Scientific Inquiry & Experimental Design in Biology
Learn to design and evaluate biology experiments: independent, dependent and controlled variables, control vs. experimental groups, if/then hypotheses, and when data support causation.
What you'll do in this lesson
A voice-first session with the Crimsora tutor on Scientific Inquiry & Experimental Design in Biology, then targeted practice and FRQs — with the tutor adapting to where you get stuck.
What this lesson covers
Biology only counts as a science because its claims can be checked. When a classmate says "listening to music helps plants grow," the interesting question is not whether it sounds true — it's how you would find out. This lesson gives you the toolkit: how to turn a curious observation into a hypothesis you can actually test, how to sort the pieces of an experiment into independent, dependent, and controlled variables, and how a control group lets you claim that your treatment (and nothing else) caused the difference you measured.
You will also practice the harder half of the objective: judging other people's investigations. Many experiments you meet in class, in the news, and later in this course are technically completed but not convincing — one plant per group, two things changed at once, or a conclusion about cause drawn from data that only show a pattern. Learning to spot those flaws is exactly the skill you'll use for every lab report you write this year.
You will also practice the harder half of the objective: judging other people's investigations. Many experiments you meet in class, in the news, and later in this course are technically completed but not convincing — one plant per group, two things changed at once, or a conclusion about cause drawn from data that only show a pattern. Learning to spot those flaws is exactly the skill you'll use for every lab report you write this year.
Turning an Observation into a Testable Hypothesis
Scientific inquiry starts with an observation and a question, not with a guess. You notice that mold grows faster on bread left on the counter than on bread in the refrigerator. The question follows: does temperature affect the growth rate of bread mold?
A hypothesis is a proposed, testable explanation or prediction — not a restatement of the question and not a wish. The most useful classroom format is if/then, because it forces you to name both variables at once: "If bread mold grows faster at warmer temperatures, then slices stored at 25 degrees Celsius will develop a larger area of mold in 5 days than slices stored at 4 degrees Celsius."
Three things make a hypothesis testable. It must be falsifiable — there has to be a possible result that would show it wrong. It must name something you can measure with the tools you have. And it must be specific about direction (more, faster, larger), because "temperature will affect mold" is true no matter what happens and therefore teaches you nothing.
Where students go wrong: writing "Mold likes warm places" (not measurable), "I think the warm bread will grow more mold because I saw it once" (an anecdote, not a prediction), or a hypothesis about something untestable, such as whether mold prefers being on the counter. Also avoid the word "prove." A single experiment can support or fail to support a hypothesis; it cannot prove it. That is why scientists repeat and revise rather than declare things finished.
Finally, a hypothesis should rest on prior knowledge. Knowing from earlier lessons that enzyme-driven reactions speed up with temperature makes the mold prediction reasonable rather than random, and reasoning like that is what your teacher looks for in the introduction of a lab report.
A hypothesis is a proposed, testable explanation or prediction — not a restatement of the question and not a wish. The most useful classroom format is if/then, because it forces you to name both variables at once: "If bread mold grows faster at warmer temperatures, then slices stored at 25 degrees Celsius will develop a larger area of mold in 5 days than slices stored at 4 degrees Celsius."
Three things make a hypothesis testable. It must be falsifiable — there has to be a possible result that would show it wrong. It must name something you can measure with the tools you have. And it must be specific about direction (more, faster, larger), because "temperature will affect mold" is true no matter what happens and therefore teaches you nothing.
Where students go wrong: writing "Mold likes warm places" (not measurable), "I think the warm bread will grow more mold because I saw it once" (an anecdote, not a prediction), or a hypothesis about something untestable, such as whether mold prefers being on the counter. Also avoid the word "prove." A single experiment can support or fail to support a hypothesis; it cannot prove it. That is why scientists repeat and revise rather than declare things finished.
Finally, a hypothesis should rest on prior knowledge. Knowing from earlier lessons that enzyme-driven reactions speed up with temperature makes the mold prediction reasonable rather than random, and reasoning like that is what your teacher looks for in the introduction of a lab report.
The Three Kinds of Variables
Every experiment has exactly one thing you deliberately change, one thing you measure, and a long list of things you refuse to let change.
A memory hook: the dependent variable depends on the independent one. If you can say "I set the temperature and then read the mold area," you have them in the right order.
