Serial Position Effect in Studying: The Recency Trap

The serial position effect is the finding that people recall items from the start and the end of a list better than items from the middle. On a delayed test the end-of-list advantage reverses, which is called negative recency, and an exam is a delayed test.

Every study below used word lists or an ordering task, except one 2024 conference paper. None of these abstracts reports a trial of how to order a revision session, so read the practical steps here as our advice rather than as something the studies measured.

What is the serial position effect?

It is the shape you get when you plot how often an item is recalled against where it sat in the list. Murdock (1962) described that shape for immediate free recall of word lists in the Journal of Experimental Psychology: “The serial position curve is characterized by a steep, possibly exponential, primacy effect extending over the 1st 3 or 4 words in the list, an S-shaped recency effect extending over the last 8 words in the list, and a horizontal asymptote spanning the primacy and recency effect.”

Three regions, then: a steep primacy effect over the first three or four words, an S-shaped recency effect over the last eight, and a flat line spanning the two. The words in the middle sit on that flat line.

The pattern reaches past lists you have only just studied. Kelley, Neath and Surprenant (2015) open with a general claim: “Serial position functions with marked primacy and recency effects are ubiquitous in episodic memory tasks.” They then went looking for it in general knowledge, asking people to put exemplars from a category in order along a dimension. Their result: “In all cases, a serial position function was observed: People were more accurate to order the youngest and oldest actors, the lightest and heaviest animals, the shortest and tallest basketball players, the smallest and largest countries, and the smallest and largest planets, relative to intermediate items.”

That task asked people to order things they already knew, with nothing revised for it, so it widens where the shape turns up without touching the question of how to study.

Why do the ends of a list do better?

Kelley, Neath and Surprenant give their own account of it: “The serial position function arises because the first and last items enjoy a benefit of having no competitors on 1 side and therefore have enhanced distinctiveness relative to mid-dimension items, which suffer by having many competitors on both sides.” On that reading, an item at either edge has neighbours in one direction only and is easier to pick out, while an item in the middle is crowded from both sides.

That is the authors’ interpretation of their own ordering data, so hold it loosely. Their abstract reports an ordering task and nothing about revision, which makes the parallel to the middle of a study session ours rather than theirs.

What happens when the test is delayed?

The recency half of the curve turns over. Kuhn, Lohnas and Kahana (2019) state it in the first line of their abstract: “The well-known recency effect in immediate free recall reverses when subjects attempt to recall items studied and tested on a series of prior lists, as in the final-free-recall procedure (Craik, 1970).” The consequence follows straight after: “In this case, the last few items on each list are actually remembered less well than are the midlist items.”

Sheaffer and Levy (2022) report the same direction on a recognition test. Writing in Memory and Cognition, they put it this way: “While items learned immediately before testing are generally remembered better than prior items in a study list, in delayed testing this relationship is reversed, yielding a negative recency effect.” Their delayed recognition method was word lists again: “we examined performance of 169 participants on a delayed recognition test following multiple sessions requiring the study and immediate free recall testing of 16 lists of 16 words”.

Kuhn and colleagues found that reversal at one level only. In the same study they “replicate the negative-recency effect for the within-list serial position curve and the positive-recency effect for the between-list serial position curve.” Position inside a list turned over, while the list studied most recently still showed a recency advantage on that same delayed test.

Why does the end of a list lose its advantage?

Both papers point at spacing. Kuhn, Lohnas and Kahana propose it as the mechanism: “These results suggest that negative recency may reflect the beneficial effects of spaced practice, in which end-of-list items recalled early constitute massed repetitions and end-of-list items recalled late are spaced repetitions.” On their account, an end-of-list item that somebody recalls early gets its second exposure almost on top of the first, and one recalled later gets a gap in between.

Sheaffer and Levy tested that account and reported: “This revealed a strong effect of the amount of spacing between initial study position and initial free recall position on the degree of negative recency, supporting the spacing account.”

