A Flanker round puts one target in the center and surrounds it with distracting symbols. Your job is deliberately narrow: answer the direction of the center target and ignore the surrounding flankers. The interesting part is not whether arrows are visually difficult. It is whether a surrounding response tendency can pull attention away from the one location that actually determines the answer.
Every trial has a left- or right-facing center target. That target alone determines the correct response. Congruent flankers point the same way. Incongruent flankers point the opposite way. Neutral flankers do not map to either response. The independent verifier reconstructs these display relationships from the materialized trial instead of trusting a prewritten answer flag.
An overall accuracy number can hide where errors happened. Flanker keeps congruent and incongruent accuracy separate, shows neutral accuracy when neutral trials exist, and computes reaction-time medians from correct trials. When enough correct congruent and incongruent trials are available, it also shows the session conflict cost: incongruent median RT minus congruent median RT.
| Level | Trials | Condition mix | Response window | Display |
|---|---|---|---|---|
| Easy | 48 | 32 congruent, 16 incongruent | 900 ms | 4 flankers around one target |
| Standard | 64 | 28 congruent, 28 incongruent, 8 neutral | 700 ms | 4 flankers around one target |
| Challenge | 80 | 28 congruent, 36 incongruent, 16 neutral | 550 ms | 6 flankers around one target |
Difficulty does not change what counts as correct. It changes the number of trials, condition balance, response window and amount of surrounding visual material. Challenge intentionally contains more incongruent than congruent trials, so a player cannot coast on a mostly-agreeing display pattern.
Start with overall accuracy. If many answers are wrong or missed, reaction-time differences are hard to interpret because the run is not yet stable. Next compare congruent and incongruent accuracy. A large accuracy gap means the conflicting context affected correctness, not merely speed. Then look at median correct RT, which is more resistant than a mean to one unusually slow trial.
The conflict cost is shown only when the run contains enough correct observations in both relevant conditions. A positive value means correct incongruent responses were slower in the median than correct congruent responses in that session. A small, zero or negative value is not automatically “better,” and a large positive value is not automatically “bad.” It is a descriptive measurement from one locally generated game block, influenced by speed–accuracy choices, device input, distraction and normal run-to-run variation.
By default, the run is Clean: the target sits in the center but receives no extra cue. The optional Target Spotlight outlines the center target and visually reduces the prominence of the flankers. That can be useful while learning the rule, but the moment Spotlight is enabled during a run, the whole run becomes Assisted.
Assisted results remain in recent history because they are still useful practice records. They cannot replace a Clean best. This separation avoids a common problem in practice tools where a hint changes the task but the result is still compared directly with an unassisted attempt.
Practice uses a seed stored in the URL. If you open Practice without a seed, the page inserts one. The seed chooses a frozen source, shuffles it within explicit run constraints, can flip the global left/right mapping, and selects a display family and spacing. Reopening the same level and seed reconstructs the same concrete block. A Fresh block changes the seed.
The frozen source bank contains 90 project-local sources: 30 Easy, 30 Standard and 30 Challenge. It does not use a scraped clinical test, paid assessment bank, competitor trial sequence, population norm table or remote test API. The generator constrains the condition runs and target-direction runs before a source is accepted. The separate verifier checks those contracts again.
Suppose the center target points right. In a congruent trial, the surrounding directional symbols also point right. In an incongruent trial, the surrounding directional symbols point left even though the correct answer is still right. In a neutral trial, the surrounding symbols have no left/right answer meaning. The game can render this relationship with arrows, chevrons or triangle-like symbols and with different spacing, but the center-target response rule never changes.
This distinction also separates Flanker from the existing Stroop specialist. Both create competing information, but Stroop asks you to select an attribute such as ink color while suppressing a conflicting word response. Flanker places the response conflict spatially around a center target. DailyBrainArc keeps them as separate products and does not present either one as a clinical instrument.
On desktop, use Left Arrow or A for a left center target and Right Arrow or L for a right center target. On touch devices, the two large response buttons are the primary controls. H toggles Target Spotlight. Enter starts a new run when the current run is not active. During an active mobile run, the bottom navigation is hidden so the response controls and stimulus remain the visual priority.
The response timer is part of the level contract, so a missed deadline is recorded as an error rather than silently waiting forever. That keeps a 48-, 64- or 80-trial block finite and makes deterministic replays comparable in structure.
Recent runs, Clean bests and specialist Daily history are stored in the browser through the DailyBrainArc local state. Daily Flanker has its own history and streak inside flankerStats.daily. It does not write to the Today’s 6 day records, and adding this specialist does not alter the frozen Today’s 6 lineup.
If browser storage is cleared, that local specialist history can disappear. The page does not need a remote puzzle feed to create a Practice or Daily block. That local-first boundary is intentional: reproducibility comes from frozen project data plus deterministic materialization, not a server-side assessment score.
If you want a different interference mechanism, try Stroop. For response withholding rather than left/right distractor conflict, use Go / No-Go. For raw response speed without distractor conflict, use Reaction Time. Keeping these tasks separate makes each result easier to understand.