Planning Fair Tests and Investigations

You might wonder which paper towel is best. Before you test it, decide what “best” means. Do you want it to soak up more water, or hold more weight when wet? Those are different questions.

Turn a wondering into a question

A testable question can be checked with observations or measurements. For a fair test, name what you will change and what you will measure. Try: “How does the brand of paper towel affect the amount of water it soaks up?” You change the brand. You measure the water in milliliters.

“Do plants like fertilizer?” needs work too. A plant cannot tell you what it likes. Ask: “How does the amount of fertilizer affect bean plant height after three weeks?” Now the result is a height in centimeters.

An opinion such as “Which color is prettiest?” has no single answer to measure. A question can also be scientific but too large or unsafe for your classroom. Choose something you can check with the time and tools you have.

Match the investigation to the question

Scientists use several ways to find answers.

Choose a fair test when you can change one thing and keep other things the same. You cannot choose people's leg lengths. Measuring many people lets you look for a pattern instead.

Compare kinds or change an amount

A comparative test compares separate things, such as four cup-wrapping materials. A fair test can also change a number in steps, such as ramp heights of 10, 20, 30, 40, and 50 cm.

Check that your question does not bundle two changes together. “Do bigger, heavier balls bounce higher?” changes size and mass. To test mass, choose balls of the same size and keep the drop height and surface the same.

Imagine testing two cup wraps. One cup is large and wrapped in foil. The other is small and wrapped in wool. If one stays warmer, which change helped? The cup size and wrap both changed, so you cannot tell.

Give each variable a job

A variable is something that can change in a test. Your plan gives each variable one of three jobs.

The dependent variable is the result you are looking for. You do not choose how far the car rolls. You measure that distance after each release.

Make the comparison fair

A fair test changes one variable on purpose. It keeps other things that could affect the result the same. To test ramp height, use the same car and ramp each time. Put the back wheels on the same start line. Let go without pushing. Measure along the same floor from the bottom of the ramp to the front of the car.

To find controls, ask: “What else could affect my result?” Then say how you will keep each thing the same. For cup wraps, use matching cups. Measure 100 mL of water into each. Start them at the same temperature and leave them for the same time.

Some conditions are hard to match exactly, such as wind outdoors. Test in the same place and keep conditions as similar as possible. Record any change that could matter.

A control is a comparison setup

A control is a setup without the change being tested. In a fertilizer test, one group gets water without fertilizer. Compare the other groups with this control. The water amount is a controlled variable; the no-fertilizer group is the control.

You can record controlled quantities to check them. A starting-temperature column helps show that all cups began equally warm. The variable's job tells you its type, even when it appears in a table.

Repeated results help you judge cause and effect. Look for a difference that keeps appearing and is clearly larger than the spread of the repeat readings.

Before a test, write what you think will happen and why. This is your prediction. It links the change you will make to the result you expect.

Say which way the result will go

A useful pattern is “If … then … because …”. For example: “If I raise the ramp, then the car will roll farther, because it will be moving faster at the bottom.” The prediction names a change, a result, and a reason.

“The car will do better” is too vague. “Farther” means you will measure distance. “Faster” would need a time measurement. A prediction can also say there will be no change.

Your reason should come from something you know or have noticed. For example: “Longer helicopter wings will take longer to fall because bigger wings push against more air.” A guess has no reason behind it.

An idea and its expected result

A hypothesis is a testable idea that could explain something. “Fertilizer gives plants nutrients that help them grow” is a hypothesis. From it, you might predict: “Plants given fertilizer will be taller after three weeks than plants given only water.” Some classrooms also call a prediction with a reason a hypothesis.

Your prediction helps you get ready. If you expect the car to roll farther, you need enough clear floor and a long enough tape measure.

Let evidence change your mind

Suppose you predict that a heavier weight on a string will swing faster. You time 10 complete swings with 50 g, 100 g, and 200 g weights on a 50 cm string. Each takes about 14 seconds. Your results show no clear effect of mass.

That is useful learning. Keep the readings and change your idea. Changing the numbers to match a prediction would hide what happened.

