Reasoning
Self-Explanation Task
Also called: self-explanation effect task, self-explanation prompts, prompted self-explanation, menu-based self-explanation, explaining worked examples
Measure learning from explaining worked examples: participants study chance problems step by step, choose the rule behind each step or simply go on, then solve similar and new problems.
12-20 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Try it here first, with no account.
- Papers, 2021 to 2025
- 244
- Evidence grade
- C
- Registered clinical studies
- 0
What it measures
Measures learning from explaining worked examples. Participants learn to solve chance problems, such as drawing marbles from a bag, from 6 worked examples shown one step at a time. After each step they choose from a menu the rule the step uses, or simply go on, as the study sets. Then they solve 6 test problems: 3 like the examples and 3 new ones. The main score is the percent of test problems answered right; the percent of steps explained right comes with it.
Constructs
- Self-explanation
- Learning from examples
- Problem-solving transfer
- Probability reasoning
Research uses
- Instructional design and learning research
- Metacognition and study strategy research
- Mathematics and statistics education research
- Within-study group comparison
How the paradigm works
The self-explanation task measures learning from explaining worked examples to oneself. Chi and colleagues (1989) found that students who explained to themselves the steps of worked examples, relating them to the principles in their text, understood the principles better and relied less on the examples when solving problems. Chi and colleagues (1994) then showed that merely prompting students to explain improved learning. Aleven and Koedinger (2002) found that students who explained their solution steps in a computer tutor learned with greater understanding and did better on transfer problems. In the tutor's earlier study (Aleven, Koedinger and Cross, 1999), students explained each step by choosing the theorem or definition behind it from the tutor's glossary. Atkinson, Renkl and Merrill (2003) prompted learners to identify the principle behind each step of a worked example.
More about the paradigm
Here, a screen first teaches five rules of chance. Then six worked examples, two of each of three kinds of chance problem from the Worked Examples Task, the kinds in an order drawn from the session's seed, are shown one step at a time. With the menu (the standard), after each step the participant chooses from a menu of the five rules the one the step uses and sees whether the choice was right, with the right rule named; with no prompts (a setting), the participant reads each step and goes on. Nothing is typed while studying, so no free text is scored. One practice problem with feedback shows the answer box. The test has six problems with no time limit and no feedback: three like the examples, with new stories and numbers, and three new ones that join two kinds. Answers are typed as percentages. The problems and steps are written for these tasks in each of the eight languages.
The main score is the percent of the six test problems answered right. The percent over the three similar and the three new problems, the percent of steps explained right with the menu, the median time per test problem and the time spent studying the examples come with it. Scores support research and do not diagnose.
Scores
- Test Problems Right%
- Percent of the six test problems answered with the right percentage; blank answers count as wrong.
- Higher means more test problems solved
- Similar Problems Right%
- Percent of the three test problems like the worked examples answered right.
- Transfer to problems of a kind already studied
- New Problems Right%
- Percent of the three new test problems, each joining two kinds, answered right.
- Transfer to problems no example showed
- Steps Explained Right%
- With the menu, percent of the 18 worked steps whose rule was chosen right; empty without prompts.
- Higher means the rules behind the steps were recognized
- Median Time per Test Problemms
- Median time from a test problem's appearance to its answer, over answered problems, leaving out answers under 100 ms.
- Describes pace; the test has no time limit
- Time Studying the Examplesms
- Total time on the worked examples' step screens, each to the press of its button.
- Informational
- Steps Explainedcount
- Steps explained with the menu: 18 in a complete session with the menu, 0 without prompts.
- Informational
- Test Problems Answeredcount
- Test problems submitted with an answer.
- A session with fewer than 3 is invalid
- Blank Test Answerscount
- Test problems submitted with an empty box.
- Lower means more problems were attempted
- Answers Under 100 mscount
- Test answers submitted within 100 ms of the problem appearing.
- Expected to be 0
Use and evidence
244 papers from 2021 to 2025 name this task in their title or abstract, down 8% compared with 2016 to 2020. That ranks it 133 of 670 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 72 |
| 2024 | 46 |
| 2023 | 37 |
| 2022 | 47 |
| 2021 | 42 |
| 2020 | 65 |
| 2019 | 50 |
| 2018 | 51 |
| 2017 | 48 |
| 2016 | 50 |
| 2015 | 45 |
| 2014 | 40 |
| 2013 | 50 |
| 2012 | 42 |
| 2011 | 45 |
| 2010 | 37 |
| 2009 | 40 |
| 2008 | 28 |
| 2007 | 32 |
| 2006 | 27 |
| 2005 | 19 |
| 2004 | 17 |
| 2003 | 15 |
| 2002 | 22 |
| 2001 | 10 |
| 2000 | 8 |
Use grade C
100 to 299 papers from 2021 to 2025.
Evidence grade C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 1
- Papers on reliability or validity
- 37
- Registered clinical studies using it as an outcome
- 0
Measured on ConductCognition
Pooled across the studies that feed our platform norms, by device group. A figure needs 50 or more people.
No pooled figure yet.
How the grades work
Use: A is 1,000 or more papers from 2021 to 2025, B 300 to 999, C 100 to 299, D 30 to 99 and E fewer than 30. Evidence: A means at least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome; B, at least 8 test-retest papers, and either 5 registered clinical studies or 150 reliability or validity papers; C, at least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies; D, little published measurement evidence found. The counts measure use. They include every paper that names the task in its title or abstract. Grades describe the published research on this task, not ConductCognition's version of it.
