How to Read Scientific Papers?

Reading Scientific Papers¶

Neuroscience is a pretty vast field.

People will often tell you that, if you want to become a scientist, you should read scientific papers.

But what does that actually mean?

Which papers should you read? Why should you read them? And what should your goal be when you are reading a paper?

Similar to our advice on exploring your research interests, I think the easiest way to start is simply to start with something.

Maybe there is a cognitive process that interests you, such as reward, learning, or memory. Maybe you are interested in a particular disorder, such as addiction. Brainstorming with generative AI tools such as ChatGPT is completely fine—we discussed some ways of doing this in the previous section.

However, once you have narrowed down your interests a little, it is important that you start reading.

And, yes, there will probably be some memorization involved too. Having some important concepts and terminology in your head will actually make your life much easier.

Let me explain.

Let’s Say You Are Interested in Memory¶

Suppose you decide that you are interested in memory.

Memory is, of course, an enormous topic.

If you prompt a generative AI tool such as ChatGPT and ask it to explain the neuroscience of memory, you might suddenly receive a 500–1,500-word answer explaining how memory is a multi-scale problem. Before you know it, you are encountering words such as receptors, ion channels, kinases, proteomics, synaptic strength, engrams, “sparse” neuronal ensembles, neuronal excitability, hippocampus, cortex, encoding, consolidation, retrieval, working memory, schemas, recall, recognition, navigation, conditioning, skill learning, MVPA, fMRI, EEG, and so on.

You might not know what most of these terms mean.

That’s okay!

The question is: how do we begin organizing all of this information?

Here is one way I would organize my thoughts and readings around memory. The same general approach can be applied to almost any other cognitive process—attention, consciousness, reward, vision, and so on.

First, Learn the Basic Language¶

Before diving deeply into the neuroscience of memory, you probably need to understand what researchers actually mean by memory.

How do we define it? More importantly, how do researchers operationalize it?

To operationalize something means turning an abstract concept into something that can actually be observed or measured. “Memory” is a very broad concept. In an experiment, however, a researcher needs to decide what exactly counts as memory and how they are going to measure it.

At this stage, what you need is some basic working knowledge.

And honestly, introductory psychology textbooks can be very good for this.

For example, there is a freely available OpenStax textbook, Psychology 2e:

https://openstax.org/details/books/psychology-2e

You can download the textbook for free and look at Chapter 8: Memory.

This chapter introduces the three basic functions of memory—encoding, storage, and retrieval—as well as different models of memory storage, different forms of long-term memory such as episodic memory, relevant brain areas, different types of amnesia, and other foundational concepts.

Why do I recommend starting with a textbook?

Because sometimes, even with generative AI, it is difficult to know what you need to know.

You don’t know what you don’t know.

Simply reading one good introductory chapter gives you a basic framework. Now, when you encounter terms such as episodic memory, encoding, or retrieval in a scientific paper, you already have some idea of what the authors are talking about.

And at this level, I actually recommend trying to memorize some of these basic terms. You will encounter them again and again. Having them in your head makes everything you read afterward much easier.

Then, Pick Something That Catches Your Interest¶

Once you understand roughly what memory means, you might find that one particular aspect of memory interests you.

This part can involve a bit of trial and error.

Sometimes you don’t know what you are interested in precisely because you haven’t read enough yet to know what is out there.

Let’s say that, from your introductory reading, you become interested in memory retrieval.

The OpenStax textbook defines retrieval as the “act of getting information out of long-term memory storage and back into conscious awareness.”

At this point, rather than immediately looking for one individual experiment, I would first try to find a review article about memory retrieval.

Why?

Because when you are new to a topic, you usually do not yet know which experiments are important, which ideas are widely accepted, which terms you need to understand, or even what the major questions in the field are. A good review can give you this broader picture by bringing together findings from many different studies.

A primary research article usually reports a specific study that the authors carried out themselves. A review article, by contrast, looks across many existing studies and brings them together to explain what researchers currently know about a particular topic.

You can think of it roughly like this:

Primary paper: “Here is the experiment we did and what we found.”

Review paper: “Here is what the field has learned from many experiments, and how these findings fit together.”

So, when you are first entering an unfamiliar research area, a good review can act as a kind of map of the field. It helps you learn the language, identify important ideas and experiments, and decide what you want to explore more deeply.

