Social Cognition
Talk Bubbles Game
Also called: picture stories game, speech bubble stories game, stories in pictures game, perspective taking stories game
Measure how well adults notice when someone says something they should not have said: short stories told in pictures, with three people, speech bubbles and no time limit.
10-14 minUntimed
Free to start: 10 free participants per account, then $0.99 per participant for academic research, or any yearly plan. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 110
- Evidence grade
- C
- Registered clinical studies
- 24
Figures for the Social Slips Stories Task, the published test this game is compared with.
How these are countedWhat it measures
Measures how well adults notice when someone in a short story told in pictures says something it would have been better not to say, with no time limit.
Constructs
- Social reasoning
- Perspective taking
- Judging what others know
Research uses
- Research on noticing when someone says something they should not have said
How the paradigm works
The Talk Bubbles Game measures how well adults notice when someone in a short story says something it would have been better not to say. It is a game in the family of social slip tests, in which a person reads a story and says whether anyone spoke without knowing a fact that made their words unwelcome. Each story is told in pictures: three drawn people with their names, 3 or 4 panels, a caption above each picture and a speech bubble in some. The person reads the panels one at a time with Next. Then they answer questions with every panel of the story still in view: did anyone say something awkward, who said it, did that person know the fact, and one question about a detail of the story. A button shows the story again from the first panel.
More about the paradigm
Half of the 12 stories hold such a remark: a newcomer, a person who has never met someone, or a person who missed what happened says something that does not suit someone in the room. The other half are stories in which no one says anything of the kind, though some hold a kind remark, a question, a plan or a gripe about something nobody present is tied to, so a person who says no to every story scores half. A person who says yes to every story scores at most half, and about 17% when they guess the speaker. Two practice stories, one of each kind, come first and are explained afterwards. Nothing is timed, no feedback is shown in the test, and the game runs on a computer, a tablet or a phone. A session takes 10 to 14 minutes.
Scores
- Stories Judged Right%
- The percent of stories judged right: a story with such a remark judged yes with the right speaker, a story without one judged no. Empty below 9 judged stories.
- Higher is better
- Remark Stories Judged Right%
- The percent of the stories with such a remark judged yes with the right speaker.
- Higher is better
- Harmless Stories Judged No%
- The percent of the stories without such a remark judged no.
- Higher is better
- Speaker Did Not Know%
- Of the stories with such a remark that were judged yes, the percent where the person said the speaker did not know the fact.
- Higher is better
- Detail Accuracy%
- The percent of the detail questions answered right, in both kinds of story.
- Below two thirds invalidates
- Stories Shown Againcount
- The stories for which See the story again was used.
- Informational
- Median Judgment Timems
- The median time from the first question appearing to its answer, after the story was read.
- Informational
- Usable Storiescount
- Stories whose judgment is counted. Left out: a story in which the tab was hidden, the window changed size, or a second finger or a palm touched the screen.
- Quality check
- Share of Key Answers%
- The share of the answers made with a key (Enter or Space on a focused button) and not a tap or a click.
- Informational; the comparison of touch and keys reads this share within each look
- Share of Touch Answers%
- The share of the answers made by touch.
- Informational; the comparison of touch and keys reads this share within each look
Use and evidence
110 papers from 2021 to 2025 name this task in their title or abstract, up 9% compared with 2016 to 2020. That ranks it 266 of 711 test families in our census. These counts belong to the Social Slips Stories Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 26 |
| 2024 | 18 |
| 2023 | 26 |
| 2022 | 19 |
| 2021 | 21 |
| 2020 | 30 |
| 2019 | 18 |
| 2018 | 16 |
| 2017 | 18 |
| 2016 | 19 |
| 2015 | 17 |
| 2014 | 21 |
| 2013 | 19 |
| 2012 | 18 |
| 2011 | 11 |
| 2010 | 16 |
| 2009 | 6 |
| 2008 | 3 |
| 2007 | 5 |
| 2006 | 2 |
| 2005 | 1 |
| 2004 | 0 |
| 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
- 6
- Papers on reliability or validity
- 33
- Registered clinical studies using it as an outcome
- 24
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 the Social Slips Stories Task, the published test this game is compared with, not the game's own.
Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Stories
- 12 stories, 6 with a remark that it would have been better not to make and 6 without one, in a fixed order with no more than 2 of a kind in a row. Each story is told in 3 or 4 pictures. Three drawn people, each with a name, are in every picture, with a caption above it and a speech bubble in some. Each person speaks once. The stories are for adults and are written for each language.
- Reading
- The person reads the pictures one at a time and touches Next. The questions follow, with every picture of the story in view in a strip above them. A See the story again button shows the story from the first picture, and the number of stories shown again is kept.
- Questions
- First, did anyone say something awkward, something it would have been better not to say? After a yes, who said it, by touching one of the three people, shown with their names, and did that person know the fact that matters. Then one question about a detail of the story, with 3 choices. The questions follow those of the Social Slips Stories Task.
- Scoring
- A story with such a remark is judged right with a yes and the right speaker, and a story without one with a no, as the Social Slips Stories Task scores its own. Saying no to every story scores 50%; saying yes to every story scores at most 50%, and about 17% when the speaker is a guess.
- Time
- No time limit, and no screen shows a score. A session takes 10 to 14 minutes, with a short rest after the 6th story. It starts with a practice of 2 stories, one of each kind, after each of which the person sees what happened. The test has no feedback.