Controlled variables are where real labs succeed or fail. Suppose the warm slices sat by a sunny window and the cold slices sat in a closed refrigerator. Now temperature and light both differ, so if mold grows faster in the warm group you cannot tell which factor mattered. A variable that differs along with your independent variable and could also explain the result is a confounding variable, and one confounding variable is enough to sink a conclusion.
Two frequent mix-ups: students label the units or the measuring tool as the dependent variable (the ruler is not the variable — mold area is), and they list the control group as a controlled variable. Those are different ideas; the next section separates them. Also remember that time is usually not the independent variable in a simple design. If you measure mold every day at one temperature, you are describing a growth curve, not comparing treatments. To test temperature, temperature must be the thing that differs between groups.
| Variable | Definition | Mold example |
|---|---|---|
| Independent (manipulated) | What the experimenter changes on purpose | Storage temperature: 4, 15, or 25 degrees Celsius |
| Dependent (responding) | What is measured as the outcome | Area of mold, in square centimeters, after 5 days |
| Controlled (constant) | Everything held identical across groups | Bread brand, slice size, moisture, light, container, day of setup |
Controlled variables are where real labs succeed or fail. Suppose the warm slices sat by a sunny window and the cold slices sat in a closed refrigerator. Now temperature and light both differ, so if mold grows faster in the warm group you cannot tell which factor mattered. A variable that differs along with your independent variable and could also explain the result is a confounding variable, and one confounding variable is enough to sink a conclusion.
Two frequent mix-ups: students label the units or the measuring tool as the dependent variable (the ruler is not the variable — mold area is), and they list the control group as a controlled variable. Those are different ideas; the next section separates them. Also remember that time is usually not the independent variable in a simple design. If you measure mold every day at one temperature, you are describing a growth curve, not comparing treatments. To test temperature, temperature must be the thing that differs between groups.
Control Group Versus Experimental Group
A control group receives no treatment, or the standard treatment, and provides the baseline you compare against. An experimental group receives the level of the independent variable you are testing. Without a baseline, a result is uninterpretable: if 30 percent of treated seeds germinate, is that good? You cannot know until you see what untreated seeds did.
Controls come in two flavors worth naming. A negative control is expected to show no effect (untreated seeds, or a Petri dish with no antibiotic disk) and confirms that your setup isn't producing effects on its own. A positive control is expected to show a known effect (a disk with an antibiotic already proven to work) and confirms your procedure can detect an effect when one exists. If your positive control fails, the whole run is suspect no matter what the treated groups did.
In studies with living subjects, a placebo control matters because expectation changes behavior and even physiology. Give one group the real supplement and the other an identical-looking pill with no active ingredient, and ideally keep both the subjects and the researchers unaware of who got which — a double-blind design — so the person recording data cannot nudge the numbers.
The classic student error is to say an experiment "has no control group" when it actually compares several treatment levels. Comparing 0, 5, and 10 milliliters of fertilizer does have a control: the 0-milliliter group. The 0 level is the baseline. A genuinely missing control looks like this: every plant gets fertilizer, at different amounts, with nothing at zero — so you can compare doses to each other but never to normal growth.
A second error is treating the control group as "the group we do nothing to," including care. The control group must be treated identically in every way except the independent variable: same water, same soil, same light, same handling.
Controls come in two flavors worth naming. A negative control is expected to show no effect (untreated seeds, or a Petri dish with no antibiotic disk) and confirms that your setup isn't producing effects on its own. A positive control is expected to show a known effect (a disk with an antibiotic already proven to work) and confirms your procedure can detect an effect when one exists. If your positive control fails, the whole run is suspect no matter what the treated groups did.
In studies with living subjects, a placebo control matters because expectation changes behavior and even physiology. Give one group the real supplement and the other an identical-looking pill with no active ingredient, and ideally keep both the subjects and the researchers unaware of who got which — a double-blind design — so the person recording data cannot nudge the numbers.
The classic student error is to say an experiment "has no control group" when it actually compares several treatment levels. Comparing 0, 5, and 10 milliliters of fertilizer does have a control: the 0-milliliter group. The 0 level is the baseline. A genuinely missing control looks like this: every plant gets fertilizer, at different amounts, with nothing at zero — so you can compare doses to each other but never to normal growth.
A second error is treating the control group as "the group we do nothing to," including care. The control group must be treated identically in every way except the independent variable: same water, same soil, same light, same handling.
