← All self-study courses

Read sleep-and-memory papers with confidence (10 sessions, 15 minutes each, v1.0)

Goal of this course: Build the background to read our lab's papers on your own.

This is a self-study course for undergraduates in medicine and the sciences. In 10 sessions of 15 minutes each, the goal is to become able to explain, in your own words, the abstracts of published papers from the Department of Neurophysiology.

A Japanese version of this course is available (講座_研究室の論文を読む基礎_v1.0.md). This English version has the same content.

Introduction (how to use this course)

What this is

• This is a teaching material published for students' self-study by the Department of Neurophysiology, Kobe University Graduate School of Medicine.
• Our lab's papers sit where four topics overlap: "sleep", "memory", "neurons newly born in the adult brain", and "how to measure neural activity". Even if you know one of them, without the others you will get stuck on the second line of an abstract. This course lays that foundation, 15 minutes at a time.
• Each session comes with one of our lab's papers in which that session's topic is actually used (all published, all with a DOI).

Who it is for

• Roughly first- to third-year undergraduates in medicine or the sciences.
• High-school biology (the basics of neurons and genes) is enough background. The only math is averages and division in Session 8.
• It suits people who would like to try reading our lab's papers, or who want to prepare before the Basic Research Placement (kiso-haizoku) or a lab visit.

What you need

• A conversational AI (one that answers questions in a chat). The one you usually use is fine.
• Something to take notes on (paper or screen). Sessions 1, 7, and 8 include simple calculations by hand.
• In Session 8 only, you can run a short piece of code (Python) if you wish. A free environment that runs in the browser is enough. You can complete the course without running it.
• A way to open paper web pages. Putting https://doi.org/ in front of a DOI (the identification number of each paper) opens the paper's page.

How to proceed

1. Load this file into your conversational AI (attach the file or paste it in).
2. Type "Please start Session 1."
3. Answer the AI's questions in your own words. It is fine to be wrong. The explanation starts from where you went wrong.
4. Stop after 15 minutes. The next day, load the same file again and type "Please start Session 2." - Each session is sized to finish in 10–12 minutes. Use the remaining time to ask again about anything you did not understand.

Cautions

• The AI can make mistakes. Check important points against textbooks or the papers themselves.
• Do not enter personal information about yourself or others, medical-care information, or other people's data into the AI.
• This course is not intended as medical advice. For concerns about your body or sleep, please consult a medical institution.
• You do not need to paste the text of papers into the AI. Read the abstract on the paper's page, and show the AI a summary you have written in your own words.

Request for feedback (please be sure to send it)

This course is revised every time based on the voices of its users. If you used it, please be sure to send feedback.

• Send to: the address listed on the Contact page of the lab website
• Subject: "Course feedback: Basics for Reading Our Lab's Papers"
• What we would like you to write (4 points): (1) how many sessions you completed (2) what was hard to understand (3) what was good (4) how long one session took

Feedback from people who stopped partway is also very helpful.

Terms of use

• Use for personal study or in classes is free.
• If you distribute a modified version, please contact the lab (via the Contact page).
• The content is current as of October 2026.

List of all 10 sessions

1. Sleep stages and EEG — humans and mice
2. Stages of memory — the hippocampus and amygdala
3. Sleep and memory consolidation
4. Neurons newly born in the hippocampus
5. Ensembles of cells that hold a memory (engrams)
6. How to measure and manipulate neural activity
7. How to read behavioral experiments — fear conditioning
8. How to read statistical figures
9. Reading one of our lab's papers (1)
10. Reading one of our lab's papers (2), and asking your own question

Instructions for the AI (how to play the teacher)

This section is for the AI that has loaded this file. You do not need to read it aloud to the learner.

• Role: You are a teacher giving one-on-one lessons to an undergraduate. Teach using only the "Session N" section of this file.
• Time: Each session is cut off at 15 minutes. The explanation is sized to finish in 10 minutes. If it looks like it will go over 15 minutes, tell the learner that the rest can be moved to the next session.
• First question (only once, at the start of Session 1): "What is the one thing you most want to understand through this course? Please tell me one thing." Match the answer to the "Correspondence table" below, and tell the learner in one line which session will be especially useful. In that session, tailor your analogies to the learner's interest.
• Start of Session 2 onward: Ask just one recall question from the "Summary" of the previous session.
• Flow of a session: state the aim in one line → give the key points of the explanation one at a time → ask one question per key point → wait for the answer → hands-on task → check question → at the end, 3 key points and 2 mini check questions → ask about understanding on a 3-level scale (understood well / mostly / still unsure) → end by previewing the title of the next session in one line.
• Ask only one thing at a time: After asking a question, wait for the learner's answer. Do not give the answer first.
• Rephrase: If the learner seems not to understand, change the analogy, rephrase one step more simply, and ask one more question. If correct, acknowledge briefly and move on.
• Be short: Each turn is within 100 words. Go longer only when the learner says "tell me more." Do not move on while staying long-winded.
• Do not move to the next session on your own: Do not start the next session until the learner says "Please start Session N."
• Terms: Briefly paraphrase specialist terms when they first appear. When introducing an English abbreviation, give the full term followed by the abbreviation (e.g. conditioned stimulus (CS)).
• Do not make things up: Answer numbers and conclusions of papers only within what is written in this file. If asked about something not written here, answer "Let's check it together in the text of the paper," and indicate where to read (which figure, which section). Mark general statements with "according to textbooks." Say so when your confidence is low.
• About the lab: If asked about the lab's research, clinical studies, or specific people, do not answer by guessing; direct the learner: "Please check within what is published on the Research and Publications pages of the lab website."
• Decline: If the learner writes a medical consultation (their own or a family member's symptoms, medications, sleep problems) or personal information, do not deal with its content; say "This course cannot handle medical consultations or personal information. For medical consultations, please contact a medical institution," and return to the course.
• Text of papers: If the learner tries to paste the full text of a paper, tell them they do not need to. Ask them to read the abstract on the paper's page and show a summary written in their own words.
• End of the final session: Work with the learner to put together the 4 points of the "Request for feedback," and give it to them as text they can send by e-mail as is.

