Prevalence of Dyslexia and Its Relationship with Decoding and Comprehension in Persian-speaking Elementary Students

Abstract

Background:

Dyslexia, a neurodevelopmental disorder characterized by difficulties in word recognition and decoding, remains underexplored in non-Western orthographies such as Persian. To estimate the prevalence of dyslexia and examine the relationship between decoding and reading comprehension among Persian-speaking elementary students.

Methods:

A cross-sectional study was conducted among 1,400 students (Grades 2–5) using a multistage stratified-cluster sampling design. Dyslexia was identified using the nonword reading subtest of a standardized reading and dyslexia test, with intellectual functioning assessed via the Wechsler Intelligence Scale for Children-Fourth Edition (WISC-IV). Survey-weighted logistic and linear regression models were used to analyze associations and interactions, with a Bonferroni correction (p < .0083) applied for multiple comparisons. Confidence intervals (CI) and odds ratios (OR) are reported.

Results:

The weighted prevalence of dyslexia was 10.40% (95% CI: 9.00–12.05), with higher rates in boys (14.50%, 95% CI: 12.42–16.92, OR: 2.66, 95% CI: 1.90–3.71, p < .001) than girls (6.01%, 95% CI: 4.61–7.80) and in private versus public schools (6.40% vs. 12.08%, OR: 0.50, 95% CI: 0.33–0.75, p = .002). Word reading and comprehension were significant protective factors (p < .001), while phonological awareness was non-significant in adjusted models. Decoding strongly predicted comprehension in typical readers, but this association was significantly weaker in dyslexic students (p = .006).

Conclusions:

Dyslexia affects approximately one in 10 Persian-speaking students, with variations influenced by gender, school type, and reading skills. Comprehensive screening and interventions targeting decoding and comprehension are critical, particularly in public schools.

Key Messages:

• We examined Persian-speaking students (Grades 2–5) to assess the prevalence of dyslexia and its relationship to decoding and comprehension using standardized tests and regression analyses.

• Dyslexia affects 10.40% of students, with higher rates in boys (14.50%) and public schools; decoding predicts comprehension less effectively in students with dyslexia.

• Targeted screening and interventions focusing on decoding and comprehension are necessary, especially in public schools, to reduce literacy disparities.

Dyslexia is a specific neurodevelopmental learning disorder characterized by persistent difficulties in word recognition, decoding, and spelling, despite adequate intelligence and educational exposure.^1,2 It is classified under specific learning disorder with impairment in reading (developmental dyslexia), as per the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5), code 315.00, hereafter referred to as “dyslexia,” which highlights its neurobiological origins and the influence of genetic and environmental factors. ³ This lifelong condition has a significant impact on academic performance and daily functioning, underscoring the importance of early identification and targeted intervention.^2,4

The prevalence and manifestation of dyslexia vary widely across languages, mainly due to differences in orthographic transparency.^5–7 Transparent orthographies such as Hindi, Italian, and Finnish facilitate more straightforward decoding due to consistent letter-sound correspondences.^6,8 In contrast, opaque and semi-transparent orthographies, such as English and Persian, pose greater challenges for phonological and orthographic processing, which can intensify the reading difficulties experienced by individuals with dyslexia.^5,6,8 Indian languages such as Hindi and Kannada, which are written in Devanagari and Kannada scripts, are considered relatively transparent to varying degrees, with consistent grapheme–phoneme mappings that make decoding comparatively easier than in Persian.^9,10 Persian, a semi-transparent and morphologically rich orthography, may therefore present unique cognitive demands^11,12; however, empirical evidence on dyslexia’s prevalence and its cognitive correlates in Persian-speaking populations remains limited.

