on each measure with primary language and grade as predictors. Across both \u003cem\u003e\u003cstrong\u003eb.\u003c/strong\u003e\u003c/em\u003e and \u003cem\u003e\u003cstrong\u003ec. \u003c/strong\u003e\u003c/em\u003ewe see that visual measures are more unbiased than the reading outcome measures. Error bars represent confidence intervals.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/e8e5bbef9450284c7f1fac23.png"},{"id":85179034,"identity":"3a68a929-8fdc-46e1-ad07-1c12b729a5c1","added_by":"auto","created_at":"2025-06-23 07:06:31","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":587635,"visible":true,"origin":"","legend":"\u003cp\u003eshows the Correlation between visual measures and reading outcome for each grade, respectively. The light and dark blue correspond to the group split based on FRPM eligibility presented in Figure 1a respectively. \u003cem\u003eLeft \u003c/em\u003e(MEP-L), \u003cem\u003emiddle\u003c/em\u003e (MEP-P), and \u003cem\u003eright\u003c/em\u003e(Motion).\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/f174830f470aeb0828875b82.png"},{"id":85180428,"identity":"a1d404d3-cf6f-47e3-94c9-04ad6179fe14","added_by":"auto","created_at":"2025-06-23 07:14:31","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":500177,"visible":true,"origin":"","legend":"\u003cp\u003e\u003cstrong\u003ea.\u003c/strong\u003eTop panel: \u003cem\u003eleft\u003c/em\u003e(MEP-L), \u003cem\u003emiddle\u003c/em\u003e (MEP-P), and \u003cem\u003eright\u003c/em\u003e(Motion) shows the correlation between visual measures and reading outcome for each grade, respectively. The ivory and green correspond to the group split based on primary language presented in Figure 2a respectively.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/d2fdd0db3c655889ad3507d7.png"},{"id":85179037,"identity":"9307417d-d2a3-46c9-bd5d-639383f72ccd","added_by":"auto","created_at":"2025-06-23 07:06:31","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":469348,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 4. a. and b. shows feature importance plots for predicting risk at end-of-the-same-year across both grades. c. and d. show feature importance plots for predicting risk a year later for the same kids who were longitudinally followed up. Supplementary Figure S2 shows the exact plot for simulated class-balanced data. Across both the class- balanced and imbalanced datasets the MEP measures are important predictors of risk alongside other language-based measures.\u003c/p\u003e","description":"","filename":"5.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/201a61d14376407c0a883a6d.png"},{"id":85179035,"identity":"5354c385-4bf1-4368-9f9d-2661d630faaa","added_by":"auto","created_at":"2025-06-23 07:06:31","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":620449,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 5. Latent profile analysis identifies cognitive subgroups with distinct reading trajectories in Kindergarten children. \u003cstrong\u003ea\u003c/strong\u003e. Performance profiles derived from latent profile analysis based on z-scored performance across visual processing and language measures; \u003cstrong\u003eb\u003c/strong\u003e. Distribution of Woodcock–Johnson Letter–Word Identification (WJ–LWI) scores one year later, showing stratified reading outcomes across subgroups; \u003cstrong\u003ec\u003c/strong\u003e. Individual reading growth trajectories from baseline to follow-up, with bold lines indicating group-level means; \u003cstrong\u003ed–f\u003c/strong\u003e. density plots showing subgroup distributions for basic reading skills (d), broad reading (e), and reading composite scores (f).\u003c/p\u003e","description":"","filename":"6.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/b1df7476c21e647b390cb7ff.png"},{"id":85181046,"identity":"6cee8967-b345-48a1-ba64-42e0ea20bbee","added_by":"auto","created_at":"2025-06-23 07:22:31","extension":"png","order_by":7,"title":"Figure 7","display":"","copyAsset":false,"role":"figure","size":1274100,"visible":true,"origin":"","legend":"\u003cp\u003eFigure 6. Performance based subgroups and reading outcomes in first grade children (n = 241) mirror kindergarten profiles with developmental shifts. \u003cstrong\u003ea\u003c/strong\u003e. Latent profile analysis reveals five subgroups consistent with the Kindergarten cohort; \u003cstrong\u003eb\u003c/strong\u003e. Distribution of WJ–LWI scores at follow-up one year later, showing continued stratification in reading ability; \u003cstrong\u003ec\u003c/strong\u003e. Longitudinal reading trajectories for each child, with group-level means shown in bold; \u003cstrong\u003ed–g\u003c/strong\u003e. density plots display subgroup differences in basic reading (d), broad reading (e), reading composite scores (f), and reading fluency (g).\u003c/p\u003e","description":"","filename":"7.png","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/6b1846634a8e1b1499007f74.png"},{"id":89473476,"identity":"c20852a1-70b2-430e-8807-bf1a4f116e40","added_by":"auto","created_at":"2025-08-20 09:59:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":6075092,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/11a89f0d-cbee-4a6a-8c67-5747f4795b58.pdf"},{"id":85180427,"identity":"4ba500c4-ddf1-4c3f-b2c8-d1cdcd50a511","added_by":"auto","created_at":"2025-06-23 07:14:31","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":1426162,"visible":true,"origin":"","legend":"Supplementary Material: The link between vision and reading: A language-agnostic window into heterogeneity in early reading development.","description":"","filename":"SupplementaryMaterial.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/a63d9f7b50af4cb6ed3e51d7.pdf"},{"id":85179031,"identity":"e8053c65-f25f-439c-a060-d970bbbe057f","added_by":"auto","created_at":"2025-06-23 07:06:31","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":19341,"visible":true,"origin":"","legend":"","description":"","filename":"Table3.docx","url":"https://assets-eu.researchsquare.com/files/rs-6804845/v1/40343d79e73b14b6f1700959.docx"}],"financialInterests":"There is \u003cb\u003eNO\u003c/b\u003e Competing Interest.","formattedTitle":"\u003cp\u003eThe Link Between Vision and Reading: A Language-agnostic Window into Heterogeneity in Early Reading Development\u003c/p\u003e","fulltext":[{"header":"Main","content":"\u003cp\u003eTheories linking deficits in visual processing in individuals with dyslexia date back to the 1850s, when dyslexia was first conceived as a visual problem and was referred to as \u0026lsquo;congenital word blindness\u0026rsquo; in the first case study published\u003csup\u003e1\u003c/sup\u003e. Ever since, myriad theories have been proposed linking deficits in visual processing to reading development \u003csup\u003e2\u0026ndash;6\u003c/sup\u003e. Work by Lovegrove and colleagues in the 1980s showed that people with dyslexia have challenges in visual processing of transient stimuli \u003csup\u003e7\u003c/sup\u003e. Subsequent studies revealed deficits in visual sensitivity to transient and moving stimuli across a wide range of experimental conditions \u003csup\u003e8\u0026ndash;12\u003c/sup\u003e. In the 1990s, physiological evidence came from Livingstone and colleagues\u003csup\u003e13\u003c/sup\u003e who studied the postmortem brains of five individuals with dyslexia and compared it to five Controls and reported that cell bodies in the magnocellular layers of the lateral geniculate nucleus (LGN) were fewer in terms of density and were approximately 27% smaller in dyslexic brains compared to control brains. Together, these observations led to the magnocellular theory of dyslexia \u003csup\u003e4,13\u003c/sup\u003e, which posits that difficulties in learning to read are the consequence of a low-level deficit in the magnocellular visual pathway, and that this deficit can be detected with either physiological responses \u003csup\u003e12\u0026ndash;15\u003c/sup\u003e or psychophysical thresholds \u003csup\u003e9\u0026ndash;11,16,17\u003c/sup\u003e to rapid, transient or moving stimuli. A number of studies that followed have failed to replicate the relationship between reading skills and motion processing \u003csup\u003e18\u0026ndash;22\u003c/sup\u003e. Importantly, intervention studies revealed that intense training in reading neither improved visual motion sensitivity\u003csup\u003e10\u003c/sup\u003e nor did it hinder one\u0026rsquo;s ability to show improvements in reading \u003csup\u003e23\u003c/sup\u003e, leaving this as the most contested hypothesis in the field, till date. Later researchers reformulated the magnocellular hypothesis in more general terms, since the M-pathway reaches the dorsal striate cortex, implicating a more general issue with attention \u003csup\u003e24,25\u003c/sup\u003e. In the last decade, there has been growing interest in the visual-spatial attention hypothesis, which posits that people with dyslexia shift spatial attention more slowly than skilled readers. This hypothesis fits the \u0026ldquo;sluggish attention shift hypothesis,\u0026rdquo; which is purported to explain various other sensory deficits reported in children with dyslexia across visual and auditory domains when processing rapidly presented information \u003csup\u003e26\u0026ndash;29\u003c/sup\u003e\u003csup\u003e,\u003c/sup\u003e\u003csup\u003e30\u003c/sup\u003e. \u003c/p\u003e\n\u003cp\u003eDespite five decades of research, the field lacks clear, empirically validated visual tasks that can be reliably administered to younger children (ages 5\u0026ndash;6) \u003csup\u003e2\u0026ndash;6\u003c/sup\u003e at scale. On one hand, much of the existing literature is based on small- and homogenous- sample studies where samples are not sufficiently large and diverse to discern the contribution of different risk factors, with only a couple of studies exceeding N \u0026gt; 100 \u003csup\u003e17,31,32\u003c/sup\u003e. On the other hand, there is growing consensus that reading development is heterogeneous shaped by multiple, interacting risk factors. Differences in temporal visual attention, visual crowding, motion sensitivity, and oculomotor control have all been linked to early reading outcomes, independent of language ability \u003csup\u003e33\u0026ndash;35\u003c/sup\u003e. Similarly, variability in auditory processing\u0026mdash;such as rapid auditory discrimination and temporal sensitivity\u0026mdash;has been shown to contribute to individual differences in phonological development and reading acquisition \u003csup\u003e36,37\u003c/sup\u003e. These findings suggest that early reading trajectories are shaped by more than just phonological processing abilities. But capturing this heterogeneity in practice, specifically early in development, requires robust, scalable tools that can be reliably administered across diverse populations early in their reading development. \u003c/p\u003e\n\u003cp\u003eA critical step towards understanding heterogeneity in early reading development is the systematic implementation of reliable and equitable screening tasks administered in school settings. Such an approach is particularly powerful because it not only reaches a large, diverse, and representative population but also enables longitudinal tracking of how early performance in sensory and cognitive tasks relates to later reading outcomes. The growing recognition that early deficits in visual processing might confer risk for dyslexia is reflected in recent dyslexia screening legislation across different states in the United States of America. Developmental dyslexia, with a global prevalence of 7\u0026ndash;17% \u003csup\u003e2,38\u003c/sup\u003e substantially impacts children\u0026apos;s educational, socioemotional, and subsequent socioeconomic outcomes \u003csup\u003e6\u003c/sup\u003e. This widespread occurrence across languages and cultures underscores the necessity for universal early screening tools. Reading acquisition not only represents a complex neurobiological developmental process that depends on cognitive, linguistic, and sensory neural systems⁹, but is significantly shaped by environmental and instructional context that the child experiences \u003csup\u003e39,40\u003c/sup\u003e. Consequently, building effective screening tools require substantial investment in developing age-, culture-, and language-appropriate \u003csup\u003e41\u0026ndash;43\u003c/sup\u003e, assessment items that demonstrate no socioeconomic or linguistic bias upon implementation. Such equitable screening instruments provide significant value in elucidating the heterogeneity in reading development by facilitating the identification and characterization of distinct cognitive profiles underlying reading difficulties. \u003c/p\u003e\n\u003cp\u003eDespite well-documented measurement biases in various cognitive and linguistic processes\u0026sup1;⁰⁻\u0026sup1;\u0026sup2;, the relationship between visual processing and reading acquisition remains inadequately characterized at the population level. Over the past five decades, two candidate visual tasks have emerged as central to this debate. The first, global motion coherence task (Motion), stems from the magnocellular theory of dyslexia, this theory was initially supported by studies demonstrating impaired motion sensitivity in individuals with dyslexia and was tested using both physiological \u003csup\u003e12\u0026ndash;15\u003c/sup\u003e and psychophysical approaches \u003csup\u003e9\u0026ndash;11,16,17\u003c/sup\u003e. However, replication failures \u003csup\u003e18\u0026ndash;22\u003c/sup\u003e and theoretical reformulations \u003csup\u003e24,25\u003c/sup\u003e have since shifted the focus toward broader attentional mechanisms \u003csup\u003e26\u0026ndash;29\u003c/sup\u003e\u003csup\u003e,\u003c/sup\u003e\u003csup\u003e30\u003c/sup\u003e. The second candidate task, multi-element processing (MEP), has gained traction more recently and centers on the ability to rapidly extract information from a string of visual symbols\u0026mdash;a skill ecologically linked to reading. In this study we administered the MEP task that was optimized for reliability in kindergarten and first-grade children (see \u003csup\u003e44\u003c/sup\u003e) and the Motion task adapted from a previous study \u003csup\u003e31\u003c/sup\u003e and tailored for administration to kindergarten and first grade children (see Methods). While both tasks are widely studied, their utility in identifying reading-related risk remains unclear, especially in large, diverse, representative samples.