Predict a value between two results

A 20 cm ramp sends a car about 150 cm. A 30 cm ramp sends it about 210 cm. What might happen at 25 cm?

1. The new height is halfway between 20 and 30 cm.
2. Halfway between 150 and 210 cm is 180 cm.
3. Predict about 180 cm, then test the 25 cm ramp.

The new value is a prediction until you measure it.

A good plan lets someone else carry out the same test. Write your question, prediction, variables, equipment, steps, and safety actions before you start.

Choose what to test

For a number such as ramp height, plan at least five values in equal steps: 10, 20, 30, 40, and 50 cm. These values cover a range from 10 to 50 cm. Testing several heights helps you see a pattern. Repeating one height checks how much its results vary; it does a different job.

For separate things, such as four wrapping materials, test each kind. Decide when to measure and how many repeats to take. For example, measure cup temperatures after exactly 20 minutes, or release the car three times at each height.

List suitable equipment with its size or type: a 5 m tape measure, one toy car, one ramp, supports, and tape for start lines.

Write actions someone can follow

A method is your set of steps. Start each step with an action and include exact amounts or positions.

1. Mark a line on the floor at the bottom of the ramp.
2. Set the top of the ramp 10 cm above the floor.
3. Put the car's back wheels on the ramp's start line.
4. Let go without pushing.
5. Measure from the floor line to the front of the stopped car.
6. Record the distance in centimeters. Repeat twice more.
7. Do the same steps at each remaining ramp height.

“Use some water” and “wait a bit” leave people guessing. Write “100 mL of water” and “wait 20 minutes.” Prepare a table with space for every value and repeat.

Check risks and try the method

A risk is a chance that something could cause harm. Match a safe action to each risk. An adult pours hot water, and cups stand on a flat table. Stand on the floor when dropping objects; keep the space below clear. Wipe up spills. Wear gloves to handle fertilizer, never taste it, and wash your hands afterward.

Do a trial run, a quick try of your method. A car that reaches the wall needs more clear floor or a smaller height range. Make that change before collecting your full results. During the test, note accidents or changes, such as a tipped cup that had to be refilled.

A measurement needs a number and a unit. “45 cm” tells you a length. “45” alone could mean a time, a mass, or something else.

Choose a tool that fits

Choose a suitable size too. A small 10 mL cylinder lets you read 8 mL more closely than a 1 L jug. An accurate measurement is close to the true value.

Read the marks carefully

Start a ruler at its 0 mark, which may be inside its end. Keep the ruler beside the length you are measuring. For rolling distance, always measure from the same start line to the same part of the car.

Count the spaces between numbered marks on a scale. If 0 to 10 has five equal spaces, each space is worth 2. The marks go 0, 2, 4, 6, 8, 10. Counting lines instead of spaces gives the wrong step.

For water in a measuring cylinder, put your eyes level with the water. Read the bottom of its curved surface. With a liquid thermometer, keep the bulb in the water. Wait for the reading to settle, then read at eye level. For a digital thermometer, read the display once it settles.

Give your stopwatch enough time

Your hand reacts a little late when you start or stop a timer. That matters a lot for a fall lasting less than 2 seconds. For a swinging weight, time 10 complete swings together. A complete swing goes there and back.

If 10 swings take 14.2 s, divide by 10:

One swing takes about 1.42 s. This reduces the effect of your timing delay on each swing.

A data logger uses sensors to take and store readings by itself. A sensor recording once a minute from 0 through 20 minutes takes 21 readings, including the start. Record hand readings straight away too. Keep the same units throughout a comparison.

The same test rarely gives exactly the same reading every time. Small changes in release or timing can change the result. Take repeat readings with the same setup to see this spread.

Find a mean for each setting

Close readings give more confidence than readings spread far apart. Three repeats at each setting are common. Find a separate mean, or average, for each setting.

A helicopter takes 1.8 s, 1.9 s, and 1.7 s to fall.

