Promoting Academic Vocabulary and Reading Comprehension Through Blended Learning: Evidence from the Golden Triangle Intervention Among Arabic-Speaking Middle School Students

Arabic-speaking adolescents face distinctive literacy challenges arising from diglossia, morphological complexity, and limited access to academic vocabulary. This quasi-experimental, non-randomized mixed-methods study examined intervention-associated changes following the Golden Triangle, a five-month blended-learning program integrating academic-vocabulary instruction, reading comprehension, digital practice, and teacher professional development. Twenty-three Arabic-medium middle schools participated (17 intervention; 6 comparison), and 731 seventh-grade students were recruited. Primary maximum-likelihood mixed-effects models included time, condition, and their interaction, with student and school random intercepts. Model-based Wald inference indicated additional gains in receptive vocabulary (B = 6.25, 95% CI [1.82, 10.67], nominal p = 0.006) and Grade-7 reading comprehension (B = 4.88, 95% CI [0.39, 9.37], nominal p = 0.033), but no statistically detectable additional gain on the Productive Academic Vocabulary Test (PAVT; B = 1.54, 95% CI [−1.92, 4.99], p = 0.384). These intervals and tests did not incorporate a small-cluster correction. A separate school-clustered sensitivity analysis yielded positive estimates but confidence intervals that included zero for all three outcomes. Because PAVT combines lexical retrieval with morphological and sentence-context demands, its nonsignificant result does not establish an absence of productive vocabulary development. Grade-8 follow-up used a different, non-equivalent reading-comprehension measure and therefore concerns an association rather than maintenance on the same scale. Selective retention limits interpretation. Thematic content analysis of 14 teacher interviews identified perceived pedagogical benefits alongside technological, scheduling, and resource constraints. The findings are suggestive and model-dependent, rather than definitive evidence of causal or sustained literacy gains.