Attention
Middle Flag Game
Also called: selective attention game for children, flag touch game
Measure how much flags that point the other way slow a reach, in children aged 5 and over and in adults. The player holds a spot, a row of five flags shows, and the player lets go and touches the pad on the side the middle flag points to.
8-11 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
- 1,472
- Evidence grade
- A
- Registered clinical studies
- 334
Figures for the Eriksen Flanker Task, the published test this game is compared with.
How these are countedWhat it measures
A game for research on selective attention in children aged 5 and over and in adults, with an adult reading the screens aloud to children under 8. The player holds a spot, a row of five flags shows, and the player lets go and touches the pad on the side the middle flag points to. The main score is how much flags that point the other way slow the reach to the pad. It is a raw research measure with no percentile.
Constructs
- Selective attention
- Interference control
- Speed of response to a target among distractors
Research uses
- Research on selective attention and interference control in children and adults
How the paradigm works
The Middle Flag Game is a game for research on selective attention and interference control in children and adults aged 5 and over. The player holds a round spot at the foot of the screen. After a wait of 600, 900 or 1,200 ms a row of five flags shows, and the player lets go of the spot and touches the pad, in a top corner, on the side the middle flag points to. The four flags beside it point the same way (congruent) or the other way (incongruent), and the flags stay up until the touch. The main score is the cost in response time of the incongruent flags: the mean time from the flags to the touch on the pad over the correct incongruent trials, less the mean over the correct congruent ones. The start of the reach (the flags to the let go) and its move (the let go to the touch) are scored the same way, add up to the response-time cost, and are reported beside it. The paradigm is the flanker task of Eriksen and Eriksen (1974), and the release-and-press reach, with its three times, follows Erb and colleagues (2025). The flags, the spot, the pads, the lists, the forms and the words are drawn or written for this game: no flag is a country's, and no image, font or sound comes from outside.
More about the paradigm
The paper counts and grades on this page are those of the Eriksen Flanker Task, the published test the game is compared with, not the game's own. The validation plan also compares the game with its own plain look and, at ages 5 and 6, with the Day-Night Task. The cost of a conflict task retests poorly (Hedge and colleagues, 2018), so the plain look comes first, and the Stroop task is the check that should agree less. A demonstration of 2 trials with an animated hand and a practice of 12 trials come first, the right pad glowing for 800 ms after each practice touch; the practice runs once more when fewer than 9 of the 12 are right. The game then runs 4 blocks of 36 trials as standard, with a rest between blocks, and shows no score, count, timer or speed figure. A let go before the flags show is an early start and restarts the wait, and a third early start leaves the trial out. A let go under 100 ms after the flags is an anticipation and is left out. A second finger, a palm, a mouse button other than the main one, a hidden tab or a resized window leaves the trial out. Each session takes one of 2 alternate forms of 216 trials, and a third form serves Try it now and no study.
Scores
- Response-Time Costms
- The mean response time of correct incongruent trials less the mean of correct congruent ones, from the frame that drew the flags to the touch on the pad. The first trial of a block, errors, the trial after an error, a trial with a second contact, a stray touch, a hidden tab or a resized window, a trial whose times came from the handler's clock or whose first frame came late, a leave under 100 ms, a response time outside 200 to 4,000 ms and a response time more than 3 SD from the person's mean in its condition are left out. Empty below 20 trials in a condition.
- Lower means flags that point the other way slow the reach less; compared within one input type (mouse, touch or pen)
- Initiation Costms
- Secondary: the mean time from the flags to the moment the finger or the button lets go, incongruent trials less congruent ones, over the trials the response-time cost keeps.
- Secondary: the start of the reach; with the movement cost it adds up to the response-time cost
- Movement Costms
- Secondary: the mean time from the moment the finger or the button lets go to the touch on the pad, incongruent trials less congruent ones, over the same trials.
- Secondary: the move of the reach; with the initiation cost it adds up to the response-time cost
- Initiation Cost, Release Startsms
- The initiation cost over the trials that started by a release of the finger or the button. Empty below 20 trials in a condition.