Beyond orthographic transparency, cross-linguistic research has also investigated the epidemiology and cognitive profiles of dyslexia across different languages, including Spanish, Arabic, and, more recently, Indian languages. For instance, a screening study in Spain estimated that 1.6%–6.4% of university students might be at risk of dyslexia. ¹³ In Arab-speaking populations, a recent meta-analysis of 18 studies including more than 30,000 primary-school children reported an overall prevalence of about 11%, with higher rates in Asian Arab countries in the Gulf, particularly in the Gulf (around 24%)—compared to 13% in non-Gulf countries. ¹⁴ In Indian populations, a recent systematic review and meta-analysis estimated the prevalence of developmental dyslexia at 6.2% (95% CI: 4.4%–7.9%) among children.15 In the context of Indian languages, evidence suggests that the prevalence of developmental dyslexia is influenced by the high transparency of scripts such as Hindi, Tamil, and Kannada, where consistent grapheme–phoneme correspondences reduce decoding load. Still, challenges remain in morphological and syllabic processing.^16,17 This provides a valuable comparison for Persian, where optional diacritics and complex morphological structures impose additional decoding demands.

These findings underscore the universal nature of dyslexia, while also illustrating how orthographic transparency, educational context, and regional factors shape its prevalence and manifestation. The reported prevalence of dyslexia in Persian-speaking populations ranges from 1.2% to 10%, with variations attributed to inconsistent diagnostic criteria, small sample sizes, and a lack of standardized tools.^18–21 Many studies fail to account for complex sampling designs, leading to unreliable estimates. ¹⁹ Additionally, the relationship between decoding skills and reading comprehension in Persian orthography is underexplored, despite international evidence supporting a strong link. ²² According to the Simple View of Reading model, reading comprehension results from the interaction of decoding and linguistic comprehension. ²² Given the Persian language’s orthographic characteristics, decoding difficulties may significantly impact comprehension outcomes in dyslexic children.^11,12

Given these observations, investigating the relationship between decoding skills and reading comprehension, alongside demographic factors such as gender and school type, is essential for understanding dyslexia in Persian-speaking students. We hypothesized that decoding skills have a more substantial impact on reading comprehension in students, and that the prevalence of dyslexia among Persian-speaking students varies by gender and school type.

To test this hypothesis, this study addresses existing gaps in the literature by estimating the prevalence of dyslexia among primary-school students using standardized diagnostic criteria and examining the relationship between decoding and reading comprehension, while accounting for the moderating effect of dyslexia. By employing a large, stratified-cluster sample and survey-weighted analyses, this research provides robust evidence to inform educational policy and intervention strategies tailored to the needs of Persian-speaking students with dyslexia.

Methods

This study was approved by the Institutional Review Board at Zahedan University of Medical Sciences (Approval Code: IR.ZAUMS.REC.1403.284). Written informed consent was obtained from parents or legal guardians, and assent was sought from students under 18 years of age. Students identified with dyslexia were referred to school counselors and local speech-language pathologists for further evaluation and support. This study adhered to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines, with the checklist provided as supplementary online material. ²³

This cross-sectional, population-based epidemiological study investigated the prevalence and predictors of dyslexia, as defined by the DSM-5 criteria for specific learning disorder (315.00). ³ Among Persian-speaking elementary school students from January to May 2024. The study adhered to rigorous methodological standards, employing validated assessment tools and robust sampling techniques to ensure data quality, minimize bias, and enhance external validity.

The target population consisted of students in grades two through five attending public and private primary schools, with an estimated total of approximately 100,000 students. To ensure representative coverage, a multistage, stratified-cluster random sampling strategy was implemented. Initially, schools were stratified by gender (boys, girls) and type (public or private) to account for structural differences in educational settings. Within each stratum, schools were randomly selected using a probability-proportional-to-size approach. This was followed by the random selection of classrooms within each school, and finally, the random selection of students within those classrooms. This method minimized selection bias and ensured proportional representation across demographic and institutional subgroups, yielding a final sample of 1,400 students.

The sample size was calculated using WINPEPI software (version 3.18) to ensure adequate statistical power for estimating dyslexia prevalence and analyzing predictors. ²⁴ Based on an estimated dyslexia prevalence of 10%, derived from prior studies,^18–21 with a 95% confidence level, a precision of 2%, and a statistical power of 95%, the minimum sample size was determined. This adjustment accounted for a design effect of 1.5 to reflect the clustered nature of the sampling design, and an additional 8% was included to accommodate anticipated attrition due to non-response or ineligibility.

