Why cramming fails: spacing, gaps and the forgetting curve

The Short Version

Cramming works for tomorrow and fails for next month. Sleep scientist Matthew Walker's lab found that people who stayed up to study did get some benefit the next day, but a month later they had lost almost everything, while those who slept retained far more [1]. Memory researcher Charan Ranganath says the same about massed study in general: it helps only "if you need it very quickly," and spaced sessions win for anything you want to keep [3]. Andrew Huberman adds the small-scale version, with pauses of seconds inside a session, short stories mixed into hard material, and study split into two or three daily blocks at fixed times [2]. Put together, the sources describe gaps at every scale, from seconds to nights to a week, as the place where memory gets built, which makes when you study as important as how.

Table of contents

What They Said

Charan Ranganath: the forgetting curve and the spacing effect

Ranganath, a UC Davis psychologist and neuroscientist, starts from the oldest quantitative result in memory research. Hermann Ebbinghaus tried hard to memorize arbitrary nonsense, and "within a day, he lost about 60% of that information" [3]. Ranganath notes that Ebbinghaus was measuring generously, and that nobody since has managed to overturn the basic fact that people forget most of their experiences [3]. If you expect to remember everything, he says, you are already off, because that is not what human brains are designed to do [3]. His summary: "you don't want to remember more, you want to remember better" [3].

Spacing is his first tool for remembering better. He gives a concrete case. With an hour to memorize as many Spanish words as possible, doing half an hour now and another half hour later the same day will not hold the words as long as "half an hour today and half an hour one week from now" [3]. The total time is the same; only the gap differs.

He is also frank about why people cram anyway. It works in the short run. He used to study in the hallway right before a test, and the information was right there when he needed it: "actually not spacing can really help you if you need it very quickly" [3]. Then the cost: "But the problem is that you tend to forget it later on" [3]. He speaks from experience, saying he "can't remember so much from college and high school because I crammed" [3]. He calls this a boundary condition: the right schedule "depends on when you need that information" [3].

Host Lex Fridman describes using the spaced repetition app Anki every day, and Ranganath ties spacing to a related finding, the testing effect, in which testing yourself on material holds it better over time than just studying it again [3].

Matthew Walker: the all-nighter, measured

Walker, of UC Berkeley, tested the classic student strategy directly. Healthy participants either slept or stayed awake all night and then tried to learn new facts inside a brain scanner [1]. The sleep-deprived group showed a 40 percent deficit in forming new memories, and replications with different material have found deficits of 20 to 40 percent [1]. In rested people the hippocampus, which Walker describes as the brain's informational inbox, lit up during learning; in sleep-deprived people there was almost no signal, as if new files were bouncing back [1].

On cramming specifically, he concedes the short-term point. Staying up to study does help somewhat the next day [1]. But in unpublished data from his lab, when people were tested a month later, those who had slept retained far more, while the sleep-deprived group had lost almost everything [1]. Huberman, in that conversation, calls this the cramming effect: information that never makes it into long-term memory [1]. Walker's advice is to treat sleep the night before learning as an investment, not lost time [1]. He once wrote a Harvard Crimson editorial arguing that end-of-semester exam crunches force students into sleep deprivation, which he says drew a chilly response [1]. His line on the general principle: "When sleep is abundant, minds flourish" [1].

Walker also shows why one long push runs out of road within a single day. His lab had two groups learn a large set of facts at midday and then learn new material six hours later, at 6 p.m. [1]. The group that stayed awake learned worse in the evening, while a group that took a 90-minute nap was restored and even slightly better, a difference of about 20 percent [1]. He compares the hippocampus to a USB stick with limited space; sleep transfers files to the cortex, the hard drive, so there is room to learn again [1].

His motor-skill work makes the same case for practice. People practiced a finger-tapping sequence in 12 trials. After 12 hours awake, performance was unchanged; after a night of sleep, speed improved about 20 percent and accuracy almost 37 percent [1]. His conclusion is that it is "practice with a night of sleep that makes or leads to perfection" [1].

Andrew Huberman: gaps inside the session, and a schedule around it

Huberman's solo episode on studying reframes the whole task as offsetting forgetting: most new information is naturally discarded, so good studying is whatever protects what you want to keep [2]. In his episode on memory he describes Ebbinghaus's method from the other side. Ebbinghaus repeatedly read sequences of words or numbers and wrote down as many as he could recall; early on it took many repetitions, later far fewer, a reduction he called savings [4]. The material got no more interesting with each pass, yet repetition alone was enough [4]. Huberman defines the result: "memory is simply a bias in which perceptions will be replayed again in the future" [4].

