How to Write a Falsifiable Hypothesis for ISEF (With Examples)
June 2, 2026 · 7 min read
If you only get one thing right in a science-fair project, make it the hypothesis. ISEF scores 100 points across five criteria, and the two largest — Scientific Thought (30) and Creative Ability (30) — both hinge on the quality of the question you set out to answer. A sharp, falsifiable hypothesis makes the rest of the project almost write itself. A vague one undermines everything downstream, no matter how clean your data is.
What "falsifiable" actually means
A hypothesis is falsifiable if there is a possible experimental result that would prove it wrong. This is the line, drawn by the philosopher Karl Popper, between science and everything else. If no outcome could contradict your claim, you cannot test it — and judges will notice immediately.
"Plants need sunlight to be healthy" is not a useful hypothesis — it is a known fact, and it is hard to imagine the result that would refute it. "Radish seedlings grown under blue light will produce more chlorophyll per gram than those under red light" is falsifiable: run the experiment, and the chlorophyll measurement either supports it or it does not.
The anatomy of a strong hypothesis
A competition-grade hypothesis usually contains four things. Miss one and it gets weaker.
- An independent variable — the one thing you deliberately change.
- A dependent variable — the specific, measurable thing you predict will respond.
- A predicted direction — not just "it will affect," but "it will increase / decrease."
- Ideally a mechanism — a reason you expect that result, which is what lifts a guess into scientific thought.
Weak versus strong: three examples
The difference is almost always specificity plus a mechanism. Compare:
Weak: Caffeine affects heart rate.
Strong: Daphnia exposed to a 0.1% caffeine solution will show a higher mean heart rate than those in spring water, because caffeine acts as a stimulant on the organism's nervous system.
Why it works: Names the organism, the dose, the measurable outcome, the direction, and a mechanism. It is testable in an afternoon and impossible to misread.
Weak: AI can detect cancer.
Strong: A convolutional neural network trained with density-stratified mammograms will achieve higher sensitivity on dense (BI-RADS 3–4) breast tissue than a standard model, because density-aware training corrects for the bias that hides tumors in dense tissue.
Why it works: Turns an impossible wish into a single, defensible claim with a clear baseline to beat and a reason it might be true.
Weak: Salt is bad for plants.
Strong: Radish seedlings watered with a 50 mM NaCl solution will have lower mean dry biomass after 14 days than seedlings watered with deionized water, because osmotic stress reduces water uptake.
Why it works: Specific species, concentration, timeframe, measurable outcome, direction, and mechanism — everything a judge needs to evaluate the design.
A template you can reuse
When you are stuck, fill in this sentence: "We predict that [changing the independent variable] will [increase / decrease] [the dependent variable], measured by [your metric], because [mechanism]." If you cannot complete it, you have found exactly what your project is still missing.
Common ways a hypothesis fails
- Not falsifiable: no result could prove it wrong.
- Too vague: "affects" instead of a predicted direction and a specific metric.
- Predicts everything: so flexible that any outcome confirms it.
- No mechanism: a guess with no reasoning behind it.
- Not measurable: you cannot actually quantify the dependent variable with your tools.
- Already known: a textbook result re-run — it clears no novelty bar.
From hypothesis to category and methods
A good hypothesis quietly decides your next two steps. Its variables point to the ISEF category where it competes best, and its measurable outcome dictates your methodology — what you control, how many samples you need, and what you record. That is why the hypothesis is worth getting right before anything else: it is the decision the whole project is built on.
Frequently asked questions
What is the difference between a research question and a hypothesis?
A research question is what you want to find out ("Does light color affect chlorophyll production?"). A hypothesis is your testable, directional prediction of the answer ("Blue light will produce more chlorophyll than red, because..."). You need both; the hypothesis is what your experiment actually tests.
Does my hypothesis really need a mechanism?
It is not strictly required, but including a plausible mechanism is one of the easiest ways to score higher on Scientific Thought. It shows you understand why you expect the result, not just that you do.
How specific should a hypothesis be?
Specific enough that two people reading it would design the same experiment. Name the organism or system, the concentration or condition, the metric, and the predicted direction.
Get your hypothesis right — with a mentor.
Finalia walks you from a vague interest to a falsifiable hypothesis and a competition-ready paper. Phases 1–4 are free; founding rate $99/mo.