Alright, let’s start with what actually matters. You left your phone somewhere. No, not in your pocket. No, not on the table. In the kitchen? The bathroom? Damn it. You have no idea. You’ve done three laps of the house, you’ve lifted the sofa cushion as if the phone were going to be under there (it’s never under there), and you’re about to give up and use “find my device” from some other gadget. But hey. You know how to ride a bike. You haven’t touched one in five years, but if you got on one right now, your body would pedal, balance, and turn as if it had never done anything else. How is it possible that the brain can remember an insanely complex motor choreography, and yet be incapable of holding onto where you left a half-kilo object that literally vibrates, for ten whole minutes? The answer isn’t that you’re losing your marbles (okay, maybe a little). The answer is that your brain isn’t a hard drive. It’s more like a company with different departments that don’t talk to each other. And the department that stores how to ride a bike is completely different from the one that… well, the one that should be storing where you left your phone, but has apparently decided to go on strike.
The two big camps: declarative vs. procedural
To understand this mess, we need to start with the most basic division of long-term memory: declarative (or explicit) memory and procedural (or implicit) memory [1]. They’re so different that they literally run on different hardware [2][4]. Procedural memory is the one you need for the bike. It’s the memory of “how it’s done.” It stores motor skills, habits, and procedures that we execute automatically, without thinking [1][5]. It’s what lets you drive, tie your shoelaces, swim, or play an instrument without having to consciously remember every single movement. When you learned to ride a bike, it was torture at first: falls, clumsy pedaling, the anxiety of your dad letting go of the seat… But with practice, the brain recorded movement patterns in specific circuits, and the skill became automatic and unconscious. It got consolidated into a system that doesn’t need consciousness to function [3]. That’s why, even years later, the “program” is still there. Declarative memory, on the other hand, is the memory of “what.” It’s what stores facts and events that we can consciously recall and “declare” [4]. It splits into two subtypes: Semantic memory: general knowledge of the world. Knowing that Paris is the capital of France, that water boils at 100°C, or that a fork has four prongs [5]. Episodic memory: the record of our personal experiences. It’s what lets you remember what you had for breakfast yesterday, your last birthday, or, in theory, where the hell you left your phone ten minutes ago [1][5]. And here’s the problem: episodic memory, the one that’s supposed to tell you where your phone is, is one of the most fragile and context-dependent [4]. Procedural memory, the bike one, is an oak tree.
The (neural) treasure map: where each memory lives
The reason for this difference in reliability lies in the brain circuits each one uses. Declarative memory (both semantic and episodic) depends heavily on the medial temporal lobe and, especially critically, on the hippocampus [4][5]. The hippocampus acts as a kind of conductor, or an indexing system. When you live through an experience, the hippocampus “packages” it and sends it off to the cerebral cortex for long-term storage. But the hippocampus is also a bottleneck: if it’s damaged, forming new episodic memories goes straight down the drain [10]. Procedural memory, on the other hand, doesn’t need the hippocampus at all. Its main circuits are the basal ganglia, the cerebellum, and the motor cortex [5]. These are older structures, more automatic, and in a way more bomb-proof. A 2024 study, for instance, found that the white matter microstructure in the basal ganglia-cerebellar circuit is linked to individual differences in procedural learning, in both adults and children [7]. In other words, the quality of this “wiring” affects how quickly we pick up skills like riding a bike. And the cerebellum, on top of that, seems to be the hippocampus’s equivalent for sensorimotor memory, playing a key role in forming long-term motor memories [9].
The hippocampus crashes the motor party too
Procedural memory doesn’t use the hippocampus? Turns out it’s not that simple. In a neuroimaging study published in November 2024 in the Journal of Neuroscience [10], led by Rhys Yewbrey and Katja Kornysheva [3], researchers trained 24 participants to learn finger-movement sequences until they could execute them with high precision. Then they put them in an fMRI scanner and watched what happened in their brains [3]. The expectation was that activity would be concentrated in the basal ganglia and the cerebellum. And that’s exactly what happened during execution. But it turned out that when the planning of the finger movement began, activity in the hippocampus increased — in a spot where you’d expect the basal ganglia and cerebellum to be doing the work — and it even predicted the order of the sequence of key presses the participant was about to perform. The authors conclude that the hippocampus, traditionally seen as the center of episodic and spatial memory, also takes part in the higher-order control of skilled movements that require flexible retrieval [3]. In other words, the hippocampus doesn’t just help you remember where your phone is — it also helps you plan the pedaling sequence to dodge a curb. The separation between systems isn’t a wall, it’s a membrane full of holes.
