Language
Scalar Implicature Task
Also called: some but not all task, underinformative sentence task, pragmatic inference task, sentence verification of some and all
Measure how often people read "some" as "some but not all": participants judge short English sentences such as "Some ducks have feathers" as true or false, scored by the share read pragmatically, with response times against plain controls.
4-5 minTimed: desktop or laptop with a mouse or trackpad
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
- 250
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how often people read "some" as "some but not all". Participants press F or J to judge 48 short English sentences about everyday facts, such as "Some boots have laces" or "All cows have wings", as true or false. Twelve, such as "Some ducks have feathers", are true in logic but say less than they could; judging them false is the pragmatic reading, and neither reading is wrong. The score is the share read that way, with response times against plainly true and false sentences.
Constructs
- Pragmatic inference
- Scalar implicature
- Sentence comprehension
Research uses
- Experimental pragmatics research
- Development of pragmatic inference
- Within-study group comparison
How the paradigm works
The Scalar Implicature Task measures how often people read "some" as "some but not all". A sentence such as "Some ducks have feathers" is true if "some" means at least one, its logical meaning, and false if it means some but not all, because every duck has feathers: that pragmatic reading is a scalar implicature. Neither reading is wrong. The task follows Noveck (2001), whose third experiment, after Smith (1980), said facts that hold for all of a kind, for some of it or for none of it with "some" or with "all", and Bott and Noveck (2004), who timed true-or-false judgments of such sentences. Noveck's adults judged these underinformative sentences true 41% of the time, against 85% and more for 8- and 10-year-olds. With no guidance on what "some" means, about 60% of answers to them in Bott and Noveck's third experiment were pragmatic, and pragmatic answers were slower than logical ones and than answers to the control sentences.
More about the paradigm
Each session shows 6 practice sentences with feedback, none of them underinformative, then 48 test sentences without feedback: 12 underinformative ("Some ducks have feathers"), 12 universal facts said with "all" ("All zebras have stripes"), 6 partial facts said with "some" ("Some boots have laces") and 6 with "all" ("All towels are blue"), and 6 false facts said with "some" ("Some cows have wings") and 6 with "all". The controls are half true and half false within each quantifier, and no more than three sentences in a row take the same answer under either reading. Each fact appears once in a session; two lists drawn from the session's seed decide which facts are said with "some" and which with "all". A cross shows for 500 ms, then each sentence shows alone for a reading time set before the session (standard 1,000 ms) until the key reminder appears and F or J answers, within 8 seconds. With that one-second wait, Ronderos and Noveck's (2023) web version gave about as many pragmatic as logical answers, the pragmatic ones slower; with no wait it gave about half Bott and Noveck's rate of pragmatic answers. Response times run from the moment the keys work. Which key means true is drawn from the seed. Every sentence was written for this library from facts any adult knows, none from a published item set. The sentences are English in every session; the instructions follow the session's language.
The main score is the pragmatic rate: the share of underinformative sentences judged false, over answers at or after 100 ms. The task also reports the median times of pragmatic and of logical answers, the median times of right answers to true and to false "some" sentences, the slowing of each reading against the control that takes the same key, and accuracy on the 36 controls. An answer to an underinformative sentence is never counted as an error, and the pragmatic rate describes how a person read "some" in this task, not an ability. Scores support research and do not diagnose.
Scores
- Pragmatic Rate%
- Share of underinformative sentences, such as "Some ducks have feathers", judged false, over answers at or after 100 ms.
- Higher means "some" was read as "some but not all" more often; neither reading is wrong
- Pragmatic Answer Timems
- Median time of false answers to underinformative sentences, from the moment the keys work.
- Lower is faster; empty when no answer was pragmatic
- Logical Answer Timems
- Median time of true answers to underinformative sentences, from the moment the keys work.
- Lower is faster; empty when no answer was logical
- True Some Sentence Timems
- Median time of right answers to true "some" sentences about partial facts.
- Lower is faster
- False Some Sentence Timems
- Median time of right answers to false "some" sentences about false facts.
- Lower is faster
- Pragmatic Slowingms
- Pragmatic answer time minus the time of right answers to false "some" sentences; both answers are false.
- Positive values mean the pragmatic reading took longer
- Logical Slowingms
- Logical answer time minus the time of right answers to true "some" sentences; both answers are true.
- Positive values mean the logical reading took longer
- Control Accuracy%
- Right answers as a share of right and wrong ones over the 36 control sentences, whose truth is plain.
- Below 0.7 is invalid
- Underinformative Sentences Answeredcount
- Underinformative sentences answered at or after 100 ms, the pragmatic rate's denominator.
- 12 when every one was answered
- Missed Sentencescount
- Test sentences not answered within 8 seconds of the keys working.
- Expected to be 0
- Anticipationscount
- Answers under 100 ms after the keys started working, counted in no score.
- Expected to be 0
- Session In Another Language0/1
- 1 when the session ran in a language other than English; the sentences are English in every session.
- Flagged for review
Use and evidence
250 papers from 2021 to 2025 name this task in their title or abstract, down 10% compared with 2016 to 2020. That ranks it 113 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 70 |
| 2024 | 52 |
| 2023 | 39 |
| 2022 | 33 |
| 2021 | 56 |
| 2020 | 41 |
| 2019 | 77 |
| 2018 | 61 |
| 2017 | 50 |
| 2016 | 50 |
| 2015 | 45 |
| 2014 | 57 |
| 2013 | 49 |
| 2012 | 38 |
| 2011 | 39 |
| 2010 | 32 |
| 2009 | 34 |
| 2008 | 20 |
| 2007 | 30 |
| 2006 | 26 |
| 2005 | 11 |
| 2004 | 11 |
| 2003 | 7 |
| 2002 | 2 |
| 2001 | 5 |
| 2000 | 1 |
Use grade C
100 to 299 papers from 2021 to 2025.