\u003c/p\u003e\n\u003cp\u003eTo address this gap in understanding the role of visual processing abilities in early reading development, our goal is to investigate whether developmentally appropriate, scalable visual processing measures\u0026mdash;used alongside state-of-the-art language-based tasks\u0026mdash;can predict reading difficulties independent of a child\u0026apos;s home or instructional language\u0026sup1;⁵⁻\u0026sup1;⁹. We aim to explore the predictive validity of these visual measures, assess their equity across diverse linguistic backgrounds, and uncover heterogeneity in reading outcomes. To address these aims, in this study, we leveraged the state of California-funded Multitudes initiative to take some theory driven, carefully validated\u003csup\u003e44\u003c/sup\u003e, visual measures to a large and diverse sample of kindergarten and first graders in California public schools. Alongside the visual measures, children were administered conventional language-based measures followed by standardized reading assessments at the end of the school year. \u003c/p\u003e"},{"header":"Results","content":"\u003ch2\u003eVisual measures are agnostic to a child\u0026rsquo;s primary language and socioeconomic status.\u003c/h2\u003e\n\u003cp\u003eWe first asked if children in kindergarten and first-grade exhibit performance differences in the visual-processing measures across a) their eligibility for free or reduced-price meals (an indicator of socioeconomic status) and b) reported primary home language. Our motivation to investigate this question stems from the fact that socioeconomic status impacts educational opportunity and academic outcomes \u003csup\u003e39,45\u003c/sup\u003e. A child\u0026rsquo;s financial and social resources are among the strongest predictors of performance on reading assessments. Children growing up in lower socioeconomic environments have lower scores on common measures of reading, language, and executive function compared to children growing up in higher-resourced environments \u003csup\u003e46\u0026ndash;49\u003c/sup\u003e. The socioeconomic disparity in elementary school reading scores has grown by over 40% in the second half of the 20th century \u003csup\u003e50\u003c/sup\u003e. Screening measures often reflect various biases, resulting in unequal performance in diverse school settings \u003csup\u003e51,52\u003c/sup\u003e. Oftentimes these measures reflect disparities in a child\u0026rsquo;s experience prior to formal schooling and are not informative in terms of potential neurobiological intervention targets \u003csup\u003e53,54\u003c/sup\u003e. So our first goal was to investigate if these visual measures exhibited bias in :\u003c/p\u003e\n\u003cp\u003e\u003cem\u003ea. Free and reduced-price meal (FRPM):\u003c/em\u003e The National School Lunch Program (NSLP) provides free meals for children whose household income is less than 130% of the poverty line and reduced-price meals for income between 130%\u0026ndash;185% of the poverty line. FRPM metrics are highly correlated with county poverty rates (0.91 - 0.95) \u003csup\u003e55\u003c/sup\u003e. FRPM eligibility data is a student-level indicator of home income. We compared performance across students who were eligible or ineligible for FRPM. Ineligibility indexes higher socioeconomic status and was coded as 0 in our dataset.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eb. Student\u0026rsquo;s primary language:\u003c/em\u003e All students whose parents had indicated to the school that English is the only language spoken at home were categorized as English-only (coded as EO). Those students labeled as English learners (coded as EL) whose reported primary language was Spanish were classified as Spanish speakers. These two groups combined comprised ~90% of students in the dataset.\u003c/p\u003e\n\u003cp\u003eThe correlation between FRPM eligibility and a student\u0026rsquo;s primary language was 0.267 (phi coefficient for binary predictors). We examined how students across different grades performed in the visual tasks and found no group differences in terms of FRPM (Figure 1) or primary language (Figure 2). We performed a \u003cem\u003et\u003c/em\u003e-test to compare across groups and adjusted the significance threshold for multiple comparisons [Bonferroni adjusted p threshold: 0.004 (0.05 (p-cutoff) / 12 comparisons)]. Figure 1a shows how children from the FRPM Eligible and Ineligible groups perform across all visual tasks (first column). This can be visually compared to group differences in reading outcome measure (presented second column). We observed no significant difference in group performance between the two FRPM groups for the visual measures but a large effect size (Cohen\u0026apos;s d) is observed for the reading outcome measure (WJ-LWI\u0026mdash;see Methods for details). Figure 1b further shows how the effect size of the difference between the Eligible and Ineligible groups compares across visual, reading, and language-based measures.\u003c/p\u003e\n\u003cp\u003eThe number of participants with student-level data was about half the sample size for each measure because new regulation makes it more difficult for schools to share student-level data, so we also used the openly available school-level data on percentage of students with FRPM eligibility. The school-level FRPM eligibility data was available for the school year 2022-2023 from the California Department of Education \u003csup\u003e56\u003c/sup\u003e. We fit a linear model to the school-level median scores regressed on these measures with percentage of FRPM eligibility reported for each school and grades as predictors. As shown in Figure 1c, the coefficients of FRPM as predictor was close to zero for the visual measures compared to other measures, most evidently for the reading outcome measures as substantiated in Figure 1a.\u003c/p\u003e\n\u003cp\u003eSince English learners in the United States (and California specifically) are predominantly Spanish speakers \u003csup\u003e57\u003c/sup\u003e, we next compared children whose reported primary spoken-language at home was English with those who reported speaking Spanish. Figure 2a shows how English and Spanish-speaking children perform across all visual tasks (first column) and reading outcome measure (second column). Children as young as kindergarten exhibited no group differences in task performance across all visual-processing measures. This is in stark contrast to the Woodcock-Johnson reading outcome measures in our dataset, which showed group differences as early as kindergarten. First-grade children showed a significant group difference only in the letter version of the MEP task. Since the MEP-L task involves letter knowledge, it is possible that factors like pre-school attendance, home environment and parental education could mediate the group difference we notice between the English and Spanish proficient groups by first grade. No such group difference was noticed in the pseudo-letter version of the task. Figure 2b, shows how the effect size of the difference between the English and Spanish groups compare across visual, reading, and language measures. Visual-processing measures are unbiased to differences in the primary language of the participants for kindergarten and the non-alphabetic visual tasks are unbiased to differences in the primary language of the participants in first grade. Note that all the visual tasks were administered in English but children were screened for task understandability (see Methods). \u003c/p\u003e\n\u003cp\u003eSimilar to Figure 1c, we fit a linear model to the school-level median scores regressed on these measures with groups based on the primary language of participants from each school and grade as predictors. As shown in Figure 2c, the coefficients of primary language as predictor was close to zero for the model predicting visual-processing measures compared to other measures, most evidently for the reading outcome measure (WJ-LWI) as substantiated in Figure 2a as boxplots. \u003c/p\u003e\n\u003cp\u003eTogether, we show that visual-processing measures, notably, the pseudo-letter version of the MEP task (MEP-P) and the global motion coherence tasks (Motion), which are non-alphanumeric visual tasks, show no group differences based on FRPM and primary language across kindergarten and first graders. This is crucial to establish, as we next aim to understand the correlation between these visual-processing measures and reading outcomes without the influence of environmental factors that typically affect reading assessments. By confirming the absence of group differences in these visual tasks, we can more confidently examine their relationship to reading performance. \u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003e\u003cem\u003eThe relationship between visual measures and reading outcome is equivalent across groups.\u003c/em\u003e\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eWe then asked how performance in the visual measures relates to the end-of-year reading outcome (WJ-LWI\u0026mdash;see Methods for details) for each grade. The development of visual processing abilities precedes formal reading instruction, so we hypothesize that individual differences in visual processing ability could account for individual differences in reading outcome as early as kindergarten, independent of SES and home language. To address this question, we first filtered those students who took all the visual measures and had end-of-the-year reading outcome measure (n= 539) and then fit two linear regression models. The first model regresses the reading outcome on all three visual measures and their interactions with FRPM eligibility,\u003c/p\u003e\n\u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eDo children with different FRPM eligibility show different associations between visual measures and reading outcomes? \u003c/em\u003e\u003c/p\u003e\n\u003cp\u003eWe investigated the relationship between visual-processing measures and reading outcome in kindergarten students (\u003cem\u003en\u003c/em\u003e = 108) and first grade (\u003cem\u003en \u003c/em\u003e= 240), for whom FRPM data was available. Across both grades, we fit two linear regression models, i) a visual-only model (we included a model with just the non-alphabetic visual tasks presented in Table 1) and ii) a visual with FRPM eligibility interactions as summarized in Table 1 and correlations are presented in Figure 3. \u003c/p\u003e\n\u003cp\u003eFirst, the model including MEP-L, MEP-P, and Motion as predictors explained 18.22% of the variance in reading outcome in kindergarten and 9.42% in first grade. Only MEP-L performance was a significant predictor for kindergarten (\u003cem\u003e\u0026beta;\u003c/em\u003e = 0.218, \u003cem\u003ep\u003c/em\u003e = 0.0253) and first grade (\u003cem\u003e\u0026beta;\u003c/em\u003e = 0.307, \u003cem\u003ep\u003c/em\u003e = 0.00063). This is because performance in the MEP-L and MEP-P tasks are highly correlated (\u003cem\u003er\u003c/em\u003e = 0.7) and correlated with the motion task (\u003cem\u003er\u003c/em\u003e = 0.3); however, each measure is a significant univariate predictor of reading outcome across both grades (see Supplementary Table S1). An ANOVA comparison between the visual-only model and the model with visual and FRPM interactions demonstrated that including the FRPM eligibility improved model fit for both kindergarten [\u003cem\u003eF\u003c/em\u003e(4,100)= 3.930, \u003cem\u003ep\u003c/em\u003e = 0.00526] and first graders [\u003cem\u003eF\u003c/em\u003e(4, 232) = 5.396, \u003cem\u003ep\u003c/em\u003e = 0.00035]. \u003c/p\u003e\n\u003cp\u003eThe lack of significant interactions indicates that the relationship between visual measures and reading outcomes does not vary based on FRPM eligibility. While children with different FRPM eligibility did not differ significantly in their visual processing abilities, FRPM eligibility emerged as a strong predictor of reading outcome. This highlights the importance of socioeconomic factors