Language
Self-Paced Reading Task
Also called: SPR, moving window task, word-by-word reading task
Measure where English sentences become hard to read, word by word, with garden-path sentences against unambiguous versions of the same sentences.
6-9 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. Preview any test free after you sign up.
- Papers, 2021 to 2025
- 740
- Evidence grade
- C
- Registered clinical studies
- 1
What it measures
Measures where English sentences become hard to read. Participants read 60 sentences one word at a time, pressing the space bar for each word, and answer yes or no questions after some of them. Reading times on garden-path sentences, against the same sentences made unambiguous, show the cost of revising a first reading.
Constructs
- Sentence comprehension
- Syntactic parsing
- Reading speed
Research uses
- Psycholinguistics and sentence processing research
- Second-language and bilingualism research
- Within-study group comparison
How the paradigm works
The self-paced reading task measures where a sentence becomes hard to read. Participants read sentences one word at a time, pressing the space bar for each word, and the time spent on each word shows where reading slows. The words appear in a moving window (Just, Carpenter and Woolley 1982): each sentence starts as a row of dashes, one per character, and each press shows the next word in its place while the word before it turns back into dashes.
More about the paradigm
The 24 experimental sentences are garden-path sentences with a reduced relative clause, such as "The guest greeted by the host took off his coat." The verb first reads as the main verb, and the phrase "by the host" shows that this reading was wrong (Ferreira and Clifton 1986). Each sentence also has an unambiguous version with "who was" before the verb. The first noun always names a person, the case in which readers show clear difficulty (Trueswell, Tanenhaus and Garnsey 1994). Two lists, drawn from the session's random seed, show every participant each experimental sentence once, 12 in each version, among 36 filler sentences, in a seeded order in which two experimental sentences never follow each other. Yes or no questions follow 20 of the 60 sentences. The sentences and questions were written for this task; no published sentence set is used. The sentences are English.
The main score is the garden-path effect: the mean reading time per word on the "by" phrase in the ambiguous sentences minus the same words in the unambiguous sentences. The same difference on the next two words, comprehension accuracy, the mean reading time per word and reading times by region are also reported. Reading times under 100 ms or over 2,500 ms are left out (Enochson and Culbertson 2015). Scores support research and do not diagnose.
Scores
- Garden-Path Effectms
- Mean reading time per word on the by-phrase, ambiguous minus unambiguous sentences.
- Larger values mean more slowing where the first reading had to be revised
- Main Verb Effectms
- The same difference on the two words after the by-phrase.
- Larger values mean slowing that carried over
- Comprehension Accuracy%
- Share of questions answered correctly.
- Higher is better; below 0.8 is flagged
- Mean Reading Time per Wordms
- Mean reading time per word over all sentences, after trimming.
- Lower is faster
- By-Phrase Time, Ambiguousms
- Mean reading time per word on the by-phrase in ambiguous sentences.
- Lower is faster
- By-Phrase Time, Unambiguousms
- Mean reading time per word on the by-phrase in unambiguous sentences.
- Lower is faster
- Verb Time, Ambiguousms
- Mean reading time on the verb in ambiguous sentences.
- Lower is faster
- Verb Time, Unambiguousms
- Mean reading time on the verb in unambiguous sentences.
- Lower is faster
- Main Verb Time, Ambiguousms
- Mean reading time per word on the two words after the by-phrase in ambiguous sentences.
- Lower is faster
- Main Verb Time, Unambiguousms
- Mean reading time per word on the two words after the by-phrase in unambiguous sentences.
- Lower is faster
Use and evidence
740 papers from 2021 to 2025 name this task in their title or abstract, up 46% compared with 2016 to 2020. That ranks it 34 of 395 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 209 |
| 2024 | 158 |
| 2023 | 135 |
| 2022 | 122 |
| 2021 | 116 |
| 2020 | 114 |
| 2019 | 114 |
| 2018 | 101 |
| 2017 | 94 |
| 2016 | 84 |
| 2015 | 112 |
| 2014 | 92 |
| 2013 | 79 |
| 2012 | 51 |
| 2011 | 59 |
| 2010 | 49 |
| 2009 | 36 |
| 2008 | 27 |
| 2007 | 27 |
| 2006 | 28 |
| 2005 | 17 |
| 2004 | 21 |
| 2003 | 20 |
| 2002 | 17 |
| 2001 | 12 |
| 2000 | 8 |
Use grade B
300 to 999 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
- 70
- Registered clinical studies using it as an outcome
- 1
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 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Sentences
- 24 garden-path sentences with a reduced relative clause, each also written with "who was" before the verb, and 36 filler sentences.
- Lists
- Two lists drawn from the session's seed show each experimental sentence once: 12 ambiguous and 12 unambiguous.
