Beyond the Score: What Contextual Variables Reveal About Foundational Learning

FLS 2022 did more than measure what Grade 3 children could read or calculate. Its background questionnaires recorded homes, classrooms, teachers and school conditions within which foundational learning was taking place.

Two children may receive the same numeracy score, yet arrive at it through very different circumstances. One may have storybooks at home, a short walk to school and a classroom with electricity. Another may learn in a different language from the one spoken at home. The child may travel farther, have fewer reading materials and study in a school with limited facilities. The score records what each child demonstrated during the assessment. It does not, by itself, describe the opportunities that preceded the response. Contextual variables help restore that missing part of the picture. They include the circumstances reported by children, teachers and school heads. They do not replace achievement data. Nor do they lower the importance of proficiency benchmarks. They help explain where learning takes place and which conditions are associated with different patterns of performance. This is why the Foundational Learning Study (FLS) 2022 included more than literacy and numeracy tasks. Alongside the one-to-one assessment of approximately 86,000 Grade 3 children from about 10,000 schools, it administered questionnaires to pupils, teachers and school heads. More than 18,000 teachers participated in the study. The sample covered State government, government-aided, private recognised and central government schools. The assessment established a baseline for the National Initiative for Proficiency in Reading with Understanding and Numeracy, or NIPUN Bharat. The questionnaires placed that baseline within the lived conditions of learning. A learning level tells us where the child stands. Context helps us see the ground on which the child has been asked to stand.

The three questionnaires looked at different parts of that ground. According to the FLS 2022 Analytical Report, the Pupil Questionnaire asked about home language and the language of instruction, attendance at pre-primary classes or Anganwadi centres, family size, activities enjoyed, activities undertaken with family members, classroom participation and travel to school. The Teacher Questionnaire covered teaching experience, educational and professional qualifications, training, familiarity with learning outcomes and Lakshyas, use of teaching-learning materials, classroom conditions, print-rich environments, assessment procedures, the presence of children with special needs and the teacher’s commute. The School Questionnaire, answered by head teachers, covered facilities, health check-ups, classroom spaces, functional toilets, drinking water, accessibility for children with special needs, School Management Committees, Parent-Teacher Meetings and community support. These were not test questions with correct and incorrect answers. They recorded reported conditions, practices, views and activities. The analytical report used the information in two ways. First, it profiled the backgrounds of students, teachers and schools. Second, it examined whether selected groups differed in literacy or numeracy performance. The national association results in this chapter were computed specifically for the analytical report. Similar analyses appeared at the State level in the State Report Cards. This distinction matters. A national relationship may not be equally strong, or even move in the same direction, in every State or language setting. These are also distinct uses. A profile may show how common a resource or practice was. An association may show that two groups had different mean scores. Neither, on its own, establishes that one condition caused the difference. The contextual chapter is therefore a map of patterns, not a verdict on children, families, teachers or schools.

The method matters because the word association is easily mistaken for explanation. Most questionnaire variables were nominal or categorical. Several had two possible responses. Questions with more than two response categories were reduced to two groups for analysis. For example, responses such as “never” and “sometimes” could be placed in one group, while “most of the time” and “almost always” formed another. The report then calculated the difference between the groups’ mean achievement scores. It used Cohen’s d to judge practical or educational significance. Values of 0.2 and above were treated as small, 0.5 and above as moderate, and 0.8 and above as large. The reduction to two categories made comparisons easier to interpret. It also compressed differences within each group. A child who read occasionally and one who never read could, depending on the item, be placed together. The analysis also differed across domains. Literacy associations were calculated separately for subtasks such as decoding words, oral reading fluency, phonological awareness and oral reading fluency with comprehension. Numeracy associations used a composite numeracy score. A contextual variable could therefore appear in relation to selected literacy components without showing a comparable association with overall numeracy. These results should be read at the scale on which they were calculated. A small effect on one literacy subtask is neither an explanation of all literacy performance nor evidence about every child.

