Reaction time rises by tens of milliseconds per remembered item

What it is

In standard Sternberg tasks, the reaction-time regression slope for memory-set size is often on the order of tens of milliseconds per item. This is the conditional mean increment associated with one additional remembered item, not a universal timing constant for interfaces or devices.

Why it happens

The slope separates set-size-related processing from fixed stages. If each item has a similar comparison cost, the mean increment is stable; the intercept includes probe perception, response choice, and key movement. Training, encoding material, and speed–accuracy strategy alter both.

Studying it

Use several set sizes, estimate each participant's slope and intercept, and report intervals rather than one grand mean. Keep response mode, discriminability, and presentation time consistent; inspect accuracy and residuals. Slopes can be compared across studies only when paradigm, material, and processing rule are comparable.

Where it stops holding

“Tens of milliseconds” is an order-of-magnitude cue, not a precise design threshold. Small samples, few set sizes, or ignored errors make slopes unstable. In real tasks, search, reading, network, and motor costs often exceed this increment.

Related

• Same group: B1.06.1 Judging membership in a memory set takes longer linearly as the set grows · B1.06.2 Scanning is exhaustive rather than self-terminating · B1.06.3 The finding concerns memory judgment, not visual search · B1.06.5 Similar slopes for yes and no responses indicate exhaustive scanning · B1.06.6 The result concerns small short-term sets, not long-term retrieval · B1.06.7 Using the law to set an interface-option limit is a cross-paradigm error
• Nearby: B1.04 Hick-Hyman law · B1.07 Power law of practice
• Search terms: Sternberg memory scanning · scan slope · reaction time