Dialogic reading restructures shared reading so that children actively describe, recall, predict, explain, and connect ideas while the adult prompts, evaluates, expands, and invites another response. Reviews of the technique indicate that these routines can improve young children’s oral language, particularly vocabulary, when prompts and feedback are implemented coherently rather than added as occasional questions. The relevance to Uganda goes beyond importing an established method: the national Primary 1 Luganda teacher guide and pupil book already include picture description, oral literature, speaking and listening, pair activity, continuous assessment, and a gradual release sequence described as “I do, We do, You do.” In practice, however, these affordances are often compressed into teacher explanation, short display questions, choral repetition, and copying. The pedagogical challenge was therefore to make the materials’ interactive potential routine and feasible in ordinary classrooms.

The trial was conducted in central Uganda, where twenty schools, each contributing one Luganda-medium Primary 1 classroom, were randomized after baseline assessment to either the dialogic reading adaptation or usual instruction. Twenty-three children in the intervention arm and 231 in the comparison arm formed the baseline sample of 463. An independent statistician with no role in trial delivery generated and implemented the concealed allocation sequence, and outcome assessors were masked to school allocation. Teachers could not be masked because the intervention required them to change their classroom practice. The intervention was delivered for ten instructional weeks from October to December 2024, with post-test assessment completed in December 2024.

Intervention teachers retained the approved Luganda pupil book and teacher guide. Across the ten weeks, they revisited six thematic bundles during regular literacy and oral-language periods. Each twenty to twenty-five-minute routine previewed a focal picture or short text, introduced three or four target words, elicited increasingly independent child talk, and ended with a brief retell, partner response, or contextual sentence. The prompt repertoire drew on the PEER framework of Prompt, Evaluate, Expand, Repeat and the CROWD categories of Completion, Recall, Open-ended, Wh-question, and Distancing, but translated these categories into familiar classroom moves. During the “I do” phase, the teacher modelled naming, noticing, and a complete response. During “We do,” the teacher prompted shared and partner responses, paused for thinking time, evaluated meaning, and expanded short answers. During “You do,” children pointed, named, explained, connected a picture or event to their experience, or retold to a peer. Teachers were encouraged to accept linguistically appropriate Luganda variants and to extend rather than replace a child’s response.

Teachers attended a two-day practical workshop involving demonstration, prompt rehearsal, collaborative lesson planning, and microteaching. Three brief coaching visits followed during implementation, each focused on one observable practice: the coach modelled or discussed the move, observed the teacher use it, reviewed the children’s responses, and agreed one focus for the next lesson. Comparison classrooms continued usual Luganda literacy instruction with the same pupil book, teacher guide, school timetable, and assessment expectations. Usual instruction commonly comprised teacher explanation or reading from the front, pointing to print or pictures, short-answer questioning, choral repetition, and closely directed oral or written practice. Some teachers also used brief discussion, so the condition was not defined as an absence of interaction. Comparison teachers did not receive the dialogic prompt training, coaching, structured prompt cycles, or planned gradual-release routines during the study.

The researcher-developed outcome measures were reviewed by three early-grade literacy and curriculum experts at Makerere University, Kyambogo University, and the Uganda National Institute of Teacher Education for linguistic accuracy, content coverage, and cultural appropriateness. They were then piloted with thirty non-participating Primary 1 children from a comparable Luganda-medium school. Vocabulary was assessed through twenty Luganda items spanning receptive identification and expressive naming or explanation, with fifteen items sampling words taught directly in the thematic bundles and five using untaught pictures or contexts to assess transfer. Expressive responses were coded two for a full correct explanation or contextual sentence, one for a correct isolated target word without expansion, and zero for an incorrect, irrelevant, or absent response. Internal consistency was Cronbach’s alpha of 0.85. Listening comprehension was assessed after an unfamiliar Luganda story through twelve orally presented items covering literal recall, sequence, and simple inference, scored on a zero to twelve scale with internal consistency of 0.78. Assessors completed three days of training and had to achieve at least ninety percent accuracy on a calibration test before assessing participants. Ten percent of assessments were double-scored, yielding inter-rater agreement of Cohen’s kappa of 0.88.

Analyses followed the intention-to-treat principle, with every child analyzed in the arm to which the child’s school was assigned regardless of intervention dosage or teacher fidelity. Linear mixed-effects models adjusted for baseline score, age, recorded sex, and classroom clustering, using the Kenward-Roger correction for small-sample inference because only twenty clusters were randomized. Missing post-test outcomes were addressed by multiple imputation with twenty imputations, and a complete-case analysis was retained as a sensitivity check. Post-test data were obtained for 439 children, or 94.8 percent of the baseline sample, with 24 children lacking a post-test measure.