So perhaps you go to Google and type:

memory retrieval review

You will probably encounter a mixture of papers. Some might not be open access, meaning that you need access through a university or institution—or you may be asked to pay—to read the full article.

But keep looking.

As of August 11, 2026, the fourth result in this particular search is an article called “The neurobiological foundation of memory retrieval” by Frankland and colleagues, published in Nature Neuroscience in 2019:

https://pmc.ncbi.nlm.nih.gov/articles/PMC6903648/

The title already gives us a clue that this might be useful. The neurobiological foundation of memory retrieval sounds relatively broad. It does not sound like a paper describing one very specific experiment. It sounds like the authors are going to explain and synthesize an area of research.

And that is exactly what a review article is useful for.

When you are new to an area, good expert reviews can be extremely useful because someone has already done part of the work of organizing the field for you.

In neuroscience, you will often find influential reviews in journals such as Nature Reviews Neuroscience, Nature Neuroscience, Trends in Cognitive Sciences, Trends in Neurosciences, Annual Review of Neuroscience, and other review journals and review series. Depending on your area, you might also encounter reviews in more specialized journals—for example, Neuropsychopharmacology or major psychiatry journals if you are interested in clinical or psychiatric neuroscience.

Journal names and metrics are certainly not everything, but when you are completely new to a field, they can provide some useful clues about where to begin.

Most importantly, however, look at what the paper is actually trying to do. A broad title, an article described as a review, and a paper that brings together many studies are all good signs that you may have found a useful place to start.

Follow the Story of the Field¶

So, let’s start reading the Frankland et al. review.

The authors begin by describing seminal work by Tulving and Pearlstone in 1966 and the idea that a failure to remember something does not necessarily mean that the memory itself has disappeared. Instead, memory failure could also reflect a problem with retrieval.

As you continue, the authors introduce another term: ecphory.

Then we encounter the concept of an engram, and we begin learning how these ideas relate to memory retrieval.

Already, just from carefully reading the introduction, and doing some searching whenever necessary, you have started learning the basic concepts of memory retrieval and engrams. You have also begun to see, at least roughly, how thinking in the field developed over time.

So far, so good.

Then you reach the first section called “Manipulating retrieval”.

Or perhaps you look at “Box 1 | Approaches for tagging and manipulating engrams in rodents.”

And suddenly you think:

“Oh shoot... there are so many biological terms I don’t know.”

For example, the authors describe experiments asking “whether it is possible to prevent ecphory in the presence of external sensory retrieval cues.”

Okay.

Then they explain that researchers used a “tetracycline-based system (TetTag) to label a contextual fear memory engram,” such that “CA1 neuronal ensembles that were active during conditioning expressed an inhibitory opsin (ArchT).”

...Huh?

Maybe you don’t know what TetTag is.

You read the box above, and perhaps that doesn’t really help either.

Maybe you don’t even know what contextual fear memory means. You look at Figure 2 and see terms such as CA1 and dentate gyrus.

What are those?

This is where, for me, the fun part begins.

Branch Out When You Don’t Understand Something¶

Let’s say you are very interested in behavioral neuroscience and memory paradigms.

When I was studying neuroscience, I would often take a paper like this and start branching out from it.

I might think:

Okay, what are these brain regions?

Then I would read a little about the anatomy of the hippocampus.

That might lead me to CA1, CA3, and the dentate gyrus (DG). And sometimes I might get completely lost in the CA1/CA3/DG literature because I suddenly find that more interesting.

That’s fine.

Or I might return to the experiment and search:

tetracycline-based system TetTag labeling neuron

or:

optogenetic inhibition

Why optogenetic inhibition?

Because the next part of the review says:

“Critically, optogenetic inhibition of the ArchT-tagged neuronal ensemble during this test session reduced conditioned freezing levels (indicating impairment in memory retrieval).”

If I find this type of experiment fascinating and think that I might want to perform studies like this one day, then I might decide that I want to understand the genetics and molecular biology behind it.

How are researchers able to label the neurons that become active during a particular task?

How can they see those labels?

How can they later control those specific neurons?

I might branch out again and learn about those techniques.

Then I might come back to the original review and try to understand the behavioral experiment illustrated in Figure 2, including the difference between the questions that Figure 2a and Figure 2b can answer, which the authors explain below the figure.