- Quality check: a release and a move of the contact are not the same start
- Initiation Cost, Slide Startsms
- The initiation cost over the trials that started by moving the contact more than 8 px from where it was when the picture showed. Empty below 20 trials in a condition.
- Quality check: a release and a move of the contact are not the same start
- Error Cost%
- The share of incongruent trials answered on the wrong side less the share of congruent trials answered on the wrong side, in percentage points, over the answered test trials.
- Secondary: higher means more errors with flags that point the other way
- Mean Response Time, Congruentms
- The mean response time of the scored congruent trials.
- Informational
- Mean Response Time, Incongruentms
- The mean response time of the scored incongruent trials.
- Informational
- Median Response Timems
- The median response time of all scored trials.
- Informational: read it beside the cost, since a slow person shows a smaller cost
- Median Initiation Timems
- The median time from the flags to the moment the finger or the button lets go, over all scored trials.
- Quality check: above 1,500 ms flags a person who waits until sure
- Accuracy, Congruent%
- The share of answered congruent test trials in which the pad on the middle flag's side was touched.
- Quality check: chance is .5
- Accuracy, Incongruent%
- The share of answered incongruent test trials in which the pad on the middle flag's side was touched.
- Quality check: chance is .5; touching the side most of the flags point to scores 0
- Accuracy%
- The share of answered test trials in which the pad on the middle flag's side was touched.
- Quality check: chance is .5; below .70 makes the session invalid
- Scored Trialscount
- The test trials that count towards the costs.
- Quality check: the cost needs 20 in each condition
- Early Startscount
- Times the finger or the button let go, or the contact rolled out of the hold zone, before the flags showed, over the test trials. Each one restarts the wait.
- Quality check
- Leaves Under 100 mscount
- Test trials in which the finger or the button let go less than 100 ms after the flags showed. They are left out and not repeated.
- Quality check
- Timed-Out Trialscount
- Test trials with no let go within 4,000 ms of the flags, or no touch on a pad within 4,000 ms of the let go.
- Quality check
- Share of Slide Starts%
- The share of scored trials that started by moving the contact more than 8 px from where it was when the picture showed, not by a release.
- Informational: a mouse, a finger and a pen start differently
- Stray Touchescount
- First contacts that landed off the spot before the hold, and touches off both pads after the let go.
- Quality check
- Trials with a Second Contactcount
- Test trials with a second finger, a palm or a mouse button other than the main one. They are left out of every score.
- Quality check
- Trials Timed by the Pagecount
- Trials whose let go or touch time came from the page's clock and not from the event, which adds the page's queue delay. They are left out of the costs.
- Quality check
- Share of Fixed-Delay Starts%
- The share of scored trials with an initiation time within 40 ms of the person's median initiation time.
- Quality check: above .60 flags a let go at a fixed delay, where the initiation cost reads nothing
- Share on One Side%
- The share of answers on the side the person touched most often.
- Quality check: .70 or more with low accuracy flags one side
- Share of Touch Answers%
- The share of scored answers given by a finger on a touch screen (the rest by a mouse or a pen).
- Informational: the costs are read within one input type
Use and evidence
1,472 papers from 2021 to 2025 name this task in their title or abstract, up 17% compared with 2016 to 2020. That ranks it 17 of 711 test families in our census. These counts belong to the Eriksen Flanker 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 | 377 |
| 2024 | 309 |
| 2023 | 240 |
| 2022 | 271 |
| 2021 | 275 |
| 2020 | 284 |
| 2019 | 280 |
| 2018 | 271 |
| 2017 | 234 |
| 2016 | 190 |
| 2015 | 206 |
| 2014 | 189 |
| 2013 | 160 |
| 2012 | 143 |
| 2011 | 123 |
| 2010 | 133 |
| 2009 | 80 |
| 2008 | 94 |
| 2007 | 44 |
| 2006 | 51 |
| 2005 | 38 |
| 2004 | 33 |
| 2003 | 21 |
| 2002 | 22 |
| 2001 | 7 |
| 2000 | 13 |
Use grade A
1,000 or more papers from 2021 to 2025.
