The Atkinson–Shiffrin Model and the Lag Effect
Why does a fact you learned thirty seconds ago vanish while another survives? The 1968 multi-store model gave memory research its most influential answer — and a task simple enough to test it on yourself.
The multi-store model
Sensory register
Incoming information persists for a fraction of a second in a modality-specific sensory store. Whatever is attended to moves on; the rest is lost.
Short-term store (STS)
A limited-capacity rehearsal buffer. Atkinson and Shiffrin modelled it as holding only a few items at once: each new item entering the buffer can displace one already there.
Long-term store (LTS)
While an item sits in the buffer, information about it is gradually copied into a durable long-term store. The longer it stays buffered, the more is transferred — and the better later recall becomes.
The lag effect
What is a study–test lag?
In the continuous paired-associate procedure, study exposures and recall probes are interleaved in one long stream. The lag is the number of intervening events between studying a pair and being tested on it. At lag 1 the pair is almost certainly still in the short-term buffer; at lag 16 it has long since been displaced, so recall depends entirely on what transferred to the long-term store.
What does the model predict?
A steep fall over the first few intervening events (buffer displacement) that flattens toward an asymptote set by long-term transfer. The theoretical curve used on this site — roughly 65% correct at lag 1 declining to about 28% at lag 16 — follows the published predictions of Atkinson and Shiffrin (1968).
What does a flat curve mean?
Recall that barely declines with lag suggests items are being encoded into long-term storage almost immediately — for example through effortful elaboration — rather than being held only in the buffer. That is a legitimate strategy difference, not an error.
How should scores from this task be read?
Results reflect performance relative to Atkinson-Shiffrin (1968) theoretical predictions and should be interpreted directionally. This task uses a small trial count per lag condition; individual scores are subject to variability and should not be used in isolation for clinical conclusions.
Source
Atkinson, R. C., & Shiffrin, R. M. (1968). Human memory: A proposed system and its control processes. In K. W. Spence & J. T. Spence (Eds.), The psychology of learning and motivation (Vol. 2, pp. 89–195). Academic Press.