in early reading development, independent of visual processing skills and its ineluctable role in reading development. \u003c/p\u003e\n\u003cp\u003eTable 1. Regression models with visual measures and FRPM as predictors of end-of-year reading outcome. For each grade i. a visual-only model, ii. a model with just the non-alphabetic visual tasks and iii. a model with visual tasks and FRPM eligibility as predictors is presented. \u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"624\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\"\u003e\n \u003cp\u003eKindergarten (n =108)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\"\u003e\n \u003cp\u003eFirst Grade (n=240)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L + MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L * FRPM + MEP-P * FRPM + Motion* FRPM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L + MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L * FRPM + MEP-P * FRPM + Motion* FRPM\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e(Intercept)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.084 (0.069)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.105 (0.070)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.279 (0.083)**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.141 (0.068)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.167 (0.069)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.122 (0.087)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.218 (0.096)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.138 (0.107)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.307 (0.089)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.341 (0.111)**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-P\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.032 (0.093)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.174 (0.070)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.041 (0.101)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.041 (0.085)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.211 (0.071)**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.043 (0.102)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMotion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.112 (0.068)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.149 (0.068)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.121 (0.080)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.040 (0.074)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.064 (0.076)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.002 (0.089)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eFRPM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.530 (0.136)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.589 (0.134)***\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-L * FRPM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.242 (0.211)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.083 (0.175)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-P * FRPM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.008 (0.211)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.164 (0.173)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMotion * FRPM\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.055 (0.139)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.052 (0.152)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMultiple R\u0026sup2;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.182\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.142\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.293\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.094\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.048\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.171\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eF-statistic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e7.73 (3,104)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e8.67 (2,105)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e5.93 (7,100)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e8.19 (3,236)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.01 (2,237)**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.86 (7,232)***\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"7\"\u003e\n \u003cp\u003eStandard errors are in parentheses. Significance levels: *** p\u0026lt;0.001, ** p\u0026lt;0.01, * p\u0026lt;0.05, . p\u0026lt;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cem\u003eDo children with different primary languages show different associations with reading outcome?\u003c/em\u003e \u003cbr\u003eWe next investigated the relationship between visual-processing measures and reading outcome across different primary language groups in kindergarten (n=157) and first-grade (n=381). We fit two linear regression models, i) a visual-only model (we included a model with just the non-alphabetic visual tasks presented in Table 2) and ii) a visual with primary language interactions as summarized in Table 2, and correlations are shown in Figure 4. The visual-only model explained 16.3% of the variance in reading outcome measure in kindergarten and 11.9% in first grade. Only MEP-L performance was a significant predictor for kindergarten (\u0026beta; = 0.212, p = 0.0107) and first grade (\u0026beta; = 0.316, p =2.51x10\u003csup\u003e-6\u003c/sup\u003e). An ANOVA comparison between the visual-only model and the model with visual and primary language interactions demonstrated that including the Primary language improved model fit for both kindergarten [\u003cem\u003eF\u003c/em\u003e(4,149)= 2.993, \u003cem\u003ep\u003c/em\u003e =0.021] and first graders [\u003cem\u003eF\u003c/em\u003e(4, 373)= 7.896, \u003cem\u003ep \u003c/em\u003e=4.058x10\u003csup\u003e-06\u003c/sup\u003e].\u003c/p\u003e\n\u003cp\u003eThe lack of significant interactions indicates that the relationship between visual-processing measures and reading outcomes does not vary between English and Spanish speakers. Primary language, however, emerged as a strong predictor of reading outcome measure. \u003c/p\u003e\n\u003cp\u003eTable 2.Regression models with visual measures and child\u0026rsquo;s primary language as predictors of end-of-year reading outcome. For each grade i. a visual-only model, ii. a model with just the non-alphabetic visual tasks and iii. a model with visual tasks and child\u0026rsquo;s home language as predictors is presented. \u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"616\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\"\u003e\n \u003cp\u003eKindergarten (n = 157)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"3\"\u003e\n \u003cp\u003eFirst Grade (n=381)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eVariable\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L + MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L * Primary Language + MEP-P * Primary Lang + Motion* Primary Language\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L + MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-P + Motion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eReading outcome ~ MEP-L * Primary Language + MEP-P * Primary Language + Motion* Primary Language\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e(Intercept)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.102 (0.060).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.121 (0.060)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.333 (0.092)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.037 (0.052)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.049 (0.053)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.303 (0.080)***\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-L\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.212 (0.082)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.173 (0.130)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.317 (0.066)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.293 (0.100)**\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-P\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.063 (0.079)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.194 (0.061)**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.017 (0.126)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.062 (0.068)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.252 (0.057)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.011 (0.107)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMotion\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.109 (0.061).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.137 (0.062)*\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.167 (0.086).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.043 (0.055)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.065 (0.057)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.098 (0.079)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003ePrimary Language \u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.386 (0.119)**\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.571 (0.104)***\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-L * Primary Language \u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.048 (0.165)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.044 (0.131)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMEP-P * Primary Language\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.142 (0.159)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.084 (0.136)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMotion * Primary Language\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.087 (0.120)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-0.107 (0.107)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eMultiple R\u0026sup2;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.163\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.126\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.225\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.119\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.066\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.188\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003eF-statistic\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e9.894 (3,153)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e11.09 (2,154)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.165 (7,149)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e17.00 (3,377)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e13.31 (2,378)***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e12.33 (7,373)***\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"7\"\u003e\n \u003cp\u003eStandard errors are in parentheses. Significance levels: *** p\u0026lt;0.001, ** p\u0026lt;0.01, * p\u0026lt;0.05, . p\u0026lt;0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eOverall, visual-processing measures accounted for ~16\u0026ndash;18% of variance in reading outcome in kindergarten and ~10\u0026ndash;12% in first grade. \u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003e\u003cem\u003eVisual measures are important predictors of reading difficulties.\u003c/em\u003e\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eWe next examined the potential application of the observed relationship between visual processing and reading outcomes, focusing on its utility for early screening and its ability to illuminate the heterogeneity in reading development. Two important considerations guide this analysis: (i) reading difficulties are not primarily visual processing disorders, although visual deficits may co-occur and hold predictive value; and (ii) robust evaluation of early visual markers requires a multi-year longitudinal design and a sufficiently large sample. We first built a risk prediction model with visual processing measures and language-based measures administered during the winter quarter of the school year. Gold-standard reading outcomes (Woodcock Johnson Letter Word Identification, WJ-LWI) were assessed at the end of the same school year (Spring 2023), with a subset of children retested one year later (Spring 2024). Risk is defined as children with scores on outcome measures below the 20th percentile. \u003c/p\u003e\n\u003cp\u003eThe sample sizes of children with all measures varied across years. Concurrent reading outcome measures (Spring 2023) were available for 108 kindergarteners and 216 first grade children. Longitudinal follow-up (Spring 2024), were available for K: 53, G1: 112. We performed multiple imputations to address missing data within language, visual and reading outcome domains before merging imputed datasets. The final sample size, after imputations, was K: 157 and G1: 241 (see Methods). In our sample, only ~20% of children were identified as at risk (Kindergarten: 32/157; Grade 1: 51/241), resulting in a notable class imbalance. In such cases, traditional classification metrics like sensitivity and specificity become less informative, as they are highly sensitive to class distribution and can yield unstable or misleading results in small samples. Importantly, class imbalance is an expected feature when the condition of interest\u0026mdash;here, reading risk\u0026mdash;has a relatively low base rate, and it does not compromise the validity of the predictors themselves. Given our current sample size, we intentionally chose not to emphasize on sensitivity and specificity metrics. As we continue to add cohorts and our sample grows, the absolute number of at-risk children will increase, allowing for more robust estimation of these metrics. Instead, our primary aim was to evaluate the predictive contribution of visual measures relative to conventional language-based predictors.