Data

Test-retest evidence for 451 cognitive tests

A test given twice should rank people the same way twice. Test-retest papers check that. Here is how many each test has, set against how much it is used.

Test families with 25 or more test-retest papers: 37 of 495
Most used test families with fewer than 25 test-retest papers: 22 of 50; See the well-used tests
Test families with no test-retest paper found: 177 of 495

Use set against test-retest evidence

Each dot is one test family on this site. The higher a dot sits for its place across, the more test-retest work it has for how much it is used.

One dot per test family on this site. Both axes are logarithmic, and a count of zero sits on the bottom line. Dots high and to the left have more of this evidence for how much they are used. Hover a dot to name it; the lists below give every number.

Most test-retest papers

The tests whose reliability over time has been checked most often, in papers from any year.

1Verbal Fluency186 test-retest papers, 5,242 papers
2Trail Making Test A and Trail Making Test B161 test-retest papers, 3,944 papers
3Stroop Color-Word Test151 test-retest papers, 4,956 papers
4Digit Span Test136 test-retest papers, 2,755 papers
5Symbol Coding135 test-retest papers, 2,557 papers
6Verbal Learning134 test-retest papers, 2,503 papers
71-Back Task and 2-Back Task104 test-retest papers, 3,210 papers
8Matrix Reasoning89 test-retest papers, 1,106 papers
9Go/No-Go Task84 test-retest papers, 3,039 papers
10Picture Naming Test82 test-retest papers, 1,759 papers
11Contrast Sensitivity Test81 test-retest papers, 463 papers
12Implicit Association Test (IAT)68 test-retest papers, 1,497 papers
13Category Fluency56 test-retest papers, 2,471 papers
14Continuous Performance Test54 test-retest papers, 1,014 papers
15Short Story Recall52 test-retest papers, 1,038 papers
16Card Sorting Test (Berg)51 test-retest papers, 1,227 papers
17Vocabulary Test49 test-retest papers, 550 papers
18Delay Discounting and Food Delay Discounting48 test-retest papers, 2,088 papers
19Dichotic Listening Task48 test-retest papers, 280 papers
20Paced Visual Serial Addition Test45 test-retest papers, 397 papers

Well used, rarely retested

Tests with 300 or more papers from 2021 to 2025, fewest test-retest papers first: widely used, with little published work on whether they rank people the same way twice. Tests whose name has other uses are left out, since their paper counts run high.

Well-used tests with the fewest test-retest papers
Test	Papers, 2021 to 2025	Test-retest papers
Metaphor Interference Task	436	1
Self-Paced Reading Task	740	2
Ball-Toss Game	358	2
Evaluative Conditioning Task	315	2
Ultimatum Game	592	3
Visuomotor Adaptation Task	405	3
Grammaticality Judgment Task	1,140	4
Judgments of Learning Task	379	4
Attentional Blink Task	332	4
Gaze Cueing Task	304	4
Biological Motion Task	624	5
Serial Naming Task	415	6
Social Value Orientation	348	7
Perceptual Confidence Task	306	7
Novel Word Learning Task	304	7

Strong effects, weaker rankings

A task can show the same effect in nearly every group and still rank individuals poorly, because a difference score such as the Stroop cost subtracts away much of what is stable about each person. Hedge, Powell and Sumner retested seven classic tasks three weeks apart. Values are intraclass correlations, with 95% confidence intervals; 1 means people kept the same order.

