Social Cognition
Vocal Emotion Task
Also called: vocal emotion recognition task, emotional prosody recognition task, vocal affect recognition task
Measure how accurately people recognize emotion in a voice: sentences recorded by actors, each answered anger, disgust, fear, happiness, sadness or none, scored by percent correct and the unbiased hit rate for each emotion.
6-8 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
- 120
- Evidence grade
- C
- Registered clinical studies
- 6
What it measures
Measures how accurately people recognize emotion in a voice. Participants hear 48 short sentences, each spoken by an actor with anger, disgust, fear, happiness, sadness or no emotion, and choose which one they hear. The voices are recordings of actors from the CREMA-D set; the sentences are English and neutral in meaning, so the emotion is in the voice. Plays sound through speakers or headphones. Results are research measures.
Constructs
- Emotion recognition
- Emotional prosody perception
- Social cognition
Research uses
- Emotion perception research
- Social cognition studies in clinical and nonclinical groups
- Within-study group comparison
How the paradigm works
The Vocal Emotion Task measures how accurately people recognize emotion in a voice. Each trial plays one short sentence spoken by an actor, and the participant chooses the emotion the voice expresses: anger, disgust, fear, happiness, sadness or none. It follows the method of Banse and Scherer (1996), in which professional actors portrayed emotions in meaningless sentences and listeners heard each portrayal once and chose one of a set of emotion words.
More about the paradigm
The voices are recordings of actors from CREMA-D, the Crowd-sourced Emotional Multimodal Actors Dataset (Cao and colleagues, 2014), whose raters chose among the same five emotions or no emotion, for the voice alone, the face alone or both. The test plays 54 of them, chosen by a stated rule from those ratings: for each emotion, 9 clips spoken at the set's unspecified intensity whose voice was most often judged to show the emotion acted, one per actor and per sentence within an emotion, from 39 actors, as many women as men. One clip per emotion is for practice and is followed by the answer; the 48 others, 8 per emotion, form the test and play once each, in an order drawn from the session's seed with no emotion more than twice in a row and never one actor twice in a row. Every clip plays at the same speech level, so loudness is not a cue. The participant presses Play, hears the sentence once, and answers when it ends, with no time limit. Speakers or headphones both work.
The primary score is percent correct; chance is 1 in 6 (16.7%). Percent correct and the unbiased hit rate of Wagner (1993) are also reported for each emotion: the hits squared, divided by the trials of that emotion times the number of times its answer was chosen, which allows for how often each answer is used. The confusion matrix counts the answers chosen for each emotion acted. The clips are the ones CREMA-D's raters agreed on most, so accuracy here is expected to run higher than across the whole set, which CREMA-D's own figures describe. The portrayals are acted, not spontaneous.
The sentences are English in every session; the instructions and answers follow the session language, and participants are told they do not need to understand the words. Listeners recognize emotion in voices from other languages and cultures well above chance, but less well on average than in voices from their own (Scherer, Banse and Wallbott, 2001; Laukka and Elfenbein, 2021), so sessions run in other languages are flagged for review and norms are kept apart by language. Vocal emotion recognition is studied in research on schizophrenia (Hoekert and colleagues, 2007) and autism (Leung and colleagues, 2022), among others; scores support research and do not diagnose.
Scores
- Percent Correct%
- Percent of answered test trials given the emotion the actor portrayed. Chance is 16.7% (1 in 6).
- Higher is better
- Mean Unbiased Hit Rate%
- Mean of the unbiased hit rates of the six emotions.
- Higher is better
- Percent Correct, Anger%
- Percent of anger trials answered anger.
- Higher is better
- Percent Correct, Disgust%
- Percent of disgust trials answered disgust.
- Higher is better
- Percent Correct, Fear%
- Percent of fear trials answered fear.
- Higher is better
- Percent Correct, Happiness%
- Percent of happiness trials answered happiness.
- Higher is better
- Percent Correct, Sadness%
- Percent of sadness trials answered sadness.
- Higher is better
- Percent Correct, No Emotion%
- Percent of trials with no emotion acted that were answered none.
- Higher is better
- Unbiased Hit Rate, Anger%
- Wagner's unbiased hit rate for anger: hits squared divided by (anger trials x times anger was chosen). 0 when no anger trial was answered anger.
- Higher is better
- Unbiased Hit Rate, Disgust%
- Wagner's unbiased hit rate for disgust, computed the same way.
- Higher is better
- Unbiased Hit Rate, Fear%
- Wagner's unbiased hit rate for fear, computed the same way.
- Higher is better
- Unbiased Hit Rate, Happiness%
- Wagner's unbiased hit rate for happiness, computed the same way.
