Atkinson–Shiffrin
Memory Battery — Free Online
A short browser-based memory task built on the classic multi-store model. You complete three paradigms in turn — paired-associate learning, a brief serial display, and free recall — and see how your performance maps onto the sensory register, short-term store, and long-term store. No signup needed to take it.
What you'll get
The battery runs three classic memory paradigms back to back, each tied to a part of the Atkinson & Shiffrin (1968) multi-store model:
- Paired-associate learning — your encoding accuracy and a lag-decay slope (how accuracy changes as the gap between studying a pair and being tested on it grows), an indicator of transfer from short-term to long-term storage
- Serial display — your partial-report accuracy for letters shown very briefly, with an exploratory exposure-duration slope tied to the fast-fading sensory register
- Free recall — your total recall plus primacy and recency effects, the classic serial-position signature of separate long-term and short-term stores
Every figure is a raw, within-subject score. There is no percentile, no T-score, and no Higher/Average/Lower band, because no published reference table matches this exact list length, presentation rate, and item type. The primacy and recency effects are measured against your own total recall, so they stand on their own. A single session is indicative, not diagnostic.
Common questions
What is the multi-store model?
Atkinson and Shiffrin (1968) described memory as a flow through three stores: a brief sensory register, a limited-capacity short-term store, and a durable long-term store, linked by control processes such as rehearsal. Each of the three sub-tasks here probes a different part of that architecture.
What are the primacy and recency effects?
In free recall, people tend to remember the first few items well (the primacy effect, attributed to extra rehearsal into long-term storage) and the last few items well (the recency effect, attributed to the short-term store still holding them at test). The dip in the middle produces the familiar U-shaped serial-position curve.
How long does it take?
Usually about eight minutes for all three paradigms. There are no adaptive stopping rules — you complete a fixed set of trials in each sub-task.
Why is there no percentile or IQ-style score?
We deliberately do not attach borrowed test norms to this task. Published memory norms come from specific, standardized administrations with their own list lengths, timing, and item types; applying them to a different browser task would misrepresent your result. We report honest raw scores and within-subject effects instead.
Can I retake it?
Yes, as many times as you like. Memory scores are noisy and a little practice effect across the first attempts is normal.