How to Answer Data Analysis and Evaluation Questions in A Level Biology (9700) Paper 5

Paper 5 in Cambridge International AS & A Level Biology (9700) is split almost evenly between planning and the skills that come after data have been collected. Under the current 2025-2027 syllabus, Paper 5 is a 1 hour 15 minute written paper worth 30 marks and 11.5% of the full A Level. Around 14-16 marks are allocated to planning and another 14-16 to analysis, conclusions and evaluation. NeuraLearn already has a separate guide to Paper 5 planning; this guide focuses on the other half: processing data, choosing and interpreting statistics, drawing conclusions from evidence and evaluating how much confidence the investigation deserves.

Start with the data, not with the biological story

When the context is familiar, it is tempting to explain what you expect before you have described what the results actually show. Paper 5 rewards the opposite order. First identify the pattern in the data. Then quantify it. Then decide whether the variation or statistical analysis supports the comparison. Only after that should you give a biological explanation. Cambridge's June 2024 examiner report recommends highlighting and annotating the information supplied, identifying key points and looking for trends before answering. This is particularly useful when a question combines a graph, a results table and statistical information. Treat those sources as one data set rather than answering from whichever number catches your eye first.

Click to enlarge

Describe the complete trend, including groups and exceptions

A trend answer should cover the variables and groups named in the question. If a graph shows two species responding to temperature, saying 'conductance decreases as temperature increases' may be incomplete if the question expects you to recognise that the trend occurs in both species. Cambridge's 2024 report repeatedly notes cases where candidates identified the direction of a trend but lost credit because they did not apply it to all relevant groups. Also notice plateaus, optima, thresholds and departures from the overall pattern. Do not describe a curve as increasing if it increases only to a particular value and then levels off or falls.

Quote data only when the numbers prove the point

A good data quote is selective. It uses the values that directly support the conclusion and identifies the relevant groups or conditions. Quoting four unrelated numbers does not strengthen an answer if the comparison between them is unclear. Read axes, units, error bars and table headings carefully. Cambridge's report warns that candidates can lose otherwise valid conclusions by quoting inaccurate values from a graph. Where a question asks you to justify a conclusion, use the smallest number of well-chosen figures needed to show the pattern.

Show every calculation clearly

Paper 5 can ask for rates, percentage change, standard deviation, standard error, confidence intervals and partly completed statistical calculations. Cambridge's examiner report explicitly says candidates should show their working because credit may still be available when the final answer is incorrect. Check the required significant figures or decimal places before you finish. In the June 2024 Paper 51 report, a calculated rate had to be given to three significant figures, and a t value had to match the significant figures specified in the question. A correct method can still lose the final mark if the requested precision is ignored.

Know what standard deviation, standard error and confidence intervals are telling you

Standard deviation describes the spread of observations around the mean. Standard error describes uncertainty in the estimate of the mean and becomes smaller as sample size increases, all else being equal. The syllabus also uses 95% confidence intervals, approximated as the mean plus or minus two standard errors. Do not use the word 'reliable' automatically whenever you see a small error bar. Ask what the statistic represents. A small standard error indicates a more precise estimate of the mean; it does not by itself prove that the entire method was valid or accurate. Likewise, overlapping or non-overlapping intervals can inform whether a difference is likely to be meaningful, but a formal statistical test is stronger evidence when one is provided or required.

Choose a statistical test from the purpose and data type

The current syllabus requires four named tests: chi-squared, the t-test, Pearson's linear correlation and Spearman's rank correlation. Do not memorise them as four interchangeable formulas. Start with the question being asked. Use chi-squared to test whether observed and expected frequencies of nominal data differ significantly. Use a t-test to test the significance of a difference between two samples of continuous data when the populations are normally distributed and their standard deviations are approximately the same. Pearson's linear correlation is for two sets of normally distributed continuous data where a scatter graph suggests a linear relationship. Spearman's rank correlation is for ranked or ordinal data, or data converted to ranks, and is useful where the data are not normally distributed.

Click to enlarge

Write the null hypothesis to match the test

For a t-test or chi-squared test, the null hypothesis concerns no significant difference between the relevant groups, frequencies or distributions. For Pearson or Spearman correlation, it concerns no significant correlation between the two variables. Cambridge's June 2024 report highlighted a common error where candidates wrote 'no correlation' for a t-test question that actually compared two means. Name the variables or groups specifically. 'There is no significant difference' is weaker than 'there is no significant difference in mean percentage leaf death between treatment A and treatment B'.

Interpret the calculated value against the correct critical value

Do not stop after calculating t, chi-squared, r or rs. The purpose of the statistic is to make a decision. Use the correct degrees of freedom where required, identify the probability threshold specified or conventionally tested, and compare the calculated statistic with the critical value in the table supplied. In a June 2024 t-test question, the relevant threshold was p = 0.05. Strong responses compared the calculated t with the correct critical value, accepted the null hypothesis because the calculated value was smaller, and concluded that there was no significant difference. Candidates who selected the p = 0.10 critical value lost credit even if the rest of the method was sound. Focus only on the probability threshold relevant to the question rather than discussing every critical value in the table.