Correspondence table (what you want to understand → session)

• Sleep stages, EEG, REM sleep, sleep in mice → Session 1
• Stages of memory, hippocampus, amygdala, extinction → Session 2
• The relation between sleep and memory, brain rhythms during sleep → Session 3
• Neurons born in the adult brain (adult neurogenesis) → Session 4
• Cells that hold a memory, tools using genetically modified mice → Session 5
• Calcium imaging, optogenetics, electrophysiology, closed-loop → Session 6
• Fear conditioning, freezing, control groups → Session 7
• Error bars, n, p-values, effect size → Session 8
• How to read papers, the flow of the lab's research → Sessions 9 and 10
• Asking your own research question → Session 10

Session 1　Sleep stages and EEG — humans and mice

Lab paper: Tezuka T, Kumar D, Singh S, Koyanagi I, et al. Real-time, automatic, open-source sleep stage classification system using single EEG for mice. Scientific Reports. 2021;11:11151. DOI: 10.1038/s41598-021-90332-1 (the full text can be read free of charge)

Aims of this session

1. Distinguish the sleep stages (the three NREM stages and REM) by brain waves, muscle activity, and eye movements.
2. Draw the shape of a night's sleep (cycles of about 90 minutes) as a figure.
3. Name three ways in which mouse sleep differs from human sleep.

Explanation

Question: Does the "sleep record" on a wristwatch measure the stages of sleep? Answer: Most of them "estimate" stages from heart rate and body movement. The yardstick that decides the stages is still the EEG.

• Key point 1　There are two kinds of sleep: Stages are decided by looking at three things: the electroencephalogram (EEG), the electromyogram (EMG), and eye movements. In humans, one stage is assigned to each 30-second segment (an epoch).
• NREM (non-rapid eye movement) sleep: sleep in which the EEG becomes slow and large.
• REM (rapid eye movement) sleep: the EEG is close to that of wakefulness, yet the eyes move rapidly and the muscles below the neck are almost completely relaxed. Vivid dreams are reported more often in REM.
• Key point 2　Human NREM has three stages: N1 = drowsiness. N2 = real sleep, in which waves with shapes called sleep spindles and K-complexes appear. N3 = sleep with many large, slow waves (slow waves).
• "N3 is deep sleep" holds when measured with two yardsticks: (1) it is hard to wake up from (you hardly wake without a loud sound), and (2) the EEG is slowest and largest. But compared with another yardstick, (3) how relaxed the muscles are, REM is deeper. "Depth" is not a single thing; the ranking changes depending on what you measure. When a paper says "deep sleep", check which yardstick it means.
• The standard is for trained people to score stages by eye, following the manual of the American Academy of Sleep Medicine (AASM).
• Key point 3　The shape of a night: The NREM → REM cycle repeats 4–5 times in a night. One cycle is about 90 minutes. N3 is abundant in the first half, and REM becomes longer toward morning. A figure that plots the stages of a night in time order is called a hypnogram.
• Key point 4　Mice sleep "in the daytime, in short pieces": Mice are nocturnal and sleep mostly during the light period. Time is written in ZT (Zeitgeber time, a clock set by light): ZT0 = the time the lights come on, ZT12 = the time they go off. Mice repeat sleeping and waking a few minutes at a time throughout the day (polyphasic sleep), and stages are divided only into NREM and REM. In REM, the hippocampal theta waves (regular waves of around 6–9 Hz) stand out.
• Our lab's paper (Tezuka et al. 2021) built a system that automatically scores the stage on the spot (in real time) from a single EEG channel in mice. Mouse REM is short and comes only sparsely, so knowing "it is REM now" on the spot is a prerequisite for experiments that examine something only during REM (this connects to Sessions 6 and 9).

Figure　A night's hypnogram (human, schematic)

Hands-on task (hand calculation, 3 minutes)

Here is one person's night. They got into bed at 23:00, fell asleep at 23:20, were awake for a total of 30 minutes during the night, and woke up at 6:00 and got out of bed.

1. How many minutes did it take to fall asleep (sleep onset latency)?
2. What percentage of the time in bed was spent asleep (sleep efficiency)? (to one decimal place)
3. In a mouse room where the lights come on at 8:00, what ZT are 14:00 and 21:00?

Check questions

1. When N3 is called "deep sleep", in what sense is it deep? Answer by comparing it with REM.
2. Which stage becomes longer toward morning?
3. Give two ways in which mouse sleep differs from human sleep.

Summary

• Stages are decided by EEG, EMG, and eye movements: NREM (N1–N3) and REM.
• About 90-minute cycles, 4–5 times a night. N3 in the first half, REM toward morning.
• Mice sleep in short pieces during the light period, and sleep is counted as two stages, NREM and REM. Time is given in ZT.

Session 2　Stages of memory — the hippocampus and amygdala

Lab paper: Sakaguchi M, Kim K, Yu LMY, Hashikawa Y, et al. Inhibiting the Activity of CA1 Hippocampal Neurons Prevents the Recall of Contextual Fear Memory in Inducible ArchT Transgenic Mice. PLOS ONE. 2015;10(6):e0130163. DOI: 10.1371/journal.pone.0130163 (the full text can be read free of charge)

Aims of this session

1. Explain the four stages of memory (acquisition, consolidation, retrieval, extinction) using the example of studying for an exam.
2. Say what the hippocampus and the amygdala are each responsible for.
3. Distinguish "cannot recall" from "the memory has disappeared."

Explanation

Question: Has a formula you could not recall on the day of the exam disappeared from your head? Answer: Not necessarily. If it suddenly comes back to you on the way home, the memory was still there, and only the retrieval failed.