Evidence grade A
At least 25 papers report its test-retest reliability and at least 25 registered clinical studies use it as an outcome.
- Papers on test-retest reliability
- 38
- Papers on reliability or validity
- 223
- Registered clinical studies using it as an outcome
- 334
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 Eriksen Flanker 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.
- Flags
- Five flags in a row, each a pole with a swallow-tailed pennant, drawn in code. A flag that points left is the mirror of one that points right, so the pennant's direction is the only difference between them. The middle flag counts, and nothing marks it: it is told by its place. The two flags on each side of it point the same way (congruent) or the other way (incongruent). The flags sit a quarter of a flag's width apart, each is at most 150 px across, and no flag is a country's. The game look adds a soft hillside behind the screen; the plain look shows the same flags on white. Both looks draw every flag the same way.
- Spot and pads
- The player holds a round spot of 80 px at the foot of the screen. Two pads sit in the top corners, each at least 120 by 96 px, and the player touches the pad itself, never a flag. The hold counts while the contact stays inside a zone 1.6 times the spot's radius, so a roll of the finger is not a let go. The layout is worked out from the window before each trial, and a window under 480 by 440 px is held with a line asking for a taller window or an upright tablet, and takes no answer until it fits.
- Trial
- The player holds the spot for 600, 900 or 1,200 ms, in the order of the form's list. The five flags then show and stay up until the touch. The player lets go (a release, or a move of more than 8 px from where the contact was when the flags showed) and touches the pad on the middle flag's side. Only a new touch on a pad answers: a contact dragged from the spot to a pad and lifted there gives no answer, and the trial waits for a new touch until it times out; in practice that drag shows a note asking the player to let go first, and the trial starts over. A 400 ms gap follows. The player has 4,000 ms to let go and 4,000 ms to touch a pad, and no screen shows a score, a count, a timer or a speed figure. A let go before the flags show is an early start: the wait starts again, and a third early start leaves the trial out. A let go less than 100 ms after the flags is an anticipation, and the trial is left out and not repeated.
- Blocks
- A demonstration of 2 trials with an animated hand and a practice of 12 trials come first, the right pad glowing for 800 ms after each practice touch. The first 4 practice trials are congruent, and the practice runs once more when fewer than 9 of the 12 are right. The test is 4 blocks of 36 trials as standard (2 to 6), 144 trials, with a rest between blocks. Each block holds each of the four displays, a left or right middle flag with congruent or incongruent flankers, 9 times, and starts with a congruent trial.
- Answers
- Only a touch on a pad answers, with a mouse, a finger or a pen; no key does. A first contact off the spot before the hold, and a touch on no pad after the let go, are counted as stray touches and the trial goes on. A second finger, a palm or a mouse button other than the main one is counted, and that trial is left out of every score. A hidden tab or a window that changed size leaves the trial out, and the layout refits before the next one.
- Forms
- 2 alternate forms for studies, each a list of 216 trials in 6 blocks, and a third form for Try it now that no study serves. Each participant gets the forms in an order drawn once for them, and each session takes the next form. In each block every hold time comes 3 times in each display; no more than 3 trials in a row share a condition or a middle direction, and each of the 16 moves from one display to the next comes 1 to 3 times. Touching the side most of the flags point to is right on every congruent trial and wrong on every incongruent one, and touching one pad every time or guessing is right half the time: each scores .50 over a list and is invalid as not matching the task (under .70).
- Scores
- The response-time cost (the main score): the mean time from the flags to the touch on the pad over the correct incongruent trials, less the mean over the correct congruent ones, so a lower cost means flags that point the other way slow the reach less. The start of the reach (the flags to the let go) and the move (the let go to the touch) are scored the same way and add up to it. The first trial of a block, errors, the trial after an error, a trial with a second contact, a stray touch, a hidden tab or a resized window, a let go under 100 ms, a response time outside 200 to 4,000 ms and one more than 3 SD from the person's mean in its condition are left out. The error cost, the accuracy in each condition, the median times, the early starts and the shares of slide starts, fixed-delay starts, one-sided answers and touch answers come with it. A condition with fewer than 20 scored trials has no cost.
