Which Problem Do We Choose After a Wrong Answer?

You have answered a calculation problem incorrectly. The screen offers a choice: work on the same concept again or move to another concept. From the outside, that choice merely changes the order of questions. For the learner, it follows a particular failure. Someone else following the identical sequence may not have the same learning need. Evaluating a study schedule therefore requires separating the sequence itself from the process through which it was chosen.

Research published by Paulo F. Carvalho and colleagues in 2016 examined this distinction within university students' actual coursework. Students practiced calculating the mean, median, and mode in an online assignment and later answered relevant questions in their regular course exam. The study does not prove that repeating one concept is always the best approach. It asks a narrower question: can a learner's chosen repetition pattern relate to learning differently from an identical pattern assigned to somebody else? [1]

Three concepts, two decisions

Studying examples of the same concept consecutively constitutes blocking; moving among examples of different concepts constitutes interleaving. The relevant distinction concerns the concepts represented. Repeating a concept need not mean repeating a question containing identical numbers. Similarly, changing a problem's story need not involve changing its concept.

The tutorial separated these decisions. A self-regulating student could select whether the next question concerned the mean, median, or mode. They could also choose to retain the story or switch to another one. The same story could support questions about different concepts, while modified data could support further work on the same concept. Conceptual repetition and surface similarity were therefore not treated as interchangeable variables. [1]

This distinction makes behavior more interpretable. One student might repeatedly choose the same concept while working with datasets in different contexts. Another might retain a story to compare three calculations. Labeling both behaviors simply as repetition conceals which similarities the learner can use. The researchers accordingly tracked concept repetition and story repetition separately rather than collapsing them into one measure of sameness.

Feedback within a bounded assignment

Students first answered four pretest questions without feedback. They then reviewed calculation rules and began the tutorial problems. The first problem concerned the mean; subsequent steps depended on the student's group. Each response received correctness feedback. An option to finish appeared after at least five problems had been completed for each concept. This threshold did not require five correct answers. Students could continue beyond the minimum if they wished. [1]

The interface also supported comparison by displaying previous answers and making changes to datasets visible. Calculator use was encouraged. Consequently, this was not a test of unaided mental arithmetic speed or learning without assistance. The behavior of interest was arranging subsequent practice within a particular instructional interface. Its supports form part of the setting in which students made their decisions and should not disappear from an explanation of the study.

Following another learner's decisions

In the comparison condition, a student followed a sequence chosen by someone who had already completed the tutorial. The number and ordering of problem types were thus matched. Pretest scores were used in pairing students as well. Equal starting scores, however, do not establish identical strengths and weaknesses. The conspicuous difference was that the second student could not choose the next concept or story after seeing feedback on their own response. [1]

The value of this arrangement is its attempt to separate sequence from control. Two students can encounter the same problem order without being assumed to make the same errors. A repetition selected in response to the first learner's difficulty may not fit the second learner's immediate need. Establishing why this matters for learning remains a separate issue. A yoked design helps test particular explanations; it does not equalize every aspect of two learners' mental activity.

Errors were followed by more repetition

Students choosing their own sequences generally preferred to remain with the same concept. The paper reports an average concept repetition rate of 64 percent. Within the eligible subset used for comparing response outcomes, repetition followed incorrect answers on 72 percent of transitions and correct answers on 66 percent. That conditional analysis included relevant students who produced both kinds of answer, so it is not the same calculation as the overall rate. [1]

The pattern suggests that repetition need not be only an unthinking habit. Students may remain with a concept after discovering a difficulty. But sensitivity to feedback does not establish that their decisions are optimal. Repeating after an error and learning better because of that repetition are different claims. Observing the first in a behavioral record does not automatically supply causal evidence for the second.

A small, conditional connection with exam performance

The researchers analyzed subsequent exam performance alongside pretest scores, tutorial accuracy, problem count, and repetition measures. In the self-regulating group, greater concept repetition was associated with higher subsequent scores. The corresponding relationship was not statistically significant among students following someone else's sequence. This does not show that repetition harmed the latter group. A comparable positive relationship with achievement was not established for repeating stories either. [1]

The authors emphasize that the interaction between group and concept repetition was very small. They also explicitly characterize the analyses as correlational. Because students were not randomly assigned to particular repetition rates, the study cannot establish that greater repetition caused higher scores. Accounting for pretest performance is valuable, but does not remove every unmeasured difference in motivation or engagement. Moving every learner to blocked study would go well beyond these findings.

Many students, a limited outcome measure

The paper reports 2,061 students across eighteen course sections. They were not all assigned randomly to two groups during the same period. Everyone in the first collection period chose their own sequence, while random assignment to the two conditions occurred in the later period. Studying actual coursework adds relevance, but the outcome included only four exam questions targeting this material. The exam occurred at least two weeks after the assignment became available, not exactly two weeks after every individual's completion. [1]

The source also contains reporting inconsistencies. Its stated group sizes do not sum to the reported total, and the exact significance value for the interaction differs between the results and discussion. These discrepancies have not been reconciled here through raw-data analysis, and no explanation is invented for them. Interpretation consequently rests on the modest, correlational pattern the authors describe rather than a precise numerical claim of superiority. Four outcome questions also do not provide a detailed diagnostic map of each student's conceptual knowledge.

The large sample and narrow assessment solve different problems. More participants can help reveal a small group-level association in a noisy classroom setting. They do not turn four answers from one person into an extensive test of that person's understanding. Likewise, a result obtained with central-tendency calculations is not automatically a result about every school subject. Those boundaries are part of the evidence, not obstacles to be removed by an enthusiastic summary.

Whose sequence is it?

The study's useful contribution is the possibility that an apparently identical schedule does not perform the same function for two learners. Choosing the next question may involve evaluating an error and deciding what needs another look. Attention, engagement, and perceived learning needs are possible explanations; they were not isolated as independently demonstrated mechanisms here. The defensible conclusion is narrower than declaring one study order the winner: evaluating a learning schedule requires examining both the arrangement of examples and the learner's control over that arrangement.

Sources

[1] Paulo F. Carvalho, David W. Braithwaite, Joshua R. de Leeuw, Benjamin A. Motz, and Robert L. Goldstone (2016), An In Vivo Study of Self-Regulated Study Sequencing in Introductory Psychology Courses, PLOS ONE 11(3): e0152115. Open-access primary research.

Source access and factual review: 15 October 2026. The main methods, results, and discussion were directly read; raw data and OSF materials were not independently reanalyzed. Original explanation based on the Creative Commons Attribution-licensed research, attributed to Carvalho and colleagues. Internal discrepancies in sample counts and exact significance values are acknowledged rather than silently resolved.