The case of H.M. and Clive Wearing
The best evidence that these systems are independent comes from amnesia patients. The most famous case is that of Henry Molaison (known as H.M.). In the 1950s, parts of his medial temporal lobes were removed, including most of his hippocampus, to treat his epilepsy [10]. The result was severe partial amnesia: he was unable to form new episodic or semantic memories. He couldn’t remember what he’d eaten an hour earlier, nor recognize people he’d just met [10]. But there was one thing H.M. could still do: learn new motor skills. Psychologists had him practice drawing a star while looking only at its reflection in a mirror (a task that requires visuomotor coordination and is tricky at first). Day after day, H.M.’s performance on the task improved, just like a healthy person’s would. But the striking part was that he didn’t remember ever having practiced it before [10]. His declarative memory was wiped, but his procedural memory kept running at full capacity [11]. The case of Clive Wearing, a British musician who contracted herpes encephalitis that wrecked his hippocampus, is even more extreme. His episodic memory shrank down to a few seconds. He lives in a perpetual present, waking up every few minutes thinking he’s just regained consciousness for the first time. And yet, he can hold conversations, read music, and play piano and conduct orchestras with a skill that remained completely intact [2]. His musical procedural memory survived the disaster.
Semantic and episodic memory aren’t so different
A 2023 study published in PMC [8] found that there’s actually a surprisingly large overlap between the neural correlates of semantic memory and episodic memory [8]. What this means is that the brain doesn’t have two perfectly separate drawers for “facts” and “experiences.” Instead, there’s a continuum. Autobiographical memories can become more “semantic” over time (you stop remembering the exact day you learned that Paris is the capital and you simply “know” it), and semantic knowledge is often anchored in the original episodic experiences that gave rise to it.
What if procedural memory could “teach” declarative memory?
This is speculation: if procedural memory is so resilient and can work without the hippocampus, could it be used to rehabilitate declarative memory? For example: instead of trying to get an amnesia patient to consciously memorize something, we teach them a procedural skill — like playing a sequence on the piano — which they’re actually capable of. In doing so, we’re activating the basal ganglia and the cerebellum. But as we saw in the Yewbrey and Kornysheva study [3], the hippocampus also fires up during the planning of that sequence. We’re, in a sense, “inviting” the hippocampus to the motor party. Could that repeated hippocampal activation, even in a procedural context, help “wake up” its capacity to form new episodic memories? Or could we anchor declarative information onto motor processes, so a patient learns new facts as if they were a motor routine? There’s no solid evidence for this — but given that we know the hippocampus isn’t just a spectator in procedural memory [3], it’s a possibility, not a crazy one.
Your brain isn’t broken, it’s just weirdly organized
Back to the phone: you don’t remember where you left it because episodic memory is demanding — it needs attention, context, and above all, a hippocampus willing to do its indexing job [4][5]. If you were thinking about something else when you put the phone down, the hippocampus said “this isn’t a priority” and never built a solid index. The memory faded before it ever really formed. Procedural memory, on the other hand, doesn’t ask for permission or attention. It gets recorded through repetition, in circuits that are basically automation machines [1][3]. Once the program is installed, it doesn’t uninstall. It’s not that you “don’t forget” how to ride a bike — it’s that your brain doesn’t even register it as something to be remembered. It’s a reflex, a habit, a routine embedded in the flesh of your basal ganglia [5]. And then there’s declarative memory’s other sister, the semantic one, which is the worst of the three at settling in. Sure, you know Paris is the capital of France, but you don’t even remember when you learned it — it became a loose fact, with no date, no context, floating there for so long it lost all connection to the moment it came in [8]. The problem shows up when the fact is new: a history date, a formula, a physics concept you just saw in class. If you only read it twice the night before the exam, it doesn’t anchor to anything. It’s isolated information with no hooks, and the brain treats it as background noise, not as something worth keeping. That’s why re-reading your notes, even though it feels productive, barely helps: a thorough review of study techniques, which compared dozens of different methods, found that re-reading and highlighting are among the least effective strategies out there, precisely because they generate a feeling of familiarity that gets mistaken for actually knowing something [12]. It sounds familiar, so you think you know it — but that’s not the same thing. What actually works — and this isn’t a single study, it’s meta-analyses backed by hundreds of studies and thousands of students — is trying to actively recall the fact without looking at your notes, even if you get it wrong. That forces your brain to rebuild the connection from scratch, and it leaves it far stronger than ten straight re-reads [13]. It’s counterintuitive: it takes more effort, it feels worse while you’re doing it, and it still works better. Semantic memory doesn’t anchor through passive exposure. It anchors through the effort of having to go dig it up. Your brain isn’t a computer that stores everything on the same hard drive. It’s an ecosystem of different departments competing for resources, each with its own rules: one installs itself through motor repetition alone, no permission needed; another needs your attention at the exact moment something happens; and the third one — the one you use to study — needs you to force it to work if you want it to stick. And the next time you can’t find your phone, or can’t remember a date you swore you’d studied, don’t panic: now you know exactly why it happened, and what to do about it next time. The bike, at least, will always be there.