Test-retest reliability of classic tasks, from Hedge, Powell and Sumner 2018
Task	Score	Study 1	Study 2	Study 3
Flanker	Reaction time cost	.40 (.12 to .61)	.57 (.36 to .72)	Not run
Flanker	Error cost	.58 (.35 to .74)	.72 (.57 to .83)	Not run
Stroop	Reaction time cost	.60 (.31 to .78)	.66 (.26 to .83)	Not run
Stroop	Error cost	.48 (.23 to .67)	.44 (.20 to .63)	Not run
Go/no-go	Commission errors, not a difference score	.76 (.58 to .87)	.76 (.60 to .86)	Not run
Stop-signal	Stop-signal reaction time, mean method	.47 (.21 to .67)	.43 (.19 to .62)	Not run
Stop-signal	Stop-signal reaction time, integration method	.36 (.08 to .59)	.49 (.26 to .66)	Not run
Posner cueing	Cueing effect	Not run	Not run	.70 (.50 to .83)
SNARC	Effect on reaction time	Not run	Not run	.22 (0 to .49)
SNARC	Effect on errors	Not run	Not run	.03 (0 to .34)
Navon	Local reaction time cost	Not run	Not run	.14 (0 to .43)
Navon	Local error cost	Not run	Not run	.82 (.69 to .90)
Navon	Global reaction time cost	Not run	Not run	0 (0 to .18)
Navon	Global error cost	Not run	Not run	.17 (0 to .46)
Navon	Global precedence effect, reaction time	Not run	Not run	0 (0 to .29)

Source: Hedge, Powell and Sumner 2018, The reliability paradox, Behavior Research Methods. https://doi.org/10.3758/s13428-017-0935-1. Samples of 50, 62 and 42 undergraduates. Task names link to the library's version of each task.

Published reliability figures

Figures the census quotes from their sources, each read in the paper before it was printed here.

Attention Network Test: Split-half reliability: executive control .81, orienting .55, alerting .38. (MacLeod et al. 2010) https://doi.org/10.1037/a0019803
Random Dot Motion Task: Test-retest ICC .60 for motion coherence thresholds. (Clark et al. 2022) https://doi.org/10.1167/jov.22.8.18
Dichotic Listening Task: A smartphone version retested at ICC .78 and showed the expected right-ear advantage. (Bless et al. 2013) https://doi.org/10.3389/fpsyg.2013.00042
Line Bisection Test: Test-retest ICC .97 in stroke patients with neglect. (Bailey et al. 2004) https://doi.org/10.1080/09602010343000282

How the count is made

Test-retest papers are OpenAlex records from any year whose title or abstract names the task together with “test-retest” or “retest reliability”. Papers whose main topic is in the physical sciences are left out.

The count measures how much attention a test's stability has had, not how stable it is: as the table above shows, a test with many test-retest papers can still have modest reliability for some scores. Evidence grade A needs at least 25 test-retest papers and 25 registered clinical studies.

Source: ConductCognition test census 0.22.0, counted 31 October 2026, 2 November 2026 and 3 November 2026 from OpenAlex and ClinicalTrials.gov.

Check reliability in your own sample

ConductCognition computes split-half reliability from your study's own data for supported measures, such as the Stroop and Flanker costs.

Start free

Data

Test-retest evidence for 451 cognitive tests

A test given twice should rank people the same way twice. Test-retest papers check that. Here is how many each test has, set against how much it is used.

Test families with 25 or more test-retest papers: 37 of 495
Most used test families with fewer than 25 test-retest papers: 22 of 50; See the well-used tests
Test families with no test-retest paper found: 177 of 495

Use set against test-retest evidence

Each dot is one test family on this site. The higher a dot sits for its place across, the more test-retest work it has for how much it is used.

Most test-retest papers

The tests whose reliability over time has been checked most often, in papers from any year.

1Verbal Fluency186 test-retest papers, 5,242 papers
2Trail Making Test A and Trail Making Test B161 test-retest papers, 3,944 papers
3Stroop Color-Word Test151 test-retest papers, 4,956 papers
4Digit Span Test136 test-retest papers, 2,755 papers
5Symbol Coding135 test-retest papers, 2,557 papers
6Verbal Learning134 test-retest papers, 2,503 papers
71-Back Task and 2-Back Task104 test-retest papers, 3,210 papers
8Matrix Reasoning89 test-retest papers, 1,106 papers
9Go/No-Go Task84 test-retest papers, 3,039 papers
10Picture Naming Test82 test-retest papers, 1,759 papers
11Contrast Sensitivity Test81 test-retest papers, 463 papers
12Implicit Association Test (IAT)68 test-retest papers, 1,497 papers
13Category Fluency56 test-retest papers, 2,471 papers
14Continuous Performance Test54 test-retest papers, 1,014 papers
15Short Story Recall52 test-retest papers, 1,038 papers
16Card Sorting Test (Berg)51 test-retest papers, 1,227 papers
17Vocabulary Test49 test-retest papers, 550 papers
18Delay Discounting and Food Delay Discounting48 test-retest papers, 2,088 papers
19Dichotic Listening Task48 test-retest papers, 280 papers
20Paced Visual Serial Addition Test45 test-retest papers, 397 papers

