Writing as Memory: What Herman's Blog Gets Right About Active Recall

Herman's recent post, Active recall, arrives disguised as a personal essay about journaling and ends up making a claim that cognitive scientists have spent the better part of two decades defending. His thesis is deceptively modest: he is not unusually gifted at remembering things, he simply writes about them. The compliments he collects from friends who marvel at his memory are, by his own account, a kind of accidental sleight of hand. He has been practicing retrieval all along, and the rest of us mistook a trained habit for an innate trait.

What makes the piece worth taking seriously is that the personal anecdote is anchored to real evidence. Herman cites the 2011 study by Jeffrey Karpicke and Janell Blunt, Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping, published in Science. The experiment is elegant precisely because it pits intuition against measurement. College students read short educational texts and then studied them three ways: rereading the material, building elaborate concept maps while looking at the source, and free recall, which means reading once and then writing down everything they could remember on a blank page. A week later, the free-recall group retained more than either of the others.

The detail that does the real work

The finding that should unsettle anyone who has built a study routine around highlighting and rereading is the one Herman flags as the kicker. In one version of the experiment, the final test itself was a concept-mapping task. You would expect the students who practiced by making concept maps to win, since their study method matched the test format exactly. They did not. The free-recall group still outperformed them. The effortful act of pulling information out of memory beat the more elaborate, more visually impressive technique even when the deck was stacked against it.

This is the part that resists common sense. Rereading and concept mapping both feel productive. They produce a sensation of fluency, the comfortable recognition that yes, I know this. That feeling is the trap. Recognition is not retrieval. When you reread a passage, your brain coasts along a path that already exists, and the ease of the journey gets misread as mastery. Free recall offers no such comfort. It is effortful, sometimes frustrating, and the struggle to reconstruct an idea from a blank page is exactly the mechanism that strengthens the memory. The difficulty is not a side effect. It is the point.

Writing as thinking, not just recording

What elevates Herman's post above a summary of a study is the connection he draws between recall and articulation. He opens with a line from Haruki Murakami's What I Talk About When I Talk About Running: the sense that one cannot grasp much of anything without putting thoughts down in writing. Herman treats writing not as a container for finished thoughts but as the process by which thinking happens. The act of forcing a vague mental impression into legible, ordered sentences is itself the work of understanding.

This reframes what a journal or a notes file actually is. Herman admits, almost in passing, that he rarely rereads his notes or his journal. For a decade he has kept a daily journal and a single large notes.txt file full of scribbles, and the value he extracts from them has almost nothing to do with going back to consult them. The writing did its job at the moment of writing. The notes are the residue of a process, not a resource to be mined later. That distinction matters because it inverts how most people justify note-taking. We tell ourselves we are building an archive for future reference. The evidence suggests the archive is beside the point and the construction of it is everything.

The implication for how we use tools

There is a thread here that Herman touches and then sets aside, but it deserves to be pulled. He mentions, with visible worry, that the education system has had the entire concept of writing-as-study upended by AI, and points to declines in literacy and comprehension. The Karpicke and Blunt result gives that worry a sharp edge. If retrieval is the engine of durable learning, then any tool that performs the retrieval on your behalf removes the very friction that produces understanding.

When a student asks a language model to summarize a text, the model does the reconstructive work, the effortful pulling-together of scattered ideas into coherent prose, and hands back a clean result. The student receives the residue without ever running the process. By the logic of this research, that is close to the worst possible trade. It feels like learning, it looks like a finished product, and it produces the fluent recognition that masks the absence of real retention. The same caution applies to anyone who reaches for AI-assisted note generation or automatic summaries as a study aid. The output may be excellent. The cognitive transaction that would have made it stick never took place.

This is not an argument against the tools, and it would be a misreading of Herman to turn his post into one. The useful inference is narrower and more practical. These systems are genuinely valuable for tasks where the goal is the artifact: drafting boilerplate, translating, searching, organizing. They are actively counterproductive when the goal is for a human brain to come away changed. Knowing which situation you are in is the skill worth cultivating.

The counter-case worth holding

It would be too neat to declare free recall the universal answer. The Karpicke and Blunt study measured retention of factual and conceptual text material over roughly a week, and retrieval practice is strongest precisely in that domain. There are kinds of understanding it captures poorly. Procedural skills, the embodied knowledge of playing an instrument or writing production code, are built through deliberate practice and feedback loops that look different from sitting with a blank page. Concept maps, dismissed in the study as the weaker technique, genuinely help when the goal is to see structural relationships between ideas rather than to remember discrete facts, and a person who already has the material in memory may use mapping to reorganize it productively.

There is also a survivorship element in Herman's own account that he is honest enough to leave visible. He describes himself as a poor student who never learned to study, and he stumbled into retrieval practice almost by accident through journaling and blogging, driven by genuine interest rather than discipline. That is the crucial variable. He writes when something has been turning over in his mind, not on a schedule, not to satisfy a requirement. The technique worked because it was attached to authentic curiosity. Mandating free recall as a school exercise, stripped of that intrinsic motivation, might recover some of the benefit, but it would not reproduce the conditions that made it effortless for him. The method is necessary; it is probably not sufficient on its own.

Herman's closing line, that he writes for his readers but suspects he also writes for himself, lands as something more than a sentimental flourish. It is a precise description of how durable learning actually works. The audience gives the writing its shape and its standard, the obligation to be legible to someone else, while the private benefit, the deepening of his own memory and understanding, accrues silently underneath. The two are not in tension. The effort he spends making his thoughts clear to us is the same effort that makes them permanent for him, which may be the most practical reason any of us has to keep writing things down.