From there he offers gaps at three scales.

Seconds. Gap effects are brief pauses of 5 to 30 seconds during learning, during which, he reports, neurons in the hippocampus replay the new information at 20 to 30 times the normal rate [2].

Minutes. Interleaving, as he uses the term, means sprinkling small, loosely related anecdotes among challenging material, like his stories about his bulldog Costello [2]. These breaks seem to help the brain repeat the core material and connect it to existing knowledge [2].

Days. In a survey of close to 700 medical students, the top performers scheduled dedicated study time, studied about three or four hours a day split into two or three sessions, and did so at least five days a week [2]. Huberman acknowledges the limits of self-report data and of reading causes into correlations [2]. He explains why regular timing matters: focus is a limited but renewable resource, because adenosine builds up the longer you are awake and resets with sleep, and studying at the same times each day lets the brain entrain to that rhythm, typically within two or three days [2].

He applies the same logic to self-testing. In experiments comparing a test right after first exposure, evenly spaced tests, or tests only shortly before the final exam, the group that tested itself very soon after first exposure did best, and the group that waited a long time and then crammed two tests back to back did worst, with spaced testing in between [2].

And he puts the night in the schedule. The actual rewiring of connections happens during deep sleep and sleep-like states, not while you study, and research on the first night effect shows that material learned on a given day is mostly consolidated during that night's sleep [2]. So he urges students to protect the night after learning and to be wary of late caffeine and all-nighters, which undermine the learning they are meant to serve [2].

Where They Agree, and Where They Don't

Agreement. All three reach the same verdict on cramming, and with the same qualification. It buys something for the next day [1][3] and costs nearly everything over weeks [1][3]. All three treat forgetting as the default condition that study has to work against [2][3]. And all three locate the real work of memory in the intervals: Ranganath in the space between sessions [3], Walker in the sleep before and after [1], Huberman in pauses, sleep and rest [2].

Differences and loose ends.

• How long should the gap be? Ranganath's example spaces two sessions a week apart [3]. The top medical students Huberman cites study two or three times a day, most days [2]. And for self-tests, the experiments Huberman describes favor a first test the same day or the next over evenly spaced ones [2]. These are answers to slightly different questions (how to space review, how to arrange a study day, when to retrieve first), and the sources do not give one combined schedule.
• Which sleep does the work? Huberman says rewiring happens particularly in rapid eye movement sleep in the latter part of the night [2]. Walker credits fact memory to deep non-REM slow waves and sleep spindles, which move memories from the hippocampus to the cortex, and assigns REM sleep to linking memories into creative associations [1]. For motor skills he points to stage 2 spindles, densest in the last quarter of the night [1]. Both end up telling students not to cut the late hours of sleep, but they give different reasons.
• How fast is replay? Huberman says the hippocampus replays material during waking pauses at 20 to 30 times normal speed, much as in REM sleep [2]. Walker reports replay 10 to 20 times faster during non-REM sleep in rats, and says that in REM sleep replay slowed to about half of waking speed [1]. The two accounts do not match on REM, and the sources do not resolve it.
• Caffeine. Walker says there is some evidence that caffeine enhances hippocampal encoding, though no one has tested whether it rescues learning after sleep deprivation [1]. Huberman reports that the best window for an adrenaline-raising aid such as caffeine is late in a learning session or right after it, not before [4], and separately warns against late caffeine because of sleep [2].
• How soon to sleep. Huberman stresses the first night [2]. Walker agrees it matters but adds that there is no need to rush to bed: the brain can hold new memories for about 16 hours until sleep consolidates them [1].

Why It Works

Forgetting is steep, and relearning is cheap. Ebbinghaus lost about 60% within a day [3], but each return to the material took fewer repetitions [4]. Steep loss with cheaper relearning is the pattern a spaced schedule relies on.

New context each time. Ranganath's lab has a computational model, which he notes is still under revision, of why spacing helps [3]. Any study session is tied to a particular place and time, with all its background cues [3]. Recalling the material in different contexts strips away those context-specific details and strengthens the core content, which is why well-rehearsed memories feel like facts and not vivid scenes [3]. A single long session supplies only one context.

Errors need time to appear. Ranganath links spacing to the testing effect, where the mismatch you discover on trying to recall is what improves retention [3]: "If you are feeling like you're coasting, then you're actually not learning" [3].