Evidence grade D
Little published measurement evidence found.
- Papers on test-retest reliability
- 0
- Papers on reliability or validity
- 15
- Registered clinical studies using it as an outcome
- 0
Note on the counts: includes theoretical linguistics papers that run no task.
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.
- Sentences
- Facts written for this test, true of all of a kind, of some of it or of none, each said with "some" or "all": 12 underinformative ("Some ducks have feathers"), 12 true with "all", and 6 of each other kind, so the controls are half true and half false within each quantifier. They are English in every session.
- Lists
- Two lists drawn from the session's seed decide which facts are said with "some" and which with "all"; each fact appears once in a session.
- Trial
- A cross for 500 ms, then the sentence alone for 1,000 ms (a setting from 0 to 2,000 ms), then the key reminder; F or J answers within 8,000 ms. Which key means true is drawn from the seed.
- Session
- 6 practice sentences with feedback, none underinformative, then 48 with none, with no more than 3 of one answer in a row under either reading of "some".
- Scoring
- The pragmatic rate is the share of underinformative sentences judged false. Median times of pragmatic and logical answers are compared with right answers to "some" controls that take the same key. An underinformative sentence is never an error. Under 70% right on the controls is invalid.
The sentences were written for this test, not taken from Noveck's (2001) or Bott and Noveck's (2004) sets. Each sentence shows alone for a second before the keys work, a setting that is 0 in Bott and Noveck's procedure: Ronderos and Noveck (2023) found that this wait brings a web version's results close to theirs. Results describe how often "some" was read as "some but not all", a reading, not an ability.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: scalar_implicature_test from the scored run, F or J for true or false once the keys work, within 8 seconds. The 36 control sentences have a right answer (chance 50%, half true and half false within each quantifier); the 12 underinformative ones have none and are never counted as errors. Below chance on the controls invalidates and not above chance is flagged; one key on 90% of answers invalidates when the controls are not above chance. Missed responses and responses under 100 ms follow the timed-test rule. Under 70% right on the controls invalidates. A session 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). Missed responses and responses under 100 ms keep the platform's timed-test rule: more than max(3, 25%) missed responses or more than max(2, 20%) responses under 100 ms invalidate, and any is flagged. A response under 100 ms cannot be a reaction to the stimulus. An underinformative sentence is true on the logical reading of "some" and false on the pragmatic one, so its answer describes a reading, not accuracy. Ronderos and Noveck (2023) set aside participants under 70% right on the control sentences, whose readings of "some" cannot be told from not reading the sentences. 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. These limits are ConductScience choices.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Reading time before the keys work (ms) | 1,000 | 0 to 2,000 |
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
- Timed test: participants need a desktop or laptop with a mouse or trackpad. 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 Scalar Implicature Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Noveck, 2001; Bott & Noveck, 2004), with the standard settings (reading time before the keys work (ms): 1,000). The main score was pragmatic rate (share of underinformative sentences, such as "Some ducks have feathers", judged false, over answers at or after 100 ms).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Noveck, I. A. (2001). When children are more logical than adults: Experimental investigations of scalar implicature. Cognition, 78(2), 165-188. https://doi.org/10.1016/S0010-0277(00)00114-1The sentence types: universal, partial and absurd facts said with "some" or "all", in two counterbalanced sets; adults judged underinformative sentences true 41% of the time.
- Bott, L., & Noveck, I. A. (2004). Some utterances are underinformative: The onset and time course of scalar inferences. Journal of Memory and Language, 51(3), 437-457. https://doi.org/10.1016/j.jml.2004.05.006Timed true-or-false judgments with no guidance on "some": pragmatic answers slower than logical ones and than the controls.
- Ronderos, C. R., & Noveck, I. (2023). Slowdowns in scalar implicature processing: Isolating the intention-reading costs in the Bott & Noveck task. Cognition, 238, 105480. https://doi.org/10.1016/j.cognition.2023.105480A web version: a one-second wait before answering gave about as many pragmatic as logical answers, the pragmatic ones slower, and participants under 70% right on the controls were to be set aside.
Source papers
- Noveck IA (2001). When children are more logical than adults: Experimental investigations of scalar implicature. Cognition, 78(2):165-188.Paradigm source, Children and adults
- No normative data for this task.Paradigm source, N/A
Questions
How long does the Scalar Implicature Task take?
About 4 to 5 minutes, as listed in the test library.
What device does a participant need?
Timed test: participants need a desktop or laptop with a mouse or trackpad. The browser must be Chrome, Edge, Firefox or Safari.
What scores does it produce?
Pragmatic Rate, Pragmatic Answer Time, Logical Answer Time, True Some Sentence Time, False Some Sentence Time, Pragmatic Slowing, Logical Slowing, Control Accuracy, Underinformative Sentences Answered, Missed Sentences, Anticipations and Session In Another Language. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Scalar Implicature Task used in research?
250 papers from 2021 to 2025 name it in their title or abstract, down 10% compared with 2016 to 2020. That ranks it 113 of 495 test families in the ConductCognition test census (use grade C). Source: ConductCognition test census 0.22.0, counted 3 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Scalar Implicature 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?
Reading time before the keys work (ms) (0 to 2,000). 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.