Working Memory
Sternberg Memory Scanning Task
Also called: Sternberg task, Sternberg paradigm, Sternberg item recognition task
Measure how fast people search a short list held in memory: digits one at a time, a short wait, then a probe digit, scored as the slope of response time on list length.
10-11 minTimed: desktop or laptop with a mouse or trackpad
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
- 170
- Evidence grade
- C
- Registered clinical studies
- 38
What it measures
Measures how fast people search a short list held in memory. Participants see a list of 2, 4 or 6 digits one at a time, wait 2 seconds, and decide whether a last digit was in the list. The main score is the slope of response time on list length.
Constructs
- Short-term memory search
- Working memory
- Speed of comparison
Research uses
- Working memory research
- EEG and brain imaging studies of memory
- Within-study group comparison
How the paradigm works
Sternberg memory scanning measures how fast people search a short list held in memory. On each trial a list of 2, 4 or 6 different digits appears one digit at a time at the center of the screen, each for 1,200 ms. A cross then holds the screen for 2,000 ms, and a last digit, the probe, stays until the participant answers or 2,000 ms pass. Participants press one key if the probe was in the list and the other if it was not, as quickly and accurately as they can. Sternberg (1966) found that response time rises in a straight line with the length of the list, and read the slope, about 38 ms per digit, as the time to compare the probe with each digit in memory.
More about the paradigm
The timing follows Sternberg's varied-set procedure as Corbin and Marquer (2008) describe it, with a new list on every trial and the list lengths in random order. Digits come from 1 to 9. At each list length half of the probes were in the list, with every position in the list probed equally often, and half were digits that were not. Each session's order comes from its seed, and the F and J keys for yes and no are counterbalanced across sessions. Six practice trials give feedback; the test gives none. The digits and the cross switch on screen frames, and each display's measured duration is saved on every trial. Sternberg's participants also recalled the list aloud after answering. This version leaves that step out, as many later studies do; in Corbin and Marquer's replication the slope was about 47 ms per digit without recall and 62 ms with it.
The primary score is the scan slope: the slope, in ms per digit, of the straight line fitted to the median correct response time at each list length, with probes in and not in the list together. Scores also give the line's intercept, separate slopes for probes in and not in the list, the median correct response time and accuracy at each list length, and anticipations (responses under 100 ms, left out of the response times). The slope is meant for group comparisons rather than for one person's score: Corbin and Marquer (2009) found individual slopes from 6 to 139 ms per digit, and straight lines that fit some participants well and others poorly. Scores support research and do not diagnose.
Scores
- Scan Slopems/item
- Slope of the straight line fitted to the median correct response time at each list length, with probes in and not in the list together.
- The time each extra digit adds. Lower means faster scanning, but a flat slope can also come from guessing
- Scan Interceptms
- Height of the same line at a list length of 0: the time for everything other than the search.
- Lower is faster
- Slope, Probe in the Listms/item
- The same slope for probes that were in the list.
- Context for the scan slope
- Slope, Probe Not in the Listms/item
- The same slope for probes that were not in the list.
- Context for the scan slope
- Median RT, 1 Digitms
- Median correct response time for lists of one digit. Empty if the session did not run this length.
- Lower is faster
- Median RT, 2 Digitsms
- Median correct response time for lists of two digits. Empty if the session did not run this length.
- Lower is faster
- Median RT, 3 Digitsms
- Median correct response time for lists of three digits. Empty if the session did not run this length.
- Lower is faster
- Median RT, 4 Digitsms
- Median correct response time for lists of four digits. Empty if the session did not run this length.
- Lower is faster
- Median RT, 5 Digitsms
- Median correct response time for lists of five digits. Empty if the session did not run this length.
- Lower is faster
- Median RT, 6 Digitsms
- Median correct response time for lists of six digits. Empty if the session did not run this length.
- Lower is faster
- Accuracy, 1 Digit%
- Share of trials with lists of one digit answered correctly. Empty if the session did not run this length.
