Attention
Gradual-Onset Continuous Performance Task
Also called: gradCPT, gradual onset CPT, gradual-onset continuous performance test
Measure sustained attention and its moment-to-moment ups and downs: scenes of cities and mountains fade into one another, and participants press for cities and hold back for rare mountains.
5-6 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
- 36
- Evidence grade
- C
- Registered clinical studies
- 19
What it measures
Measures sustained attention and its moment-to-moment ups and downs. Round scenes of cities and mountains, drawn for this test, fade into one another every 800 ms; the person presses for each city (90%) and holds back for each rare mountain. Scores sensitivity (d prime), both kinds of error and how response times vary over time.
Constructs
- Sustained attention
- Attentional fluctuation
- Response inhibition
Research uses
- Sustained attention and mind-wandering research
- Attention across the lifespan
- Within-study group comparison
How the paradigm works
The Gradual-Onset Continuous Performance Task (gradCPT) measures sustained attention and how it rises and falls from moment to moment. Round scenes of cities and mountains fade into one another, one every 800 ms, and the participant presses the space bar for each city and holds back for each mountain. Cities are 90% of the scenes, so pressing becomes a habit and the rare mountain has to be caught. Because each scene fades in rather than appearing at once, no sudden change pulls attention to the screen: the person has to keep it there.
More about the paradigm
The task follows Esterman, Noonan, Rosenberg and DeGutis (2013) and the web version of Fortenbaugh and colleagues (2015): 10 city and 10 mountain scenes in grayscale inside a circle, a linear fade of 800 ms, mountains on 10% of scenes, the same scene never twice in a row, a scrambled city to begin and end the run, and 4 minutes as standard (4 to 12 in the settings; the lab version ran 8). The original scenes are photographs; here every scene is drawn by ConductScience code, with the same sky, ground and grays in both kinds, so cities and mountains differ only in their shapes. Drawings carry less detail than photographs, which may make the scenes easier to tell apart than in published studies, so scores are best compared within this version. Mountains are never first and never next to each other, and a 30 second practice with a message after each mistake comes first.
Presses are given to scenes afterwards by the published rule: a press from 70% of a scene's fade in to 40% of the next scene's is that scene's, and a press in between goes to whichever scene has none, to the city when a mountain is the other, and otherwise to the scene more visible. Response times run from the start of a scene's fade in. The main score is d prime, from the share of mountains held back and the share of cities missed (Fortenbaugh and colleagues, 2015), with criterion as the strategy beside it. The test also reports commission errors (mountains pressed for) and omission errors (cities missed), the mean and the coefficient of variation of response times, and error rates in the steady and the variable half of the run: the variance time course of Esterman and colleagues (2013) splits the run by how far each response time strays from the person's mean, smoothed over neighboring scenes. Scores support research and do not diagnose.
Scores
- d-Primed'
- z of the share of mountains held back minus z of the share of cities missed, with a rate of 0 or 1 moved half a trial in.
- Higher means cities and mountains were told apart better
- Commission Error Rate%
- Share of mountains pressed for.
- Lower is better
- Omission Error Rate%
- Share of cities with no press.
- Lower is better
- Criterionz
- Minus the mean of the two z values behind d prime.
- Positive means a readier press, negative a more cautious one
- Mean Response Timems
- Mean time from the start of a city's fade in to its press.
- Informational: about 800 ms is a press when the city is fully clear
- Response Time Variabilityscore
- Standard deviation of the cities' response times over their mean (coefficient of variation).
- Lower means steadier responding
- Commission Error Rate, in the Zone%
- Share of mountains pressed for in the steady half of the run, by the variance time course.
- Lower is better
- Commission Error Rate, out of the Zone%
- Share of mountains pressed for in the variable half of the run, by the variance time course.
- Usually higher than in the zone
- Omission Error Rate, in the Zone%
- Share of cities missed in the steady half of the run.
- Lower is better
- Omission Error Rate, out of the Zone%
- Share of cities missed in the variable half of the run.
- Usually higher than in the zone
- Commission Errorscount
- Mountains pressed for.
- Lower is better
- Omission Errorscount
- Cities with no press.
- Lower is better
- Mountainscount
- Mountain scenes drawn.
- Informational
- Citiescount
- City scenes drawn.
- Informational
- Ambiguous Pressescount
- Presses between two scenes' windows, given to a scene by the published rule.
- Quality check
- Extra Pressescount
- Presses no scene took: a second press for a scene, or a press outside every window.
- Quality check
- Longest Pauseseconds
- Longest time without a press during the run.
- 30 seconds or more invalidates the session
- Scenes Showncount
- Scenes drawn while fading in.
- Under 90% of the planned scenes is invalid
- Planned Scenescount
- Scenes the session planned, 75 a minute.
- Set by the settings
Use and evidence
36 papers from 2021 to 2025 name this task in their title or abstract, up 64% compared with 2016 to 2020. That ranks it 277 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 10 |
| 2024 | 8 |
| 2023 | 6 |
| 2022 | 3 |
| 2021 | 9 |
| 2020 | 4 |
| 2019 | 4 |
| 2018 | 5 |
| 2017 | 5 |
| 2016 | 4 |
| 2015 | 3 |
| 2014 | 1 |
| 2013 | 2 |
| 2012 | 1 |
| 2011 | 2 |
| 2010 | 0 |
| 2009 | 0 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 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
- 8
- Registered clinical studies using it as an outcome
- 19
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.