Spacing is the lever the rest of our study-techniques writing turns on, which makes this a quieter result than it first sounds. It is one of the four desirable difficulties Robert Bjork named, and the practical layer under it is a calendar: our guide to building a spaced repetition schedule sets out the 2357 method, where you review on day 2, day 3, day 5 and day 7.

Should you save the most important topic for last?

Our answer is no, and the delayed test is the reason. The advantage you feel over the last thing you read is the within-list recency effect, and that is the part that turned over in both of the delayed studies above. The between-list advantage that survived in Kuhn and colleagues is the counterweight, so whether a revision topic behaves like a position in a list or like a list of its own is the weakest join in this argument, and the join is ours. An exam sits days or weeks after the studying, which puts it on the delayed side of the same comparison.

Keep the limits in view. Both delayed studies used lists of words, the positions they varied were positions inside a list, and neither abstract reports a revision timetable or a classroom. The step from one to the other is ours.

Our guide on how to cram for an exam lands in the same place from a different direction. Cramming is yes for tomorrow and no for anything after that, because material learned in one burst fades fast. Parking the topic that matters most in the final hour of the final night stacks both problems on top of each other.

What can you do with this when you study?

Four habits, extended from list studies. None of these papers measured them.

1. Give the middle extra passes. On an immediate test the middle of a list is the flat part of the curve, the stretch with neither effect helping it. So the material that lands in the middle of a session is the material to come back to, and marking it while you are still in the session beats trying to spot it afterwards.
2. Treat the topic you saw last as unfinished. Inside a list, that is the position whose advantage turned over once the test was delayed, and reading a topic as a position is our step. Whatever it feels like at the end of a session, check it again before you call it done.
3. Change the order between passes. What sat in the middle on Monday can sit at the start on Wednesday. No abstract here reports a study that varied the order across passes, so tying it to serial position is our step. Mixing the order is well-trodden ground on its own, and our guide to the interleaving study method covers what it buys.
4. Settle it with retrieval. Close the notes and try to produce the answer, since the position that feels most secure at the end of a session is, on the reading above, the one to distrust. Our comparison of active recall and spaced repetition sets out why the retrieval action and the review timing compound when you run both together.

Has the serial position effect been tested as a study strategy?

One conference paper put it alongside other techniques and reported that it did not move the result. Atas and colleagues (2024) presented A Fight against the Forgetting Curve through the International Association for Development of the Information Society. Its abstract says: “The study compared the effectiveness of these strategies, revealing that self-testing and spaced repetition improve learning and combat forgetting.” Then: “However, in this context, the serial position effect did not show a noticeable impact on learning outcomes.”

Weigh it lightly. It is a conference paper, and its abstract reports no sample size and no design detail, so there is little in it to check. Taken at its word, it is one comparison in which self-testing and spaced repetition moved learning while the serial position effect did not. That sits comfortably with how we would use the effect: as a map of where recall runs thin, with the work still done by retrieval and spacing.

The title of that paper points at a neighbouring question, which is how memory fades with time rather than with position. Our guide to the forgetting curve covers Ebbinghaus, the shape of the drop, and why an early review buys more than a late one of the same length.

Where GeniusPal fits

GeniusPal turns your own material into questions, which is the retrieval half of the advice above. Upload a PDF, Word, PowerPoint, plain text, Markdown or CSV file of 10 MB at most, or paste a link, and it writes a study set: 10 questions per set on the Free plan, and up to 30 on paid plans. Free sets run as a multiple-choice quiz, and the paid Student and Genius plans add flashcard mode, recall mode and written quiz answers.

Position matters inside a set too, which is why every attempt reshuffles both the order of the questions and the order of the options within a question, from a seed minted on the server. Sit the same set again and the items do not come back in the same places. An earlier version randomised the options once at generation and then held them fixed, which let a learner learn the slot an answer sat in.

It does not schedule reviews and it has no due dates. The daily review, free on every plan, is a round of up to 20 questions drawn from across your sets, putting first the ones you got wrong or have not tried yet. It cannot read photos, or scans with no text layer, so typed notes or a PDF with real text in it is what to upload.