- Higher is better
- Accuracy, 2 Digits%
- Share of trials with lists of two digits answered correctly. Empty if the session did not run this length.
- Higher is better
- Accuracy, 3 Digits%
- Share of trials with lists of three digits answered correctly. Empty if the session did not run this length.
- Higher is better
- Accuracy, 4 Digits%
- Share of trials with lists of four digits answered correctly. Empty if the session did not run this length.
- Higher is better
- Accuracy, 5 Digits%
- Share of trials with lists of five digits answered correctly. Empty if the session did not run this length.
- Higher is better
- Accuracy, 6 Digits%
- Share of trials with lists of six digits answered correctly. Empty if the session did not run this length.
- Higher is better
- Overall Accuracy%
- Share of all test trials answered correctly. A missed response is incorrect.
- Under 80% is flagged for review
- Fewest Correct Trials at a List Lengthtrials
- Correct response times at the list length with the fewest. Each point of the line is a median of these.
- Fewer than 10 is flagged for review
- Anticipationscount
- Responses under 100 ms, left out of the response times.
- Context for the scores
- Missed Responsescount
- Test trials with no response within 2,000 ms.
- Context for the scores
Use and evidence
170 papers from 2021 to 2025 name this task in their title or abstract, up 15% compared with 2016 to 2020. That ranks it 140 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 36 |
| 2024 | 30 |
| 2023 | 33 |
| 2022 | 36 |
| 2021 | 35 |
| 2020 | 32 |
| 2019 | 24 |
| 2018 | 29 |
| 2017 | 23 |
| 2016 | 40 |
| 2015 | 38 |
| 2014 | 28 |
| 2013 | 32 |
| 2012 | 38 |
| 2011 | 39 |
| 2010 | 29 |
| 2009 | 32 |
| 2008 | 31 |
| 2007 | 30 |
| 2006 | 16 |
| 2005 | 20 |
| 2004 | 24 |
| 2003 | 20 |
| 2002 | 12 |
| 2001 | 19 |
| 2000 | 15 |
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
- 5
- Papers on reliability or validity
- 64
- Registered clinical studies using it as an outcome
- 38
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 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Trial
- An empty panel for 500 ms, the digits one at a time for 1,200 ms each, a cross for 2,000 ms, then a probe digit until a response or 2,000 ms.
- Lists
- 2, 4 or 6 different digits from 1 to 9, with a new list on every trial. Researchers can choose 1, 3 and 5, or 1 to 6.
- Trial mix
- 24 trials at each list length, 72 in all. Half of the probes were in the list, with every position probed equally often, and half were not.
- Trial order
- List lengths in a seeded random order, saved with the trial data, with no more than 3 probes of one kind in a row.
- Keys
- F and J for in the list and not in the list, counterbalanced across sessions.
- Practice
- 6 trials with feedback. The test gives none.
- Timing
- The digits and the cross switch on screen frames, and each display's measured duration is saved on every trial.
- Response times
- Correct responses of 100 ms or more enter the medians.
This version leaves out the recall step of Sternberg's procedure, as many later studies do. The slope is meant for group comparisons rather than one person's score, because individual slopes vary widely.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: sternberg_scanning_test from the scored run, F or J for a probe in the list or not in it (chance 50%, half the probes are in the list) within 2,000 ms, correctness from the key for the probe. Below chance invalidates and not above chance is flagged. One key on 90% of answers invalidates when accuracy is not above chance. Missed responses and responses under 100 ms follow the timed-test rule: more than max(3, 25%) missed or more than max(2, 20%) under 100 ms invalidate, and any is flagged. Completing fewer than half the planned trials invalidates, and any other incomplete run is flagged. Accuracy under 80%, or fewer than 10 correct responses at any list length, is flagged for review.
