A corrected mistake is not yet stable learning. Stability appears when the learner can perform the repaired move again, under changed conditions, after enough delay that the correction is no longer sitting in short-term memory.
Students, parents and tutors searching for mastery learning, revision strategies, learning progress and how to know when a topic is mastered need evidence that separates immediate correction from reliable performance.
This casebook sits beside How Learning Transfer Works and The Confirmation; it owns the learner-facing stability casebook, not the general transfer mechanism.
Stable enough is an evidence condition, not a feeling
Look for accurate performance on fresh work, reduced instructional support, survival after delay and acceptable reliability under the conditions that matter next.
1. Current evidence
For deciding whether a learning repair is stable, current evidence is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for current evidence. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects current evidence. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from current evidence. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
2. Learner goal
For deciding whether a learning repair is stable, learner goal is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for learner goal. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects learner goal. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from learner goal. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
3. Prerequisites
For deciding whether a learning repair is stable, prerequisites is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for prerequisites. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects prerequisites. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from prerequisites. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
4. Weak-link location
For deciding whether a learning repair is stable, weak-link location is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for weak-link location. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects weak-link location. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from weak-link location. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
5. Error frequency
For deciding whether a learning repair is stable, error frequency is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for error frequency. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects error frequency. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from error frequency. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
6. Error cost
For deciding whether a learning repair is stable, error cost is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for error cost. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects error cost. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from error cost. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
7. Task difficulty
For deciding whether a learning repair is stable, task difficulty is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for task difficulty. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects task difficulty. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from task difficulty. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
8. Support dose
For deciding whether a learning repair is stable, support dose is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for support dose. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects support dose. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from support dose. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
9. Prompt dependence
For deciding whether a learning repair is stable, prompt dependence is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for prompt dependence. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects prompt dependence. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from prompt dependence. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
10. Accuracy
For deciding whether a learning repair is stable, accuracy is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for accuracy. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects accuracy. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from accuracy. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
11. Explanation quality
For deciding whether a learning repair is stable, explanation quality is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for explanation quality. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects explanation quality. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from explanation quality. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
12. Method selection
For deciding whether a learning repair is stable, method selection is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for method selection. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects method selection. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from method selection. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
13. Representation
For deciding whether a learning repair is stable, representation is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for representation. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects representation. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from representation. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
14. Retrieval
For deciding whether a learning repair is stable, retrieval is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for retrieval. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects retrieval. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from retrieval. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
15. Application
For deciding whether a learning repair is stable, application is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for application. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects application. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from application. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
16. Transfer
For deciding whether a learning repair is stable, transfer is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for transfer. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects transfer. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from transfer. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
17. Delayed return
For deciding whether a learning repair is stable, delayed return is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for delayed return. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects delayed return. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from delayed return. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
18. Speed
For deciding whether a learning repair is stable, speed is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for speed. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects speed. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from speed. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
19. Stamina
For deciding whether a learning repair is stable, stamina is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for stamina. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects stamina. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from stamina. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
20. Checking
For deciding whether a learning repair is stable, checking is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for checking. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects checking. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from checking. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
21. Completion
For deciding whether a learning repair is stable, completion is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for completion. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects completion. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from completion. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
22. Confidence calibration
For deciding whether a learning repair is stable, confidence calibration is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for confidence calibration. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects confidence calibration. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from confidence calibration. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
23. Independent attempt
For deciding whether a learning repair is stable, independent attempt is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for independent attempt. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects independent attempt. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from independent attempt. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
24. Fresh task
For deciding whether a learning repair is stable, fresh task is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for fresh task. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects fresh task. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from fresh task. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
25. Mixed practice
For deciding whether a learning repair is stable, mixed practice is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for mixed practice. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects mixed practice. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from mixed practice. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
26. Cumulative review
For deciding whether a learning repair is stable, cumulative review is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for cumulative review. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects cumulative review. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from cumulative review. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
27. Old learning
For deciding whether a learning repair is stable, old learning is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for old learning. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects old learning. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from old learning. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
28. New learning
For deciding whether a learning repair is stable, new learning is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for new learning. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects new learning. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from new learning. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
29. Exam proximity
For deciding whether a learning repair is stable, exam proximity is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for exam proximity. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects exam proximity. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from exam proximity. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
30. Time available
For deciding whether a learning repair is stable, time available is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for time available. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects time available. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from time available. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
31. Study workload
For deciding whether a learning repair is stable, study workload is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for study workload. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects study workload. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from study workload. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
