Serial position effect

The serial position effect is the finding that how well people recall an item from a list depends on where it appeared. Items at the start are remembered well (the primacy effect), items at the end are remembered well (the recency effect), and items in the middle are remembered worst. The term is credited to Hermann Ebbinghaus, who studied memory through experiments on himself.

Why it matters

The best-known demonstration is Bennet Murdock's 1962 study of free recall. Participants were given lists of common words, from 10 to 40 words long, presented at one or two seconds per word, and then recalled as many as they could in any order. Every condition showed both primacy and recency. As summarized on Wikipedia, recency was strongest for the last eight positions and primacy extended over the first four.

A common explanation is that early items get more processing and are stored in long-term memory, while the last items are still in working memory at the moment of recall. For design, the takeaway is simple: position is not neutral. What sits at the beginning and end of a list, a page or a sequence gets more attention and is more likely to be remembered.

How to apply it

• Do put the most important items at the start and end of navigation, lists and menus. Laws of UX suggests placing key actions at the far left and right of navigation bars.
• Do lead with the main point in content and messages, and close with the next step.
• Do place lower-priority items in the middle when you must include them.
• Do repeat a critical instruction at the end of a long sequence, such as onboarding, rather than relying on people remembering step 2.
• Don't bury the key action or warning in the middle of a long list or paragraph.
• Don't expect people to remember details from a long spoken or read-out list, such as phone menus with many options.

Common mistakes

• Applying memory findings to scanning. The research measures recall after a list is gone. When a list stays on screen, people can look back, so ordering matters less for recall and more for what gets noticed first.
• Over-reading the numbers. The size of primacy and recency effects depends on list length, presentation speed and delay before recall, so treat them as tendencies, not exact rules.
• Forgetting the middle exists. Accepting that middle items are weaker is a reason to keep lists shorter or grouped, not to fill the middle with filler.

In AI products

Long AI answers are lists in disguise. A chat assistant that buries the actual answer in paragraph four, or a meeting-notes summariser that puts the one decision that matters halfway down a list of twenty points, works against how people remember. Lead with the answer or decision, end with the next action, and keep supporting detail in between or behind an expandable section.

The same applies to AI suggestions. If an assistant offers several options, the first and last will get the most attention, so order them deliberately and avoid putting the safest or most relevant option in the middle of a long set.