Attention
Face Pareidolia Task
Also called: illusory face detection, faces in noise, false face detection, seeing faces in noise
Measure how often people see a face in noise that holds none, beside how well they see faint drawn faces in the same noise.
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. Try it here first, with no account.
- Papers, 2021 to 2025
- 157
- Evidence grade
- D
- Registered clinical studies
- 1
What it measures
A measure of seeing faces in noise. Pictures of gray noise, like clouds, flash for 600 ms; most hold no face and one in four holds a faint drawn face of two eyes, a nose and a mouth. Participants press F if they saw a face or J if they did not, with no time limit. The main score is the share of noise-only pictures answered as a face, with sensitivity to the faint faces beside it. The scores are research measures.
Constructs
- Face pareidolia
- Face detection
- Perceptual decision bias
- Visual sensitivity
Research uses
- Face perception research
- Research on illusory perception and perceptual bias
- Within-study comparison of face reports across conditions or groups
How the paradigm works
The Face Pareidolia Task measures how often people see a face in pictures that hold none, beside how well they see a faint face that is there. In the illusory face detection studies of Rieth, Lee, Lui, Tian and Huber (2011) and Liu and colleagues (2014), people judged brief pictures of blotchy noise after practice with real faces hidden in noise. Told that half the pictures held a face, Liu's participants reported a face on 34% of pictures of noise alone. When an oval marked where a face would be, the noise behind face reports held dark patches in the places of the eyes, nose and mouth (Rieth and colleagues, Experiment 3). Signal detection studies of faces in noise separate the tendency to report a face from the ability to see one (Partos, Cropper and Rawlings, 2016).
More about the paradigm
Each trial shows a cross for 500 ms, then a picture of gray noise for 600 ms (36 frames of a 60 Hz screen), drawn by our code from three blurred noise fields. Three pictures in four hold noise alone; the fourth holds a faint drawn face, two dark eyes, a nose and a mouth, added to the noise at a fixed strength. No photograph or real face is used. Participants press F if they saw a face or J if they did not, with no time limit. The pictures are a fixed set, the same in every session, drawn from seeds the data save: 24 rounds of four (standard; 16 to 40) give 96 pictures in a seeded order, with a short break after every 48, after 6 practice pictures with feedback, the first two with a clearer face. Every display time is set before the session, and no answer changes a later picture. Displays switch on screen frames, the measured durations are saved on every trial, and trials more than one frame off plan leave the score. Desktop and laptop only; about 5 to 6 minutes.
The main score is the share of noise-only pictures answered as a face. The share of faint faces answered as a face, sensitivity (d') and response bias (criterion c) are reported beside it. The scores are research measures: the task gives no diagnosis, no cut-off and no risk of any condition.
Scores
- Face Reports on Noise%
- Share of pictures of noise alone answered as holding a face.
- Informational; a research measure with no better direction
- Face Reports on Faint Faces%
- Share of pictures with a faint drawn face answered as holding a face.
- Higher means more faint faces seen
- Sensitivity (d')d'
- z of the face report rate on faint faces minus z of the rate on noise alone, both log-linear corrected.
- Higher means faint faces were told apart from noise better; 0 is no difference
- Response Criterion (c)z
- Minus the mean of the two z scores; where a person sets the line between face and no face.
- Below 0 leans toward answering face; above 0 leans away
- Noise Pictures Scoredtrials
- Usable answers to pictures of noise alone.
- Informational
- Faint-Face Pictures Scoredtrials
- Usable answers to pictures with a faint face.
- Informational
- Median Answer Timems
- Median time from the question to the answer.
- Informational; the answer has no time limit
- Anticipationscount
- Answers under 100 ms from the question, which are not scored.
- Lower is better
- Early Key Pressescount
- F or J pressed before the question appeared, which are not taken as answers.
- Informational
- Frame Intervalms
- Median time between screen frames during the trials.
- About 16.7 on a 60 Hz screen
- Refresh RateHz
- Screen refresh rate from the frame interval.
- Informational
- Pictures on Time%
- Share of answered trials whose picture showed within one frame of plan; only these are scored.
- Under 0.9 is flagged for review
- Completed Trialstrials
- Trials answered in the scored run.
- Fewer than planned is flagged for review
- Planned Trialstrials
- Trials the run set out to show.
- Informational
Use and evidence
157 papers from 2021 to 2025 name this task in their title or abstract, up 134% compared with 2016 to 2020. That ranks it 167 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 53 |
| 2024 | 29 |
| 2023 | 27 |
| 2022 | 26 |
| 2021 | 22 |
| 2020 | 21 |
| 2019 | 12 |
| 2018 | 11 |
| 2017 | 13 |
| 2016 | 10 |
| 2015 | 5 |
| 2014 | 5 |
| 2013 | 1 |
| 2012 | 1 |
| 2011 | 0 |
| 2010 | 1 |
| 2009 | 3 |
| 2008 | 0 |
| 2007 | 0 |
| 2006 | 1 |
| 2005 | 0 |
| 2004 | 1 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 1 |
| 2000 | 0 |
Use grade C
100 to 299 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 1
- Papers on reliability or validity
- 10
- Registered clinical studies using it as an outcome
- 1
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 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Pictures
- On a dark panel: a cross for 500 ms, then a picture of gray noise 320 px across, drawn by our code at 128 by 128 pixels from three blurred noise fields (blur SD 6, 12 and 24 pixels). No photograph or real face is used.