\u003c/p\u003e\n\u003cp\u003eUsing a random forest classification model with leave one out cross validation (LOO-CV), we evaluated the combined predictive utility of the visual and language-based measures for identifying children at risk of word reading difficulties (WCJ-LWI \u0026lt; 20th percentile were classified as at risk, based on a previous study\u003csup\u003e58\u003c/sup\u003e). For concurrent-year risk classification the model achieved accuracies of 87.96% (kindergarten) and 85.19% (first grade). Longitudinal prediction accuracies were 79.73% (kindergarten) and 80.32% (first grade). Given the imbalance in risk prevalence, we focus primarily on feature importance rankings to determine the unique predictive value of visual processing measures. \u003c/p\u003e\n\u003cp\u003eTo interpret how individual measures contributed to the predictions, we computed feature importance using the Boruta package in R \u003csup\u003e59\u003c/sup\u003e. Feature importance plots (Figures 4a-d) quantify the relative importance of each predictor - such as MEP, deletion, blending etc., - on the model\u0026rsquo;s ability to classify children at risk. Higher median importance values indicate that the measure was frequently used to split the decision nodes across the different decision trees in the random forest and thereby consistently improves prediction accuracy when included in the prediction model. These findings offer mechanistic insights into the cognitive processes that are most relevant to the reading outcome at developmentally different grade levels. Our results revealed visual measures, particularly MEP, as consistently important predictors across both time points and grades. \u003c/p\u003e\n\u003cp\u003eIt should be noted that since we have fewer children at-risk, models may prioritize variables that simply distinguish the majority-class patterns thereby skewing feature importance estimates. To verify that our findings reflected true patterns rather than imbalance artifacts, we replicated our analysis with class-balanced datasets using the ROSE package in R \u003csup\u003e60\u003c/sup\u003e that simulated a balanced dataset by synthetically oversampling minority-class instances (at-risk) and undersampling the majority class. Supplementary Figure S2 is identical to Figure 4 but, plotted with the simulated balanced data set. \u003c/p\u003e\n\u003cp\u003eWe found that the visual measures, especially MEP, emerged as an important and stable predictor of risk, in relation to the conventional language-based measures across both grades. This pattern held true for risk predicted at the end of the school year as well as at the one-year follow-up. \u003c/p\u003e\n\u003cp\u003eTo further validate these findings, we conducted a complementary analysis identifying predictors that maximized area under the curve (AUC) and accuracy (Table 3). We prioritized AUC over sensitivity and specificity metrics, as AUC provides a threshold-independent measure of model performance that is particularly well-suited to imbalanced datasets like ours. Risk status was computed as below the 20th percentile for each reading outcome measure (see Methods for a detailed description of composite scores here). The AUC-accuracy profile presented in Table 3, reinforces the role of MEP-L as one of the key predictors consistent with the ranking in the feature importance plots in Figure 4. This aligns with their prominence in feature importance analyses, reinforcing the role of visual processing measures in predicting reading risk. MEP-P, however, became a stable predictor in the absence of MEP-L (Supplementary Table S2), likely due to their high disattenuated correlation (r = 0.91 \u003csup\u003e44\u003c/sup\u003e). The stability of visual measures\u0026rsquo; importance, especially MEPs, in both imbalanced and balanced scenarios (Figures 4 vs. Supplementary Figure S2) confirms their robustness as predictors, independent of class distribution. \u003c/p\u003e\n\u003ch2\u003e\u003cstrong\u003e\u003cem\u003eLatent profile analysis reveals subgroups of children with specific visual strength and specific visual challenge. \u003c/em\u003e\u003c/strong\u003e\u003c/h2\u003e\n\u003cp\u003eTo investigate heterogeneity in reading outcomes, we first looked at performance profiles across language-based and visual measures, to identify subgroups of children, guided by latent profile analysis (LPA). Bayesian Information Criterion (BIC)-optimized LPA, was implemented via the \u003cem\u003eMclust\u003c/em\u003e package in R \u003csup\u003e61\u003c/sup\u003e, on the kindergarten data (n=157). This revealed five latent profiles as the maximum number of subgroups based on the BIC criterion and such that each cluster has a sample size that is more than 10% of the total number of kindergarteners (Supplementary Figure S3). Subgroup criteria derived from the LPA solution on Kindergarten children were subsequently applied to define profiles consistently across children from both grades (see Methods). \u003c/p\u003e\n\u003cp\u003eFigure 5a shows the subgroup performance profiles. As is common with LPA, three of the subgroups represented typical performance differences: (1) a high performing group that performed ~0.25 standard deviations above the mean on all tasks, (2) an average performing group that performed around the mean, and (3) a low performing group that performed ~0.25 standard deviations below the mean. Beyond these groups that differed in performance across all tasks, two interesting subgroups emerged: children with high performance in MEP tasks while demonstrating lower abilities in other domains\u0026mdash;the Specific Visual Strength group (10% of total kindergarten children)\u0026mdash;and children with low performance on MEP tasks who exhibited high performance across other domains\u0026mdash;the Specific Visual Challenge group (15% of total kindergarten children). The Specific Visual Strength group showed longitudinal reading outcomes comparable to the High-Performing group, indicating that strengths with MEP can be a protective factor for reading development. The Specific Visual Challenge group, on the other hand, showed longitudinal reading outcomes close to the average-performing group, despite the fact that they had strong language skills (see Figure 5b). \u003c/p\u003e\n\u003cp\u003eFigure 5c shows the reading outcome (WCJ-LWI) trajectories across time \u0026mdash; from concurrent-year to the one year follow up. A linear mixed effects model [specified as: reading outcome ~ Time * Subgroups + (1 | student_id)] revealed a significant main effect of time, \u003cem\u003eF\u003c/em\u003e(1, 152) = 23.720, \u003cem\u003ep\u003c/em\u003e = 2.769x10\u003csup\u003e-6\u003c/sup\u003e, indicating overall improvement in reading outcome across the year. There was also a significant main effect of subgroups, \u003cem\u003eF\u003c/em\u003e(4, 152) = 6.80, \u003cem\u003ep \u003c/em\u003e= 4.608x10\u003csup\u003e-05\u003c/sup\u003e, with high-performers and specific visual strength groups showing significantly higher baseline scores than low-performers (\u003cem\u003ep\u003c/em\u003e = 0.000489 and \u003cem\u003ep\u003c/em\u003e = 0.0418, respectively). Although the time \u0026times; group interaction was not significant (\u003cem\u003eF\u003c/em\u003e(4, 152)= 1.9403; \u003cem\u003ep\u003c/em\u003e = 0.107), estimated marginal means revealed that all subgroups\u0026mdash;except low-performers\u0026mdash;exhibited statistically significant reading gains over time. Importantly, initial disparities in reading performance persisted, with low-performers showing no measurable growth.\u003c/p\u003e\n\u003cp\u003eIn the first grade cohort (n= 241), similar trends were observed (see Figure 6a and 6b). Figure 6c shows the reading outcome (WCJ-LWI) trajectories across time\u0026mdash;from concurrent-year to the year after. Reading outcome improved significantly over the one-year period, \u003cem\u003eF\u003c/em\u003e(1, 236) = 39.937, \u003cem\u003ep\u003c/em\u003e =1.296x10\u003csup\u003e-09\u003c/sup\u003e and there were marked baseline differences among subgroups, \u003cem\u003eF\u003c/em\u003e(4, 236) = 15.027, \u003cem\u003ep\u003c/em\u003e = 5.880x10\u003csup\u003e-11\u003c/sup\u003e: low-performers \u0026lt; specific visual strength group \u0026lt; high-performers. Similar to Kindergarten, the interaction between time and group was not significant, \u003cem\u003eF\u003c/em\u003e(4, 236) = 1.071, \u003cem\u003ep\u003c/em\u003e = 0.372, suggesting that although all groups improved, their rates of growth were comparable and the initial group differences persisted over time. Notably, the Specific Visual Challenge group continued to demonstrate significantly lower reading outcomes over time despite improving in the WCJ-LWI standard score by ~7 (\u003cem\u003ep\u003c/em\u003e = .0102), whereas the Specific Visual Strength group maintained reading levels on par with Average Performers, reinforcing their more resilient developmental trajectory.\u003c/p\u003e\n\u003cp\u003eThese results show that children with strong language but poor visual processing abilities exhibited a persistent risk for lower reading development that would have been completely missed if the screening battery did not include visual processing measures. Similarly, children with strong visual processing abilities demonstrated comparable reading performance to High Performers suggesting that there might be compensatory mechanisms early in development that those with Specific Visual Strengths can leverage to build reading skills. The subgroup profiles in our dataset and their developmental trajectories suggest that these patterns emerge early and remain stable across the foundational years of reading acquisition, underscoring the need for early identification and targeted support tailored to distinct profiles.\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eDespite 40 years of controversy, the debate over whether individual differences in visual processing can explain individual differences in reading ability remains active \u003csup\u003e3\u0026ndash;5,7,34\u003c/sup\u003e. This might seem surprising in light of hundreds of studies that report visual processing deficits in children with dyslexia \u003csup\u003e31,33,34,37,62,63\u003c/sup\u003e. Progress has been hindered by small, homogenous samples. Despite best efforts to recruit representative samples, the families who respond to university recruitment efforts are rarely representative of regional or national diversity, yet it is precisely this heterogeneity that must be understood to develop appropriate screening tools and intervention strategies. This dual constraint\u0026mdash;methodological and conceptual\u0026mdash;has kept visual processing at the periphery of reading science, despite long-standing debate of its relevance in dyslexia and reading difficulties. Here, we addressed these limitations: the visual-processing measures were selected based on prior evidence of their ability to identify struggling readers missed by conventional language-based assessments and were rigorously validated through item response theory for use with young children in a school setting \u003csup\u003e44\u003c/sup\u003e. We then administered these measures to a diverse and representative sample of kindergarteners and first-graders in California public schools and showed that visual processing abilities predict variation in early reading development.