Attention
Speeded Classification Task
Also called: Garner task, Garner paradigm, Garner speeded classification, Garner interference task, filtering task, speeded sorting task
Measure how much a part that should be ignored slows a judgment: participants press a key for one of two curves while the other stays the same, follows it or varies at random.
7-9 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. Try it here first, with no account.
- Papers, 2021 to 2025
- 23
- Evidence grade
- D
- Registered clinical studies
- 0
What it measures
Measures how much a part that should be ignored slows a simple judgment. Two curves appear side by side, and participants press F or J for the curve on the right while, block by block, the curve on the left stays the same, changes with it or changes at random.
Constructs
- Selective attention
- Perceptual grouping
- Dimensional interaction
Research uses
- Selective attention research
- Perceptual organization research
- Within-study group comparison
How the paradigm works
The Speeded Classification Task measures whether people can attend to one part of a display and ignore another. In Garner's speeded classification (Garner, 1974), people sort stimuli by one feature in three kinds of block: the other feature stays the same (baseline), changes along with the sorted feature (correlated) or changes at random (filtering). Slower sorting in filtering than in baseline, now called Garner interference, means the other feature could not be ignored. Pomerantz and Garner (1973) used pairs of curves like parentheses. All 16 of their participants sorted decks of 32 cards by one curve more slowly when the other curve varied at random, 16.99 s per deck against 14.42 s, and gained nothing when the two curves varied together. When one curve was turned on its side, the interference was gone.
More about the paradigm
Two white curves appear side by side, each bending like ( or like ). Participants press F if the curve on the right bends like ( and J if it bends like ), within 2,000 ms, while the curve on the left stays the same for a block, changes with the right curve or changes at random. A cross shows for 500 ms before each pair. Each of five blocks (two baseline, two correlated and one filtering block with twice the trials) runs twice, the second time in the reverse order, 288 trials in all at the standard setting.
The primary score is Garner interference: mean correct response time in the filtering blocks minus the baseline blocks. Pomerantz and Garner timed whole decks of cards; this version times single answers, as Felfoldy (1974) and Fitousi (2016) did, so the interference comes in milliseconds per answer. It falls with practice (Melara and Mounts, 1993) and as the time between trials grows (Fitousi, 2023), so compare people tested with the same setting.
Scores
- Garner Interferencems
- Mean correct response time when the left curve changed at random (filtering) minus when it stayed the same (baseline).
- Positive values mean the left curve could not be ignored
- Redundancy Gainms
- Mean correct response time in the baseline blocks minus when the left curve changed with the right one (correlated).
- Positive values mean the second curve helped
- Response Time, Baselinems
- Mean correct response time when the left curve stayed the same, after trimming.
- Lower is faster
- Response Time, Correlatedms
- Mean correct response time when the left curve changed with the right one, after trimming.
- Lower is faster
- Response Time, Filteringms
- Mean correct response time when the left curve changed at random, after trimming.
- Lower is faster
- Response Time, Parallel Pairsms
- Mean correct response time in the correlated block whose two curves bend the same way.
- Lower is faster
- Response Time, Mirror Pairsms
- Mean correct response time in the correlated block whose two curves mirror each other.
- Lower is faster
- Garner Interference, Mediansms
- Median correct response time in the filtering blocks minus the baseline blocks, without trimming.
- Positive values mean the left curve could not be ignored
- Accuracy%
- Share of all test trials answered with the key for the curve on the right.
- Higher is better
- Accuracy, Baseline%
- Share of baseline trials answered correctly, leaving out responses under 100 ms.
- Higher is better
- Accuracy, Correlated%
- Share of correlated trials answered correctly, leaving out responses under 100 ms.
- Higher is better
- Accuracy, Filtering%
- Share of filtering trials answered correctly, leaving out responses under 100 ms.
- Higher is better
- Error Interference%
- Accuracy in the baseline blocks minus accuracy in the filtering blocks.
- Positive values mean more errors when the left curve changed at random
- Baseline Trials Usedtrials
- Correct baseline trials at or after 100 ms, before trimming.
- Fewer than 32 is flagged
- Correlated Trials Usedtrials
- Correct correlated trials at or after 100 ms, before trimming.
- Informational
- Filtering Trials Usedtrials
- Correct filtering trials at or after 100 ms, before trimming.
- Fewer than 32 is flagged
- Anticipationscount
- Test responses under 100 ms from the pair's onset.
- Lower is better
- Missed Responsescount
- Test trials with no key press within 2,000 ms.
- Lower is better
- Test Trialstrials
- Test trials in the scored run.
- Informational
Use and evidence
23 papers from 2021 to 2025 name this task in their title or abstract, down 4% compared with 2016 to 2020. That ranks it 391 of 495 test families in our census. See the counts for every test.
Show the numbers
| Year | Papers |
|---|---|
| 2025 | 6 |
| 2024 | 5 |
| 2023 | 3 |
| 2022 | 4 |
| 2021 | 5 |
| 2020 | 3 |
| 2019 | 6 |
| 2018 | 4 |
| 2017 | 6 |
| 2016 | 5 |
| 2015 | 7 |
| 2014 | 9 |
| 2013 | 14 |
| 2012 | 10 |
| 2011 | 6 |
| 2010 | 10 |
| 2009 | 7 |
| 2008 | 4 |
| 2007 | 2 |
| 2006 | 2 |
| 2005 | 2 |
| 2004 | 2 |
| 2003 | 1 |
| 2002 | 2 |
| 2001 | 1 |
| 2000 | 1 |
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
- 6
- 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 2 November 2026 from OpenAlex and ClinicalTrials.gov.