The child-level findings offer one of the clearest examples. Children who reported that they asked questions in class had higher mean scores on several literacy subtasks than those who reported that they did not. The reported effect sizes were 0.26 for decoding words, 0.27 for decoding non-words, 0.20 for oral reading fluency, 0.23 for phonological awareness, 0.21 for oral reading fluency with comprehension and 0.25 for decoding letters. Children who reported reading material in addition to textbooks also had higher scores across the same broad group of literacy subtasks. The effect sizes were 0.30 for decoding words, 0.29 for decoding non-words, 0.24 for oral reading fluency, 0.27 for phonological awareness, 0.26 for oral reading fluency with comprehension and 0.26 for decoding letters. All were in the report’s small-effect range. The profile data provide useful context. Nationally, 82% of children said they liked reading books other than textbooks. Yet preference is not the same as access, frequency or sustained engagement. The reported association does not prove that additional reading alone produced the higher scores. Children who read more may also have stronger prior skills, greater access to print, more encouragement or better-resourced schools. Asking questions may reflect confidence, classroom safety, teacher responsiveness or existing understanding. Still, the pattern carries an important educational message. A classroom that invites questions and a learning environment that extends reading beyond the textbook are visible parts of the foundational-learning ecosystem.

Other child and home variables produced a more mixed picture. The relationship between the language spoken at home and the medium of instruction differed across literacy subtasks. Where the two languages were different, the report recorded lower performance for oral reading fluency with comprehension, but an opposite pattern for oral reading fluency and decoding letters. Each of these effects was small. This prevents a simplistic claim that language difference affected every literacy skill in one direction. It also confirms why literacy results must remain connected to the language, task and assessment design. Attendance at pre-primary classes or an Anganwadi did not reach the report’s 0.2 threshold on the listed literacy subtasks. Nor did playing games, storytelling with family members, playing games with family members or travel time reach that threshold in the national literacy association tables. Their numeracy effect sizes were also below 0.2. These findings do not show that early childhood education, family interaction, play or school access are unimportant. A single self-reported item may not capture the duration, regularity or quality of an experience. The profile itself showed a rich world of children’s interests. Ninety-three per cent said they liked playing games. Ninety-one per cent liked art and craft. Eighty-five per cent liked playing with toys, and the same share liked caring for plants or animals. Seventy per cent liked exercise or yoga. Fifty-five per cent reported walking to school. Contextual analysis is strongest when it opens a question for closer study, not when it closes the question with a convenient conclusion.

Teacher variables require the same care. At the national level, 56% of teachers reported more than five years of experience at the foundational stage, while 45% reported being graduates. Blackboards were the most frequently used teaching-learning material, reported by 88% of teachers. Textbooks followed at 86%. Twenty-eight per cent reported that computer resources or audio-visual aids were unavailable, and 14% reported unavailable library resources. Assessment practice remained varied. Paper-and-pencil class tests were the most frequently reported technique at 49%. Group activities were reported by 44%, oral work by 41%, peer work by 40% and observation by 39%. In the literacy association analysis, frequent use of observation reached a small effect only for oral reading fluency, with Cohen’s d of 0.20. The listed differences for class tests, group activity, peer work, oral work, maintaining a teacher diary and using portfolios did not reach 0.2 across the literacy subtasks. Most teacher-related numeracy effects were also below the threshold. In-service training, as represented by the relevant questionnaire item, did not show a practically significant national association with the assessed outcomes. This should not be read as proof that training or formative assessment is ineffective. The analysis did not randomly assign teachers to a programme. It did not measure the quality, duration, content or classroom transfer of every training experience. Nor did a single frequency item show how well an assessment technique was used. The defensible conclusion is limited: the selected teacher variables, as grouped and measured in FLS 2022, generally showed small or negligible national associations with the reported outcomes.