\u003c/p\u003e\n\n\u003cp\u003eWe first examined whether the visual-processing measures exhibit bias related to students\u0026apos; home language or eligibility for free and reduced-price meals (an indicator of family income). This investigation addresses a fundamental challenge in early screening: many measures are biased by socioeconomic disparities. Consequently, even when measures demonstrate good predictive power for reading challenges, it often remains unclear whether this predictive ability stems from the measure assessing a construct fundamental to reading development or simply from capturing the same differences in educational opportunity that are known to influence reading outcomes \u003csup\u003e45\u003c/sup\u003e. We observed that visual-processing measures are invariant to socioeconomic status. \u003c/p\u003e\n\u003cp\u003eWe filtered for task understandability\u0026mdash;blinded from the demographics of the participating students, we first cleaned all the data from the MEP and Motion tasks based on a predetermined rubric for performance (see Methods). We then investigated if our visual measures were biased to the child\u0026apos;s home language. Our results show that in kindergarten children grouped by home language did not show differences in task performance for both the MEP and the Motion tasks. However, in first grade children whose home language was English performed significantly higher than the Spanish speakers only in the MEP-L (letter version of the task). These group differences, by first grade, may be mediated by factors such as preschool attendance, home environment, and parental education, given that the MEP-L task involves letter knowledge. We did not observe group differences in the pseudoletter version of the MEP task (MEP-P). Thus, our visual-processing measures, especially the non-alphanumeric based, are both scalable and equitable\u0026mdash; free from biases linked to language exposure at home or socioeconomic background. \u003c/p\u003e\n\n\u003cp\u003eThese visual-processing measures explained unique variance in reading outcomes and consistently emerged as important predictors across both feature extraction and AUC-based approaches. Our initial aim was to directly compare prediction models built exclusively with visual measures against those using state-of-the-art language-based tasks, to determine which children were commonly or uniquely identified as at-risk. However, such comparisons require substantially larger samples\u0026mdash;especially ensuring sufficient representation of children classified as at-risk. Until these sample sizes are achieved, a more practical and informative approach is to assess whether visual-processing measures remain consistently important predictors within models combining both language-based and visual predictors. \u003c/p\u003e\n\u003cp\u003eBoth our random forest classification model and our AUC-based analysis of predictor performance showed that MEPs are a consistent predictor of reading risk\u0026mdash;both at the end of the same school year and at a one-year follow-up. This pattern aligns with longstanding debates about the visual pathways that underpin reading. Global motion discrimination tasks, often invoked to probe magnocellular (M‑pathway) integrity, have yielded mixed evidence for a direct mechanistic role in reading acquisition \u003csup\u003e64,65\u003c/sup\u003e. By contrast, MEP tasks are ecologically closer to the demands of word reading: they index how efficiently a child can encode multiple elements within a single fixation\u0026mdash;precisely the perceptual bottleneck \u003csup\u003e66,67\u003c/sup\u003e that fluent reading must overcome. It is therefore unsurprising that MEP outperforms the Motion task in forecasting later reading outcomes, providing convergent support for the view that rapid, fixation‑bound visual encoding\u0026mdash;not low‑level motion processing\u0026mdash;is the more critical constraint on early reading development. \u003c/p\u003e\n\n\u003cp\u003eWe then examined the performance profiles across language-based and visual measures, to identify subgroups of children with distinct profiles. Five distinct profiles emerged based on task performance, three of the five profiles are common to most clustering analysis. The high-, average- and low performers were unique sets of children who in general had scores across language-based and visual tasks that were high-, average- and low. A year later, their median reading outcomes mirrored these tiers, with high performers achieving the highest median scores, low performers the lowest, and average performers falling in between. Interestingly our analysis revealed two novel subgroups of children: a Specific Visual Challenge group, whose reading outcomes lagged despite average to high language skills, and a Specific Visual Strength group, whose reading outcomes matched those of High-Performers despite low to average language skills. Without our visual measures, the Specific Visual Challenge group would have mimicked the High Performers in terms of their scores in the language-based tasks, making it hard to reconcile their poor reading outcomes a year later. In contrast, we found a subgroup of children with Specific Visual Strengths who mimicked the Low Performers in terms of their scores in the language-based tasks but exhibited reading outcomes comparable to the High Performers. \u003c/p\u003e\n\u003cp\u003eWe were curious to examine the FRPM eligibility and home language composition of these subgroups (see Supplementary Figure S4). First, based on socio-economic status, we did not observe marked distinctions across most subgroups. While we initially hypothesized the High Performers would predominantly come from higher SES backgrounds, all groups had substantial proportions of FRPM-eligible children, with the highest percentages appearing in the Low Performer group by Grade 1. In contrast, subgroup composition based on home language revealed notable differences. As expected, High Performers were predominantly English-speaking, whereas Low Performers were overwhelmingly Spanish-speaking, reinforcing the cumulative disadvantage faced by linguistically marginalized students and the fact that the reading outcome measures were all in English. Average Performers maintained a balanced language composition. \u003c/p\u003e\n\u003cp\u003eMost intriguing, however, were the home language dynamics within the two novel visually-defined subgroups. In Kindergarten, the Specific Visual Strength group largely consisted of Spanish-speaking children, suggesting visual strengths might offset early language-related disadvantages. Indeed, these Kindergarten Spanish-speaking children with visual strengths went on to achieve strong reading outcomes a year later (as shown in Figure 5b). Conversely, the Specific Visual Challenge group at kindergarten primarily consisted of English-speaking children\u0026mdash;a reversal of the typical demographic seen in struggling readers, suggesting that the visual challenge they face is independent of linguistic advantage. These children with visual challenges went on to show considerably poor reading outcomes a year later (Figure 5b). These patterns raise critical questions about the nature of the specific visual processing difficulties faced by these children and the kinds of early interventions that might mitigate their impact on future reading outcomes. Although addressing these questions lies beyond the scope of our current work, future studies that characterize the underlying visual deficits in these subgroups and evaluate tailored, early-stage interventions could yield critical insights\u0026mdash;both theoretically and practically\u0026mdash;into optimizing reading outcomes for all learners. \u003c/p\u003e\n\u003cp\u003eNotably by Grade 1, the Specific Visual Strength group was predominantly English-speaking children, and the Specific Visual Challenge group flipped to being predominantly Spanish-speaking similar in composition to Low-Performers, reversing the pattern observed in Kindergarten. This cross-sectional shift suggests that the protective influence of visual strengths may be most salient early in development for multilingual learners but may erode over time as reading demands escalate and mismatches between home and instructional language accumulate. Critically, children facing dual vulnerabilities\u0026mdash;limited proficiency in the language of instruction and reduced visual processing abilities\u0026mdash;appear disproportionately represented among those struggling with reading, underscoring the need for early, multidimensional screening approaches that go beyond language-focused metrics alone. \u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eBy identifying distinct subgroups of children within a general population, our study provides compelling empirical evidence that visual processing abilities contribute meaningfully to heterogeneity in early reading development. This has profound implications for how we design, implement, and interpret early screening programs. As these visual measures have now been adopted into two of the four universal screeners recommended by the state of California, forthcoming cohorts will offer valuable opportunities to evaluate the predictive utility\u0026mdash;particularly the sensitivity and specificity\u0026mdash;of various combinations of predictors. A central limitation of the current study is that the language-based measures administered showed considerable variability in empirical reliability (0.61\u0026ndash;0.97), as they were not originally designed with the linguistic and demographic diversity of California public schools in mind. This may have attenuated estimates of language skill contributions and limits the generalizability of findings based on these measures alone. However, as part of the state-funded Multitudes initiative, language assessments have since been redesigned and calibrated for California\u0026rsquo;s diverse learner population, complementing the equitable visual measures introduced here.\u003c/p\u003e\n\n\u003cp\u003eTogether, these results challenge long-standing assumptions about the role of vision in reading and lay the foundation for a reconceptualization of variability in reading development. By establishing a population-level framework for longitudinal follow-up, this work enables future studies to trace how visual and language-based risk factors interact over time. In doing so, this research lays the groundwork for a new generation of early identification tools\u0026mdash;ones that are inclusive, developmentally appropriate, and empirically grounded\u0026mdash;opening the door to precision interventions that can transform outcomes for young readers worldwide. \u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003e\u003cem\u003eParticipating school recruitment\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eParticipants were recruited by the University of California, San Francisco (UCSF) Dyslexia Center\u0026rsquo;s Multitudes project, an initiative supported by the State of California to develop a universal screener for reading challenges and dyslexia in California public schools. The sample was drawn from schools in locations and communities intended to be representative of the state in terms of race, ethnicity, socioeconomic status, and home language. Care was taken to reach remote rural, urban, and suburban communities to ensure that the sample was truly representative of the diversity of lived experiences in California. The study aimed to reduce selection bias with a passive consent process. The Institutional Review Board of UCSF determined that a process of informing parents about the study and how their children would be participating, with clear instructions on how to opt-out through communications with their school administrators, was appropriate for the research going into a universal screener. Parents of eligible kindergarten and first-grade students received an information sheet describing the project, which involved normal classroom activities and presented no more than minimal risk to participants. Children whose parents wished to opt out did not participate in study activities.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eGeneral Procedure.\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003eA team of 6-10 UCSF-trained proctors administered these assessments to kindergarten and first-grade students in each of the participating schools based on the school\u0026rsquo;s convenience. Reading outcome measures were individually administered to each student. In each school, kindergarten and first-grade children were brought in batches to complete a battery of language, reading, and visual measures. All children were assigned a unique student tracking ID and were randomly grouped in batches. Thus, missing data means that either a child was absent on the day or did not wish to participate. Demographics reported by the participating school consisted of reported primary language of the student, age, English proficiency designation, grade, as well as race and ethnicity information. The student-level data on FRPM eligibility reported in this study was discontinued after the 2022-2023 school year. This data was available only for a selected set of children from the schools that provided FRPM eligibility information. Supplementary Figure S5 shows the percentage of children reported as eligible compared to the eligibility percentage reported by the California Department of Education.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eBattery of assessments\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eStandardized Reading measures.\u003c/strong\u003e The Woodcock-Johnson IV Tests of Achievement were administered to all children at the end of the school year to assess key literacy skills through Letter-Word Identification, Word Attack, and Spelling tests. For this study, we used the letter word identification (WCJ-LWI) subtest as the reading outcome for both English and Spanish-speaking children. It should be noted these reading outcome measures in English are a measure of children\u0026rsquo;s classroom reading demands.