How Investigable Questions Work in School Science | Why the Question Determines What Evidence You Can Collect

A child asks, “Why do plants grow better?”

That is a real scientific curiosity. It is not yet an investigation.

Better than what? Better under which condition? What kind of plant? What will count as “grow”? Height? Mass? Number of leaves? Root length? Over how much time? What can be changed? What must be held reasonably steady? What can actually be measured in a school laboratory or classroom?

Science education often jumps too quickly from curiosity to procedure. Students are given cups, soil, seeds and instructions before the intellectual job has become clear. They perform the experiment, fill in the table and write a conclusion. Yet they may never learn the question-design move that makes the evidence meaningful.

An investigable question is a question shaped so that observations, measurements, comparisons or tests can produce relevant evidence.

This is not a bureaucratic wording exercise. The question determines what data are needed, which variables matter, what comparison is fair enough to interpret, what tools are useful, what limitations remain and what claims can be made afterwards.

The difference between “Why do plants grow better?” and “How does the amount of light per day affect the increase in height of bean seedlings over fourteen days?” is not merely specificity. The second question begins to build an evidence pathway.

That pathway is the mechanism.

The 50-second route

If you only have a minute, keep this:

Curiosity is broader than investigation. A good science lesson should protect curiosity while teaching students how to turn part of it into a question evidence can address.
An investigable question usually identifies a phenomenon or relationship, what can vary, what can be observed or measured, and a scope that is feasible and safe.
The question controls the rest of the investigation: data, tools, comparison, variables, number of observations, analysis and the strength of the final claim.
“Testable” does not mean every good scientific question needs a controlled classroom experiment. Some are answered through observation, existing datasets, models, field measurements or comparison.
Students should learn to revise questions after discovering that a variable cannot be measured well, a test is unsafe, the timescale is too long or the proposed evidence would not distinguish between explanations.
A weak question often produces weak evidence even when the practical work is neat.
The final scientific habit is not “always do an experiment”. It is: choose a method that can generate evidence relevant to the question.

Why the question comes before the method

Suppose a class has thermometers, lamps, beakers and ice.

A procedure-first lesson might say:

“Put one beaker under the lamp and one away from it. Measure the temperature every minute.”

Students can complete that accurately. They may learn measurement and graphing. But unless the question is intellectually active, the procedure can feel like following a recipe.

Now begin differently:

How does distance from a lamp affect the rate at which water warms?

Immediately, the equipment acquires roles.

Distance is something to vary.
Temperature change is something to measure.
Time becomes part of the rate.
Starting temperature matters.
The amount of water matters.
Lamp power matters.
Measurement intervals matter.

The question turns objects into an investigation system.

This is why question design is not a decorative preface. It allocates jobs.

Stage 1: Start with a phenomenon worth wondering about

Investigable questions become mechanical when students never meet the thing that generated the question.

Show condensation forming on a cold bottle. Let a paper helicopter fall. Place different materials in sunlight. Observe shadows changing. Show seeds sprouting in unexpected places. Compare photographs of the same shoreline across years. Give students a strange dataset.

A phenomenon gives the question a reason to exist.

Students can begin with broad questions:

Why did this happen?
Does it always happen?
What changes it?
Which material works best?
Why are these two results different?
What would happen if…?

Do not correct broad questions too early. They reveal what students notice.

The teaching move is to say, in effect: That is a valuable question. Which part could we investigate with evidence we can collect?

This protects curiosity while introducing constraint.

Stage 2: Decide what kind of question it is

Not every scientific question has the same shape.

A useful classroom classification is:

Descriptive questions

What is happening? How often? How much? Where? When?

Example: How does the temperature of the school courtyard change from 8 a.m. to 2 p.m.?

These may require repeated observation rather than manipulating a variable.

Comparative questions

How are two or more conditions different?

Example: Which of three materials reduces heat transfer most over ten minutes?

The core design job is making the comparison interpretable.

Relationship questions

How does one measured quantity change with another?

Example: How does ramp height affect the distance a toy car travels?

These often lead to graphs and attention to patterns.

