Working Scientifically: Variables, Measurement and Evidence

A lab has glass, chemicals, heat and electrical equipment. Safe habits protect you and the people nearby.

Get ready

Enter when the teacher says to. Clear bags and stools from walkways. Tie back long hair and tuck in loose clothing. Wear safety goggles for work with chemicals, heating or glass. Keep them on until everyone has finished. A splash can travel from another bench. Use gloves or a lab coat when required.

Find the eyewash, fire blanket, fire extinguisher, shut-off switches and first-aid kit. Report spills, accidents and broken equipment at once. If chemicals reach skin or eyes, rinse with plenty of running water and tell the teacher immediately.

Handle materials

Never eat, drink or taste anything in the lab. Read each chemical label. Take the amount needed and leave unused chemicals out of the stock bottle. Returning them could contaminate its contents. Smell a chemical only when instructed: gently waft vapor toward your nose. Use the waste container provided.

Check glass for chips and cracks. Keep it back from bench edges. Report broken glass; leave the cleanup to the teacher. Hot glass looks like cold glass, so allow heated glass to cool. Keep water away from electrical equipment. Use the low-voltage supply provided and switch off before changing a circuit.

Light and use a Bunsen burner

The barrel is the upright tube. Gas enters through the hose near its base. Turning the collar opens or closes the air hole.

1. Put the burner on a heat-resistant mat and clear nearby paper.
2. Close the air hole.
3. Light the splint or lighter before turning on the gas.
4. Turn on the gas and bring the lit source just above the barrel.

Having a flame ready prevents unburned gas from collecting. The closed air hole gives a visible yellow safety flame. Use it while the burner is waiting. Half-opening the hole gives a blue heating flame. Fully opening it gives a noisy blue roaring flame. Its hottest part is just above the pale inner cone. Blue flames are hard to see, so return to yellow when heating stops. Turn off the gas at the tap when finished.

Heat a test tube

Stand while heating so you can move away quickly. Put liquid in only about one-third of the tube. Hold it at an angle with a test-tube holder. Point its opening away from everyone. Move it gently through the flame so one spot does not overheat and spit. Use tongs for hot objects.

At the end, turn off gas and electricity. Let equipment cool, clean the bench and wash your hands.

A hazard can cause harm. A risk includes how likely that harm is and how serious it could be. A hot beaker is a hazard; touching it could cause a burn.

Recognize the nine symbols

GHS hazard pictograms have black pictures on white backgrounds inside red diamonds. GHS is an international system for chemical labels. Learn the picture and its meaning together.

The marked-chest symbol also covers lung damage when a liquid enters the airways. Some serious effects can follow one exposure; others develop over time. A symbol identifies a hazard type. The label explains which hazards apply.

Read the whole label

Look for the product name, pictograms, supplier and signal word. “Danger” marks more severe hazards; “Warning” marks less severe ones. Hazard statements, or H statements, describe harm. Precautionary statements, or P statements, explain protective actions. A Safety Data Sheet, or SDS, has fuller details in 16 sections, including first aid, storage and disposal.

Read the label on the bottle you will use. Different concentrations can have different hazards. Concentrated acid may be corrosive, while a much more dilute solution may be irritating. A substance without a pictogram can still harm you: hot water scalds.

Make a risk assessment

Before starting, list the hazards, possible harms and controls. Include what to do if something goes wrong.

For heating 50 mL of water, one row could be:

Other rows would cover the flame, hot water and breakable glass. “Be careful” does not say what action prevents harm.

First remove or replace a hazard where possible. A water bath can replace a flame. A less hazardous chemical or lower concentration can reduce the hazard. Smaller amounts, safe procedures and suitable equipment lower risk. Goggles and gloves provide a final layer of protection.

Protect living things too. Handle animals gently and return them where found. Keep chemical waste in its proper container, away from drains. Wash your hands after handling soil, plants or microorganisms.

A measurement needs a number and a unit. “12 cm” gives a length. The number “12” alone leaves its meaning unclear. Length, time and mass are examples of quantities: things we measure.

