Designing a fair test
Six in every ten marks on the Science 9 PAT are for skills rather than facts, and many of those questions ask the same thing: given an investigation, say what was changed, what was measured and what was kept the same — and whether the result means anything.
Why does a test have to be fair?
Because when two things change at once, the result has two possible causes, and no amount of careful measuring afterwards can tell you which one it was.
In 1747 a Royal Navy surgeon, James Lind, had twelve sailors sick with scurvy on his ship. He kept them in the same part of the ship on the same diet, split them into six pairs, and gave each pair a different treatment. The pair given oranges and lemons recovered far faster than any other. Holding everything else the same is what made the result mean something: if the citrus pair had also had better food, nobody could have said what helped.
Grade 9 gives the idea three names. The manipulated variable is the one thing you change on purpose, the responding variable is what you measure, and the controlled variables are everything you hold still.
Where it turns up
- New medicines — a clinical trial compares patients given the drug with a similar group given a placebo, assigned at random so the groups differ only in the treatment
- Prairie farms — a new fertilizer or seed variety is tested on side-by-side strips of one field, so soil and weather are the same for both
- Websites and apps — an A/B test shows two versions of a page to randomly chosen visitors, with one thing different, such as a button's wording
- Fuel-consumption ratings — new cars sold in Canada are rated with the same standardized laboratory tests, so models can be compared fairly
- 1. The words, first
- 2. One question, three kinds of variable
- 3. Spotting the unfair test
- 4. Repeats, averages and odd results
- 5. Reading the results
- 6. The same skill in all five units
The words, first
The idea: Control and controlled sound alike and mean different things. Two of these words change name in later courses, so both names are here.
| Word | What it means |
|---|---|
| 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 kept the same, so only the manipulated variable can explain a difference. |
| Control | One whole setup with the manipulated variable left out — the plant with no fertilizer — so there is something to compare against. |
| Hypothesis | A testable prediction with a reason: If … then … because … |
| Trial | One run of the procedure. Repeated trials are the same run done again. |
| Mean | The average: add the results and divide by how many there are. |
| Outlier | A result that does not fit the pattern, often from a mistake in that trial. |
| Interpolate / extrapolate | Estimate a value between measured points / beyond them. The first is fairly safe; the second is a guess. |
One question, three kinds of variable
The idea: Write the question as “How does ___ affect ___?” The first blank is manipulated, the second is responding, and everything else must be controlled.
How does [manipulated] affect [responding]?
Worked. How does water temperature affect how quickly an antacid tablet reacts?
- Manipulated: water temperature — 10, 20, 30, 40 and 50 °C.
- Responding: the time until the fizzing stops, in seconds.
- Controlled: volume of water, brand and size of tablet, tablet kept whole, the same container, no stirring.
Why “kept whole” is on the list. A broken tablet has more surface area and reacts faster, so one crushed tablet would make its temperature seem to give a faster reaction than it really does. That is how to find controlled variables: ask what else could change the responding variable, and hold each of those still.
A hypothesis with a reason. “If the water is warmer, then the tablet will stop fizzing sooner, because the particles move faster and collide more often.” The because turns a guess into something the experiment can support or contradict.
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.
A common skills question gives several designs and asks which is fair. Check each against one rule: does anything differ besides the manipulated variable?
| Question | Design | Fair? |
|---|---|---|
| Does fertilizer make bean plants taller? | Fertilized plant on a sunny windowsill; unfertilized plant in a dim corner | No — light differs too |
| Which metal wire has the least resistance? | 1 m of copper and 50 cm of nichrome, same thickness, same battery | No — length differs too |
| Does salt water freeze more slowly than tap water? | Equal volumes in identical cups, same freezer shelf, checked every 10 minutes | Yes |
| Which paper towel absorbs the most water? | Three brands cut to the same size, each dipped for 10 s, one piece of each | Fair, but weak — one trial each |
Control versus controlled. Controlled variables are conditions kept the same in every setup. A control is one whole setup — the bean plant with no fertilizer — that shows what happens without the change. Without it, a fertilized plant that grew 12 cm tells you nothing: the unfertilized one might have grown 12 cm too.
Comparing designs is a fair test too. Which insulation keeps a model habitat warmest? Test every design under identical conditions, and judge it against the criteria the question gives.
Repeats, averages and odd results
The idea: One result could be luck. Repeats show how much results vary, and the mean evens the variation out — once any result that is clearly wrong has been dealt with.
Worked. Three trials at 20 °C: 44 s, 46 s and 45 s. Mean = (44 + 46 + 45) ÷ 3 = 45 s, and the trials agree to within a second, so 45 s can be trusted.
At 40 °C: 25 s, 24 s and 41 s. The mean of all three is (25 + 24 + 41) ÷ 3 = 30 s, but 41 s fits neither the other trials nor the trend. Perhaps the water had cooled or the tablet was chipped. Do not quietly delete it: call it an outlier, say what you suspect, and repeat the trial. If the repeat gives 26 s, the mean is (25 + 24 + 26) ÷ 3 = 25 s.
Sample size with living things. One bean plant per group is not enough, because individuals vary — that is the whole of Unit A — and one naturally tall plant could make a useless fertilizer look good. Ten plants per group lets the variation even out.
Reading the results
The idea: Manipulated variable along the bottom, responding variable up the side. Describe the trend with numbers, estimate between points, and be careful beyond the ends.
Line or bar? A line graph when the manipulated variable is a number that could take any value in between, like temperature. A bar graph when it is a set of categories, like brands of paper towel — there is nothing between one brand and the next.
Describing a trend. “As the temperature increased from 10 °C to 50 °C, the time fell from 64 s to 20 s, falling by less with each 10° step.” Numbers from the data, and the shape. “It went down” earns less.
Interpolating and extrapolating. At 35 °C, between two measured points, the curve gives about 29 s — a reasonable estimate. At 90 °C, far beyond the data, nobody knows whether the curve keeps its shape, and a good answer says so.
What the conclusion may claim. “The results support the hypothesis: warmer water made the tablet react faster between 10 °C and 50 °C.” Support, not prove, and the range stated. Then one sentence on the weakness you would fix next time.
When nothing was manipulated. Survey data — more mayfly nymphs where a river has more oxygen — show a pattern, but nobody changed anything on purpose, so they cannot show that one thing causes the other.
The same skill in all five units
The idea: Every unit has its own version of this question. The words change; the rule does not.
| Unit | Manipulated | Responding | Some controlled variables |
|---|---|---|---|
| A · Diversity | Fertilizer per pot | Bean plant height after three weeks | Light, water, soil, seed variety |
| B · Matter | Water temperature | Time for a tablet to stop fizzing | Water volume, tablet size |
| C · Environment | Salt in the water, in ppm | Radish seeds that sprout | Number of seeds, water volume, temperature |
| D · Electricity | Length of wire | Current on an ammeter | Wire material and thickness, battery |
| E · Space | Angle of a solar panel to a lamp | Current the panel produces | The lamp and its distance, the panel |
Once you can fill in a row like these for a scenario you have never seen, the skills questions stop being the hard ones. Science 10's toolkit Predict, observe, conclude takes the next step.