Should spaced repetition algorithms consider your chronotype?

Does when you review your Anki cards matter? The answer is yes, and over a sustained period of time the timing of when you review your spaced repetition material probably affects the strength of your memory in a large and sustained manner.

But while academic research on spaced repetition has accumulated a compelling body of evidence, there is another burgeoning line of inquiry which suggests that when we do intense cognitive tasks like active recall can have large impacts on our performance in those tasks.

Given that, even small improvements in spaced repetition quality could lead to large improvements in user memory (by compounding over years), getting this right is extremely important for power (and regular) users of spaced repetition.

The idea behind spaced repetition is simple: the human brain creates memories by forming and strengthening the neural pathways that link cues to their target responses. By default, most such associations weaken over time, causing us to forget these. Spaced repetition hacks this by strengthening the associations that pair the respective cues and responses on successful retrieval, to reduce forgetting. The act of successful, active retrieval directly strengthens the neural associations in the brain. Up to a point, the more effortful this is, the stronger those neural pathways become.

By the above two principles, what becomes clear is that if someone actively recalled specific facts from their memory at appropriately spaced intervals, those specific facts would be almost etched into their memory by repeated effortful active recall.

Given that the act of effortful retrieval is so important to encoding memory, doing it in a way that optimizes the conditions for retrieval practice would be essential to ensuring the success of spaced repetition. But if effortful retrieval is so crucial to memory formation, when we perform this retrieval might be just as important as how often we do it.

—- Most people intuitively describe themselves as “morning people” or “night owls”. This intuitive feeling is supported by decades of established research. Since the 1970s, researchers have discovered that there are differences in people’s cognitive capacities over the course of the day. Some people actually do their best work in the morning, and others do it at night. Working in sync with our “chronotype” – our natural circadian rhythm patterns and peak timing for cognitive performance – improves our productivity. But the evidence is mixed and is highly dependent on the specific task.

In a landmark 1998 study, researchers May and Hasher discovered something surprising: when participants tried to learn new information at their non-optimal times of day, they couldn’t suppress their initial, incorrect responses. For example, after being prompted with the phrase “turn off the…,” a participant’s instinctive answer of “lights” remained highly active in their working memory, even after being told the new, correct answer was “stove.” As a result, they not only failed to learn the correct answer but sometimes even strengthened their memory for the wrong one, showing that the ability to learn from our mistakes is critically dependent on our optimal timings. This wasn’t a subtle effect. The timing difference alone explained more than one-fifth of the variation in whether participants successfully learned the correct information (ηp² = 0.203)

However, when it came to tasks that relied on accessing well-learned knowledge, which were simple and required little inhibition, participants didn’t do any better or worse at their optimal times. Their performance on things like vocabulary tests and basic judgments was remarkably consistent, demonstrating that the effect is not a global performance boost, but a specific benefit for the difficult work of managing distraction and controlling thought.

A similar, more memory-specific effect was powerfully demonstrated in a study by Maylor and Badham (2018). In their experiment, they tested participants’ ability to remember individual words versus their ability to remember the specific links, or associations, between word pairs. They found only minor and non-significant synchrony effects on the “easy” task of item memory, where participants simply had to recognize if they had seen a single word before. However, they found large and dramatic synchrony effects on the “hard” task of associative memory, which required participants to recall if two specific words had been presented together as a pair.

The effect size for this difficult associative task was substantial, explaining over a fifth (23.3%) of the performance difference between chronotypes based on the time of day. The researchers hypothesized that this is because item memory relies on a fast, automatic feeling of familiarity, while associative memory demands slow, effortful recollection; a controlled executive process highly vulnerable to low circadian arousal. Crucially, this distinction between automatic familiarity and controlled recollection is the exact same mechanism of “controlled vs. automatic” processes first identified by May and Hasher in their 1998 paper, now applied with greater precision to the specific functions of memory.

This was also seen in a systematic review by Chauhan et al. (2025), which analyzed 65 studies on the topic. The review found that while a majority of the studies (over 50%) reported no significant overall effect, a clear pattern emerged: the synchrony effect was reliably found only in the subcategory of cognitively demanding tasks requiring attention, inhibition, and memory. —- What does this mean for spaced repetition? I think at the very minimum it means that there’s a large boost to improving performance in a single review for difficult flashcards by just doing it at the right time of the day. While we can’t actually translate the percentage of variance explained into perfect accuracy calculations, we can do a small back of the envelope calculation. For the sake of argument assume that timing effects might create a half standard deviation variance in performance. This is a fair assumption considering that: (1) the studies showed timing explaining over 20% of performance variance, (2) this translates to what researchers consider large effect sizes, and (3) assuming half a standard deviation provides a conservative estimate for practical planning purposes.

Let’s imagine a user reviewing a batch of 100 difficult cards, where the overall average success rate is 70%. A user reviewing at their non-optimal time (trough) might perform half a standard deviation below the average. If a standard deviation is ~15 percentage points, their accuracy could drop to around 62.5%. They would fail 37 of the cards.

A user reviewing at their optimal time (peak) might perform half a standard deviation above the average. Their accuracy could rise to around 77.5%. They would only fail 22 of the cards.

The difference is stark. In a single session, the user reviewing at their worst time has to re-learn almost 70% more failed cards than the user reviewing at their best time.

This single-session difference is where the true, long-term impact begins. Spaced repetition is a compounding system. The user at their peak strengthens memories more effectively, leading to longer future intervals and a more efficient learning process. The user at their trough constantly resets card intervals, gets bogged down re-learning failed material, and risks falling into “leech hell,” sabotaging the very efficiency the algorithm is designed to provide. If these effects compound over an extended period of time, then the user who uses it only at the peak outperforms the user who uses it only at the trough substantially.

The above is just an extreme, illustrative example with made up numbers. While it is likely that the extent of it might be lower or higher, I think that this is directionally correct and provides an easy low-cost change to your spaced repetition workflow for a potentially large benefit.