1. Add the times: $1.8+1.9+1.7=5.4$
2. Count the readings: 3.
3. Divide: $5.4÷3=1.8$

The mean is 1.8 s. Keep the unit with the answer. Round to the detail of your readings: tenths of a second here. A mean always falls between the smallest and largest readings, including those ends. A mean of 5.4 s would be too large; that is the total.

Keep surprising readings and check them

An anomalous result, or outlier, stands apart from the others. A car rolls 152 cm, 148 cm, and 95 cm from the same ramp height. Circle 95 cm, keep it in the table, and check what happened. Repeat the test if you can.

If you saw the car hit a table leg, you know that trial went wrong. Leave 95 cm out of the mean and explain why. Add 152 and 148 to get 300. Divide by the two valid readings to get 150 cm.

Including all three gives about 132 cm. The known fault pulls the mean down. If you do not know why a result differs, keep it and investigate. A reading that disagrees with your prediction can still be real.

Use several living things

Plants differ even when you care for them alike. Use several plants in each group. Four plants without fertilizer reach 14, 13, 15, and 14 cm: a total of 56 cm and a mean of 14 cm. Four with fertilizer reach 19, 20, 18, and 19 cm: a total of 76 cm and a mean of 19 cm.

Measuring one plant four times does not show how different plants grow. Repeats also cannot fix an unfair setup. Starting every ruler reading at 1 cm makes all the readings wrong in the same way.

A results table gives each reading a place. Make it before you begin, so you can record numbers as soon as you measure them.

Build the columns

Put what you change in the first column. Give each repeat its own column. Add a mean column at the end. Headings name the quantity and put its unit in parentheses. With “Wing length (cm)” at the top, write 4 in the box below, not “4 cm.”

Here are helicopter results ready to use:

Read across one row to compare repeats at one wing length. Read down the mean column to see how fall time changes with wing length. Keep the wing lengths in number order. Jumbled rows make the pattern harder to spot.

Record a comparison of materials

For a materials test, the first column holds names. Four cups start at 60 °C. Their temperatures after 20 minutes are shown below. Subtract the final temperature from the starting temperature to find each drop.

For wool, the calculation is $60-52=8$. The drop is 8 °C. The start column also lets you check that the cups began at the same temperature.

Preserve the full record

Write an odd reading down too. Circle it and add a note about anything you noticed. Keep the individual repeats when you calculate a mean. Someone reading only the mean cannot tell whether the repeats were close together.

A table helps you look up exact readings. A graph of those readings helps you see the pattern quickly. Check the headings and rows before you turn your table into a graph.

Use a bar graph, also called a bar chart, to compare separate kinds. Each wrapping material gets one bar. There is no material halfway between wool and foil.

Set up the axes

Draw a horizontal axis across the bottom and a vertical axis up the side. Name the graph “Temperature drop for different cup coverings.” Label the bottom “Cup covering” and the side “Temperature drop (°C).”

The drops are 12 °C for foil, 8 °C for wool, 10 °C for bubble wrap, and 19 °C for no cover. The largest value is 19. A vertical scale from 0 to 20 will fit every bar.

Use equal steps of 2 °C: 0, 2, 4, 6, and so on. Every grid space must mean the same amount. Starting at 0 matters because the full bar height shows the size of the drop.

Draw one bar for each kind

1. Write the four covering names along the bottom.
2. Draw the foil bar up to 12 °C.
3. Draw wool to 8 °C and bubble wrap to 10 °C.
4. Draw no cover to 19 °C, halfway between 18 and 20.

Make all bars the same width. Leave a gap between them to show they are separate kinds. The names can go in any order. You might put the bars from tallest to shortest to make comparison easier.

Read what the height means

The tallest bar shows the greatest temperature drop. That is the cup with no cover. The shortest bar is wool: it lost the least heat, shown by the smallest drop in temperature. Wool kept the water warmest of these four setups.

Compare the drops by subtracting:

The uncovered cup dropped 11 °C more than the wool-wrapped cup. Read the axis label before deciding what “tallest” means. A taller temperature-drop bar means more cooling, not warmer water.