Causal test questions

Does changing X cause a change in Y under these conditions?

Example: How does the concentration of salt in water affect the time taken for a fixed ice cube to melt?

These usually require stronger control of alternative explanations.

Explanatory questions

Why does the phenomenon occur?

Example: Why does a metal spoon feel colder than a wooden spoon in the same room?

Explanatory questions often cannot be settled by one classroom test. Students may need a combination of observations, models, prior scientific ideas and targeted investigations.

Knowing the question type helps students choose a method rather than automatically chanting “independent variable, dependent variable, control variables” for every scientific problem.

Stage 3: Turn vague words into observable quantities or categories

Students often use words that feel meaningful but are not yet operational.

“Better.” “Faster.” “Healthier.” “Strong.” “Clean.” “More effective.” “Good soil.” “Big plant.”

Science does not require every concept to become a perfect number, but an investigation needs a defensible observation rule.

Suppose the question is:

Which paper towel is strongest?

What is strength?

Maximum mass held before tearing?
Number of coins supported when wet?
Force required to tear a strip of fixed width?
Resistance to puncture?

Each definition creates a different test.

This is a profound scientific lesson. A measurement does not simply reveal a concept. It represents the concept through an operational choice.

Students should learn to ask:

What exactly will we observe that counts as evidence for the word we are using?

That question prevents many beautiful but meaningless experiments.

Stage 4: Identify what can vary—and what variation would answer the question

A question such as:

How does water temperature affect the time sugar takes to dissolve?

suggests at least two variables:

water temperature;
dissolving time.

But the deeper reasoning is relational.

We are not interested in temperature alone or time alone. We want to know whether changes in one are systematically associated with changes in the other under the test conditions.

This helps students understand variables as roles in a question rather than vocabulary to label after the procedure has already been written.

For a causal classroom test, the student asks:

What will I deliberately change?
What outcome will I observe?
What other differences could also change the outcome?

For an observational study, the first variable may not be deliberately changed at all. The important issue is still how the quantities are defined and related.

Stage 5: Check whether the question can actually be answered with the available world

A scientifically interesting question can be a poor classroom investigation.

“How does sleep affect exam performance?” matters, but a student cannot ethically assign classmates to severe sleep deprivation.

“How does carbon dioxide affect global temperature?” is central to climate science, but the full causal question is not resolved by putting two bottles under lamps for twenty minutes.

“How does exercise affect heart health over a lifetime?” exceeds the timescale of a school project.

This is where feasibility becomes part of scientific judgement.

Students should test a proposed question against several constraints:

Time: Can a meaningful change occur during the available period?
Measurement: Can the outcome be observed with sufficient resolution?
Control: Can major alternative explanations be managed?
Ethics: Is it acceptable to manipulate this condition?
Safety: Can the investigation be performed without unreasonable risk?
Resources: Are suitable tools and materials available?
Scale: Is the phenomenon too large, small, slow or complex for direct classroom testing?
Replication: Can enough observations be collected to avoid treating one odd result as the pattern?

A question that fails this screen should be revised, not forced.

Question revision is science, not defeat.

Stage 6: Make the comparison fair enough for the claim you want to make

“Fair test” is useful language for young learners, but it can become a slogan.

The deeper idea is alternative explanations.

Suppose two plants receive different amounts of light, but one also receives more water and starts larger. If it grows more, the evidence cannot clearly separate the effect of light from the other differences.

Controls matter because we are trying to make competing explanations less plausible.

This does not mean a school experiment must achieve perfect laboratory control. Real experiments have noise. Biological organisms vary. Instruments have limits. Classrooms are not research institutes.

The important habit is to ask:

If the outcome changes, what else could reasonably have caused the change?

Then control the most important alternatives where feasible and state the remaining limitations honestly.

That is more intellectually useful than memorising “keep everything the same”, which is often impossible and sometimes conceptually wrong.

Stage 7: Ask whether the evidence will be able to distinguish between plausible answers

This is one of the most neglected question-design tests.

Imagine a student asks:

Does fertiliser help plants grow?

