Skip to content

AP Biology · technique

How to answer experimental-design questions

On this page · 5 sections

Experimental design is the single most reusable skill on the AP Biology exam, because it is not tied to any one unit. The College Board can hang it on enzymes, on photosynthesis rates, on gene regulation in bacteria, on allele frequencies in a beetle population, on nutrient runoff in a pond. The biology changes; the moves you make do not.

It shows up in two places. In Section I, blocks of the 60 multiple-choice questions are attached to a described experiment, a data table, or a graph, and ask you to pick the appropriate control, the variable that was manipulated, the conclusion the data actually support, or the next logical test. Section II is where it dominates, and its six free-response questions have fixed, named types. Question 1 is Interpreting and Evaluating Experimental Results. Question 2 is Interpreting and Evaluating Experimental Results with Graphing, and it requires you to construct a graph by hand. That is the single most predictable fact about this exam, which is why the graphing rules below are not optional points. Those two are the long questions, worth 8 to 10 points each. The four short questions are worth 4 points each: Scientific Investigation, Conceptual Analysis, Analyze Model or Visual Representation, and Analyze Data. Experimental reasoning drives questions 1, 2, 3, and 6 outright, so it carries most of the free-response half of your score.

Scoring is point-by-point, not holistic. A reader works down a rubric with a specific earnable point for each task: one for stating a valid null hypothesis, one for identifying the control and saying what it establishes, one for correct axis labels with units, one for the calculation, one for a claim that matches the data, one for reasoning that links the result to a biological mechanism. Parts are scored independently, so a wrecked part (a) does not sink part (c). Two things do cost you: writing a correct statement next to a contradictory one, which can cancel the point, and describing data when the prompt asked you to explain or justify. You are given a formula sheet with mean, standard deviation, standard error of the mean, and chi-square, plus a chi-square critical value table. Nothing statistical is memorized. It is all applied.

The steps

  1. 1

    Strip the stem down to variables

    Before you read the data, find three things in the setup paragraph and mark them. The independent variable is what the researchers deliberately changed and set the levels of: drug concentration, temperature, light wavelength, presence or absence of a gene. The dependent variable is what they measured as a response: oxygen produced per minute, percent germination, number of offspring, band intensity on a gel. Everything else they deliberately held the same is a controlled variable, and the question often hides its answer there. Say the relationship out loud in one sentence: they changed X and measured Y. If you cannot say that sentence, you have not understood the experiment, and every later part will drift. On a graph, this also tells you the layout, because the independent variable goes on the x-axis and the dependent variable on the y-axis.

  2. 2

    Find the control and name what it rules out

    A control group is not just "the normal one." It exists to eliminate a specific alternative explanation, and the point is awarded for naming that explanation. A negative control gets everything except the independent variable: buffer with no enzyme, cells with no signaling molecule, the vector plasmid with no inserted gene. It establishes the baseline and proves the effect you measured did not come from the setup itself. A positive control gets a treatment already known to produce the effect. It proves the assay would have detected a response if one existed, which is what saves you when your experimental group shows nothing. Ask which one the question needs. If a researcher reports no response to a new hormone, the missing piece is a positive control. If a researcher reports a response, the missing piece is a negative control. Never call the untreated experimental level "the control" without explaining the comparison it makes possible.

  3. 3

    State the hypothesis as something that could fail

    The null hypothesis is always the boring one: there is no difference between the groups, or any observed difference is due to chance alone. That phrasing matters, because chi-square and error-bar reasoning both test the null, not your hunch. Write it in the terms of the actual experiment rather than in the abstract. "There is no difference in the mean rate of oxygen release when yeast catalase is mixed with hydrogen peroxide in the presence versus the absence of the inhibitor" earns the point. "The inhibitor will not do anything" does not. Name the assay, not just the organism, so the reader can see what was measured. The alternative hypothesis states there is a difference, and if the prompt asks for a prediction, give it a direction and a reason: the treated group will produce less oxygen because the compound blocks the active site. A prediction with a mechanism attached is doing double duty, since reasoning points show up later in the same question.

