IB Biology IA guide: how to plan, write and score well (2025 criteria)

What is the IB Biology IA?

The Biology internal assessment (IA) is a single scientific investigation. You choose a question, gather or select quantitative data to answer it, analyse that data, and write a report of at most 3,000 words. Your teacher marks it and the IB moderates a sample. It is worth 20% of the final grade, and the task and criteria are the same at SL and HL.

The IB recommends about 10 hours of teaching time for the whole task. Most students need more outside lessons, and Biology often needs the most, because living material has to be grown or collected before you can measure it. The question does not need to go beyond the syllabus. It needs to be yours, measurable, and big enough that every criterion has something to mark.

The rules that trip students up

• Your raw data are yours. You may not hand in the same data set as a classmate.
• Groups are allowed, but reports are not shared. Up to three students may work together, but each one must vary a different independent variable, measure a different dependent variable, or use different data. Each writes an individual report.
• One draft. Your teacher can read and advise on one draft, orally or in writing, but does not edit it. The next version is the final one.
• Reference everything. Missing or improper referencing counts as academic malpractice.
• No reuse. The same piece of work cannot count for both the IA and the Extended Essay.

What changed from the pre-2025 criteria

If a website or an old exemplar mentions Personal engagement or Communication, it describes the previous course. The 2025 course has four equal criteria and no separate marks for presentation or personal engagement. Old exemplars still show good biology, but they were marked on a different scheme, so do not copy their structure blindly.

Marks that once rewarded effort and presentation now reward reasoning: explained choices, uncertainty carried through, and a conclusion that stays inside the numbers. Biology adds one more habit: showing that you understand how much living things differ from one another.

How the Biology IA is marked: the four criteria

Each criterion is worth 6 marks and is marked in three bands: 1 to 2, 3 to 4 and 5 to 6. Your teacher gives whole numbers, choosing the band that fits best. A top mark does not need a perfect report. It needs a few things done clearly and cleanly.

The bands are separated by command terms. Describe means say what you did. Explain means say what you did and why. Look for the word “because” in the right places.

Criterion A: Research design (6 marks)

What it rewards. A research question set in context, thoughtful choices about how you collect or select data, and a method that someone else could repeat.

What top work does. The question names both variables and the system (which organism, tissue, habitat or data set), with background theory that bears directly on it. Each choice, such as range, repeats, sample size, instrument or control, comes with a reason, and the method is clear enough to repeat without guessing.

Where marks go. A topic instead of a question; a narrow range; “some leaves” instead of a described sample; no sampling method or sample size; controls named but not measured; ethics copied from a worksheet.

Our tip. Next to every number in your method, write “because” and finish the sentence. If you cannot, the choice is not yet explained.

Criterion B: Data analysis (6 marks)

What it rewards. Data that are recorded, processed and presented so that the reader can follow them (clear) and so that the conventions are right (precise): units, decimal places, significant figures, labelled tables and graphs. Uncertainties must be handled properly.

What top work does. Raw tables carry units and uncertainties. Counts and rates state their sample size and base (per minute, per gram, per square metre). Each calculation has a worked example, error bars say what they show, and any statistical test is chosen for a reason with its conditions checked.

Where marks go. Means with no spread; an uncertainty found once and never used again; a test with no check of its conditions; “significant” used loosely; points joined dot to dot; outliers dropped with no reason.

Our tip. Pick one processed number and trace it back to the raw data. If any step is missing, the reader cannot follow it either.

Criterion C: Conclusion (6 marks)

What it rewards. How well you answer your question in line with your analysis (including its uncertainties) and with accepted science.

What top work does. The first sentence answers the question using both variables and numbers. The claim is justified by the data and their uncertainties, limited to the range and sample you measured, and compared with a cited value whose source and conditions the reader can trace.

Where marks go. Retelling the method; a trend claimed when the error bars allow a flat line; “consistent with the literature” with no number or source; a value for a different organism or temperature; conclusions about a whole species from a tiny sample; the word “proves”.

Our tip. Underline the variables in your question, then check that your first conclusion sentence contains both.

Criterion D: Evaluation (6 marks)

What it rewards. An honest look at your method: specific weaknesses and limitations, how much each one matters, and realistic improvements.

What top work does. Every weakness belongs to this method and shows in this data (scatter, an offset, an outlier, a large error bar). The weaknesses are weighed against each other, ideally with numbers, and each improvement says what changes and what you would expect to see.