Notice what is happening here.

I’m not reading the paper from the first word to the last word without stopping.

I’m reading. I encounter something I don’t understand. I leave. I learn something else. I come back. I understand a little more. Something else catches my attention. I leave again.

Scientific reading can be very non-linear.

Sometimes, Go Back to the Original Study (i.e., the Primary Article)¶

At some point, I might become curious about the original study that conducted these experiments cited by Frankland and colleagues’ review.

I might wonder:

How did the researchers who actually performed this experiment describe what they did?

So I follow the citation.

For example, the Figure 2 caption in the Frankland review says:

“In this experiment [19], neuronal ensembles in the CA1 region of the hippocampus were tagged with the inhibitory opsin, ArchT, during contextual fear conditioning (left).”

So, what is citation 19?

If you look it up, it takes you to Tanaka et al. (2014), published in Neuron:

https://pubmed.ncbi.nlm.nih.gov/25308331/

Now you have found one of the primary/original research papers behind the review.

Depending on your interests and your current level of knowledge, you may still want to finish—or at least understand much more of—the review first.

As you can see, the process is quite non-linear. But it still involves a careful reading of at least one key review, combined with what I would call curiosity reading: branching out whenever you need more context or whenever something catches your interest.

This is also why choosing a good review can matter so much. A good review can become your starting map of an unfamiliar field.

And if you are unsure where to start, you can always approach our community on Discord:

https://tinyurl.com/discordclematis

Our co-chairs and community will be happy to answer questions and help point you toward useful papers and resources.

Eventually, Read a Primary Paper [Very, Very Carefully!]¶

After building up some knowledge, you may finally feel brave and confident enough to tackle a primary research paper.

Or perhaps you still feel completely intimidated.

It doesn’t actually matter. At some point, you should try anyway.

Ideally, find a primary paper that helped define the field or had an important influence on the research that followed.

In neuroscience, you will often encounter influential primary research in journals such as Neuron and Nature Neuroscience, and sometimes Nature, Cell, or Science, as well as journals such as PNAS, Cell Reports, and Current Biology, among many others.

Again, the journal and its metrics are not everything.

Another very useful clue is whether a paper is repeatedly cited and discussed in the expert reviews you have been reading. Tanaka et al., for example, did not appear out of nowhere. We found it because the Frankland review pointed us toward it.

Now comes something that I think is important, but VERY HARD.

At least once, try to spend many hours reading one influential primary paper in very careful detail.

And I really mean many hours.

There have been papers that I have personally spent 40–60 hours trying to understand properly.

Try to understand every figure. Ask yourself what the authors are doing, how they are doing it, and why they are doing it. What question does each experiment answer? Why did they need that control? How does one figure lead to the next?

Eventually, you want to reach the point where someone can show you one of the figures and you can explain what was done and what it means.

Even better, try presenting the paper as though you had done the work yourself.

If you are interested in systems neuroscience, and particularly in engram research, the Tanaka paper we just found would be a very good example to work toward.

Another Example: Drosophila, Memory, and Connectomics¶

Let’s take a very different example.

If you are interested in Drosophila, the mushroom body, learning and memory, and connectomics, you might eventually encounter this primary research article:

https://pmc.ncbi.nlm.nih.gov/articles/PMC5806122/

If you are completely new to this area, reading this paper carefully will be very hard.

But instead of giving up, you can ask:

What concepts do I need to understand before I can understand this paper?

Perhaps you need to learn about the fruit fly brain first.

How is neuroscience typically done in Drosophila? And actually, why do neuroscientists use Drosophila in the first place?

Perhaps you need a review on learning and memory in Drosophila. Then you realize that you need to understand Kenyon cells and the mushroom body.

Because this is also a connectomics paper, perhaps you need to read something introducing synapse-level connectomics.

So you leave the paper.

You read.

You learn.

Then you come back.

Hopefully, you are now in a much better position to understand it.

It will probably still be difficult. You may still need to push very, very hard to get through the paper.

That’s normal.

What If You Realize You Don’t Care About This Approach?¶

Let’s return to our memory example.

Suppose you started reading Frankland et al. (2019) on the neurobiology of memory retrieval. You learned about Tulving’s work in 1966. You learned about engrams. You started learning about CA1, TetTag, optogenetics, and contextual fear conditioning.