At post-test, children assigned to dialogic reading had higher adjusted scores on both prespecified outcomes. For vocabulary, adjusted means were 12.6 in the intervention arm and 9.4 in the comparison arm, an adjusted difference of 3.2 points on a twenty-point scale, with a 95 percent confidence interval of 2.3 to 4.1 and p less than .001, corresponding to Hedges’ g of 0.46. For listening comprehension, adjusted means were 7.8 and 6.3 respectively, an adjusted difference of 1.5 points on a twelve-point scale, with a 95 percent confidence interval of 0.4 to 2.6 and p of .009, corresponding to Hedges’ g of 0.31. The larger standardized effect for vocabulary than for listening comprehension is consistent with the distinction between proximal and broader outcomes reported in dialogic-reading and emergent-literacy research. Fifteen vocabulary items sampled directly taught content while five assessed transfer, so the results show substantial learning on an intervention-aligned composite but are not interpreted as evidence of generalized language development beyond the tested transfer contexts. Listening comprehension used an unfamiliar story and thus provides a broader, though still short-term, outcome.

Across the three observation visits, mean intervention fidelity was 82 percent, with classroom means ranging from 61 to 94 percent. Teachers most consistently used the “I do, We do, You do” structure and introduced target words, while expansion of child responses and distancing prompts were less consistent. Because fidelity varied only across ten intervention clusters and no component-level mediation analysis was prespecified, fidelity was not modelled as a predictor of child outcomes. The observation data therefore describe implementation and do not identify the mechanism of the treatment effect. No unintended effects or adverse events were reported by head teachers or recorded during coaching observations.

The findings are compatible with a social-constructivist account in which guided participation supports increasingly independent language use, but the trial did not test mediation. The study advances the literature through contextual integration rather than by claiming a new prompting taxonomy: established PEER and CROWD principles were aligned with the Luganda curriculum’s own gradual-release sequence, picture-description activities, and oral-literature periods, and the same books and timetable were present in both arms. Accordingly, the contrast primarily concerned the systematic use of prompts, response expansion, partner talk, and child-led practice, supported by training and coaching. Beyond Uganda, the findings offer a curriculum-implementation lesson rather than a transportable effect estimate. Familiar-language materials may contain participatory activities, but these affordances can remain underused unless teacher preparation and implementation support specify how interaction should unfold. For curriculum developers and school systems, distributing texts should therefore be accompanied by teacher guides, professional learning, coaching, and observation tools that make prompt-feedback cycles, wait time, response expansion, partner talk, and gradual release visible in practice. For language education in multilingual systems, the transferable principle is alignment among classroom language, curriculum materials, teacher support, and assessment, not the direct export of Luganda prompts or the present effect estimates. Implementation feasibility should be interpreted cautiously: the adaptation avoided the cost of replacement pupil books, but it was not resource-free, as teachers attended a two-day workshop and received three coaching visits, had to protect twenty to twenty-five-minute talk routines within crowded timetables, and needed skill in managing multiple child responses. Before national scale-up, future research should estimate training and coaching costs, test lighter-touch support, document teacher workload, examine sustainability after coaching ends, and assess whether material availability is adequate across schools. Several strengths support interpretation: randomization occurred only after recruitment and baseline testing, sequence generation and assignment were independent, allocation was concealed, assessors were masked, attrition was low, analysis followed intention to treat with small-sample correction and imputation, and measures showed acceptable internal consistency and strong inter-rater agreement. Limitations remain, as only twenty clusters were randomized, so precision depends on the Kenward-Roger adjustment and findings require replication, and the study involved Luganda-medium Primary 1 classrooms in central Uganda, which limits generalization to other languages, grades, regions, and school systems. Future studies may use larger cluster samples, include delayed measures and direct reading outcomes, examine the taught-versus-transfer structure of vocabulary effects, and compare coaching models. The trial supports the promise of using existing familiar-language materials more interactively; it does not establish the causal contribution of individual prompt types, long-term reading effects, or national scalability. Replication with larger samples, preregistration, longer follow-up, and implementation-cost evidence is the appropriate next step.

Subject of Research: Education

Article Title: Effects of dialogic reading with Luganda primary 1 materials on children’s vocabulary and listening comprehension

Article References: Kabugo, D. (2026). Effects of dialogic reading with Luganda primary 1 materials on children’s vocabulary and listening comprehension. Discover Education, 5(1), Article 1182. https://doi.org/10.1007/s44217-026-02271-6

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02271-6

Keywords: dialogic reading, Luganda, Uganda, vocabulary, listening comprehension, cluster-randomized trial, early literacy, language education, Effects, dialogic, reading, primary