Teacher–Student Dialogue During Classroom Teaching: Does It Really Impact on Student Outcomes?

Abstract

It is now widely believed that classroom dialogue matters as regards student outcomes, with optimal patterns often regarded as requiring some or all of open questions, elaboration of previous contributions, reasoned discussion of competing viewpoints, linkage and coordination across contributions, metacognitive engagement with dialogue, and high student participation. To date, however, the relevance of such features has been most convincingly examined in relation to small-group interaction among students; little is known about their applicability to teacher–student dialogue. This article reports a large-scale study that permits some rebalancing. The study revolved around 2 lessons (covering 2 of mathematics, literacy, and science) that were video recorded in each of 72 demographically diverse classrooms (students’ ages 10–11 years). Key measures of teacher–student dialogue were related to 6 indices of student outcome, which jointly covered curriculum mastery, reasoning, and educationally relevant attitudes. Prior attainment and attitudes were considered in analyses, as were other factors (e.g., student demographics and further aspects of classroom practice) that might confound interpretation of dialogue–outcome relations. So long as students participated extensively, elaboration and querying of previous contributions were found to be positively associated with curriculum mastery, and elaboration was also positively associated with attitudes.

Acknowledgments

This research was conducted while all authors were affiliated with the Faculty of Education, University of Cambridge. We wish to thank the Economic and Social Research Council of Great Britain, the participating students, the teachers and head teachers, and the large number of colleagues and postgraduates who assisted with project design, sample recruitment, data collection, and data preparation (especially Ayesha Ahmed, Annabel Amodia-Bidakowska, Sarah Baugh, Elisa Calcagni, and Helen Lancaster, who helped with everything).

Notes

¹ Analyses conducted when two thirds of the classrooms had been visited indicated that dialogue indices computed from two lessons almost perfectly predicted indices computed from three, so long as the lessons covered different subjects (i.e., 90%–95% of the variance was covered). Thus, to expedite data collection and coding, two lessons only were recorded thereafter.

² The desirability of including outcome measures that policymakers and hence schools regard as highly significant was a major reason for the study’s focus on Year 6.

³ As noted, the background research literature and associated professional development emphasize boosting the frequency with which theoretically productive dialogue is used. Therefore, to address this background, the analyses reported here were necessarily also based on frequencies. For this reason, correlations between totals are the appropriate indicators of reliability rather than turn-based indicators like kappa or Krippendorff’s alpha.

⁴ Data reduction across this article is intended merely to simplify measured variables, so following Field (2013) principal component analysis was regarded as appropriate throughout. However, given the controversy within the literature over techniques, all relevant analyses were repeated using factor analysis. Reported results were always replicated (e.g., here the same strong, three-way solution emerged with no cross-factor loadings).

⁵ Two-stage linear step-up procedures (Benjamini, Krieger, & Yekutieli, 2006) confirmed that this approach resulted in no inappropriate exclusions.

⁶ In the interest of brevity, dialogue–outcome relations are not reported where p >.1, and the effects of confound and baseline variables are also omitted. Full models will be supplied on request, as will analyses reported in summary form elsewhere in the article.

⁷ Recall that the lower the ratio, the higher the frequency of Reasoned (or Elaborated) relative to Non-Dialogic.

Additional information

Funding

This research was supported through Grant No. ES/M007103/1 by the Economic and Social Research Council of Great Britain.