From Correct Answer to Transfer

For digitally delivered instructional feedback, the practical problem is not installing a faster response; it is choosing a feedback configuration that serves the learning outcome and survives the conditions in which learners actually study. This chapter turns the evidence into a bounded design method: start with the work the learner must do, select a message that supports the next action, choose timing around the opportunity to process it, and evaluate later performance or transfer. The method is deliberately conditional because a controlled advantage in one retrieval task does not automatically travel to every classroom.

Start with the outcome, not the timestamp

The first design decision is to name what “better learning” means. Immediate accuracy, delayed retention, application, and transfer are related outcomes, but they test different capabilities. If a learner is practicing the retrieval of factual information, the relevant question may be whether they can produce the answer later without seeing it. If they are learning a procedure, the question may be whether they can select and execute the steps on a new problem. If they are writing an argument, the important change may be in organization, evidence, or reasoning rather than in the correctness of one sentence.

The measurement must match that purpose. A score taken immediately after a correction can show that the learner can respond while the answer is still salient. A one-week or one-month test asks whether something remained available after time passed. A changed problem or unfamiliar passage asks more about transfer than a repeated item. The retrieval-practice experiment is useful partly because it did not stop at the immediate test; it compared performance immediately, after one week, and after one month. The immediate-feedback group remained ahead at the later tests in that configuration, but the design still measured retention of one prose passage through one short-answer format. A different learning goal would require a different later test.

This choice changes what feedback should contain. If the goal is immediate accuracy and the learner already has the relevant method, a confirmation may be enough to stabilize a retrieval. If the goal is delayed retention, the learner must encounter information in a way that supports later recall rather than only recognition. If the goal is application, a message should connect the current response to the principle that determines the next case. If the goal is transfer, the learner needs an opportunity to use that principle under changed conditions. Feedback cannot create an outcome that the assessment never measures.

Process-focused guidance is particularly relevant when a task must travel beyond the current item. The meta-analysis observed a large effect for process-focused feedback, Hedges’ g = 1.25, but that estimate came from only three interventions. Its value is therefore directional: process guidance deserves consideration when the learner needs a reusable method, yet the number is too unstable to justify a universal prescription. The same evidence reported a task-focused effect of g = 0.53 across 67 interventions. A local correction may be the right design when the local task is the intended target. The question is not whether process focus is always superior; it is whether the message addresses the level at which future performance will be judged.

A concrete contrast shows the choice. Imagine an online language exercise in which a learner retrieves a grammatical form. A quick correction may be useful if the immediate aim is to strengthen the correct form.