They compare one fertilised plant with one unfertilised plant. The fertilised plant ends taller.

Could the result distinguish:

a real fertiliser effect;
a difference in seed quality;
uneven watering;
measurement error;
random growth variation?

Not very well.

Now imagine several plants in each condition, standardised procedures, repeated measurements and an agreed growth measure. The evidence becomes more discriminating.

The question has not changed much in words. The evidence design has.

Students need to see that evidence quality depends on whether plausible rival explanations would produce distinguishable patterns.

This is how school science begins to move from demonstration toward inference.

Stage 8: Decide how much data the claim needs

A single measurement can answer some questions.

“What is the current temperature of this water?” may need one reading if the instrument is adequate.

A relationship claim usually needs more.

If students are investigating how ramp height affects travel distance, two heights provide only a crude comparison. Several heights can reveal whether the relationship looks linear, curved, threshold-like or noisy.

Repeated trials help separate a repeatable tendency from an accidental outcome.

The exact number of observations depends on context, but the reasoning principle is universal:

The strength and shape of the claim determine how much evidence you need.

This is an important antidote to classroom rituals such as “always do three trials”. Three may be sensible in some lessons and arbitrary in others.

Students should know why repetition exists.

A concrete example: from “Which soil is best?” to an investigable question

A Primary 5 class wants to know which soil is best for plant growth.

The original question contains three hidden problems:

“soil” varies along many dimensions;
“best” is undefined;
plant growth takes time and varies between individual plants.

The teacher does not replace the question immediately. The class improves it.

What plant? Mung bean seedlings.
What outcome? Increase in stem height over ten days.
Which soil categories? Standard potting mix, sandy soil and clay-rich soil.
How much water? Same measured amount.
How much light? Same location, pots rotated.
How many plants? Several per soil condition.
How will height be measured? From soil surface to highest point of stem, at the same time each day.

The question becomes:

How does soil type affect the mean increase in stem height of mung bean seedlings over ten days under the same watering and light conditions?

This is more cumbersome than “Which soil is best?” but now the evidence has a job.

The class can still discuss what the question does not establish. Height is not the whole of plant health. Ten days is short. The soil categories may differ in nutrients and drainage simultaneously. The conclusion should remain scoped to the conditions tested.

This is scientific maturity: improved precision followed by honest limits.

Another example: paper helicopters

Students make paper helicopters and ask:

Does blade length affect how long a paper helicopter takes to fall?

This is a strong classroom question because the variables are accessible, the timescale is short and multiple trials are possible.

But design choices still matter.

If changing blade length also changes total paper area and mass, what exactly is being manipulated? If the drop height varies, flight time changes. If one student releases with a push and another lets go gently, launch conditions differ. If timing is manual, reaction time introduces measurement noise.

The point is not to eliminate every imperfection. It is to make students see the chain from question to evidence.

They may refine the question:

For paper helicopters made from the same template and paper, how does blade length affect mean fall time from a fixed height of two metres?

Then they can decide how many blade lengths, how many trials and how to record variation.

Why a hypothesis is not the same as an investigable question

Students are often taught to write:

“If X increases, then Y will increase because…”

That can be useful. But the hypothesis answers a different job.

The question says what relationship is being investigated.
The hypothesis states an expected outcome or explanation.

A class can have a good investigable question without knowing enough to make a confident directional prediction. In exploratory work, forcing a hypothesis can encourage students to pretend they already know.

Conversely, a beautifully written hypothesis cannot rescue a question that cannot generate relevant evidence.

Keep the jobs separate.

Why “testable” does not always mean “do a controlled experiment”

This misconception narrows science.

Some questions are investigated observationally:

How does lichen coverage vary with distance from a road?
How does shadow length change during the day?
How do bird species counts differ between two habitats?

Some use existing data:

How has local rainfall changed across decades?
Is there a relationship between latitude and average temperature across selected cities?

Some use models:

How would changing one parameter affect a simulated epidemic?
Which structural feature makes a bridge model more stable under load?