  4. 4

    Read the data before you write a word

    Spend thirty seconds mining the figure. Get the direction of the trend, whether it is linear or levels off, and where it changes. Get magnitude with real numbers, including units, because rubrics reward specific data over the phrase "went up." Then check the uncertainty. If error bars are drawn as 95% confidence intervals, roughly two standard errors of the mean, the rule is mechanical: overlapping bars mean you cannot claim a significant difference, and non-overlapping bars mean the difference is likely real and not chance. If a chi-square value is given alongside a critical value, compare them. Note the sample size, because n tells you how much a single odd measurement could be distorting the mean. Do this before writing and your claim will already be defensible; do it after and you will end up defending a sentence you guessed at.

  5. 5

    Do the mechanical tasks by the rulebook

    Graphing and calculating are the cheapest points on the exam because they are scored against a fixed list. When constructing a graph: independent variable on x, dependent on y, both axes labeled with units, a scale that is evenly spaced and uses most of the grid, and bars for discrete categories versus a line or scatter for continuous data. Add error bars if the standard error is given, and a key if there is more than one series. For calculations, standard error of the mean is the sample standard deviation divided by the square root of n, so quadrupling your sample size halves the error bar. Chi-square sums the quantity observed minus expected, squared, divided by expected, across every category, and it runs on counts, never percentages. Degrees of freedom is the number of categories minus one. Show the substituted numbers, not just the final value, and carry the units.

  6. 6

    Answer in claim, evidence, reasoning, keyed to the verb

    Match the prompt verb to the amount of writing it wants. "Identify" and "describe" want the observation only: the mean rate fell from 12.4 to 3.1 mL per minute as concentration rose. "Explain" wants the biological mechanism behind the observation. "Justify" or "support a claim" wants all three parts in order: the claim, the specific numbers that back it, and the reasoning that connects those numbers to the claim. The reasoning sentence is the one students skip and the one readers are looking for, and it usually starts with "because." Keep everything on the point being asked. Extra correct biology is neutral, but a hedge that contradicts your own answer can cancel an earned point, so pick one position and commit to it.

  7. 7

    Close the loop to the concept and the next experiment

    The last part of a long free-response question almost always widens out. It asks you to connect the result to a course-level idea, or to propose what you would do next. For the connection, name the concept explicitly: enzyme active-site specificity, negative feedback, differential reproductive success, ATP yield from the electron transport chain. For the next investigation, do not propose a whole new topic. Change exactly one variable from the experiment you were just given, keep everything else identical, state the control, and say what result would support which conclusion. That final clause is where the point lives. A proposal without a stated expected outcome is a plan, not an experiment, and rubrics treat it that way.

Worked through

The setup. Researchers test whether compound X inhibits catalase, the enzyme that breaks hydrogen peroxide into water and oxygen. Each trial mixes a fixed volume of liver extract with a fixed volume of hydrogen peroxide at pH 7 and 25 degrees Celsius, and they measure the oxygen released in 60 seconds. Five replicates per group. Means are reported with error bars of two standard errors of the mean. The sample standard deviation for the 0 micromolar group was 1.2 mL per minute.

The data, in mL of oxygen per minute: Buffer with peroxide, no extract: 0.2, plus or minus 0.3 Extract, 0 micromolar X: 12.4, plus or minus 1.1 Extract, 10 micromolar X: 11.9, plus or minus 1.3 Extract, 50 micromolar X: 7.6, plus or minus 0.9 Extract, 100 micromolar X: 3.1, plus or minus 0.8

Variables. Independent variable: concentration of compound X, set at four levels. Dependent variable: volume of oxygen released per minute. Controlled: pH, temperature, extract volume, peroxide volume, reaction time.

Controls. This experiment runs two, and they do different jobs. The extract at 0 micromolar X is the negative control for the independent variable. It gets everything except compound X, so it is the baseline the treated means are measured against. It rules out the possibility that a drop in oxygen came from the handling, the timing, or the extract itself rather than from X. The buffer-with-peroxide-and-no-extract tube is a separate assay control. It contains no enzyme and produces essentially no oxygen, which rules out spontaneous peroxide breakdown and stray bubbles from the apparatus. It establishes that any output above 0.2 mL per minute is enzyme-catalysed. If a rubric asks for the control for the effect of compound X, name the 0 micromolar extract group.

Null hypothesis. There is no difference in mean oxygen production between catalase reactions run with and without compound X, and any difference observed is due to chance.

Checking one error bar. The 0 micromolar group has a sample standard deviation of 1.2 with n equal to 5. Standard error is 1.2 divided by the square root of 5, which is 1.2 divided by 2.24, or 0.54. Two standard errors is 1.1, matching the reported bar.