• Key point 1　Stages of memory: acquisition (when you learn) → consolidation (over hours to days, a fragile memory becomes stable) → retrieval (remembering). In research, the meaning of a result changes depending on which stage was manipulated.
• Key point 2　What the hippocampus and amygdala handle:
• Hippocampus: memories of events ("when, where, and what happened") and memories of places. It stores the surroundings (the context) as one bundle.
• Amygdala: attaches emotional value, such as "frightening" or "pleasant", to events. For memories of frightening experiences, it is easy to think of the hippocampus as handling "where" and the amygdala as handling "it was frightening."
• Key point 3　Extinction is not "overwriting": If a cue associated with a frightening experience is presented repeatedly with nothing happening, the fear response weakens. This is called extinction. Extinction is thought not to erase the original memory; instead, a new memory, "it is safe now", is laid on top. That is why the response can come back as time passes (spontaneous recovery).
• Key point 4　In our lab's paper: Sakaguchi et al. (2015) made mice whose neurons in a part of the hippocampus called CA1 carry a protein (ArchT) that suppresses neural activity in response to light. After the mice had associated a place with a frightening experience, shining light on CA1 only during the recall test made the mice much less fearful. However, the next day, when they were put in the same place without light, they did show fear.
• You might think "the memory was destroyed." But since it came back the next day, what was disrupted was not the memory but its "retrieval."

Figure　Stages of memory and when they are manipulated

Hands-on task (write a table, 3 minutes)

Which stage of memory does each of the following four experiments manipulate? On paper, write a table with the columns "experiment" and "stage."

1. Give a drug just before learning.
2. After learning, keep the animal from sleeping until the test the next day.
3. Stop neural activity only during the test.
4. Every day, play a tone that was associated with a frightening experience 20 times, with nothing happening.

Check questions

1. In memories of frightening experiences, what are the hippocampus and the amygdala each responsible for?
2. In Sakaguchi et al. (2015), what can we conclude from the fact that the mice showed fear on the day after the test?
3. Why can the response come back after extinction?

Summary

• Memory goes acquisition → consolidation → retrieval. Extinction adds a new memory on top.
• The hippocampus handles "where and when"; the amygdala handles "frightening or pleasant."
• "Cannot recall" does not necessarily mean "has disappeared."

Session 3　Sleep and memory consolidation

Lab paper: Koyanagi I, Akers KG, Vergara P, Srinivasan S, et al. Memory consolidation during sleep and adult hippocampal neurogenesis. Neural Regeneration Research. 2019;14(1):20-23. DOI: 10.4103/1673-5374.243695 (the full text can be read free of charge; a short review)

Aims of this session

1. Give one piece of evidence for "sleep stabilizes memory."
2. Say where in the brain three rhythms during sleep are generated.
3. State each of two theories about sleep and memory in one sentence.

Explanation

Question: Is sleeping after learning just a way to "rest" the brain? Answer: It is not only rest. During sleep, the brain is thought to be organizing the memories of the day.

• Key point 1　Evidence: In both humans and animals, it has been reported again and again that a group that sleeps after learning retains the memory better than a group that stays awake for the same length of time. In animals, it is known that preventing sleep after learning weakens the memory.
• Key point 2　Three rhythms during sleep and where they are generated:
• Slow oscillation: large waves of about one per second or slower, generated in the cerebral cortex. Prominent in NREM.
• Sleep spindles: bursts of fast waves at a dozen or so per second, generated in the thalamus (prominent in human N2).
• Sharp-wave ripples: waves of the hippocampus. The sharp wave comes from a region called CA3, and the very fast small waves riding on it (ripples) are easy to record in CA1.
• Bonus: theta waves, prominent in REM, are hippocampal waves whose rhythm is paced by input from a region called the medial septum.
• Key point 3　Memory replay: In rats and mice, it has been observed that hippocampal cells that were active when the animal passed through places in a certain order while awake become active again in the same order during subsequent sleep (mainly during sharp-wave ripples in NREM).
• Key point 4　Two theories:
• The "active consolidation" theory (as organized by Born and colleagues): when cortical slow oscillations, thalamic spindles, and hippocampal ripples occur in time with one another, hippocampal memories are replayed and gradually transferred to the cerebral cortex.
• The "synaptic homeostasis" theory (Tononi and Cirelli): neural connections (synapses) strengthened during waking are weakened overall during sleep, so that only the prominent connections remain.
• Many researchers think the two are not so much in conflict as looking at different aspects. Much about the role of REM sleep remains unknown, and our lab is investigating it (Sessions 9 and 10).
• In our lab's paper: In a short review, Koyanagi et al. (2019) summarize the following: neurons in a hippocampal region called the dentate gyrus (including neurons born in adulthood) are needed for memory; rhythms and replay during sleep may support consolidation; and animal studies show that depriving sleep makes it harder for new neurons to be born (though this is hard to separate from the effects of stress). They argue that experiments are needed that manipulate new neurons only during specific sleep stages.

Figure　Rhythms during sleep and where they are generated

Hands-on task (hand calculation, 3 minutes)

Take one cycle to be about 90 minutes (this is an average; each person sleeps differently).

1. If you sleep 7 and a half hours, about how many cycles is that?
2. If the same person wakes after 4 and a half hours, how many cycles is that? Combined with "REM is longer toward morning" from Session 1, which stage is especially likely to be reduced when sleep is short?

Check questions

1. Where in the brain are slow oscillations, sleep spindles, and sharp-wave ripples each generated?
2. What is memory replay?
3. State the "synaptic homeostasis" theory in one sentence.

Summary

• Sleep after learning helps memory consolidation.
• Slow oscillations come from the cortex, spindles from the thalamus, and sharp-wave ripples from the hippocampus. Theta waves also come from the hippocampus.
• A "transfer" theory and a "tidy-up" theory. The role of REM still holds many mysteries.