Some scientific questions are historical:

What evidence supports a past mass-extinction event?
How do rock layers indicate earlier environments?

The scientific principle is not “manipulate a variable whenever possible”. It is “choose a method that can produce relevant evidence for the question.”

Evidence from current education guidance

UNESCO’s 2026 teaching resources explicitly identify the formulation of investigable questions, analysis of evidence and active scientific practices as important parts of classroom science. The Next Generation Science Standards similarly frame asking questions, planning and carrying out investigations, analysing data, constructing explanations and arguing from evidence as connected practices rather than isolated skills.

That connection matters.

A question should not be taught as the first box on a worksheet. It determines the later boxes.

If the question changes, the relevant data may change.
If the data cannot answer the question, the method must change.
If the method produces ambiguous evidence, the claim must weaken.
If the claim exceeds the evidence, the scientific chain has broken.

Question design is therefore part of evidence literacy.

Caveat: neat school investigations are models of science, not miniature copies of research science

School science simplifies.

Variables are cleaner. Timescales are shorter. Equipment is limited. Teachers often know the likely outcome. Safety rules restrict possibilities. Curriculum goals shape the phenomenon selected.

Students should not be told that real science is simply a bigger version of the five-step school method.

Real research can involve exploratory observation, instrument development, model building, fieldwork, messy datasets, competing theories, replication, statistical inference and years of revision.

The classroom goal is more modest and more defensible: teach students the logic connecting a question to evidence.

That logic transfers even when the methods become more sophisticated.

Common failure mode 1: the question contains the answer

“Does more sunlight make plants grow taller?”

This can invite confirmation rather than investigation, especially if students already know what result the teacher expects.

A more open formulation may be:

“How does daily light exposure affect the change in seedling height over fourteen days?”

The difference is subtle. The relationship is still specified, but the result is not embedded as a desired conclusion.

Common failure mode 2: the variables are measurable but scientifically uninteresting

Students can measure many things.

“How does pencil colour affect the time a toy car takes to roll down a ramp?” is testable. It is also unlikely to have a meaningful mechanism unless colour is connected to another relevant property.

Testability is not enough.

A worthwhile question should connect to a phenomenon, concept, model or plausible relationship.

Science education should teach both feasibility and significance.

Common failure mode 3: “keep everything the same” becomes impossible perfectionism

No two seedlings are identical. No two hand timings are identical. Room temperature drifts. Instruments have tolerances.

Students can become confused when teachers demand that “everything except one variable” be identical.

Use better language:

“Keep the important alternative causes as similar as reasonably possible, measure what matters, repeat where useful, and describe what still varied.”

That is closer to scientific reasoning.

Common failure mode 4: measurement precision is mistaken for conceptual quality

A digital sensor can record to three decimal places and still measure the wrong thing.

If the question is about “plant health” and the only measure is height, more precise height data does not solve the validity problem.

Before asking, “How accurately can we measure?” ask, “Does this measurement represent the outcome we claim to care about?”

Precision without validity is false confidence.

Common failure mode 5: students collect data before deciding what pattern would matter

The class records twenty numbers and only afterwards asks what to do with them.

A better design asks in advance:

If X matters, what pattern would we expect?
Would we compare means?
Look for a trend?
Examine a threshold?
Count categories?
Graph change over time?

This does not predetermine the result. It gives the data a planned analytical role.

Common failure mode 6: the conclusion answers a bigger question than the investigation tested

One classroom test shows that a black paper surface warmed more than a white paper surface under one lamp.

The conclusion becomes: “Dark colours always absorb more heat.”

That leap may align with a wider scientific principle, but the experiment itself tested a narrower condition.

Teach students to scale the conclusion:

“In our setup, the black paper surface showed a larger temperature increase than the white surface over ten minutes under the same lamp.”

Then connect that observation to established scientific knowledge carefully.

A strong conclusion knows the size of its evidence.

The learner route: a question-builder you can use

When you are given a science project, start with seven prompts:

Phenomenon: What am I curious about?
Relationship: What might vary or differ?
Evidence: What could I observe or measure?
Comparison: What would I compare?
Feasibility: Can I do this safely with the time and tools available?
Alternative explanations: What else could change the result?
Scope: What exactly could my evidence justify saying?