School conditions produced a more visible set of literacy associations. In the national school profile, 97% of schools reported safe windows and ventilation in classrooms at the foundational stage. The same proportion reported textbooks or reading materials. Electricity and proper lighting were reported for 91%. Storybooks were reported for 89%, locally available or developed teaching-learning materials for 88%, and toys or play equipment for 84%. Computers were reported for 63% and internet access for 56%. School Management Committees were reported by 95% of head teachers. Monthly Parent-Teacher Meetings were reported by 56% of schools, while 17% reported that such meetings were not organised. The association tables showed higher scores on selected literacy subtasks in schools reporting electricity and proper lighting. Effect sizes were 0.21 for decoding non-words, 0.20 for oral reading fluency and 0.23 for phonological awareness. Basic drinking water showed effects of 0.21 for oral reading fluency and 0.24 for phonological awareness. Basic handwashing facilities showed effects of 0.24 for decoding non-words, 0.22 for oral reading fluency and 0.24 for phonological awareness. A readily available medical room was associated with higher performance on five literacy subtasks. The effects ranged from 0.20 for phonological awareness to 0.32 for oral reading fluency. Schools reporting computers had small positive effects across decoding words, decoding non-words, oral reading fluency, phonological awareness and decoding letters. The largest among these was 0.35 for oral reading fluency. Internet access showed a similar pattern, including 0.35 for oral reading fluency. Textbooks and reading materials showed an effect of 0.22 for oral reading fluency. Yet the school-related effect sizes for composite numeracy were all below 0.2. Electricity, toilets, drinking water, accessibility, storybooks, computers and internet access therefore did not produce the same reported pattern across the two domains. These findings do not say that hardware teaches a child to read. They show that resources often travel together with broader institutional capacity, learning opportunities and school conditions.

One result demonstrates why causal language would be especially unsafe. Schools that reported serving mid-day meals daily had lower mean literacy scores than schools that did not, with a moderate effect of -0.52 for oral reading fluency and smaller negative effects for several decoding-related subtasks. The corresponding numeracy effect was -0.12, below the report’s practical-significance threshold. It would be wrong to infer that a school meal reduces learning. Mid-day meals are a social-support intervention, and the schools or children covered by such support may differ in many other ways from the comparison group. The simple association does not adjust for all those differences. The same caution applies when a facility, qualification or practice appears positively related to a score. Computers may be accompanied by stronger infrastructure, staffing, location or funding. Children who ask questions may already be confident readers. Schools with medical rooms may differ systematically from schools without them. Contextual variables can also interact. Home language may meet classroom language, teacher practice and print access in different combinations. Geography, school management and the social composition of the served population may influence both the presence of a facility and the observed score. A two-group comparison cannot separate these pathways. A background questionnaire identifies where relationships appear. It does not isolate the mechanism that produced them. For policy, this is not a weakness to be ignored. It is a guide to the next level of inquiry. Associations can identify conditions that deserve disaggregated analysis, adjusted statistical models, qualitative follow-up or targeted monitoring. State-level reports are important here because a national average can conceal different patterns across languages and jurisdictions. Repeated evidence across settings would strengthen a hypothesis. A single isolated association should invite verification. The results should never be used to blame a child’s family, dismiss a welfare measure or declare one classroom practice universally superior.

The larger value of contextual variables lies in moving the public conversation from Who scored more? to What opportunities surrounded the score? The point is not to explain away low achievement. It is to identify the conditions that a system can improve. Responsibility remains collective because many of those conditions lie beyond the control of a Grade 3 child. The FLS 2025–26 Assessment Framework identifies reliable, valid and comparable evidence on foundational learning as a central purpose. It aims to monitor progress against the FLS 2022 baseline, identify learning gaps and support policy formulation, resource allocation, teacher training, curriculum development and early-grade interventions. These objectives make context relevant even when the headline remains an achievement level. Context can help distinguish a universal need from a concentrated one. It can show whether a reading-resource gap calls for materials, whether a classroom-participation pattern calls for pedagogic support, or whether a facility gap requires administrative action. It can also prevent a national average from becoming a one-size-fits-all prescription. When future contextual evidence is reported, it should be read alongside the linked literacy and numeracy results. The useful questions are precise. Have learning gains reached children across languages and school-management types? Are improvements shared where reading materials, digital resources or basic facilities remain limited? Do States see the same patterns as the national analysis? Which associations remain after related background conditions are considered together? The answers may differ across States, languages and subtasks. That variation is evidence, not inconvenience. FLS 2022 established that foundational learning is measurable. Its contextual analysis added an equally important insight: learning is never produced by the child alone. A fair reading of FLS must keep achievement, opportunity and evidence in the same frame. The score may be the visible result. Context is the system around it.