Session 4　Neurons newly born in the hippocampus

Lab paper: Arruda-Carvalho M, Sakaguchi M, Akers KG, Josselyn SA, Frankland PW. Posttraining Ablation of Adult-Generated Neurons Degrades Previously Acquired Memories. The Journal of Neuroscience. 2011;31(42):15113-15127. DOI: 10.1523/JNEUROSCI.3432-11.2011 (the full text can be read free of charge)

Aims of this session

1. Say where adult neurogenesis (the generation of new neurons in the adult brain) takes place.
2. Say that newly generated neurons have special properties during a "young" period.
3. Explain what we learn differently from removing these neurons "before learning" versus "after learning."

Explanation

Question: Are new neurons generated in the adult brain? Answer: In mice and rats at least, they continue to be generated. The old textbook statement that "neurons do not increase in the adult brain" was a rule with exceptions.

• Key point 1　Where: In a deep layer of the hippocampal dentate gyrus, the cells that give rise to neurons (neural stem cells) divide, and new neurons (granule cells, the main neurons of the dentate gyrus) continue to be generated. In mice there is also a stream from the wall of the ventricle toward the olfactory bulb (the entry point for smell). How long this continues into adulthood in the human hippocampus is still debated.
• Key point 2　Young neurons are special: A newly generated neuron takes several weeks to mature in shape and connect into the existing circuit. During the young period of a few weeks after birth, it is thought to respond readily to input and to change its connections readily. They are only a small subpopulation of the neurons of the dentate gyrus, but they are thought to be involved in memories such as telling apart similar places.
• Key point 3　How to see "only the new neurons": A method is used that places a genetic label only on the cells that were dividing during the period when a drug (e.g. tamoxifen) was given. Counting from the day the label was placed, one can tell "how many weeks old" a neuron is (the tools are covered in detail in Session 5).
• Key point 4　In our lab's paper: Arruda-Carvalho, Sakaguchi, et al. (2011; the two are co-first authors who contributed equally) developed a diphtheria toxin-based strategy in mice to selectively ablate a population of predominantly mature, adult-generated neurons, without affecting ongoing neurogenesis.
• Ablating them before learning did not prevent the formation of new contextual fear memories or water maze (a task of remembering the location of a platform in a pool) memories.
• In contrast, ablating an equivalent population after learning degraded the already-formed memories. Memory that does not rely on the hippocampus was not degraded.
• You might think "new neurons are probably for learning new things." But this result shows that the new neurons that were present at the time of learning become an enduring component of the formed memory. When they were ablated beforehand, the animals could still learn, which is thought to be because other neurons stepped in to compensate.

Figure　The life of a new neuron and the timing of removal

Hands-on task (table of predictions, 3 minutes)

Consider the following three groups and predict next week's memory test result as "normal / weak." Give a one-line reason as well.

1. A group with no intervention (control)
2. A group in which new neurons were removed before learning
3. A group in which new neurons were removed after learning Furthermore, to answer the objection "maybe the removal operation itself (introducing the toxin, the animal becoming unwell) did the damage," which group should be compared with which?

Check questions

1. What is the name of the place in the hippocampus where adult neurogenesis occurs?
2. In Arruda-Carvalho et al. (2011), how did the results differ between removal before and after learning?
3. From that difference, what can we say about new neurons?

Summary

• In the adult hippocampus (dentate gyrus) too, neurons continue to be generated (in mice and rats at least).
• Young new neurons are changeable and are involved in memory.
• Removing them after learning degrades the already-formed memory = they are a component of the memory.

Session 5　Ensembles of cells that hold a memory (engrams)

Lab paper: Sakaguchi M, Hayashi Y. Catching the engram: strategies to examine the memory trace. Molecular Brain. 2012;5:32. DOI: 10.1186/1756-6606-5-32 (the full text can be read free of charge; a review)

Aims of this session

1. State in one sentence what an engram (memory trace) is.
2. Explain with a figure the mechanism of "labeling only the cells that were active."
3. State the three conditions for calling a given ensemble of cells an engram.

Explanation

Question: Is a memory "stored" somewhere in the brain? Answer: In the first half of the 20th century there was a period when, no matter where the brain was lesioned, memory was not completely erased, and it was thought that "the place cannot be found." Today, an ensemble of cells that were active at the time of learning is thought to hold the memory.

• Key point 1　Engram: This is the physical trace that a memory leaves in the brain. It was named at the beginning of the 20th century by the German researcher Semon. In current research it is studied as "a small ensemble of cells that are active at the time of learning and active again at the time of recall."
• Key point 2　Labeling the cells that were active: When neurons are strongly active, there are genes that switch on immediately (immediate early genes). Representative ones are c-fos and Arc.
• Using the "switch" (promoter) of such a gene, a marker protein is produced only in the cells that were active.
• A drug is further used to set the "time window during which labeling is allowed." For example, in the Tet system, no label is placed while doxycycline (Dox) is in the diet, and a label is placed only on cells that were active during the period when Dox was withdrawn.
• Key point 3　"This type of cell" versus "the cells active at this time": Even with the same Tet system, the way cells are selected changes depending on what is used as the switch. Sakaguchi et al. (2015) in Session 2 used the switch of a gene called CaMKII, which operates in excitatory neurons, and selected cells by "type." With the c-fos switch, one can select by "whether a cell was active at that time."
• Other systems: Cre-loxP (a mechanism in which an enzyme called Cre excises the region between markers called loxP) selects "only this type of cell," and CreERT2 (Cre works only when tamoxifen is given) sets "only this time period." The "label only the new neurons" in Session 4 is of this type.
• Key point 4　Conditions for calling something an engram: (1) it was active at the time of learning; (2) it is active again at the time of recall; (3) silencing those cells makes the memory harder to express (necessity), or activating them produces the memory (sufficiency). Observation alone reaches only (1) and (2). (3) requires manipulation (Session 6).
• In our lab's paper: Sakaguchi and Hayashi (2012) wrote a review that organizes, in order, the methods for searching for the memory trace — from classical lesion experiments and drug experiments, to labeling with immediate early genes, to methods that bias which cells are used for a memory (increasing a protein called CREB), to manipulating labeled cells with light, and further to methods for looking at how synapses are arranged along the dendrites (the branches of a neuron that receive input).

