How to Actually Do Research as a Master's Student and Early PhD

Research is not reading more. It's not collecting papers. It's not understanding everything perfectly before you start. Research is the process of identifying something the world doesn't know yet, and then finding it out.

That sounds obvious. But most students spend their first year doing everything except that. They read obsessively, take notes in Notion, attend seminars, and never produce a single original result.

The distinction between a Master's and a PhD matters here. At the Master's level, you're usually applying existing methods to a new domain or refining something that already works. That's fine. At the PhD level, the bar shifts: you need to produce something that nobody knew before. New method, new insight, new framework.

The output mindset

Train yourself to think in outputs. Every week, ask yourself: what did I produce? Not read, not think about, not plan. Produce. Code, results, written paragraphs, a sketch of an argument. If your week left no trace, it was a weak week.

You will never read everything in your field. Accept this now. The goal is to read strategically, not comprehensively.

The three-pass method

Where to find papers

Start from a survey paper in your area. Use its references to find the foundational work. Then use Google Scholar's "cited by" to find what came after. Follow the trail forward and backward.

Take useful notes

Don't summarize what a paper says. Write what it means for your work. After each paper, answer three questions: What problem does it solve? What does it not solve? How does it change what I'm doing?

This is the hardest part. Finding a good problem takes longer than solving it. Most early researchers pick problems that are either too broad, too solved, or too boring to anyone but themselves.

What makes a problem good?

How to find your problem

Read the limitations section of every paper in your area. That's where researchers tell you what they couldn't do. That's your list of open problems. Read it seriously.

Talk to people. PhD students two years ahead of you know which problems are promising and which ones are graveyards. Ask them. Buy them coffee. This is the fastest research shortcut that exists.

Look for the assumption everyone makes that nobody has tested. Every field has sacred cows. Those are often research problems hiding in plain sight.

Narrowing your scope

Your instinct will be to tackle something ambitious and broad. Resist it. Narrow the problem until it feels almost too small. Then narrow it again. A tight, well-solved small problem is worth more than a sprawling unsolved big one.

In empirical research, your experiments are your arguments. A sloppy experiment is a bad argument, regardless of what it shows.

Before you run anything

Write down what you expect to happen before you run the experiment. This forces you to think clearly and protects you from unconsciously interpreting results to match what you hoped. Keep these predictions in a lab notebook, dated.

Baselines and ablations

Your method needs to beat something. "It works" is not a result. "It works better than X, Y, and Z on this benchmark" is a result. Set up strong baselines early, not as an afterthought.

Ablations tell you why your method works. If you remove component A and performance drops, A matters. If it doesn't drop, A doesn't matter. Run ablations on every design choice you claim is important. Reviewers will ask. Have the answers ready.

Statistical rigor

Run experiments multiple times with different seeds. Report mean and standard deviation. One run is not a result in most fields. Know whether you need significance tests and use them correctly.

Don't cherry-pick your best run. Report typical performance. If your method is fragile to hyperparameters, that's information the reader needs.

When results don't cooperate

Negative results happen constantly. They're not failures; they're information. Before you abandon an approach, understand why it failed. That understanding is often more publishable and more useful than a positive result you don't understand.

Research that isn't written doesn't exist. Writing is not what you do after you finish the work. Writing is how you finish the work. It forces clarity that thinking alone never produces.

Structure of a research paper

The writing process

Write ugly drafts first. Your first draft's only job is to exist. You edit drafts, not blank pages. Set a daily word goal small enough that you'll always hit it: 300 words, 500 words. Do it every day.

Separate writing from editing. Write a section, then leave it alone for a day. Come back and cut ruthlessly. The best writing in research is concise. Say it once, clearly.

Getting your paper rejected

You will get rejected. Most papers do, including good ones. Read every reviewer comment even when it stings. Reviewers who are wrong are still showing you where your writing failed to communicate clearly. Fix the writing.

Your advisor relationship determines much of your graduate experience. You don't get to choose this perfectly, but you do get to manage it.

What advisors want

Advisors want students who make progress, communicate clearly, and don't create problems. They're busy. They're juggling grants, papers, teaching, and multiple students. Your job is to make your work easy for them to engage with.

Come to every meeting with a written update. Two or three bullet points: what you did, what you found, what you need. Don't make them drag information out of you.

When you disagree

You will disagree with your advisor. On approach, on scope, on what result means what. That's normal. State your view clearly, explain your reasoning, and be willing to be wrong. But don't just defer automatically. Part of your training is learning to defend your scientific positions.

If the relationship is broken

Sometimes advisor-student relationships fail. If yours is heading there, document everything, talk to the graduate director, and understand your program's policies. Changing advisors is hard and takes time, but staying in a bad situation indefinitely is worse.

Graduate school is a long project. The people who finish are not always the smartest; they're the ones who didn't break down or give up. Managing yourself is as important as managing your research.

Time structure

Research time is not like coursework time. There are no deadlines until there are enormous deadlines. You have to create your own structure. Block specific times for deep work and protect them. Treat them like classes you can't skip.

Track what you actually do, not what you plan to do. One week of honest tracking will tell you where your time goes. Most people are shocked.

The PhD slump

Almost every PhD student goes through a period where nothing works, they feel like a fraud, and they can't see the point. This is normal. It's not a sign you should quit; it's a sign you're in the hard middle.

Keep a record of what you've done. On bad days, read it. You have done more than you feel like you have.

Burnout is real

Working 80-hour weeks is not a badge of honor. It's a path to burnout that will cost you months of productivity and potentially your degree. Work intensely during work hours. Shut it off. Sleep. Exercise. See people outside academia.

By the end of your PhD, you need to be known for something. Not famous, but identifiable. When someone in your field hears your name, they should be able to say what you work on.

Present your work

Give talks whenever you can. Lab meetings, seminars, conferences. Presenting forces you to understand your own work better than writing does. It also puts your name in front of people who might hire you, collaborate with you, or cite you.

Be findable online

Have a simple webpage with your name, your papers, and your contact information. Link your Google Scholar profile. Put your thesis on arXiv if your field does that. Make it easy for the right people to find you.

Peer review

Review papers when you're asked. It's how the field works. It's also how you learn to evaluate research critically and understand what good work looks like from the other side of the table.

Build relationships, not networking

Networking is transactional and people can smell it. Be genuinely curious about other people's work. Ask real questions at seminars. Follow up on interesting conversations. The relationships that matter most in your career will come from genuine intellectual connection, not business card exchanges.