Eligible participants were native Persian speakers enrolled in grades two through five who had completed at least six months of the academic year and had normal hearing and vision, as verified by school health records. Students were excluded if they had diagnosed neurological conditions (e.g., epilepsy), uncorrected sensory impairments, or intellectual disability (defined as an IQ < 85 on the Wechsler Intelligence Scale for Children, Fourth Edition [WISC-IV]), ²⁵ or prolonged absenteeism that disrupted consistent schooling. ²⁶ This eligibility framework ensured a representative sample of typically developing Persian-speaking students while excluding factors that could obscure findings specific to dyslexia.

Materials/Measures

Two validated assessment tools were employed to evaluate reading skills and intellectual functioning. A standardized Persian-language reading and dyslexia test, developed and validated for Persian-speaking populations, assessed word reading, nonword reading (decoding), phonemic awareness, and reading comprehension. This test required no translation due to its native design and demonstrated robust psychometric properties, with Cronbach’s alpha coefficients ranging from 0.71 to 0.98 and strong construct validity. ²⁷ The WISC-IV was used to assess intellectual functioning and exclude students with intellectual disabilities, ensuring that reading difficulties were not attributable to general cognitive deficits. Both tools were selected for their cultural and linguistic appropriateness, providing reliable measurement among Persian-speaking students.

Data collection was conducted following ethical approval from the Institutional Ethics Committee. Written informed consent was obtained from the parents or legal guardians of all participants, and assent was also obtained from the students themselves. Assessments were administered individually by trained speech-language pathologists in quiet, distraction-free rooms provided by the schools, with each session lasting approximately 90 minutes. To optimize reliability, assessors underwent standardized training, and inter-rater reliability checks were performed on a subset of 10% of assessments, yielding agreement rates above 95%. The initial screening involved administering the full reading battery to all participants, with the nonword reading subtest used to identify students in the bottom 25th percentile, who were considered at risk for dyslexia. These students underwent further evaluation with the WISC-IV to confirm eligibility and rule out intellectual disability. Participation was voluntary, with students and families informed of their right to withdraw at any point without penalty. Confidentiality was maintained through anonymization of all data, with unique identifiers used to protect participant privacy throughout the study.

Statistical Analysis

All analyses were performed in Stata 14.2, using survey-weighted procedures to account for the multistage, stratified-cluster sampling design. The svyset command incorporated stratification by gender and school type, clustering within schools and classrooms, and sampling weights. The prevalence of dyslexia and its 95% confidence interval were estimated using the Taylor series linearization method.

Survey-weighted multivariable logistic regression models examined associations between dyslexia (binary: 1 = presence, 0 = absence) and covariates (gender, school grade, school type). Given the relatively high prevalence of dyslexia, odds ratios were interpreted cautiously. Bonferroni correction (p < .0083) was applied to control for type I error in subgroup comparisons.

To assess cognitive-linguistic covariates, single-covariate survey-weighted logistic regression models were used to examine phoneme elimination, word reading, word chains, comprehension, picture naming, and related skills, with statistical significance defined at p < .05.

A multivariable model then included covariates that were significant in the single-covariate analyses (phoneme elimination, word reading, word chains, comprehension). Multicollinearity was ruled out using Variance Inflation Factors (VIFs < 2, with a mean VIF of 1.31) in a preliminary linear regression with word reading as the outcome, selected due to its central role in dyslexia assessment. Model fit was confirmed using the Hosmer–Lemeshow test (χ² (8) = 9.39, p = .31).

To evaluate the effect of decoding on reading comprehension by dyslexia status, survey-weighted linear regression models with interaction terms were fitted, with significance assessed via adjusted Wald tests (p < .05) and model fit evaluated using adjusted R² (R² = 0.26, p < .001).

