A Vigilance Task in Your Budget

Your battery covers the intoxicant domains well but is blind to fatigue's signature — erratic lapses in sustained attention. This lesson designs a vigilance micro-task that fits your under-2-minute phone budget, and shows the two or three things you must not get wrong.

Recall first

What the PVT actually is (stripped to essentials)

The Psychomotor Vigilance Task is deceptively simple: a stimulus appears at random intervals; you respond as fast as you can; it records your reaction time over and over. That's it. Its power comes from three properties you must preserve if you shrink it:

• Random inter-stimulus interval (ISI). The gap between stimuli varies unpredictably (classically 2–10 s). Randomness is non-negotiable — it stops users from anticipating and building a rhythm, which is what forces genuine sustained attention.
• No learning curve. There's nothing to strategize, so scores barely improve with practice[1] — the property your richer tasks lack (Lesson 3).
• No ceiling. Reaction time always has headroom; you can't "max it out," so impairment always has room to show.

The metrics — this is where the sensitivity lives

Don't just log mean RT. The evidence on what best discriminates sleep loss[1] points to:

• Response speed mean(1/RT) — the reciprocal of reaction time. More normally distributed and more sensitive across the whole range than raw mean RT; it's the recommended primary metric.[1]
• Variability SD or CoV of RT — the direct readout of wake-state instability. Often the earliest sign.
• Lapses count(RT ≥ 500 ms) — momentary attention failures. In a very short task few accumulate, so pair with "minor lapses" (≥ 355 ms) to keep signal.
• False starts RT < 100 ms or pre-stimulus — responding before you could have seen the stimulus. Flags guessing/rhythm-tapping; count them separately.

The budget problem — and an honest answer

The full PVT is 10 minutes. The validated brief version (PVT-B) is 3 minutes and keeps much of the sensitivity to sleep loss.[2] You want your whole battery under ~2 minutes, so a vigilance slice of ~45–60 s. That's shorter than anything formally validated — which is fine if you treat the shortening as a hypothesis to test, not a given. At least one study found a 3-minute version already diverging from the 10-minute reference under some conditions.[3] The shorter you cut, the more you must validate.

A concrete starting spec

A defensible v1 to prototype and then validate — the full version lives in the Vigilance Micro-Task Spec reference card:

Duration

~60 s (fits budget; validate against a 3-min PVT-B)

Stimulus

One salient target appears after a random gap; tap ASAP

ISI

1–4 s, uniform random — never fixed

Trials

~15–20 (as many as fit)

Primary

mean(1/RT) and CoV of RT

Secondary

lapses ≥500 ms, minor lapses ≥355 ms, false starts

Feed into

the per-user baseline + RCI from Lesson 3

Gamify carefully — the core is sacred

You're building "games," and that instinct is good for engagement — but the vigilance task's validity lives in exactly the properties gamification tends to destroy. The rule: decorate around the core, never alter the core.

• Keep: random ISI, simple immediate response, RT as the measurement, boring-by-design so there's nothing to master.
• Safe to add: a theme/skin, per-trial RT feedback ("312 ms!"), a gentle progress bar.
• Avoid: anything that rewards strategy or lets the player learn a rhythm — a "beat your combo" mechanic, predictable timing, or a difficulty ceiling that makes failure impossible. Each of those quietly reintroduces the practice and ceiling effects from Lesson 3.

Before you trust it — validate

Three checks, in order:

1. Concurrent validity: do scores on your 60-s version correlate with a 3-min PVT-B in the same sitting?
2. Test–retest reliability: do un-impaired users score consistently day to day (aim ICC > 0.8)? Without this you can't separate impairment from noise.[2]
3. Sensitivity to the real thing: does it drop after real sleep loss (e.g., a rested vs. sleep-restricted comparison)? That's the only test that matters in the end.

Check yourself

Your single win

You can now spec a vigilance micro-task that closes your fatigue blind spot inside a 60-second budget: random ISI, dead-simple response, score on 1/RT and variability, decorate but never alter the core, and validate the short version against a longer one. Combined with Lessons 1–3, you now have the full loop — the right domains, the fingerprints, a trustworthy decision rule, and the one task you were missing.

References

1. [1] Basner & Dinges, Maximizing Sensitivity of the PVT to Sleep Loss, SLEEP (2011).
2. [2] Basner, Mollicone & Dinges, Validity and Sensitivity of a Brief PVT (PVT-B) (2011).
3. [3] The 3-Minute PVT Demonstrates Inadequate Convergent Validity…, Frontiers in Neuroscience (2022).