Social Cognition
Word-Sentence Association Task
Also called: word sentence association paradigm, WSAP, interpretation bias task, word-sentence relatedness task
Measure how people read ambiguous everyday situations: a word suggesting a threatening or a harmless reading appears, then a sentence, and participants say whether they are related.
6-9 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
- 21
- Evidence grade
- C
- Registered clinical studies
- 18
Your message has been read but not answered.
What it measures
Measures how people read ambiguous everyday situations in research. A word that suggests a threatening or a harmless reading appears, then a short sentence, and the participant says whether the two are related. The words and sentences are in English.
Constructs
- Interpretation bias
- Resolving ambiguity
- Emotional processing
Research uses
- Interpretation bias research at the group level
- Studies of interpretation training, compared within a study
- Group comparisons in clinical research, without diagnostic use
How the paradigm works
The Word-Sentence Association Task measures how people read ambiguous everyday situations, an interpretation bias measured in research. Beard and Amir (2009) introduced it: a word that suggests a threatening or a harmless reading appears before an ambiguous sentence, and the participant decides whether the word and the sentence are related. For example, "ignored" or "driving" appears before "Your message has been read but not answered."
More about the paradigm
On each trial a cross appears for 500 ms, then the word for 500 ms, then the sentence, which stays until the participant has read it and presses the space bar or Next. The participant then answers whether the word and the sentence are related, with Related or Not related (the F or J key on a keyboard), with no time limit and no feedback. The 32 sentences were written for this test, 26 about other people's reactions and 6 about everyday mishaps, each with one word that suggests a threatening reading and one that suggests a harmless reading. Each sentence appears once with each word, in two halves with a short break between them. Eight more trials pair a neutral sentence with an unrelated word (on by default), which shows whether answers depended on the sentences, a check Beard and Amir proposed. Two practice trials come first. The session's seed sets which word each sentence shows first and the order, with no more than three trials of one kind in a row. No published item is used, and the words and sentences are English.
The main score, interpretation bias, is the share of threatening words called related minus the share of harmless words called related, from -1 to 1. Higher values mean threatening readings were endorsed more, relative to harmless ones. The task also reports both shares, the source's two response-time bias scores (the time to reject threatening words minus the time to endorse them, and the time to endorse harmless words minus the time to reject them), the four median decision times behind them, the share of unrelated pairs called related and the median reading time. In the source study the two shares were not strongly correlated (r = -.25), so both are reported beside their difference. These are research measures of how ambiguous situations were read in a session. They do not diagnose and do not measure anxiety or any other condition. Sessions run in a language other than English are flagged for review.
Scores
- Interpretation Bias%
- Share of threatening words called related minus the share of harmless words called related, from -1 to 1.
- Higher means threatening readings were endorsed more, relative to harmless ones
- Threat Endorsement%
- Share of the threatening words called related to their sentence.
- Higher means more threatening readings were endorsed
- Harmless Endorsement%
- Share of the harmless words called related to their sentence.
- Higher means more harmless readings were endorsed
- Threat Response-Time Biasms
- Median time to reject threatening words minus the median time to endorse them. Needs at least 3 answers of each kind.
- Higher means threatening readings were accepted more easily than rejected
- Harmless Response-Time Biasms
- Median time to endorse harmless words minus the median time to reject them. Needs at least 3 answers of each kind.
- Higher means harmless readings were accepted less easily than rejected
- Endorsing Threat, Median Timems
- Median time from the question to a Related answer for threatening words, from answers at or after 100 ms.
- Informational
- Rejecting Threat, Median Timems
- Median time from the question to a Not related answer for threatening words.
- Informational
- Endorsing Harmless, Median Timems
- Median time from the question to a Related answer for harmless words.
- Informational
- Rejecting Harmless, Median Timems
- Median time from the question to a Not related answer for harmless words.
- Informational
- Unrelated Pairs Called Related%
- Share of the 8 unrelated check pairs called related. Empty with the checks off.
- Lower supports interpretation
- Median Reading Timems
- Median time from the sentence to the space bar or Next.
- Informational
- Answered Wordsitems
- Threatening and harmless words answered in the scored run, out of 64.
- Quality check
- Anticipationscount
- Answers within 100 ms of the question.