Sample Size, Replication, and the Limits of a Causal Claim
One plant, one mouse, or one bread slice per group is a story, not evidence. Individual organisms vary, so a difference between two individuals may be nothing but natural variation. Increasing sample size () averages that variation out, so a real difference between group means can stand out from the noise. Twenty seeds per group () makes a much stronger case than , and repeating the entire experiment on a different day — replication — checks that the result wasn't a fluke of that particular setup.
How big is big enough? There is no magic number, but ask whether the difference between groups is larger than the spread of values within a group. If treated plants averaged 12 centimeters and controls averaged 11 centimeters, while individual plants in each group ranged from 6 to 17 centimeters, the group difference is buried in the variation and the honest conclusion is "no clear effect."
Design type also limits what you may claim. A controlled experiment with random assignment can support a causal claim, because randomizing spreads unmeasured differences evenly across groups. An observational study — surveying people who already chose to drink coffee, or comparing wild populations — can show correlation only, because the groups may differ in dozens of unmeasured ways.
Where students go wrong: writing "this proves caffeine causes faster heart rate" from a survey, or from an experiment with three subjects. Match the strength of your wording to the strength of your design — "the data support," "suggest," or "are consistent with" are the phrases of a careful biologist.
How big is big enough? There is no magic number, but ask whether the difference between groups is larger than the spread of values within a group. If treated plants averaged 12 centimeters and controls averaged 11 centimeters, while individual plants in each group ranged from 6 to 17 centimeters, the group difference is buried in the variation and the honest conclusion is "no clear effect."
Design type also limits what you may claim. A controlled experiment with random assignment can support a causal claim, because randomizing spreads unmeasured differences evenly across groups. An observational study — surveying people who already chose to drink coffee, or comparing wild populations — can show correlation only, because the groups may differ in dozens of unmeasured ways.
| Design feature | What you may conclude |
|---|---|
| Random assignment, control group, large | The treatment likely caused the difference |
| Control group but tiny | Suggestive; needs replication |
| No control group | Cannot interpret the outcome |
| Observational, groups self-selected | Association only, not cause |
Evaluating Someone Else's Investigation
Half of scientific practice is critique. Run any described study through the same short sequence and the flaws surface quickly.
First, name the independent and dependent variables in your own words; if you cannot find exactly one independent variable, two things were changed at once. Second, find the baseline — is there a control or a zero level? Third, hunt for confounding variables by asking what else differed between groups besides the treatment: location, time of day, who handled the organisms, starting size or age. Fourth, check and whether the trial was repeated. Fifth, check that the measurement is quantitative and consistently made; "the plants looked healthier" is not data, while "mean height in centimeters, measured from soil line to apical tip" is.
Finally, compare the conclusion to the design. A conclusion that reaches beyond the sample studied is overgeneralization: testing one fertilizer on radishes tells you about radishes under those conditions, not about "all plants." A conclusion that names a cause when the study only observed a pattern is a correlation-causation error.
When you find a weakness, propose the fix. That is what turns criticism into design skill: "Because all warm samples sat by the window, light was confounded with temperature; place both groups in identical dark incubators set to different temperatures." Or, "With two mice per group, increase to at least fifteen per group and repeat the trial."
This habit carries directly into the rest of the course. When you later measure how temperature or pH changes enzyme activity, the same questions decide whether your data mean anything: what did you change, what did you hold constant, what was your baseline, and how many trials did you run?
First, name the independent and dependent variables in your own words; if you cannot find exactly one independent variable, two things were changed at once. Second, find the baseline — is there a control or a zero level? Third, hunt for confounding variables by asking what else differed between groups besides the treatment: location, time of day, who handled the organisms, starting size or age. Fourth, check and whether the trial was repeated. Fifth, check that the measurement is quantitative and consistently made; "the plants looked healthier" is not data, while "mean height in centimeters, measured from soil line to apical tip" is.
Finally, compare the conclusion to the design. A conclusion that reaches beyond the sample studied is overgeneralization: testing one fertilizer on radishes tells you about radishes under those conditions, not about "all plants." A conclusion that names a cause when the study only observed a pattern is a correlation-causation error.
When you find a weakness, propose the fix. That is what turns criticism into design skill: "Because all warm samples sat by the window, light was confounded with temperature; place both groups in identical dark incubators set to different temperatures." Or, "With two mice per group, increase to at least fifteen per group and repeat the trial."
This habit carries directly into the rest of the course. When you later measure how temperature or pH changes enzyme activity, the same questions decide whether your data mean anything: what did you change, what did you hold constant, what was your baseline, and how many trials did you run?
