Serial Position Effect

The serial position effect is the robust finding that, when people are asked to recall a list of items in any order, items at the beginning of the list (primacy) and items at the end (recency) are remembered better than items in the middle, producing a characteristic U-shaped recall curve. The pattern is one of the most reliable observations in memory research and has been reproduced for words, digits, pictures, and a wide range of materials across many decades. Although Hermann Ebbinghaus described related order effects in his pioneering 1885 studies of serial learning — where items in the middle of a list of nonsense syllables were learned more slowly than items at the beginning or end — the canonical free-recall serial position curve is the demonstration reported by Murdock (1962).

Beyond its empirical reliability, the serial position effect has been central to memory theory because the two halves of the curve appear to come apart in informative ways. Primacy and recency can be selectively disrupted by different variables, a double dissociation that historically provided some of the strongest evidence for the distinction between a brief, limited-capacity store and a more durable long-term memory (Glanzer & Cunitz, 1966; Atkinson & Shiffrin, 1968). Subsequent work has both refined that interpretation and produced serious alternatives in which a single principle — the relative distinctiveness of items in time — generates both effects without invoking separate memory stores (Brown et al., 2007).

The Free-Recall Serial Position Curve

The canonical demonstration of the effect comes from Murdock's (1962) free-recall experiments. Participants heard lists of unrelated words read at a steady rate and were then asked, immediately after the last word, to recall as many words as possible in any order. Plotting the proportion of recalls against the position at which each word had appeared in the list produces the U-shaped curve that now appears in almost every memory textbook: a sharp rise of recall for the first few items, a flat low region across the middle of the list, and a steeper rise across the last few items. The shape is remarkably consistent: it appears across list lengths, presentation rates, and a wide range of materials, which is why it serves as a benchmark phenomenon against which theories of memory are tested.

Primacy and Recency

The two halves of the curve have long been attributed to different processes. The primacy effect is usually attributed to the greater amount of rehearsal that early items receive: when a list begins, early items can be rehearsed alone or against just a few competitors and so receive more opportunities to be encoded into long-term memory. As more items arrive, rehearsal time per item decreases. The recency effect, by contrast, has classically been attributed to the last few items still being held in short-term memory (or working memory) at the moment of recall and so being available for immediate retrieval.

The strongest support for this dual-process account is a striking dissociation. If participants must perform a brief, demanding distractor task — typically counting backwards by threes for around 30 seconds — between the end of the list and recall, the recency effect is essentially eliminated, while the primacy effect and middle of the curve are left intact (Glanzer & Cunitz, 1966; see also the earlier demonstration by Postman & Phillips, 1965). On the dual-store interpretation, the distractor task displaces the recency items from short-term memory while leaving long-term traces untouched. Conversely, variables that affect encoding into long-term memory — slower presentation, deeper or more meaningful processing, greater familiarity — enhance primacy and the middle of the curve without much affecting the last few items. This double dissociation has been treated as a textbook example of two memory systems revealing themselves in one task.

Theoretical Significance

The serial position effect provided central evidence for the modal (multi-store) model of memory proposed by Atkinson and Shiffrin (1968), in which information moves from a sensory register through a limited-capacity short-term store and, with rehearsal, into a more permanent long-term store. The dissociability of primacy and recency mapped naturally onto the model: primacy reflected successful transfer to the long-term store, while recency reflected items still active in the short-term store at retrieval. The combination of robust empirical curve and theoretically interpretable dissociation made the serial position effect one of the most influential phenomena in the cognitive revolution and a centrepiece of how memory has been taught ever since.

Beyond the Modal Model

Subsequent findings complicated the simple dual-store story. Craik (1970) demonstrated the negative recency effect: when recall is delayed by a filled distractor task, the items from the end of the list — which had been recalled best on the immediate test — are now recalled worse than primacy items. The dual-store account explains this by noting that recency items, retrieved easily from the short-term store on the immediate test, never had to be encoded deeply into long-term memory, while primacy items had been thoroughly encoded and so survive the delay.

