Executive Function
Transitive Inference Task
Also called: transitive inference test, TI task, relational inference task, premise pair task
Measure relational inference: participants learn which shape wins in five overlapping pairs, then choose between new pairs that only the hidden order can answer.
9-12 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 97
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures whether people combine learned relations into an order. Participants learn which of two shapes is correct in five overlapping pairs, then choose between new pairings with no feedback. The main score is accuracy on the new pairs that only the hidden order can answer, with accuracy on the trained pairs and awareness of the order next.
Constructs
- Relational memory
- Transitive inference
- Reinforcement learning
- Awareness of task structure
Research uses
- Research on relational memory and the hippocampus
- Schizophrenia research at the group level
- Within-study comparison of learned pairs and inferred pairs
How the paradigm works
Transitive inference is concluding that if A beats B and B beats C, then A beats C. In the laboratory task, people learn by trial and error which of two shapes is correct in pairs that overlap, without being told that the pairs form one order, then choose between pairs they never saw (Bryant and Trabasso, 1971; Smith and Squire, 2005). Six shapes hold a hidden order, A over B over C over D over E over F. Training shows only the five neighboring pairs, AB, BC, CD, DE and EF, and in each the shape higher in the order is always correct. A is always correct and F never, but B, C, D and E are each correct in one pair and wrong in another, so the new pairs BD, CE and BE are answered from the order. Smith and Squire found that people who became aware of the order chose B over D nearly always while people who did not stayed near chance; others found inference without awareness (Greene and colleagues, 2001), or only on the pair two places apart, B over E (Frank and colleagues, 2005), and Ellenbogen and colleagues found that inference grew over a night's sleep.
More about the paradigm
Each trial shows two shapes side by side until the participant chooses one by clicking or tapping, or with the F and J keys, with no time limit; after 4 seconds a line asks for an answer. In training, Correct or Incorrect shows for 1 second with the shapes still on screen. Training builds up in stages of 40 trials (Greene and colleagues, 2001 and 2006): each pair 4 times in a row, then 2 times in a row, then once each in a fixed order, then in a mixed order, with no pair sharing a shape with the next; the settings add up to two more mixed stages. The test shows the five trained pairs, the three inference pairs and the end pair AF, 6 times each (4 or 8 in the settings), half with each shape on the left, 54 trials with no feedback. At the end the participant says whether the shapes seemed to follow one order and puts the six in order. Practice comes first: 4 trials with two practice shapes, one always correct. The shapes are smooth outlines drawn by code for this task in place of the sources' Japanese characters and patterns, all with the same area and color, and which shape takes which place comes from the session's seed. About 9 to 12 minutes.
The main score is inference accuracy: the share of choices of the higher shape in the three inference pairs. Accuracy on the trained pairs and on the inner trained pairs BC, CD and DE that the inferences rest on, the difference between inference and inner trained pairs, accuracy on BD and CE (one place skipped) and on BE (two places skipped), the end pair AF, training accuracy, whether an order was reported, how far the stated order agrees with the hidden one, and awareness of the order come next. A session with under 75% right on the inner trained pairs in the test is flagged for review, because its inferences rest on pairs it did not learn. Scores support research and do not diagnose.
Scores
- Inference Accuracy%
- Share of test choices of the shape higher in the order in the new pairs BD, CE and BE, which only the order can answer.
- Higher is better; 0.5 is chance
- Trained Pair Accuracy%
- Share of right choices in the five trained pairs during the test.
- Higher is better; 0.5 is chance
- Inner Trained Pair Accuracy%
- Share of right choices in the trained pairs BC, CD and DE during the test: the pairs the inferences rest on.
- Under 0.75 is flagged for review
- Inference Minus Inner Trained Pairs%
- Inference accuracy minus inner trained pair accuracy.
- Near 0 means the inferences were as accurate as the pairs they rest on
- Inference Accuracy, One Place Skipped%
- Share of right choices in BD and CE.
- Higher is better; 0.5 is chance
- Inference Accuracy, Two Places Skipped%
- Share of right choices in BE.
- Higher is better; 0.5 is chance
- End Pair Accuracy%
- Share of choices of A over F, a new pair that needs no inference because A was always correct and F never.
- Higher is better; 0.5 is chance
- Training Accuracy%
- Share of right choices across all training trials.
- Higher is better; 0.5 is chance
- Last Training Stage Accuracy%
- Share of right choices in the last 40 training trials, which mix the pairs.
