Paper 5 is a written paper worth 30 marks with no apparatus: one question asks you to plan an investigation, the other hands you results to process and evaluate. Work through eight sets built from what the examiner reports say candidates get wrong: naming the variables and a control, dilution tables, a method in order with a risk assessment, means and rates with the anomaly left out, a best-fit line and a gradient, standard error and 95% confidence limits, a chi-squared test with its null hypothesis, and the difference between reliability, validity and accuracy. Every question uses fresh numbers, is marked as you go, and has a worked solution. Exam mode mixes the sets with no feedback until the end. Allow 10 to 15 minutes per set.
This trainer needs JavaScript.
The independent and dependent variables named with how they are changed and measured, the key variables standardised with a method, a prediction linked to the biology, a real control experiment (not a repeat), at least five values of the independent variable over a sensible range, and a risk assessment that names the hazard, judges the risk and gives a matching precaution.
A table with headings and units for every dilution, the difference between serial and proportional dilution (9700 June 2025 examiner report: candidates confuse them), solutions equilibrated in a water bath before mixing (the report rejects “acclimatised”), and replicates justified by a mean and the spread, not by “finding anomalies” (June 2024 report).
Means with the anomaly left out, rates as 1 ÷ time, percentage changes with their sign (June 2024 report: the minus sign on a decrease was often missing), consistent decimal places, SE = s ÷ √n, 95% CI = mean ± 2SE, and a graph read with precision and units.
A null hypothesis of no difference, expected numbers (not ratios), each (O − E)² ÷ E, the statistic to 2 or 3 significant figures, the right degrees of freedom, and a conclusion that compares with the p = 0.05 critical value and says what happens to the null hypothesis (June 2023 report: many rejected a null hypothesis their value supported).
Specific limitations and matching improvements that increase confidence, reliability (repeatable) kept apart from validity (the right thing measured, other variables controlled) and accuracy (the instrument), percentage error from an uncertainty, and systematic errors (shift everything, keep the trend) told from random ones.
The marking follows the conventions in published Cambridge mark schemes. The experiments, readings and wording on this page are our own.