- Higher is better
- Unbiased Hit Rate, Sadness%
- Wagner's unbiased hit rate for sadness, computed the same way.
- Higher is better
- Unbiased Hit Rate, No Emotion%
- Wagner's unbiased hit rate for no emotion, computed the same way.
- Higher is better
- Confusion Matrixcount
- Answers for each emotion acted and answer chosen, keyed acted:chosen, for example fear:sad.
- Informational
- Median Correct RTms
- Median time from the end of the sentence, when the answers turn on, to a right answer, over answers at or after 100 ms.
- Context for the scores
- Quick Answerscount
- Answers under 100 ms after the answers turned on. They still count, since the answer can be chosen while the sentence plays.
- Half or more invalidates the session
- Answered Trialstrials
- Test trials answered.
- Fewer than 24 invalidates the session
- Planned Trialstrials
- Test trials the session set out to play: 48.
- Informational
- Longest Run of One Answertrials
- Most consecutive answers with the same emotion.
- 8 or more is flagged for review
- Session in Another Language0/1
- 1 when the session ran in a language other than English, the language of the sentences.
- 1 is flagged for review
- Reported Output Delayms
- The sound output delay the browser reported for the device.
- Context for the scores
Use and evidence
120 papers from 2021 to 2025 name this task in their title or abstract, up 35% compared with 2016 to 2020. That ranks it 196 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 47 |
| 2024 | 20 |
| 2023 | 16 |
| 2022 | 20 |
| 2021 | 17 |
| 2020 | 24 |
| 2019 | 20 |
| 2018 | 18 |
| 2017 | 13 |
| 2016 | 14 |
| 2015 | 15 |
| 2014 | 16 |
| 2013 | 15 |
| 2012 | 15 |
| 2011 | 11 |
| 2010 | 8 |
| 2009 | 9 |
| 2008 | 14 |
| 2007 | 4 |
| 2006 | 4 |
| 2005 | 2 |
| 2004 | 1 |
| 2003 | 0 |
| 2002 | 2 |
| 2001 | 0 |
| 2000 | 0 |
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
- 1
- Papers on reliability or validity
- 18
- Registered clinical studies using it as an outcome
- 6
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 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Recordings
- 54 sentences spoken by 39 actors, from CREMA-D (Cao and colleagues, 2014): 9 for each emotion, chosen by a stated rule from the set's voice-only ratings, as many by women as by men, all played at one speech level. The sentences are English and neutral in meaning.
- Answers
- Six buttons in a fixed order: anger, disgust, fear, happiness, sadness and none. They turn on when the sentence ends. No time limit.
- Trials
- 48: each emotion 8 times, each recording once.
- Order
- Drawn from the session's seed, with no emotion more than 2 times in a row and never one actor twice in a row.
- Sound
- Play starts each sentence, heard once, through speakers or headphones. A steady tone sets the volume first.
- Practice
- 6 recordings, one per emotion, each followed by the answer for 2,000 ms. The test gives no feedback.
The voices are recordings of actors from CREMA-D, used under its Open Database License, and the emotions are acted. The test plays the clips its raters agreed on most. The sentences are English; sessions in other languages are flagged for review. Scores support research and do not diagnose.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: vocal_emotion_test from the scored run, one of six answers for each recording (chance 1 in 6; each emotion is acted on 8 of the 48 trials), with no time limit, graded from the button pressed and the emotion acted. Below chance invalidates and not above chance is flagged; one answer on 90% of trials invalidates when accuracy is not above chance. Answers under 100 ms from the buttons turning on still count, and at least half the answers that fast invalidate. Fewer than 24 answered test trials (four per emotion) invalidate; fewer than planned withhold the percentile. The same answer on 8 or more trials in a row, or a session run in a language other than English, 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). The answers turn on when the sentence ends, and an answer can be chosen while it plays, so a quick answer is not a guess; only a session where at least half the answers came under 100 ms is invalid. The order never plays one emotion on more than two trials in a row, so eight answers in a row with one label include at least two errors and suggest answering without listening. The scores for each emotion rest on that emotion's trials, so a session with fewer than four per emotion is invalid. The sentences are English: listeners tend to recognize emotion in voices of their own language a little better than in others (Scherer, Banse and Wallbott, 2001), so a session in another language is flagged and its norms are kept apart by language.
Running this test
Settings you can change
This test runs with fixed settings.
Need a different version of this test?
Devices and languages
- Untimed test: runs on desktops, laptops, tablets and phones. Participant devices
- Instructions and screens: English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. Words and sentences: English.