Where marks go. “Human error” with no sign it affected your results; ten weaknesses and no ranking; “more repeats” with no number.

Our tip. For each weakness, ask where in your data a stranger could see it. If the answer is nowhere, cut it.

Three habits separate the bands in all four criteria: give reasons, not lists; give numbers, not adjectives (a stated uncertainty beside a stated difference); and link the criteria, so that a prediction in A is tested in C, a spread from B is used in C, and a weakness in D shows in the graph in B.

How to choose a Biology IA topic

A good topic is less about how impressive it sounds and more about whether it can produce clean, repeatable numbers from material you can get and keep consistent. A task that gives no trend and nothing to criticise is too small. Use these seven tests.

1. Clear variables. You can name what you will change and what you will measure, and your school equipment can measure it to a useful precision.
2. Range and repeats. You can vary the independent variable over a wide range, at least five values in our experience, and you can afford a good sample at each value in the time you have.
3. Theory to test against. A known relationship, such as an enzyme optimum or a water-potential balance, gives you a prediction in Criterion A, a comparison in C and assumptions to examine in D.
4. Feasible, safe and ethical. It fits your time, equipment and school rules, and it harms no person or animal. A pilot run in the first fortnight saves weeks.
5. Yours. Not the standard school practical with the same variables and ranges, which every classmate will also pick.
6. Room for uncertainty work. If the spread between organisms or samples is the main source of error, you can analyse it. If every value is identical and nothing varies, there is little to say.
7. Honest scope. Small and clean beats broad and thin.

Topics that tend to fail: anything that cannot give five measured values; a yes-or-no question; material or equipment you cannot get; work that harms animals or people; slow-growing material with no time to repeat it; and a question already answered by a standard practical.

Start from the material you can handle reliably, then ask what you could vary widely and what theory says should happen. Our full list of ideas is in 95 IB Biology IA ideas and EE topics, and the free 25 Biology IA topic guides each come with a set-up diagram and a sample graph. Treat ideas as starting points. The final question, plan and data must be your own.

How to write a Biology IA research question

The IB describes a question “with context” as one that names the dependent and independent variables (or two correlated variables), briefly describes the system, and has background theory of direct relevance. In practice that means four parts, shown below.

Weak, better, strong

These examples are our own, written to show the difference. They are not topics to copy.

What changed each time? The strong versions give a range with units, name how the dependent variable is measured, fix the system and sample, and point to something you can compare your result with in Criterion C. They are longer, and that is fine if every part earns its place.

The 5-point check

Before you commit, score your question against five points. This is the same check our Coach applies to a research question: focus, variables, feasibility, data, and fit with the criteria.

Variables, controls, replicates and sample size

Every Biology IA has three kinds of variable. The independent variable is what you change. The dependent variable is what you measure. Control variables are everything else that could change the result. A control is only a control if you say how you hold it constant and how you would notice if it drifted. Biology adds a fourth thing you cannot fully control: the organisms themselves.

The catalase question from Figure 4 shows what a good plan writes down.

Biological variation and why you need more repeats

Two identical test tubes of acid behave almost alike. Two potatoes, two seeds or two quadrats do not. They differ in age, size, genes and place, and none of that is your mistake. The consequence is simple: you need a bigger sample before the mean means anything, and you must report how big it was.

Two kinds of repeat do different jobs:

• Technical repeats measure the same sample again. They show how good your technique is.
• Biological repeats use new, independent organisms, discs or plots. They show how much the living material varies, and only they count towards your sample size n. Ten readings from one leaf is a sample of one leaf.

The IB gives no magic number, and these figures are our working advice, not an IB rule. Five values of the independent variable show the shape of the trend. At each value we suggest at least five independent samples, and eight to ten where the material is cheap and quick. Run your pilot, find the standard deviation, and let it tell you how many more you need.

The standard error of a mean falls with the square root of n, so to halve it you need four times the sample. Going from 3 to 10 samples gives a big gain. Going from 20 to 30 gives a small one, and the hours are better spent controlling a variable.

Reduce the variation you can

• Use one source for each batch: one potato, one seed packet, one culture.
• Standardise size: a cork borer for cylinders and discs, plants of the same age.
• Record what you could not hold: room temperature, light, humidity, time of day.
• Randomise the run order, so a slow drift in the room does not line up with one condition.
• Run a blank or control group, so you know what “no effect” looks like.