Reading significance. The 0 micromolar interval spans 11.3 to 13.5. The 10 micromolar interval spans 10.6 to 13.2. Those overlap, so no inhibition is detectable at 10 micromolar. The 50 micromolar interval spans 6.7 to 8.5 and the 100 micromolar interval spans 2.3 to 3.9. Neither overlaps the 0 micromolar interval, so both differences are likely real.

The justified answer. Claim: compound X inhibits catalase at 50 and 100 micromolar. At 10 micromolar no inhibition was detectable, so these data do not support an effect at that concentration. Evidence: mean oxygen output fell from 12.4 to 7.6 to 3.1 mL per minute as X rose from 0 to 50 to 100 micromolar. The 95% confidence intervals for the 50 and 100 micromolar groups do not overlap the 0 micromolar group, while the 10 micromolar group's does. Reasoning: non-overlapping confidence intervals mean the difference is unlikely to be produced by random variation among replicates, so the drop in oxygen reflects fewer peroxide molecules being converted per second, which is what reduced enzyme activity looks like. Do not upgrade the 10 micromolar result into a claim that X does nothing there. With only five replicates, the study may be too small to catch a small effect, and failing to reject the null is not evidence for it.

Connecting out. Catalase works because peroxide fits a specific active site. Anything that blocks that site or distorts the enzyme's shape lowers the reaction rate, which is why a small molecule can shut down a protein that is otherwise abundant and undamaged.

The next experiment. Hold compound X at 100 micromolar and vary hydrogen peroxide concentration across a wide range, with a matching series at 0 micromolar X as the control and everything else unchanged. If high peroxide restores the rate to the uninhibited maximum, X is competing for the active site and the inhibition is competitive. If the rate stays depressed no matter how much peroxide is added, X is binding elsewhere and the inhibition is noncompetitive.

Common mistakes

  • Describing the data and stopping there. "Oxygen production decreased as concentration increased" is the evidence, not the answer. A justify prompt still needs the claim and the because-clause that ties the numbers to a mechanism, and readers award those as separate points.

  • Ignoring the error bars. Two means that look far apart on a graph are not significantly different if their 95% confidence intervals overlap. Claiming a real effect from overlapping bars, or dismissing a real effect because the means are numerically close, both lose the analysis point.

  • Turning a non-significant result into a positive claim of no effect. Overlapping error bars mean the difference was not detected, not that the difference is zero. Write "no detectable effect at this concentration," and remember that a small n can hide a real one.

  • Mislabeling the control. Calling the untreated experimental level "the control" without saying what it rules out earns nothing, and offering a negative control when the experiment produced no response is the wrong fix. No response calls for a positive control, to show the assay could have detected one.

  • Botching chi-square mechanics. Running the test on percentages instead of raw counts, using n minus one where n is the number of individuals rather than the number of phenotype categories, or writing "we accept the null hypothesis" instead of failing to reject it.

  • Proposing a next experiment with no control and no predicted outcome. Changing several conditions at once, or writing a plan that ends without "if the result is A, that supports B," leaves the reader nothing to score.

  • Hedging into a contradiction. Writing that the compound inhibits the enzyme and then adding that the results might show no effect gives the reader two answers. A contradictory statement sitting next to a correct one can cancel the point you already earned.

Checklist

  • Can you read any experimental stem and state, in one sentence, what was changed, what was measured, and what was held constant?
  • Can you tell a negative control from a positive control, and say which one a given flawed experiment is missing and why?
  • When an experiment carries more than one control, can you say which one is the comparison group for the independent variable and which one validates the assay?
  • Can you write a null hypothesis in the specific terms of the experiment in front of you, naming the assay rather than just the organism?
  • Can you calculate the standard error of the mean from a standard deviation and a sample size, and explain what happens to the error bar when n goes up?
  • Can you look at two means with 95% confidence interval error bars and say correctly whether the difference is significant, without turning an overlap into proof of no effect?
  • Can you run a chi-square on raw counts, get the degrees of freedom right, compare to the critical value at p equals 0.05, and state the conclusion in null-hypothesis language?
  • Can you construct a graph from a data table with correct axis assignment, labeled units, an even scale, the right mark type, and error bars?
  • Can you write a three-sentence claim, evidence, and reasoning answer where the reasoning names an actual biological mechanism?
  • Can you propose a follow-up experiment that changes one variable, keeps a control, and states which result would support which conclusion?

More AP Biology technique