Feedback about self as a person is the least effective feedback level; task, process, and self-regulation feedback support learning and combine for greater impact¶

Claim · All claims
Evidence · 3 studies (4 entries) · 2 review r1, 1 quant-synthesis r2 · q2–q4 · i0 negligible
Currency · newest study 2025 · 2 of 4 entries checked against their sources (automated) · not reviewed by a person

Subclaims¶

q2 i? Of Hattie and Timperley's four feedback levels, feedback about the self as a person is least effective because it contains no information related to the task; the task, process, and self-regulation levels each support learning. → Fusco 2025 q2 i? Combining the three useful feedback levels increases positive impact on learning. → Fusco 2025 (2)

q2 i0 In the review that proposed the four levels, feedback about the self as a person ("Good girl", "Great effort") is argued to be the least effective level because it carries little task information and draws attention away from the task; the meta-analyses it cites put praise at little or no relationship to achievement (teacher praise 0.12; praise 0.09 against 0.34 for feedback without praise). → Hattie & Timperley 2007

q2 i0 In Kluger and DeNisi's moderator data as Hattie and Timperley tabulate it, feedback designed to discourage the learner averaged −0.14 (49 effects) against 0.33 otherwise, and feedback under much threat to self-esteem averaged 0.08 (102 effects) against 0.47 under little threat (170 effects). → Hattie & Timperley 2007

q2 i? Whether feedback is positive or negative matters less than the level it is aimed at: at the self level negative feedback is more potent than positive, and learners low in self-efficacy are more likely to meet negative feedback with negative affect, lower motivation on the next task and attributions to ability rather than effort (the review's account of earlier studies; no effect size). → Hattie & Timperley 2007

q4 i? Across 607 effect sizes feedback interventions improved performance on average, but over a third lowered it, which sampling error and whether the feedback was positive or negative did not explain; moderator analyses suggested that effectiveness falls as feedback moves attention up from the task toward the self. → Kluger & DeNisi 1996

Evidence¶

Fusco 2025¶

Fusco, J., DiMaria, S., Yeadon, B., Kolodner, J., & Shell, A. (2025). Assessment During Collaborative Learning. Digital Promise. https://doi.org/10.51388/20.500.12265/252

q2 · i? · review · r1

The primer's appendix summarizes Hattie and Timperley's (2007) meta-analytic review of formative assessment and feedback, which identifies four levels: task, processes, self-regulation, and self as a person. No effect size is printed in the primer.

"Of the four levels, feedback about “self as a person” is the least effective for helping further learning, as it contains no information related to the task."

Fusco 2025 (2)¶

Fusco, J., DiMaria, S., Yeadon, B., Kolodner, J., & Shell, A. (2025). Assessment During Collaborative Learning. Digital Promise. https://doi.org/10.51388/20.500.12265/252

q2 · i? · review · r1

The appendix recommends combining the task, process, and self-regulation feedback levels for greater impact, citing Hattie and Timperley (2007). Self-regulation feedback helps learners create internal feedback routines making them more effective self-regulated learners.

"If possible, combine the three levels of feedback about the task, processing, and self-regulation for increased positive impact on learning."

Hattie & Timperley 2007¶

Hattie, J., & Timperley, H. (2007). The power of feedback. Review of Educational Research, 77(1), 81–112. doi:10.3102/003465430298487

q2 · conceptual analysis and narrative review drawing on earlier meta-analyses · i0 · negligible effect of praise, 0.09 against 0.34 without praise (Kluger & DeNisi's data as tabulated); discouraging feedback −0.14 · n=N/A (a review, not a study of individual participants) · review · r?

Read in full text. The review proposes that feedback answers three questions (Where am I going? How am I going? Where to next?) at four levels: the task, the processes needed to do it, self-regulation, and the self as a person. It argues that self-level feedback is the least effective, because it "usually contains little task-related information" and "too often deflects attention from the task". It cites Wilkinson's (1981) meta-analysis of teacher praise (overall effect 0.12) and its Table 3 summary of Kluger & DeNisi (1996): praise 0.09 (80 effects) against 0.34 without; feedback designed to discourage −0.14 (49 effects) against 0.33; much threat to self-esteem 0.08 (102 effects) against 0.47 for little threat (170 effects). It distinguishes praise aimed at the person from praise directed at the effort, self-regulation or processes of the task, which it says can raise self-efficacy and so has more, though still limited, effect. On valence it argues that the level at which feedback is aimed matters more than whether it is positive or negative, and, from earlier studies, that learners low in self-efficacy are more likely to react to negative feedback with negative affect, less motivation on a later task and ability attributions. These are the review's readings of other studies; it pools no new data.