Share the same units

Scientists use the International System of Units, or SI. Common base units are meter (m) for length, kilogram (kg) for mass, second (s) for time, kelvin (K) for temperature and ampere (A) for electric current. The mole and candela complete the seven base units.

School thermometers usually show degrees Celsius (°C). A change of 1 °C has the same size as a change of 1 K. Their zero points differ: 0 °C equals 273.15 K.

Some units combine base units. Area uses m², volume uses m³, and speed uses m/s. Density can use kg/m³ or g/cm³. Force uses newtons (N), and energy uses joules (J). Liters and milliliters are accepted for use with SI:

• 1 L = 1000 mL.
• 1 mL = 1 cm³.
• 1 L = 1000 cm³.

Use prefixes

A prefix changes a unit's size.

So 1 m = 100 cm = 1000 mm, and 1 mm = 1000 µm. Also, 1 kg = 1000 g and 1 g = 1000 mg. Mega means one million times; nano means one billionth.

Convert one step at a time

Smaller units need a larger number for the same amount. To change 2.5 kg to grams, multiply by 1000: 2.5 × 1000 = 2500 g. To change 450 mL to liters, divide by 1000: 450 ÷ 1000 = 0.45 L. To change 35 mm to centimeters, divide by 10: 35 ÷ 10 = 3.5 cm.

Before adding measurements, put them in the same unit. For 1.2 m plus 30 cm, first change 1.2 m to 120 cm. Then 120 + 30 = 150 cm.

Keep symbols clear

Write 5 kg, with a space before the unit. Unit symbols have no plural ending or final period. Use g and s, not “gm” and “sec”. Capitals matter: m means meter, while M means mega. Choose units that give useful numbers, such as grams for a paper clip and kilometers between towns.

Mass is the amount of matter, measured with a balance in g or kg. Weight is gravity's pulling force, measured with a newton meter or spring scale in N. On Earth, a mass of 1 kg has a weight of about 9.8 N.

Choose an instrument by asking three questions: What quantity am I measuring? Will the value fit its range? Can it show the differences I need?

Match instrument and quantity

A graduated cylinder is also called a measuring cylinder. Beakers and flasks have rough volume marks. Use them to hold liquids; choose a cylinder, syringe or pipette to measure the volume.

Check range and resolution

The range runs from the lowest value to the highest value the instrument measures. A 100 mL cylinder cannot hold a 250 mL measurement in one go.

The resolution is the smallest change the instrument shows. A ruler with millimeter marks shows changes of 1 mm. A balance displaying 12.47 g has a resolution of 0.01 g. Each change in its last digit represents 0.01 g.

Suppose you need 25 mL of water. A 50 mL cylinder with 1 mL marks holds it and gives useful detail. A 250 mL cylinder with 2 mL marks holds it too, but its scale is coarser. Pick the smallest suitable cylinder that holds the whole volume.

A seedling grows about 2 mm in one day. Centimeter marks are too far apart to show that small change clearly. A millimeter ruler is a better match. The biggest instrument is not automatically the best choice.

Think about how readings are taken

A toy car takes about 1.5 s to travel down a ramp. A hand-operated stopwatch may show hundredths of a second. But you still need to see the car move and press the button. Human reaction delays are about 0.2 s and vary between trials.

Starting late makes the recorded interval shorter. Stopping late makes it longer. These delays rarely match exactly, so times can vary in either direction. A light gate detects the car electronically and removes the operator's reaction time. A longer journey also makes a small delay matter less compared with the total time.

A data logger records readings from sensors automatically. It can collect rapid measurements or monitor temperature overnight. Sensors still need correct setup and checks. Extra display digits cannot fix a poorly adjusted instrument.

A suitable instrument only helps when you read it correctly. Check its scale and zero before taking a measurement.

Find the value of one division

Find two numbered marks. Subtract the smaller value from the larger one. Divide that difference by the number of equal spaces between them.

For marks at 10 mL and 20 mL with five equal spaces:

1. The difference is 20 − 10 = 10 mL.
2. Each space is 10 ÷ 5 = 2 mL.
3. Three spaces above 10 mL gives 10 + 6 = 16 mL.

Count spaces, not the lines that bound them. Six lines make five spaces.

