Attention
Voice Identity Task
Also called: voice discrimination task, speaker identity task, learned voice recognition task
Measure how accurately people recognize newly learned voices and distinguish the same voice from a different one.
5-8 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
- 67
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how accurately people recognize newly learned voices and tell two voices apart. Participants learn two synthetic voices, recognize them in new sentences among two other voices, then judge whether pairs of sentences have the same or different speakers. Uses speech in the session language over headphones. Results describe these synthetic voices.
Constructs
- Voice identity discrimination
- Voice recognition
- Auditory identity memory
Research uses
- Auditory perception research
- Voice learning research
- Within-study comparisons
How the paradigm works
Measure voice discrimination and recognition in two parts. The first asks whether a voice belongs to the two identities learned earlier; the second asks whether two different sentences have the same or different voices. The design draws on listener judgments in voice identity research.
More about the paradigm
Participants learn two synthetic voices through two sentences each, then answer 16 recognition questions using four new sentences from four voices. Half the trials use learned voices, half use voices absent from learning. New means absent from the learning part, even when that voice appears again during recognition. Recognition comes before pair comparisons so comparison clips cannot expose the new identities first.
The pair part has 16 trials, or 32 with two repetitions. Same and different pairs occur equally often, with both sentence orders and every first identity balanced. Different pairs stay within the two lower-pitched or two higher-pitched voice variants. Every pair uses two different sentences and a 500 ms gap. Four practice choices use separate voices and give feedback; measured trials do not.
Scores
- Voice Pair Accuracy%
- Share of usable same/different judgments answered correctly.
- Higher means more pairs distinguished correctly
- Voice Recognition Accuracy%
- Share of usable studied/new voice judgments answered correctly.
- Higher means more studied voices recognized correctly
- Voice Pair Sensitivityd'
- Signal sensitivity with 0.5 added to each contingency-table cell; null if a signal class is missing.
- Higher means better separation
- Voice Recognition Sensitivityd'
- Signal sensitivity with 0.5 added to each contingency-table cell; null if a signal class is missing.
- Higher means better separation
- Voice Pair Criterionscore
- Response criterion with log-linear correction. Positive values favor Same.
- Describes answer bias
- Voice Recognition Criterionscore
- Response criterion with log-linear correction. Positive values favor New voice.
- Describes answer bias
- Voice Pair Hit Rate%
- Share of signal trials correctly identified: different voices.
- Informational
- Voice Pair False Alarm Rate%
- Share of nonsignal trials incorrectly answered as different voices.
- Informational
- Voice Recognition Hit Rate%
- Share of signal trials correctly identified: a studied voice.
- Informational
- Voice Recognition False Alarm Rate%
- Share of nonsignal trials incorrectly answered as a studied voice.
- Informational
- Usable Voice Pairstrials
- Answered pair trials, excluding anticipations.
- Informational
- Usable Voice Questionstrials
- Answered recognition trials, excluding anticipations.
- Informational
- Median Answer Timems
- Median answer time after the sound ends, excluding anticipations.
- Informational; no time limit
- Anticipationscount
- Answers under 100 ms after the answer buttons turn on.
- Excluded from accuracy and sensitivity
- Completed Trialstrials
- Answered measured trials across both parts in the selected run.
- Fewer than planned is flagged
- Planned Trialstrials
- Recognition trials plus planned voice-pair trials.
- Informational
- Headphone Check Passed0/1
- 1 if the six-trial check passed, 0 if it failed.
- A failed check is flagged
- Sound Output Delayms
- Output delay reported by the browser, when available.
- Informational
Use and evidence
67 papers from 2021 to 2025 name this task in their title or abstract, down 7% compared with 2016 to 2020. That ranks it 316 of 582 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 20 |
| 2024 | 16 |
| 2023 | 6 |
| 2022 | 12 |
| 2021 | 13 |
| 2020 | 16 |
| 2019 | 12 |
| 2018 | 19 |
| 2017 | 13 |
| 2016 | 12 |
| 2015 | 14 |
| 2014 | 12 |
| 2013 | 10 |
| 2012 | 7 |
| 2011 | 4 |
| 2010 | 10 |
| 2009 | 7 |
| 2008 | 7 |
| 2007 | 4 |
| 2006 | 3 |
| 2005 | 2 |
| 2004 | 2 |
| 2003 | 3 |
| 2002 | 1 |
| 2001 | 0 |
| 2000 | 0 |
Use grade D
30 to 99 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 1
- Papers on reliability or validity
- 12
- 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.
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.
- Learning
- Two of four main voices are selected from the session seed, one from each lower- or higher-pitched pair. Each speaks two learning sentences: four presentations.
- Recognition
- 16 trials: four voices each say four new sentences. Eight trials use learned voices and eight use voices absent from learning. The words are matched across voices.