- Scenes
- 10 cities and 10 mountains in grayscale inside a circle 256 px across, drawn by code with the same sky, ground and grays, so the two kinds differ only in shape.
- Fade
- Each scene fades into the next over 800 ms by a linear mix, set on every screen frame from the run's clock; a scrambled city opens and closes the run.
- Order
- Mountains on 10% of scenes, never first and never next to each other; the same scene never twice in a row. 75 scenes a minute.
- Answer
- The space bar for every city, nothing for a mountain. Instructions ask for accuracy, and the pace sets the deadline.
- Length
- 4 minutes (standard; 4 to 12) without a break, saved one row per minute.
- Practice
- 40 scenes, 4 of them mountains, with a message after each mountain pressed for and each city missed.
- Scoring
- Presses go to scenes by the published windows: from 70% of a scene's fade in to 40% of the next scene's, with a rule for the presses in between. d prime from mountains held back and cities missed; error rates in and out of the zone by the variance time course.
The published task uses photographs of cities and mountains. These drawn scenes carry less detail, which may make them easier to tell apart, so compare scores within this version rather than with published values.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: gradual_onset_cpt_minute from the scored run, one per minute, read as one trial per scene drawn, with presses given to scenes by the published gradCPT rule: a press is needed for each city (90%) and none for a mountain (10%). Presses mostly on mountains invalidate, and presses that do not separate the two are flagged. Missing more than max(3, 25%) of the cities invalidates; any miss is flagged. Under 90% of the planned scenes drawn, or 30 seconds or more without a press, invalidates. By the published windows no press is given to a scene less than 320 ms into its fade in, so the 100 ms anticipation rule never applies.
Why these rules
Commission errors on mountains are the measure, so pressing on every scene is flagged rather than invalidated. Responses concentrated on trials that needed none (one-sided Fisher exact test, p < .05) mean the instructions were reversed and invalidate; responses that do not separate the two trial types are flagged. Responding is the common, prepotent response here, so missing more than max(3, 25%) of response trials signals non-engagement (the platform's timed-test rule); any miss is flagged. Fortenbaugh and colleagues (2015) left out sessions with 30 seconds or more without a response, since cities come about every second. The 90% limit is a ConductScience choice, as for the Clock Vigilance Test. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Test length (minutes) | 4 | 4 to 12 |
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.
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Gradual-Onset Continuous Performance Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Esterman et al., 2013), with the standard settings (test length (minutes): 4). The main score was d-prime (z of the share of mountains held back minus z of the share of cities missed, with a rate of 0 or 1 moved half a trial in).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Esterman, M., Noonan, S. K., Rosenberg, M., & DeGutis, J. (2013). In the zone or zoning out? Tracking behavioral and neural fluctuations during sustained attention. Cerebral Cortex, 23(11), 2712-2723. https://doi.org/10.1093/cercor/bhs261
- Rosenberg, M., Noonan, S., DeGutis, J., & Esterman, M. (2013). Sustaining visual attention in the face of distraction: A novel gradual-onset continuous performance task. Attention, Perception, & Psychophysics, 75(3), 426-439. https://doi.org/10.3758/s13414-012-0413-xThe first gradCPT, with faces fading into one another.
- Fortenbaugh, F. C., DeGutis, J., Germine, L., Wilmer, J. B., Grosso, M., Russo, K., & Esterman, M. (2015). Sustained attention across the life span in a sample of 10,000: Dissociating ability and strategy. Psychological Science, 26(9), 1497-1510. https://doi.org/10.1177/0956797615594896The 4 minute web version, its d prime and criterion, and the variance time course.
- Fortenbaugh, F. C., Rothlein, D., McGlinchey, R., DeGutis, J., & Esterman, M. (2018). Tracking behavioral and neural fluctuations during sustained attention: A robust replication and extension. NeuroImage, 171, 148-164. https://doi.org/10.1016/j.neuroimage.2018.01.002The variance time course step by step.
- Brau, J. M., Sugarman, A., Rothlein, D., DeGutis, J., Esterman, M., & Fortenbaugh, F. C. (2022). The impact of image degradation and temporal dynamics on sustained attention. Journal of Vision, 22(4), 8. https://doi.org/10.1167/jov.22.4.8How the look of the scenes changes performance.
Source papers
- Esterman M, Noonan SK, Rosenberg M, DeGutis J (2013). In the zone or zoning out? Tracking behavioral and neural fluctuations during sustained attention. Cerebral Cortex, 23(11):2712-2723.Paradigm source, Adults
Questions
How long does the Gradual-Onset Continuous Performance Task take?
About 5 to 6 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?
d-Prime, Commission Error Rate, Omission Error Rate, Criterion, Mean Response Time, Response Time Variability, Commission Error Rate, in the Zone, Commission Error Rate, out of the Zone, Omission Error Rate, in the Zone, Omission Error Rate, out of the Zone, Commission Errors, Omission Errors, Mountains, Cities, Ambiguous Presses, Extra Presses, Longest Pause, Scenes Shown and Planned Scenes. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Gradual-Onset Continuous Performance Task used in research?
36 papers from 2021 to 2025 name it in their title or abstract, up 64% compared with 2016 to 2020. That ranks it 277 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 Gradual-Onset Continuous Performance Task have?
In the published research, 1 paper reports its test-retest reliability and 19 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Which settings can a researcher change?
Test length (minutes) (4 to 12). 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.