Key terms
- Hypothesis.
- A testable, falsifiable proposed explanation or prediction, often written in if/then form and based on prior knowledge or observation.
- Independent variable.
- The single factor the experimenter deliberately changes or sets between groups; also called the manipulated variable.
- Dependent variable.
- The outcome that is measured and that may respond to the independent variable; it must be quantifiable.
- Controlled variable.
- Any factor deliberately held constant across all groups so it cannot influence the outcome.
- Control group.
- The group receiving no treatment or the standard treatment, providing the baseline for comparison with experimental groups.
- Confounding variable.
- An uncontrolled factor that varies along with the independent variable and could also explain the observed result.
- Sample size (n).
- The number of individuals or trials in each group; larger samples reduce the chance that natural variation is mistaken for a real effect.
- Correlation versus causation.
- A correlation is an association between two variables; only a controlled experiment with random assignment supports the claim that one caused the other.
Worked example
A student reads that a plant hormone spray increases growth. She sprays 15 radish seedlings with the hormone solution each morning and leaves 15 other seedlings unsprayed. All 30 seedlings are the same variety, planted the same day in identical pots of the same soil, given 50 milliliters of water daily, and kept in the same growth chamber. After 14 days she measures stem height in centimeters. Sprayed seedlings average 9.8 centimeters (range 8.5 to 11.0); unsprayed seedlings average 7.1 centimeters (range 6.4 to 8.0). Identify the variables and groups, write an if/then hypothesis, and evaluate whether she may claim the hormone caused the extra growth.
Step 1 — Independent variable: the factor she chose to change is the presence of hormone spray (sprayed versus not sprayed). It has two levels.
Step 2 — Dependent variable: stem height in centimeters at day 14. It is measured, quantitative, and could respond to the treatment.
Step 3 — Controlled variables: radish variety, planting date, soil type and amount, pot size, water volume (50 milliliters daily), and growth-chamber light and temperature. Holding these constant means they cannot explain a difference between groups.
Step 4 — Groups: the 15 sprayed seedlings are the experimental group; the 15 unsprayed seedlings are the control group and supply the baseline for normal growth.
Step 5 — Hypothesis: "If the hormone spray promotes stem elongation, then radish seedlings sprayed daily will have greater mean stem height after 14 days than unsprayed seedlings."
Step 6 — Evidence strength: the difference in means is centimeters. The sprayed range (8.5 to 11.0) does not overlap the unsprayed range (6.4 to 8.0), so the group difference is larger than the variation within groups. With per group and tight control of other factors, this is reasonably strong support.
Step 7 — Limits and a fix: one flaw remains. Only the treated plants had liquid sprayed on their leaves, so leaf wetting is confounded with the hormone. A better control would be spraying the second group with the same solvent minus the hormone — a placebo-style control. Her conclusion should also stay within the sample: the data support the claim that this hormone increased stem height in these radish seedlings under these conditions, not in all plants. Replicating the trial once more would strengthen the claim further.
Step 2 — Dependent variable: stem height in centimeters at day 14. It is measured, quantitative, and could respond to the treatment.
Step 3 — Controlled variables: radish variety, planting date, soil type and amount, pot size, water volume (50 milliliters daily), and growth-chamber light and temperature. Holding these constant means they cannot explain a difference between groups.
Step 4 — Groups: the 15 sprayed seedlings are the experimental group; the 15 unsprayed seedlings are the control group and supply the baseline for normal growth.
Step 5 — Hypothesis: "If the hormone spray promotes stem elongation, then radish seedlings sprayed daily will have greater mean stem height after 14 days than unsprayed seedlings."
Step 6 — Evidence strength: the difference in means is centimeters. The sprayed range (8.5 to 11.0) does not overlap the unsprayed range (6.4 to 8.0), so the group difference is larger than the variation within groups. With per group and tight control of other factors, this is reasonably strong support.
Step 7 — Limits and a fix: one flaw remains. Only the treated plants had liquid sprayed on their leaves, so leaf wetting is confounded with the hormone. A better control would be spraying the second group with the same solvent minus the hormone — a placebo-style control. Her conclusion should also stay within the sample: the data support the claim that this hormone increased stem height in these radish seedlings under these conditions, not in all plants. Replicating the trial once more would strengthen the claim further.
