Inadequate foundational decoding skills constrain global literacy goals for pupils in low-and middle-income countries

Learning to read is paramount for primary education success, impacting future educational attainment and overall societal development. However, a significant percentage of children in LMICs (approximately 57%) cannot read with basic comprehension by age 10, despite substantial investment in global education. This research departs from studies focusing on systemic factors (teacher training, infrastructure) and instead examines the issue through the lens of the science of reading. This science emphasizes the importance of foundational decoding skills—understanding that letters represent sounds and fluently retrieving letter-to-sound mappings—as the basis for reading comprehension. High-income countries often prioritize these skills using systematic phonics and regular assessments, a contrast to the Global Proficiency Framework used in LMICs, which focuses more on comprehension and lacks a strong phonics emphasis. The absence of regular decoding assessments in LMICs hinders understanding of poor reading outcomes. This study uses the Early Grade Reading Assessment (EGRA), a tool modeled on DIBELS but adapted for LMICs, to address this gap.

The existing literature highlights the significant problem of low literacy rates in LMICs. While previous research has concentrated on systemic factors such as teacher quality, school resources, and infrastructure, this study focuses on the foundational skills necessary for reading acquisition, namely decoding. Research in high-income countries demonstrates the crucial role of systematic phonics instruction in developing decoding skills, an area largely neglected in LMICs. The Global Proficiency Framework, the main framework for assessing reading progress in LMICs, focuses heavily on comprehension, with limited attention to decoding. This study aims to bridge this gap by utilizing data from the EGRA assessment to evaluate decoding skills in LMICs and their relationship to reading comprehension.

This research utilizes data from hundreds of EGRA surveys conducted over 15 years, assessing over half a million pupils in 48 LMICs, across 96 languages. The EGRA, modeled on the DIBELS assessment, includes tasks evaluating letter name identification, letter sound identification, non-word reading, and oral reading fluency. These tasks are comparable to DIBELS measures, allowing for benchmark comparisons. The researchers analyzed EGRA data across the first three instructional years, comparing performance against DIBELS benchmarks for substantial and severe risk. Statistical analyses, using a Bayesian alternative to a t-test, investigated pupil progress across years, deviation from benchmarks, and the relationship between decoding and reading comprehension. Data included scores from EGRA surveys in local languages and former colonial languages (English, French, Spanish, Portuguese). The analysis addressed potential concerns by examining data subsets: nationally representative samples, English-only samples, and samples with all four decoding measures. The researchers developed a database comprising 230 EGRA surveys, encompassing 694 subsurveys from 22,656 schools. The analysis focuses on alphabetic writing systems. Benchmarks from the DIBELS 8th edition were utilized for comparison. Bayesian methods were employed to account for uncertainty in the estimates and allow for a more robust analysis, particularly given the potential for outliers in the data. Skewness in the data was addressed through further analysis utilizing a skewed t-distribution. Pearson correlations examined the relationship between decoding and reading comprehension.

The study revealed strikingly poor performance on all decoding measures, with performance falling further below benchmarks each year. By year 3, a substantial percentage of pupils scored below the substantial-risk benchmark (64-99%) and the severe-risk benchmark (55-96%). While some improvement occurred across years, this was insufficient to meet benchmarks, indicating a widening performance gap. Statistical analyses confirmed that subskill scores improved across years, although uncertainty and variability across education systems remained high. Importantly, comparison against substantial-risk benchmarks showed that the performance gap widened with each year, signifying pupils falling further from the trajectory needed for proficient reading. A strong positive correlation (r > 0.60) existed between all four decoding tasks and reading comprehension, highlighting the crucial role of decoding in comprehension. The absolute gains in oral reading fluency, though exhibiting medium-to-large effect sizes, were small in absolute terms compared to benchmarks, emphasizing the need to evaluate progress using absolute measures rather than just effect sizes. Large effect sizes from low baselines can provide a misleading impression of significant progress.

The findings demonstrate a significant failure in public policy regarding reading instruction in LMICs. The lack of focus on foundational decoding skills, as evidenced by the Global Proficiency Framework's emphasis on comprehension, hinders progress towards literacy goals. The crucial role of decoding in reading comprehension is underscored by the strong correlation between decoding skills and comprehension scores. The insufficient progress in decoding skills directly impedes children's ability to access higher-level text comprehension. The observed learning gains, while statistically significant, were insufficient to place students on a trajectory toward proficient reading. This emphasizes the need for a shift in policy, focusing on improving foundational decoding skills rather than solely emphasizing comprehension. The use of US benchmarks is acknowledged as a pragmatic choice, though language-specific benchmarks are desirable for future research.

This study highlights the critical need for improved reading instruction in LMICs, focusing on foundational decoding skills. The lack of progress in decoding severely undermines the ability to achieve literacy goals. Rigorous systematic phonics programs accompanied by regular assessments of decoding abilities are essential. Future research should develop language-specific benchmarks and address the challenges faced in multilingual and low-resource contexts. Policy responses must address the failure to translate the science of reading into effective educational practices in LMICs.

The study's conclusions are limited by the nature of the available data. The EGRA data, while extensive, has limitations concerning consistency in administration and sampling methodologies. The data represent averaged scores for populations, hindering inferences about individual pupil proficiency. The use of US benchmarks is a limitation, and language-specific benchmarks would provide a more nuanced analysis. While efforts were made to mitigate these limitations through data scrutiny and robust statistical methods, these limitations should be considered when interpreting the results.