"We argue that FS is the least effective, FR and FP are powerful in terms of deep processing and mastery of tasks, and FT is powerful when the task information subsequently is useful for improving strategy processing or enhancing self-regulation (which it too rarely does)."

Kluger & DeNisi 1996¶

Kluger, A. N., & DeNisi, A. (1996). The effects of feedback interventions on performance: A historical review, a meta-analysis, and a preliminary feedback intervention theory. Psychological Bulletin, 119(2), 254–284. doi:10.1037/0033-2909.119.2.254

q4 · meta-analysis with moderator analyses · i? · the abstract prints no effect size for the self-attention moderator (overall d = .41) · n=607 effect sizes (23,663 observations) · quant-synthesis · r2

Read from the abstract only. A meta-analysis of feedback interventions on task performance in laboratory and field settings, not only education. Feedback improved performance on average (d = .41), but over a third of interventions lowered it, and this could not be explained by sampling error, by feedback sign (positive or negative) or by existing theories. The authors' feedback intervention theory holds that feedback shifts attention among three levels of control (task learning, task motivation, and meta-task processes including the self), and their moderator analyses suggest that effectiveness decreases as attention moves up toward the self and away from the task. The moderator effect sizes are not in the abstract; the figures in the Hattie & Timperley entry above are their tabulation of this study.

"The results suggest that FI effectiveness decreases as attention moves up the hierarchy closer to the self and away from the task."

Discussion¶

What the claim rests on. The ranking of the four levels is Hattie and Timperley's argument, not a comparison any study here ran: no meta-analysis recorded on this page coded feedback by the four levels. What is measured is narrower. Praise and feedback that discourages or threatens self-esteem have near-zero or negative average effects in Kluger and DeNisi's moderator data, against 0.33 to 0.47 in the comparison groups, and the same meta-analysis found effectiveness falling as feedback drew attention toward the self. Those are between-study moderator contrasts across laboratory, workplace and school tasks, so they show association with these feedback features rather than the effect of changing one feedback message. The general claim with both studies' overall effects is Feedback improves learning by a medium amount on average.

Positive or negative is the wrong question. Kluger and DeNisi report that feedback sign did not explain why so many interventions lowered performance, and Hattie and Timperley argue that the level at which feedback is aimed matters more than its valence. For design, the useful contrast is between a comment on the person ("you're a weak writer", "you're so smart") and a comment on the work and what to do next, not between warm and critical wording as such.

Who is hit harder. Hattie and Timperley's account of earlier studies is that learners low in self-efficacy are more likely to react to negative feedback with negative affect, less motivation and ability attributions, while confident learners treat disconfirming feedback as information. No effect size is attached and no claim in the wiki tests the moderation directly; the nearest evidence is an experiment in which negative feedback lowered post-task self-efficacy and raised threat (feedback valence affects self-efficacy, challenge and threat).

Two discrepancies in the sources. Hattie and Timperley summarise Kluger and DeNisi as 131 studies, 470 effect sizes, an average of 0.38 and 32% negative, while the 1996 abstract reports 607 effect sizes, d = .41 and over a third negative. Their text attributes the praise figure to "Kluger and DeNisi (1998)" while Table 3 credits the 1996 paper. The figures above are given as Hattie and Timperley print them.

Learner Variables¶

• Self-Efficacy — moderator: an instructional effect differs with it

Related Claims¶

• Praise for intelligence, rather than for effort, after success undermines children's motivation and performance after a later failure — related
• Feedback valence affects post-task self-efficacy, challenge, and threat — related
• The variability of feedback-intervention effects, including negative effects, exceeds sampling error and is robust to exclusion of a single prolific lab — related
• Feedback Addressing Task Improves Learning — related
• Feedback Answers Three Questions — related
• Feedback improves learning by a medium amount on average, but its effect varies widely with the information it carries, and a substantial share of feedback effects are negative — related
• Self-regulated learning improves achievement — related
• Self Assessment Improves Self Regulated Learning — related
• Established feedback frameworks do not address the epistemic status of feedback tasks relevant to automation decisions — related