How Concurrent Feedback Works | Guide the Action While It Is Happening Without Making the Display Part of the Skill

eduKateSG Learning Node Series · 0273

A line moves across a screen while the learner moves. A tone rises when force is too high. A wearable vibrates when posture drifts. The correction arrives before the action is finished.

This is concurrent feedback: augmented information delivered during performance rather than after it. It is attractive because the learner can correct in real time. The movement can become safer, smoother and more accurate immediately. In rehabilitation, sport, simulation and technical training, that can be extremely useful.

But concurrent feedback changes the task. A learner practising balance while following a moving cursor is not practising exactly the same task as balancing without the cursor. A trainee steering while a haptic device continuously nudges the hand is learning inside a coupled human–feedback system. If the display disappears and performance collapses, the practice system may have trained dependence as effectively as it trained skill.

Concurrent feedback works when real-time information improves the learner’s perception and correction of the action—then progressively becomes unnecessary.

The 50-Second Read

Concurrent feedback arrives during the movement.
It can be visual, auditory, haptic, verbal or multimodal.
It often improves immediate performance because error can be corrected before the action ends.
That immediate advantage is not automatically durable learning.
The feedback display can become part of the practised task.
Continuous guidance may reduce the need to interpret intrinsic sensory information.
Complex or safety-critical tasks may justify more concurrent support, especially early.
The useful question is not “concurrent or never concurrent?” but “what information should be available now, and how will it later be withdrawn?”
Retention should be tested without the same concurrent aid.
Transfer should test whether the learner can adapt when the environment, target or feedback channel changes.
Real-time data should show only information the learner can act on.
The final capability should live in the learner, not in the display.

Canonical Owner Boundary

This node owns augmented feedback delivered while the action is still unfolding. How Feedback Frequency Works owns how often feedback enters practice. How Knowledge of Performance Works owns information about how an action was executed. How Knowledge of Results Works owns outcome information. The next node, How Terminal Feedback Works, owns feedback delivered after the action is complete. This article asks one precise question: what changes when external information is available online, during performance itself?

1. Concurrent Feedback Changes the Control Loop

Without augmented feedback, a learner acts using intention, prior knowledge and intrinsic sensory information. During the movement, vision, proprioception, touch, sound and balance signals provide a partial estimate of what is happening. The learner compares that estimate with the desired state and adjusts.

Concurrent feedback inserts an additional measurement into that loop. A cursor shows position error. A tone represents speed. A vibration signals that the body has crossed a posture threshold. The learner can now regulate the movement using both intrinsic and augmented information.

This can improve control because the augmented channel may be more precise, faster or easier to interpret than the learner’s internal estimate. It can also change what the learner attends to and which signals the learner learns to trust.

2. Real-Time Correction Is Especially Attractive in Complex Tasks

In a simple laboratory task, the learner may be able to infer error from the outcome. In a complex task, several variables can change at once. Rowing technique, surgical hand movement, gait retraining, postural control and equipment operation can all involve timing, force, sequence and spatial coordination.

Concurrent feedback can reduce the search space. Instead of asking a novice to infer which of twelve variables caused the error, the system can highlight one relation as the movement unfolds.

The benefit is strongest when the real-time information maps cleanly to an actionable variable. A real-time display that says everything may be no better than no display because the learner cannot convert the data into one useful correction.

3. Immediate Performance and Learning Are Different Outcomes

Concurrent feedback can make training look successful. Errors shrink. Movements become smoother. The learner appears more competent.

The critical test comes when the feedback disappears. If the learner cannot reproduce the performance without the display, then some of the apparent competence belonged to the feedback system.

This is why motor-learning research distinguishes acquisition performance from retention and transfer. Immediate performance answers “how well did the learner perform under the training conditions?” Retention asks whether improvement persists after time. Transfer asks whether the capability survives a change in task conditions.

4. The Display Can Become Part of the Skill

Imagine learning to keep a force trace inside a green band. After enough practice, the learner may become extremely good at moving the trace rather than accurately sensing the force.

The practised skill has quietly become “control the body while watching the trace”. If the real performance context has no trace, the practice task and target task are no longer identical.