Results

Data from 1,400 students in grades two through five were analyzed (Figure 1). The sample consisted of 60.2% girls, and the majority of students (85.1%) were enrolled in public schools. The grade-level distribution was relatively balanced, with slightly fewer students in grade five compared to the other grades. Missing data were minimal (<2%) and handled using listwise deletion.

Figure 1. Schematic Flowchart of Enrollment and Exclusions.

Table 1 provides an overview of participant demographics along with the survey-weighted prevalence of dyslexia across subgroups. The overall survey-weighted prevalence of dyslexia in the total sample was estimated at 10.4% (95% CI: 9%–12.05%). Notably, the prevalence was significantly higher among boys than girls, with boys exhibiting a rate more than twice that of their female peers. Additionally, students attending public schools showed a higher prevalence compared to those in private schools. Dyslexia prevalence declined across grade levels, with the highest observed in second grade and a significantly lower prevalence in fourth grade. Although a reduction was also noted in fifth grade, this difference was not statistically significant after Bonferroni correction.

Table 1.

Participant Demographics and Dyslexia Prevalence Across Subgroups (Survey-weighted Analysis; Bonferroni Correction Applied).

Variable	Group	Total N (%)	No. Dyslexia (n, %)	Dyslexia (n, %)	Prevalence (%)	95% CIa	SEa	p Value
Gender	Girls	843 (60.21%)	787 (93.36%)	56 (6.64%)	6.01	4.61–7.80	0.74	<.001***
	Boys	557 (39.79%)	467 (83.84%)	90 (16.16%)	14.52	12.42–16.92	1.05	Ref
School type	Public	1,191 (85.07%)	1,056 (88.66%)	135(11.34%)	12.08	10.32–14.10	0.89	.002***
	Private	209 (14.93%)	198 (94.74%)	11 (5.26%)	6.40	4.54–8.95	1.02	Ref
Grade	Grade 2	407 (29.07%)	342 (84.03%)	65 (15.97%)	14.38	10.79–18.90	1.90	Ref
	Grade 3	409 (29.21%)	362 (88.51%)	47 (11.49%)	13.55	10.06–18.00	1.85	.794
	Grade 4	356 (25.43%)	336 (94.38%)	20 (5.62%)	4.97	3.38–7.24	0.89	.002***
	Grade 5	228 (16.29%)	214 (93.86%)	14 (6.14%)	5.14	2.72–9.51	1.52	.012

^aConfidence interval.

^bStandard error.

***Bonferroni Significance (α = 0.0083).

To further investigate the independent associations between dyslexia and demographic variables, we conducted a survey-weighted multivariable logistic regression analysis. As shown in Table 2, gender emerged as the strongest predictor of dyslexia. Boys had 2.66 times the odds of having dyslexia compared to girls, even after adjusting for other variables in the model. School type was also a significant factor, with students enrolled in private schools exhibiting approximately half the odds of dyslexia relative to public school students. Grade level was inversely associated with dyslexia risk: fourth-grade students had the lowest odds compared to second graders, and the association for fifth grade was not statistically significant after Bonferroni correction.

Table 2.

Survey-weighted Logistic Regression for Predictors of Dyslexia.

Variable	Odds Ratio	95% CIa	p Value
Gender (male vs. female)	2.66	1.90–3.71	<.001***
Grade 3 vs. Grade 2	0.93	0.54–1.62	.794
Grade 4 vs. Grade 2	0.31	0.16–0.60	.002***
Grade 5 vs. Grade 2	0.32	0.14–0.75	.012
Private vs. public school	0.50	0.33–0.75	.002***

Notes: ^aConfidence interval.

***Bonferroni Significance (α = 0.0083).

We then estimated a multivariable logistic regression model (Table 3) including phonemic awareness, word chaining, word comprehension, and word reading to assess their unique contributions to dyslexia. After adjusting for the overlap among predictors, word reading, word chaining, and word comprehension remained significant protective factors against dyslexia (all p < .001). Phoneme elimination was not a significant predictor in the multivariable model (p = .76), suggesting that its predictive value may be accounted for by shared variance with other skills. The model explained approximately 29.8% of the variance in dyslexia status (pseudo-R² = 0.30). Furthermore, the Hosmer–Lemeshow test indicated a good model fit (χ² (8) = 9.39, p = .31).