Ethics in the Biology IA: animals, people and the environment

Recognising safety, ethical and environmental issues is part of Research design, and your teacher must make sure you have thought about them before you collect any data. In Biology this is not a formality, because your material is alive. Write about your actual organisms and steps.

The IB has its own guidance on experiments with animals and with people, and your school has rules on top. Check the current guidelines and your school’s ethics process with your teacher before you start. What follows is our general approach.

Animals

No investigation should harm an animal or cause it distress. Our default is to steer students away from live-animal work, above all vertebrates. If an animal is involved, your teacher must agree first, and the work should rest on observing natural behaviour, not changing it with stress, injury, heat or chemicals. Plants, fungi, microorganisms, existing data and simulations are good alternatives, and you do not need a dissection to do a good IA.

Human participants

Using classmates as subjects (reaction time, resting heart rate) brings its own rules. Participants should give informed consent: they know what you will do, they know their data will appear in your report, and they can stop at any point without giving a reason. Keep every measurement non-invasive, and avoid drugs, fasting, injury risk or private health information. Whether younger participants may take part, and how consent must be recorded, depends on the current IB guidance and your school, so check with your teacher. Use codes, not names, in your report.

Microorganisms

Use only organisms your school has approved as safe, and never grow unknown environmental samples, which could contain pathogens. Keep dishes taped shut, do not open them after incubation, and dispose of cultures as your school instructs.

Fieldwork and the environment

Get permission to enter and sample a site, especially a park, a reserve or private land. Do not collect protected species, trample plants or leave equipment behind, and count in place instead of taking samples away. Plan for weather and heat, and for how you dispose of cultures and solutions.

Safety

Name the hazard and the control for each real item in your plan: hydrogen peroxide, stains, acids, ethanol, glassware, hot lamps and any allergens in your material. A short table of hazard and precaution, specific to your steps, beats a general paragraph.

Methods that give clean data

Every technique has one classic trap that quietly costs marks. Here are the five we see most in Biology, each with a fix you can build into the plan.

Colorimetry

The trap. Reading an unknown outside the calibration, or letting the cuvette and wavelength change between readings. It shows up in beetroot pigment, enzyme product and chlorophyll work.

The fix. Make at least five standards across the range you expect, zero with a blank, choose a wavelength where your substance absorbs strongly, and place the cuvette the same way each time. Keep unknowns inside the calibrated range and dilute any sample that reads above your top standard.

Gas production

The trap. Gas that leaks from a bung, dissolves in the water, or is counted as bubbles of different sizes. Timing that starts late because the first few seconds are hard to catch.

The fix. Use a gas syringe or an inverted measuring cylinder over water that has already been saturated with the gas, test the apparatus for leaks in your pilot, and write one fixed start procedure. Measure volume, not bubble count. For a rate, use the early, straight part of the curve, and say what you used as the time base.

Quadrats and transects

The trap. Too few quadrats, placed wherever it is easy to stand, so the sample cannot represent the habitat.

The fix. Decide the quadrat size in a pilot. Use random coordinates for a random sample, or a regular line for a gradient, as in Figure 8. Keep adding quadrats until the running mean stops moving, and say how you reached your final number. Use the same observer and one rule for plants that lie on the edge of the frame.

Microscope counts

The trap. Counting cells in one convenient field of view, counting cells twice at the border, or comparing slides at different magnifications.

The fix. Count several fields per slide, chosen in a fixed pattern, and several slides per condition. Use one rule for cells touching the edge (for example, count those touching the top and left edges and ignore the bottom and right). Keep the magnification and the stain time the same, and record the field area so you can report cells per unit area.

Data loggers and sensors

The trap. A sensor that drifts, an interval that misses the change, or a huge file that is never processed.

The fix. Check the sensor against a known standard, and choose an interval that gives enough points where the curve changes. Note the start time and events, then reduce the data to a few values that answer your question, such as an initial rate, a peak or a time to a threshold.

Statistics for the Biology IA

Biological data scatter, so a bare mean tells the reader very little. Statistics answer two questions: how much do the values vary, and is the difference I see bigger than that variation would produce on its own? Which tests your course expects is something to check with your teacher and your Biology guide. The ideas below are general statistics, and we have checked the numbers.