\u003c/p\u003e\n\u003cp\u003eThe Letter-Word Identification test evaluates basic reading skills by measuring, untimed, word recognition and decoding abilities, it requires them to read and correctly identify letters and words that range from high-frequency to those of increasing difficulty. The LWI tests are integral components of the broader WCJ IV battery, designed to comprehensively evaluate single-word reading abilities and written language capabilities across various developmental stages \u003csup\u003e68\u003c/sup\u003e\u003csup\u003e,\u0026nbsp;\u003c/sup\u003e\u003csup\u003e69,70\u003c/sup\u003e\u003csup\u003e,\u0026nbsp;\u003c/sup\u003e\u003csup\u003e71\u003c/sup\u003e. To capture more comprehensive profiles of reading ability, we also examined composite scores from combining subtests derived from the WCJ IV battery:\u003c/p\u003e\n\u003cp\u003ei. \u003cem\u003eBasic Reading Skills:\u0026nbsp;\u003c/em\u003eComposed of the letter-word identification and word attack subtests, this composite evaluates fundamental decoding skills, including sight word recognition, phonemic decoding, and structural analysis. It reflects a child\u0026rsquo;s capacity to accurately read both familiar and novel words using a combination of visual and phonological strategies.\u003c/p\u003e\n\u003cp\u003eii. \u003cem\u003eBroad Reading composite:\u0026nbsp;\u003c/em\u003eThis composite integrates performance across letter-word identification, passage comprehension, and sentence reading fluency subtests. It provides a multidimensional index of reading achievement, encompassing decoding accuracy, reading fluency (rate), and comprehension. The composite score offers insight into how effectively students read connected text and extract meaning under time constraints.\u003c/p\u003e\n\u003cp\u003eiii. \u003cem\u003eReading composite:\u0026nbsp;\u003c/em\u003eComprising letter-word identification and passage comprehension, this composite encompasses decoding and reading comprehension (only administered to first graders). It captures the ability to both recognize individual words and understand meaning in sentences and brief passages, thereby offering a reliable estimate of core reading competence.\u003c/p\u003e\n\u003cp\u003eiv. \u003cem\u003eReading Fluency:\u0026nbsp;\u003c/em\u003eThis reflects performance in the sentence reading fluency subtest for first graders.\u003c/p\u003e\n\u003cp\u003eAcross all these reading outcome measures, it is important to acknowledge that for Spanish-speaking children, performance may partially reflect English knowledge or ongoing language acquisition rather than pure decoding or automaticity. This conflation of linguistic and reading skills is a known limitation when using English-based assessments in emergent bilingual populations and underscores the need for interpreting reading scores in light of children\u0026apos;s instructional language context and exposure.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReading Risk.\u003c/strong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003eWe chose the 20th percentile of the within-sample distribution of the WJ-LWI as the cut-off for risk classification (this was based on a previous study on the same dataset \u003csup\u003e58\u003c/sup\u003e).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eVisual-processing measures.\u003c/strong\u003e Included rapid visual processing of multiple elements of letters and pseudo letters and global motion processing ability. All visual measures were administered using a touchscreen Chromebook and all instructions were through an engaging audio-visual illustration of the task.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMulti-element Processing letters (MEP-L)\u003c/em\u003e \u0026amp; \u003cem\u003eMulti-element Processing pseudo-letters (MEP-P)\u003c/em\u003e: For task description, trial structure, design, and validation please see \u003csup\u003e44\u003c/sup\u003e. The assessment was gamified to make it more engaging and fun for younger children. The task simply involved a string of elements that were briefly presented for 240ms. After the elements disappeared a blue line appears below one of the element positions and participants were required to identify the element that was presented above the blue line from a set of six choices. The game sets children on a mission to help a lost whale in the sea, by playing this game of elements that is the secret map to the lost treasures and friends.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMotion\u003c/em\u003e: This experiment was built using JS psych and launched online on the ROAR app platform \u003csup\u003e72\u003c/sup\u003e. The stimuli for the motion task were adapted from a previous study \u003csup\u003e31\u003c/sup\u003e. In this single-interval forced-choice experiment, on each trial random dots with varying proportions of signal and noise were presented. The assessment was gamified to make it more engaging and fun for younger children. The task simply involves participants judging the global direction of motion of the entire dot field. The game sets children on a mission to help Beatrice the bee, in finding where the colony of bees is headed in a dense forest. If the colony of bees is headed to the apple tree they have to click on the apple tree on the left or the lemon tree on the right.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eTask understandability:\u003c/em\u003e For the visual measures, we wanted to distinguish task understandability from task performance. Blinded from the demographics of the participating students, we first cleaned all the data from the MEP and Motion tasks based on a predetermined rubric for performance.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMEP task filtering criterion\u003c/em\u003e: For the MEP tasks only children who attempted and engaged with at least 24 trials were considered for further analysis. In addition, those participants whose raw scores were below 4 correct out of the 24 attempted trials had ability estimates that were overestimated by the IRT model. These participants are highlighted in green in Supplementary Figures S1a and S1b and were excluded from further analysis. Supplementary Figure 1c shows the reading outcome distribution for the excluded children 115 (MEP-L) and 135 (MEP-P) confirming that we were not categorically excluding poor readers.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMotion task filtering criterion\u003c/em\u003e: This task had a non-response contingent trial succession, meaning trials are a fixed length time and the game automatically will progress to the next trial. Therefore, it is critical to define a performance cut-off that would inform us about the child\u0026rsquo;s task understandability. All participants who surpassed a performance cut-off of \u0026gt; 57% in the highest signal coherence conditions of 96% were filtered and included. This cut-off in performance was to verify that children understood the task and performed above chance in at least the easiest trials. This criterion excluded around 220 participants. The excluded participants reading outcome distributions were normal and we weren\u0026rsquo;t categorically eliminating poor readers (see Supplementary Figure S1d).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLanguage-based measures\u003c/strong\u003e.\u003cstrong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eA pilot version of the \u003cem\u003eReach Every Reader (RER) digital\u003c/em\u003e assessment battery \u003csup\u003e73\u003c/sup\u003e\u003cstrong\u003e\u003cem\u003e\u0026nbsp;\u003c/em\u003e\u003c/strong\u003ewith individual tasks that were widely used in screening was administered \u003csup\u003e74\u003c/sup\u003e. These tasks tapped into constructs of phonological awareness (deletion, blending), decoding (letter naming, word reading, non-word reading), phonological working memory (digit span, non-word repetition), rapid object naming, and oral language (expressive vocabulary, sentence repetition). We selected the Woodcock-Johnson Letter-Word Identification (WCJ LWI) subtest as our reading outcome measure, as it assesses word reading skills\u0026mdash;a core component of decoding, therefore the additional decoding measures were not included in the language model.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eAdministration and data collection of RER language-based measures:\u003c/em\u003e All language measures were administered on iPads with headphones. The administration involved a proctor-facing iPad and a student-facing iPad. Students were asked to wear headphones for instructions and prompts. All instructions were very engaging and child-friendly audio-visual illustrations of the task. Data were collected by UCSF proctors working individually with each child, ensuring the accuracy and efficiency of administration and data collection. Most measures were designed as computer adaptive testing and the platform returns ability estimates after each child completes the task. For measures that are not computer-adaptive like rapid automatic naming of objects, the platform recorded raw scores.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eProcedure:\u003c/em\u003e Phonological awareness was measured using the blending (BLE) and deletion (DEL) tasks. For the blending task, students join two phonemes or words to form a single word and remove one phoneme or word from a given word for the deletion task. Decoding skills were assessed using Word Reading (WRE) and Non-Word Reading (NRE), where students read the word or nonsense word presented on screen. Language and vocabulary skills were assessed with expressive vocabulary (EVO) and sentence repetition (SRT) tasks. EVO involved naming the visually presented object correctly and in SRT students repeated a set of sentences. Letter Naming context (LNC) and rapid automatic naming (RAO) skills were both time-sensitive measures and were scored based on the total number of correct responses in 45 seconds. In the RAO task, children were presented with a grid of 5x5 objects and a practice to name all the objects. The task is to name the objects in the 5x5 grid as quickly as possible. Similarly, for the LNC task, children were presented with random upper and lowercase letters and were asked to read out the grid of 5x5 letters as quickly as possible. Phonological working memory was assessed using the digit span (DGS) task and non-word repetition (NWR) tasks. In the DGS task, students repeated two-digit number sequences that incrementally went up to six digits. The task was discontinued after three consecutive incorrect responses. In the NWR task, students repeated a list of nonsense words. Reliability of language-based measures: The empirical reliability of the language measures used in this study ranged from .61 to .97. Most of the criterion-referenced reliability for the language measures ranged from .54 to .78. These numbers indicate moderate to high consistency across the majority of measures (see Appendix B Table B1 in a previous study \u003csup\u003e58\u003c/sup\u003e\u003cstrong\u003e).\u003c/strong\u003e\u003c/p\u003e\n\n\u003cp\u003e\u003cstrong\u003eData Imputations.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo address missing values and preserve statistical power across a wide range of predictors, we performed multiple imputation using the \u003cem\u003emice\u003c/em\u003e package in R \u003csup\u003e75\u003c/sup\u003e. Imputations were conducted separately for kindergarten and first grade cohorts before each analysis. Imputations were applied to distinct subtests of variables based on their domain: the domains were: i. language-based measures (deletion, bending, non-word repetition, sentence repetition, digit span, rapid automatic naming, evocative vocabulary); ii. rapid visual processing measures (MEP-L and MEP-P); iii. Motion (is its own domain); iv. reading outcome at the end of the same school year (WCJ: LWI, WA, ORF, RF, SRF) and v. reading outcome measures at the end of the next school year (WCJ: LWI, WA, ORF, RF, SRF, Spelling). Within each subset, missing values were imputed using predictive mean matching (PMM) and a fixed random seed to ensure reproducibility. After imputation, the subsets were merged using a common student identifier to generate a single, fully imputed dataset per grade. The imputed kindergarten and first grade datasets were then combined for all downstream analyses. Importantly, imputation was performed prior to the calculation of outcome-based classifications (e.g., risk status) to avoid introducing bias. This approach enabled the integration of diverse behavioral and cognitive data while preserving the natural variance structure and ensuring the reproducibility of the analysis pipeline\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFeature importance plots.\u0026nbsp;\u003c/strong\u003eTo identify which language-based and visual predictors most reliably distinguished children at risk for reading difficulties, we employed the Boruta feature selection algorithm \u003csup\u003e59\u003c/sup\u003e, a wrapper-based method built on random forests that iteratively compares the importance of actual features against permuted \u0026ldquo;shadow\u0026rdquo; features. This approach allows for the identification of all relevant features without relying on arbitrary thresholds or assumptions about linearity.