Attention
Speech-to-Song Task
Also called: speech-to-song illusion, speech to song illusion, speech-to-song transformation, speech-song illusion, repetition and music perception
Measure how repetition turns speech into song: a short phrase from a synthetic voice, rated once, then rated again after it repeats.
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
- 27
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how much repetition makes speech sound like song. A synthetic voice says a short everyday phrase, and participants rate it from exactly like speech to exactly like singing, then hear the same phrase repeat and rate it again. The main score is how much the rating rises.
Constructs
- Perception of speech and song
- Effects of repetition on listening
- Pitch perception in speech
Research uses
- Music and speech perception research
- Cross-language listening research
- Within-study comparison of listening experience
How the paradigm works
The Speech-to-Song Task measures how much repetition makes speech sound like song. Deutsch, Henthorn and Lapidis (2011) found that a spoken phrase heard over and over comes to sound sung: listeners rated it more like singing after ten repeats than after one, but not when the repeats were moved in pitch or had their syllables put in another order, so the change needs the same recording each time. This version plays short everyday phrases written for the platform, said by a synthetic voice in the participant's language. It needs normal hearing, and headphones or the device's speakers both work. The test starts with the sound setup: the phrases load and the participant turns the device's volume until a steady tone is comfortable.
More about the paradigm
On each trial the participant presses Play, hears a phrase once, and rates how it sounded on a scale from 1, exactly like speech, to 5, exactly like singing, the scale and anchors of Deutsch and colleagues (2011) and Margulis, Simchy-Gross and Black (2015). Then the same recording plays eight times in a row with 400 ms of silence between the plays, as recent online studies played them (Rathcke and Smit, 2025; Sadakata and colleagues, 2026), and the participant rates it again. One practice phrase comes first, then nine phrases in an order drawn from the session's seed, 18 ratings. The ratings have no time limit and no right answer, and nothing adapts to them. Each language has ten phrases, the same ten meanings, spoken by that language's open text-to-speech voice and set to one speech level; every phrase plays 12 dB below the volume-setting tone.
The main score is the rating change: the mean over the nine phrases of the rating after the repeats less the rating after one play. The mean of each rating and the share of phrases rated higher after the repeats are reported beside it. Not every phrase changes, and not every listener hears the change. Phrases with steadier pitch within syllables and more melodic pitch movement changed more in Tierney, Patel and Breen (2018), and listeners heard more change in languages they did not know than in their own in Margulis and colleagues (2015). For speakers of tone languages the findings are mixed: less change in Jaisin and colleagues (2016), no reliable difference in Kachlicka and colleagues (2024), and a small but reliable change in Mandarin listeners hearing Mandarin in Rathcke and Canzi (2026). So results are compared within a language. The strength of the change has been linked to musical aptitude more than to musical training (Tierney and colleagues, 2021). The voices are synthetic: the published studies used recorded speech, and how strongly these phrases turn into song has not been measured. The test does not measure musical ability or hearing: scores are research measures and do not diagnose.
Scores
- Rating Changescore
- Mean over the nine phrases of the rating after the repeats less the rating after one play, on the scale from 1, exactly like speech, to 5, exactly like singing.
- Positive means the phrases sounded more like singing after the repeats; neither direction is better
- Rating After One Playscore
- Mean rating after each phrase played once.
- Informational
- Rating After the Repeatsscore
- Mean rating after each phrase's repeats.
- Informational
- Phrases Rated Higher%
- Share of the phrases rated more like singing after the repeats than after one play.
- Informational
- Median Rating Timems
- Median time from when the ratings opened, after the sound, to the rating.
- Informational; the rating has no time limit
- Anticipationscount
- Ratings within 100 ms of the ratings opening, which leave the scores with their phrase.
- Half or more of the ratings makes the session invalid
- Completed Ratingstrials
- Ratings given in the scored run, two per phrase.
- Fewer than planned is flagged for review
- Planned Ratingstrials
- Ratings the run set out to collect.
- Informational
- Reported Output Delayms
- Output delay the device reported to the browser; empty when the browser does not report it.
- Informational
Use and evidence
27 papers from 2021 to 2025 name this task in their title or abstract, up 59% compared with 2016 to 2020. That ranks it 375 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 7 |
| 2024 | 10 |
| 2023 | 4 |
| 2022 | 1 |
| 2021 | 5 |
| 2020 | 3 |
| 2019 | 4 |
| 2018 | 6 |
| 2017 | 2 |
| 2016 | 2 |
| 2015 | 3 |
| 2014 | 1 |
| 2013 | 0 |
| 2012 | 0 |
| 2011 | 1 |
| 2010 | 1 |
| 2009 | 0 |
| 2008 | 1 |
| 2007 | 0 |
| 2006 | 0 |
| 2005 | 0 |
| 2004 | 0 |
| 2003 | 0 |
| 2002 | 0 |
| 2001 | 0 |
| 2000 | 0 |
Use grade E
Fewer than 30 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 1
- Registered clinical studies using it as an outcome
- 0
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.
- Sound setup
- The phrases load and the participant turns the device's volume until a steady 1 kHz tone is comfortable. Headphones or the device's speakers both work, so there is no headphone check.
- Phrases
- Ten short everyday statements per language, the same ten meanings, written for this test and said by a synthetic voice in the participant's language, not recorded from a person; 1.0 to 2.1 s each.
- Trial
- Press Play to hear a phrase once and rate it from 1, exactly like speech, to 5, exactly like singing. Then press Play to hear the same recording 8 times, with 400 ms of silence between the plays, and rate it again.