Mean, standard deviation and error bars

• Mean. The centre of your sample. Report it with the sample size n.
• Standard deviation (SD). How far individual values typically sit from the mean. It describes the spread of the data. Use the sample version (STDEV.S in a spreadsheet), not the population one.
• Standard error (SE). SD divided by √n. It describes how well you know the mean, and it shrinks as the sample grows.

An error bar can show an SD, an SE or a range, so say which on the graph or in the caption. For small samples, plot the individual points beside the bars, as in Figure 10. Overlapping SD bars do not prove there is no difference, and clear separation does not replace a test.

Which test?

A t-test asks whether two group means differ by more than chance would explain. Use the paired version when the same subjects are measured twice, and the unequal-variance (Welch) version when the spreads differ. A chi-squared test asks whether counts fit an expected pattern or differ between groups. A correlation asks whether two measured variables rise and fall together. Use Pearson’s r for points around a straight line, and Spearman’s ρ for a steady rise or fall along a curve: an invented set of eight light-intensity points that climbs and levels off gives r = 0.98 but ρ = 1.00. Neither shows that one variable causes the other.

A worked t-test

Take the catalase question with invented numbers. Eight independent potato extracts are tested at 20 °C and eight at 30 °C. The volumes of oxygen in 60 s are:

• 20 °C: 3.1, 3.6, 2.8, 3.4, 3.0, 3.9, 3.3, 3.5 cm³
• 30 °C: 4.6, 5.2, 4.1, 4.9, 5.5, 4.4, 5.0, 4.7 cm³

1. Say what you are testing. The null hypothesis is that the two true means are equal. The alternative is that they differ.
2. Mean and SD. 20 °C: mean 3.33 cm³, SD 0.35. 30 °C: mean 4.80 cm³, SD 0.45. The difference in means is 1.48 cm³.
3. Standard error of the difference. √(0.35²÷8 + 0.45²÷8) = 0.20 cm³, using the unrounded SDs.
4. The t statistic. 1.475 ÷ 0.2016 = 7.32. The Welch degrees of freedom come to about 13.
5. The p-value. A spreadsheet gives it as =T.TEST(range1, range2, 2, 3), which means two-tailed, unequal variance. Here p is about 5 × 10⁻⁶, so we write p < 0.001.
6. Interpret it. If the true means were equal, a gap this large would almost never turn up by chance. A 95% confidence interval for the difference runs from about 1.04 to 1.91 cm³. The test assumes independent samples and no strong outliers, and the dots in Figure 10 show no sign of either problem.

The test does not say why the rate rose, or whether the size of the effect matters biologically. The 44% rise over 10 °C (4.80 ÷ 3.33 = 1.44) is a number to compare with a published value for this enzyme, and that comparison belongs in your Conclusion.

A worked chi-squared test

A student surveys 40 quadrats on a shaded path edge and 40 in open lawn, and records whether a small plant species is present (invented counts).

The expected count for each cell is (row total × column total) ÷ 80, which gives 21 and 19. Then χ² is the sum of (observed − expected)² ÷ expected over the four cells: 2 × (36 ÷ 21) + 2 × (36 ÷ 19) = 7.22. With 1 degree of freedom, the critical value at the 5% level is 3.84, so 7.22 is well beyond it (p = 0.007). The conditions are met: the data are counts, the quadrats are independent, and every expected count is above 5. Presence and shade are associated, but that does not show shade is the reason, because moisture or trampling might differ too.

Decide on your test before you see the results, and never run several tests and report only the one that worked.

Uncertainties in the Biology IA: instruments and biological variation

Every measurement is a range, not a point. You record it as a ± value, carry it through each calculation, show it on the graph and use it to interpret your result. In Biology keep two sources apart. Instrument uncertainty is how well your equipment reads. Biological variation is how much the living things differ. The second is usually far bigger, and a good report says so.

Use the same conventions as the other sciences: the uncertainty has the units and decimal places of the value, usually to one significant figure, and comes from half the smallest division, the stated tolerance, or the spread of your repeats if that is larger. Round the result to the place of its uncertainty.

Typical instrument uncertainties

These are typical school values. Prefer the tolerance printed on your own equipment or in the manual.

A worked example: percentage change in mass

A potato cylinder has an initial mass of 1.84 g and a final mass of 1.62 g, each read on a balance good to ±0.01 g.

