Task switching test: method

What it measures

Cognitive flexibility (task-set switching): speed–accuracy efficiency when the rule changes unpredictably, plus switch cost (switch versus repeat trials) and mixing cost (repeat trials in the mixed block versus single-task blocks).

Procedure

Stimuli: a blue or orange circle or square (all 4 combinations, balanced within each block). The left answer means blue or circle, the right answer orange or square (keys F or left arrow, J or right arrow, or two on-screen buttons). Each trial: the cue word (COLOR or SHAPE, in the page language) appears alone for 600 ms, then the shape appears under it; the reaction time and the 3,000 ms response window start at the shape. After an error or a timeout, 600 ms of feedback. 400 ms between trials. Practice (default for new players, skippable): 6 colour, 6 shape and 8 mixed trials. Counted blocks in fixed order: 12 colour-only, 12 shape-only, 48 mixed, with a 1.5 s 'next' screen between blocks. In the mixed block the first task is random, half of the 47 transitions (24) are switches in a random order, and no task runs more than 3 trials in a row. Leaving the browser tab ends the run. About 4 minutes.

Parameters

Scoring

A trial is correct when the right key is pressed at 150 ms or later; a timeout is an error with a reaction time of 3,000 ms. Primary: rate correct score in the mixed block = 60,000 x correct answers / sum of all reaction times (ms) in the block, in correct answers per minute; higher is better. Also: the same rate for the single-task blocks; accuracy for mixed, single-task, switch and repeat trials; timeouts. Switch cost = mean RT of correct switch trials minus mean RT of correct repeat trials (mixed block, the previous trial also correct, first trial excluded). Mixing cost = mean correct repeat RT minus mean correct single-task RT (first trial of each single-task block excluded). Both costs are shown only with at least 5 usable switch and 5 usable repeat trials. The rate score is the main number because difference scores such as switch costs have low reliability. No player comparison is shown.

Valid-run rules

The server regenerates every trial from the run's seed; exactly 72 counted answers are needed. The server rebuilds every run from the raw log (and the run's seed where there is one) and computes the score itself; anything the page says about right or wrong is ignored. A log that breaks the game's rules is rejected, and nothing is stored. Stored runs can carry quality flags that are never shown and never block a result: 10 or more runs from one IP address, a missing visitor id, tab switches, and automation signals (automated browser, impossible speed, no input events, machine-regular timing).

Differences from the classic task

Rogers and Monsell (1995) used predictable alternating runs; this game uses random cued switching (Meiran 1996), with one fixed 600 ms cue interval. Lab versions usually have hundreds of trials and vary the cue interval; here the mixed block has 48 trials. About half of the trials are response-congruent (blue circle, orange square: both rules give the same key); they are not analysed separately.

Limitations

Few trials, so switch and mixing costs are noisy and shown as rough guides only. Keyboard versus touch input changes speed. The cue is a word, so reading in the page language is part of the task. Practice effects across runs. Unsupervised setting. Not a clinical test.

Languages and review status

Colour and shape stimuli are language-independent; the cue words (COLOR, SHAPE) and instructions are shown in all site languages (English plus 99 other languages).