The inbox fills. Walker's nap study suggests the hippocampus loses capacity across a waking day and sleep restores it [1]. Animal work he cites shows that sleep restriction makes hippocampal synapses resistant to forming new connections [1].

Consolidation happens offline. Walker describes two mechanisms for facts: transfer from hippocampus to cortex during deep non-REM sleep, and replay of waking firing patterns [1]. He cites the Jenkins and Dallenbach study, which he dates to 1929 and which had only two participants but has been replicated many times: nonsense syllables were rapidly forgotten across waking hours, but after two to three hours of sleep the memories stopped decaying [1]. Huberman's gap effects are the waking counterpart, by his account [2].

Chemistry sets what gets kept. Ranganath says the brain runs on something like 12 to 20 watts and so prioritizes, with neuromodulators released during surprise, fear or stress promoting lasting memories [3]. He adds that memories stored in overlapping pools of neurons compete and can wipe each other out [3]. Huberman's memory episode makes a parallel point about adrenaline and timing: "adrenaline is the final common pathway" by which experiences are stamped in, the useful spike comes late in or just after learning, and staying amped up all the time blunts the effect because what matters is the change relative to the previous hour or two [4]. He cites work showing chronic stress impairs learning and memory while acute spikes enhance them [4]. By that logic, a sleepless, caffeinated exam week looks more like the first condition than the second, though Huberman does not put it in those terms.

Where It Applies, and Where It Doesn't

• If you only need it tomorrow, cramming can work. Both Ranganath and Walker say so [1][3]. The advice here is for anything you want to keep.
• It covers facts, languages and skills. Ranganath mentions languages and facts [3]; Walker's finger-tapping results extend it to physical practice, where sleep does more than preserve and actually improves performance [1].
• A bad night is not a catastrophe. Huberman says it does not erase what you learned [2], and Walker says memories can wait about 16 hours for sleep [1].
• Some of the evidence is thin or provisional. Walker's month-later result is unpublished [1]. Ranganath's model is under revision [3]. The medical student survey is correlational [2]. The Edina SAT figures Walker mentions in the same episode have been questioned, as he notes [1].
• Schedules are individual. Walker says the best learning window depends on your own circadian rhythm: perhaps 10 to 11 a.m. for an early riser, closer to midday or 1 p.m. for him, with a dip sometime between 1 and 4 p.m. and a second window after it [1].
• Stimulants and cold exposure are optional and carry risk. Huberman says the minimal effective dose of caffeine can be zero and that people should check with their doctor, and he opposes non-prescribed use of drugs like Ritalin, Adderall or modafinil [4]. Anyone with a sleep problem or a medical condition should get a professional's advice before changing sleep, caffeine or medication habits.

What To Do With It

1. Split the hours. Given a fixed amount of study time, divide it across days instead of spending it in one sitting; Ranganath's example is half an hour today and half an hour a week later [3].
2. Build a daily rhythm. Follow the pattern of the top medical students: three or four hours split into two or three sessions, at least five days a week, at consistent times [2].
3. Pause for 5 to 30 seconds during study. Stop taking in new material and let the last piece settle [2].
4. Break up hard material. Mix in a short, loosely related story or example, as Huberman does in his own teaching [2].
5. Retrieve early, then again later. Test yourself the same day or the next [2], and keep returning with a spaced repetition tool such as Anki, which Fridman uses daily [3].
6. Sleep before you learn. Walker's all-nighter group formed 40 percent fewer new memories [1].
7. Protect the night after, including its last hours. That is when the day's material is consolidated [1][2].
8. Use a nap or rest between sessions. A 90-minute nap restored afternoon learning in Walker's study [1], and Huberman recommends 10 to 20 minutes of non-sleep deep rest [2].
9. Study at your circadian peak and plan around the afternoon dip [1].
10. Cram only on purpose. If you need material for a single near deadline and never again, a last push will serve [3]; for anything cumulative, it is the schedule most likely to leave you with nothing a month on [1].

Sources

1. Dr Matt Walker Using Sleep To Improve Learning Creativity Memory · Huberman Lab
2. Optimal Protocols for Studying & Learning · Huberman Lab
3. Charan Ranganath on How Memory Works: Imagination, Deja Vu and False Memories · Lex Fridman Podcast
4. Understand & Improve Memory Using Science-Based Tools · Huberman Lab