Your Brain on Encoding: Why How You Study Matters More Than How Long

By Minerva Next Team | | 10 min read

Harvard's latest memory research reveals why deep encoding beats hours of rereading - and why AI shortcuts may weaken your recall.

You just spent four hours reviewing your notes. You highlighted the important parts, reread the tricky sections, and even reorganized your study guide. But when the exam lands on your desk, your mind goes blank.

Sound familiar? You are not alone. A 2024 Kahoot survey found that 96% of students rely on rereading notes as their primary study method - the same strategy that decades of research consistently ranks as one of the least effective. The problem is not that you are not studying enough. It is that your brain never truly encoded the information in the first place.

New research from Harvard is shedding light on exactly how memories form at the molecular level - and the findings have direct implications for how you should be studying. Here is what the science says, and what you can do about it.

What Encoding Actually Means (And Why It Matters)

When you encounter new information, your brain does not simply record it like a camera. It actively constructs a memory through a process called encoding - transforming sensory input into a neural representation that can be stored and later retrieved.

Think of it like this: reading a fact is like writing on water. Encoding that fact - connecting it to what you already know, questioning it, generating it from memory - is like carving it into stone.

This idea has deep roots in cognitive science. In 1972, psychologists Fergus Craik and Robert Lockhart proposed the levels of processing framework, arguing that the depth at which you process information determines how well you remember it. Shallow processing (recognizing a word's font or color) produces weak memories. Deep processing (understanding meaning, making connections) produces strong ones.

A 2024 study published in Memory & Cognition by Peng, Logie, and Della Sala put hard numbers on this. They found that deep processing produced substantially better memory performance (d' = 2.61-2.83) compared to shallow processing (d' = 1.83-2.11), with large effect sizes. The critical finding? The rate of forgetting was the same regardless of encoding depth. Deep encoding does not slow down forgetting - it gives you a fundamentally higher starting point that persists over time. The gap between deep and shallow processing never closes.

In other words, how you study in the first hour determines what you will still know in a week.

Inside Your Synapses: Harvard's Memory Breakthrough

In May 2025, Harvard researchers published a groundbreaking study in Nature Neuroscience that mapped memory formation at a level of detail never achieved before.

The team developed a technique called EPSILON (Extracellular Protein Surface Labeling in Neurons) that tracks proteins critical to synaptic plasticity - specifically AMPARs, the receptors that strengthen connections between neurons when you learn something new. Using sequential fluorescent dye labeling and advanced microscopy, the researchers could visualize where and how much synaptic strengthening had occurred.

"We can look at the history of the synaptic plasticity, studying where and how much of the synaptic potentiation has happened," explained Doyeon Kim, a Harvard graduate student and co-author of the study.

What does this mean for students? It confirms something cognitive scientists have long suspected: memory is not a passive recording. It is an active construction process that physically reshapes your neural connections. The stronger the encoding event - the more deeply you engage with material - the more robust the synaptic changes. Weak engagement produces weak synaptic traces that fade quickly.

Adam Cohen, senior co-author and Harvard professor of Chemistry and Physics, described the technique as providing "a lens into the synaptic architecture of memory, something previously unattainable in such detail."

This is not abstract neuroscience. Every time you passively reread a textbook, you are producing minimal synaptic change. Every time you struggle to recall a concept from memory, explain it in your own words, or connect it to something you already know, you are building the kind of robust synaptic architecture that persists.

The AI Shortcut Problem

Here is where things get uncomfortable for the 70% of students now using AI tools for studying.

A 2026 study by Andre Barcaui tested 120 undergraduates, randomly assigning them to study with ChatGPT or traditional methods. On a surprise retention test 45 days later, traditional learners scored 68.5% while ChatGPT users scored just 57.5% - an 11 percentage point gap. The AI group also studied significantly less (3.2 hours vs. 5.8 hours), and the performance gap widened for technical topics.

"There is an abysmal difference between delivering a piece of work and understanding the process of its creation," Barcaui noted.

A separate study from Corvinus University of Budapest found even more alarming results: students with unrestricted AI access showed a 20-40 percentage point decline in genuine understanding. On paper tests without AI, students could answer only about 10-20 out of 50 true-or-false questions correctly - barely above random guessing.

The pattern is consistent across studies. A review in Frontiers in Psychology found that while AI platforms can produce short-term test score improvements, students who relied on ChatGPT for practice actually performed worse on subsequent exams compared to non-users.

As learning science writer Carl Hendrick put it bluntly: "LLMs are engineered for frictionless and user-friendly task completion, not for the friction-filled process of learning."

This is the encoding problem in action. When AI removes the cognitive effort from studying - summarizing a chapter for you, generating flashcard answers, explaining concepts so clearly you never have to struggle - it simultaneously removes the deep processing that creates durable memories. The experience feels productive. The synaptic changes tell a different story.

Five Encoding Strategies That Actually Work

The good news is that decades of research, synthesized in John Dunlosky's landmark review of study techniques, points to specific strategies that force deep encoding. Here are five you can start using today.