- Taps
- One rule holds for every button: a tap within 300 ms of the last accepted tap, or of a screen change, changes nothing and is counted, so a double tap never answers the next question. Enter or Space on a focused button works too, unless the key was held down from before the screen opened. A story in which the tab was hidden, the window changed size, or a second finger or a palm touched the screen while it was read or while its first question was answered is left out of the scores.
- Forms
- 2 forms of 12 stories, on one skeleton: the same kinds of story, numbers of pictures, places of the speaker and order. A form changes the stories, the people, their names and the settings. Each person gets the forms in an order drawn once for them, and each session takes the next form.
- Looks
- The game look adds a drawn setting behind the people and a face on each person, the same for every story, and the plain look draws the same people, names, bubbles and captions on a plain panel. The researcher chooses the look for a study; the default is the game look.
- Adults
- For adults aged 18 and over. In older-adult mode words and buttons are 20% larger. Buttons are at least 64 px high (77 px in that mode), each person to touch at least 96 px wide (115 px) and speech bubble text at least 14 px. On a phone of 360 by 640 the pictures of a story stack and the question screen scrolls up and down, never sideways. A screen too short for a story's picture, as on a phone held sideways, shows a line asking the person to turn the phone upright or make the window taller, and the story waits until the screen is tall enough.
- Schedule
- Two visits: a baseline on one form and a retest on the other form 7 to 28 days later, with the forms in each person's own order. A third visit would show a person the stories of their first visit again. Keep each person on one device class at every visit, in a quiet room. English-language studies can add the Social Slips Stories Task, whose stories differ, in either order. The game takes 10 to 14 minutes.
- Checks
- A session is marked invalid when fewer than 9 of the 12 stories were judged, when fewer than two thirds of the detail questions were right (the answers then do not show that the stories were read), when answers under 200 ms make up at least half of the answers, when the judgments fall significantly below chance, or when the run holds two forms. The same answer to at least 90% of the stories is flagged for review.
- Scores
- The share of stories judged right (the main score, empty below 9 judged stories), the stories with the remark and the stories without it apart, how often the person said the speaker did not know, the detail questions answered right, the stories shown again and the median time to judge a story. The share of answers made with a key and by touch is also kept, so the two can be compared. No percentile is computed.
The design follows the story tests of Stone, Baron-Cohen and Knight (1998), in which a speaker says something unwelcome without knowing a fact; the stories, pictures, people, settings and layout are our own, and the stories are written for each language. This game is a new measure for adults: its reliability and norms come from use, the figures published in those studies are not its own, and it computes no percentiles. Its comparison with the Social Slips Stories Task runs in English-language studies for now, since that test's stories are English. This page shows no story of the game and the game has no public demonstration, so a study's participants meet the stories in the study.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: the first answer to each story's judgment question in talk_bubbles_game, yes or no, with half of the stories holding the remark, so chance is 50%. Accuracy significantly below chance invalidates. The same answer on at least 90% of 12 or more stories is flagged for review and never invalidates. Answers under 200 ms on at least half of 4 or more answers invalidate. Fewer than two thirds of the detail questions right invalidates, and so does judging fewer than 9 of the 12 stories. A run of two forms is invalid.
Why these rules
Seeing the remark in every story, or in none, is a pattern this game can show, so the same answer throughout is flagged for review and never invalidates. An answer under 200 ms was too fast to follow a story (ours). A detail question asks a plain fact of the story, with three choices, so a person who did not read the stories is right on about a third of them; below two thirds right, the answers do not show that the stories were read (the Social Slips Stories Task's rule). With fewer than three quarters of the stories judged, the scores rest on too few stories to compare with a full session. A run that holds two forms cannot be read against one table and is not scored. The stories are written for each language, so a session in another language is not flagged for it.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Look | Game | Game, Plain |
Every plan can change the look, which never changes how a result is scored.
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 Talk Bubbles Game in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Stone, Baron-Cohen and Knight, 1998), with the standard settings (look: Game). The main score was stories judged right (the percent of stories judged right: a story with such a remark judged yes with the right speaker, a story without one judged no).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Stone, V. E., Baron-Cohen, S., & Knight, R. T. (1998). Frontal lobe contributions to theory of mind. Journal of Cognitive Neuroscience, 10(5), 640-656. https://doi.org/10.1162/089892998562942The story tests whose design the game follows: a speaker says something unwelcome without knowing a fact (Crossref record only; no figure is taken from it).
- Poos, J. M., Zinzen, I., Kalisvaart, M., Assendelft, L., Marticic Giljevic, K., Ruitenberg, M., Bidarra, R., & van den Berg, E. (2025). A gamified Faux Pas Test: Comparing psychometric properties to the pen-and-paper version in a Dutch population. Journal of Neuropsychology. https://doi.org/10.1111/jnp.70015A game version of a story test of this kind, compared with its paper version in a Dutch sample; this game has its own stories and its own scoring (Crossref record only; no figure is taken from it).
Source papers
Questions
How long does the Talk Bubbles Game take?
About 10 to 14 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?
Stories Judged Right, Remark Stories Judged Right, Harmless Stories Judged No, Speaker Did Not Know, Detail Accuracy, Stories Shown Again, Median Judgment Time, Usable Stories, Share of Key Answers and Share of Touch Answers. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Social Slips Stories Task used in research?
These counts belong to the Social Slips Stories Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. 110 papers from 2021 to 2025 name the task in their title or abstract, up 9% compared with 2016 to 2020. That ranks it 266 of 711 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Social Slips Stories Task have?
These counts belong to the Social Slips Stories Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. In the published research, 6 papers report the task's test-retest reliability and 24 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?
Look (Game, Plain). Every plan can change the look, which never changes how a result is scored.
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.
Related tests