Check the picture is fair

A scale starting at 6 °C would make an 8 °C bar look very small beside a 12 °C bar. Start this bar graph at 0. Unequal steps, such as 0, 2, 5, 10 at equal spaces, also give the wrong picture. Check both your numbers and the gaps between them.

A line graph shows how a measured result changes across numerical values. Use it when values between your test settings make sense. You could test a 7 cm wing between 6 and 8 cm. You cannot test a kind halfway between wool and foil.

Label and scale both axes

Put what you change on the horizontal x-axis. Put what you measure on the vertical y-axis. For a change over time, put time on the x-axis.

For the helicopter results, label the axes “Wing length (cm)” and “Mean time to fall (s).” Use 0 to 12 cm in steps of 2 on the x-axis. Use 0 to 4 s in steps of 0.5 on the y-axis. Each axis has its own even scale.

Plot pairs of values

For the first point, move across to 4 cm. Then move up to 1.8 s and mark a small cross. Plot the other pairs: 6 cm with 2.3 s, 8 cm with 2.8 s, 10 cm with 3.2 s, and 12 cm with 3.6 s.

Plot means when you have repeat readings. Join the points in order with ruler lines. If points follow a smooth curve, you can draw a smooth line through them. Keep an anomalous point visible but unjoined while you check it.

Describe the shape

The helicopter line rises: longer wings took longer to fall. A falling line shows that the result gets smaller as the value across the bottom increases. A flat line shows no change.

A cooling-water graph usually falls steeply at first, then more gently. A steeper fall means more cooling in the same amount of time.

Estimate between points

To estimate the fall time for a 7 cm wing:

1. Find 7 cm halfway between 6 and 8 cm.
2. Move up to the line joining those points.
3. Move across to the time axis.

The time is halfway between 2.3 and 2.8 s: about 2.55 s, or 2.6 s to the nearest tenth. This is an estimate, because you did not measure a 7 cm wing. Test it to check.

A value far beyond the tested range is less secure. The pattern could change outside your measurements. Also check that equal distances along each axis always stand for equal numerical steps.

Your conclusion answers the question you started with. Use the pattern and numbers from your results to explain what happened.

Build an answer from evidence

Suppose you asked how ramp height affects rolling distance. Your means rise from 80 cm at a 10 cm ramp to 310 cm at a 50 cm ramp.

A useful conclusion is: “Raising the ramp made the car roll farther. The mean distance rose from 80 cm at 10 cm high to 310 cm at 50 cm high. This matched my prediction. From higher up, the car moved faster at the bottom and kept rolling longer.”

The first sentence answers the question. The numbers support it. The last sentences compare the result with the prediction and give a science reason. Retelling how you released the car would describe the method, not answer the question.

Check how strong the difference is

For a fair test, check that you changed one thing and kept other important things the same. Look at the repeats. When the difference keeps appearing and is clearly bigger than their spread, you can say the change caused the result.

A pattern found by observing people works differently. Children with bigger feet often read better. Bigger feet do not make reading easier. Older children tend to have bigger feet and more reading practice. Another change can explain why two measurements go together.

“No clear difference” is useful

For different masses on a swinging string, mean times for 10 swings are 14.1 s, 14.2 s, and 14.2 s. Suppose the repeats at each mass vary by more than 0.1 s. The tiny gaps between means are smaller than that spread. Conclude that mass made no clear difference in this test.

A prediction can be wrong while the test still teaches you something. Report the pattern you found, including an unexpected one.

Keep the claim within the test

Say “Wool kept the water warmest of the four coverings we tested.” Other materials were not compared. Likewise, the ramp results describe heights from 10 to 50 cm. A much higher ramp needs a new test.

Use words such as “The results support our idea” when judging an explanation. One investigation gives evidence; further tests help you check whether the finding holds.

After a test, ask how much you trust the answer. This is evaluating your investigation. Use what actually happened to decide what worked and what needs changing.

Look back at the evidence

Check your plan and results together.

• Did another variable change, such as the drop height?
• Were repeats close together or spread apart?
• Were any readings unusual? Could you explain them?
• Did you test enough values or enough plants?
• Was a measurement difficult to take?

