Key research papers on student engagement

How the literature is organised

The modern literature dates from Fredricks, Blumenfeld and Paris (2004), a review of definitions, measures, precursors and outcomes that argued for studying engagement as a multifaceted construct with behavioural, emotional and cognitive facets. It carries 12,345 citations in the snapshot behind this page, about four times the next work on the topic. Before it, the motivational tradition of Ellen Skinner and colleagues had shown that teacher behaviour and student engagement influence each other: Skinner and Belmont (1993) followed 144 children and 14 teachers across a school year and found not only that teacher involvement, structure and autonomy support predicted engagement, but that behaviourally disengaged students subsequently received less of all three — a feedback loop worth stating in any study of teacher support. Furrer and Skinner (2003), with 641 children, found relatedness to parents, teachers and peers each independently predicting engagement.

School-based research then developed along two lines: measurement (Appleton et al., 2006; Appleton, Christenson and Furlong, 2008; and the handbook edited by Christenson, Reschly and Wylie, 2012) and relationships, summarised by Roorda et al. (2011), whose meta-analysis of 99 studies found medium-to-large associations between teacher-student relationships and engagement but only small-to-medium ones with achievement. Higher education followed a different route, through institutional surveys of educationally purposeful activities — Carini, Kuh and Klein (2006) found many engagement measures linked positively to critical thinking and grades, though most relationships were weak — and Kahu's (2013) framework, which places engagement within its wider social and institutional context.

Main debates

The main debate is about the definition, and it is not a pedantic one. Appleton, Christenson and Furlong (2008) list three things the field still needs: consensus on the name of the construct, reliable measures of its dimensions, and the construct validation studies to support intervention work. Kahu (2013) is sharper still, identifying poor definitions and a failure to distinguish the state of engagement from the factors that influence it and from its consequences, and proposing a framework to fix that. Wong and Liem (2022) review the current state of the construct and propose refinements.

The practical consequence is that two studies measuring "engagement" may not be measuring the same thing, which is why a literature review should quote each study's definition rather than its label. Whether disengagement is the opposite of engagement or something else is a live question: Lo, Hew and Jong (2024) treat engagement and disengagement as separate outcomes within each of the three dimensions, and find different evidence for each.

Where recent work is heading

Since 2021 the most-cited studies concern engagement in online and blended settings, usually framed by self-determination theory. Chiu (2021), with 426 Grade 11 students, found digital support engaged students better than teacher support by satisfying autonomy, relatedness and competence; Chiu (2022), with 1,201 Grade 8 and 9 students, found all three needs predicting engagement, with relatedness support particularly important. Heilporn, Lakhal and Bélisle (2021) interviewed 20 university teachers and sorted their strategies into course structure and pace, choice of activities, and the teacher's role and relationships.

Meta-analyses connect engagement to outcomes. Tao et al. (2022) put the correlation between perceived teacher support and achievement at r = 0.16 across 71 articles, moderated by grade level and by the type of support, with engagement partially mediating it. Li and Xue (2023) pool 148 effects from 93,188 participants and rank the teacher-student relationship and positive teacher behaviour highest among external influences. Wong et al. (2024) meta-analyse engagement against both achievement and subjective wellbeing.

The newest strand asks how AI affects engagement. Lo, Hew and Jong (2024) reviewed 72 empirical studies published within a year of ChatGPT's release and found robust but narrow evidence on behavioural engagement and on academic dishonesty, mixed evidence on emotional engagement, and broad but weak evidence on cognitive engagement — including reduced critical thinking and overreliance as forms of disengagement. Wang and Xue (2024) and Almusaed et al. (2023) are the intervention-side counterparts.