The Flanker test is one of the standard selective-attention tasks in cognitive psychology. You look at a row of letters or arrows. The middle one is the target. The ones flanking it are distractors. Your job is to respond based on the target — and only the target — as quickly and accurately as you can.
When the flankers match the target, the task is easy. When they don't, you slow down. The difference is the flanker effect, and it indexes how well your attention can stay narrow while irrelevant information presses on it from the sides.
This page is a plain-English explainer. Where the paradigm comes from, what congruent and incongruent trials measure, what a "good" interference score looks like, and how the test relates to the broader Attention Network Test that grew out of it.
What the Flanker test measures
The Flanker test measures selective attention — specifically, the ability to focus on a small region of the visual field while suppressing similar but task-irrelevant information nearby. It's closely related to what the Stroop test measures, but the conflict in Flanker is purely spatial: the distractors are physically next to the target rather than embedded in it.
Two trial types do most of the work:
- Congruent — flankers point or read the same direction as the target (→→→→→). Response is fast.
- Incongruent — flankers point or read the opposite direction (→→←→→). Response is slow.
The flanker interference score is the response-time difference between incongruent and congruent trials. Larger interference scores mean nearby information had a stronger pull on your response system; smaller scores mean attention stayed tighter on the target.
A third trial type, neutral (flankers that are letters or symbols not mapped to any response, like XXSXX), is sometimes included as a clean baseline.
Where the paradigm comes from
Barbara A. Eriksen and Charles W. Eriksen introduced the task in 1974 in Perception & Psychophysics. The original paradigm used letter targets (H and S) with flanker letters that either matched the target letter (and thus mapped to the same response) or disagreed (and mapped to the opposite response). The clear behavioral effect was the response-time cost on incongruent trials, even when subjects were explicitly instructed to ignore the flankers and could perfectly identify what they were.
The result was important because it showed that visual attention isn't an all-or-nothing spotlight. Even when you fixate on a single letter and intend to attend only to it, nearby objects within a small visual region (roughly 1 degree of visual angle, in the original studies) automatically influence your response system.
The test got a major modern extension in Jin Fan and colleagues' 2002 Journal of Cognitive Neuroscience paper introducing the Attention Network Test (ANT). The ANT embeds the Flanker task inside a larger paradigm that also measures two other attentional networks: alerting (preparedness for an upcoming target) and orienting (shifting attention based on a spatial cue). Each ANT trial produces three scores — alerting, orienting, and conflict (which is essentially the flanker interference score). The ANT became one of the most widely-used computerized attention measures in research because it produces three theoretically distinct attention scores from a single 20-minute session.
How the modern test works
A typical computerized Flanker test:
- A fixation cross appears in the center.
- A row of five characters appears, briefly (typically 100–250 ms).
- The center character is the target. Flanking characters are distractors.
- You press one of two keys based on the target.
- The trial sequence mixes congruent, incongruent, and (often) neutral trials in random order.
Modern arrow-flanker versions use → and ← arrows instead of letters because they translate cleanly across languages and are intuitive even for children. Most adult ANT studies use the arrow version.
A complete Flanker test typically runs 50 to 100 trials and takes 5 to 10 minutes. The Attention Network Test version runs longer (around 20 minutes) because it includes the additional cue-related conditions.
Scoring typically produces:
- Mean response time on congruent trials.
- Mean response time on incongruent trials.
- Flanker interference score = incongruent RT minus congruent RT.
- Accuracy on each trial type.
What "good" looks like
For healthy adults on a standard arrow-flanker task:
- Congruent RT: roughly 400–550 ms.
- Incongruent RT: roughly 500–650 ms.
- Flanker interference score: typically 50–120 ms.
- Accuracy: 95%+ on both conditions for most adults.
Children have larger flanker interference scores than adults, and the score decreases through development as selective attention matures. Older adults show somewhat larger interference scores than mid-life adults, consistent with the general pattern of attentional control declining gradually with age.
As with most reaction-time-based attention measures, single-session results are noisy. A 30-trial test produces an interference estimate that varies session to session by 20–40 ms easily, before any real underlying change has occurred. Treat absolute numbers from a single sitting with skepticism.
How to interpret your own result
Three things to keep in mind.
Interference scores are noisy. They're the difference between two noisy reaction-time estimates, so they inherit noise from both. Individual differences in flanker interference do correlate with attention measures in large samples, but a single at-home estimate is not a stable indicator of your attention.
A small interference score isn't necessarily "better." A very small score can reflect strong selective attention; it can also reflect that you weren't responding as fast on congruent trials as you could have. Researchers usually look at congruent RT, incongruent RT, and the difference together, not the difference in isolation.
The Flanker test is not a stand-alone clinical instrument. Like the Stroop test, the Flanker task is used in research on conditions where selective attention is studied — ADHD research, schizophrenia research, aging research — but it is never used as a stand-alone clinical tool. Clinical decisions involve interview, history, and multiple tests.
Related Senwitt content
- The Stroop test is the most closely related task; both measure selective attention and interference control through different forms of stimulus conflict.
- The simple reaction time test is the underlying speed measure on which Flanker interference builds.
- The reading skill page is Senwitt's daily practice surface for sustained reading attention.
If you want a daily practice habit that exercises reading attention alongside the other thinking skills — without measuring you — that's what Senwitt is for.
How the Attention Network Test extended the Flanker paradigm
Fan and colleagues' 2002 design takes the Flanker task and adds cue conditions that separately probe two other attentional functions: alerting (a warning signal before the target) and orienting (a spatial cue indicating where the target will appear). Each ANT trial includes a cue or no cue, then the flanker stimulus.
By cross-tabulating cue type with flanker congruency, you get three scores:
- Alerting = no-cue RT minus double-cue RT. How much a warning signal speeds you up.
- Orienting = center-cue RT minus spatial-cue RT. How much a location hint helps.
- Conflict = incongruent RT minus congruent RT. The flanker interference score from the original Eriksen task.
The ANT became popular in research because it generates three theoretically independent attention scores from one 20-minute session. Whether the three networks are truly independent is a debated point in the cognitive-neuroscience literature, but the ANT is the dominant computerized attention test of the 2000s and 2010s and is still widely used.
A note on online Flanker tests
Most browser-based Flanker tests use the arrow version and run 30 to 60 trials. They're fine for getting a feel for the paradigm. The same caveats apply that apply to any online RT-based test: browser timing latency adds noise to the absolute milliseconds; relative-self trends are meaningful, absolute comparisons against published norms are not.