Figure　Labeling only the cells that were active (example of the Tet system)

Hands-on task (write a schedule, 3 minutes)

Suppose you want to "label only the memory cells learned on Monday, and see on Friday's test whether those cells are active again." Using the c-fos switch and the Tet system (a label is placed only during the period when Dox is withdrawn), write the "Dox on / off" schedule from Sunday to Friday.

Check questions

1. What is an engram? In one sentence.
2. How does cell selection differ between the c-fos switch and the CaMKII switch?
3. Why is an observation-only experiment not enough to say "this ensemble of cells is an engram"?

Summary

• An engram = an ensemble of cells active during learning and active again during recall.
• Immediate early genes (c-fos, Arc) + a drug set "when" to place the label.
• There are three conditions. The third (necessity / sufficiency) requires manipulation.

Session 6　How to measure and manipulate neural activity

Lab paper: Carrier-Ruiz A, Sugaya Y, Kumar D, Vergara P, et al. Calcium imaging of adult-born neurons in freely moving mice. STAR Protocols. 2021;2(1):100238. DOI: 10.1016/j.xpro.2020.100238 (the full text can be read free of charge; a protocol)

Aims of this session

1. Name two tools each for "observing" (electrical, optical) and "manipulating" (optical, pharmacological).
2. Distinguish what "observing" can tell you (correlation) from what "manipulating" can tell you (close to causation).
3. Explain the difference between closed loop and open loop with a familiar example.

Explanation

Question: How can we tell that a neuron "was active"? Answer: Broadly, in two ways: listening with electricity, or watching with light.

• Key point 1　Measuring with electricity: The electroencephalogram (EEG) and electromyogram (EMG) determine the sleep stage. The local field potential (LFP) is the electrical signal of a local population in the brain, and theta waves and sharp-wave ripples are measured with it (Session 3). The activity of a single cell is measured with single-unit recording, or with patch clamp (a method that presses a fine tube against the cell membrane to measure voltage or current directly). Electricity can follow changes as fast as a thousandth of a second.
• Key point 2　Watching with light (calcium imaging): When a neuron is active, calcium inside the cell increases. A protein that glows brightly when it binds calcium (such as GCaMP) is placed in the neuron, and its light is watched.
• Mounting a miniaturized microscope (miniscope) on the head lets you see the activity of individual cells in a freely moving mouse. For deep brain regions, a thin rod-like lens (a GRIN lens) is inserted to reach them.
• It is slower than electricity, but its strength is that it can follow the same cells over many days.
• Our lab's paper (Carrier-Ruiz et al. 2021) is a protocol that describes, from surgery to analysis, a method of giving a glowing protein to the new neurons of the hippocampus with tamoxifen and watching them with a miniaturized microscope while simultaneously recording sleep with EEG and EMG. Because new neurons are few and rarely active, it also describes techniques for analyzing long recordings by dividing them into short segments and overlaying them.
• Key point 3　Manipulating with light (optogenetics): A protein that opens and closes in response to light is placed in the neuron, and light is delivered through a thin optical fiber. ChR2 excites neurons with blue light, while ArchT, GtACR, and others suppress them with light. There are two strengths: (1) you can manipulate "only at this moment" on the scale of a thousandth of a second; (2) you can manipulate in an animal while it is behaving.
• Key point 4　Manipulating with a drug (chemogenetics): An artificial receptor called DREADD is placed in the neuron, and a drug that acts only on that receptor is introduced. No optical fiber is needed and the effect is broad, but because the effect lasts tens of minutes to hours, it is not suited to fine temporal manipulation such as "only during REM."
• Key point 5　Closed loop and open loop: Take an air conditioner as an example. A timer that "runs for 10 minutes every hour" is an open loop. Measuring the room temperature and running only when it is hot is a closed loop.
• It is the same in experiments. Delivering light at a predetermined time without watching the brain's state is open loop. Reading the EEG on the spot and delivering light only when it is judged "this is REM now" is closed loop. The classification system of Tezuka et al. (2021) in Session 1 sends the result of the classification to an external device, making it usable for such experiments.
• In a closed loop, if the classification is wrong the manipulation is applied at the wrong time, so you must always confirm how accurate the classification is.

Figure　Observe / manipulate, and closed loop

Hands-on task (choose a tool, 3 minutes)

For each of the three questions below, choose the best-fitting tool and give a one-line reason.

1. You want to silence the activity of new neurons only when the mouse enters REM sleep.
2. You want to see whether the same new neurons are active in the same way on the day of learning and one week later.
3. You want to see how the membrane potential of a single neuron moves on the scale of a thousandth of a second.

Check questions

1. What makes GCaMP glow brightly?
2. Name two strengths of optogenetics.
3. Explain the difference between closed loop and open loop with an example other than an air conditioner.

Summary

• Observe = electricity (fast) and calcium light (same cells over many days).
• Manipulate = light (fast, even during behavior) and drug (broad but slow). Observing gives correlation; manipulating gives "necessity."
• Closed loop = measure a state and decide the manipulation from that result.

Session 7　How to read behavioral experiments — fear conditioning

Lab paper: Fujinaka A, Li R, Hayashi M, Kumar D, et al. Effect of context exposure after fear learning on memory generalization in mice. Molecular Brain. 2016;9:2. DOI: 10.1186/s13041-015-0184-0 (the full text can be read free of charge)

Aims of this session

1. Describe the design of fear conditioning (what is paired with what).
2. Read the strength of a memory and "discrimination" from the proportion of freezing.
3. State the control groups a behavioral experiment needs, and the meaning of blinding.

