Cognitive Load & Memory

The central constraint of UX: working memory is tiny and attention is expensive. Most "confusing" interfaces are working-memory failures.

Try it — hold ten digits in your head. Memorize a plain ten-digit number and type it back; then repeat the test with fresh digits chunked like a phone number. This is a demonstration, not a quiz — most people feel the second round get easier.

Round 1 of 2 — plain, unchunked digits.

Ten digits, three seconds. Ready when you are.

Cognitive Load Theory (Sweller, 1988)

Three load types, adapted to UX:

Intrinsic load — inherent difficulty of the user's task. You can't remove it, only sequence it (chunk complex tasks into steps).
Extraneous load — effort spent on the interface itself: decoding layout, hunting for controls, parsing jargon, re-orienting after inconsistency. This is the load design exists to eliminate.
Germane load — effort building understanding (schemas). Worth protecting: leave capacity for the user's actual work.

Design moves that cut extraneous load: consistency (reuse learned patterns), recognition over recall, sensible defaults, plain language, progressive disclosure, inline help at point of need, visual grouping instead of prose instructions.

Working memory limits

Practical capacity ≈ 4±1 chunks (Cowan 2001), not 7 (Miller 1956 was about memory span for familiar chunks). Never require carrying info between screens: show the comparison side by side, keep the cart visible, echo the email address on the "check your inbox" screen.
Chunking raises effective capacity: format phone numbers, card numbers, and codes; group settings into labeled sections of ~5–7 items.

Long-term memory & learning

Recognition is far easier than recall (a robust memory finding) — hence menus beat command lines for novices, autocomplete beats blank fields, recently-used lists beat searching again.
Spacing effect (Ebbinghaus 1885; Cepeda et al. 2006 meta-analysis): distributed practice beats massed — onboarding that drips features over sessions outperforms a one-shot tour.
Serial-position effect: first/last list items are remembered best.
Picture-superiority effect: concepts presented as images are remembered better than words alone (Paivio's dual-coding theory) — icons WITH labels beat labels alone for memorability, but icons alone are usually ambiguous (NN/g: label your icons).

Interruptions & task resumption

Working memory holds the current goal state; interruptions evict it. Resuming after an interruption carries a measurable "resumption lag," and lag grows with interruption length and complexity (Altmann & Trafton's memory-for-goals work; Monk, Trafton & Boehm-Davis 2008). Interrupted workers often finish tasks by working faster — at the cost of higher stress and frustration (Mark, Gudith & Klocke 2008). Design moves: autosave everything, restore exact state on return ("resume where you left off"), never let a session timeout or modal destroy in-progress work, batch/defer non-urgent notifications during focused tasks, and after any forced detour (auth, verification) return the user to the exact point of departure with context intact.

Multimedia & instruction (Mayer)

Mayer's principles for explaining anything in-product: people learn better from words + relevant pictures than words alone (multimedia principle); better when extraneous material is excluded (coherence); when words sit next to the graphics they describe (spatial contiguity); and when content is segmented into learner-paced chunks. Apply to onboarding, tooltips, docs.

Practical checklist

Can every step be completed with what's visible on screen? If not, why?
Any screen with >2 unfamiliar terms: rewrite or define inline.
Any form asking for info the system could infer/remember: remove the field.
Any instruction longer than a sentence: redesign so it isn't needed.

Sources

Sweller, J. (1988). "Cognitive load during problem solving." Cognitive Science, 12(2); Sweller, Ayres & Kalyuga (2011). Cognitive Load Theory.
Cowan, N. (2001). "The magical number 4 in short-term memory." Behavioral and Brain Sciences, 24(1). Miller, G. A. (1956). Psych. Review, 63(2).
Cepeda, N. et al. (2006). "Distributed practice in verbal recall tasks." Psychological Bulletin, 132(3).
Paivio, A. (1971/1986). Dual-coding theory. Mental Representations.
Altmann, E. M. & Trafton, J. G. (2002). "Memory for goals." Cognitive Science, 26(1); Monk, Trafton & Boehm-Davis (2008). JEP: Applied, 14(4).
Mark, G., Gudith, D. & Klocke, U. (2008). "The cost of interrupted work." Proc. CHI '08.
Mayer, R. E. (2001; 3rd ed. 2020). Multimedia Learning. Cambridge UP.
Johnson, J. Designing with the Mind in Mind (3rd ed. 2020) — memory chapters; Weinschenk, S. (2011). 100 Things Every Designer Needs to Know About People. New Riders.