This does not make concurrent feedback bad. It means designers must plan a handoff from external information to intrinsic control. The handoff should be treated as part of the training architecture, not an optional final step.

5. Visual Feedback Is Powerful—and Visually Expensive

Screens, traces, mirrors and overlays are common because vision can represent position, trajectory, force and error with high precision. But the visual system may already be needed for the task itself.

If a tennis player must watch the ball, asking them to monitor a complex visual dashboard at the same time can compete with task-relevant vision. If a clinician must visually inspect tissue, a second visual stream may increase attentional load.

The design question is therefore not merely whether visual feedback is informative. It is whether the task can afford another visual demand.

6. Auditory Feedback Can Free the Eyes

Sound can encode timing, force, velocity or deviation without requiring the learner to look away. Movement sonification converts a changing movement variable into changing sound.

A 2026 systematic review of movement sonification highlights the growing use of auditory mappings in augmented-feedback research. Auditory feedback can be especially attractive when vision is already committed to the task, although mapping design remains critical: the learner must be able to understand what a pitch, rhythm or timbre change means.

Sound can also become distracting or cognitively expensive if the mapping is arbitrary. A useful sonification should become interpretable through practice rather than behave like a second puzzle layered onto the first.

7. Haptic Feedback Can Put Information Into the Body

Vibration, resistance, force guidance and tactile cues can communicate error without occupying vision or hearing. A wearable can vibrate on the side toward which posture is drifting. A robot can provide force information as the learner reaches.

Haptic feedback has a special advantage: the information arrives through a modality already closely linked to movement. It also has a special risk: guidance forces can physically move the learner into the correct path, reducing the need to generate the path independently.

A cue and an assist are not the same. A vibration that says “drifting left” preserves the learner’s responsibility to correct. A device that physically pushes the limb right may improve the movement while reducing error-generation and self-correction demands.

8. Multimodal Feedback Is Not Automatically Better

It is tempting to combine a visual trace, a sound and a vibration because each carries useful information. More channels can increase redundancy and accessibility. They can also overload attention or create conflicting cues.

Use multimodal feedback when each channel has a clear job or when redundancy protects performance under noisy conditions. Avoid duplicating the same low-value signal across three modalities merely because the technology can do it.

9. Error Bandwidth Can Reduce Unnecessary Real-Time Correction

Concurrent feedback does not have to be continuous. One design is to keep the display silent while performance remains inside an acceptable region and trigger a cue only when error crosses a threshold.

This approach preserves periods of self-regulation while still protecting against large drift. The threshold should reflect task demands. An error band that is too narrow turns almost every natural fluctuation into an alarm; one that is too wide allows meaningful error to persist.

10. Fade the Display, Not the Standard

A common training strategy is to begin with rich concurrent feedback and progressively remove it. The mistake is to lower the performance target at the same time.

The goal is the opposite: preserve the target while reducing external information. First show the full trace. Then show only threshold crossings. Then show intermittent windows. Then remove the display and ask the learner to predict what it would have shown.

This creates a bridge from externally guided control to internally regulated performance.

11. Ask for Prediction Before Reveal

Concurrent systems can accidentally prevent the learner from discovering whether their internal estimate is accurate because the answer is always visible.

One solution is intermittent masking. Hide the feedback briefly. Ask the learner to estimate position, force or timing. Then reveal the trace. The gap between prediction and measurement becomes a calibration signal.

Learning now includes both movement control and better self-measurement.

12. Safety Can Justify More Concurrent Support

In rehabilitation, novice technical training or hazardous environments, preventing a dangerous movement can matter more than creating desirable difficulty.

Concurrent feedback may therefore remain dense during safety-critical phases. The correct comparison is not always “more feedback versus better learning”. It may be “more feedback versus unacceptable risk”.

Safety requirements should be explicit so that the instructor knows which feedback can be faded and which must remain as an operational safeguard.

13. Cross-Domain Comparison: Lane-Keeping Assistance

A vehicle can warn or steer when it drifts from the lane. The system may make immediate driving safer. But a driver who learns only with continuous lane assistance may have fewer opportunities to detect small drift independently.