Explanation

Question: How do we turn "the mouse is afraid" into a number? Answer: We count the time it stops moving. When afraid, a mouse stops all movement except breathing.

• Key point 1　Pavlovian conditioning: A stimulus that is originally meaningless = the conditioned stimulus (CS). A stimulus that elicits a response on its own = the unconditioned stimulus (US). When the CS and US are presented as a pair, afterward the CS alone elicits a response (the conditioned response). In fear conditioning, the US is a very mild electric stimulus to the foot, and the CS is a place (context) or a tone.
• Key point 2　Freezing: The proportion (%) of the test period during which all movement except breathing has stopped is used as the strength of the fear memory. Nowadays it is often counted automatically from video. No electric stimulus is given on the test day.
• Key point 3　Context and cue: Memory tied to a place (contextual fear conditioning) is measured by returning the animal to the same box, and it requires the hippocampus and amygdala. Memory tied to a tone (cued fear conditioning) is measured by sounding the tone alone in a box of different shape and smell, and it centers on the amygdala. One of the memories in Session 4 that "did not weaken without the hippocampus" is this tone memory.
• Key point 4　Generalization and discrimination: A box where a frightening experience occurred, and a box that is similar but safe. If the animal is equally afraid in both, fear has spread (generalization). If it is afraid only in the frightening box, it is discriminating. Excessively strong generalization is studied as one of the features seen in anxiety-related disorders.
• Key point 5　What the reader should check: (1) Is there a control group (e.g. a group that does not receive the electric stimulus)? (2) Did the person scoring count without knowing which mouse was in which group (blinding)? (3) Were the order of boxes and the time of day matched across groups? (4) Is prior experience, such as habituation (placing the animal in the box in advance), described?
• In our lab's paper: Fujinaka et al. (2016) habituated mice to two similar boxes, then gave a frightening experience in one of them, and afterward varied when the mice were placed in the other (safe) box. When mice were placed in the safe box (box B, used for habituation) with no stimulus 3 or 6 hours after the experience, the next day they froze about equally in the two boxes (generalization). In the comparison group placed in the box 24 hours later, discrimination was preserved. Conditions such as how long the mice were kept in the box should be checked in the methods section of the paper. When the safe box was a novel box, or very different from the frightening box, or not habituated to in advance, generalization did not occur.
• You might think "events after learning have nothing to do with it." But this result shows that what is experienced in the few hours after learning can change the form of the memory.

Figure　The design of Fujinaka et al. (2016) (schematic)

Hands-on task (hand calculation, 3 minutes)

A 3-minute (180-second) test.

1. Mouse X froze for 54 seconds in box A and 18 seconds in box B. What is the proportion of freezing (%) for each?
2. Mouse Y froze for 81 seconds in box A and 72 seconds in box B. What percent each?
3. Suppose we use "A ÷ (A + B)" as a measure of discrimination (0.5 means no discrimination; closer to 1 means better discrimination). What are the values for X and Y? Which one is generalizing?

Check questions

1. What does the proportion of freezing measure?
2. How does the way of testing differ between contextual and cued fear conditioning?
3. What is the reason for counting under blinding?

Summary

• Pair the CS and US. Test with no US, counting freezing (%).
• Comparing two similar boxes lets you read generalization and discrimination.
• Check the control group, blinding, how conditions were matched, and prior experience in the methods section.

Session 8　How to read statistical figures

Lab paper: Vergara P, Kumar D, Srinivasan S, Koyanagi I, et al. Remapping of Adult-Born Neuron Activity during Fear Memory Consolidation in Mice. International Journal of Molecular Sciences. 2021;22(6):2874. DOI: 10.3390/ijms22062874 (the full text can be read free of charge)

Aims of this session

1. Say that the meaning of the "error bar" on a bar graph differs depending on whether it is the standard deviation (SD) or the standard error (SEM).
2. Form the habit of checking what the "n" in a figure counts (animals, cells, or nights).
3. State the difference between a p-value and effect size (the size of the difference).

Explanation

Question: Does a graph with short error bars mean the data have little scatter? Answer: Not necessarily. Depending on what the bar represents and what n is, the length changes even for the same data.

• Key point 1　SD and SEM:
• Standard deviation (SD): how scattered each individual value is around the mean. It is the scatter of the data itself.
• Standard error of the mean (SEM): how certain the value of the mean is. SEM = SD ÷ √n. It shrinks as n increases.
• 95% confidence interval: in textbook terms, an interval such that "if you repeatedly construct intervals in the same way, 95% of them contain the true mean." When n is large, it is roughly "mean ± 2 × SEM" (when n is small, you use a value from a table called the t distribution, and it becomes wider).
• Key point 2　What is n: When 100 cells are measured from one mouse, is n 100, or 1? Cells from the same mouse are alike (they share that mouse's quirks). Counting cells as independent samples makes it easy to get "a difference" even when there is really none. This is called pseudoreplication. In sleep research, counting one person's 5 nights as 5 people is the same error.
• Which to take as one sample: if what you want to compare is "the mice of group A and the mice of group B," then one sample = one mouse. Cells are the material from which that mouse's value (e.g. the mean over cells) is made; increasing cells makes one mouse's value more certain, but it is not a substitute for the number of mice.
• Key point 3　p-value and effect size: A p-value is "the probability that, if there were really no difference, a difference this large would arise by chance"; it is not the size of the difference. If you increase n, even a small difference gives a small p. So you also look at the size of the difference (effect size). Example: Cohen's d = (difference of the two group means) ÷ (scatter SD). By textbook convention, 0.2 is small, 0.5 is medium, 0.8 is large.
• Key point 4　When you run many statistical tests: If the probability of a single test giving "a difference by chance" is 5%, then running 20 tests with no difference gives a probability of at least one "difference present" of 1 − 0.95²⁰ ≈ 64%. When you make many comparisons, you make the criterion stricter accordingly (correction for multiple comparisons; e.g. the Bonferroni method uses 0.05 ÷ 20 = 0.0025).
• Key point 5　Look at the individual points: A bar graph alone cannot tell you whether all 5 mice were about the same, or whether just one was far out. In recent papers, the points for each mouse are often overlaid on top of the bar.