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). Missed responses and responses under 100 ms keep the platform's timed-test rule: more than max(3, 25%) missed responses or more than max(2, 20%) responses under 100 ms invalidate, and any is flagged. A response under 100 ms cannot be a reaction to the stimulus. Participants in Sternberg's procedure answer more than 90% correctly at every list length (Corbin and Marquer, 2008, https://doi.org/10.1080/09541440701688793), so accuracy under 80% means the response times may not reflect a search of the list. Each point of the slope is a median of correct response times, so it needs enough of them at every list length. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| List lengths | 2, 4 and 6 (standard) | 2, 4 and 6 (standard), 1, 3 and 5, 1 to 6 |
| Trials at each list length | 24 | 16 to 40 |
| Time for each digit (ms) | 1,200 | 500 to 2,000 |
| Wait before the probe (ms) | 2,000 | 1,000 to 4,000 |
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
- Timed test: participants need a desktop or laptop with a mouse or trackpad. 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 Sternberg Memory Scanning task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Sternberg, 1966), with the standard settings (list lengths: 2, 4 and 6; trials at each list length: 24; time for each digit (ms): 1,200; wait before the probe (ms): 2,000). The main score was scan slope (slope of the straight line fitted to the median correct response time at each list length, with probes in and not in the list together).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Sternberg, S. (1966). High-speed scanning in human memory. Science, 153(3736), 652-654. https://doi.org/10.1126/science.153.3736.652
- Sternberg, S. (1969). The discovery of processing stages: Extensions of Donders' method. Acta Psychologica, 30, 276-315. https://doi.org/10.1016/0001-6918(69)90055-9Memory scanning as a processing stage.
- Sternberg, S. (1975). Memory scanning: New findings and current controversies. Quarterly Journal of Experimental Psychology, 27(1), 1-32. https://doi.org/10.1080/14640747508400459Review of the findings.
- Corbin, L., & Marquer, J. (2008). Effect of a simple experimental control: The recall constraint in Sternberg's memory scanning task. European Journal of Cognitive Psychology, 20(5), 913-935. https://doi.org/10.1080/09541440701688793The varied-set procedure and its timing, with and without recall.
- Corbin, L., & Marquer, J. (2009). Individual differences in Sternberg's memory scanning task. Acta Psychologica, 131(2), 153-162. https://doi.org/10.1016/j.actpsy.2009.04.001How far individual slopes vary.
- Jungeblut, H. M., Hagemann, D., Löffler, C., & Schubert, A.-L. (2021). An investigation of the slope parameters of reaction times and P3 latencies in the Sternberg memory scanning task: A fixed-links model approach. Journal of Cognition, 4(1). https://doi.org/10.5334/joc.158List lengths of 1, 3 and 5, and the low reliability of slopes.
Source papers
- Sternberg S (1966). High-speed scanning in human memory. Science, 153(3736):652-654.Paradigm source, Adults
Questions
How long does the Sternberg Memory Scanning Task take?
About 10 to 11 minutes, as listed in the test library.
What device does a participant need?
Timed test: participants need a desktop or laptop with a mouse or trackpad. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Scan Slope, Scan Intercept, Slope, Probe in the List, Slope, Probe Not in the List, Median RT, 1 Digit, Median RT, 2 Digits, Median RT, 3 Digits, Median RT, 4 Digits, Median RT, 5 Digits, Median RT, 6 Digits, Accuracy, 1 Digit, Accuracy, 2 Digits, Accuracy, 3 Digits, Accuracy, 4 Digits, Accuracy, 5 Digits, Accuracy, 6 Digits, Overall Accuracy, Fewest Correct Trials at a List Length, Anticipations and Missed Responses. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Sternberg Memory Scanning Task used in research?
170 papers from 2021 to 2025 name it in their title or abstract, up 15% compared with 2016 to 2020. That ranks it 140 of 395 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Sternberg Memory Scanning Task have?
In the published research, 5 papers report its test-retest reliability and 38 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?
List lengths (2, 4 and 6 (standard), 1, 3 and 5, 1 to 6), trials at each list length (16 to 40), time for each digit (ms) (500 to 2,000) and wait before the probe (ms) (1,000 to 4,000). 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.
Related tests