This toolkit is the product of ongoing collaboration among a large community of scholars, practitioners, government officials, and education development professionals to advance the cause of early reading assessment and acquisition among primary school children in low-income countries. Extensive peer review comments for this toolkit and suggestions for instrument development were provided by Marcia Davidson, Sandra Hollingsworth, Sylvia Linan-Thompson, and Liliane Sprenger-Charolles. Development of EGRA would not have been possible without the support of the nongovernmental organization and Ministry of Education EGRA Evaluation Teams of Afghanistan, Bangladesh, Egypt, The Gambia, Guyana, Haiti, Honduras, Jamaica, Kenya, Liberia, Mali, Nicaragua, Niger, Peru, Senegal, and South Africa. Our deepest gratitude goes to the teachers, the students, and their families for their participation and continued faith in the benefits of education. In repayment we will diligently seek to improve reading outcomes for all children around the world. Amber Gove is responsible for primary authorship of the toolkit, with contributions from Luis Crouch, Amy Mulcahy-Dunn, and Marguerite Clarke; editing support was provided by Erin Newton. The opinions expressed in this document are those of the authors and do not necessarily reflect the views of the United States Agency for International Development or the World Bank. Please direct questions or comments to Amber Gove at agove@rti.org.

Over the last decade, USAID has been supporting the Government of Tajikistan (GOT) to improve learning outcomes in primary education. The GOT has not previously used a national objective assessment to measure those learning outcomes. USAID introduced the Early Grade Reading Assessment to measure the impact of its investments. An EGRA has been implemented about every two years since 2013.

Aligning donor investment with country priorities and effective approaches of engagement are essential for long-term impact. Over the last decade, USAID has been supporting the Government of Tajikistan (GOT) to improve literacy and numeracy skills of all primary education students. Numerous other development agencies also fund projects in the education sector, with an average total annual contribution to education of roughly (based on OECD data). The impact of these investments is less than it could be, in part because there needs to be greater alignment between the government’s priorities and development partner activities.

The research on first language learning is the premise for the Philippines Mother-Tongue-based Multi-Lingual Education (MTB-MLE) Policy which requires schools to deliver the Kindergarten to Grade 3 curriculum in the mother tongue (home language) of the school’s community (Corder, 1983; Walter & Dekker, 2011. Salmona, 2014; Yadav, 2014). Considering that the national curriculum requires children to transition to and learn in Filipino and English at the start of Grade 4, the question of mother tongue’s effect on second and third language acquisition is not academic, but central to the policy debate on MTB-MLE efficacy. The MTB-MLE policy consists of five discrete areas: curriculum, learning resources, assessment, teacher recruitment and training, and community support. MTB-MLE has proven challenging with respect to the myriad languages and dialects. Out of the 180 plus languages spoken, only 19 have been formally supported with an official orthography, standard learning resources and teacher professional development materials. In communities with non-supported languages, teachers contextualize the teaching and learning materials, often translating and adapting from the linguistically nearest mother tongue to their own. In a 2019 study on MTB-MLE, the Philippines Institute of Development Studies found inconsistent implementation across schools, including teachers’ negative attitudes toward MTB-MLE, linguistic diversity of learners and classrooms, and lack of teaching and learning materials being key factors hindering its implementation (PIDS, 2019). The Bicol Region poses a particular challenge. In an area slightly larger than Connecticut with a population of just under 4 million, thirteen different languages plus numerous dialects are spoken in the various provinces, cities and towns that dot this volcanic region (Lobel, 2019). The standard language of Central Bikol, which is the mother tongue of approximately half the population consists of six different local dialects depending on the locale. Wedged in the middle of the Central Bikol-speaking area is a cluster of five distinct languages: Rinconada, Buhi-non, Bikol Libon, West Albay Bikol, and Miraya, with only one or two municipal communities each that speak these languages. Under the USAID Advancing Basic Education in the Philippines (ABC+), RTI conducted a longitudinal study that provides new evidence on the efficacy of MTB-MLE. The study tracked the language and literacy acquisition of four groups of learners: those Central Bikol learners who are learning in a fully supported language; Buhi-non speaking learners who are learning in an unsupported language (ie, Buhi-non); Central Bikol speakers who’s language of instruction is Tagalog and Tagalog learners who’s language of instruction is Tagalog. The findings show evidence that the Central Bikol learners whose LOI is Tagalog are performing at par or worse in nearly all domains of reading in their first (Central Bikol), second (Tagalog/Filipino) and third languages (English). The findings show a flattening of their trajectory in terms of the pace of language and literacy acquisition, as well as significant equity gaps in comparison to their Tagalog peers. The evidence points toward continued support to MTB-MLE, despite the challenges in implementation.