Put your eye in the right place

Look straight at a scale with your eye level with the reading. From above or below, the pointer seems to shift against the marks. This is parallax error.

Water curves at its surface inside a narrow tube. This curve is the meniscus. Stand a graduated cylinder on a flat bench. Bring your eye level with the water and read the bottom of the curve. If marks at 20 and 30 mL contain ten spaces, each space is 1 mL. A bottom three spaces above 20 mL means 23 mL.

For temperature, immerse the thermometer bulb without touching the container's bottom or sides. Stir gently and wait for a steady reading. Read while the bulb is still in the liquid. Lifting it out makes it start measuring the air.

Check the starting point

Line up a ruler's zero mark with the object's end. The ruler's physical edge may have a blank margin or be worn. Keep the ruler flat along the object. For a curved line, lay string along it and then measure the string.

An instrument should read zero when measuring nothing. A balance can include the container's mass unless you tare it:

1. Place the empty container on the balance.
2. Press “tare” or “zero”.
3. Add the material and read its mass.

Hang a newton meter vertically and adjust it to 0 N before adding a load. A wrong starting reading is a zero error.

Record what the instrument shows

Write a balance display of 12.40 g as 12.40 g. The final zero shows that it reads in 0.01 g steps. A millimeter ruler could give 45 mm, or 4.5 cm. It cannot justify writing 4.5372 cm.

Keep readings of the same quantity in the same units and with the same decimal places. Put units in a results-table heading so each entry has a clear meaning.

An investigation begins with something noticed. Sugar seems to disappear faster in hot tea than iced tea. That observation can lead to a question we can answer with evidence.

Make the question testable

A testable question can be answered by observations or measurements. “Is hot water better?” is too vague. “How does water temperature affect the time 5 g of sugar takes to dissolve?” says what to change and measure.

Not every question needs a fair test. “Which rocks are igneous?” can be answered by observing and classifying. A question about which drink tastes nicer asks for a personal preference, not a scientific explanation.

Explain, then predict

A hypothesis is an explanation we can test. It gives a scientific reason. For example: “Sugar dissolves faster in hotter water because the water particles move faster and hit the sugar more often.”

A prediction states an expected result of a particular test. For example: “Raising the water temperature from 20 °C to 60 °C will reduce the time for 5 g of sugar to dissolve.” This names the variables and the direction of change. It does not need an exact time to be checkable.

You can combine the two: “If the water is hotter, then the dissolving time will decrease, because faster-moving water particles hit the sugar more often.” The expected change is the prediction. The explanation supplies the reasoning.

An aim states what you intend to find out: “To find how water temperature affects dissolving time.” Keep the original prediction after collecting results. Results that disagree with it are useful evidence. They show that the explanation needs checking or changing. One investigation cannot prove it true everywhere.

Choose how to collect evidence

Choose the approach that answers the question. You cannot control every difference between people, so a study of leg length and jumping is pattern seeking.

A variable is a factor that can change. Naming variables clearly makes a plan easier to follow and a comparison fairer.

Identify the three roles

In a fair test, the independent variable is changed on purpose. The dependent variable is measured to find its effect. Control variables are other factors that could affect the result. Keep these the same.

For the question “How does water temperature affect sugar's dissolving time?”:

Suppose hot water is stirred but cool water is not. A shorter time could result from heat, stirring or both. Keeping stirring the same lets the comparison test temperature.

Name measurable quantities. “The water” is unclear: its temperature and volume are different variables. Write “water volume in mL”. The dependent variable is measured, so it cannot also be held fixed to hide any change.

Recognize variable types

Categorical variables have names as values, such as soil type or light color. Discrete variables are counts in whole numbers, such as how many seeds sprout. Continuous variables are measured and can take values throughout a range, such as length or temperature.

A temperature can fall between 20 and 21 °C. A count of sprouted seeds cannot fall between 8 and 9 seeds. This difference helps decide how to display data. Categories use separate bars; continuous changes can use a line graph.