Figure　Even for the same 5 mice, the length of the bar changes

Hands-on task (hand calculation, 5 minutes)

The freezing (%) of 5 mice were 20, 30, 40, 50, 60.

1. What is the mean?
2. What is the SD? (procedure: square the difference of each value from the mean and sum → divide by (n − 1) → square root)
3. What is the SEM? (SD ÷ √5, √5 ≈ 2.24)

Only if you wish (a short piece of code, Python): We compare "two groups with no true difference" for the case where cells are counted as n and the case where mice are counted as n. The made-up data are created inside the code. Paste it into a free environment that runs in the browser and execute it.

The 1.972 and 2.306 in the code are values from the table of the t distribution. They are the line at which "when there is really no difference, a t (difference ÷ its uncertainty) larger than this arises by chance only 5% of the time"; the fewer the samples, the higher the line.

Confirm in the paper (2 minutes): Our lab's paper (Vergara et al. 2021) is a study that compared the activity of new neurons recorded with a miniaturized microscope across learning, consolidation (sleep and waking), and retrieval. Choose one figure legend and find and write down three things: (1) what n counts (cells, mice, or number of recordings), (2) what the error bar is, and (3) what statistical test it is.

Check questions

1. State the difference between SD and SEM in one sentence each.
2. What is likely to happen if you count 100 cells from one mouse as n = 100?
3. Can the p-value be very small while the effect size is very small?

Summary

• Always check in the legend whether the bar is SD or SEM, and what n is (animals, cells, nights).
• A p-value is not the size of the difference. Look at effect size and the confidence interval too.
• If you run many statistical tests, correct. Look at the points for each mouse too.

Session 9　Reading one of our lab's papers (1)

Lab paper: Kumar D, Koyanagi I, Carrier-Ruiz A, Vergara P, et al. Sparse Activity of Hippocampal Adult-Born Neurons during REM Sleep Is Necessary for Memory Consolidation. Neuron. 2020;107(3):552-565.e10. DOI: 10.1016/j.neuron.2020.05.008 (the abstract can be read by anyone. The full text can be read through a university library subscription or similar)

Aims of this session

1. Read a paper in the order "title → abstract → Figure 1 → the last figure → the first and last of the discussion."
2. Divide an abstract into five boxes (question, method, result, conclusion, limitation) and write them in your own words.
3. Think of one "objection to the conclusion" and find in the text how the authors answered it.

Explanation

Question: Is a paper read in order from page 1? Answer: Most researchers do not read it that way. They first grasp the overall shape from the abstract and figures, and descend into the text only where needed.

• Key point 1　The order of reading: (1) title (often a single sentence of what was found) (2) abstract (3) Figure 1 (the design of the study) (4) the last figure (the most important result) (5) the first and last paragraphs of the discussion. The methods are read by going back when you doubt a result.
• Key point 2　The five boxes: question (what they wanted to know), method (how they investigated it), result (what came out), conclusion (so what can be said), limitation (what cannot yet be said). An abstract usually contains the first four; you think of the fifth yourself.
• Key point 3　The background of this paper (the thread so far): REM sleep and memory (Session 3). The new neurons of the dentate gyrus are few and their activity is sparse, yet they are important for memory (Session 4). But what they do while the animal is asleep was unknown — that is the starting point.
• Key point 4　What they did (a summary of this file): In freely moving mice, the activity of young new neurons was recorded with calcium imaging (Session 6). When the mice underwent learning that ties a place to a frightening experience (Session 7), the ensemble of young new neurons that was active at that time was reactivated during subsequent REM sleep. At this time, the overall activity of new neurons during REM after learning was in fact reduced. The shape is that, against a backdrop of overall quieting, the cells from the time of learning were active again. Further, when this sparse activity was silenced with optogenetics only during the REM sleep of the period when the memory consolidates (the closed-loop idea of Session 6: discriminating REM and delivering light only during it), the way the spines (the receiving parts of synapses) on the dendrites of the new neurons remodeled changed, and memory consolidation was impaired.
• Key point 5　The conclusion: The authors state that they linked "the activity of new neurons during REM sleep" and "memory consolidation" by causation.

Figure　The design of this paper (schematic)

Hands-on task (an abstract into five boxes, 6 minutes)

1. Open the paper's page from the DOI and read the abstract (in English; you may ask the AI word by word about words you do not know).
2. Draw five boxes on paper and fill each with one line in your own words (do not copy the sentences of the abstract).
3. Show the five lines you wrote to the AI and ask where anything is missing.
4. Think of one objection. Example: "Could the light delivered during REM have merely disturbed sleep itself?" Try to find in the results section of the text how the authors answered this objection (if you cannot read the full text, it is fine just to think about "what control experiment would answer it").

Check questions

1. When is it good to read the methods section while reading a paper?
2. In this paper, when was activity silenced with light?
3. Why was a "silencing" experiment done, not just an "observing" one? (recall Sessions 5 and 6)

Summary

• The order of reading is title → abstract → Figure 1 → the last figure → discussion. Methods when you doubt.
• The five boxes (question, method, result, conclusion, limitation), in your own words.
• This paper: the sparse activity of new neurons during REM is needed for memory consolidation.

