How Knowledge of Performance Works | Show the Learner How the Action Was Executed Without Turning Feedback Into a Script

eduKateSG Learning Node Series · 0269

A learner can know that the shot missed and still have no idea what in the movement produced the miss.

That gap separates knowledge of results from knowledge of performance. The result says the ball landed 20 centimetres left. The performance information says the release came late, the racket face opened, the balance shifted, or the movement path curved. One describes the outcome. The other describes features of the action that produced it.

In motor learning, knowledge of performance—often shortened to KP—is a form of augmented feedback about movement execution or movement quality. It can be verbal, visual, video-based, kinematic, acoustic or sensor-generated. It can be descriptive, such as “your trunk rotated before the arm accelerated”, or prescriptive, such as “keep the target line stable through the release”.

KP is powerful because many important errors are difficult for novices to feel accurately. It is risky because external information can become so detailed, frequent or body-focused that the learner stops developing an internal error-detection system.

Knowledge of performance works when external information helps the learner connect what the action felt like with what the action actually did—then becomes less necessary as that connection improves.

The 50-Second Read

Knowledge of results tells the learner about the outcome.
Knowledge of performance tells the learner about the movement or action pattern that produced the outcome.
KP is useful when the critical performance feature is hard to perceive intrinsically.
Video, force traces, joint-angle displays and coach observations can all become KP.
Descriptive KP says what happened; prescriptive KP suggests what to change.
More information is not automatically better information.
Feedback about ten movement features can exceed what the learner can use on the next attempt.
Body-part feedback can also shift attention internally; this can interact with task and expertise.
Retention and transfer without feedback matter more than practice performance with feedback.
The learner should increasingly estimate the error before seeing the external feedback.
KP should protect the target skill, not replace the learner’s own perception and correction.
The end state is not perfect feedback delivery. It is useful performance when the feedback source is absent.

Canonical Owner Boundary

This node owns augmented information about how an action was executed. How Knowledge of Results Works owns outcome feedback. How Attentional Focus Instructions Work owns where attention is directed. How Self-Controlled Feedback Works owns learner control over when feedback is requested. This article asks a narrower question: what should information about the action itself do for learning?

1. Outcome and execution answer different questions

Imagine an archer whose arrow lands low. Knowledge of results says “eight centimetres below centre”. That is useful because it defines the outcome error. It does not identify whether the bow moved, the release timing changed, posture collapsed or the sight picture was wrong.

KP might add one of those missing execution details. The information is especially valuable when several different movement patterns can produce the same outcome error. Two learners can miss low for different reasons and therefore require different next attempts.

The distinction is not absolute. In some tasks, the outcome reveals the movement. In others, execution quality matters even when the visible result succeeds. A gymnast can land a skill while using a compensatory pattern. A musician can reach the correct note while producing excessive tension. A surgical trainee can complete a sequence but use a hand path that becomes fragile under pressure.

2. The learner already has intrinsic feedback

Movement generates information before a coach says anything. Vision, proprioception, touch, balance, sound and task outcome all provide intrinsic feedback. The learner feels force, sees trajectory, hears contact and senses timing.

Augmented KP should therefore be understood as additional information, not the only information. This matters because learning involves improving how intrinsic signals are interpreted. If the coach immediately announces every correction, the learner can become highly efficient at following the coach while remaining poor at detecting the same error alone.

A useful practice sequence is often: attempt → notice → estimate → receive selected external information → compare → adjust. The estimation step forces the learner to interrogate intrinsic feedback before an external answer arrives.

3. Descriptive KP and prescriptive KP are not the same

Descriptive KP reports a feature of the completed performance: “your contact point was behind the body” or “the second phase took longer than the first”. Prescriptive KP proposes a future change: “contact farther forward” or “keep the two phases closer in duration”.

Prescriptive feedback can reduce search cost for a novice. It can also make the learner passive if every error comes with a complete correction. Descriptive feedback leaves more inference work to the learner but can be unhelpful when the learner lacks the knowledge needed to generate a repair.

The choice should depend on what the learner can already do. If they can accurately recognise the problem and propose a valid correction, descriptive KP may be enough. If they cannot yet distinguish the relevant feature from irrelevant noise, a small prescriptive cue may be justified.