Set up a control group

A control group provides a comparison without the factor being tested. To test fertilizer, give fertilizer to one set of plants and none to the other. Use matching plant types, light, water, soil and pot sizes.

The no-fertilizer plants are the control group. The shared water volume is a control variable. These names describe different parts of the investigation.

Living things differ even when treated alike. Use many plants or animals so one unusual individual has less influence. Some conditions, such as outdoor weather, are hard to control. Record them and consider how they could affect the pattern. When conditions cannot be controlled, pattern seeking may fit the question better than fair testing.

A complete plan includes the question, hypothesis, prediction, variables, equipment, steps, blank results table and risk assessment.

Choose values and repeats

The range runs from the lowest to the highest value tested. The interval is the gap between neighboring test values. Temperatures of 20, 30, 40, 50 and 60 °C give a range of 20–60 °C and an interval of 10 °C.

For a continuous independent variable, choose at least five values to reveal the pattern. Keep the range safe and practical. Take at least three repeat measurements at each value. Repeats show how much the readings vary and help reveal unusual results. They also allow a mean. With living things, include many individuals: 20 seeds give more useful evidence than two.

A short trial run checks that the plan works. It can reveal that a process takes too long or that the scale is too coarse. Adjust the plan before collecting the main results.

Make the steps exact

Use numbered steps in the order they happen. Begin with action words such as “Measure”, “Add” and “Record”. Give amounts, equipment and a clear stopping point.

A sugar investigation could use these steps:

1. Bring water to 20 °C in a water bath.
2. Measure 100 mL into a 250 mL beaker. Check its temperature.
3. Measure 5.0 g of granulated sugar with a balance.
4. Add the sugar, start the timer and stir steadily.
5. Stop when no sugar crystals remain visible. Record the time.
6. Carry out three trials at each temperature: 20, 30, 40, 50 and 60 °C.

Use fresh water and sugar for each trial. Keep beaker size, sugar type and stirring rate the same. Five temperatures with three trials each give 15 measured times. “Add some sugar” would leave its mass uncertain and make the method hard to repeat.

Check validity and care

A valid investigation answers its intended question. It measures the right outcome and keeps other relevant factors controlled. Times could agree closely even if water volume changed at every temperature. That would still be an unfair comparison.

A clear method makes repetition possible. Similar results from someone else following it provide evidence of reproducibility.

Include risks before practical work begins. Investigations must not harm living things or cause distress. Get people's agreement before surveys and keep their answers private. Follow the teacher's rules about organisms and materials.

Record results as soon as you collect them. A prepared table keeps the changed values and repeat measurements together.

Know what kind of data you have

Qualitative data are descriptions, such as “The solution turned blue.” Quantitative data are measurements or counts, such as “The temperature rose by 12 °C” or “18 seeds sprouted.” A count is quantitative even without a measurement unit.

Primary data are collected by you or your group. Secondary data come from someone else, such as another class or a science database. These labels describe where data came from. A website's temperature measurements are both secondary and quantitative.

Put units in the headings

Place the independent variable first, with its values in order. Follow it with the dependent variable's repeats and mean. Each measurement heading names the quantity and unit. “Temperature (°C)” and “Temperature / °C” both work. The entries then need numbers only.

Here are dissolving times for 5 g of sugar:

The shared quantity for the trial columns is time to dissolve. A table of observations can instead have headings such as “Color before”, “Color after” and “Bubbles observed”. Use short, exact descriptions.

Calculate the mean

The mean is the total divided by the number of readings. At 40 °C:

1. Add the times: 52 + 48 + 50 = 150 s.
2. Count the readings: three.
3. Divide: 150 ÷ 3 = 50 s.

Use the number of readings included, not the number originally planned. Keep original readings in the table even when a confirmed mistake is left out of a mean.

The range of repeats is largest minus smallest. At 20 °C, 102 − 98 = 4 s. A small range shows close agreement. A mean alone does not show whether readings were close together.

