Skills Beta

Experimental design and argumentation: the one-page sheet

Independent and dependent variables

A controlled experiment changes one factor, the independent variable, and measures one response, the dependent variable. Everything else is held constant as controlled variables, a control group shows the baseline, and repeated runs show chance variation. A testable question names both variables, and an if-then prediction links them with a reason.

Worked example. "Dishes of 50.0 g water at 20, 30 and 40 °C; mass lost in 24 h measured." Independent variable: temperature (set). Dependent variable: mass lost (measured). Controlled: dish, starting mass, time, air movement. Control group: unheated dishes. If-then: if temperature rises, then more water evaporates, because more molecules escape their hydrogen bonds.
Quick test: the independent variable is what you set before the run; the dependent variable is what you read off at the end.
A control group is a whole group given the baseline; a controlled variable is one factor held the same in every group. Exams mark answers down for mixing them up.
If two factors change together (temperature and a fan), the result cannot be traced to either one. Look for that flaw in every described experiment.

The question names the factor you will change (temperature) and the result you will measure (mass lost). Temperature is the independent variable: you choose its values. Mass lost is the dependent variable: it may respond to temperature. These controlled variables cannot explain a difference, so a change in mass lost can be traced to temperature. Repeats show chance variation, and the control group shows the baseline, so you can judge whether temperature really made a difference.

control group: The group that does not receive the treatment being tested (or gets the normal, baseline condition), used as the comparison for the experimental (treatment) groups.
controlled experiment: An experiment in which the investigator deliberately changes one factor and keeps everything else the same, so a difference in the result can be traced to that one factor.
independent variable: The factor the investigator deliberately changes or sets to different values (for example, water temperature). Also called the manipulated variable.
dependent variable: The result the investigator measures to see whether it responds to the independent variable (for example, mass of water lost). Also called the responding or measured variable.
controlled variable: A factor kept the same in every group so it cannot be the reason the results differ (for example, the volume of water and the shape of the dish).
repeated trials: Running each condition several times (replicates). Repeats show how much results vary by chance and make the average more reliable.
testable question: A question that can be answered by changing one factor and measuring a result, such as "Does water temperature affect how much water evaporates in 24 hours?"
hypothesis: A proposed, testable explanation for an observation, which leads to predictions an experiment can check.
if-then prediction: A prediction that links the two variables: "If [independent variable is changed this way], then [dependent variable will change this way]", usually followed by "because" and the proposed explanation.

Controls and what they rule out

A negative control (should be "no") rules out false positives from the method, materials or contamination. A positive control (should be "yes") rules out false negatives from a test that is not working. If either control fails, the run's results are not trusted. A confounding variable is a second difference between groups that could explain the result; hold it constant. When asked to justify a control, say what it rules out.

Worked example. Grease-spot test. Negative control: water, treated the same way. Justification: "Water has no lipid, so if it leaves no spot after 30 minutes, a lasting spot from a food is not caused by moisture or handling." Positive control: vegetable oil. Justification: "Oil is a known lipid, so a spot shows the test can detect fat today."
To earn the point, a justification names what the control rules out. "Water is the control because it is the control" scores nothing.
A confounding variable differs between groups and could change the dependent variable. Different pans, different sample masses or different reading times are classic examples.
The control must go through every step the test samples go through. An untouched sheet of paper is not a negative control for a test that involves rubbing and wetting.

You need one control for each kind of error. This negative control should give "no"; if it does, a "yes" from a food is not caused by the paper, handling or contamination. This positive control should give "yes"; if it does, a "no" from a food is believable, because the test is working. Every result from that run is suspect, so you find the cause (such as a contaminated wash tub) and repeat. That confounding factor could explain the difference, so it must be held constant before the result can be credited to the independent variable.

negative control: A group or sample that should NOT give the response (for example, water in a test for fat). If it stays negative, the response in other samples was not caused by the procedure, the materials or contamination.
positive control: A group or sample known to give the response (for example, vegetable oil in a test for fat). If it turns positive, the test works under these conditions, so a negative result elsewhere is believable.
confounding variable: A factor other than the independent variable that differs between groups and could also affect the dependent variable, so the result cannot be credited to the independent variable alone.

Null and alternative hypotheses

A null hypothesis states that the independent variable has no effect; the alternative states that it does (optionally in a named direction). Because measurements vary by chance, you judge a difference against the variation within groups. A difference much larger than that variation leads you to reject the null; otherwise you fail to reject it. Data support or fail to support hypotheses; they never prove them, and a null is never accepted.

Worked example. Question: does heating temperature affect leakage from beet? H₀: heating temperature has no effect on leakage. H₁: heating temperature affects leakage (directional: higher temperature increases leakage). 20 vs 80 °C: no overlap, reject H₀. 20 vs 40 °C: heavy overlap, fail to reject H₀.
The null hypothesis says no effect, no difference, no relationship. The alternative says there is one, with or without a direction.
The only two decisions are reject or fail to reject the null. Never write "accept the null" or "prove the hypothesis".
Failing to reject is not the same as showing there is no effect: a small effect, few repeats or noisy data can hide a real difference.