1. Practice Testing (Retrieval Practice)

Close your notes and try to recall what you just learned. Write it down, say it out loud, or quiz yourself. This is not just a way to check what you know - the act of retrieving information from memory physically strengthens the neural pathways involved.

Dunlosky's review rated practice testing as the highest-utility study strategy available: "Taking practice tests (versus merely rereading the material) can substantially boost student learning." A study of 623 biology students found that self-quizzing correlated with 4-5% higher exam scores, while passive strategies like rereading showed no benefit or even negative effects.

2. Elaborative Interrogation

Instead of accepting a fact at face value, ask why it is true and how it connects to what you already know. If you are studying that mitochondria produce ATP, ask: why do cells need a dedicated organelle for energy? How does this relate to what happens when you exercise?

This forces your brain to integrate new information with existing knowledge - exactly the kind of deep processing that produces strong encoding. Dunlosky rated it as a moderate-to-high utility strategy.

3. Self-Explanation

As you work through problems or read new material, explain each step to yourself. Why does this step follow from the last one? What principle is being applied here? How would I explain this to someone else?

Research shows this is remarkably effective: "Final test performance was three times better for students who self-explained," according to Dunlosky's review. The key is that self-explanation forces you to identify gaps in your understanding - and filling those gaps is where deep encoding happens.

4. The Generation Effect

Generating information from memory is consistently more effective than passively receiving it. A meta-analysis of 86 studies found that generated information is remembered approximately 30-50% better than passively read material, with an effect size of d = 0.40.

In practice, this means: instead of rereading a definition, try to write it from memory first. Instead of reviewing solved examples, attempt the problem before looking at the solution. The struggle is not a sign of failure - it is the encoding process at work.

5. Spaced and Interleaved Practice

Distribute your study sessions across multiple days rather than cramming. Mix different topics or problem types within a single session. Both strategies feel harder in the moment - Dunlosky notes that "interleaved practice slows learning but leads to much greater retention."

This is counterintuitive. Massed practice (studying one topic until you feel confident, then moving on) feels more productive. But the feeling of fluency is deceptive. Spacing and interleaving force your brain to repeatedly reconstruct memories from scratch, strengthening the encoding each time.

Making This Work in the Real World

None of this means you should throw away your AI tools or burn your highlighters. The goal is to be strategic about when you engage in deep processing versus when you use shortcuts.

Here is a simple framework: use AI and passive methods for the discovery phase - getting an overview of a topic, finding resources, organizing your notes. But when it is time to actually learn the material, switch to active encoding strategies. Close the AI chat. Put away your notes. Retrieve, generate, explain, and question.

The Kahoot survey found that 54% of students believe better study habits would improve their academic performance. They are almost certainly right. The research is clear: it is not about studying more. It is about studying in ways that force your brain to do the hard work of encoding.

Your synapses will thank you on exam day.

At Minerva Next, we're building a learning platform that helps you study smarter - not longer. By connecting your study materials into one structured workflow, we help you spend less time organizing and more time actually learning.

References

  1. Kahoot! (2024). Study Habits Snapshot 2024. https://kahoot.com/press/2024/10/29/study-habits-snapshot-2024/

  2. Peng, N., Logie, R.H., & Della Sala, S. (2024). Effect of levels-of-processing on rates of forgetting. Memory & Cognition. https://pmc.ncbi.nlm.nih.gov/articles/PMC11868305/

  3. Harvard Gazette. (2025). Tracking precisely how learning, memories are formed. Nature Neuroscience. https://news.harvard.edu/gazette/story/2025/05/tracking-precisely-how-learning-memories-are-formed/

  4. PsyPost. (2026). ChatGPT acts as a "cognitive crutch" that weakens memory, new research suggests. Based on Barcaui, A. (2026), Social Sciences & Humanities Open. https://www.psypost.org/chatgpt-acts-as-a-cognitive-crutch-that-weakens-memory-new-research-suggests/

  5. Benedek, M. & Sziklai, B.R. (2025). Impact of AI Tools on Learning Outcomes: Decreasing Knowledge and Over-Reliance. Corvinus University of Budapest. https://arxiv.org/html/2510.16019v1

  6. Jose, B. et al. (2025). The cognitive paradox of AI in education: between enhancement and erosion. Frontiers in Psychology. https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2025.1550621/full

  7. Hendrick, C. (2025). The Algorithmic Turn: The Emerging Evidence on AI Tutoring. https://carlhendrick.substack.com/p/the-algorithmic-turn-the-emerging

  8. Dunlosky, J. (2013). Strengthening the Student Toolbox: Study Strategies to Boost Learning. American Educator, Fall 2013. https://www.aft.org/ae/fall2013/dunlosky

  9. Walck-Shannon, E.M., Rowell, S.F., & Frey, R.F. (2021). To What Extent Do Study Habits Relate to Performance? CBE Life Sciences Education, 20(1). https://pmc.ncbi.nlm.nih.gov/articles/PMC8108503/

  10. Structural Learning. The Generation Effect: Why Creating Information Beats Reading. https://www.structural-learning.com/post/generation-effect-active-learning