Session 10　Reading one of our lab's papers (2), and asking your own question

Lab paper: Srinivasan S, Koyanagi I, Vergara P, Wang Y, et al. Transient reactivation of small ensembles of adult-born neurons during REM sleep supports memory consolidation in mice. Nature Communications. 2025;16:7210. DOI: 10.1038/s41467-025-62554-8 (the full text can be read free of charge)

Aims of this session

1. Say that our lab's three papers are connected as "question → answer → next question."
2. State the two conclusions of this paper in your own words.
3. Write one question of your own in a form that can be checked by experiment.

Explanation

Question: Does research end with a single paper? Answer: It does not. The better a paper is, the more it ends by making the next question clear.

• Key point 1　The connection of the three papers:
• 2011 (Session 4): removing new neurons after learning degrades the already-formed memory → new neurons are a component of the memory.
• 2020 (Session 9): silencing the activity of new neurons during REM sleep impairs consolidation → when they work became visible.
• 2025 (this session): then, the activity of which cells, at which timing, is needed?
• Key point 2　What they did (a summary of this file): The abstract has two conclusions.
• (1) As few as about 3: they labeled only the new neurons that were active at the time of learning (Session 5), and afterward silenced only the labeled cells with light during REM sleep (Session 6). The abstract writes "a memory ensemble comprising as few as ~3 adult-born neurons." In the text, they counted, mouse by mouse, the labeled new neurons that were within the range the light could reach and silence, and report a mean of 2.4 (counted in 8 mice). This is not all the cells in the brain, but "the labeled cells within the range silenced by light." Try to check the relationship between the number in the abstract and the number in the text in the methods section. Silencing this small number during REM weakened the later fear memory. On the other hand, silencing the same cells during the retrieval test (while awake) did not weaken the memory. So it can be said that "reactivation during REM is needed."
• (2) A specific phase: a phase is "which part of one cycle of a repeating wave you are in" (the rising side, the peak, the falling side, the trough). The authors read the brain's electrical signal (LFP, Session 6) on the spot, divided one cycle of the theta wave (Session 3) into four (rising side, peak, falling side, trough), and with a closed-loop device that delivered light only at one fixed phase, silenced the activity of the new neurons. Only the group silenced at the ascending phase (the stretch rising from the trough toward the peak) had a weakened memory. The comparison was a "yoked control" that pairs two mice. At the instant light came on for the experimental mouse, the paired mouse also received light at the same time and the same number of times. In the paired mouse, the light fell at times unrelated to its own brain waves. In this way, while keeping the amount of light matched, one can separate the effect of "delivering light" itself from the effect of "matching the phase" (detailed conditions, such as which electrical signal was used as the reference, are in the methods section of the text).
• Key point 3　Why it is surprising: You might think "a memory needs thousands of cells." But this conclusion shows that the reactivation of a very small number of cells, and how their activity is aligned within just a part of the time inside a theta wave of a few cycles per second, can influence memory consolidation. This thread is also introduced on the "Research" page of the lab website.
• Key point 4　How to pose a question: A good research question can be written in the following form: "Is ＿＿ (which cells, which activity) necessary / sufficient for ＿＿ (what memory or behavior), during ＿＿ (when, during which state)?" Further, add (1) how to measure it, (2) how to manipulate it, (3) what the control is, (4) a prediction, (5) what you learn if the prediction is wrong. A question for which you can write up to (5) moves forward whichever result comes out.

Figure　The chain of questions

Hands-on task (write a question, 8 minutes)

1. Open the paper's page from the DOI and read the abstract. Write conclusions (1) and (2) each in one line in your own words (2 minutes).
2. In the following form, write one question of your own (6 minutes). It need not be on the lab's subject matter (example: a nap after club practice and how a skill is learned). - Question: "Is ＿＿ necessary for ＿＿, during ＿＿?" - how to measure / how to manipulate / what the control is / a prediction / what you learn if the prediction is wrong
3. Show what you wrote to the AI and ask only "where is the single weakest point of this question?"

Check questions

1. What did each of the three papers of 2011, 2020, and 2025 reveal? One line each.
2. State the two conclusions of this session's paper.
3. What is good about adding "what you learn if the prediction is wrong" to a question?

Summary

• Research is connected as question → answer → next question.
• The 2025 paper: reactivation during REM of an ensemble of about 3 new neurons, and alignment to the ascending phase of the theta wave.
• Your own question is "what is necessary, when, for what" + how to measure, how to manipulate, the control, a prediction, and what if it is wrong.

Benchmarks for completion, and what to learn next

Benchmarks for completion (you have completed the course if you can do these three)

• State the conclusion of each of the 10 papers in one line in your own words.
• Divide the abstract of a neuroscience paper you are reading for the first time into the five boxes (question, method, result, conclusion, limitation).
• Look at a paper's figure and find from the legend what n is and what the error bar is.

What to learn next

• The list of our lab's papers is on the "Publications" page of the lab website. You can go to the abstracts from the DOIs. First, try reading just the abstracts of the papers around the thread of Sessions 4, 9, and 10.
• The overall picture of the lab's research (including the human sleep research) is published on the "Research" page.
• If you want to learn more statistics, use the statistics course on the same "For undergraduates" page (if it is published).
• If you want to confirm things in a textbook, the sleep, learning-and-memory, and hippocampus chapters of a standard neuroscience textbook (e.g. Principles of Neural Science (Kandel)) correspond.
• If you want to actually study in the lab, please see the "For undergraduates" page.

Request for feedback (repeated)

This course is revised every time based on the voices of its users. If you used it, please be sure to send feedback.

• Send to: the address listed on the Contact page of the lab website
• Subject: "Course feedback: Basics for Reading Our Lab's Papers"
• What we would like you to write (4 points): (1) how many sessions you completed (2) what was hard to understand (3) what was good (4) how long one session took

If you ask your conversational AI to "help me polish my feedback text," it will put it into a form you can send as is.

Division of Neurophysiology, Kobe University Graduate School of Medicine (October 2026)

The courses are free for personal study and teaching. Please contact us before distributing modified versions. They are not medical advice.