- Higher is better
- Order Reported0/1
- 1 when the participant said the shapes seemed to follow one order; 0 for no or not sure.
- Informational
- Stated Order Agreement%
- Share of the 15 pairs of shapes that the order stated at the end puts the right way round.
- 1 is the hidden order; 0.5 is chance
- Aware of the Order0/1
- 1 when an order was reported and the stated order has at most one pair of shapes the wrong way round.
- Informational; read inference accuracy with it
- Median Test Choice Timems
- Median time from the shapes' onset to the choice in the test.
- Informational; the task has no time limit
- Choices After the Remindercount
- Choices made after the 4 second reminder, in training and in the test.
- Informational
- Anticipationscount
- Presses within 100 ms of the shapes' onset. They are not taken as choices.
- Quality check
- Test Choices Madecount
- Test choices in the scored run.
- Fewer than planned is flagged for review
- Planned Test Trialscount
- Test trials in the session: each of the 9 test pairs 6 times at the standard setting.
- Informational
Use and evidence
97 papers from 2021 to 2025 name this task in their title or abstract, up 21% compared with 2016 to 2020. That ranks it 190 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 22 |
| 2024 | 20 |
| 2023 | 22 |
| 2022 | 14 |
| 2021 | 19 |
| 2020 | 12 |
| 2019 | 18 |
| 2018 | 31 |
| 2017 | 10 |
| 2016 | 9 |
| 2015 | 20 |
| 2014 | 24 |
| 2013 | 22 |
| 2012 | 23 |
| 2011 | 15 |
| 2010 | 21 |
| 2009 | 20 |
| 2008 | 15 |
| 2007 | 12 |
| 2006 | 20 |
| 2005 | 10 |
| 2004 | 14 |
| 2003 | 9 |
| 2002 | 12 |
| 2001 | 9 |
| 2000 | 5 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 15
- Registered clinical studies using it as an outcome
- 0
Note on the counts: includes animal studies.
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 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Shapes
- 6 smooth shapes drawn by code for this task, with the same area and color, and 2 more for the practice. Which shape takes which place in the hidden order (A to F) is drawn from the session's seed.
- Trained pairs
- AB, BC, CD, DE and EF, the five neighboring pairs of the order. The shape higher in the order is always correct.
- Training
- Stages of 40 trials, each pair 8 times per stage and 4 with each shape on the left: each pair 4 times in a row, then 2 times in a row, then once each in a fixed order, then in a mixed order, 1 mixed stage (standard; 1 to 3), 160 trials at the standard setting. One pair never shares a shape with the next.
- Trial
- The shapes stay until a choice by click, tap, or the F and J keys, with no time limit; after 4,000 ms a line asks for an answer. In training, Correct or Incorrect shows for 1,000 ms with the shapes still shown, then a 500 ms blank.
- Test
- The 5 trained pairs, the inference pairs BD, CE and BE, and the end pair AF, 6 times each (standard; 4 or 8) and half with each shape on the left, 54 trials in a seeded order with no pair twice in a row and no feedback.
- End questions
- Whether the shapes seemed to follow one order (yes, no or not sure), then the six shapes put in order, one choice at a time.
- Practice
- 4 trials with two practice shapes, one always correct.
The design follows the six-shape order of Smith and Squire (2005) and Frank and colleagues (2005), with the staged training of Greene and colleagues (2001, 2006) in place of training to a criterion, so every session gets the same number of trials. The shapes are our own, so accuracies are not directly comparable with studies that used Japanese characters or patterns. Whether people infer the order may depend on noticing it, so read inference accuracy with the awareness measures. 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: transitive_inference_test from the scored run, a choice between two shapes with no time limit and no feedback. A choice is right when it picks the shape higher in the hidden order (chance 50%). Accuracy below chance invalidates and accuracy not above chance is flagged. One side on 90% of choices invalidates when accuracy is not above chance; the higher shape is on each side equally often. Answers under 200 ms on at least half the choices invalidate. Under 75% right on the inner trained pairs (BC, CD and DE) and a run short of its planned choices are flagged for review and withhold the percentile.