A more serious challenge came from the long-term recency effect. Bjork and Whitten (1974) used the continuous-distractor paradigm, in which a brief distractor task is inserted between every pair of list items and again at the end. The simple short-term-store account predicts that the end-of-list distractor should wipe out recency, just as it does in Glanzer and Cunitz's procedure. In fact, a clear recency effect remained — and indeed reappeared at much longer time scales, including over hours and days in everyday memory. Recency, on this evidence, is not the property of a single short-lived store but a more general feature of memory retrieval (Greene, 1986).

These findings motivated a class of distinctiveness or temporal-context accounts that explain serial position effects without invoking separate memory stores. The Temporal Context Model of Howard and Kahana (2002) treats memory retrieval as a process in which a slowly-changing internal context vector cues items associated with similar contexts; primacy and recency emerge from the structure of this context representation. The SIMPLE model of Brown et al. (2007) goes further, deriving both effects from a single principle — items are easier to recall when they are temporally distinctive relative to their neighbours — applied across very different time scales. Defenders of the modal account have answered with computational analyses showing that recency effects can still be modelled by a working-memory-like buffer in conjunction with long-term retrieval (Davelaar et al., 2005). The contemporary theoretical landscape is therefore one of active debate rather than a settled dual-store consensus.

Applications

The serial position effect has practical implications across many domains. In education, it suggests that important material is best placed at the beginning and end of a lecture rather than in the middle. In marketing and consumer research, items at the beginning or end of a list of options receive more attention and are more likely to be remembered. In legal settings, the order of witnesses or arguments can influence how well jurors recall them, and political speakers and presenters intuitively exploit primacy and recency by placing their strongest points at the opening and close.

The effect also has a notable clinical application: the shape of an individual's serial position curve is a sensitive marker of certain memory disorders. Patients with early Alzheimer's disease show a characteristic reduction of the primacy effect with relatively preserved or even exaggerated recency on immediate free recall — a pattern attributed to early hippocampal decline that impairs encoding into long-term memory while leaving short-term retrieval relatively intact (Bayley et al., 2000). This profile has been studied as an early indicator of cognitive decline and a way to differentiate Alzheimer's-type dementia from other patterns of memory impairment.

A current and somewhat surprising application has emerged in artificial intelligence. Large language models, when asked to retrieve specific information from long contexts, show a U-shaped accuracy curve over the position of the relevant information in the input — performance is high when the answer sits near the beginning or end of the context window and drops sharply when it is buried in the middle (Liu et al., 2024). The phenomenon, now widely referred to as "lost in the middle," strikingly mirrors the human serial position effect despite arising in systems whose self-attention mechanism is in principle capable of accessing any position equally well. It has become a substantial topic of study in its own right, both as a practical engineering problem for retrieval-augmented systems and as an unexpected parallel between human and machine memory.

Key Researchers

The following researchers made foundational contributions to the discovery, interpretation, and modern theory of the serial position effect, ordered alphabetically by surname.

