Tertiary Retention: Our Externalized Memories

The memory we think we are losing

Try the test: how many phone numbers do you still know by heart? For most of us, the answer fits on the fingers of one hand — and it includes numbers that no longer exist. Second scene: your photo gallery “resurfaces” a memory from five years ago, and you realize you would never have found it on your own, nor even looked for it. Third scene, more recent: you reopen a conversation with an AI assistant to ask it — the phrasing has become commonplace — “what did we conclude again?”

Around these scenes hovers a small shame, sustained by two symmetrical discourses. One laments: the digital makes us amnesiac, we can no longer retain anything. The other waves it away: why memorize what the machine knows better than you? As always in this series, the two camps split the truth between them and miss the question.

The previous episode ended on a question: where, exactly, do the kinds of knowledge we delegate go — into which media, into whose hands? To answer it, Stiegler did something unusual: he rewrote the theory of memory.

The three memories

Husserl and the melody

The starting point is a famous analysis by the philosopher Edmund Husserl. When you listen to a melody, you never perceive an isolated note: the note that has just passed remains “held” in your present perception — without which you would hear nothing but a string of unconnected sounds. Husserl calls this primary retention: immediate memory, embodied in perception itself. Then there is secondary retention: recollection — the song you recall the next day, already sorted, colored, interpreted by who you are.

Two memories, then, both lodged in the living. This is where Stiegler adds his floor.

The third floor

For there exists a third memory, which inhabits no body: the recording of the melody. The score, the record, the file. Stiegler calls it tertiary retention: memory fixed in a technical medium, which outlives the individual and is transmitted across the generations. Without it, no culture, no law, no science — each generation would start again from zero. Readers of this series will recognize the gesture: it is the grammatisation of episode 2 that produces these media, and it is the “originary default” of episode 1 that makes the matter vital — the human, this being without essence, is human only with its prostheses included. Our memory has always overflowed our brain. Forgetting phone numbers is not a modern decline: it is the normal functioning of a species that has delegated its retentions since the invention of writing.

If externalization is our condition, where is the problem? Stiegler puts it in a phrase that has only grown more timely: the great platforms want to become “THE tertiary retention of our secondary retentions.” Translation: the memory medium is no longer an inert object waiting to be consulted — it is an active system that decides what is presented to you, when, in what order, and that learns from your consultations. Whoever controls tertiary retentions shapes secondary retentions: what you will remember, what you will believe you lived through, what will count. The attention economy, crossed with the previous episode, is only the visible part of this mechanism.

Otto's notebook and the rented notebook

The best objection: the extended mind

There is, however, an optimistic — and philosophically serious — way of seeing all this. In 1998, Andy Clark and David Chalmers published a paper that became a classic, “The Extended Mind.” In it they tell of Otto, who suffers from a memory disease and writes everything down in a notebook: addresses, appointments, important facts. When Otto wants to go to the museum, he consults his notebook just as Inga consults her biological memory. Their thesis: since the notebook plays exactly the functional role of memory, it is part of Otto's mind. The mind does not stop at the skull; it extends into tools.

At first glance, this thesis dissolves our worry: to externalize is not to lose — it is to extend. Your phone is not the crutch of your failing memory; it is a part of your memory, full stop. And this reading is right on the essential point: it describes what Stiegler describes too, the human as a being of cognitive prostheses.

What parity forgets

But Otto's example contains a detail that philosophical discussion long treated as trivial, and that has become the central question: Otto owns his notebook. No one rereads it during the night. No one reorganizes its pages so that he lands first on certain addresses. No one slips advertisements into it between two appointments, studies its consultation habits, or can change its lock or alter its contents remotely.

Our notebooks, by contrast, are rented. The gallery that “resurfaces” your memories chooses which ones. The feed that reminds you of an event decides what deserves an anniversary. The service can shut down, change its terms, throttle a feature — and the part of your memory that resides there follows the fate of the landlord. The extended-mind thesis remains true; it is simply incomplete. If the notebook is part of your mind, then the question of who owns the notebook is not a commercial question. It is a question of mental integrity.

Language models push this limit case a notch further. They do not store your memories: they have retained — in the most Stieglerian sense — our collective ways of phrasing, and replay them on demand. And the history of your conversations with them becomes a notebook of a new kind: the one where you deposit your reasoning in progress, your hesitations, your naive questions. A notebook that the landlord, depending on the terms of the lease, also reads.

I write from firsthand knowledge: I have literally named my own working system Extended Mind. This article was written with it: a base of records, concepts, and notes accumulated over years, without which neither this series nor this site would exist. But the difference comes down to a single sentence, and it is the whole thesis of this episode: this memory runs on my own machine, I can export it in full, and no one else decides what surfaces from it. It is not a technical feat — it is a choice, one that has a cost, and one I would make again.

The notebook audit

From this episode, take away a tool: three questions to put to each of your important external memories — the gallery, the cloud, the team's document base, the assistant history.

1. Can I export it today, in full, in a format readable elsewhere? If the answer is no, you do not have a memory: you have a subscription to your own memories.
2. Who besides me reads it — and to do what with it? The question calls not for paranoia: it calls for an answer. If you do not know it, it is set out in the terms you accepted.
3. If it disappeared tomorrow, what would I lose — and what have I stopped retaining because it existed? This last part of the question links this episode to the previous one: it is the inventory of delegations, applied to memory.

At the scale of a team, the same questions sketch a policy: internal documentation that the team owns and can take with it is not the same object as a proprietary silo queried on request — even if the user experience feels alike. The collective memory of an organization is a tertiary retention like any other: it belongs to someone, and it is better to know to whom.

Neither nostalgia, then, nor resignation. To externalize one's memory is not a fault: it is our way of being a species. But we choose our prostheses — and we may prefer, whenever possible, the notebook one owns to the notebook one rents.

What memory makes possible

There remains the most beautiful question, which this series has saved for the end. Externalized memories that circulate freely — from books to the digital commons — have not only preserved the past: they have allowed generations to think together, to transmit knowledge to one another, and to become, each of them, a little more than the one before. Stiegler has a word for this process, and it is the most optimistic in his vocabulary. Final episode: transindividuation — and what contributory technologies would be.