Well used, rarely retested

Well-used tests with the fewest test-retest papers
Test	Papers, 2021 to 2025	Test-retest papers
Metaphor Interference Task	436	1
Self-Paced Reading Task	740	2
Ball-Toss Game	358	2
Evaluative Conditioning Task	315	2
Ultimatum Game	592	3
Visuomotor Adaptation Task	405	3
Grammaticality Judgment Task	1,140	4
Judgments of Learning Task	379	4
Attentional Blink Task	332	4
Gaze Cueing Task	304	4
Biological Motion Task	624	5
Serial Naming Task	415	6
Social Value Orientation	348	7
Perceptual Confidence Task	306	7
Novel Word Learning Task	304	7

Strong effects, weaker rankings

Test-retest reliability of classic tasks, from Hedge, Powell and Sumner 2018
Task	Score	Study 1	Study 2	Study 3
Flanker	Reaction time cost	.40 (.12 to .61)	.57 (.36 to .72)	Not run
Flanker	Error cost	.58 (.35 to .74)	.72 (.57 to .83)	Not run
Stroop	Reaction time cost	.60 (.31 to .78)	.66 (.26 to .83)	Not run
Stroop	Error cost	.48 (.23 to .67)	.44 (.20 to .63)	Not run
Go/no-go	Commission errors, not a difference score	.76 (.58 to .87)	.76 (.60 to .86)	Not run
Stop-signal	Stop-signal reaction time, mean method	.47 (.21 to .67)	.43 (.19 to .62)	Not run
Stop-signal	Stop-signal reaction time, integration method	.36 (.08 to .59)	.49 (.26 to .66)	Not run
Posner cueing	Cueing effect	Not run	Not run	.70 (.50 to .83)
SNARC	Effect on reaction time	Not run	Not run	.22 (0 to .49)
SNARC	Effect on errors	Not run	Not run	.03 (0 to .34)
Navon	Local reaction time cost	Not run	Not run	.14 (0 to .43)
Navon	Local error cost	Not run	Not run	.82 (.69 to .90)
Navon	Global reaction time cost	Not run	Not run	0 (0 to .18)
Navon	Global error cost	Not run	Not run	.17 (0 to .46)
Navon	Global precedence effect, reaction time	Not run	Not run	0 (0 to .29)

Published reliability figures

Figures the census quotes from their sources, each read in the paper before it was printed here.

Attention Network Test: Split-half reliability: executive control .81, orienting .55, alerting .38. (MacLeod et al. 2010) https://doi.org/10.1037/a0019803
Random Dot Motion Task: Test-retest ICC .60 for motion coherence thresholds. (Clark et al. 2022) https://doi.org/10.1167/jov.22.8.18
Dichotic Listening Task: A smartphone version retested at ICC .78 and showed the expected right-ear advantage. (Bless et al. 2013) https://doi.org/10.3389/fpsyg.2013.00042
Line Bisection Test: Test-retest ICC .97 in stroke patients with neglect. (Bailey et al. 2004) https://doi.org/10.1080/09602010343000282

How the count is made

Source: ConductCognition test census 0.22.0, counted 31 October 2026, 2 November 2026 and 3 November 2026 from OpenAlex and ClinicalTrials.gov.

Check reliability in your own sample

ConductCognition computes split-half reliability from your study's own data for supported measures, such as the Stroop and Flanker costs.

Start free

Test-retest evidence for 451 cognitive tests

Use set against test-retest evidence

Most test-retest papers

Well used, rarely retested

Strong effects, weaker rankings

Published reliability figures

How the count is made

More on use and evidence

Check reliability in your own sample

Test-retest evidence for 451 cognitive tests

Use set against test-retest evidence

Most test-retest papers

Well used, rarely retested

Strong effects, weaker rankings

Published reliability figures

How the count is made

More on use and evidence

Check reliability in your own sample