- Quality check
Use and evidence
21 papers from 2021 to 2025 name this task in their title or abstract, up 50% compared with 2016 to 2020. That ranks it 398 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 8 |
| 2024 | 5 |
| 2023 | 3 |
| 2022 | 1 |
| 2021 | 4 |
| 2020 | 1 |
| 2019 | 6 |
| 2018 | 3 |
| 2017 | 0 |
| 2016 | 4 |
| 2015 | 4 |
| 2014 | 3 |
| 2013 | 1 |
| 2012 | 2 |
| 2011 | 1 |
| 2010 | 0 |
| 2009 | 1 |
| 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 C
At least 2 test-retest papers, 30 reliability or validity papers or 3 registered clinical studies.
- Papers on test-retest reliability
- 2
- Papers on reliability or validity
- 14
- Registered clinical studies using it as an outcome
- 18
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.
- Items
- 32 ambiguous sentences written for this test, 26 about other people's reactions and 6 about everyday mishaps, each with one word that suggests a threatening reading and one that suggests a harmless reading. The words and sentences are English.
- Trial
- A cross for 500 ms, the word for 500 ms, then the sentence until the space bar or Next; then the question Are the word and the sentence related?, answered with Related or Not related (F or J on a keyboard), with no time limit and no feedback; then a 500 ms blank screen.
- Session
- Each sentence once with each word, in two halves with a short break between them. The seed draws which word each sentence shows first and the order, with no more than 3 trials of one kind in a row, and all of it is saved.
- Unrelated checks
- 8 trials that pair a neutral sentence with an unrelated word, 4 in each half. On by default; off runs the 64 trials of the source's design.
- Practice
- 2 trials with neutral sentences, each followed by a note that any answer is fine.
The sentences were written for this test; the source's 110 sentences are not published, and none of its examples is used. The response times include moving a pointer or a finger on devices without a keyboard, so compare them within one kind of device.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: word_sentence_association_trial from the scored run, one answer per word and sentence, with no correct answer. Fewer than 48 of the 64 threatening and harmless words answered, or none of one kind, invalidate. Calling at least half of the unrelated check pairs related, the same answer on every threatening and harmless word, or a session run in another language is flagged for review. Answers under 200 ms on at least half the trials invalidate. A hidden tab withholds the percentile.
Why these rules
There is no correct option to compare with chance, so accuracy is not judged against chance. Whether a word fits an ambiguous sentence is the participant's reading, so no answer is wrong and repeated answers are not judged against chance; they only flag the session when every word got the same answer, because the bias then rests on no difference. The score compares two shares, and with fewer than three quarters of the words answered one answer moves it too far. Unrelated pairs called related suggest answers that did not depend on the sentences, the check Beard and Amir (2009) proposed for the task. 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. The word lists are English. A session run in another language shows English words to a participant who may not read English fluently, so it is flagged for review. A hidden tab lets the study, delay or response timers run without the participant, so a percentile is withheld after one.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Unrelated check pairs | On | On or off |
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
- 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.
- 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 Word-Sentence Association Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Beard C, Amir N, 2009), with the standard settings (unrelated check pairs: On). The main score was interpretation bias (share of threatening words called related minus the share of harmless words called related, from -1 to 1).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Beard, C., & Amir, N. (2009). Interpretation in social anxiety: When meaning precedes ambiguity. Cognitive Therapy and Research, 33(4), 406-415. https://doi.org/10.1007/s10608-009-9235-0
Source papers
- Beard C, Amir N (2009). Interpretation in social anxiety: When meaning precedes ambiguity. Cognitive Therapy and Research, 33(4):406-415.N 52, Adults
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Word-Sentence Association Task take?
About 6 to 9 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?
Interpretation Bias, Threat Endorsement, Harmless Endorsement, Threat Response-Time Bias, Harmless Response-Time Bias, Endorsing Threat, Median Time, Rejecting Threat, Median Time, Endorsing Harmless, Median Time, Rejecting Harmless, Median Time, Unrelated Pairs Called Related, Median Reading Time, Answered Words and Anticipations. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Word-Sentence Association Task used in research?
21 papers from 2021 to 2025 name it in their title or abstract, up 50% compared with 2016 to 2020. That ranks it 398 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 Word-Sentence Association Task have?
In the published research, 2 papers report its test-retest reliability and 18 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?
Unrelated check pairs (On or off). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings.
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.