The challenges of conducting educational assessments in low- and middle-income environments during the pandemic can be eased by AI-powered speech recognition technology that offers a promising solution to enhance assessments. By utilizing advanced algorithms and machine learning techniques, this technology accurately transcribes spoken language into written text. Reading fluency and comprehension can be efficiently measured by integrating AI speech recognition into assessments, without the need for physical presence. From the safety of their homes, students can perform the assessments using their smartphones or computers, assisting schools in organizing complex logistics. AI speech recognition technology has a great edge in providing instant feedback, which is one of its main benefits. While students are speaking out loud, the AI system can swiftly assess their intonation, pronunciation, and tempo, rendering quick guidance and identifying areas for refinement. This personalized feedback effortlessly assists students in boosting their reading abilities, even in the absence of in-person teacher interactions. Moreover, AI-backed evaluations can be carried out on a wider scale, enabling educators to collect extensive data on reading patterns and tackle specific issues that are commonly seen among students. The objective of this presentation is to feature the self-administered AI Speech recognition Computer-based reading assessment that RTI developed at the request of the Philippines Department of Education (DepEd), under the USAID All Children Reading (ACR). Throughout the school years of 2020-2022, the COVID-19 pandemic posed significant challenges to conducting face-to-face assessments, particularly in remote learning environments. As a result, teachers faced constraints in terms of time and resources to individually assess learners' reading skills against crucial learning competencies. The proposed automated assessment technology offered a potential solution to alleviate this burden and streamline the evaluation process, allowing educators to efficiently gauge students' reading abilities remotely. In February 2022, ACR-Philippines initiated discussions with USAID and the Philippines Department of Education (DepEd) to produce a ‘proof of concept’ that explores the feasibility of a self-administered computer-based reading assessment (CoBRA) in English and Filipino for students in the Philippines. The concept found resonance with the DepEd leadership as the adoption of a computer-based format for assessments aligns with international practices and provides an excellent opportunity to ascertain students' preparedness to take computer-based tests, such as the Program for International Student Assessment (PISA). The result of this intervention generated a prototype solution piloted and tailored to fit DepED's existing platforms for supporting remote learning and delivery. The pilot provided insights on the feasibility of a computer-based assessment in the context of the Philippines for students in grades 4 -6. The research findings examined the performance, reliability, and results of the AI Speech recognition technology reading assessment, compared to the assessor-administered approach of the assessment. The research generated key design considerations, feedback from end users, recommendations regarding implementing similar approaches, and the future development of similar technology for other languages within and outside the Philippines.

This reference document is a companion to the Language Complexity Study conducted by RTI International under the All Children Reading–Philippines project in 2020. The study was a secondary analysis of Early Grade Reading Assessment (EGRA) data at multiple time points, looking specifically at the effect of language complexity on reading acquisition of second and third languages (L2 and L3, respectively). The study used the 2013 and 2019 National EGRA data sets to analyze performance according to categories of language complexity based on syllabic complexity, orthographic depth, and other related items. The study was guided by the methodology described by Brunette et al. (2019), who studied the effects of language complexity on reading outcomes in Uganda. The languages selected for analysis were those among the officially supported mother tongue (MT) languages of instruction in the Philippines. Analysts assessed whether the level of complexity was in any way predictive of the average increase or decrease in L2 and L3 scores across schools as measured in 2019 (provided there were schools that reported using MT in 2019, which was not the case for three of the languages: Chavacano, Ivatan, and Sambal.

This study uses national Grade (G) 3 Early Grade Reading Assessment (EGRA) data from 2013, when G3 students learned to read in Filipino and English rather than a mother tongue, and comparable data from 2019, when G3 students would have, according to policy, first learned to read in their mother tongue. The data were used to better understand the role of L1 complexity in L2 and L3 reading acquisition. Sample: 241 schools; 232 schools were the same in 2013 and 2019. Final sample used for analysis: 2,264 G3 students in 2019 and 2,267 G3 students in 2013. Children were assessed in Filipino (L2) and English (L3). Secondary analysis of the data set looked at reading performance and changes in reading performance according to language complexity.

The purpose of this activity was to develop and pilot a CB-EGRA in the Khmer language for validation and use in Cambodia. The CB-EGRA is a paper-based assessment that can be administered with little training to a group of students at one time. It is therefore an arguably more efficient and cost-effective method of obtaining early grade reading data, compared to a traditional EGRA.

This report presents the findings of a deep dive impact evaluation of the United States Agency for International Development (USAID) Uzbekistan Education for Excellence Program (The Program), which was initiated on December 9, 2019, and originally scheduled to end on December 8, 2023, then extended through March 31, 2024. The Program is implemented by a consortium of partners including RTI International (RTI) as the consortium lead and Florida State University and Mississippi State University as partners. For this report, the Program used a comparative analysis of Early Grade Reading and Mathematics Assessments (EGRA and EGMA, respectively) administered at baseline in 2021 and at endline in 2023. The Program evaluation focused on students in grades 2 and 4. In November and December of the 2021-22 school year, the baseline EGRA and EGMA were administered to students who completed grades 2 and 4 and had been in grades 3 and 5 for three months. The endline assessments were administered to students in grades 2 and 4 in May of 2023, a few weeks prior to the end of that school year. This meant that students at endline were being assessed after having completed 68 fewer days of schooling than those who were assessed at baseline.

The Uzbekistan Education for Excellence Program is a 4-year program (December 9, 2019–December 8, 2023) funded by the United States Agency for International Development (USAID) and implemented by the RTI International Consortium in collaboration with the Uzbekistan Ministry of Public Education. The Program conducted baseline Early Grade Reading and Mathematics Assessments (EGRA and EGMA, respectively) in November and December 2021. The purpose of the baseline was to measure the pre-intervention levels of mathematics and Uzbek language reading achievement for students at the end of grade 2 and grade 4.