Why these rules
Accuracy significantly below chance (one-sided exact test, p < .05) cannot come from guessing; it means the response mapping was reversed or the task was not followed (Slick, Sherman and Iverson 1999 treat below-chance forced-choice performance as definite response bias). Chance-level accuracy can be a real result in impaired participants, so it is only flagged. One option on at least 90% of answered trials, when the correct options are balanced and accuracy is not above chance, means the answers did not depend on the items (the task counterpart of long-string responding in surveys; Curran 2016). 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. An inference pair can only be answered from the trained pairs it rests on, so studies of this task train to a criterion on them first (Ellenbogen and colleagues, 2007, required more than 75% on the inner pairs; Smith and Squire, 2005, 90% on every pair); this test trains for a fixed number of trials and flags sessions in which the inner pairs were not learned.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Mixed training stages | 1 | 1 to 3 |
| Test trials per pair | 6 (standard) | 4, 6 (standard), 8 |
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.
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Transitive Inference Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Bryant & Trabasso, 1971; Smith & Squire, 2005), with the standard settings (mixed training stages: 1; test trials per pair: 6). The main score was inference accuracy (share of test choices of the shape higher in the order in the new pairs BD, CE and BE, which only the order can answer).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Bryant, P. E., & Trabasso, T. (1971). Transitive inferences and memory in young children. Nature, 232(5311), 456-458. https://doi.org/10.1038/232456a0Transitive inference from trained neighboring pairs.
- Greene, A. J., Spellman, B. A., Dusek, J. A., Eichenbaum, H. B., & Levy, W. B. (2001). Relational learning with and without awareness: Transitive inference using nonverbal stimuli in humans. Memory & Cognition, 29(6), 893-902. https://doi.org/10.3758/BF03196418Training on overlapping pairs without telling people about the order, and inference without awareness.
- Smith, C., & Squire, L. R. (2005). Declarative memory, awareness, and transitive inference. Journal of Neuroscience, 25(44), 10138-10146. https://doi.org/10.1523/JNEUROSCI.2731-05.2005The six-shape order, the test of BD, CE, BE and AF, and the questions about awareness.
- Frank, M. J., Rudy, J. W., Levy, W. B., & O'Reilly, R. C. (2005). When logic fails: Implicit transitive inference in humans. Memory & Cognition, 33(4), 742-750. https://doi.org/10.3758/BF03195340BD, BE and CE in people kept unaware of the order.
- Greene, A. J., Gross, W. L., Elsinger, C. L., & Rao, S. M. (2006). An fMRI analysis of the human hippocampus: Inference, context, and task awareness. Journal of Cognitive Neuroscience, 18(7), 1156-1173. https://doi.org/10.1162/jocn.2006.18.7.1156The four training stages of 40 trials, from runs of one pair to mixed pairs.
- Ellenbogen, J. M., Hu, P. T., Payne, J. D., Titone, D., & Walker, M. P. (2007). Human relational memory requires time and sleep. Proceedings of the National Academy of Sciences, 104(18), 7723-7728. https://doi.org/10.1073/pnas.0700094104Six shapes in random-order training, inferences one and two places apart, and the role of time and sleep.
- Titone, D., Ditman, T., Holzman, P. S., Eichenbaum, H., & Levy, D. L. (2004). Transitive inference in schizophrenia: Impairments in relational memory organization. Schizophrenia Research, 68(2-3), 235-247. https://doi.org/10.1016/S0920-9964(03)00152-XThe task in schizophrenia research.
Source papers
- Smith C, Squire LR (2005). Declarative memory, awareness, and transitive inference. Journal of Neuroscience, 25(44):10138-10146.N 19, Young adults
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Transitive Inference Task take?
About 9 to 12 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?
Inference Accuracy, Trained Pair Accuracy, Inner Trained Pair Accuracy, Inference Minus Inner Trained Pairs, Inference Accuracy, One Place Skipped, Inference Accuracy, Two Places Skipped, End Pair Accuracy, Training Accuracy, Last Training Stage Accuracy, Order Reported, Stated Order Agreement, Aware of the Order, Median Test Choice Time, Choices After the Reminder, Anticipations, Test Choices Made and Planned Test Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Transitive Inference Task used in research?
97 papers from 2021 to 2025 name it in their title or abstract, up 21% compared with 2016 to 2020. That ranks it 190 of 395 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.14.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Transitive Inference Task have?
In the published research, 0 papers report its test-retest reliability and 0 registered clinical studies use it as an outcome (evidence grade D in the ConductCognition test census).
Which settings can a researcher change?
Mixed training stages (1 to 3) and test trials per pair (4, 6 (standard), 8). 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. 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.