Source: ConductCognition test census 0.48.0, counted 12 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Problems
- The Worked Examples Task's chance problems: 3 kinds, each solved in 3 steps with 5 rules of chance. Every answer is a whole percentage, and no two answers are the same.
- Examples
- A screen of the 5 rules, then 6 worked examples, 2 of each kind, the kinds in an order drawn from the session's seed, each shown one step at a time.
- Explaining
- With the menu (standard), after each step a menu of the 5 rules: the choice is marked, the line under it says whether it was right or names the right rule, and the step keeps its rule's name. With no prompts, a Next step button. Nothing is typed while studying, so no free text is scored.
- Practice
- 1 problem with feedback shows the answer box before the test.
- Test
- 6 problems with no time limit and no feedback: 3 like the examples, with new stories and numbers, then 3 new ones that join two kinds, each group in an order drawn from the seed.
- Answer
- A typed percentage, in a box that opens the number keypad on phones, with a % sign beside it. No calculator.
- Languages
- Each problem, step and rule is written in all eight languages, with its numbers passed in so a translation cannot drop one.
Give each group of a study one condition and compare the groups, as Chi and colleagues (1994) compared students prompted to explain with students who were not; one person's score also depends on what they knew before. The scores support research and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: self_explanation_test from the scored run, one typed percentage for each of the 6 test problems. Typed answers have no chance level, so accuracy is not judged against chance, and test problems have no time limit, so none can be missed. Fewer than 3 answered invalidates; a run short of its 6 problems is flagged. Answers under 200 ms on at least half of 4 or more answered problems invalidate. One identical answer on 4 or more problems is flagged for review. With the menu, one rule chosen for 12 or more of the 18 steps is flagged for review. A hidden tab is flagged.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. A score over fewer than half the test problems cannot be compared with a score over all of them, so it invalidates. A low score is the result the test measures and never invalidates. The 6 answers all differ, so one answer typed for most of the problems cannot follow them; it is flagged. An answer under 200 ms is faster than a simple visual reaction, so it cannot reflect the item. Following response-time effort screening for untimed tests (Wise and Kong 2005), the session is invalid only when at least half the answers are that fast. The right rules are spread over the steps, no rule right for more than 6 of the 18, so one rule chosen for most steps cannot follow them; it is flagged. Problems stay on screen until they are answered, so a hidden tab does not change what was shown, but it can mean a calculator or help was used; it is flagged.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Explanation prompts | Menu of rules (standard) | Menu of rules (standard), No prompts |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
Need a setting this test does not have?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Languages: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese.
- Available on: Every plan, including the free plan.
- No percentile. Scores and quality flags only. How norms work
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Self-Explanation Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Chi MTH, Bassok M, Lewis MW, Reimann P, Glaser R, 1989; Chi MTH, De Leeuw N, Chiu MH, LaVancher C, 1994; Aleven V, Koedinger KR, 2002), with the standard settings (explanation prompts: Menu of rules). The main score was test problems right (percent of the six test problems answered with the right percentage; blank answers count as wrong).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Chi, M. T. H., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). Self-explanations: How students study and use examples in learning to solve problems. Cognitive Science, 13(2), 145-182. https://doi.org/10.1207/s15516709cog1302_1
- Chi, M. T. H., De Leeuw, N., Chiu, M.-H., & LaVancher, C. (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18(3), 439-477. https://doi.org/10.1207/s15516709cog1803_3
- Aleven, V. A. W. M. M., & Koedinger, K. R. (2002). An effective metacognitive strategy: Learning by doing and explaining with a computer-based Cognitive Tutor. Cognitive Science, 26(2), 147-179. https://doi.org/10.1207/s15516709cog2602_1
- Aleven, V., Koedinger, K. R., & Cross, K. (1999). Tutoring answer explanation fosters learning with understanding. In S. P. Lajoie & M. Vivet (Eds.), Artificial Intelligence in Education (pp. 199-206). IOS Press. Explanations chosen from a glossary of rules.
- Atkinson, R. K., Renkl, A., & Merrill, M. M. (2003). Transitioning from studying examples to solving problems: Effects of self-explanation prompts and fading worked-out steps. Journal of Educational Psychology, 95(4), 774-783. https://doi.org/10.1037/0022-0663.95.4.774Prompts to name the principle behind each worked step.
Source papers
- Chi MTH, De Leeuw N, Chiu MH, LaVancher C (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18(3):439-477.Paradigm source, School students
Questions
How long does the Self-Explanation Task take?
About 12 to 20 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Test Problems Right, Similar Problems Right, New Problems Right, Steps Explained Right, Median Time per Test Problem, Time Studying the Examples, Steps Explained, Test Problems Answered, Blank Test Answers and Answers Under 100 ms. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Self-Explanation Task used in research?
244 papers from 2021 to 2025 name it in their title or abstract, down 8% compared with 2016 to 2020. That ranks it 133 of 670 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.48.0, counted 12 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Self-Explanation Task have?
In the published research, 1 paper reports its test-retest reliability and 0 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Explanation prompts (Menu of rules (standard), No prompts). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
Which languages can participants use?
Participants can take this test in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese, with its words in the same language. English is on every plan, and the other languages come with yearly plans. Have a native speaker read the machine-translated text before launch.
Is this a clinical test?
No. ConductCognition is for research use only. It runs, scores and exports the test; clinical interpretation is outside the platform.
Every test: the full test list.
Run this test in your next study
Start free with 10 free participants on any test, then use participant credits or a yearly plan.