- 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 Vocal Emotion Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Banse & Scherer, 1996; Wagner, 1993; Cao et al., 2014). The main score was percent correct (percent of answered test trials given the emotion the actor portrayed).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Banse, R., & Scherer, K. R. (1996). Acoustic profiles in vocal emotion expression. Journal of Personality and Social Psychology, 70(3), 614-636. https://doi.org/10.1037/0022-3514.70.3.614
- Wagner, H. L. (1993). On measuring performance in category judgment studies of nonverbal behavior. Journal of Nonverbal Behavior, 17(1), 3-28. https://doi.org/10.1007/BF00987006The unbiased hit rate.
- Cao, H., Cooper, D. G., Keutmann, M. K., Gur, R. C., Nenkova, A., & Verma, R. (2014). CREMA-D: Crowd-sourced emotional multimodal actors dataset. IEEE Transactions on Affective Computing, 5(4), 377-390. https://doi.org/10.1109/TAFFC.2014.2336244The recordings and their ratings.
- Scherer, K. R., Banse, R., & Wallbott, H. G. (2001). Emotion inferences from vocal expression correlate across languages and cultures. Journal of Cross-Cultural Psychology, 32(1), 76-92. https://doi.org/10.1177/0022022101032001009Recognition of vocal emotion across languages.
- Laukka, P., & Elfenbein, H. A. (2021). Cross-cultural emotion recognition and in-group advantage in vocal expression: A meta-analysis. Emotion Review, 13(1), 3-11. https://doi.org/10.1177/1754073919897295Higher accuracy within than across cultures.
- Hoekert, M., Kahn, R. S., Pijnenborg, M., & Aleman, A. (2007). Impaired recognition and expression of emotional prosody in schizophrenia: Review and meta-analysis. Schizophrenia Research, 96(1-3), 135-145. https://doi.org/10.1016/j.schres.2007.07.023Use in schizophrenia research.
- Leung, F. Y. N., Sin, J., Dawson, C., Ong, J. H., Zhao, C., Veić, A., & Liu, F. (2022). Emotion recognition across visual and auditory modalities in autism spectrum disorder: A systematic review and meta-analysis. Developmental Review, 63, 101000. https://doi.org/10.1016/j.dr.2021.101000Use in autism research.
Source papers
- Banse R, Scherer KR (1996). Acoustic profiles in vocal emotion expression. Journal of Personality and Social Psychology, 70(3):614-636.Paradigm source, Adults
- Wagner HL (1993). On measuring performance in category judgment studies of nonverbal behavior. Journal of Nonverbal Behavior, 17(1):3-28.Paradigm source, Adults
- Cao H, Cooper DG, Keutmann MK, Gur RC, Nenkova A, Verma R (2014). CREMA-D: Crowd-sourced emotional multimodal actors dataset. IEEE Transactions on Affective Computing, 5(4):377-390.Paradigm source, Adults
- Scherer KR, Banse R, Wallbott HG (2001). Emotion inferences from vocal expression correlate across languages and cultures. Journal of Cross-Cultural Psychology, 32(1):76-92.Paradigm source, Adults
- Laukka P, Elfenbein HA (2021). Cross-cultural emotion recognition and in-group advantage in vocal expression: A meta-analysis. Emotion Review, 13(1):3-11.Paradigm source, Adults
Questions
How long does the Vocal Emotion Task take?
About 6 to 8 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?
Percent Correct, Mean Unbiased Hit Rate, Percent Correct, Anger, Percent Correct, Disgust, Percent Correct, Fear, Percent Correct, Happiness, Percent Correct, Sadness, Percent Correct, No Emotion, Unbiased Hit Rate, Anger, Unbiased Hit Rate, Disgust, Unbiased Hit Rate, Fear, Unbiased Hit Rate, Happiness, Unbiased Hit Rate, Sadness, Unbiased Hit Rate, No Emotion, Confusion Matrix, Median Correct RT, Quick Answers, Answered Trials, Planned Trials, Longest Run of One Answer, Session in Another Language and Reported Output Delay. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Vocal Emotion Task used in research?
120 papers from 2021 to 2025 name it in their title or abstract, up 35% compared with 2016 to 2020. That ranks it 196 of 495 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.22.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Vocal Emotion Task have?
In the published research, 1 paper reports its test-retest reliability and 6 registered clinical studies use it as an outcome (evidence grade C in the ConductCognition test census).
Which languages can participants use?
The instructions and screens are in English, Spanish, French, German, Italian, Portuguese (Brazil), Chinese (Simplified) and Japanese. The words and sentences stay in English, so sessions in other languages are flagged for review. 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.