Record each column with consistent decimal places. A thermometer reading in tenths gives 12.0, 15.5 and 18.0 °C. Keep means to the same decimal places as the readings. A calculator's extra digits do not add measurement detail.

A graph makes a pattern easier to see. First check the kinds of variables, then choose a graph and set up its axes.

Match the graph to the question

For four soils, 18, 15, 12 and 9 seeds sprout out of 20. Draw separate bars labeled with the soil names. Gaps show that the soils are separate categories. Joining the tops would suggest meaningful values between soil types.

For temperature and dissolving time, both variables are continuous. Plot the results on a line graph. For height and arm span, each person supplies a pair of measurements. Use a scatter graph, with one point per person.

A histogram groups a continuous variable into intervals and shows how often values fall in each. Class heights grouped into equal 5 cm intervals are an example. Its bars touch because the intervals follow on without gaps.

Set up the axes

Put the independent variable on the horizontal x-axis. Put the dependent variable on the vertical y-axis. For sugar, use “Temperature (°C)” and “Mean time to dissolve (s)”. A useful title names both quantities.

Choose equal scale steps that are easy to count: 1, 2, 5 or 10 units per square. Use enough of the page for the points to fill more than half the grid in both directions. A scale of 0, 5, 20, 100 in equally spaced marks is uneven and changes the apparent pattern.

A bar chart's numerical axis starts at zero so bar lengths compare fairly. A line or scatter graph may start above zero if its starting value is clearly numbered.

Plot and check

For the sugar data, a scale from 0 to 60 °C horizontally and 0 to 100 s vertically works. Use 10 °C and 10 s per major division. The first point is at 20 °C and 100 s. The next is at 30 °C and 72 s.

Plot every mean as a small dot or cross. Check each against its table row. The smooth line or curve you add should show the overall pattern; it is not a way to hide misplaced points.

Experimental points rarely form a perfect line. A line of best fit shows their overall trend, using one straight line or a smooth curve.

Draw the overall pattern

Use a ruler when points lie roughly in a straight line. Draw a smooth curve when they bend. Keep roughly equal numbers of ordinary points on either side. Do not join experimental points dot to dot. Small reading differences should not become sharp bends.

A best-fit line need not pass through the first point, last point or origin. Use the origin only when it is a measured or known point that belongs to the pattern. A point far away may be anomalous: check it and repeat that measurement if possible.

Describe the relationship

A positive relationship means both variables increase together. A negative relationship means one decreases as the other increases. No clear trend means there is no clear relationship in the data.

For sugar, write: “As water temperature increases, the mean dissolving time decreases.” Add numbers: “It falls from 100 s at 20 °C to 30 s at 60 °C.” The fall is 28 s between 20 and 30 °C, but only 6 s between 50 and 60 °C. These differences show the curve leveling off.

On a scatter graph, points close to a best-fit line show a stronger correlation than widely spread points. Correlation describes a link. It does not establish its cause.

Test for direct proportion

A spring has these force and extension pairs: 0 N and 0.0 cm; 1 N and 2.0 cm; 2 N and 4.1 cm; 3 N and 5.9 cm; 4 N and 8.0 cm; 5 N and 10.1 cm.

The points lie close to a straight line through the origin. In this tested range, extension is directly proportional to force. Doubling force roughly doubles extension. A straight line that starts above zero would not show direct proportion.

Estimate inside or outside the range

Interpolation estimates between measured values. On the spring's best-fit line, 2.5 N gives about 5.0 cm. Read from 2.5 N up to the line, then across to the extension axis.

Extrapolation estimates beyond the tested range by extending the line. It is less dependable because the pattern may change. Stretching a spring far enough can stop it following the straight-line pattern. An estimate outside the measured range has less support than one between nearby measurements.

Readings can agree with one another and still be wrong. Checking data quality means asking both how close they are and what they are close to.

Compare with a reference

The true value is what a perfect measurement would give. We usually cannot know it exactly. Instead, we compare with an accepted value, a value supported by careful measurements.

A reading is accurate when it is close to the true value. A set of readings is precise when they are close to one another. Their range is small. Think of darts: a tight cluster away from the bullseye is precise but misses the target.