- Devices
- Computers and touch-screen computers at release, with a mouse, a finger or a pen; a touch-screen computer counts as a computer and records touch. Tablets open once two filmed tablets pass the device timing check, and phones do not run the game. A mouse, a finger and a pen do not move alike, so read each cost within one input type. For a child under 8, an adult sits beside the child, reads each screen aloud and neither points nor answers. A person who cannot use a pointer takes the plain flanker task, which runs on keys.
- Schedule
- The standard 4 blocks give 144 trials. Use 3 blocks for children aged 5 to 8, and 6 blocks for adults who need the cost for individual differences. Keep each person on one device class and one input at every visit. For a retest, run a second visit of the same study 7 to 28 days later, and the game takes the next form.
The paradigm is the flanker task of Eriksen and Eriksen (1974), with flags in place of letters. The reach (hold a spot, let go, touch a pad, with the response, initiation and movement times) follows Erb and colleagues (2025), in which a touch-screen release-and-press flanker task gave a split-half reliability of .85 for the response-time effect in 51 adults. The flags, the spot, the pads, the 36-trial blocks, the three hold times, the lists, the forms, the words and the looks are our own. This is a new measure for research with children aged 5 and over and with adults: its reliability and norms come from use, the figures published in those studies are not its own, and it computes no percentiles. The cost of a conflict task retests poorly (Hedge and colleagues 2018) and may not agree with another task's cost (Rouder and Haaf 2019), so the response-time cost is reported with the start and the move of the reach beside it, and the validation plan reads the game against its own plain look first. A calibrated flanker task gave reliable estimates in under 100 trials (Kucina and colleagues 2023); this game keeps the plain conflict and 144 trials as standard. A person who waits until sure shows a small cost, as in every difference score, so read the cost beside the median times and the quality flags, and read each cost within one input type. For a retest 7 to 28 days later, add a second visit to the same study.
Quality checks
Each session is marked valid, needs review or invalid, with the reason, before you see its scores.
Scored rows: middle_flag_game test trials, one touch on one of two pads (chance 50%), with the condition, the side and the hold time taken from the form's table. Fewer than 60% of the planned trials answered cleanly (no second contact, stray touch, hidden tab, resized window or void) invalidates, and so does accuracy under 70%. One side on 70% of the answers with accuracy under 80% is flagged. Early starts and leaves under 100 ms on more than 15% of the trials are flagged and on more than 30% invalidate. Also flagged: more than 20% of answered trials let go under 200 ms after the flags, a median time to let go above 1,500 ms, more than 60% of the trials within 40 ms of the same time to let go, and second contacts on more than 10% of the trials. Missed responses and responses under 100 ms follow the timed-test rule.
Why these rules
A trial has one right pad among two, so guessing gives half; under 70% right is far enough above chance that it leaves room for a careful player, and below it the answers did not follow the flags, which is a ConductScience choice. 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. 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 primary score is the response time from the flags to the touch, which keeps its cost when a player lets go early and decides in the air, lets go at a fixed delay or waits until sure (Erb 2025 splits that time into the start and the move of the reach, which add up to it); the checks name those habits (fast, slow and fixed-delay starts) so that a researcher reads the flags beside the cost. What they cannot tell is a person who varies a long wait from one who is slow, which every difference score shares, the plain flanker's and Stroop's included. A second finger, a palm or a touch off the pads is how a young child's hands interrupt a trial, and the trial is left out of every score. The limits are ours.
No percentile. Scores and quality flags only.
Running this test
Settings you can change
| Setting | Standard | Allowed |
|---|---|---|
| Look | Game | Game, Plain |
| Blocks | 4 | 2 to 6 |
Yearly plans can change settings within these limits. Free and pay-as-you-go studies use the standard settings.
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
- 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 Middle Flag Game in a web browser (ConductCognition task version 1.0.0, scoring version 1.0.0; Eriksen and Eriksen, 1974), with the standard settings (look: Game; blocks: 4). The main score was response-time cost (the mean response time of correct incongruent trials less the mean of correct congruent ones, from the frame that drew the flags to the touch on the pad).
