Attention
Uncanny Valley Task
Also called: Uncanny valley ratings, human likeness and eeriness
Measure how human and how eerie faces look across five appearances, from stylized to photographic, with real photographs for comparison.
5-7 minUntimed
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
- 593
- Evidence grade
- C
- Registered clinical studies
- 0
What it measures
Measures how human and how eerie faces look as their appearance changes from stylized to photographic. Participants give two ratings for each face. Real photographs provide a comparison.
Constructs
- Human likeness
- Eeriness
- Face perception
Research uses
- Research on artificial faces and avatars
- Research on the uncanny valley
How the paradigm works
The uncanny valley describes a proposed rise in eeriness as an artificial face approaches a human appearance. This task measures the relation directly with separate ratings of human likeness and eeriness.
More about the paradigm
Participants rate 36 pictures by default: five appearances of each of six generated identities and six real photographs. A shorter setting uses 24 pictures. Each generated portrait is blended with its own flat stylization at five fixed weights. Real photographs are separate comparison pictures. The order is seeded, with no identity shown twice in a row. Two practice pictures introduce both seven-point scales.
The primary measure compares eeriness at the near-real level with the mean of the two endpoints within each identity. Positive values show a near-real rise. Mean ratings at all five levels, human-likeness change, real-photo ratings and descriptive response times show the wider pattern.
This research version uses new generated faces and code-made stylizations, rather than published morphs. Ratings need not form a valley. Norms are off; compare raw measures within a study.
Scores
- Near-real eeriness contrastpoints
- Within-identity eeriness at level 3 minus the mean of levels 0 and 4, averaged across identities with all three ratings. Positive means a near-real rise in eeriness.
- Informational
- Human likeness, 0% photopoints
- Mean 1-to-7 human likeness rating at level 0 (0% photo weight).
- Informational
- Human likeness, 25% photopoints
- Mean 1-to-7 human likeness rating at level 1 (25% photo weight).
- Informational
- Human likeness, 50% photopoints
- Mean 1-to-7 human likeness rating at level 2 (50% photo weight).
- Informational
- Human likeness, 75% photopoints
- Mean 1-to-7 human likeness rating at level 3 (75% photo weight).
- Informational
- Human likeness, 100% photopoints
- Mean 1-to-7 human likeness rating at level 4 (100% photo weight).
- Informational
- Eeriness, 0% photopoints
- Mean 1-to-7 eeriness rating at level 0 (0% photo weight).
- Informational
- Eeriness, 25% photopoints
- Mean 1-to-7 eeriness rating at level 1 (25% photo weight).
- Informational
- Eeriness, 50% photopoints
- Mean 1-to-7 eeriness rating at level 2 (50% photo weight).
- Informational
- Eeriness, 75% photopoints
- Mean 1-to-7 eeriness rating at level 3 (75% photo weight).
- Informational
- Eeriness, 100% photopoints
- Mean 1-to-7 eeriness rating at level 4 (100% photo weight).
- Informational
- Real-photo human likenesspoints
- Mean human-likeness rating for separate real photographs.
- Informational
- Real-photo eerinesspoints
- Mean eeriness rating for separate real photographs.
- Informational
- Human-likeness gainpoints
- Mean within-identity human-likeness difference between 100% and 0% photo weight.
- Informational
- Median rating timems
- Median time for valid ratings, excluding responses under 100 ms.
- Informational
- Pictures ratedcount
- Known pictures with both valid ratings.
- Informational
- Planned picturescount
- 24 or 36 test pictures.
- Informational
- Ratings answeredcount
- Unique picture/scale pairs with valid 1-to-7 ratings.
- Informational
- Anticipationscount
- Presses under 100 ms, including those the task ignores.
- Informational
Use and evidence
593 papers from 2021 to 2025 name this task in their title or abstract, up 59% compared with 2016 to 2020. That ranks it 40 of 582 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 207 |
| 2024 | 122 |
| 2023 | 109 |
| 2022 | 71 |
| 2021 | 84 |
| 2020 | 87 |
| 2019 | 92 |
| 2018 | 77 |
| 2017 | 68 |
| 2016 | 48 |
| 2015 | 75 |
| 2014 | 64 |
| 2013 | 43 |
| 2012 | 27 |
| 2011 | 28 |
| 2010 | 25 |
| 2009 | 24 |
| 2008 | 11 |
| 2007 | 14 |
| 2006 | 11 |
| 2005 | 7 |
| 2004 | 1 |
| 2003 | 1 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade B
300 to 999 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
- 0
- Papers on reliability or validity
- 49
- Registered clinical studies using it as an outcome
- 0
Measured on ConductCognition
Pooled across the studies that feed our platform norms, by device group. A figure needs 50 or more people.
