Practice Is Not Repetition: Build a Loop That Changes Performance

Practice must change a defined performance

Producing another article is not automatically practice. Practice begins when a person names a capability precisely enough to observe, attempts a task that elicits it, receives information about the gap, changes the approach, and tries again. “Improve writing” is too broad; “state the decision answer in the first 120 words without losing an important caveat” creates a performance that can be reviewed. The unit of learning is the changed attempt, not the volume of content surrounding it.

How People Learn II synthesizes research on prior knowledge, feedback, context, and metacognition. It supports treating learning as an interaction between the learner and the situation rather than a simple accumulation of hours. That does not yield a universal practice recipe for editorial work. A useful loop must still define the local task, evidence of improvement, and conditions under which the skill should transfer.

Evidence: National Academies of Sciences, Engineering, and Medicine

The loop has five distinct jobs

A complete loop includes a target, an eliciting attempt, an observation, a correction, and a new attempt. The target focuses attention. The task makes the capability visible. Observation separates what happened from why it may have happened. Correction chooses one change rather than rewriting everything. The second attempt tests whether the change improves the same performance. A later transfer task asks whether the adjustment survives outside the rehearsal.

Skipping a job creates predictable confusion. Feedback without a retest is advice, not evidence of learning. A second attempt on an identical prompt may measure memory. A delayed task in a different topic or channel is harder but more informative about transfer. The Scrum Guide offers a primary example of short inspection and adaptation cycles around work artifacts; the article adapts that logic without claiming editorial skill practice is Scrum.

• Target one observable behavior.
• Elicit it with a representative task.
• Separate the trace from the interpretation.
• Change one controllable feature.
• Retest now and later in a different context.

Evidence: National Academies of Sciences, Engineering, and Medicine; Scrum Guides

Difficulty should expose error without destroying the signal

A task that is effortless produces little diagnostic information; one that overwhelms the learner may reveal only overload. Calibrate practice so the person can complete the task while making meaningful choices. Reduce complexity by shortening the source set, narrowing the audience, or practicing one section. Increase it by adding contradictory evidence, a tighter length constraint, or an unfamiliar reader situation. Keep the target stable while adjusting difficulty, otherwise improvement cannot be interpreted.

CMU’s active-learning guidance includes worked examples, peer review, polling, and self-assessment. Those techniques matter here because they make thinking observable rather than leaving the learner passive. Yet activity is not equivalent to difficulty calibration. The practice designer must still ask whether the exercise elicits the exact reasoning or communication move being developed and whether the learner can act on the resulting information.

Evidence: Carnegie Mellon University Eberly Center; Scrum Guides

Feedback needs direction, timing, and an action path

Useful feedback points to evidence in the attempt, relates it to a criterion, and suggests what to examine next. “Clearer” is too vague; “the claim arrives before the scope condition, so a scanning reader can misread it as universal” identifies a trace and a repairable relationship. Immediate feedback can prevent repeated misunderstanding, while delayed self-review can expose whether the learner recognizes the issue independently. Combining them may be more informative than declaring one timing universally best.

Feedback also has access and power dimensions. A learner should know the criterion before judgment, have a way to ask questions, and be able to demonstrate the skill in an appropriate mode. CAST’s Universal Design for Learning Guidelines support multiple means of engagement, representation, and action or expression. That does not mean every practice must use every mode; it means the chosen format should not introduce an irrelevant barrier that obscures the capability.

Evidence: Carnegie Mellon University Eberly Center; CAST

Artifact: the practice–feedback–transfer log

Use one row per cycle. Record date, capability target, task conditions, predicted challenge, observable trace, feedback source, selected correction, immediate retest, delayed transfer task, and next decision. Attach examples or links when appropriate, but minimize personal or sensitive information. Score only a small rubric with behavioral anchors. A log full of hours and outputs but empty correction fields is evidence that activity has not been converted into a learning loop.

Review the log every few cycles for patterns. If the same correction recurs, the task may be too difficult, the feedback may be unusable, or the criterion may not be understood. If performance improves only on identical prompts, vary topic, audience, time pressure, or tool access. If it transfers but at a high cognitive cost, continue until the behavior becomes more reliable or decide that external support is a better system design.

Evidence: National Academies of Sciences, Engineering, and Medicine; Carnegie Mellon University Eberly Center; CAST

Improvement claims need modest boundaries

One cleaner revision does not establish retained skill. Familiarity, extra time, reviewer prompting, or an easier topic can explain the difference. Use repeated observations across comparable tasks, preserve unsuccessful attempts, and state uncertainty. Some capabilities depend on team coordination or domain knowledge rather than individual technique; practicing sentence structure will not repair missing evidence governance or unrealistic workload.

Begin with a real task due this week and isolate a ten-to-twenty-minute rehearsal that precedes production. Complete an immediate correction and schedule a different transfer task several days later. This timing is illustrative, not a validated prescription or TenMultigure performance result. Revisit the method by 2027-02-10 or sooner if the learner’s role, accessibility needs, or target work changes.

Evidence: National Academies of Sciences, Engineering, and Medicine; CAST