<p>Despite decades of research, the impact of feedback timing on learning outcomes remains debated. The most recent meta-analysis directly comparing immediate and delayed feedback was published in 1988, preceding the transformation of computer-assisted learning environments. This study reevaluates the effect of feedback timing on learning outcomes within computer-assisted learning environments. Three key aspects were examined: (a) the overall effectiveness of immediate versus delayed feedback, (b) the influence of varying definitions of feedback delay on learning outcomes, and (c) study characteristics and feedback features that may account for prior inconsistencies. A systematic meta-analysis was conducted on 51 studies published from 1988 to 2024, analyzing 160 effect sizes through meta-regression with robust variance estimation. Across the included studies, delays implemented in delayed feedback conditions ranged from 1 second to 7 days in time-based designs and from 1 to 60 intervening items in item-based designs. Results indicate that feedback timing does not significantly influence learning outcomes on average (<InlineEquation ID="IEq1"> <EquationSource Format="TEX">\(g = 0.03\)</EquationSource> </InlineEquation>, 95% CI [<InlineEquation ID="IEq2"> <EquationSource Format="TEX">\(-0.08\)</EquationSource> </InlineEquation>, 0.13], <InlineEquation ID="IEq3"> <EquationSource Format="TEX">\(p =.61\)</EquationSource> </InlineEquation>). Substantial heterogeneity between studies suggests the effect is moderated by key factors. Moderator analyses reveal that educational level, learning domain, and response time constraints significantly influence feedback timing effectiveness, providing a partial explanation for inconsistencies across studies. Yet, partial confounding of moderators limits the ability to fully disentangle their individual contributions. These findings suggest that while feedback timing alone is not a decisive factor, its impact is context-dependent. Although approximately half of the effects involved multi-hour or <InlineEquation ID="IEq4"> <EquationSource Format="TEX">\(\ge\)</EquationSource> </InlineEquation> 10-item delays, there were relatively few studies involving delays of one day or longer. As a result, the reviewed studies provide limited evidence concerning the common multi-day delays typical of many classroom settings.</p>

A Meta-Analysis of the Impact of Feedback Timing on Learning Outcomes in Computer-Assisted Learning

期刊论文
发表日期：

2026年1月26日

查看全文

Erva Nihan Kandemir,
Emanuele Esposito,
Lilas Gurgand,
Franck Ramus

摘要

Despite decades of research, the impact of feedback timing on learning outcomes remains debated. The most recent meta-analysis directly comparing immediate and delayed feedback was published in 1988, preceding the transformation of computer-assisted learning environments. This study reevaluates the effect of feedback timing on learning outcomes within computer-assisted learning environments. Three key aspects were examined: (a) the overall effectiveness of immediate versus delayed feedback, (b) the influence of varying definitions of feedback delay on learning outcomes, and (c) study characteristics and feedback features that may account for prior inconsistencies. A systematic meta-analysis was conducted on 51 studies published from 1988 to 2024, analyzing 160 effect sizes through meta-regression with robust variance estimation. Across the included studies, delays implemented in delayed feedback conditions ranged from 1 second to 7 days in time-based designs and from 1 to 60 intervening items in item-based designs. Results indicate that feedback timing does not significantly influence learning outcomes on average ( \(g = 0.03\) , 95% CI [ \(-0.08\) , 0.13], \(p =.61\) ). Substantial heterogeneity between studies suggests the effect is moderated by key factors. Moderator analyses reveal that educational level, learning domain, and response time constraints significantly influence feedback timing effectiveness, providing a partial explanation for inconsistencies across studies. Yet, partial confounding of moderators limits the ability to fully disentangle their individual contributions. These findings suggest that while feedback timing alone is not a decisive factor, its impact is context-dependent. Although approximately half of the effects involved multi-hour or \(\ge\) 10-item delays, there were relatively few studies involving delays of one day or longer. As a result, the reviewed studies provide limited evidence concerning the common multi-day delays typical of many classroom settings.