But then you realize:

“Actually... I don’t really care about rodent work.”

That’s useful information too!

Maybe you are still fascinated by the idea of an engram. Maybe you are interested in forgetting and retrieving memories, but you want to understand these processes in humans.

And now that you know CA1 is part of the hippocampus, you start wondering:

What about the hippocampus and memory retrieval in humans?

So you search again:

hippocampus and memory retrieval in humans review

When I search this, one of the first results is “A closer look at the hippocampus and memory” (Voss et al., 2018), published in Trends in Cognitive Sciences: https://pmc.ncbi.nlm.nih.gov/articles/PMC5659202/

It has more than 300 citations. Again, citation counts are not a perfect measure of quality or importance. But when you are completely new to an area and trying to decide where to start, they can be one useful clue.

And Trends in Cognitive Sciences is one of the review journals I mentioned earlier.

So: this looks like an expert review on exactly the topic I’m trying to understand.

Let’s read it.

And Then the Process Starts Again¶

The authors (Voss et al., 2018) introduces episodic memory, a concept we already encountered in our OpenStax textbook.

But now the discussion becomes much more sophisticated.

The authors emphasize that episodic memory requires us to bind together different stimuli and their spatial, temporal, and conceptual relationships to form coherent memory representations.

In simpler terms, a coherent memory is not necessarily one isolated piece of information. We need to bind different things together: what happened, where it happened, when it happened, how different pieces of information relate to one another, and so on.

Then the authors start discussing how the hippocampus may be critical for the “bi-directional interaction of memory and exploration processes that are iteratively engaged over the course of learning in order to build episodic memories.”

And perhaps you read that sentence and think:

“...What?”

That’s fine.

Sit with it.

Break it apart.

Ask ChatGPT to unpack the sentence for you. Look up the concepts you don’t understand. Then go back and read it again.

As you continue through the review, you may start encountering fMRI paradigms and discussions of what researchers can and cannot infer from fMRI studies.

Then you realize:

“Wait... I don’t actually understand how fMRI works.”

So now perhaps you need an expert review or a textbook chapter introducing fMRI for cognitive neuroscience.

And off you go again.

You learn about fMRI.

Then you come back.

And so on, and so on.

It is an iterative process.

Hopefully, after doing this for a while, you eventually encounter a difficult and influential primary research paper on memory and fMRI that genuinely interests you.

Then you do the same thing we discussed earlier: spend many hours critically reading that one paper. Understand the experimental paradigm. Understand the methods. Work through every figure. Ask what each analysis is testing and why the authors performed it.

Once you invest that time, trust me: the steepest part of the learning curve starts to pass.

The next paper will contain concepts you recognize.

Then the next paper will use a method you have already seen.

Then another paper will mention a theory you already understand.

Slowly, you stop feeling as though every sentence contains five completely new ideas.

So, How Should You Read Scientific Papers?¶

If I had to summarize the approach above, I would suggest something like this:

1. Start with a broad topic that genuinely interests you. Memory, attention, reward, addiction, consciousness, vision ... anything is fine.

2. Learn the basic vocabulary first. A good introductory textbook can be incredibly useful. Learn the major concepts and try to remember the terms that appear repeatedly.

3. Narrow your interest a little. Maybe “memory” becomes “memory retrieval”, or “addiction” becomes “maladaptive reward learning”.

4. Find a good expert review. Use the review to understand the major concepts, important experiments, history of the field, and questions researchers are currently asking.

5. Read slowly and branch out whenever you need to. If you don’t understand a brain region, technique, behavioral paradigm, mathematical concept, or piece of biology, go and learn about it. Then come back.

6. Follow citations when an experiment interests you. Reviews can lead you directly to important primary papers.

7. Let your interests change as you read. You might begin with rodent memory and discover that you actually care about human fMRI. You might begin with behavior and discover that you love molecular biology. That is part of the process.

8. Eventually, choose one influential primary paper and go very very deep. Spend the time required to understand what the researchers did, why they did it, what every figure means, and what conclusions the experiments actually support.

9. Try to explain the paper yourself. If you can look at the figures and explain the study as though you had performed the experiments yourself, you have moved far beyond simply “reading” the paper.

And then?

Find another paper.

The process starts again! But this time, you know a little more than you did before.