- Faces
- One picture in four holds a faint drawn face, two dark eyes, a nose and a mouth, added at 1.5 noise SDs with its center within 6 pixels of the picture's center. The other three hold noise alone.
- Duration
- Each picture shows for 36 frames of a 60 Hz screen, 600 ms, in whole frames of the screen's own rate.
- Answer
- Press F if you saw a face or J if you did not, with no time limit. Keys pressed before the question are counted and not taken as answers.
- Trials
- 24 rounds of four pictures (standard; 16 to 40): 96 pictures, the same fixed set in every session, in a seeded order, with a short break after every 48.
- Practice
- 6 pictures with feedback for 1,500 ms; the first two faces are clearer, at 3 noise SDs.
- Fixed times
- Every display time is set before the session, and no answer changes a later picture: there is no staircase and no training mode.
- Score
- The share of noise-only pictures answered as a face, with the share of faint faces answered as a face, d' and criterion c beside it. Trials whose picture ran more than one frame off plan, and answers under 100 ms, leave the score.
The noise and the faces are drawn by our code after the three-scale noise of the illusory face studies, not taken from them, and the faces are schematic, so rates are not comparable with published ones. Participants are told truthfully that most pictures hold no face, where the earlier studies told them half did. The scores are research measures: the task gives no diagnosis, no cut-off and no risk score.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: face_pareidolia_test from the scored run, F (a face) or J (no face) after each picture, with no time limit, read from the saved picture. Face answers to pictures of noise alone are what the test measures, so accuracy is not judged against chance. Face answers significantly more often on noise alone than on faint faces (one-sided Fisher exact test, p < .05) invalidate; face answers that do not separate the two kinds of picture are flagged. Responses under 100 ms follow the timed-test rule. A run short of its planned trials, or fewer than 90% of trials with the picture within one frame of plan, is flagged for review.
Why these rules
A face answer to noise alone is not an error to count against chance: it is the illusory face report the test exists to measure. The faint faces give the answers a direction, so face answers that come mostly on noise alone mean the keys were reversed, and answers that do not tell faint faces from noise mean the faces were not seen or the task was not followed. An answer under 100 ms from the question was pressed before the question could be read. The picture's 600 ms carries the measure, so trials whose picture ran late leave it.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Rounds of pictures | 24 | 16 to 40 |
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 Face Pareidolia Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Rieth, Lee, Lui, Tian and Huber, 2011; Liu and colleagues, 2014), with the standard settings (rounds of pictures: 24). The main score was face reports on noise (share of pictures of noise alone answered as holding a face).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Rieth, C. A., Lee, K., Lui, J., Tian, J., & Huber, D. E. (2011). Faces in the mist: Illusory face and letter detection. i-Perception, 2(5), 458-476. https://doi.org/10.1068/i0421The illusory face detection task: brief pictures of noise made from dark blobs at three scales, a 600 ms display, and a classification image of eyes, nose and mouth when an oval marked the face's place.
- Liu, J., Li, J., Feng, L., Li, L., Tian, J., & Lee, K. (2014). Seeing Jesus in toast: Neural and behavioral correlates of face pareidolia. Cortex, 53, 60-77. https://doi.org/10.1016/j.cortex.2014.01.013Face reports on pictures of noise alone as the measure of face pareidolia, with a 600 ms display; 34% of pure-noise pictures reported as faces.
- Partos, T. R., Cropper, S. J., & Rawlings, D. (2016). You don't see what I see: Individual differences in the perception of meaning from visual stimuli. PLoS ONE, 11(3), e0150615. https://doi.org/10.1371/journal.pone.0150615Faces of varying visibility in noise scored by signal detection, separating response bias from sensitivity; expecting more faces than were shown raised false alarms.
- Cropper, S. J., McCauley, A., Gwinn, O. S., Bartlett, M., & Nicholls, M. E. R. (2020). Flowers in the attic: Lateralization of the detection of meaning in visual noise. Journal of Vision, 20(10), 11. https://doi.org/10.1167/jov.20.10.11Faces in noise answered yes or no and scored by signal detection; with three noise-only pictures for each face, false alarms did not change.
- Mamiya, Y., Nishio, Y., Watanabe, H., Yokoi, K., Uchiyama, M., Baba, T., Iizuka, O., Kanno, S., Kamimura, N., Kazui, H., Hashimoto, M., Ikeda, M., Takeshita, C., Shimomura, T., & Mori, E. (2016). The pareidolia test: A simple neuropsychological test measuring visual hallucination-like illusions. PLoS ONE, 11(5), e0154713. https://doi.org/10.1371/journal.pone.0154713A published picture test with faces hidden in noise. This task uses none of its pictures, scoring or cut-offs.
Source papers
- Liu J, Li J, Feng L, Li L, Tian J, Lee K (2014). Seeing Jesus in toast: Neural and behavioral correlates of face pareidolia. Cortex, 53:60-77.N 20, 18-25
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Face Pareidolia 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?
Face Reports on Noise, Face Reports on Faint Faces, Sensitivity (d'), Response Criterion (c), Noise Pictures Scored, Faint-Face Pictures Scored, Median Answer Time, Anticipations, Early Key Presses, Frame Interval, Refresh Rate, Pictures on Time, Completed Trials and Planned Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Face Pareidolia Task used in research?
157 papers from 2021 to 2025 name it in their title or abstract, up 134% compared with 2016 to 2020. That ranks it 167 of 495 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Face Pareidolia Task have?
In the published research, 1 paper reports its test-retest reliability and 1 registered clinical study uses it as an outcome (evidence grade D in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Rounds of pictures (16 to 40). 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.