4. Video becomes useful only when the learner knows what to look for

Slow-motion video can reveal information that disappears during a fast movement. But a high-resolution replay contains far more information than a learner can use. Without a search target, the learner may notice clothing, speed, apparent awkwardness or a visually dramatic feature that is not causally important.

Good video KP narrows the comparison. Mark one line, one event, one timing relation or one external effect. Compare an attempt with a reference or with the learner’s own successful trial. Ask the learner to predict what the replay will show before playing it.

The objective is to turn video from entertainment into calibration. Eventually, the learner should be able to say “that release felt late” before seeing the replay—and be increasingly correct.

5. Instrumented KP can reveal hidden variables

Force plates, motion capture, pressure sensors, timing systems and wearable devices can generate KP that humans cannot observe directly. A graph might show asymmetry, peak force, timing, path deviation or acceleration.

This does not mean every measured variable deserves feedback. Measurement capability can outrun instructional usefulness. A system can calculate twenty biomechanical features while the learner can productively change one.

Select the variable because it is tied to the target performance and because changing it is plausibly actionable. Otherwise the technology becomes an information generator rather than a learning system.

6. More precise feedback can create less precise learning

Suppose a sensor reports the knee angle after every jump to one decimal place. The number looks exact. The learner may begin chasing 92.4° instead of understanding the movement effect the angle is supposed to support.

Measurement precision is not instructional precision. If a range is acceptable, a binary or bandwidth message may be more useful than continuous correction. If the underlying estimate itself carries error, displaying many decimals can create false certainty.

Feedback should preserve the tolerances of the real skill. Good performance often has many viable movement solutions rather than one geometrically perfect template.

7. KP can improve practice while weakening error detection

This is the central dependency risk. During practice, continuous external information can stabilise performance. When the information disappears, performance may deteriorate because the learner never learned to recognise and correct the error independently.

The classic guidance hypothesis proposed that excessive augmented feedback can guide performance so strongly that learning becomes dependent on the guidance. Later evidence is mixed rather than universally confirming one rule. A 2021 systematic review of augmented feedback in motor learning describes substantial variation across task, population, feedback form and retention design. That heterogeneity is itself important evidence against a universal “less feedback is always better” recipe.

The appropriate question is not “How little feedback can we give?” It is “What information does this learner need now, and what evidence will show that the learner can later perform without it?”

8. KP can accidentally direct attention to the body

Many KP statements naturally reference joints and body segments: “bend the knee”, “rotate the shoulder”, “keep the wrist firm”. That can be necessary when body configuration is itself the target—for example in some rehabilitation, safety or technical correction contexts.

But attentional-focus research shows that instructions and feedback can alter performance depending on whether attention is directed internally toward body movement or externally toward movement effects. The relationship is task-dependent and should not be simplified into a ban on anatomical information.

A useful translation is to ask whether the same execution information can be expressed through its external consequence. “Extend the arm faster” may sometimes become “accelerate the implement through the target”. The better wording depends on what information the learner actually needs.

9. One corrective feature at a time can beat a complete movement diagnosis

A novice swimming stroke can contain ten visible errors. Correcting all ten after one length creates a memory problem before it creates a movement solution.

Select the error with the greatest leverage: the one that drives several downstream problems, creates a safety issue, blocks the next stage, or can be changed with a clear cue. Let the learner practise that relation long enough to test whether the predicted downstream change appears.

This is not a law that feedback must contain exactly one point. It is a design principle: feedback bandwidth should match the learner’s capacity to act on it.

10. KP should connect sensation to consequence

The most valuable feedback often creates a three-part link: what I felt → what actually happened → what consequence followed.

For example: “You felt as though you stayed centred. The pressure trace shows that you moved left just before release. That early shift is why the implement path also moved left.” The learner now has a sensory hypothesis to test on the next attempt.

Over time, the external measurement becomes less necessary because the learner’s own estimate improves. This is the calibration role of KP.

11. Cross-Domain Comparison: Debugging Software

A program produces the wrong output. The output error is knowledge of results. A stack trace showing where execution diverged is knowledge of performance.

