Primacy Effect: Why the First Items Stick and Set the Frame

This page is about the start of the curve: why the opening items win, and what first information does to a judgment. The serial-position effect is the whole U shape that primacy forms one half of, and the recency effect is the raised end at the other side of it.

In list recall the classic account is rehearsal. When the second word arrives you can still repeat the first; by the tenth word the opening items have been gone over many times and the newest once. Dewey Rundus measured this directly by having students rehearse aloud while they studied, and the count fell steeply with position: the first word of a list appeared in an average of 7.5 separate rehearsal sets, the twelfth in 2.3, the last in 1.0. On the classic two-store reading, all of that extra repetition moves the early items into long-term memory, where they survive the end of the list.

The decisive part of Rundus's result is what happened when he held rehearsal constant. Comparing items that had received the same amount of rehearsal, words from the opening positions were recalled no better than words from the middle. The advantage was in the rehearsal, not in the position, which is why anything that equalizes rehearsal flattens primacy and anything that clears short-term memory leaves it standing. Glanzer and Cunitz showed the second half of that in 1966, as the next section describes.

Outside the laboratory the same word names something related but not identical: the extra weight of information that arrives first, which is what the demo above tests. Solomon Asch showed it in 1946 by reading two groups the same six traits in opposite orders. One group heard intelligent, industrious, impulsive, critical, stubborn, envious; the other heard the list reversed. The first group described an able person with some faults, the second a person whose difficulties got in the way of their abilities. The same words produced different people, because the opening terms set a direction and the later ones were fitted to it.

The two senses share a cause only loosely. Order effects in judgment depend on when the judgment is made and how much the material invites a running impression, and in some settings the last information dominates instead. What holds across both is that order is information the sender controls and the receiver rarely notices.

Bennet Murdock's "The serial position effect of free recall" in the Journal of Experimental Psychology in 1962 is the standard reference for the shape that primacy belongs to. Across six conditions that crossed list length with presentation rate, recall was highest at the end of the list, next highest at the start, and lowest in the middle, and a slower rate lifted the early and middle positions while leaving the end much as it was.

Murray Glanzer and Anita Cunitz separated the two ends in 1966 in the Journal of Verbal Learning and Verbal Behavior. Their manipulation was a delay filled with counting aloud between the last word and the recall test. After ten seconds of counting most of the end-of-list advantage had gone, and after thirty seconds, in that experiment, nothing of it remained, while recall of the early items was essentially unchanged. One end of the curve could be destroyed without touching the other, which became the standard evidence that primacy and recency rest on two separate processes.

Dewey Rundus supplied the mechanism in work reported in 1970 as a Stanford technical report and published in the Journal of Experimental Psychology in 1971. Twenty-five students studied eleven lists of twenty unrelated nouns, five seconds per word, rehearsing aloud into a tape recorder, so that every repetition could be counted by position. The number of rehearsal sets an item appeared in fell steadily from 7.5 at the first position to 1.0 at the last, and the probability of recalling an item rose with the number of rehearsals it had received. Items from the first four positions and items from the middle were recalled equally well once the amount of rehearsal was matched.

The demo on this page follows Asch's 1946 order experiment rather than a word list: one visitor's recall curve cannot separate primacy from the rest of the serial-position pattern, while two groups reading the same six words in opposite orders, 34 in one and 24 in the other, show the first word's weight directly.

The rows below are illustrations of where the first item gets more weight than its share, not findings from the studies cited on this page.

Knowing about the effect does not switch it off, so the practical counters work on order and on records.

• Record a judgment on each item as it arrives. Deferred ratings are ratings of memory, and memory favors the opening.
• Rotate the order when fairness matters, in interviews, tasting panels, demo days and review stacks. Where rotation is impossible, log each item's position so the scores can be checked against it.
• Read or present the material twice in different orders when a decision is important. Anything that changes between the two passes is a candidate for an order effect and worth a second look.
• Use the effect deliberately when you are the sender. State the point you want remembered in the first sentence rather than building to it.
• Give the middle its own pass. In a long list or document, the middle is usually the part that gets the least rehearsal, so it is where unexamined claims survive.
• For material you want to learn, split long lists into short ones. Each short list has its own opening, so more items get the opening's rehearsal.

The primacy effect is one named piece of a larger pattern, and it is easy to confuse with the pieces next to it.

• It is not the whole serial-position effect, which is the U-shaped curve made of a raised start and a raised end. Primacy is the start on its own, and in free recall it is usually the smaller of the two humps.
• It is not the recency effect, the better recall of the last few items, held in short-term memory on the classic account; in Glanzer and Cunitz's task it disappeared after thirty seconds of counting. The two respond to different manipulations.
• It is not anchoring, where a first number pulls a later numerical estimate toward it. Anchoring is about the value of a judgment; primacy is about what is recalled and how later information is interpreted.
• It is not the halo effect, in which one trait colors how every other is read; Asch studied that in the same 1946 paper by swapping warm for cold in an otherwise identical list. Primacy is about which trait arrives first, the halo about one trait spreading whatever its position.