Designing a fair test
Every Science 7 unit has experiments in it — bean plants in different soils, cups of hot water cooling, paper bridges loaded until they collapse. Each one is only worth doing if the test is fair: one thing changed on purpose, one thing measured, and everything else kept the same.
Why does a test have to be fair?
Because if two things change at once, you cannot tell which one caused the result. Put a bean plant with fertilizer on a sunny windowsill and one without fertilizer in a dim corner, and the first grows taller — but was it the fertilizer, or the light? No amount of careful measuring afterwards can say.
A fair test changes just one thing, the manipulated variable, measures its effect, the responding variable, and keeps everything else the same: the controlled variables. Then, if the results differ, only one thing can explain it.
Where it turns up
- New crop varieties — before a new wheat or canola variety is recommended to Alberta farmers, it is grown beside older varieties at test sites around the province for several years, so only the variety differs
- Insulation labels — insulation sold in Canada is labelled with an RSI value measured by a standard test, so any two products can be compared fairly
- Hockey helmets — helmets sold in Canada must meet a CSA standard that puts every design through the same impact tests, so a pass means the same thing whatever the brand
- Weather stations — official thermometers sit in shaded, ventilated boxes between 1.25 and 2 metres above the ground, so a reading in Lethbridge can be compared with one in Grande Prairie
- 1. The words, first
- 2. From a question to three kinds of variable
- 3. Spotting the unfair test
- 4. Repeat it, then average it
- 5. What the results can and cannot tell you
- 6. The same skill in every unit
The words, first
The idea: Controlled and control sound alike and mean different things: many conditions kept the same, versus one whole setup to compare against.
| Word | What it means |
|---|---|
| Testable question | A question you can answer by changing something and measuring what happens. |
| Manipulated variable | The one thing you change on purpose. Later courses call it the independent variable. |
| Responding variable | What you measure to see the effect. Later called the dependent variable. |
| Controlled variables | Everything you keep the same, so only the manipulated variable can explain a difference. |
| Control | One setup with the manipulated variable left out — the plant with no fertilizer, the cup with no wrapping — to compare against. |
| Hypothesis | A prediction you can test, with a reason: If … then … because … |
| Trial / mean | One run of the experiment / the average: add the results and divide by how many there are. |
From a question to three kinds of variable
The idea: Write the question as “How does ___ affect ___?” The first blank is the manipulated variable, the second is the responding variable, and everything else must be controlled.
How does [the thing I change] affect [the thing I measure]?
| Not testable yet | Testable |
|---|---|
| Which soil is best? | How does the type of soil affect the height of bean plants after three weeks? |
| Which wrapping is the best insulator? | How does the material wrapped round a cup affect how much the water in it cools in 20 minutes? |
| Are triangles strong? | How does the number of triangles in a paper bridge affect the load it holds before it fails? |
| Should we build a dam here? | Not a fair-test question: science can predict what a dam would do, but not whether it is worth it |
Worked: plants. How does the amount of water each day affect the height of bean plants after three weeks?
- Manipulated: water per day — 0, 25, 50 and 100 mL.
- Responding: the height of each plant after three weeks, in centimetres.
- Controlled: the kind of bean, the pot size, soil from the same bag, the same windowsill, the same planting day.
Finding the controlled variables. Ask what else could make a bean plant grow taller — light, soil, temperature, the seed — and hold each one the same. If the 100 mL pot sits closest to the window, a taller plant might be the light's doing, not the water's.
A hypothesis with a reason. “If a bean plant gets more water, then it will grow taller, because a plant needs water to make sugar by photosynthesis.” The because is what the experiment puts to the test.
Spotting the unfair test
The idea: Look for a second difference between the setups. If there is one, the test cannot say which difference caused the result.
| Question | Design | Fair? |
|---|---|---|
| Does a black can of water warm faster in sunlight than a white one? | Black can on a sunny windowsill, white can on a shady shelf | No — the sunlight differs too |
| Which wrapping keeps a cup of hot water warmest? | Cups wrapped in wool, foil and newspaper, plus one unwrapped, each with 200 mL of 80 °C water and a lid, read after 20 minutes | Yes |
| Does the number of triangles affect the load a bridge holds? | A 4-triangle bridge of paper, an 8-triangle bridge of cardboard | No — the material differs too |
Why a control helps. Suppose the cup wrapped in wool cooled by 12 degrees in 20 minutes. Is that good? Without an unwrapped cup tested at the same time, there is nothing to compare it with. The unwrapped cup is the control.
Repeat it, then average it
The idea: One result could be luck. Repeating each setup shows how much the results vary, and the mean evens that out.
Worked: heat. Four cups each start with 200 mL of water at 80 °C and a lid; one is unwrapped and three are wrapped. The temperature after 20 minutes, in °C, three trials each:
| Wrapping | Three trials | Mean |
|---|---|---|
| None (the control) | 58, 57, 59 | 58 |
| Foil | 59, 58, 60 | 59 |
| Newspaper | 64, 63, 65 | 64 |
| Wool | 68, 67, 69 | 68 |
wool: (68 + 67 + 69) ÷ 3 = 68 °C, a drop of 80 − 68 = 12 degrees · control: a drop of 80 − 58 = 22 degrees
When is a difference real? Compare the gap between two means with how far each set of trials spreads. Wool and newspaper differ by 4 degrees and none of their trials overlap, so wool really did better. Foil and the control differ by 1 degree, and their trials overlap — 58 and 59 turn up in both — so this test found no clear difference between foil and no wrapping.
An odd result. If one newspaper trial had read 55, do not quietly drop it. Say it does not fit, look for a reason — perhaps the lid was left off — and run that trial again.
Graphing it. Wrappings are categories, so the means go on a bar graph; a number such as layers of wool would go on a line graph.
What the results can and cannot tell you
The idea: A fair test supports a conclusion about what was tested, in the conditions it was tested in. It does not prove anything beyond that.
What you may claim. “In this test, wool kept the water warmer than newspaper, foil or no wrapping over 20 minutes.” The data support that. “Wool is the best insulator” goes too far: only three materials were tested, for 20 minutes, on cups of water. Results that match a hypothesis support it; they do not prove it.
Structures: testing until it fails. A paper bridge loaded until it collapses can never be tested again. Build several of each design, and treat each one as a trial.
When nothing was changed on purpose. A class might count more dandelions in sunny quadrats than in shady ones. Nobody changed the sunlight; they only counted. That shows a pattern, not a cause — the sunny spots might also have different soil. Only changing one thing on purpose can show that it causes another.
The same skill in every unit
The idea: Every unit has its own version of this question: what is manipulated, what is responding, what is controlled. The words change; the rule does not.
| You change | You measure | You keep the same |
|---|---|---|
| A · Ecosystems: hours of light each day | Number of duckweed plants after two weeks | Water volume, temperature, number of plants at the start |
| B · Plants: type of soil | Height of bean plants after three weeks | Water, light, pot size, kind of bean |
| C · Heat: colour of a can | Temperature rise of the water inside after 15 minutes in sunlight | Water volume, starting temperature, can size |
| D · Structures: number of triangles in a paper bridge | Load held before it fails, in newtons | Paper, span, glue, how the load is added |
| E · Planet Earth: minutes spent shaking rock pieces in a jar of water | Mass of sediment broken off | Type and mass of rock, water volume, how hard it is shaken |
Every experiment has a weakness. How hard a jar is shaken is almost impossible to keep the same by hand. Naming a weakness like that, and how you would fix it, belongs in a good conclusion. Science 9's toolkit of the same name takes the skill further.