Give someone a list of twenty words and ask them to recall as many as they can, in any order. The result has been the same for sixty years: the first few come back, the last few come back, and the middle is a hole. Plot recall against position and you get a U — high at both ends, sagging through the centre.
Murdock demonstrated the curve systematically in 1962. It's called the serial position effect, and its two halves have separate causes.
Primacy and recency are different mechanisms
Primacy — the early items win because they got rehearsal time. When item one arrives there's nothing competing with it, so it's repeated more and has the best chance of reaching long-term storage.
Recency — the last items win because they're still in the workspace when you're asked. They haven't been consolidated; they're simply still there.
Glanzer and Cunitz separated the two with one elegant change: insert a distractor task — counting backwards for thirty seconds — between the list and the recall test. The recency advantage vanished. Primacy survived untouched. Delay empties the short-term workspace but doesn't touch what's already been filed.
That's the finding with teeth. Recency is a rented advantage. It feels like knowing and expires on contact with the next thing you do.
Where the curve shows up in real life
Once you know the shape you see it everywhere:
- Lectures and talks. You remember the opening framing and the closing summary. The three arguments in between — usually the actual content — sag.
- Meetings. The first agenda item gets real attention; the last gets urgency; items three through six get decided badly.
- Reading sessions. The first chapter after you sit down and the last pages before you stop.
- Your own notes. The top of the page is careful, the bottom is a sprint, the middle is bullets you can no longer parse.
- Lists of options. Which is why the order of items on a menu, a ballot, or a shortlist is never neutral.
The curve also compounds with length. A ninety-minute session doesn't have a bigger middle than a thirty-minute one — it has a deeper one.
What to do about it
Break the session into more sessions. Every break creates a new beginning and a new end. Two forty-minute halves with a real pause between them have four privileged positions instead of two — the cheapest possible intervention, and one of the reasons the Pomodoro technique works for reasons its inventor didn't claim.
Note the middle deliberately. Since the ends will survive on their own, spend your effort where the curve sags. If you're flagging in the middle of a talk, that's exactly when to write more, not less.
Don't trust the end-of-session glow. Walking out feeling like it all landed is recency talking, and Glanzer and Cunitz showed how fast that expires. The check is a delayed one: try recalling it tomorrow, not on the way to the car. This is the same illusion that makes re-reading feel productive, and the same fix — retrieval, not exposure.
Reverse the order on review. Study your notes from the bottom up sometimes, or shuffle sections. Reviewing in the original sequence rebuilds the same curve every time, so the middle stays weak on the tenth pass. Varying the order is a small dose of interleaving.
Put the important thing first or last — deliberately. When you're the one presenting, the sag is a design constraint. Bury nothing that matters in the middle third.
How it connects
The serial position effect is one of a family of findings about which items survive rather than how many. Its close relative is the Von Restorff effect — the item that stands out gets remembered regardless of position — and both sit downstream of the capacity limit described by working memory. The general lesson runs through all of memory and focus: what you retain isn't a fair sample of what you experienced, so the fix is structural, not effortful. Write down the parts your memory is quietly declining to keep.