Three thermometers measure pure boiling water at standard atmospheric pressure. Its accepted temperature is 100 °C.

A is precise, but its readings are about 2 °C too low. B has an accurate mean, but its readings are more spread out. C combines a small spread with a mean at the accepted value. Taking a mean helped B's random variation average out. It did not correct A's low readings.

Repeat or reproduce

Results are repeatable when the same person gets similar results with the same method and equipment. Results are reproducible when a different person, method or set of equipment gives similar results.

Suppose you measure the same object several times. Similar values show repeatability. A second group gets similar values using its own equipment. That adds evidence of reproducibility and increases confidence in the result.

Keep resolution separate

Resolution describes an instrument: the smallest change it shows. Precision describes a set of readings: how closely they agree. A balance showing hundredths of a gram has fine resolution, but a wrong zero can shift every reading.

The word “reliable” is sometimes used for dependable results. Say exactly what has been checked: close repeats, agreement between groups or closeness to an accepted value. These are different kinds of evidence.

Measurements have uncertainty: we cannot give a perfectly exact value. A measurement error is a difference from the true value. It can occur even when nobody makes a mistake.

Match the response to the error

Reaction delays in timing vary between trials. Judging when the last sugar crystal disappears can also vary. These random effects may make readings too high or too low. Averaging repeats reduces their effect.

A balance reading 0.3 g when empty has a systematic zero error. Every displayed mass is 0.3 g too high. A 5.6 g display needs the correction 5.6 − 0.3 = 5.3 g. A 10.4 g display becomes 10.1 g. Repeating the wrong reading does not remove the added 0.3 g. Reset the zero before measuring again.

Investigate a possible anomaly

An anomalous result, or outlier, does not fit the pattern. It may lie far from other repeats or a best-fit line.

Dissolving times of 52, 74 and 48 s make 74 s worth checking. Look for a cause and repeat the measurement. If a recorded mistake explains it, leave that result out of the mean. Keep it visible in the table, mark it and explain the decision.

For a confirmed mistake at 74 s, the remaining mean is (52 + 48) ÷ 2 = 50 s. Using all three gives 58 s. The choice affects the reported result, so it needs a reason. Disagreeing with a prediction is not a reason to remove a reading. If repeats keep giving the unexpected value, the effect may be real.

Use spread to estimate uncertainty

One simple estimate is half the range of repeats. For 98, 102 and 100 s:

1. Mean = 300 ÷ 3 = 100 s.
2. Range = 102 − 98 = 4 s.
3. Half the range = 4 ÷ 2 = 2 s.

Write the result as 100 ± 2 s. The ± sign reads “plus or minus”. This describes uncertainty estimated from the spread. It does not correct a wrong zero or guarantee where the true value lies.

A conclusion answers the investigation question. It explains what the results show. A list of equipment or a retelling of the method does not answer that question.

Build claim, evidence and reasoning

Use claim, evidence, reasoning, often shortened to CER.

• The claim answers the question in one sentence.
• The evidence gives specific observations or measurements, with units.
• The reasoning uses scientific ideas to explain why the evidence supports the claim.

For the sugar investigation:

Claim: “Sugar dissolves faster in hotter water.”

Evidence: “The mean time fell from 100 s at 20 °C to 30 s at 60 °C. It decreased at every temperature step.”

Reasoning: “Water particles move faster at higher temperatures. They hit the sugar more often and with more energy, carrying its particles away into the water more quickly.”

The evidence tells us what changed. The reasoning explains the change. “Because the water is hot” names a condition but does not explain its effect.

Compare with the prediction

State whether the results support, partly support or do not support the prediction. Give the evidence that led to that decision. For sugar: “The results support the prediction that higher temperatures reduce dissolving time.”

A result that disagrees with the prediction still answers a scientific question. Keep the original prediction and explain the disagreement. Results support an explanation; one test does not prove it true in every case.

Keep the claim within the investigation's reach. The sugar measurements cover 20–60 °C and one type of sugar. They do not establish times at 90 °C or results for salt. Those questions need further evidence.