- Order
- Drawn from the session's seed: 2 fillers first, then each experimental sentence followed by 1 to 3 fillers, so two experimental sentences never follow each other.
- Display
- Moving window: each sentence starts as dashes, one per character, on one line; each press of the space bar shows the next word and hides the one before.
- Questions
- 20 yes or no questions, 10 answered yes, after a third of the sentences; F for yes and J for no.
- Practice
- 3 sentences, 2 with a question and feedback for 1,500 ms.
- Trimming
- Reading times under 100 ms or over 2,500 ms are left out of the scores; the upper limit can be set from 2,000 to 4,000 ms.
- Score
- Garden-path effect: mean reading time per word on the by-phrase, ambiguous minus unambiguous sentences.
The sentences and questions were written for this task; no published sentence set is used. The sentences are English.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: self_paced_reading_test_question from the scored run, one yes or no answer per question; word reading times from self_paced_reading_test_word. Accuracy significantly below chance (one in two) invalidates and accuracy not above chance is flagged. The same key on 90% of at least 12 answers, with accuracy not above chance, invalidates. Answers under 200 ms on at least half of the questions invalidate. More than 20% of word reading times under 100 ms invalidate and more than 10% are flagged. Fewer than 80% of questions answered correctly is flagged and withholds the percentile. A session run in another language is flagged. A hidden tab withholds the percentile.
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). 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. A word cannot be read in under 100 ms, so a session with more than 20% of word times that fast (the platform's timed-test proportion) was not read; more than 10% is flagged, against the 2 to 2.5% of times trimmed at 100 and 2,500 ms in Enochson and Culbertson (2015). Self-paced reading studies usually exclude participants who answer more than 20% of the questions wrongly (Enochson and Culbertson 2015), so lower accuracy is flagged and the percentile withheld. 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. Reading times run while the tab is hidden, so a hidden tab withholds the percentile.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Longest reading time kept (ms) | 2,500 | 2,000 to 4,000 |
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.
- Recruiting online? See the setup guides for Prolific, Sona Systems, CloudResearch Connect and Qualtrics.
Cite and describe
Participants completed the Self-Paced Reading Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Just, Carpenter & Woolley, 1982), with the standard settings (longest reading time kept (ms): 2,500). The main score was garden-path effect (mean reading time per word on the by-phrase, ambiguous minus unambiguous sentences).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Just, M. A., Carpenter, P. A., & Woolley, J. D. (1982). Paradigms and processes in reading comprehension. Journal of Experimental Psychology: General, 111(2), 228-238. https://doi.org/10.1037/0096-3445.111.2.228
- Ferreira, F., & Clifton, C. (1986). The independence of syntactic processing. Journal of Memory and Language, 25(3), 348-368. https://doi.org/10.1016/0749-596X(86)90006-9
- Trueswell, J. C., Tanenhaus, M. K., & Garnsey, S. M. (1994). Semantic influences on parsing: Use of thematic role information in syntactic ambiguity resolution. Journal of Memory and Language, 33(3), 285-318. https://doi.org/10.1006/jmla.1994.1014
- Enochson, K., & Culbertson, J. (2015). Collecting psycholinguistic response time data using Amazon Mechanical Turk. PLOS ONE, 10(3), e0116946. https://doi.org/10.1371/journal.pone.0116946
Source papers
- Just MA, Carpenter PA, Woolley JD (1982). Paradigms and processes in reading comprehension. Journal of Experimental Psychology: General, 111(2):228-238.Paradigm source, Adults
- Ferreira F, Clifton C (1986). The independence of syntactic processing. Journal of Memory and Language, 25(3):348-368.Paradigm source, Adults
- Enochson K, Culbertson J (2015). Collecting psycholinguistic response time data using Amazon Mechanical Turk. PLOS ONE, 10(3):e0116946.Paradigm source, Adults
Questions
How long does the Self-Paced Reading Task take?
About 6 to 9 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?
Garden-Path Effect, Main Verb Effect, Comprehension Accuracy, Mean Reading Time per Word, By-Phrase Time, Ambiguous, By-Phrase Time, Unambiguous, Verb Time, Ambiguous, Verb Time, Unambiguous, Main Verb Time, Ambiguous and Main Verb Time, Unambiguous. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Self-Paced Reading Task used in research?
740 papers from 2021 to 2025 name it in their title or abstract, up 46% compared with 2016 to 2020. That ranks it 34 of 395 test families in the ConductCognition test census (use grade B). Source: ConductCognition test census 0.14.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Self-Paced Reading Task have?
In the published research, 2 papers report its test-retest reliability and 1 registered clinical study uses it as an outcome (evidence grade C in the ConductCognition test census).
Which settings can a researcher change?
Longest reading time kept (ms) (2,000 to 4,000). 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.