But giving a novice the entire trace can overwhelm them. A good debugger identifies the relevant branch, asks what the programmer expected at that point, and helps them compare expected and actual state. Eventually the programmer learns to form better diagnostic hypotheses before opening the debugger.

The analogy is useful because both systems distinguish outcome error from process error. It also has limits: human movement is adaptive, variable and embodied rather than deterministic code.

12. Cross-Domain Comparison: Instrument Flying

A pilot can know that the aircraft deviated from altitude. More useful training asks which control inputs, trim settings, scan pattern or delayed corrections generated the deviation.

Instructor feedback that simply flies the aircraft verbally for the learner may produce a smooth training flight and poor independent control. Useful process feedback improves the trainee’s ability to detect drift before the instructor speaks.

13. Classroom Translation: Written Performance Has a Process Too

KP is a motor-learning term, so it should not be copied carelessly into every academic task. But the information distinction transfers conceptually.

A Mathematics result says “the answer is wrong”. Process information identifies where the transformation changed sign or where the representation no longer matched the problem. A writing result says “the paragraph lacks evidence”. Process information identifies that the student made a claim, selected a quotation, but never explained the relation between the two.

The classroom lesson is to separate outcome from process, while preserving the canonical owners of academic feedback. The teacher should not import motor-learning evidence as if algebra and a tennis serve were identical tasks.

14. Failure Mode: The Coach Narrates Every Movement

The learner receives a continuous stream: shoulder, wrist, elbow, foot, timing, head, angle, speed.

Repair: decide what feature matters most for this attempt. Ask the learner what they noticed. Provide selected KP. Retest without the cue.

15. Failure Mode: KP Becomes a Fixed Ideal Template

Every learner is compared with one model movement, even when the task permits functional variation.

Repair: distinguish invariant task requirements from variable personal solutions. Correct what constrains outcome, safety or adaptability—not harmless stylistic difference.

16. Failure Mode: Practice Looks Excellent Only With the Display On

A visual trace guides every repetition. The trace is removed and performance collapses.

Repair: insert no-feedback trials, ask for error estimation, delay the display, vary conditions and test retention. Improvement under augmented feedback is not enough to establish learning.

17. A Practical KP Protocol

Define the task outcome and the movement feature that plausibly controls it.
Let the learner attempt before correcting.
Ask what they felt and what they think happened.
Give one or a few high-value performance cues.
Connect the cue to an external consequence where possible.
Let the learner make the next adjustment.
Repeat enough to test the hypothesis, not simply to obey the cue.
Reduce, delay or remove external KP as error detection improves.
Test retention without augmented feedback.
Test transfer under a changed condition.

18. Evidence and Limits

The distinction between KR and KP is established in motor-learning literature. A systematic review of augmented feedback summarises feedback studies across healthy and clinical populations. Earlier methodological work by Schmidt and Young on kinematic feedback helped formalise the study of movement-pattern information. A 2022 trial-sequential meta-analysis on feedback and motor-skill learning in physical education shows that feedback effects depend on design and evidence quality rather than one universal schedule.

Limits matter. Motor tasks differ. Novices and experts differ. Rehabilitation findings do not automatically generalise to healthy school learners. Practice performance can diverge from retention. KP can also change attentional focus as well as information content. A responsible application therefore treats KP as an information design problem, not a magic phrase.

19. Missing-Node Scan

The missing node may be knowledge of performance when the learner sees the outcome but cannot identify the execution feature producing it; when video replay produces observation without useful interpretation; when sensor data create more metrics than actionable cues; when a coach corrects every movement before the learner estimates the error; when practice improves only while external displays remain visible; or when one ideal technique is enforced despite several functionally successful solutions.

20. The Return Path

Return to the missed shot.

“Twenty centimetres left” identifies the result. “The release opened late” identifies one candidate performance feature. The learning begins when the learner can connect that external observation to what the movement felt like, change the next attempt and later detect the same error without waiting for someone else to say it.

The best knowledge of performance eventually becomes knowledge the performer can generate from the performance itself.

Research and Further Reading

eduKateSG Learning Node Series · 0269 · Previous: 0268 — How Knowledge of Results Works.