A clear pattern with close repeats gives more confidence. A single plant in each group gives less: the difference could come from the plants themselves. Many repeats are helpful only when the method is fair and the measurements are taken correctly.

Match an improvement to a problem

“Be more careful” does not say what to change. Name the problem first, then explain the action that would help.

You might also take five repeats instead of three. That helps you see the spread more clearly, but it will not fix a wall blocking the car.

Avoid a vague label such as “human error.” Say “We sometimes started the stopwatch late.” Someone reading that knows which part of the method needs work.

Check ideas with more evidence

Results can support an idea by agreeing with it. They can also go against it. If the evidence goes against your idea, look again at the explanation and plan a useful next test.

A helicopter test about wing length might lead to: “Does wing width affect fall time?” A ramp graph might suggest testing a 25 cm height between two earlier settings.

Compare findings with other groups too. If they follow the same method and find a similar pattern, you have more reason to trust it. If they find something different, compare the methods and conditions to work out what to check next.

Engineers make designs to solve problems. Before comparing designs, decide what success means and what limits you must follow.

Set the goal and the limits

Criteria say what a design must do. For a parachute, your chosen criterion might be the longest mean fall time from a fixed height. Constraints are limits, such as the materials, time, or money available.

Imagine square parachutes with sides of 20, 30, and 40 cm. Use plastic from the same bag, matching strings, and the same toy figure. Drop each three times from 2 m and measure its fall time. Only the canopy size changes.

The longest mean fall time shows the slowest average descent over that drop. Compare that result with your chosen criterion to select a design.

Learn where a design fails

A failure point is where or when a design stops working. A bridge test can show both how much a design holds and which part needs help.

1. Put a paper bridge across a 20 cm gap between books.
2. Add identical coins one at a time in the same place.
3. Record the last number it held before collapse.
4. Note where it bent or tore.

Suppose it repeatedly bends in the middle. That points to a place to strengthen. You might fold the paper into a zigzag to stiffen it, then test that design.

Change one feature and test again

Keep the paper type, amount of paper, gap, and loading method matched when comparing fold patterns. For another trial after collapse, make a fresh bridge to the same design. A damaged bridge would not start in the same condition.

If you change both the paper and the folds, you cannot tell which change helped. Change one feature, test, examine the evidence, and improve again. This cycle may take several tries.

Choose using the results and the criteria. A bridge that looks neat may carry fewer coins. A design that carries more coins but uses extra paper may break a materials constraint. It must meet both the goal and the limits.

Sharing your test lets other people follow your thinking and ask useful questions. You can use a report, poster, talk, slideshow, or short video.

Put the story in a clear order

A report usually follows this order:

1. Question: What did you want to find out?
2. Prediction: What did you expect, and why?
3. Equipment and method: How did you carry out the test?
4. Results: What did you record? Show the table and graph.
5. Conclusion: What answer does the evidence support?
6. Evaluation: How much do you trust it? What would you improve?

Give enough detail for someone to understand how you made the test fair. Name what you changed, measured, and kept the same. Include measurement units.

Help your audience see the evidence

For younger children, explain unfamiliar words and use a simple picture of the setup. For your class, you can use science terms you have learned together. Make graph titles, axis labels, and units large enough to read.

Point to the results as you speak. For example: “The 4 cm wings took a mean of 1.8 s to fall. The 12 cm wings took 3.6 s. The line rises as wing length increases.” Your listener can connect each number to a place on the graph.

Explain why you trust the pattern: “The repeat readings were close together.” Mention a difficulty too: “Starting and stopping the timer was tricky.” A problem is useful information for someone trying to repeat your test.

Answer questions with your record

If someone asks whether all your trials agreed, show the repeat columns. If they ask about an odd reading, show your note about it. If you did not measure something, say so and explain how you could test it next.

Listen to other groups' findings. Similar patterns give more confidence. Different patterns are a reason to compare methods and conditions. A question from your audience may help you spot a useful improvement or plan a new investigation.