References
[1] Declarative and Procedural Memory Contributions to Argument Structure and Word Order Learning (2025). Repositories.cdlib.org. Reliability: Theoretical framework (presents the Declarative/Procedural model for language, an established theoretical proposal).
[2] Cacciatori di ricordi (2025). Almanacco.cnr.it. Reliability: Science communication piece (an outreach article from Italy’s CNR explaining the anatomical difference between declarative and procedural memory, citing classic cases).
[3] The Hippocampus Preorders Movements for Skilled Action Sequences (2024). Yewbrey, R., & Kornysheva, K. Journal of Neuroscience, 44(45). Reliability: Reliable (peer-reviewed fMRI neuroimaging study showing hippocampal involvement in the planning of motor sequences).
[4] Physiology, Long Term Memory (2023). Almaraz-Espinoza, A., et al. StatPearls. Reliability: Theoretical framework (comprehensive review of long-term memory physiology, published in a medical reference database).
[5] TABLE 1: Types of Long-Term Memory (2021). PMC. Reliability: Theoretical framework (reference table on types of memory, examples, and neural substrates, from a peer-reviewed source).
[6] Memory systems (2010). Wolk, D. A., et al. Continuum (Minneap Minn). Reliability: Theoretical framework (review article on memory systems and their neuroanatomical basis, published in an American Academy of Neurology journal).
[7] Procedural learning is associated with microstructure of basal ganglia-cerebellar circuitry in children (2024). Bianco, K. M., et al. Brain and Cognition, 180. Reliability: Reliable (peer-reviewed empirical MRI study with an SRT task in children).
[8] The shared and unique neural correlates of personal semantic, general semantic, and episodic memory (2023). PMC. Reliability: Reliable (study finding overlap in the neural correlates of semantic and episodic memory).
[9] The cerebellum acts as the analog to the medial temporal lobe for sensorimotor memory (2024). PNAS. Reliability: Reliable (study demonstrating the cerebellum’s role in long-term sensorimotor memory).
[10] El caso de H.M. y la revolución en la neurociencia cognitiva (2025). RCV.hn. Reliability: Science communication piece (article summarizing the Henry Molaison case and its contribution to our understanding of procedural memory).
[11] Declarative Learning, Priming, and Procedural Learning Performances comparing Individuals with Amnestic Mild Cognitive Impairment… (2023). Reliability: Reliable (study confirming that procedural learning can operate independently of the hippocampus).
[12] Dunlosky, J., Rawson, K.A., Marsh, E.J., Nathan, M.J., & Willingham, D.T. (2013). Improving students’ learning with effective learning techniques: promising directions from cognitive and educational psychology. Psychological Science in the Public Interest, 14(1), 4-58. Reliability: Reliable.
[13] Roediger, H.L. & Karpicke, J.D. (2006). Test-enhanced learning: Taking memory tests improves long-term retention. Psychological Science, 17(3), 249-255. Reliability: Reliable (confirmed by later meta-analyses with hundreds of studies and large sample sizes: Rowland, 2014, Psychological Bulletin, g=0.50; Adesope, Trevisan & Sundararajan, 2017, Review of Educational Research, g=0.61 across 217 studies).