Resources  /  Student Research Starter Kit

Your first real research project.

Not a homework assignment. Not a rehearsed demo. An actual investigation, run by you, of something no one has quite answered before. Here is how to start.

Step 01Start with a question that bugs you.

Every good project begins with something you noticed, and quietly wondered about. The trick is to catch that wonder before it fades, and to sharpen it into a question a real experiment can actually answer.

You do not need a laboratory. You do not need a professor. You need a habit of noticing. Something at your bus stop that seems wrong. A pattern in your neighbourhood no one mentions. A claim on a shampoo bottle that sounds suspicious. That is where a project starts.

The mistake most students make is choosing a topic instead of a question. "Water pollution" is a topic. It is not testable. Compare:

A topic vs. a testable question

Weak — a topicWater pollution in the Yamuna.

Better — a questionIs there more plastic in the Yamuna near industrial discharge points than 2 km downstream?

Strong — a testable questionDoes the visible microplastic count per 500 mL of Yamuna water differ significantly between three sites within 100 m of the Sector 63 discharge outlet and three sites 2 km downstream?

The strong version tells you exactly what to measure, where, and how much. If someone asks "how would you test that?", you already know. That is the bar.

A quick check

Say your question out loud. If a friend can immediately picture the experiment you would run — the samples, the tool, the number — it is a research question. If they say "cool, but what would you actually do?", it is still a topic.

Step 02Turn curiosity into a hypothesis.

A hypothesis is a guess, but a specific one. It commits, in advance, to what you expect to find. Then the experiment either supports it or does not. Both outcomes are useful. Only one is embarrassing: not having predicted anything at all.

The simplest form works: If I do X, then Y will happen, because of Z. The "because" matters. Without it, you have a prediction. With it, you have science.

Building a real hypothesis

The questionDoes the microplastic count differ between sites near and far from the discharge point?

The hypothesisIf industrial discharge adds microplastic to river water, then samples taken within 100 m of the outlet will contain at least twice as many visible microplastic fragments per 500 mL as samples taken 2 km downstream, because currents and dilution should reduce concentration over distance.

Why it is strongIt is quantitative ("at least twice"). It names the mechanism ("dilution"). It is falsifiable — the data can prove it wrong.

Step 03Design a fair test.

A fair test changes one thing, measures one thing, and keeps everything else the same. That sounds obvious. It is not.

Independent What you change

The single thing you deliberately vary. Distance from the outlet, in this project.

Dependent What you measure

The thing you think will respond. Fragment count per 500 mL.

Controlled What you hold constant

Volume filtered, mesh size, weather, time of day, counting method. Every "not-this-one" variable.

You also need repeats, and a control. Repeats mean testing the same condition more than once so a single fluke does not fool you (three samples per site is a bare minimum, five is better). A control is the version with no treatment applied — distilled water run through the same filter, in this case — so you know your method itself is not the source of your finding.

The single most common student mistake

Changing two things at once. If you sample near the outlet at 4 pm and downstream at 8 am, you have not measured distance. You have measured distance and time-of-day mixed together. When something looks surprising in your data, ask first: did anything else change that I did not mean to change?

Step 04Write everything down.

The single most important physical object in your project is a bound notebook. It is where the research actually lives. Everything else — the poster, the abstract, the interview — is a summary of what is in this book.

You need a real, page-numbered notebook. Every session gets a date, a time, what you set out to do, what actually happened, the raw data, any deviations from your plan, and one sentence on what you would change tomorrow. In ink. Mistakes crossed out with a single line, initialled and dated. Never rewritten "cleanly" the next day.

Judges at IRIS and ISEF read logbooks. They can tell within thirty seconds whether it was kept day-by-day or reconstructed the week before the fair. A reconstructed logbook is disqualifying, and it is obvious.

We have a full worked example, a printable blank page, and the six rules that separate a real research record from a nice-looking one:

Open the Research Logbook Template →

Step 05Handle ethics before you handle data.