\u003c/p\u003e\n\u003cp\u003e\u003cu\u003eData preparation:\u003c/u\u003e We selected a comprehensive set of predictor variables based on theoretical relevance to early reading development. These included measures of rapid visual processing (MEP\u0026ndash;L; MEP\u0026ndash;P; Motion), language skills (sentence repetition and expressive vocabulary), phonological awareness (deletion, blending), rapid automatized naming (rao), and phonological working memory (digit span and non-word repetition). Each variable was z-scored prior to analysis. The binary reading risk outcome was derived based on WCJ at the end of first grade.\u003c/p\u003e\n\u003cp\u003eTwo datasets were constructed: 1. An original (imbalanced) dataset with class imbalances of those at risk; 2. A synthetically balanced dataset generated using the ROSE (Random Over-Sampling Examples) package in R \u003csup\u003e60\u003c/sup\u003e, which synthesizes new data points by bootstrapping and interpolating between existing minority class observations to counteract the effects of class imbalance.\u003c/p\u003e\n\u003cp\u003eFeature selection was conducted separately on both the imbalanced and balanced datasets using the Boruta function in R with 100 random forest iterations and 100 trees per forest. The outcome variable was reading risk status, and all language-based and visual predictors were included as candidate features. Feature importance was assessed using Z-score normalized mean decrease in accuracy values, and predictors were ranked based on their median importance scores across iterations. A custom post-processing function was applied to clean and extract the Boruta output. Importance values were visualized using boxplots ordered by median importance as shown in Figure 4 and Supplementary Figure S2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eLatent profile analysis.\u0026nbsp;\u003c/strong\u003eTo identify subgroups of children with distinct profiles based on performance, we conducted a model-based latent profile analysis (LPA) using the Mclust package in R \u003csup\u003e61\u003c/sup\u003e. This approach fits gaussian finite mixture models under various constraints to the covariance matrices. Preferred models were selected using the Bayesian Information Criterion (BIC). All continuous input variables were standardized (z-scored) prior to modeling. The LPA was first run on kindergarten children (\u003cem\u003en\u003c/em\u003e = 157). We chose the EEI model\u0026mdash;which assumes equal volume and shape for each group with diagonal covariance matrices across aligned to the coordinate axes (Supplementary Figure S3). The EEI parameterization assumes the grouping variables to be uncorrelated and offers a balanced compromise between model complexity and parsimony, particularly suitable for relatively small samples where overparameterization can result in unstable solutions. Among all models evaluated, the EEI model with five latent profiles achieved one of the best BIC values while maintaining the constraint that each cluster contain at least 10% of the sample, ensuring sufficient subgroup stability for interpretation.\u003c/p\u003e\n\u003cp\u003eThe five subgroups reflected a range of performance patterns. These included a high-performing group with scores \u0026ge; +0.25 SD across all tasks, an average-performing group with scores near the sample mean, and a low-performing group with scores \u0026le; \u0026ndash; 0.25 SD. The visual strength group exhibited scores in the MEPs \u0026ge; +0.25 SD and scores on deletion \u0026lt; -0.10 and the visual challenge group exhibited scores in deletion \u0026ge; +0.10 SD and scores on MEPs \u0026lt; -0.25 SD. These empirically derived profiles were subsequently applied to the cohort of first graders to assess the generalizability of the classification framework across samples.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eCompeting interests. The authors declare no competing interests. \u003c/p\u003e\n\u003cp\u003eData Availability. De-identified data is available in the analysis repository. \u003c/p\u003e\n\u003cp\u003eCode Availability. \u003c/p\u003e\n\u003cp\u003eThe experiment code is available at https://github.com/yeatmanlab/roar-mep\u003c/p\u003e\n\u003cp\u003eThe analysis code is available at https://github.com/yeatmanlab/Visual-Tasks-as-Language-agnostic-Early-Identification-Measures-of-Reading-Challenges \u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eSnowling, M. J. Dyslexia: a hundred years on. \u003cem\u003eBMJ\u003c/em\u003e \u003cstrong\u003e313\u003c/strong\u003e, 1096\u0026ndash;1097 (1996).\u003c/li\u003e\n\u003cli\u003ePennington, B. F. \u0026amp; Edition, W. Controversial Therapies for Dyslexia. \u003cem\u003ePerspect. Lang. Lit.\u003c/em\u003e \u003cstrong\u003eWinter\u003c/strong\u003e, 7\u0026ndash;8 (2011).\u003c/li\u003e\n\u003cli\u003eVellutino, F. R., Fletcher, J. M., Snowling, M. J. \u0026amp; Scanlon, D. M. Specific reading disability (dyslexia): What have we learned in the past four decades? \u003cem\u003eJ. Child Psychol. Psychiatry\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 2\u0026ndash;40 (2004).\u003c/li\u003e\n\u003cli\u003eStein, J. F. \u0026amp; Walsh, V. To see but not to read; the magnocellular theory of dyslexia. \u003cem\u003eTrends Neurosci.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 147\u0026ndash;152 (1997).\u003c/li\u003e\n\u003cli\u003eHulme, C. The implausibility of low-level visual deficits as a cause of children\u0026rsquo;s reading difficulties. \u003cem\u003eCogn. Neuropsychol.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 369\u0026ndash;374 (1988).\u003c/li\u003e\n\u003cli\u003eWolf, M. \u003cem\u003eet al.\u003c/em\u003e Towards a dynamic, comprehensive conceptualization of dyslexia. \u003cem\u003eAnn. Dyslexia\u003c/em\u003e (2024) doi:10.1007/s11881-023-00297-1.\u003c/li\u003e\n\u003cli\u003eLovegrove, W. J., Bowling, A., Badcock, D. \u0026amp; Blackwood, M. Specific reading disability: Differences in Contrast Sensitivity as a Function of Spatial Frequency. \u003cem\u003eScience\u003c/em\u003e \u003cstrong\u003e210\u003c/strong\u003e, 439\u0026ndash;440 (1980).\u003c/li\u003e\n\u003cli\u003eDemb, J. B., Boynton, G. M., Best, M. \u0026amp; Heeger, D. J. Psychophysical evidence for a magnocellular pathway deficit in dyslexia. \u003cem\u003eVision Res.\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 1555\u0026ndash;1559 (1998).\u003c/li\u003e\n\u003cli\u003eTalcott, J. B. \u003cem\u003eet al.\u003c/em\u003e Dynamic sensory sensitivity and children\u0026rsquo;s word decoding skills. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e97\u003c/strong\u003e, 2952\u0026ndash;2957 (2000).\u003c/li\u003e\n\u003cli\u003eTalcott, J. B., Hansen, P. C., Assoku, E. L. \u0026amp; Stein, J. F. Visual motion sensitivity in dyslexia: Evidence for temporal and energy integration deficits. \u003cem\u003eNeuropsychologia\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 935\u0026ndash;943 (2000).\u003c/li\u003e\n\u003cli\u003eBoets, B., Vandermosten, M., Cornelissen, P., Wouters, J. \u0026amp; Ghesqui\u0026egrave;re, P. Coherent Motion Sensitivity and Reading Development in the Transition From Prereading to Reading Stage. \u003cem\u003eChild Dev.\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, 854\u0026ndash;869 (2011).\u003c/li\u003e\n\u003cli\u003eEden, G. F. \u003cem\u003eet al.\u003c/em\u003e Abnormal processing of visual motion in dyslexia revealed by functional brain imaging. \u003cem\u003eNature\u003c/em\u003e \u003cstrong\u003e382\u003c/strong\u003e, 66\u0026ndash;69 (1996).\u003c/li\u003e\n\u003cli\u003eLivingstone, M. S., Rosen, G. D., Drislane, F. W. \u0026amp; Galaburda, a. M. Physiological and anatomical evidence for a magnocellular defect in developmental dyslexia. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e88\u003c/strong\u003e, 7943\u0026ndash;7947 (1991).\u003c/li\u003e\n\u003cli\u003eDemb, J. B., Boynton, G. M. \u0026amp; Heeger, D. J. Brain activity in visual cortex predicts individual differences in reading performance. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e94\u003c/strong\u003e, 13363\u0026ndash;13366 (1997).\u003c/li\u003e\n\u003cli\u003eBen-Shachar, M., Dougherty, R. F., Deutsch, G. K. \u0026amp; Wandell, B. A. Contrast responsivity in MT+ correlates with phonological awareness and reading measures in children. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cstrong\u003e37\u003c/strong\u003e, 1396\u0026ndash;1406 (2007).\u003c/li\u003e\n\u003cli\u003eCornelissen, P., Richardson, A., Mason, A., Fowler, S. \u0026amp; Stein, J. Contrast sensitivity and coherent motion detection measured at photopic luminance levels in dyslexics and controls. \u003cem\u003eVision Res.\u003c/em\u003e \u003cstrong\u003e35\u003c/strong\u003e, 1483\u0026ndash;1494 (1995).\u003c/li\u003e\n\u003cli\u003eTalcott, J. B. \u003cem\u003eet al.\u003c/em\u003e On the relationship between dynamic visual and auditory processing and literacy skills; results from a large primary-school study. \u003cem\u003eDyslexia\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 204\u0026ndash;225 (2002).\u003c/li\u003e\n\u003cli\u003ePaulesu, E., Danelli, L. \u0026amp; Berlingeri, M. Reading the dyslexic brain: multiple dysfunctional routes revealed by a new meta-analysis of PET and fMRI activation studies. \u003cem\u003eFront. Hum. Neurosci.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 830 (2014).\u003c/li\u003e\n\u003cli\u003eAmitay, S., Ben-Yehudah, G., Banai, K. \u0026amp; Ahissar, M. Disabled readers suffer from visual and auditory impairments but not from a specific magnocellular deficit. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e125\u003c/strong\u003e, 2272\u0026ndash;2285 (2002).\u003c/li\u003e\n\u003cli\u003eAmitay, S., Ben-Yehudah, G., Banai, K. \u0026amp; Ahissar, M. Reply to: Visual magnocellular deficits in dyslexia. \u003cem\u003eBrain\u003c/em\u003e \u003cstrong\u003e126\u003c/strong\u003e, 3E\u0026ndash;3 (2003).\u003c/li\u003e\n\u003cli\u003eSkottun, B. C. The magnocellular deficit theory of dyslexia: The evidence from contrast sensitivity. \u003cem\u003eVision Res.\u003c/em\u003e \u003cstrong\u003e40\u003c/strong\u003e, 111\u0026ndash;127 (2000).\u003c/li\u003e\n\u003cli\u003eSkottun, B. C. The need to differentiate the magnocellular system from the dorsal stream in connection with dyslexia. \u003cem\u003eBrain Cogn.\u003c/em\u003e \u003cstrong\u003e95\u003c/strong\u003e, (2015).\u003c/li\u003e\n\u003cli\u003eJoo, S. J., Donnelly, P. M. \u0026amp; Yeatman, J. D. The causal relationship between dyslexia and motion perception reconsidered. \u003cem\u003eSci. Rep.\u003c/em\u003e \u003cstrong\u003e7\u003c/strong\u003e, 4185 (2017).\u003c/li\u003e\n\u003cli\u003eFlint, S. \u0026amp; Pammer, K. It is the egg, not the chicken; dorsal visual deficits present in dyslexia are not present in illiterate adults. \u003cem\u003eDyslexia\u003c/em\u003e \u003cstrong\u003e0\u003c/strong\u003e, 1\u0026ndash;15 (2018).\u003c/li\u003e\n\u003cli\u003eGori, S., Seitz, A. R., Ronconi, L., Franceschini, S. \u0026amp; Facoetti, A. Multiple Causal Links between Magnocellular-Dorsal Pathway Deficit and Developmental Dyslexia. \u003cem\u003eCereb. Cortex\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 4356\u0026ndash;4369 (2016).\u003c/li\u003e\n\u003cli\u003eFacoetti, A., Paganoni, P., Turatto, M., Marzola, V. \u0026amp; Mascetti, G. G. Visual-spatial attention in developmental dyslexia. \u003cem\u003eCortex\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 109\u0026ndash;123 (2000).\u003c/li\u003e\n\u003cli\u003eFacoetti, A., Lorusso, M. L., Paganoni, P., Umilt\u0026agrave;, C. \u0026amp; Mascetti, G. G. The role of visuospatial attention in developmental dyslexia: evidence from a rehabilitation study. \u003cem\u003eBrain Res. Cogn. Brain Res.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 154\u0026ndash;164 (2003).\u003c/li\u003e\n\u003cli\u003eFacoetti, A. \u003cem\u003eet al.\u003c/em\u003e The relationship between visuo-spatial attention and nonword reading in developmental dyslexia. \u003cem\u003eCogn. Neuropsychol.\u003c/em\u003e \u003cstrong\u003e23\u003c/strong\u003e, 841\u0026ndash;855 (2006).\u003c/li\u003e\n\u003cli\u003eFacoetti, A., Ruffino, M., Peru, A., Paganoni, P. \u0026amp; Chelazzi, L. Sluggish engagement and disengagement of non-spatial attention in dyslexic children. \u003cem\u003eCortex\u003c/em\u003e \u003cstrong\u003e44\u003c/strong\u003e, 1221\u0026ndash;1233 (2008).\u003c/li\u003e\n\u003cli\u003eHari, R. \u0026amp; Renvall, H. Impaired processing of rapid stimulus sequences in dyslexia. \u003cem\u003eTrends Cogn. Sci.\u003c/em\u003e \u003cstrong\u003e5\u003c/strong\u003e, 525\u0026ndash;532 (2001).\u003c/li\u003e\n\u003cli\u003eO\u0026rsquo;Brien, G. \u0026amp; Yeatman, J. D. Bridging sensory and language theories of dyslexia: Toward a multifactorial model. \u003cem\u003eDev. Sci.\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, e13039 (2021).\u003c/li\u003e\n\u003cli\u003eValdois, S., Reilhac, C., Ginestet, E. \u0026amp; Line Bosse, M. Varieties of Cognitive Profiles in Poor Readers: Evidence for a VAS-Impaired Subtype. \u003cem\u003eJ. Learn. Disabil.\u003c/em\u003e 22219420961332 (2020) doi:10.1177/0022219420961332.\u003c/li\u003e\n\u003cli\u003eFranceschini, S., Gori, S., Ruffino, M., Pedrolli, K. \u0026amp; Facoetti, A. A causal link between visual spatial attention and reading acquisition. \u003cem\u003eCurr. Biol.\u003c/em\u003e \u003cstrong\u003e22\u003c/strong\u003e, 814\u0026ndash;819 (2012).