Then write the question.

Do not worry if it changes later. A revised question can be evidence of better scientific thinking.

The parent route: help with thinking, not with designing a perfect experiment for the child

Parents can accidentally take over science projects because the child’s first question is messy.

Resist the urge to transform it immediately into a polished research proposal.

Ask: “What do you mean by ‘better’?” “How could you see or measure that?” “What would you compare?” “What else might affect it?” “Can you actually do that in a week?” “What would the result allow you to say?”

These questions keep ownership with the learner.

If the investigation requires unsafe materials, human-subject manipulation or specialist equipment, help the child redesign the question rather than finding a way around the constraint.

Constraint is part of science.

The teacher route: teach question revision explicitly

A useful lesson can show three versions of the same question.

Version 1: What makes ice melt?

Version 2: Does salt make ice melt faster?

Version 3: How does salt concentration affect the time required for equal-mass ice cubes to melt at the same room temperature?

Ask students:

What became clearer?
What evidence became possible?
What new limitations appeared?
Is version 3 always “better”, or merely better for a particular investigation?

Then give students broad questions and let them revise.

The teaching goal is not a formulaic sentence frame. It is the judgement behind the refinement.

How to know whether students understand investigable questions

Do not only ask them to label variables.

Give them flawed questions and ask for diagnosis.

Examples:

Which music makes plants happiest?
Problem: “happiest” is not a defined plant outcome; music category also needs specification.

Does exercise make people healthier?
Problem: too broad, ethical and causal complications, “healthier” undefined, timescale unclear.

How does the number of paper clips attached to a paper spinner affect its fall time from 1.5 metres?
Potentially workable; discuss mass, release method, trials and safety.

Why is the Moon sometimes visible during the day?
Scientifically meaningful but not best handled as a controlled classroom experiment; use models and astronomical observations.

Students who can distinguish these cases understand more than a variable checklist.

Question quality and data quality are different problems

A strong investigable question can still produce poor data.

The thermometer may be badly placed. Students may record at inconsistent times. Samples may be too small. Categories may be scored differently by different observers. A sensor may saturate. A ruler may not have enough resolution for the expected change.

The reverse is also true: beautifully precise data can be collected for a weak question.

That distinction helps students diagnose investigations more intelligently.

Ask two separate questions:

Question quality: If we had excellent data, would those data actually address the scientific relationship we care about?

Data quality: Given this question, did our method produce observations accurate and consistent enough to support interpretation?

This prevents a common classroom habit in which students respond to every problem by “doing more trials”. More trials can reduce some forms of random variation. They cannot fix a measurement that represents the wrong construct or a comparison that confounds two major causes.

Science improves when learners know which layer has failed.

An investigable question should survive a prediction audit

Before students collect data, ask them to imagine several possible outcomes.

Suppose the question is:

How does water temperature affect the time required for a fixed mass of sugar to dissolve under the same stirring conditions?

What if warmer water dissolves the sugar faster?
What if there is almost no difference?
What if the results are highly variable?
What if the hottest condition produces an unexpected anomaly?

For each possibility, what could the class reasonably infer?

This “prediction audit” checks whether the planned evidence will actually be interpretable. It can expose hidden problems before materials are used.

If every possible outcome would lead students to write the same conclusion, the investigation is not functioning as a test. If an unexpected outcome would be dismissed automatically as “wrong”, the class is performing confirmation, not inquiry.

A good question leaves room for the world to answer in more than one way.

Question revision after data is a scientific strength, not cheating

Students are often taught that the question must be fixed at the beginning and never touched again.

For a formal assessment, the investigation may indeed need a stable plan. But scientifically, early evidence can reveal that the original question was poorly framed.

Perhaps the expected temperature range was too narrow to produce measurable change. Perhaps “leaf colour” was too subjective as an outcome. Perhaps the apparatus could not keep humidity stable. Perhaps a supposedly minor variable dominated the system.

