Serial-Position Effect: Why You Remember the First and Last Items

The classic account explains one curve with two memory processes. The first items on a list get rehearsed more than any others: when the second word arrives you can still repeat the first, when the third arrives you can repeat both, and by the time the list is ten words long the opening items have been silently rehearsed many times. That extra rehearsal moves them into long-term memory. This is the primacy effect.

The last items are still sitting in short-term memory when recall begins. You have not rehearsed them, but you do not need to: they are simply still there, and most people write them down first. This is the recency effect. The middle items get neither advantage. They arrived too late for heavy rehearsal and too early to still be in the buffer, so they are the ones that fade.

Three findings support the two-process account. Slowing the presentation rate improves recall of the early and middle items (more rehearsal time) but leaves the recency items unchanged. Making the list longer flattens the middle without touching the ends. And a short distraction task between the list and recall, such as counting backwards for 30 seconds, wipes out the recency effect while leaving primacy intact, because the distraction clears short-term memory but cannot undo rehearsal that has already happened.

The curve has been reproduced many times in free-recall experiments and appears with lists heard, as in Murdock's study, as well as lists read. Later models, such as Howard and Kahana's temporal context model, explain the same curve without a separate short-term store, treating recency as a matter of how similar the retrieval context is to the encoding context; the curve itself is not in dispute.

Bennet Murdock published "The serial position effect of free recall" in the Journal of Experimental Psychology in 1962. He read lists of common English words to 103 undergraduates at the University of Vermont, one word at a time, and asked them to write down as many as they could in any order. Six list conditions crossed length with rate: 10, 15 and 20 words at two seconds each, and 20, 30 and 40 words at one second each.

Plotting the percentage of participants who recalled each word against its position produced the same shape in every condition: a high start that fell over the first few positions, a flat middle, and a steep rise over the last positions to a peak at the final word. The level of the flat middle depended on the condition: longer lists at the faster rate had a lower floor, and the slower rate lifted the early and middle positions, while the rise at the end looked much the same in every condition. Longer lists stretched the flat middle without changing the ends.

The curve was not new. The experimental study of list learning goes back to Hermann Ebbinghaus's nonsense-syllable experiments of 1885, and by 1962 the serial-position curve was a familiar finding; Murdock's paper opens by citing earlier work. His contribution was to map it with free recall, modern controls and enough participants to show that the shape was stable across list length and speed. Glanzer and Cunitz extended the result in 1966 by adding a delay before recall and showing that only the recency portion of the curve disappeared, which became the standard evidence for separate short-term and long-term stores.

The rows below are illustrations of where ordered information gets judged from memory later, not findings from the studies cited on this page.

The list experiments offer no way to switch the effect off, but it can be planned around.

• When you need to remember a list, write it down. A written list has no middle slump.
• When you are the one presenting, put the most important point first and repeat it last. The middle is for supporting detail that can be looked up.
• When you are judging a sequence of people or options, record a judgment on each one immediately. Deferred ratings are ratings of memory, and in the list experiments memory favored the ends.
• Rotate the order when fairness matters: interview slots, tasting panels, ballots, demo days. If rotation is impossible, at least record where each person or option sat in the order, so that the ratings can be checked against position afterwards.
• For material you want to learn, break long lists into short ones. Each short list has its own primacy and recency, so more items get an advantaged position.

The serial-position effect is the whole U-shaped curve. Its two halves have their own names, and several neighbors are easy to confuse with it.

• It is not only the primacy effect, which is the raised beginning of the curve on its own, driven by rehearsal, and which also names the tendency for first impressions to dominate judgments.
• It is not only the recency effect, the raised end, driven by what is still in short-term memory and by the fact that the latest information tends to be weighted most in a judgment.
• It is not the Von Restorff effect, where one item that stands out from the rest (a word in red, a number among words) is recalled better regardless of its position.
• It is not the peak-end rule, which is about how an experience is evaluated afterwards, from its most intense moment and its ending, rather than about which items are recalled.