Any project involving human participants, live animals, potentially hazardous chemicals, or fieldwork on regulated land requires ethics approval before you begin. Not after. Data collected without prior approval cannot be used at IRIS, ISEF, or any credible fair.

Which forms your project needs depends on what you are doing. A survey of classmates is not the same as a study on lab mice. A soil experiment in your garden is not the same as one on protected land. There is no honour in guessing.

Our forms guide walks you through the ISEF forms with a decision matrix and links to the official 2027 versions:

Which ISEF forms do I need? →

A rule that will save you

Your Research Plan and Form 1A must be signed before any experimentation begins. If in doubt, ask. A one-week delay to get ethics right beats a disqualification the day before the fair.

Step 06Tell the story tightly.

A judging round is short. Some rounds give you five minutes, some seven. Your poster is a supporting visual, not a script. What matters is that in those five minutes, a scientist you have never met walks away with a clear sense of what you asked, what you did, what you found, and what you would do next.

The structure that always works, in order:

  1. The question, in one sentence. "I wanted to find out whether industrial discharge measurably raises microplastic concentration in the Yamuna near Sector 63."
  2. Why it matters, in one sentence. Not global. Local, specific, honest. "Local fisherfolk drink from this stretch and there is no published measurement of it."
  3. What you did, in two sentences. The method, the sample size, the controls. Do not over-explain.
  4. What you found, with the number. "Sites within 100 m of the outlet had 2.3× the fragment count of sites 2 km downstream, across 15 samples per site (p = 0.008)."
  5. What surprised you. Every honest project has one thing that did not go as expected. Judges love this. It is where curiosity is visible.
  6. What you would do next. One sentence. What is the next experiment, if you had another six months?

Practice this out loud, in five minutes flat, ten times before the fair. Not written out. Spoken. To your grandmother if that is who is around. If she understands it, a judge will.

For a deeper look at abstract structure, read four annotated ISEF-style abstracts:

Sample Abstracts (Annotated) →

Step 07Answer these six questions before the fair.

These are the questions judges ask most often. If you can answer each of them in under thirty seconds, you are ready. If any of them make you hesitate, that is where the next week of your prep should go.

What was the one thing you changed?
Your independent variable, in plain language. If you cannot say this in one sentence, your experiment probably was not fair.
What did you compare your results against?
Your control condition, or your baseline. "Compared to what?" is the single most powerful judging question.
How do you know your finding is not just chance?
Your repeats, your sample size, and if you have run one, your significance test. Even "I saw the same effect across all five samples" is a real answer.
What went wrong?
Something did. Tell them. Real projects have real setbacks and judges know it. Hiding them looks worse than owning them.
What would you do differently next time?
One concrete change. Not "work harder". Something specific: a larger sample, a different measurement tool, a different site.
What is this useful for?
Who benefits from knowing what you found? Be honest. "It helps my local municipality decide where to test" is a good answer. "It could change the world" is not.

Ten rules, distilled.

If you read nothing else on this page, read these:

Pick a question you actually care about.

Ten months is a long time to be curious about something you are not.

Make it testable.

If a friend cannot picture your experiment from your question alone, sharpen it.

Change one thing at a time.

Two changes at once means you have measured neither.

Keep a real, bound logbook.

Same day. In ink. Mistakes crossed out, not erased.

Get ethics signed before you start.

Data collected without approval cannot be used. This is not optional.

Repeat everything at least three times.

One measurement is a story. Three is a finding.

Report the number you found, not the number you hoped for.

A negative result is still a result. A dishonest result is fraud.

Practise the five-minute talk out loud.

Ten times, minimum. Not written. Spoken.

Ask your mentor early, not late.

They cannot help with a plan you never showed them.

Enter something.

An unentered project teaches you less than an imperfect one that got judged.

What to do next.

The single best next step is to enter a fair. A deadline focuses everything. Here is what to open now:

The best time to start was last week.

The second best time is now. Open your notebook, write today's date, and write one question you have been quietly wondering about. That is Day 1.