Researchers revise questions because contact with the world teaches them what can and cannot be known through the original design.

School science can model this honestly.

Students can preserve the original question, document why it proved weak, then write a revised question and explain what changed. That creates a richer record of scientific reasoning than quietly modifying the procedure and pretending the first plan was perfect.

The important rule is transparency: revision should be documented, not hidden.

From question to claim: keep the chain visible

A useful final check is to draw the whole investigation as a chain:

phenomenon → question → operational definitions → method → observations → pattern → claim → limitation

Then ask where each arrow could break.

Did the operational measure really represent the concept?
Did the method isolate the comparison?
Did the observations show enough consistency?
Did the analysis match the question?
Did the claim stay within the evidence?

This chain is powerful because it shows students that “the experiment” is not a single thing. It is a sequence of reasoning decisions.

When a conclusion is weak, the response is not merely “be more careful next time”. The class can locate the weak link.

That diagnostic habit is one of the most transferable forms of scientific literacy a school can teach.

Frequently asked questions

Must an investigable question include “How does X affect Y?”

No. That frame is useful for some relationship and causal questions, but science also uses descriptive, comparative, observational and explanatory questions.

Is “why” a bad science question?

No. “Why” questions are central to science. They may need to be decomposed into smaller questions that different forms of evidence can address.

Must students always have a hypothesis?

No. A hypothesis can be valuable when a prediction is meaningful and grounded. Exploratory investigations may begin with less specific expectations.

How many variables should change?

In a simple causal test, deliberately changing one focal factor while managing important alternatives can make interpretation easier. More advanced investigations may study multiple variables. The key is whether the design can support the intended inference.

Why repeat trials?

Repetition helps reveal variability and reduces the chance that one unusual observation dominates the conclusion. The appropriate number depends on the phenomenon and measurement.

Can internet research answer an investigable question?

Sometimes existing datasets or published observations are the appropriate evidence. But copying information from websites is not automatically an investigation. Students still need a question, source evaluation, a method of comparison or analysis, and a scoped claim.

What if the result contradicts the hypothesis?

That is useful evidence. Check the method, measurement and alternative explanations, then revise the explanation or prediction as needed. The purpose of an investigation is not to make the hypothesis win.

What if the experiment “doesn’t work”?

Diagnose what failed. Was there too little variation? Was the outcome hard to measure? Did the apparatus introduce noise? Was the timescale wrong? A failed design can teach question-method alignment extremely well if the class analyses it rather than hides it.

The deeper lesson: science begins by deciding what the world could tell us

Questions are sometimes treated as the soft, creative beginning of science and measurement as the hard, serious part.

That division is false.

A scientific question already contains decisions about what difference matters, what could count as evidence, what comparison is informative and what kind of answer is possible.

Those decisions are intellectual infrastructure.

When students learn to build investigable questions, practical work changes character. They are no longer merely following instructions that happen to involve scientific equipment. They are learning to connect curiosity to evidence through a method whose limitations they can explain.

The deepest habit is not “ask a testable question”.

It is this:

Before collecting evidence, decide what question the evidence must be capable of answering.

That habit protects science from becoming activity without inference.

Sources

UNESCO, From evidence to the classroom: new UNESCO publication provides tools for teaching Reading, Mathematics and Science (11 May 2026): https://www.unesco.org/en/articles/evidence-classroom-new-unesco-publication-provides-tools-teaching-reading-mathematics-and-science
UNESCO, Aportes para la enseñanza de la Lectura, las Matemáticas y las Ciencias (2026): https://www.unesco.org/en/articles/aportes-para-la-ensenanza-de-la-lectura-las-matematicas-y-las-ciencias-estudio-regional-comparativo
Next Generation Science Standards, Science and Engineering Practices: https://www.nextgenscience.org/glossary/science-and-engineering-practices-sep
National Academies, A Framework for K–12 Science Education — Essential Practices for K–12 Science Classrooms: https://nap.nationalacademies.org/visualizations/practices-for-k-12-classrooms/resources/practices-for-k-12-classrooms.pdf