No pooled figure yet.
Note on the counts: many papers study robots, avatars or design and run no rating task.
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 0.39.0, counted 8 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Pictures
- Six generated identities at five levels each, plus six separate London Set photographs: 36 pictures. The short setting uses four identities and four photographs: 24 pictures.
- Continuum
- Each generated portrait is blended with a flat, code-made stylization at photo weights of 0%, 25%, 50%, 75% and 100%. Real photographs are separate anchors; no real face enters a generated blend.
- Ratings
- Two seven-point scales per picture: human likeness and eeriness. One scale at a time, with the order counterbalanced across session seeds. Click, tap or press 1 to 7; no time limit.
- Order and timing
- A seeded random order, with no identity in consecutive pictures. A 400 ms blank follows the second rating. A rest comes halfway through.
- Practice and debrief
- Two separate pictures introduce both scales. Ratings have no correct answer. The end screen identifies every generated test picture and explains both practice pictures.
- Primary measure
- Within each generated identity, eeriness at 75% photo weight minus the mean at 0% and 100%, averaged across identities with all three ratings. Positive values show a near-real rise in eeriness. The five-level ratings and real-photo ratings are also reported.
This version uses our own generated portraits and code-made blends. It adapts the human-likeness and eeriness rating method, without copying the published models, morphs or image set. The task measures the observed rating curve. Raw ratings support within-study comparisons; norms are off. London Set photographs are credited to DeBruine and Jones under CC BY 4.0. Including practice, the photographs share people with these tests: Face Memory Test (2), Face-name Memory Task (5), Face Matching Task (7), Chimeric Faces Task (2), Part-Whole Face Task (5), Thatcher Illusion Task (1), Two-Tone Pictures Task (2), Verbal Overshadowing Task (4). Run face-memory tasks before this task or in a separate session, because seeing a face can affect later judgments. The Real or AI Face Task shares 6 test people, one practice person, 6 generated test identities and one generated practice identity. Including practice pictures and foils, this task shares people with Face Impressions Task (7).
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: uncanny_valley_test from the scored run. Ratings have no accuracy rule. Repeated ratings are retained. Responses under 200 ms on at least half the ratings invalidate. Missing ratings are flagged for review.
Why these rules
The two scales record subjective judgments, including flat or monotonic patterns. 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.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Generated identities | 6 (36 pictures, standard) | 6 (36 pictures, standard), 4 (24 pictures) |
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.
- 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 Uncanny Valley Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Burleigh et al., 2013), with the standard settings (generated identities: 6 (36 pictures, standard)). The main score was near-real eeriness contrast (within-identity eeriness at level 3 minus the mean of levels 0 and 4, averaged across identities with all three ratings).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Burleigh, T. J., Schoenherr, J. R., & Lacroix, G. L. (2013). Does the uncanny valley exist? An empirical test of the relationship between eeriness and the human likeness of digitally created faces. Computers in Human Behavior, 29(3), 759-771. https://doi.org/10.1016/j.chb.2012.11.021Human-likeness and eeriness ratings; the shape of the relation depends on the stimulus continuum.
- DeBruine, L., & Jones, B. (2017). Face Research Lab London Set [Data set]. figshare. https://doi.org/10.6084/m9.figshare.5047666.v5Real-photo anchors, CC BY 4.0.
Source papers
Questions
How long does the Uncanny Valley Task take?
About 5 to 7 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?
Near-real eeriness contrast, Human likeness, 0% photo, Human likeness, 25% photo, Human likeness, 50% photo, Human likeness, 75% photo, Human likeness, 100% photo, Eeriness, 0% photo, Eeriness, 25% photo, Eeriness, 50% photo, Eeriness, 75% photo, Eeriness, 100% photo, Real-photo human likeness, Real-photo eeriness, Human-likeness gain, Median rating time, Pictures rated, Planned pictures, Ratings answered and Anticipations. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Uncanny Valley Task used in research?
593 papers from 2021 to 2025 name it in their title or abstract, up 59% compared with 2016 to 2020. That ranks it 40 of 582 test families in the ConductCognition test census (use grade B). Source: ConductCognition test census 0.39.0, counted 8 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Uncanny Valley 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 C in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Generated identities (6 (36 pictures, standard), 4 (24 pictures)). 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.