1. The change. 1.62 − 1.84 = −0.22 g. A difference adds the absolute uncertainties: ±0.02 g.
2. The percentage change. −0.22 ÷ 1.84 × 100 = −11.96%, written −12.0%.
3. Percentage uncertainties. 0.02 ÷ 0.22 = 9.1% on the change, and 0.01 ÷ 1.84 = 0.54% on the initial mass. Added for the quotient: 9.6%.
4. Back to absolute. 9.6% of 12.0 is about 1.2 percentage points, so the result is −12.0 ± 1.2%.

That is the balance’s effect on one cylinder. It says nothing about how much cylinders differ from each other, which needs the spread of the repeats.

Which one dominates?

Here is a comparison with invented yeast data. A gas syringe collects carbon dioxide for 5 minutes (300 s) from five independent flasks at 30 °C: 5.8, 6.9, 6.2, 7.4 and 5.7 cm³. The mean is 6.4 cm³ and the sample SD is 0.73 cm³.

The syringe reading is good to ±0.05 cm³, which is 0.8% of the mean, and the timing error of ±0.2 s in 300 s is 0.07%. The flasks differ by 11%. If you drew error bars from the syringe alone, you would show 0.8% where the data show 11%. The honest bars come from the SD. State that biological variation dominates, and your evaluation can then target it (more flasks, one yeast batch) instead of asking for a better syringe.

Instrument uncertainty still matters for a small difference of two masses, a short time or a tiny volume. Systematic errors, such as a thermometer that reads 1 °C high or a colorimeter that was not zeroed, are in neither number and belong in the evaluation.

Data analysis: from raw data to a graph

Good data analysis is mostly good housekeeping, done the same way every time.

1. Raw data first. Quantity and unit in each heading, the instrument uncertainty stated, and decimal places that match the instrument. Add observations that help: a badly cut cylinder, a leaking flask, the room temperature.
2. Process in steps. Turn raw values into the quantity your question needs: a percentage change, a rate per minute, a count per square metre. State the base, and give one worked example for each type of calculation.
3. Summarise each condition. Mean, SD and n for every value of the independent variable. Show how many samples you excluded and why.
4. Treat outliers openly. If you leave a value out, say which one, why (a leak, a damaged sample) and what difference it makes.
5. Graph the summary. Plot means with error bars, label the bars, and fit a line or curve you can justify.
6. Test where it helps. Add a test if your question needs one, as in the Statistics section.

Here is a processed table for the potato-cylinder example, with invented numbers: five cylinders at each of six sucrose concentrations.

Read Figure 12 as a reader would. The axes carry units and the bars say what they show. The means fall close to a straight line, so a straight fit is reasonable, and it is not forced through the origin, because nothing says it should pass through it. The line crosses zero at 0.43 mol dm⁻³, where the tissue neither gains nor loses mass overall. Regression statistics on all 30 cylinders put the uncertainty of that crossing at about ±0.01 mol dm⁻³. Say how you got it.

R² is 0.994 for the line through the means, and it is tempting to write “very reliable” beside it. Do not. R² shows how closely the means follow the line, not how much the cylinders vary or whether the fit holds beyond 1.0 mol dm⁻³. Fit a line only where the biology supports one: a saturating curve, such as photosynthesis against light, should not be forced to be straight.

How to write the conclusion

A conclusion earns its marks by being specific, honest and traceable. Figure 13 shows the parts.

Answer with numbers and uncertainty

Start with a sentence that has both variables and a result, then say whether the difference is larger than the variation allows, using your test or a comparison of the spread. If your error bars would let a flat line through the data, you cannot claim a trend. Use the language of evidence (“supports”, “is consistent with”) and say whether your prediction from Criterion A held.

Give the published science, with its source and conditions

Quote an accepted value or finding, the conditions it applies to (organism, temperature, pH, location), and a reference a reader could find: author or publisher, title, edition or database, and the page or table. A bare website name will not do, and a value for 37 °C does not test a result measured at 25 °C.

Percentage error against percentage uncertainty

These are different things. The percentage error measures how far you are from a published value. The percentage uncertainty measures what your data can justify, and in Biology that is often the spread between samples. Read them together. If your result is further from the accepted value than your spread explains, look for a systematic error and name it. If it is close, say it agrees within your uncertainty. A tiny difference is not proof of accuracy.

Explain the biology

Close the loop with a biological explanation that fits your data: enzyme and substrate collide more often as the temperature rises, or the cell loses water until its solute concentration matches the solution. If the data cannot show a cause, say so. And keep to your sample: a result from one species of cress in one dish is not a statement about all plants.