Task design
The trial sequence as the test runs in ConductCognition.
- Display
- Two white curves side by side on a dark panel, each 60 px tall and bending like ( or like ), their centers 48 px apart. A curve like ( is coded L and one like ) is coded R, and a pair is named left curve first.
- Task
- F if the curve on the right is L and J if it is R, with the two keys and their curves shown under every pair.
- Blocks
- Baseline: LL and LR, or RL and RR (the left curve stays the same). Correlated: LL and RR, or LR and RL (it changes with the right curve). Filtering: all four pairs (it changes at random).
- Fixation
- 500 ms, between every answer and the next pair.
- Response window
- Until a key or 2,000 ms.
- Trial mix
- 24 trials in each two-pair block and 48 in the filtering block (standard; 16 to 48, in steps of 8), each pair equally often in a seeded random order. The five blocks run twice, the second time in the reverse order: 288 trials in 10 parts.
- Practice
- 16 trials with all four pairs and feedback for 800 ms. In the test, only a wrong, slow or early answer gets a message, for 1,000 ms.
- Score
- Garner interference: mean correct response time in the filtering blocks minus the baseline blocks.
Pomerantz and Garner (1973) had people sort decks of 32 typed cards into two piles and timed each deck. This version draws the curves and times each answer, as Felfoldy (1974) and Fitousi (2016) did, so the interference comes in milliseconds per answer and cannot be compared with the source's seconds per deck. Participants always answer for the right curve, as in Pomerantz and Schwaitzberg (1975); the source used each curve in turn. The interference faded as the curves were spread apart and was gone at 4 degrees (Pomerantz and Schwaitzberg, 1975); these sit much closer than that on a desktop screen at arm's length. It also falls with practice (Melara and Mounts, 1993), so compare people tested with the same setting.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: speeded_classification_test from the scored run, F or J for the curve on the right (chance 50%, each key on half the trials of every block) within 2,000 ms, correctness from the recorded key. Below chance invalidates and not above chance is flagged. One key on 90% of answers invalidates when accuracy is not above chance. Fewer than 32 correct trials in the baseline or the filtering blocks is flagged and withholds the percentile. Missed responses and responses under 100 ms follow the timed-test rule.
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. The interference is a difference between two means, so it needs enough correct trials in each; the limit of 32, half the trials of each kind in the shortest session, is a ConductScience choice.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Trials per block | 24 | 16 to 48 |
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
- 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 Speeded Classification Task in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Garner, 1974; Pomerantz and Garner, 1973), with the standard settings (trials per block: 24). The main score was garner interference (mean correct response time when the left curve changed at random (filtering) minus when it stayed the same (baseline)).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Pomerantz, J. R., & Garner, W. R. (1973). Stimulus configuration in selective attention tasks. Perception & Psychophysics, 14(3), 565-569. https://doi.org/10.3758/BF03211198The pairs of curves and the three kinds of block: all 16 participants sorted more slowly when the other curve varied at random.
- Garner, W. R. (1974). The processing of information and structure. Erlbaum. The speeded classification paradigm and its baseline, correlated and filtering tasks.
- Pomerantz, J. R., & Schwaitzberg, S. D. (1975). Grouping by proximity: Selective attention measures. Perception & Psychophysics, 18(5), 355-361. https://doi.org/10.3758/BF03211212Answering for the right curve only, and the interference fading as the curves are spread apart.
- Felfoldy, G. L. (1974). Repetition effects in choice reaction time to multidimensional stimuli. Perception & Psychophysics, 15(3), 453-459. https://doi.org/10.3758/BF03199285Single timed answers in place of card sorting; the interference followed how often a stimulus repeated.
- Fitousi, D. (2016). Simon and Garner effects with color and location: Evidence for two independent routes by which irrelevant location influences performance. Attention, Perception, & Psychophysics, 78(8), 2433-2455. https://doi.org/10.3758/s13414-016-1197-1A discrete-trial design with two baseline, two correlated and one filtering block of twice the length.
- Melara, R. D., & Mounts, J. R. W. (1993). Selective attention to Stroop dimensions: Effects of baseline discriminability, response mode, and practice. Memory & Cognition, 21(5), 627-645. https://doi.org/10.3758/BF03197195Garner interference fell with practice.
- Algom, D., & Fitousi, D. (2016). Half a century of research on Garner interference and the separability-integrality distinction. Psychological Bulletin, 142(12), 1352-1383. https://doi.org/10.1037/bul0000072A review of the paradigm.
Source papers
- Pomerantz JR, Garner WR (1973). Stimulus configuration in selective attention tasks. Perception & Psychophysics, 14(3):565-569.N 16, Adults (mean age 22)
Questions
How long does the Speeded Classification Task take?
About 7 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?
Garner Interference, Redundancy Gain, Response Time, Baseline, Response Time, Correlated, Response Time, Filtering, Response Time, Parallel Pairs, Response Time, Mirror Pairs, Garner Interference, Medians, Accuracy, Accuracy, Baseline, Accuracy, Correlated, Accuracy, Filtering, Error Interference, Baseline Trials Used, Correlated Trials Used, Filtering Trials Used, Anticipations, Missed Responses and Test Trials. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Speeded Classification Task used in research?
23 papers from 2021 to 2025 name it in their title or abstract, down 4% compared with 2016 to 2020. That ranks it 391 of 495 test families in the ConductCognition test census (use grade E). Source: ConductCognition test census 0.22.0, counted 2 November 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Speeded Classification 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?
Trials per block (16 to 48). 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. 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.