Levels-of-Processing Effect.

01Overview

The levels-of-processing effect (Craik & Lockhart) holds that memory strength depends on depth of cognitive processing — not on storage duration alone. Shallow processing of appearance (font colour, animation) produces fragile memory. Deep processing of meaning (implications, self-relevance, task integration) produces durable recall.

Designers often optimise for shallow exposure: banners users see, tooltips they dismiss, videos they half-watch. Levels-of-processing asks whether users thought about the content — connected it to a goal, paraphrased it, used it — or merely noticed it. Notice is not learn.

02Detailed explanation

Classic experiments show semantic questions beat rhyme or visual questions for later recognition. Product analogues:

• Onboarding carousels users swipe through without integrating produce no durable knowledge of where settings live.
• Privacy notices styled for attention without comprehension tasks fail levels-of-processing — seen, not encoded.
• Microcopy users read but do not apply (error messages ignored until deep processing forced by failure).
• Training modules with click-through completion metrics measure exposure, not depth.

Deep processing requires time, relevance, and often generation — users producing answers, not only consuming copy. Faster onboarding is not always better onboarding.

03Why it exists

Memory traces vary in strength by elaboration at encoding. Elaboration links new information to existing schemas — making retrieval cues plentiful.

Interfaces optimised for speed and aesthetics often minimise cognitive effort — accidentally minimising memory formation for critical information.

04Effects on users

Users forget where features live after tooltip tours — shallow processing. They remember workflows they struggled through once — deep processing via effort.

They confabulate understanding of terms they saw but never elaborated — shallow familiarity feels like knowledge.

05Effects on designers & teams

Teams mistake exposure metrics for learning:

• Tooltip tours as education. High view counts, low recall in follow-up tasks.
• Compliance click-through. Legal satisfied; users cannot restate what they agreed to.
• Beautiful empty states without action. Aesthetic processing without semantic task.
• Video-only help. Passive watching without practice problems.

06Practical takeaways

• Require generative interaction. Ask users to locate, paraphrase, or apply — not only acknowledge.
• Space deep processing across sessions. Pair with spacing effect — revisit meaningfully.
• Connect to user goals at encoding. "You'll need this when you invoice" beats generic feature tour.
• Test delayed recall. Week-later tasks reveal shallow onboarding success.
• Replace swipe carousels with do-one-task onboarding. Action creates depth.
• Design errors as encoding opportunities. Meaningful failure can deepen memory of correct path.

07Design examples

08Ethical risks

Shallow-processing consent flows create legal coverage without informed users — ethics of exposure vs understanding.

Assuming users remember what they saw once justifies hiding critical controls — harming those who needed depth and got gloss.

Self-test: What must users remember in your product — and did your last onboarding make them think about it, or just see it?

09Related biases