ConductCognition versions: task version 1.0.0, scoring version 1.0.0. Every export records the versions each session used, for a methods section.
- Eriksen, B. A., & Eriksen, C. W. (1974). Effects of noise letters upon the identification of a target letter in a nonsearch task. Perception & Psychophysics, 16(1), 143-149. https://doi.org/10.3758/BF03203267The paradigm: a target letter flanked by letters that call for the same response or the other one, which slowed the response more (abstract read). The game takes the paradigm, with a pennant flag in place of a letter.
- Erb, C. D., Morrison, S., & Nicholson-Brown, A. (2025). Moving beyond button presses to enhance the reliability of congruency tasks. Behavior Research Methods, 57(8), 215. https://doi.org/10.3758/s13428-025-02740-0A release-and-press arrow flanker task on a touch screen gave split-half reliabilities of .85 for the response-time effect, .94 for initiation and .78 for movement in 51 adults (full text read). The game takes the reach, its three times and the rules for the scored trials.
- Hedge, C., Powell, G., & Sumner, P. (2018). The reliability paradox: Why robust cognitive tasks do not produce reliable individual differences. Behavior Research Methods, 50(3), 1166-1186. https://doi.org/10.3758/s13428-017-0935-1The test-retest reliabilities of seven classic tasks, the Eriksen flanker and Stroop tasks among them, ranged from 0 to .82 and were low for most (full text read). The page reports the cost beside the start and the move of the reach, and each study reads its own reliability.
- Kucina, T., Wells, L., Lewis, I., de Salas, K., Kohl, A., Palmer, M. A., Sauer, J. D., Matzke, D., Aidman, E., & Heathcote, A. (2023). Calibration of cognitive tests to address the reliability paradox for decision-conflict tasks. Nature Communications, 14(1), 2234. https://doi.org/10.1038/s41467-023-37777-2A flanker task calibrated with an added manipulation gave reliable estimates of individual differences in under 100 trials (full text read). The game adds no manipulation of its own.
- Rouder, J. N., & Haaf, J. M. (2019). A psychometrics of individual differences in experimental tasks. Psychonomic Bulletin & Review, 26(2), 452-467. https://doi.org/10.3758/s13423-018-1558-yWith the data of Hedge and colleagues, Bayes-factor support for no correlation between the Stroop and flanker effects (abstract read). The game's checks name the plain flanker task as the one to agree with and the Stroop task as the one that should agree less.
Source papers
Questions
How long does the Middle Flag Game take?
About 8 to 11 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?
Response-Time Cost, Initiation Cost, Movement Cost, Initiation Cost, Release Starts, Initiation Cost, Slide Starts, Error Cost, Mean Response Time, Congruent, Mean Response Time, Incongruent, Median Response Time, Median Initiation Time, Accuracy, Congruent, Accuracy, Incongruent, Accuracy, Scored Trials, Early Starts, Leaves Under 100 ms, Timed-Out Trials, Share of Slide Starts, Stray Touches, Trials with a Second Contact, Trials Timed by the Page, Share of Fixed-Delay Starts, Share on One Side and Share of Touch Answers. Each session is also marked valid, needs review or invalid, with the reason.
How widely is the Eriksen Flanker Task used in research?
These counts belong to the Eriksen Flanker Task, the published test this game is compared with. This game is a new measure, so its own reliability and norms come from use. 1,472 papers from 2021 to 2025 name the task in their title or abstract, up 17% compared with 2016 to 2020. That ranks it 17 of 711 test families in the ConductCognition test census (use grade A). Source: ConductCognition test census 0.58.0, counted 31 October 2026 from OpenAlex and ClinicalTrials.gov.
How much reliability evidence does the Eriksen Flanker Task have?
These counts belong to the Eriksen Flanker 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, 38 papers report the task's test-retest reliability and 334 registered clinical studies use it as an outcome (evidence grade A 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) and blocks (2 to 6). Settings can be changed on yearly plans; free and pay-as-you-go studies use the standard settings. 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. 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