In 2013, John Dunlosky and colleagues published what remains the most useful survey in the learning-science literature [Dunlosky et al., 2013] Improving students' learning with effective learning techniques . They took ten common study techniques — highlighting, rereading, summarising, self-testing, and the rest — and ranked them by how well the evidence supported them as tools for remembering things.
Highlighting came out at the bottom. Rereading came out near the bottom. Summarising, when done from memory rather than from the text, did reasonably well. Self-testing and spaced practice came out on top — by a lot.
None of this was news to cognitive scientists. It has been news to almost everyone else, for about four decades.
What highlighting actually does
The appeal of highlighting is obvious. It feels active. You are marking the text, making decisions, doing something more than passively reading. Fluent readers emerge with neat yellow trails through their PDFs and a strong sense of having studied.
The problem is that highlighting does almost none of the cognitive work that produces durable memory. It is shallow processing — you mark what looks important without necessarily understanding why. It is recognition, not retrieval — when you later look at the highlights, you recognise them fluently without pulling them from memory. And crucially, it is passive generation: you haven’t written, said, or done anything with the information.
Bjork’s desirable-difficulties framework [Bjork, 1994] View in bibliography → captures the irony. Study conditions that feel easy — highlighted text that reads fluently on re-review, familiar passages re-read to the point of ease — are almost exactly the conditions that produce the worst long-term retention. Study conditions that feel harder — attempting to recall without looking, practising under conditions different from how you learned — produce the best.
The illusion of competence
This is the rub. Students — and teachers, and even the cognitive scientists who know better — consistently misjudge their own learning when they rely on recognition rather than retrieval as the signal.
You highlight a textbook. Next week, you look at the highlights. You recognise them. You feel studied. This is the illusion of competence: the experience of fluent recognition being mistaken for the ability to retrieve and apply.
On an exam — which is, structurally, a retrieval task — the illusion breaks. You could recognise the material with the book in front of you, but you cannot pull the material out without the book. The gap between “I know this when I see it” and “I know this when I have to produce it” is exactly the gap that highlighting does not close, and cannot close.
What works, briefly
If you want to skip the book, here is the short version:
- Retrieval practice. [Roediger & Karpicke, 2006] View in bibliography → Close the book. Try to state what you read, from memory. Any retrieval attempt — even a failed one — is worth more than any amount of rereading.
- Spacing. Return to the material at expanding intervals. Day 1, day 3, day 7, day 14. Each interval should include at least one night of sleep, because sleep is where the hippocampus does the consolidation work.
- Interleaving. [Rohrer & Taylor, 2007] Mix topics during practice rather than blocking them. It feels harder in the moment; it is substantially better for transfer.
- Generation. Produce the information yourself. Write in your own words. Draw the diagram. Explain to someone else.
- Productive failure. [Kapur, 2008] View in bibliography → Attempt a problem before being taught how to solve it. Even if you fail — especially if you fail — the instruction afterwards lands on prepared ground.
These are not personal preferences. These are the consensus findings of the most replicated research programme in educational psychology.
What Fluera does with this
Fluera’s entire product is an attempt to make the right things the defaults.
- The canvas is blank. No templates that would let you skip generation.
- Step 2 of our 12-step cycle forces you to reconstruct material from memory before any tool appears. That’s productive failure.
- The Socratic mode asks before it answers. That’s retrieval practice.
- Ghost Map reveals gaps against an ideal. That’s hypercorrection, which is retrieval practice under high-arousal conditions.
- The spaced-repetition scheduler returns at widening intervals. That’s spacing and successive relearning.
- Fog of War, for exam prep, masks what you almost remember. That’s retrieval under occlusion — the sharpest form of desirable difficulty.
None of these are new ideas. The cognitive science is settled. What is new is a tool built end-to-end around the evidence, rather than around whichever feature looked best in a demo.
If you want to study better, you can start today without Fluera — close the book, write what you remember, come back tomorrow. If you want a tool that makes those habits the path of least resistance, that’s what the beta is.