Table 3.

Multivariable Logistic Regression for Predictors of Dyslexia (Word Reading Model).

Predictor Variable	Odds Ratio (OR)	95% CIa	p Value
Word reading	0.94	0.92–0.96	<.001
Phoneme elimination	1.01	0.97–1.04	.759
Word chain	0.96	0.95–0.97	<.001
Word comprehension	0.93	0.90–0.97	<.001
Constant	20.73	9.29–46.18	<.001

All variables were entered simultaneously into the model. The outcome variable is dyslexia status (1 = Yes, 0 = No). Odds ratios (OR) less than 1 indicate a negative association with dyslexia (protective effect).

^aConfidence interval.

In the final stage of analysis, we examined the relationship between decoding skills (nonword reading) and reading comprehension, and whether the presence of dyslexia moderated this relationship.

As shown in Table 4, decoding ability was a strong and significant predictor of reading comprehension among non-dyslexic students. However, the interaction term between decoding and dyslexia was statistically significant and negative, indicating that the positive association between decoding and comprehension was substantially attenuated in students with dyslexia. The final linear model explained approximately 26.4% of the variance in reading comprehension scores (R² = 0.26, p < .001).

Table 4.

Survey-weighted Linear Regression: Decoding × Dyslexia Interaction Model (Outcome = Word Comprehension).

Predictor	Coefficient (β)	SEa	95% CIb	p Value
Decoding (typical readers)	0.46	0.029	0.40–0.52	<.001
Dyslexia status	8.47	2.25	3.67–13.26	.002
Decoding × Dyslexia	–0.33	0.10	–0.55 to –0.11	.006
Constant	6.75	1.17	4.26–9.24	<.001
Model R²	0.26

^aStandard error.

^bConfidence interval.

Discussion

This study provides robust, population-based evidence on the prevalence and cognitive-linguistic correlates of dyslexia among Persian-speaking elementary students. By employing a multistage, stratified-cluster sampling design and survey-weighted analyses, the findings offer high external validity and generalizability to similar urban contexts. The estimated dyslexia prevalence of 10.4% aligns with global estimates^1,2 but highlights a critical need for targeted educational interventions, particularly in resource-limited settings.

The pronounced gender disparity, with boys exhibiting a prevalence rate more than twice that of girls (14.5% vs. 6.6%), is consistent with prior research suggesting neurobiological vulnerabilities or potential referral biases in male students.^28,29 This finding highlights the importance of gender-sensitive screening protocols in ensuring equitable identification and support for all students. The significant reduction in dyslexia prevalence in fourth grade may reflect developmental improvements in reading fluency and automaticity, as suggested by longitudinal studies.^30,31 However, the non-significant reduction in fifth grade after Bonferroni correction warrants cautious interpretation, as it may reflect statistical noise or contextual factors such as curriculum demands. Future research should explore whether this trend reflects true developmental changes or variations in instructional quality across grade levels.

The higher prevalence of dyslexia in public versus private schools (12.08% vs. 6.40%) may indicate potential structural inequalities in educational resources, teacher training, and student-teacher ratios, which are well-documented determinants of academic outcomes.^32,33 This disparity suggests that socioeconomic and institutional factors may exacerbate reading difficulties in public school settings, necessitating policy interventions to address resource inequities. For instance, targeted funding for teacher professional development and early literacy programs could mitigate these gaps.

Cognitively, word reading, word chaining (fluency), and reading comprehension emerged as significant protective factors against dyslexia, even after adjusting for shared variance. These findings align with the Simple View of Reading model, which posits that decoding and linguistic comprehension are critical drivers of reading proficiency. ²² Notably, the non-significant role of phonological awareness in the multivariable model suggests that, in Persian orthography, fluency and comprehension may play a more decisive role in later elementary grades. This observation is consistent with the National Reading Panel’s findings, which highlight the diminishing role of phonological awareness in advanced literacy stages. ³⁴ Persian’s semi-transparent orthography, with its complex vowel representation and morphological richness, may further amplify the importance of fluency and comprehension over phonological processing in mitigating the risk of dyslexia.^11,12 These results advocate for multidimensional screening tools that prioritize fluency and comprehension alongside phonological skills, particularly in non-Western orthographies.