Do not confuse correlation with causation

A significant Pearson or Spearman result supports an association between two variables. It does not prove that one variable causes the other. A third variable may influence both, the relationship may be indirect, or the observational design may not establish causation. Your conclusion should match what the test can legitimately establish. 'There is a significant negative correlation between X and Y' is statistically defensible. 'Increasing X causes Y to fall' requires experimental evidence capable of supporting causation.

Make conclusions that combine pattern, evidence and biology

The syllabus expects candidates to summarise the main conclusions, identify key points in raw and processed data, discuss the extent to which a hypothesis is supported and give detailed scientific explanations. A strong conclusion therefore has three layers: what happened, how the data support that statement and why the pattern is biologically plausible. Suppose an experiment finds a higher mean enzyme rate at 30 °C than at 20 °C. A developed answer would identify the increase, quote suitable values if requested, use the supplied statistical information to say whether the difference is significant, then explain the change using kinetic energy, collision frequency and enzyme-substrate complex formation. The biological explanation should follow from the evidence rather than replace it.

Use the wording 'significant' carefully

In everyday language, 'significant' can mean large or important. In Paper 5 statistics it has a specific meaning linked to probability. Cambridge's June 2024 report warns against the term 'insignificant' when discussing statistical data. Prefer 'the difference is not statistically significant' or 'there is no significant difference' when that conclusion is supported by the test. Do not describe a visibly large difference as statistically significant unless the statistical evidence supports that statement.

Evaluation is about confidence in the conclusion

A good evaluation is not a generic list of laboratory weaknesses. The syllabus asks you to judge anomalies, replication, the range and intervals of the independent variable, the method of measuring the dependent variable, how effectively variables were controlled, the validity of the investigation and how much confidence can be placed in the conclusions. That means each evaluation point should connect a limitation to its effect. 'Only one volunteer was used' is a limitation. 'Only one volunteer was used, so the result may reflect that individual's physiology and may not be representative of the wider population' is an evaluation.

Click to enlarge

Do not use 'repeat it' as a universal improvement

Repeats are valuable when random variation is important because they allow a mean to be calculated and help reduce the influence of anomalous readings. But repeats do not fix every problem. Cambridge's 2024 examiner reports specifically warn against treating repeating as a route to accuracy. Accurate data depend on appropriate and sufficiently precise measurement, while repeated measurements primarily help assess consistency. If the main problem is a systematic bias, repeated measurements using the same flawed method can reproduce the same bias. Improve the measurement method, calibration or control instead.

Distinguish a control from a standardised variable

A control provides a comparison that helps isolate the effect being investigated. A standardised variable is kept constant across the experimental treatments so it does not confound the result. Cambridge explicitly highlighted confusion between these ideas in the June 2024 Paper 51 report. If a mosquito-repellent investigation includes a treatment with no repellent, that is a control condition. Keeping the exposure time or mosquito species the same is standardisation. Do not call every constant condition a control.

Original example: interpreting a t-test

Imagine two plant treatments produce mean growth values of 14.2 mm and 17.8 mm. A t-test gives t = 2.46. For the correct degrees of freedom, the critical value at p = 0.05 is 2.10. Because the calculated t is greater than the critical value, the null hypothesis is rejected at the 5% level and the difference between the mean growth values is statistically significant. The conclusion should not stop there. You would then return to the experimental context and state which treatment had the greater mean growth and, if asked, explain the biological reason that could account for the difference. If the experimental design had a major limitation, the evaluation should still discuss how much confidence to place in that conclusion.

Common Paper 5 analysis and evaluation mistakes

1. Explaining the biology before accurately describing the pattern in the data.
2. Giving a trend that ignores one of the groups named in the question.
3. Quoting several numbers without making the comparison they are supposed to support.
4. Hiding calculation working and losing possible method credit after a numerical error.
5. Using 'no correlation' as the null hypothesis for a test that actually compares two means.
6. Choosing a statistical test because its name is familiar rather than because the data and purpose fit it.
7. Using the wrong probability threshold or critical value when interpreting a statistical test.
8. Treating correlation as proof of causation.
9. Calling a difference statistically significant without statistical evidence.
10. Listing limitations without explaining how they weaken validity or confidence.
11. Suggesting more repeats when the real problem is systematic error or an inappropriate measurement method.
12. Confusing a control treatment with a variable that is kept constant.

A reliable Paper 5 analysis routine

1. Annotate the table, graph and written information before answering.
2. State the full pattern, including relevant groups, turning points and anomalies.
3. Use one or two precise data quotes to support the comparison.
4. Carry out the necessary calculation and show the working clearly.
5. Identify the type of data and choose the statistical test that matches the question.
6. Write a null hypothesis that matches difference or correlation appropriately.
7. Compare the calculated statistic with the correct critical value and probability threshold.
8. State what the statistical result allows you to conclude without claiming causation unless the design supports it.
9. Add the biological explanation only after the data conclusion is established.
10. Evaluate the investigation by linking each limitation to its consequence and each improvement to the weakness it addresses.

Put it into practice

Take one Biology (9700) Paper 5 question containing a graph, statistical test and evaluation section. Attempt the analysis without the mark scheme, then review it in four passes: data pattern, calculation, statistical conclusion and evaluation. If the conclusion is unsupported, add the missing evidence. If an improvement is generic, rewrite it so it addresses one named limitation. Inside NeuraGeek, practise Biology (9700) Paper 5 planning and analysis/evaluation as separate skill sets.