- Order
- 1 practice phrase, then the other 9 in an order drawn from the session's seed: 18 ratings, with no time limit and no right answer.
- Level
- Every phrase 12 dB below the volume-setting tone, a level relative to the volume the participant set.
- Score
- The rating change: the mean over the 9 phrases of the rating after the repeats less the rating after one play, with the mean of each rating and the share of phrases rated higher after the repeats. Ratings under 100 ms leave the scores.
The voices are synthetic: the published studies used recorded speech, and how strongly these phrases turn into song has not been measured. Phrases differ in how much they change, and listeners hear less change in their own language than in one they do not know, so compare results within a language and within studies run the same way. The test needs normal hearing, and it does not measure musical ability or hearing. The 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: speech_to_song_test from the scored run, two ratings per phrase from 1, exactly like speech, to 5, exactly like singing, each after the sound has played, with no time limit. A rating has no correct option, so accuracy is not judged against chance. The same rating on every trial is flagged for review; ratings under 100 ms invalidate when they are at least half; a run short of its planned ratings is flagged for review.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Some listeners hear no change, so one rating throughout can be a real result and is only flagged. The ratings open once the sound has played, so a rating under 100 ms was pressed before they opened, and only a session made mostly of such ratings is invalid.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Repeats | 8 | 4 to 10 |
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 Speech-to-Song Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Deutsch, Henthorn and Lapidis, 2011; Tierney, Patel and Breen, 2018), with the standard settings (repeats: 8). The main score was rating change (mean over the nine phrases of the rating after the repeats less the rating after one play, on the scale from 1, exactly like speech, to 5, exactly like singing).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Deutsch, D., Henthorn, T., & Lapidis, R. (2011). Illusory transformation from speech to song. The Journal of the Acoustical Society of America, 129(4), 2245-2252. https://doi.org/10.1121/1.3562174The illusion, the rating scale from exactly like speech to exactly like singing, and the need for exact repeats.
- Margulis, E. H., Simchy-Gross, R., & Black, J. L. (2015). Pronunciation difficulty, temporal regularity, and the speech-to-song illusion. Frontiers in Psychology, 6, 48. https://doi.org/10.3389/fpsyg.2015.00048Ratings before and after repeats on a 1 to 5 scale, and more change in languages listeners do not know.
- Tierney, A., Patel, A. D., & Breen, M. (2018). Acoustic foundations of the speech-to-song illusion. Journal of Experimental Psychology: General, 147(6), 888-904. https://doi.org/10.1037/xge0000455Which phrases change: steadier pitch within syllables, melodic structure and a regular beat.
- Tierney, A., Patel, A. D., Jasmin, K., & Breen, M. (2021). Individual differences in perception of the speech-to-song illusion are linked to musical aptitude but not musical training. Journal of Experimental Psychology: Human Perception and Performance, 47(12), 1681-1697. https://doi.org/10.1037/xhp0000968Differences between listeners in how strongly they hear the change.
- Jaisin, K., Suphanchaimat, R., Figueroa Candia, M. A., & Warren, J. D. (2016). The speech-to-song illusion is reduced in speakers of tonal (vs. non-tonal) languages. Frontiers in Psychology, 7, 662. https://doi.org/10.3389/fpsyg.2016.00662Less change in speakers of tone languages, in a small sample.
- Kachlicka, M., Patel, A. D., Liu, F., & Tierney, A. (2024). Weighting of cues to categorization of song versus speech in tone-language and non-tone-language speakers. Cognition, 246, 105757. https://doi.org/10.1016/j.cognition.2024.105757No reliable difference in the illusion between tone-language and other listeners.
- Rathcke, T. V., & Canzi, M. (2026). Individual differences in musical melody perception moderate the speech-to-song illusion in Mandarin Chinese listeners. Scientific Reports, 16, 10523. https://doi.org/10.1038/s41598-026-44268-zThe change in Mandarin listeners hearing Mandarin, with eight repeats 400 ms apart.
- Rathcke, T. V., & Smit, E. A. (2025). Musical aptitude moderates ease and vividness but not frequency of the speech-to-song illusion. Scientific Reports, 15, 33315. https://doi.org/10.1038/s41598-025-18612-8Eight repeats 400 ms apart, online.
- Sadakata, M., Nobbe Smyth, M., van 't Veer, M., & Tanaka, A. (2026). Pitch accent speakers exhibit speech-to-song transformation accompanied by reduced musicality ratings: A cross-linguistic study. Psychological Research, 90(2), 45. https://doi.org/10.1007/s00426-026-02262-0A rating after one play and after eight repeats 400 ms apart, online, with Japanese listeners.
Source papers
- Deutsch D, Henthorn T, Lapidis R (2011). Illusory transformation from speech to song. The Journal of the Acoustical Society of America, 129(4):2245-2252.N 54, Adults
- Tierney A, Patel AD, Breen M (2018). Acoustic foundations of the speech-to-song illusion. Journal of Experimental Psychology: General, 147(6):888-904.N 40, Adults
Questions
How long does the Speech-to-Song 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?
Rating Change, Rating After One Play, Rating After the Repeats, Phrases Rated Higher, Median Rating Time, Anticipations, Completed Ratings, Planned Ratings and Reported Output Delay. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Speech-to-Song Task used in research?
27 papers from 2021 to 2025 name it in their title or abstract, up 59% compared with 2016 to 2020. That ranks it 375 of 495 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Speech-to-Song 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 D in the ConductCognition test census).
Does it report a percentile?
No. Scores and quality flags only.
Which settings can a researcher change?
Repeats (4 to 10). 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.