From data to evidence
Data are the numbers and observations you collect. They become evidence when you use them to support or challenge a claim. A scientist’s job is to decide what the evidence actually shows, no more and no less.
A strong scientific explanation has three parts, often called claim–evidence–reasoning:
| Part | What it does | Example |
|---|---|---|
| Claim | Answers the question | Salt water freezes at a lower temperature than fresh water. |
| Evidence | Gives the data | Fresh water froze at 0 °C; salt water froze at −2 °C in all three trials. |
| Reasoning | Explains why the evidence supports the claim | Salt particles get in the way of water molecules forming ice, so it must get colder before ice forms. |
Reviewing variables
To judge evidence, first check the variables:
- The independent variable is changed on purpose.
- The dependent variable is measured.
- Controlled variables are kept constant.
If a controlled variable wasn’t actually controlled, the results may not mean what they seem to.
Correlation is not causation
A correlation means two variables change together. Causation means one actually causes the other.
Ice cream sales and sunburns both rise in summer. They are correlated, but ice cream doesn’t cause sunburns. A third factor, hot sunny weather, causes both. This hidden factor is called a confounding variable.
Only a controlled experiment, where one variable is changed and others are held constant, can show causation convincingly.
| Correlation | Possible hidden cause |
|---|---|
| Students with more books at home get higher grades | Family income, time spent reading |
| Cities with more firefighters have more fires | Bigger cities have more of both |
Is the evidence reliable?
Evidence is reliable if repeating the investigation gives similar results. To improve reliability:
- repeat trials and average the results
- use larger sample sizes
- measure carefully with suitable tools
Is the investigation valid?
An investigation is valid if it really tests what it claims to test. A test of "Which fertilizer is best?" isn’t valid if the plants given fertilizer A also got more sunlight.
Watching out for bias
Bias is anything that pushes results unfairly in one direction. Examples:
- Sampling bias: asking only your friends whether they like a new lunch menu.
- Confirmation bias: noticing only data that support what you already believe.
- Funding bias: a study paid for by a company that wants a certain result.
Scientists reduce bias by using random samples, "blind" tests where the tester doesn’t know which group is which, and peer review.
Anomalies
An anomaly (or outlier) is a result that doesn’t fit the pattern. Don’t just erase it. Check whether something went wrong, such as a measuring error. If you find a clear mistake, you may repeat that trial. If not, report the anomaly honestly.
Common mistakes
- Claiming more than the data show. "This proves…" is rarely right; "this supports…" is better.
- Confusing correlation with causation.
- Ignoring results that don’t fit.
Try it yourself
A student tests whether music helps plants grow. She plays music to plants on a sunny windowsill and keeps silent plants in a shaded corner. The music plants grow taller.
- What is her claim?
- Is her investigation valid? Why or why not?
- Name a confounding variable.
- Redesign the experiment to make it fair.
Quick recap
- Evidence is data used to support or challenge a claim.
- Use claim–evidence–reasoning to build explanations.
- Correlation doesn’t prove causation; look for confounding variables.
- Reliable results can be repeated; valid tests measure what they claim to.
- Watch for bias and report anomalies honestly.