Pause

A spoken instruction plays with the text streaming beneath it, and the last clause changes the answer: "Click the red square. Unless it is next to a circle. Then click the blue one." A third of the instructions have no twist, so waiting is not a guaranteed win and cannot be gamed by never listening to the start.

Parameters

40 to 60 items per session; the twist position varies so it cannot be anticipated; the answer controls are locked until the voice stops in weeks one and two, and the lock fades in three steps across weeks three to six.

Scored

Three buckets - acted before the end, acted after but wrong, correct. The first bucket is the number that matters. An early click shows you exactly how early: "you clicked 1.2 s before the end."

Why this shape

There is no off-the-shelf paradigm for "wait for the whole instruction", so it is assembled from pieces that do exist: the Token Test's decoy items reveal whether a listener attended only to the start or only to the end, and the fading lock is a delayed-response scaffold. The assembly is ours and is graded accordingly.

Hold the space bar. Four boxes on screen. After an unpredictable wait one of them flashes; release and hit that box. Let go early and nothing happens except a wasted trial. Later blocks add distractor flashes that are not the target.

Parameters

Two warm-up blocks of 20 and four blocks of 40; wait uniformly random between 2 and 10 seconds; target on screen 100–150 ms (the laboratory version uses 32–64 ms, which is too short for a browser).

Scored

Premature releases. Full points for fast-and-correct, half for average, a loss for premature - the reward scheme from the original task.

Why this shape

This is the cleanest laboratory model of acting before the cue, and it is the measure the whole profile is organised around. Its premature-release rate also serves as the validity check for every other drill in the session.

Arrows point left or right and you press the matching key. On a quarter of trials a sound says stop. The stop signal arrives later when you succeed and earlier when you fail, so it always sits at the edge of what you can do.

Parameters

About 200 trials in four blocks, 7–10 minutes; stop-signal delay staircases ±50 ms and is clamped between 50 and 1,150 ms; inter-trial gap random 500–1,500 ms so anticipation does not pay.

Scored

Stop-signal reaction time by the integration method. A session is discarded if the stop rate falls outside 25–75%, or if failed-stop reaction times exceed go reaction times.

Why this shape

These are the consensus parameters, and the block feedback exists to catch the classic cheat: slowing down deliberately so the signal always arrives in time. Note the honest limit - this task is strong evidence for the trained skill and near cousins, and it does not train waiting.

Faces appear one after another. Press for every face except the one with the rare expression. Faces rather than shapes, deliberately.

Parameters

500 ms on, 500 ms off; three go faces to every one hold; 150–300 trials. The gap shortens and the go ratio rises as skill rises, both of which make withholding harder.

Scored

Commission errors, omissions, raw reaction time and its variability. Commission error rate is one of the two most reliable numbers in the product.

Why this shape

Emotional content, because that is where transfer showed up: training with emotional material moved emotion regulation where neutral material did not. Running the same paradigm on faces costs nothing and is the version with the better record.

A voice says "head" and the correct key is toes; "left" means right. The mapping flips every block.

Parameters

Eight blocks of 20, a new mapping each block.

Scored

Efficiency - time divided by accuracy - rather than accuracy alone.

Why this shape

Simon Says is harder than "bear and dragon" because the same voice gives both the do and the don't cues, which is the realistic case. The efficiency score exists because bright teenagers hit the ceiling on raw accuracy: 39% of US twelve-year-olds do so on the standard measure.

A wrong answer given faster than your own median minus half a standard deviation is a rush error: its own icon, a bigger point loss, a sound, and a one-second freeze.

Parameters

The baseline median comes from the first two sessions, which are burn-in and are never counted.

Scored

The ratio of rush errors to careful errors is shown to you as a style gauge. It is not a progress metric - see page eleven, where its reliability is graded exploratory.

Why this shape

Reaction-time research usually punishes being too slow. Here the sign flips, because the costly error is being too fast. The error is made unmissable on purpose: reduced post-error slowing in attention profiles appears to come mainly from failing to notice the error, so noticing is what the design buys.