32. School workload
For deciding whether a learning repair is stable, school workload is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for school workload. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects school workload. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from school workload. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
33. Tuition workload
For deciding whether a learning repair is stable, tuition workload is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for tuition workload. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects tuition workload. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from tuition workload. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
34. Sleep and recovery
For deciding whether a learning repair is stable, sleep and recovery is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for sleep and recovery. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects sleep and recovery. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from sleep and recovery. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
35. Resource quality
For deciding whether a learning repair is stable, resource quality is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for resource quality. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects resource quality. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from resource quality. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
36. Feedback
For deciding whether a learning repair is stable, feedback is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for feedback. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects feedback. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from feedback. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
37. Teacher evidence
For deciding whether a learning repair is stable, teacher evidence is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for teacher evidence. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects teacher evidence. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from teacher evidence. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
38. Tutor evidence
For deciding whether a learning repair is stable, tutor evidence is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for tutor evidence. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects tutor evidence. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from tutor evidence. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
39. Learner self-evidence
For deciding whether a learning repair is stable, learner self-evidence is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for learner self-evidence. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects learner self-evidence. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from learner self-evidence. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
40. Parent observations
For deciding whether a learning repair is stable, parent observations is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for parent observations. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects parent observations. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from parent observations. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
41. Marks
For deciding whether a learning repair is stable, marks is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for marks. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects marks. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from marks. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
42. Work samples
For deciding whether a learning repair is stable, work samples is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for work samples. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects work samples. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from work samples. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
43. Revision history
For deciding whether a learning repair is stable, revision history is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for revision history. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects revision history. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from revision history. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
44. Decision log
For deciding whether a learning repair is stable, decision log is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for decision log. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects decision log. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from decision log. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
45. Priority ranking
For deciding whether a learning repair is stable, priority ranking is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for priority ranking. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects priority ranking. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from priority ranking. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
46. Stop rule
For deciding whether a learning repair is stable, stop rule is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for stop rule. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects stop rule. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from stop rule. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
47. Maintenance dose
For deciding whether a learning repair is stable, maintenance dose is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for maintenance dose. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects maintenance dose. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from maintenance dose. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
48. Re-entry trigger
For deciding whether a learning repair is stable, re-entry trigger is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for re-entry trigger. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects re-entry trigger. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from re-entry trigger. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
49. Handover
For deciding whether a learning repair is stable, handover is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for handover. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects handover. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from handover. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
50. Next review
For deciding whether a learning repair is stable, next review is useful only when it changes a real learning decision. The common mistake is equating one corrected example, one high mark or one fluent explanation with durable capability. That mistake often creates more activity—more worksheets, more tuition, more notes, more revision—but not necessarily a better next move. Start by stating the learner’s current task, the conditions under which the evidence was produced and the decision that must be made.
Build a small record for next review. It can be a one-line observation, a changed-task result, a support note, an error sample, a timing record or a before-and-after comparison. Do not collect evidence because portfolios look impressive. Collect it because it helps answer a question: continue this repair, reduce support, change the method, return to normal learning, revisit an old prerequisite, or seek specialist depth.
Worked learner case
Alicia has a weak result and wants to revise everything. Instead, she inspects next review. The evidence shows that one part of the task is stable while another still needs prompting. She protects the stable part and directs the next session at the fragile dependency. After the repair she attempts a changed problem after a delayed return. The new result matters more than repeating the original worksheet because it tests whether the learning travelled.
Beatrice has the opposite problem. Her marks improved, so everyone assumes the repair is finished. A delayed check reveals that the result depended on a familiar question pattern. The team does not panic or restart the whole course. They name the boundary precisely, add one changed practice set and review again. This keeps evidence proportional to the decision.
Decision protocol
Use four steps. First, describe the current state without labelling the learner. Second, identify the smallest uncertainty that blocks the decision. Third, choose a task that can reduce that uncertainty. Fourth, decide in advance what result would justify continue, change, reduce, stop or handoff. This prevents hindsight from turning every result into support for the plan that was already preferred.
Where several adults are involved, separate observation from interpretation. A school teacher may see classroom independence; a tutor may see performance under targeted practice; a parent may see homework initiation and fatigue; the learner may know which steps still feel effortful. These are different windows. Combine them carefully rather than allowing one perspective to erase the others.
Fresh-task and delayed evidence
Use a changed problem after a delayed return. Change enough surface detail that memory of the correction cannot carry the answer. Then return after delay where practical. Record support conditions explicitly. A correct response with a hint and a correct response without a hint are both useful, but they are not identical evidence. Legitimate access arrangements remain part of the intended condition; instructional rescue is what should be tracked and, where appropriate, faded.
Do not demand perfect performance before moving forward. Learning systems need thresholds. A capability can be stable enough for maintenance while still containing occasional error. The threshold depends on consequence, future dependency and assessment demands. High-risk prerequisites may deserve stronger evidence than low-cost details that can be refreshed later.
What not to infer
Do not infer a fixed ability, personality or future outcome from next review. Do not promise that a reduced prompt will automatically become a higher mark. Do not assume a high mark means every component is secure. Do not assume a low mark means every component is weak. Keep the conclusion task-specific and time-bounded. The purpose of the record is to guide the next learning move.
Parent, tutor and learner handoff
Stability evidence should travel with the learner when responsibility changes. A useful handoff says what has been attempted, what evidence changed the plan, what remains uncertain, what support is currently needed and when the next review occurs. It should be short enough to read. A handoff that requires another adult to reconstruct months of history from messages and worksheets is not yet a usable learning record.
Frequently asked questions
How many correct answers prove mastery?
There is no universal number. Use varied tasks, delay, support conditions and the consequence of error.
What if the learner makes an occasional mistake?
Stability does not require perfection. Ask whether the error rate and recovery are acceptable for the next demand.
Does speed matter?
Sometimes. Speed matters when the real task is timed or when slow retrieval overloads later steps.
Can a topic become unstable again?
Yes. Forgetting, interference and changed demands can expose fragility. Maintenance and cumulative review remain useful.
Routes and handoff
Use Learning Transfer, Capability Profile and the Learning Atlas.
The question is not whether the learner once succeeded. It is whether the success can be trusted enough for the next stage.