- Pair comparisons
- 16 pairs, or 32 with two repetitions. Half same and half different. Every first voice and both sentence orders are balanced. Different identities stay within their lower- or higher-pitched pair.
- Order
- Learning and recognition come before pair comparisons. The seed sets the learned identities and trial orders. New means not in the learning part, even on a repeated recognition presentation.
- Playback
- Press Play once. A pair contains two different sentences with a 500 ms gap. The answers turn on after playback; choices are untimed. Speech is 12 dB below the comfortable-volume tone.
- Voices and languages
- Six eSpeak NG 1.52.0 formant identities in each of eight languages: m1, m2, f1 and f2 for measured trials; m3 and f3 for practice. Original sentences; no recorded or imitated person.
- Practice
- Four practice choices with feedback: two recognition questions and two voice pairs. Practice voices and sentences are separate from the measured parts. No measured-trial feedback.
- Headphones
- Needs headphones and normal hearing. The volume step and six-trial headphone check come first. By default a failed check stops the task; a study can continue with the failure flagged.
- Scores
- Separate accuracy, hit rate, false alarm rate, sensitivity (d′) and response criterion for each part. A 0.5 correction is added to all four cells for d′ and criterion. Answers under 100 ms are anticipations and leave the scores.
These measures describe discrimination and recognition of the task’s synthetic voices. Compare results within the same speech language and setting. No percentile is given. Answers under 100 ms leave the score, and a session with at least half its answers that fast is invalid.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: voice_identity_test from the run with the most test rows, latest on ties. Each part has two balanced choices (chance 50%). Accuracy and signal detection are scored separately. Accuracy significantly below chance across the measured answers (one-sided exact p < .05) invalidates; not above chance is flagged. Answers under 100 ms leave the score; at least half that fast invalidates. Incomplete runs and failed headphone checks are flagged. A stopped headphone check leaves no task score.
Why these rules
Recognition tests the studied synthetic identities in unheard sentences. Pair judgments compare different sentences from the same or different synthetic identities. These two abilities are reported separately; no percentile is given.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Pair comparison length | 16 pairs | 16 pairs, 32 pairs |
| When the headphone check fails | Stop the test | Stop the test, Continue the test |
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. Needs headphones. 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 Voice Identity Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Lavan et al., 2021; Zäske, Awwad Shiekh Hasan and Belin, 2017), with the standard settings (pair comparison length: 16 pairs; when the headphone check fails: Stop the test). The main score was voice pair accuracy (share of usable same/different judgments answered correctly).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Lavan, N., Kreitewolf, J., Obleser, J., & McGettigan, C. (2021). Familiarity and task context shape the use of acoustic information in voice identity perception. Cognition, 215, 104780. https://doi.org/10.1016/j.cognition.2021.104780Same/different judgments on sequential voice recordings. This task uses newly generated formant voices and original sentences.
- Zäske, R., Awwad Shiekh Hasan, B., & Belin, P. (2017). It doesn’t matter what you say: FMRI correlates of voice learning and recognition independent of speech content. Cortex, 94, 100–112. https://doi.org/10.1016/j.cortex.2017.06.005Recognition of learned voices among new voices using sentences not heard during learning. Our shorter task uses synthetic voices, not the source recordings.
Source papers
- Lavan, N., Kreitewolf, J., Obleser, J., & McGettigan, C. (2021). Familiarity and task context shape the use of acoustic information in voice identity perception. Cognition, 215, 104780.N 177, Adults
Questions
How long does the Voice Identity Task take?
About 5 to 8 minutes, as listed in the test library.
What device does a participant need?
Untimed test: runs on desktops, laptops, tablets and phones. Needs headphones. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Voice Pair Accuracy, Voice Recognition Accuracy, Voice Pair Sensitivity, Voice Recognition Sensitivity, Voice Pair Criterion, Voice Recognition Criterion, Voice Pair Hit Rate, Voice Pair False Alarm Rate, Voice Recognition Hit Rate, Voice Recognition False Alarm Rate, Usable Voice Pairs, Usable Voice Questions, Median Answer Time, Anticipations, Completed Trials, Planned Trials, Headphone Check Passed and Sound Output Delay. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Voice Identity Task used in research?
67 papers from 2021 to 2025 name it in their title or abstract, down 7% compared with 2016 to 2020. That ranks it 316 of 582 test families in the ConductCognition test census (use grade D). Source: ConductCognition test census 0.39.0, counted 8 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Voice Identity Task have?
In the published research, 1 paper reports its test-retest reliability and 0 registered clinical studies use 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?
Pair comparison length (16 pairs, 32 pairs) and when the headphone check fails (Stop the test, Continue the test). 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, with its words in the same language. 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.