10 questions science fair judges ask.

This sounds like an easy opener, but a strong answer goes well past the obvious one.

Judges want to know how the student arrived at the question: what the background reading showed, what problem or gap it revealed, and why this particular question was worth investigating. A student interested in Alzheimer’s disease should be able to trace how that broad interest narrowed into something specific and testable.

Students often hear this as “have I discovered something nobody in science has ever discovered?” That is rarely what is being asked. For a pre-college project the contribution is usually modest: testing a known relationship under different conditions, comparing approaches that have not been compared in this context, applying an existing technique to a new dataset, improving a design, or following up a question raised by earlier work. What matters is that the student understands how the project fits into what was already known.

This is the question that separates knowing what was done from understanding why. A judge might ask why a particular concentration, temperature range, or number of trials; why this model, metric, or set of features; why this material, design, or performance criterion.

The student does not need to have invented the technique, since researchers use established methods constantly, but they should understand why it suits the question. The AlphaGo student described in the companion piece could answer this precisely: he had gone back to the papers that introduced the algorithms he was using and read what their authors themselves described as unresolved, and comparing feedback methods was the angle that gap opened up.

For a traditional experiment, judges may ask directly about independent and dependent variables and controls. An engineering student may instead need to give the requirements and success criteria that determine whether a prototype works. A machine-learning student may need to explain what was varied, what was measured, and which metric was used.

The vocabulary changes; the underlying question does not.

What changed, what did you measure, and how did you decide whether it mattered?

A result is only useful once other explanations have been considered. In experimental work that usually means controls, but judges probe more broadly: were the groups comparable, was there enough replication, could the measurement introduce bias, was the data collected consistently, could something other than the independent variable explain what happened? Computational projects meet the same question through training and test data, sampling, data quality, and evaluation.

Students often have many graphs, tables, and model outputs, and a judge may ask for the one that matters most. This tests whether the student can tell data from evidence. Data are the measurements collected. Evidence is the part of those measurements that helps answer the research question.

There is a real temptation to treat a supported hypothesis as a successful project and an unsupported one as a failure. That is not how research works. A hypothesis is a prediction to be tested, and when the evidence does not support it, the scientific question simply becomes why. Perhaps the original explanation was wrong, another variable matters, the effect is smaller than expected, or the design needs improvement.

Unexpected results often produce the best judging conversations, because they give the student room to reason out loud.

Students fear this question, believing that admitting a weakness makes the project look worse. Usually the opposite is true. Every research project has limitations: sample size, measurement accuracy, experimental conditions, characteristics of the dataset, assumptions built into a model, limited time, a variable that could not be controlled, or how far the findings generalize. The question is whether the student understands them.

This question tests reflection. Research rarely proceeds as planned: measurements turn out to be difficult, materials behave unexpectedly, data are messier than expected, and a method that looked reasonable on paper reveals its weaknesses once it is used. By the end of a project, the student should understand the method better than when it began.

Research rarely ends with a final answer; a good project usually generates another question. A result may suggest a mechanism, a limitation may become the next experiment, an unexpected finding may deserve its own study, or a working prototype may reveal a new performance problem. This question tells a judge whether the student sees the project as a completed assignment or as part of an ongoing investigation.