• Richard C. Atkinson — President Emeritus, University of California; Professor Emeritus of Cognitive Science and Psychology, UC San Diego (previously Professor of Psychology, Stanford University). Co-architect with Shiffrin of the modal (multi-store) model of memory, whose short-term/long-term-store dichotomy provides the dominant theoretical interpretation of primacy and recency (Atkinson & Shiffrin, 1968).
  UC San Diego
• Robert A. Bjork — Distinguished Research Professor Emeritus of Psychology, University of California, Los Angeles. With William B. Whitten, established the long-term recency effect through the continuous-distractor paradigm, showing that recency persists across delays and so cannot be the property of a single short-lived store (Bjork & Whitten, 1974).
  Google Scholar · UCLA Psychology
• Murray Glanzer (1922–2025) — Professor Emeritus of Psychology, New York University. With Anita R. Cunitz, demonstrated that a filled delay between list and recall eliminates the recency effect while leaving primacy intact, providing the canonical double-dissociation evidence for separate short- and long-term memory stores (Glanzer & Cunitz, 1966).
• Michael J. Kahana — Professor of Psychology, University of Pennsylvania; director of the Computational Memory Lab. With Marc W. Howard (Boston University), formulated the Temporal Context Model, the leading modern account in which a continuously-changing internal context representation produces serial position and contiguity effects without separate stores (Howard & Kahana, 2002).
  Google Scholar · Computational Memory Lab
• Bennet B. Murdock Jr. (1925–2022) — Professor of Psychology, University of Toronto. Provided the canonical demonstration of the free-recall serial position effect and pioneered mathematical models of human memory (Murdock, 1962).
• Ian Neath — Professor of Psychology, Virginia Tech. With Gordon D. A. Brown (University of Warwick) and Nick Chater (Warwick Business School), developed and tested the SIMPLE distinctiveness model, which derives serial position effects from a single relative-distinctiveness principle and offers the principal alternative to dual-store accounts (Brown et al., 2007).
  Google Scholar · Virginia Tech
• Richard M. Shiffrin — Distinguished Professor and Luther Dana Waterman Professor of Psychological and Brain Sciences and Cognitive Science, Indiana University Bloomington. Co-architect with Atkinson of the modal model and developer of the SAM and REM mathematical/computational frameworks that frame later interpretations of serial position effects (Atkinson & Shiffrin, 1968).
  Google Scholar · Indiana University

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, pp. 89–195). Academic Press. https://doi.org/10.1016/S0079-7421(08)60422-3
2	Bayley, P. J., Salmon, D. P., Bondi, M. W., Bui, B. K., Olichney, J., Delis, D. C., Thomas, R. G., & Thal, L. J. (2000). Comparison of the serial position effect in very mild Alzheimer's disease, mild Alzheimer's disease, and amnesia associated with electroconvulsive therapy. Journal of the International Neuropsychological Society, 6(3), 290–298. https://doi.org/10.1017/S1355617700633040
3	Bjork, R. A., & Whitten, W. B. (1974). Recency-sensitive retrieval processes in long-term free recall. Cognitive Psychology, 6(2), 173–189. https://doi.org/10.1016/0010-0285(74)90009-7
4	Brown, G. D. A., Neath, I., & Chater, N. (2007). A temporal ratio model of memory. Psychological Review, 114(3), 539–576. https://doi.org/10.1037/0033-295X.114.3.539
5	Craik, F. I. M. (1970). The fate of primary memory items in free recall. Journal of Verbal Learning and Verbal Behavior, 9(2), 143–148. https://doi.org/10.1016/S0022-5371(70)80042-1
6	Davelaar, E. J., Goshen-Gottstein, Y., Ashkenazi, A., Haarmann, H. J., & Usher, M. (2005). The demise of short-term memory revisited: Empirical and computational investigations of recency effects. Psychological Review, 112(1), 3–42. https://doi.org/10.1037/0033-295X.112.1.3
7	Glanzer, M., & Cunitz, A. R. (1966). Two storage mechanisms in free recall. Journal of Verbal Learning and Verbal Behavior, 5(4), 351–360. https://doi.org/10.1016/S0022-5371(66)80044-0
8	Greene, R. L. (1986). Sources of recency effects in free recall. Psychological Bulletin, 99(2), 221–228. https://doi.org/10.1037/0033-2909.99.2.221
9	Howard, M. W., & Kahana, M. J. (2002). A distributed representation of temporal context. Journal of Mathematical Psychology, 46(3), 269–299. https://doi.org/10.1006/jmps.2001.1388
10	Liu, N. F., Lin, K., Hewitt, J., Paranjape, A., Bevilacqua, M., Petroni, F., & Liang, P. (2024). Lost in the middle: How language models use long contexts. Transactions of the Association for Computational Linguistics, 12, 157–173. https://doi.org/10.1162/tacl_a_00638
11	Murdock, B. B. (1962). The serial position effect of free recall. Journal of Experimental Psychology, 64(5), 482–488. https://doi.org/10.1037/h0045106
12	Postman, L., & Phillips, L. W. (1965). Short-term temporal changes in free recall. Quarterly Journal of Experimental Psychology, 17(2), 132–138. https://doi.org/10.1080/17470216508416422