\u003c/li\u003e\n\u003cli\u003eVidyasagar, T. R. \u0026amp; Pammer, K. Dyslexia: a deficit in visuo-spatial attention, not in phonological processing. \u003cem\u003eTrends Cogn. Sci.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 57\u0026ndash;63 (2010).\u003c/li\u003e\n\u003cli\u003eGori, S. \u0026amp; Facoetti, A. How the visual aspects can be crucial in reading acquisition? The intriguing case of crowding and developmental dyslexia. \u003cem\u003eJ. Vis.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, 15.1.8 (2015).\u003c/li\u003e\n\u003cli\u003eTallal, P. Improving neural response to sound improves reading. \u003cem\u003eProc. Natl. Acad. Sci. U. S. A.\u003c/em\u003e \u003cstrong\u003e109\u003c/strong\u003e, 16406\u0026ndash;16407 (2012).\u003c/li\u003e\n\u003cli\u003eGoswami, U. Sensory theories of developmental dyslexia: three challenges for research. \u003cem\u003eNat. Rev. Neurosci.\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 43\u0026ndash;54 (2015).\u003c/li\u003e\n\u003cli\u003ePeterson, R. L. \u0026amp; Pennington, B. F. Developmental dyslexia. \u003cem\u003eAnnu. Rev. Clin. Psychol.\u003c/em\u003e \u003cstrong\u003e11\u003c/strong\u003e, 283\u0026ndash;307 (2015).\u003c/li\u003e\n\u003cli\u003eBradley, R. H. \u0026amp; Corwyn, R. F. Socioeconomic status and child development. \u003cem\u003eAnnu. Rev. Psychol.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 371\u0026ndash;399 (2002).\u003c/li\u003e\n\u003cli\u003ePugh, K. \u0026amp; McCardle, P. \u003cem\u003eHow Children Learn to Read: Current Issues and New Directions in the Integration of Cognition, Neurobiology and Genetics of Reading and Dyslexia Research and Practice\u003c/em\u003e. (Psychology Press, 2011).\u003c/li\u003e\n\u003cli\u003eSolano-Flores, G. \u0026amp; Li, M. Generalizability of test scores across languages: Using multiple data sources to investigate the translation of achievement tests for multilingual populations. \u003cem\u003eInternational Journal of Testing\u003c/em\u003e \u003cstrong\u003e9\u003c/strong\u003e, 98\u0026ndash;123 (2009).\u003c/li\u003e\n\u003cli\u003eSolano-Flores, G. \u0026amp; Trumbull, E. Examining language in context: The need for new research and practice paradigms in the testing of English-language learners. \u003cem\u003eEduc. Res.\u003c/em\u003e \u003cstrong\u003e32\u003c/strong\u003e, 3\u0026ndash;13 (2003).\u003c/li\u003e\n\u003cli\u003eSolano-Flores, G. This paper examines how language, dialect, and register affect the assessment of multilingual students and highlights the challenges of creating assessments that are valid across linguistic and cultural groups. \u003cem\u003eReview of Research in Education\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 117\u0026ndash;150 (2006).\u003c/li\u003e\n\u003cli\u003eRamamurthy, M. \u003cem\u003eet al.\u003c/em\u003e Design and validation of a rapid visual processing measure for screening reading difficulties in early childhood. https://osf.io/x6ysc/download.\u003c/li\u003e\n\u003cli\u003eReardon, S. F. The widening academic achievement gap between the rich and the poor. in \u003cem\u003eInequality in the 21st Century\u003c/em\u003e 177\u0026ndash;189 (Routledge, 2018). doi:10.4324/9780429499821-33.\u003c/li\u003e\n\u003cli\u003eSarsour, K. \u003cem\u003eet al.\u003c/em\u003e Family socioeconomic status and child executive functions: the roles of language, home environment, and single parenthood. \u003cem\u003eJ. Int. Neuropsychol. Soc.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 120\u0026ndash;132 (2011).\u003c/li\u003e\n\u003cli\u003eLawson, G. M., Hook, C. J. \u0026amp; Farah, M. J. A meta-analysis of the relationship between socioeconomic status and executive function performance among children. \u003cem\u003eDev. Sci.\u003c/em\u003e \u003cstrong\u003e21\u003c/strong\u003e, (2018).\u003c/li\u003e\n\u003cli\u003eChristensen, J., Aar\u0026oslash;e, L., Baekgaard, M., Herd, P. \u0026amp; Moynihan, D. P. Human Capital and Administrative Burden: The Role of Cognitive Resources in Citizen‐State Interactions. \u003cem\u003ePublic Administration Review\u003c/em\u003e \u003cstrong\u003e80\u003c/strong\u003e, 127\u0026ndash;136 (2020).\u003c/li\u003e\n\u003cli\u003eZelazo, P. D. \u0026amp; Carlson, S. M. The neurodevelopment of executive function skills: Implications for academic achievement gaps. \u003cem\u003ePsychol. Neurosci.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 273\u0026ndash;298 (2020).\u003c/li\u003e\n\u003cli\u003eReardon, S. F. The widening academic achievement gap between the rich and the poor. in \u003cem\u003eSocial stratification\u003c/em\u003e 536\u0026ndash;550 (Routledge, 2018).\u003c/li\u003e\n\u003cli\u003eValencia Sheila W \u0026amp; Riddle Buly Marsha. Behind Test Scores: What Struggling Readers REALLY Need. \u003cem\u003eThe Australian Journal of Language and Literacy\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 217\u0026ndash;233 (2004).\u003c/li\u003e\n\u003cli\u003eRathbun, A. \u0026amp; West, J. From Kindergarten Through Third Grade: Children\u0026rsquo;s Beginning School Experiences. \u003cem\u003eMathematica Policy Research Reports\u003c/em\u003e.\u003c/li\u003e\n\u003cli\u003eNorton, E. S., Beach, S. D. \u0026amp; Gabrieli, J. D. E. Neurobiology of dyslexia. \u003cem\u003eCurr. Opin. Neurobiol.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 73\u0026ndash;78 (2015).\u003c/li\u003e\n\u003cli\u003eRaschle, N. M., Chang, M. \u0026amp; Gaab, N. Structural brain alterations associated with dyslexia predate reading onset. \u003cem\u003eNeuroimage\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 742\u0026ndash;749 (2011).\u003c/li\u003e\n\u003cli\u003eCroninger, R., Rice, J. K. \u0026amp; Checovich, L. \u003cem\u003eEvaluation of the Use of Free and Reduced-Price Meal Eligibility as a Proxy for Identifying Economically Disadvantaged Students. Alternative Measures and Recommendations\u003c/em\u003e. (Maryland State Department of Education. Retrieved from http://www \u0026hellip;, 2015).\u003c/li\u003e\n\u003cli\u003eDataQuest (CA Dept of Education). https://dq.cde.ca.gov/dataquest/.\u003c/li\u003e\n\u003cli\u003eFacts about English Learners in California. https://www.cde.ca.gov/ds/ad/cefelfacts.asp#:~:text=Basic%20Facts.\u003c/li\u003e\n\u003cli\u003eSiebert, J. M. \u003cem\u003eet al.\u003c/em\u003e Differences in reading screening accuracy by percentile cutoff and English proficiency: Feature selection and group-wise prediction evaluation. \u003cem\u003eEdArXiv\u003c/em\u003e (2024) doi:10.35542/osf.io/kswau.\u003c/li\u003e\n\u003cli\u003eKursa, M. B. \u0026amp; Rudnicki, W. R. Feature Selection with theBorutaPackage. \u003cem\u003eJ. Stat. Softw.\u003c/em\u003e \u003cstrong\u003e36\u003c/strong\u003e, 1\u0026ndash;13 (2010).\u003c/li\u003e\n\u003cli\u003eLunardon, N., Menardi, G. \u0026amp; Torelli, N. ROSE: A package for binary imbalanced learning. \u003cem\u003eR J.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 79 (2014).\u003c/li\u003e\n\u003cli\u003eScrucca, L., Fraley, C., Murphy, T. B. \u0026amp; Raftery, A. E. \u003cem\u003eModel-Based Clustering, Classification, and Density Estimation Using Mclust in R\u003c/em\u003e. (Chapman \u0026amp; Hall/CRC, Philadelphia, PA, 2023).\u003c/li\u003e\n\u003cli\u003eLobier, M., Zoubrinetzky, R. \u0026amp; Valdois, S. The visual attention span deficit in dyslexia is visual and not verbal. \u003cem\u003eCortex\u003c/em\u003e \u003cstrong\u003e48\u003c/strong\u003e, 768\u0026ndash;773 (2012).\u003c/li\u003e\n\u003cli\u003eMascheretti, S. \u003cem\u003eet al.\u003c/em\u003e Visual motion and rapid auditory processing are solid endophenotypes of developmental dyslexia. \u003cem\u003eGenes Brain Behav.\u003c/em\u003e \u003cstrong\u003e17\u003c/strong\u003e, 70\u0026ndash;81 (2018).\u003c/li\u003e\n\u003cli\u003eNorcia, A. M. \u0026amp; Tyler, C. W. The role of the magnocellular pathway in dyslexia. \u003cem\u003eCurrent Directions in Psychological Science\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 200\u0026ndash;204 (2004).\u003c/li\u003e\n\u003cli\u003eVidyasagar, T. R. \u0026amp; Pammer, K. Impaired visual search in dyslexia relates to the role of the magnocellular pathway in attention. \u003cem\u003eNeuroreport\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 1283\u0026ndash;1287 (1999).\u003c/li\u003e\n\u003cli\u003eRonen, I. \u0026amp; Yeshurun, Y. Perceptual and cognitive mechanisms in reading. \u003cem\u003ePsychological Science\u003c/em\u003e \u003cstrong\u003e24\u003c/strong\u003e, 2335\u0026ndash;2344 (2013).\u003c/li\u003e\n\u003cli\u003eRey, A. M. \u0026amp; Valdois, S. The role of visual processing in reading development: Insights from the visual processing speed theory. \u003cem\u003ePsychological Bulletin\u003c/em\u003e \u003cstrong\u003e139\u003c/strong\u003e, 1224\u0026ndash;1252 (2013).\u003c/li\u003e\n\u003cli\u003eMather, N. \u0026amp; Wendling, B. J. \u003cem\u003eEssentials of WJ IV Tests of Achievement\u003c/em\u003e. (John Wiley \u0026amp; Sons, 2015).\u003c/li\u003e\n\u003cli\u003eMiciak, J., Ahmed, Y., Capin, P. \u0026amp; Francis, D. J. The reading profiles of late elementary English learners with and without risk for dyslexia. \u003cem\u003eAnn. Dyslexia\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 276\u0026ndash;300 (2022).\u003c/li\u003e\n\u003cli\u003eHajovsky, D. B., Villeneuve, E. F., Schneider, W. J. \u0026amp; Caemmerer, J. M. An Alternative Approach to Cognitive and Achievement Relations Research: An Introduction to Quantile Regression. \u003cem\u003eJournal of Pediatric Neuropsychology\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 83\u0026ndash;95 (2020).\u003c/li\u003e\n\u003cli\u003eSchrank, F. A., Mcgrew, K. S., Mather, N. \u0026amp; Woodcock-Johnson, I. V. Woodcock-Johnson IV Tests of Achievement. \u003cem\u003eTests of Achievement\u003c/em\u003e (2014).\u003c/li\u003e\n\u003cli\u003eYeatman, J. D. \u003cem\u003eet al.\u003c/em\u003e Rapid Online Assessment of Reading (ROAR). https://roar.stanford.edu/technical/ (2024).\u003c/li\u003e\n\u003cli\u003eCatts, H. W. \u0026amp; Petscher, Y. A Cumulative Risk and Resilience Model of Dyslexia. \u003cem\u003eJ. Learn. Disabil.\u003c/em\u003e \u003cstrong\u003e55\u003c/strong\u003e, 171\u0026ndash;184 (2022).\u003c/li\u003e\n\u003cli\u003eJanuary, S.-A. A. \u0026amp; Klingbeil, D. A. Universal screening in grades K-2: A systematic review and meta-analysis of early reading curriculum-based measures. \u003cem\u003eJ. Sch. Psychol.\u003c/em\u003e \u003cstrong\u003e82\u003c/strong\u003e, 103\u0026ndash;122 (2020).\u003c/li\u003e\n\u003cli\u003eVan Buuren, S. \u0026amp; Groothuis-Oudshoorn, K. \u0026ldquo;mice: Multivariate Imputation by Chained Equations in R. \u003cem\u003eJournal of Statistical Software\u003c/em\u003e \u003cstrong\u003e45\u003c/strong\u003e, 1\u0026ndash;67 (2011).\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table 3","content":"\u003cp\u003eTable 3 is available in the Supplementary Files section.\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-6804845/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6804845/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Despite decades of research, the role of visual processing in learning to read remains contested—largely due to small, homogenous samples and the lack of reliable tools to capture the true heterogeneity of reading development. In this study, we administered theory-driven, carefully validated measures of rapid visual processing to a large, socioeconomically and linguistically diverse cohort of kindergarten and first-grade children in California public schools (N ~ 1200). These visual measures proved to be equitable —performance did not vary by home language or socioeconomic status— and independently accounted for 12–18% of the variance in reading outcomes. They were also important predictors of reading risk at year-end and a year- later. Latent profile analysis revealed subgroups invisible to traditional screeners: children with strong language but poor visual skills who later struggled to read, and children with visual strengths who outperformed expectations despite phonological weaknesses. Integrating measures of rapid visual processing into early screening offers a promising path forward, towards more equitable, personalized interventions and a deeper understanding of early reading development.","manuscriptTitle":"The Link Between Vision and Reading: A Language-agnostic Window into Heterogeneity in Early Reading Development","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-06-23 07:06:26","doi":"10.21203/rs.3.rs-6804845/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"47da43fa-44e3-4eef-ab09-07e4e55fb626","owner":[],"postedDate":"June 23rd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":50228581,"name":"Biological sciences/Psychology/Human behaviour"},{"id":50228582,"name":"Biological sciences/Neuroscience/Visual system/Pattern vision"}],"tags":[],"updatedAt":"2025-08-20T09:50:56+00:00","versionOfRecord":[],"versionCreatedAt":"2025-06-23 07:06:26","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6804845","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6804845","identity":"rs-6804845","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}