Mathematical Psychology
About

Serial Position Models

Serial position models explain the characteristic U-shaped curve of recall probability across list positions, capturing the enhanced memory for early items (primacy) and recent items (recency) through distinct mathematical mechanisms.

P(recall | pos i) = P(STS | i) + [1 − P(STS | i)] · P(LTS | i)

The serial position effect, first documented by Hermann Ebbinghaus (1885), refers to the robust finding that items at the beginning and end of a studied list are recalled better than items in the middle. Mathematical models of this phenomenon have been central to memory theory, as the shape and dynamics of the serial position curve provide strong constraints on the underlying memory architecture.

The Dual-Store Account

The classic Atkinson and Shiffrin (1968) model explains the serial position curve through two memory stores. Primacy arises because early items receive more rehearsal and are therefore more strongly encoded in long-term store (LTS). Recency arises because the last few items remain in a limited-capacity short-term store (STS) at the time of test:

Dual-Store Recall Probability P(recall | i) = P(from STS | i) + [1 − P(from STS | i)] · P(from LTS | i)

This model predicted that a filled delay between study and test should eliminate recency (by displacing items from STS) while leaving primacy intact, a prediction confirmed by the classic Glanzer and Cunitz (1966) experiments.

Ratio Rule and Long-Term Recency

A challenge to dual-store models came from the discovery of long-term recency: recency effects that persist over delays of days or weeks when the temporal spacing of items is proportionally increased. Bjork and Whitten (1974) proposed the ratio rule, which states that discriminability between items depends on the ratio of the inter-presentation interval (IPI) to the retention interval (RI):

Ratio Rule Discriminability ∝ IPI / RI

When the ratio IPI/RI is held constant, recency effects appear regardless of the absolute time scale. This finding motivated single-store temporal distinctiveness models such as SIMPLE.

Buffer Models

Formal buffer models (Raaijmakers & Shiffrin, 1981; Davelaar et al., 2005) model the STS as a limited-capacity buffer of size k. Items enter the buffer sequentially; when the buffer is full, an existing item is displaced probabilistically to make room. Primacy arises because early items spend more time in the buffer (and receive more transfer to LTS), while recency arises because the last k items have high probability of remaining in the buffer at test.

Modern Contextual Accounts

Contemporary models such as TCM and CMR explain both primacy and recency through a gradually changing temporal context signal. Recency reflects the similarity between the context at test and the context at encoding for recent items. Primacy may arise from a novelty signal that enhances encoding for the first items encountered in a new context. These models account for both short-term and long-term recency within a unified framework, without requiring separate memory stores.

The Ubiquity of Serial Position Effects

Serial position effects appear across an extraordinary range of domains: word lists, musical sequences, consumer product evaluations, legal testimony, and even the order of entries on ballots. Any model of memory must account for these effects, making the serial position curve one of the most important benchmarks in mathematical psychology.

Related Topics

References

  1. 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).
  2. Glanzer, M., & Cunitz, A. R. (1966). Two storage mechanisms in free recall. Journal of Verbal Learning and Verbal Behavior, 5, 351-360.
  3. Bjork, R. A., & Whitten, W. B. (1974). Recency-sensitive retrieval processes in long-term free recall. Cognitive Psychology, 6, 173-189.

External Links