Practice questions
A researcher tests whether light color affects the rate of photosynthesis in elodea. She places identical elodea sprigs in test tubes under red, blue, green, or white light and counts oxygen bubbles released per minute. Which is the dependent variable?
- The color of light shining on each test tube
- The number of oxygen bubbles released per minute
- The temperature of the water in each test tube
- The length of the elodea sprig used in each tube
Answer: The number of oxygen bubbles released per minute
The dependent variable is the measured outcome. She sets the light color herself, so that is the independent variable. Water temperature and sprig length are held constant across tubes, making them controlled variables — and note that bubble count is the only item here that she reads off rather than sets up. A common wrong answer is water temperature, because students see the word "temperature" and think of a measurement, but she is keeping it the same on purpose so it cannot affect the comparison.
A newspaper reports that people who eat breakfast weigh less than people who skip it, and concludes that eating breakfast causes weight loss. Explain why this conclusion is not justified, and describe a study design that could support a causal claim.
Answer: The report describes an observational study in which people chose their own breakfast habits, so it can show only a correlation. Breakfast eaters may differ from skippers in many unmeasured ways — total food intake, exercise, sleep, income, or overall health habits — and any of these confounding variables could explain the weight difference. Reverse causation is also possible: people already trying to manage weight may adopt a breakfast routine. To support causation, researchers would need a controlled experiment: recruit a large group of similar volunteers, randomly assign each person to a breakfast-eating group or a no-breakfast group, keep other instructions identical, and measure weight change over several months. Random assignment with a large sample spreads unmeasured differences evenly between the groups, so a difference in outcome can reasonably be attributed to the treatment.
The key move is naming the design type first. Because the groups were self-selected rather than randomly assigned, the study cannot rule out alternative explanations, so the correct wording is "associated with," not "causes." A complete answer names at least one specific confounding variable and includes random assignment plus a large sample in the proposed fix.
A student wants to know whether adding sugar to water keeps cut flowers fresh longer. He puts one carnation in plain tap water on a windowsill and one carnation in sugar water in a cabinet, then records how many days each flower stays open. Identify two serious design flaws and state how to correct each.
Answer: Flaw 1: two factors differ between the flowers — sugar and light location — so light is a confounding variable and any difference cannot be attributed to sugar. Correction: place both vases in the same location under identical light and temperature, changing only whether sugar is present. Flaw 2: the sample size is one flower per group, so natural variation between individual carnations could produce the entire difference. Correction: use at least ten flowers of the same variety, size, and cut date in each group and compare group averages, then repeat the whole trial. A useful third improvement is defining the measurement precisely, such as the number of days until petals visibly wilt, judged by the same criterion for every flower.
This design fails the two checks used most often when evaluating an investigation: exactly one independent variable, and enough replication to separate a real effect from individual variation. Notice that the control group here (plain water) is appropriate in principle — the problem is that it was not treated identically apart from the sugar, which destroys the comparison the control group is supposed to make possible.
FAQ
- How do I tell the independent variable from the dependent variable?
- Ask which one you set up before collecting data and which one you read off afterward. The independent variable is what you deliberately change between groups; the dependent variable is the measurement that may respond to it. Test your labels in a sentence: "stem height depends on hormone spray" makes sense, but "hormone spray depends on stem height" does not, so height is dependent.
- Is the control group the same thing as a controlled variable?
- No. A controlled variable is a condition kept identical for everyone, such as temperature, soil, or water amount. The control group is a set of subjects that receives no treatment or the standard treatment, giving you a baseline to compare against. An experiment can have many controlled variables but usually just one control group.
- Why is a bigger sample size better, and how many is enough?
- Individual organisms naturally vary, so a difference between one treated and one untreated subject may be pure chance. More subjects let those random differences average out, so a real group difference shows through. There is no universal minimum, but a practical test is whether the gap between group averages is larger than the spread of values inside each group. In class labs, aim for at least ten per group and repeat the trial when possible.
- Can an experiment ever prove a hypothesis?
- No. Data can support a hypothesis, fail to support it, or contradict it, but a single study never proves it, because a future experiment or a better-controlled design could still overturn the result. Write conclusions as "the data support the hypothesis that..." and keep the claim inside the boundaries of what you actually tested — your species, your conditions, your sample.
Learn this with a teacher, not a page
The Crimsora tutor teaches Scientific Inquiry & Experimental Design in Biology live — explaining on a whiteboard, asking you questions, and adapting to where you get stuck.