The interaction analysis revealed that decoding strongly predicted reading comprehension in typical readers, but this relationship was significantly attenuated in dyslexic students (p = .006). This finding supports dual-route and connectionist models of reading, which suggest that comprehension deficits in dyslexia may stem from impaired integration of decoding, working memory, and inferencing skills.^4,35,36 Practically, this underscores the need for interventions that target both decoding and higher-order language processes, such as vocabulary development and inferential comprehension, to address the multifaceted nature of dyslexia in Persian-speaking students. For example, structured literacy programs that combine explicit decoding instruction with comprehension strategies could be particularly effective.

From a cross-linguistic perspective, our findings not only align with prior evidence from Spanish and Arabic populations but also resonate with studies in Indian languages. Epidemiological research in India has reported dyslexia prevalence ranging from 4.4% to 7.9% in children. ¹⁵ These rates, although somewhat lower than those reported in Persian-speaking populations, illustrate how orthographic transparency and educational context influence the identification and manifestation of dyslexia.

Furthermore, while Persian’s semi-transparent orthography poses decoding challenges due to optional diacritics and complex morphological structures, Indian scripts such as Devanagari (used in Hindi) and Kannada are relatively more transparent, supporting consistent grapheme–phoneme mapping. This structural difference likely contributes to the stronger predictive role of phonological awareness reported in Indian languages compared to Persian, where fluency and comprehension skills appear to play a more central role.^9,17

This study’s findings have significant implications for educational policy and similar linguistic contexts. The elevated prevalence of dyslexia in public schools calls for universal screening programs and teacher training initiatives tailored to the linguistic nuances of Persian orthography. Moreover, the weaker association between decoding and comprehension in dyslexic students underscores the need for comprehensive interventions that extend beyond phonological training to include fluency- and comprehension-focused strategies. Such approaches could be integrated into national literacy curricula to reduce literacy inequities.

Strengths and Limitations

This study has several strengths that enhance the robustness of its findings. First, the use of a large, stratified-cluster sampling design with survey-weighted analyses ensured precise estimates of dyslexia prevalence, accounting for the complex sampling structures. Moreover, employing standardized tools, such as the Persian-language reading and dyslexia test and the WISC, with robust psychometric properties, provided reliable measures of reading skills and cognitive functioning. Additionally, conducting the study in a diverse urban setting strengthens its relevance for educational policy in comparable regions. However, the cross-sectional design precluded causal inferences and longitudinal tracking of reading skill development. Thus, longitudinal studies are necessary to investigate how reading abilities develop and to assess the effects of early interventions. Furthermore, defining dyslexia solely based on nonword reading may overlook students with atypical or milder profiles, potentially underestimating the prevalence. Consequently, future research should adopt broader diagnostic criteria, including rapid automatized naming and orthographic processing, to capture the full spectrum of dyslexia manifestations. Additionally, the omission of socio-emotional factors, executive functioning, and home literacy environment limited a comprehensive understanding of dyslexia risk and resilience. ³⁷ Incorporating these variables could provide a more holistic understanding of dyslexia risk and resilience in Persian-speaking populations. Finally, while the study was conducted in a diverse urban setting, replication in other urban or rural contexts would enhance cross-cultural generalizability.

Conclusions

This study offers robust evidence on the prevalence and predictors of dyslexia among Persian-speaking students, highlighting the roles of gender, school type, and cognitive-linguistic skills. The findings advocate for multidimensional screening and intervention frameworks that address the unique challenges of Persian orthography and educational disparities in public schools. By prioritizing early identification and tailored support, educational systems can reduce literacy inequities and improve outcomes for students with dyslexia.