Describe observations and links

Evidence can be qualitative. A conclusion can describe a consistent color change or the presence of starch in leaves. Compare the observations between conditions and explain their meaning. Results from other groups or secondary sources can show whether a finding is reproduced.

A correlation is a link between variables. It does not by itself show causation, where one factor produces a change in another. Ice cream sales and sunburn both rise in summer. Hot sunny weather helps explain both; buying ice cream does not cause sunburn.

A fair test controls other relevant factors and gives stronger evidence of cause. A conclusion can also lead to a next question: after testing temperature, you might investigate sugar grain size.

An evaluation judges how much confidence to place in a result. It looks at the method and data, then explains useful changes.

Inspect the evidence

Start with the investigation's question. Did the measured outcome answer it? Were control variables kept the same? Then look at the results:

• How close were the repeats?
• Were unusual readings checked?
• Was the instrument's resolution fine enough?
• Were enough values tested over a useful range?
• Did other groups obtain similar results?
• Could random or systematic errors affect the finding?

Use details from the investigation. “The repeats differed by up to 0.3 s” gives more information than “The results were bad.” Close repeats increase confidence, but they cannot rescue an unfair comparison.

Link problem, effect and change

A useful improvement has three connected parts:

1. Problem: what went wrong or limited the method.
2. Effect: how it could change the results.
3. Change: a specific action that tackles it.

For a dissolving investigation: “The water cooled during each trial. Its changing temperature could alter the dissolving time. Keep the beaker in a water bath set to the test temperature.” The water bath addresses temperature control.

For a toy car: “The journey took about 1.5 s and was timed by hand. Different reaction delays spread the repeat times by up to 0.3 s. Use light gates and a data logger to remove the operator's reaction time.” More stopwatch digits would not remove the button-press delay.

“Be more careful” leaves the action unclear. “Use better equipment” leaves its purpose unclear. Name the change and explain the link to the weakness.

Strengthen or extend the test

More repeats help assess variation. More independent-variable values show the pattern in greater detail. If the graph changes quickly between two values, test extra values there.

Check validity too. If water volumes differed between temperatures, measure the same volume each time. Even perfectly matching repeat times would not remove that unfairness.

You can extend a question with a wider safe range or another material. Keep those new results separate from what the original test established. Evaluation feeds the next plan: ask, plan, measure, analyze, conclude, evaluate, then improve the investigation.

A scientific claim should have evidence that other people can check. Science-sounding words or a person in a lab coat do not provide that evidence.

Ask what supports the claim

An advertisement says, “Students who drink FocusFizz score 30% higher on tests!” Before accepting it, ask:

• How many students took part, and how were they chosen?
• Was there a fair comparison with students drinking water?
• Did both groups take the same tests under the same conditions?
• Who paid for the work, and who benefits from selling the drink?
• Have other investigators found similar results?

Ten drink users with no control group provide weak evidence. Differences in teaching, preparation or other conditions could explain their scores. A larger group helps with variation, but a large, unfair comparison can still mislead.

An anecdote is a personal story, such as “It worked for my cousin.” The story may be true. One case, however, cannot establish a general effect because other factors may explain it.

Check the source

Prefer sources that name their authors and show their evidence. Science organizations, universities, government science agencies and textbooks are useful starting points. Check the date and compare the claim with another source. Trace a news report back to the research it describes when possible.

In peer review, experts in the same field check a scientific report before journal publication. They examine the method, results and conclusions. This can catch errors or weak arguments. It does not guarantee that every finding is correct. Independent testing can strengthen confidence or reveal a problem that reviewers missed.

Let evidence change explanations

Scientific ideas change when new evidence does not fit. Around 1900, one atom model pictured electrons inside a ball of positive charge. Experiments firing particles at thin gold foil showed a tiny central nucleus containing most of the atom's mass. The model changed to fit the evidence.

Pseudoscience sounds scientific but avoids testable claims, ignores opposing evidence or relies on testimonials instead of controlled tests. Compare how a claim is checked, not how impressive it sounds. A good explanation must face evidence that could show it is wrong.