The analogy clarifies the difference between performance support and learning support. Some systems are designed to remain permanently available, in which case dependence may be acceptable. Training systems usually aim to create capability that survives removal of the support.

14. Cross-Domain Comparison: Live Code Linting

A programming environment can underline errors as the code is typed. This improves immediate correctness. It can also mean the programmer rarely practises detecting the same error without the tool.

The right training design depends on the final environment. If the tool will always be present, learning to work effectively with it is valid. If an examination or operational context removes it, unsupported practice becomes necessary.

15. The Tool-Permanent Question

Concurrent feedback is sometimes criticised because learners can become dependent on it. But dependence is not automatically a problem if the tool is part of the final performance system.

A pilot learns with instruments because instruments are part of flying. A surgeon may use navigation technology because the technology remains present in practice. A worker may use a wearable safety alarm because the alarm is designed to remain.

The real question is: Is the feedback a training scaffold to be removed, or an operational instrument to be mastered? Confusing those two jobs creates bad evaluation.

16. Failure Mode: The Learner Tracks the Display Instead of the Task

The cursor becomes the target. The learner stops using intrinsic cues.

Repair: reduce display density, periodically hide it, ask for self-estimation and test performance without it.

17. Failure Mode: The Feedback Arrives Too Fast to Interpret

A dashboard updates ten variables continuously. The learner sees movement but cannot convert it into action.

Repair: choose one high-value variable or encode only threshold crossings. Real-time does not mean all-data-all-the-time.

18. Failure Mode: Removal Happens Only at the Final Test

The learner practises entirely with rich concurrent guidance and encounters unsupported performance for the first time during assessment.

Repair: make withdrawal part of practice. Alternate supported and unsupported trials before high-stakes testing.

19. A Practical Concurrent-Feedback Protocol

Define the target skill without the feedback system.
Identify which error information is unavailable or unreliable intrinsically.
Select one modality that fits the task’s attentional demands.
Map one useful variable to one interpretable cue.
Begin with enough real-time information to support safe, meaningful correction.
Insert brief feedback-off windows.
Ask the learner to predict what the hidden display would show.
Compare prediction with measurement.
Fade density, frequency or precision as calibration improves.
Test delayed retention without the training display.
Test transfer under changed conditions.
If the tool will remain operationally, separately test human–tool performance as the real target system.

20. Evidence and Limits

A broad systematic review of augmented feedback in motor learning distinguishes concurrent from terminal feedback and documents substantial heterogeneity across tasks and populations. A 2024 systematic review in developmental coordination disorder notes the growing use of real-time visual, auditory and haptic technology but finds limited evidence that one feedback form is consistently superior. A 2024 systematic review and meta-analysis of real-time postural feedback in older adults found immediate balance benefits while highlighting uncertainty about longer-term learning and the distinction from terminal feedback.

The classic concurrent-versus-terminal question also varies with task complexity. In a complex rowing-type task, Sigrist and colleagues found self-controlled terminal visual feedback outperformed concurrent visual, auditory and haptic feedback under their experimental conditions. That result should not be turned into a universal rule; it demonstrates that real-time guidance can improve training performance while simultaneously changing what is learned.

21. Missing-Node Scan

The missing node may be concurrent-feedback design when a learner performs accurately only while a cursor is visible; when a visual dashboard competes with the visual demands of the real task; when a haptic device physically guides the solution so strongly that the learner never generates it; when every sensor variable is displayed despite only one being actionable; when no unsupported trials occur until the final assessment; or when the feedback system is actually intended to remain permanently but is being evaluated as though removal were the goal.

22. The Return Path

Return to the moving line on the screen.

At first, it may be exactly what the learner needs: a precise external signal that reveals an error too fast or subtle to feel. The important question is what happens next. If the learner gradually learns to anticipate the line, reproduce the action when the line is hidden and transfer the skill to a new condition, the concurrent display has done its job.

If the line disappears and the skill disappears with it, then the system trained the line.

Concurrent feedback should make the learner more perceptive, not make the feedback display indispensable.

Research and Further Reading

eduKateSG Learning Node Series · 0273 · Previous: 0272 — How Feedback Frequency Works.