Two groups will almost never have exactly the same mean, so a difference alone proves nothing. The null predicts that any difference between groups will be about as small as chance variation. At 20 vs 40 °C the gap (0.01) is smaller than the spread (0.06); at 20 vs 80 °C the groups do not even overlap. You reject the null hypothesis, and the data support the alternative. You fail to reject the null: the data do not show an effect, though a small one could still exist.

null hypothesis: The statement that the independent variable has no effect: any difference between groups is due to chance alone (H₀). It is the statement an experiment tests.
alternative hypothesis: The statement that the independent variable does have an effect, so groups really differ (H₁ or Hₐ). It may name a direction ("higher temperature increases leakage") or not ("temperature changes leakage").
reject the null hypothesis: The decision made from data: reject the null hypothesis when the difference between groups is too large to blame on chance, or fail to reject it when it is not. A null hypothesis is never "accepted" or "proved".

Claim, evidence and reasoning

A scientific argument has three parts. The claim answers the question in one sentence. The evidence quotes the specific data, with units and the comparison that matters. The reasoning explains with a biological mechanism why that evidence supports the claim. Keep the claim within what the data show, and name limitations and sources of error honestly.

Worked example. Claim: potato catalase works fastest near pH 7. Evidence: 4.8 mL O₂/min at pH 7, versus 2.1 at pH 5 and 3.0 at pH 9; the 95% CIs do not overlap. Reasoning: pH alters R-group charges in the active site, changing its shape and substrate binding. Limitation: pH was tested every 2 units, so the optimum lies between about 6 and 8.
Evidence = data (numbers, units, a comparison). Reasoning = biology (a mechanism). Restating the data as the "reason" earns no reasoning point.
Fit the claim to the evidence: overlapping error bars support "no difference shown", not "faster".
A source of error is a measurement problem (inconsistent timing). A limitation is a boundary on the conclusion (only one species, wide spacing of values).

You write a claim: one sentence that answers it ("catalase works fastest near pH 7"). You quote evidence: values with units and the comparison that matters (4.8 mL/min at pH 7 against 2.1 at pH 5 and 3.0 at pH 9, with non-overlapping error bars). You give reasoning: pH changes the charges on R groups in the active site, altering its shape and substrate binding. You name a limitation, which keeps your claim from going further than the data allow.

scientific claim: A one-sentence answer to the question that the data can support or refute, such as "Catalase from potato works fastest near pH 7."
evidence (CER): The specific data that back the claim, quoted with numbers and units and compared across groups, such as "4.8 mL O₂/min at pH 7 versus 2.1 at pH 5 and 3.0 at pH 9".
reasoning (CER): The biological explanation of why the evidence supports the claim, linking the data to a principle (for example, how pH changes the charges in an active site). It is more than restating the data.
claim-evidence-reasoning: Claim-evidence-reasoning: a structure for a scientific argument. State the claim, give the evidence from the data, then explain with biology why that evidence supports the claim.
experimental limitation: Weaknesses in a study that limit what it can conclude, such as too few repeats, an uncontrolled variable, or conditions too narrow to generalize from, and sources of error in how measurements were made.

Prediction plus mechanism

A prediction with mechanism names the variable that will change, the direction, and the chain of causes from the disruption to that result. Trace perturbations through the model you are given, step by step; say "no change" where a part is unaffected; and propose next experiments that isolate where a disruption acts, for example by using isolated mitochondria to bypass glycolysis.

Worked example. Prediction: ATP in cyanide-treated cells will fall. Mechanism: cyanide blocks the last carrier, so electrons cannot reach O₂; the chain stops pumping H⁺; the gradient runs down; ATP synthase, driven by H⁺ flow, makes less ATP.
Three parts score: direction (up, down, no change), mechanism (each cause leads to the next) and a measurable variable at the end.
Use the model you are given. Trace the change along its arrows: a compound that takes electrons before a block restores flow upstream of the block, not downstream.
Some variables do not change. Saying "no change" with a reason (DCMU does not change how much light reaches the leaf) is part of a good prediction.

Electrons can no longer be passed to O₂, so O₂ use falls. The chain stops pumping H⁺, and the H⁺ gradient across the inner membrane runs down. ATP production by oxidative phosphorylation falls sharply. The Krebs cycle stalls, and cells that can ferment make more lactate to regenerate NAD⁺ for glycolysis. You can predict: less O₂ used, less ATP, more lactate, and no change in parts the poison does not touch.

prediction with mechanism: A prediction that states what will change, in which direction, and the step-by-step biological cause: "ATP will fall, because blocking the last electron carrier stops proton pumping, so the gradient that drives ATP synthase runs down."
system perturbation: A deliberate or natural disruption of one part of a biological system (a poison, an inhibitor, a missing molecule), used to predict or test how the rest of the system responds.
follow-up experiment: An experiment proposed after the first one to test what it could not: a new variable, a missing control, a wider range, or the mechanism behind the result.
model-based prediction: A prediction made by tracing a change through a model or diagram of a system, step by step, to its effect on a measurable variable.