Core concepts
1. Observation vs Inference
An observation is what you directly detect with your senses or instruments — something you can point to in the data. An inference is an interpretation or conclusion you draw from those observations. Examiners mark these separately and penalise mixing them.
Worked example
- Observation: "The liquid turned blue-black when iodine solution was added."
- Inference: "Starch is present in the liquid."
- Observation: "The temperature rose from 22 °C to 35 °C over 5 minutes."
- Inference: "An exothermic reaction occurred."
⚠️Common trap: When a question says "State what you observe," write only what is directly seen, heard, or measured — never explain why. Adding "because…" to an observation answer costs marks.
2. Variables — IV, DV and Controlled
- Independent variable (IV) — the one factor you deliberately change. There is only one IV per experiment.
- Dependent variable (DV) — what you measure as a result of changing the IV.
- Controlled variables (CV) — all other factors kept constant so only the IV causes changes in the DV. Always name specific CVs with values or units, never just say "all other variables."
Example — testing how temperature affects enzyme activity
- IV: temperature of water bath (°C)
- DV: time for starch to be fully digested (seconds)
- CVs: volume of amylase solution (e.g. 2 cm³), concentration of starch solution (1%), pH of solution (pH 7)
✅A good hypothesis: "If [IV] increases, then [DV] will [increase/decrease] because [reason based on science]."
3. Precision, Accuracy, Reliability & Validity
- Precision — how finely an instrument can read. A thermometer with 0.1 °C divisions is more precise than one with 1 °C divisions. Smaller scale divisions = higher precision.
- Accuracy — how close a reading is to the true value. Improved by calibrating instruments and eliminating systematic errors.
- Reliability — how reproducible results are. Improved by repeating readings (≥3 times) and calculating the mean.
- Validity — whether the experiment genuinely tests the hypothesis. Requires all variables except the IV to be controlled.
⚠️Common trap: Confusing precision with accuracy. A balance not zeroed before use gives consistently wrong readings — precise but inaccurate. Repeating won't fix it.
4. Types of Error
- Random error — unpredictable, varies direction each time. Caused by slight fluctuations (e.g. reaction timing, reading a scale). Reduced by repeating and averaging.
- Systematic error — consistent offset, always in the same direction. Not reduced by repeating. Must be identified and corrected at the source (e.g. re-zero the balance, recalibrate thermometer).
- Parallax error — reading a scale at an angle instead of eye level. Causes a consistent reading error. Always read at eye level, perpendicular to the scale.
Examples
- Random: stopwatch reaction time varies ±0.2 s each trial
- Systematic: thermometer always reads 1.5 °C too high due to calibration fault
- Parallax: reading a meniscus from above — always records too high a volume
5. Identifying & Handling Anomalies
An anomalous result is one that clearly does not fit the pattern of the other results. Steps to handle correctly:
- Identify which reading is anomalous (state the actual value).
- State a plausible reason (parallax error, contamination, misread timer, equipment fault).
- Exclude it from the mean calculation — but keep it visible in the table and mark it clearly.
- State that the mean was calculated from the remaining consistent readings only.
Mean = Sum of valid readings ÷ Number of valid readings
Worked example
Readings: 21.0 cm, 21.2 cm, 27.9 cm, 21.1 cm
Anomaly: 27.9 cm (significantly higher than others — possible parallax error).
Mean = (21.0 + 21.2 + 21.1) ÷ 3 = 21.1 cm (rounded to same d.p. as readings)
⚠️Common trap: Silently dropping an anomaly or including it in the mean without comment. Examiners expect an explicit reason for any excluded value.
6. Describing Graph Trends
Step-by-step structure
- State the overall relationship (positive/negative correlation, directly proportional, no relationship, levels off).
- Describe the nature of the trend (linear, non-linear/curved, steep then flattening).
- Quote specific values with units from the graph to support your description.
- Note any anomalous points or changes in gradient.
Useful language
- "As X increases from ___ to ___, Y increases from ___ to ___."
- "The relationship is linear between ___ and ___."
- "Beyond ___, Y levels off / begins to decrease."
- "There is a positive/negative/no correlation between X and Y."
⚠️Common trap: "X goes up so Y goes up." Too vague — no marks. You must state the direction, nature (linear or curved), and cite at least two data points with units.
7. Correlation vs Causation
Two variables may correlate (rise or fall together) without one causing the other. A confounding variable — a third factor — may be causing changes in both.
Classic example
Ice cream sales and drowning rates both rise in summer. Conclusion: ice cream causes drowning? No — the confounding variable is hot weather, which independently causes both.
To establish causation, you need a controlled experiment where only the IV is changed. Survey or observational data alone cannot prove causation.
✅Always end correlation-based conclusions with: "This shows a correlation only. A controlled experiment is needed to establish causation."
8. Evaluating a Conclusion
For any given conclusion, ask:
- Does the data range actually cover the claim? (Extrapolating beyond the tested range is risky.)
- Were there enough repeats to trust the mean?
- Were all relevant variables controlled (validity)?
- Is the sample size large enough to generalise?
- Were anomalies handled correctly, or were they ignored?
- Does the conclusion match only what the data shows, or does it overreach?
9. Suggesting Improvements
Always link each improvement to the specific problem it solves:
- "Increase the number of repeats from 3 to 5" → improves reliability, reduces effect of random error.
- "Use a thermometer with 0.1 °C divisions instead of 1 °C" → improves precision of temperature readings.
- "Control [specific variable] by [specific method]" → improves validity of the fair test.
- "Extend the range of IV from 20–60 °C to 20–80 °C" → strengthens the conclusion by covering a wider range.
- "Use a data logger instead of a stopwatch" → reduces reaction-time error, improving accuracy.
⚠️Common trap: Saying "repeat the experiment" without specifying how many times or what you're improving. Always state the number and the reason.
10. Command Words — quick reference
- State — give a fact; no reason needed.
- Describe — say what you observe or what happens; no "because."
- Explain — give a reason using "because"; the mechanism is required.
- Compare — give at least one similarity AND one difference.
- Suggest — give a possible reason; apply knowledge to an unfamiliar context.
- Evaluate — judge quality/reliability; weigh evidence and give a reasoned conclusion.
- Predict — use existing data/knowledge to say what will happen next.
- Calculate — show working, state formula, substitute values with units, give answer with units.
Worked practice questions
Practice Q1
Four readings of volume are: 12.1 cm³, 12.0 cm³, 18.9 cm³ and 12.2 cm³. (a) Identify the anomalous result and suggest a reason. (b) Calculate the correct mean.
Answer
(a) 18.9 cm³ is anomalous — it is significantly higher than the other three consistent readings. Possible reason: parallax error when reading the measuring cylinder, or a contaminated sample.
(b) Mean = (12.1 + 12.0 + 12.2) ÷ 3 = 12.1 cm³
Practice Q2
A student claims that students who eat breakfast score higher in exams, based on a survey of 200 students. Evaluate this claim.
Answer
The claim cannot be established from survey data alone — this shows a correlation, not causation. Confounding variables such as family income, sleep habits, or general health may independently explain both breakfast eating and exam performance. A controlled experiment would be needed to establish causation. The sample size of 200 is reasonable, but the survey method cannot control for other variables.
Practice Q3
A graph shows reaction rate (y-axis) vs temperature (x-axis). The curve rises steeply from 10 °C to 37 °C, then falls sharply above 40 °C. Describe the trend and explain the fall.
Answer
Describe: Reaction rate increases as temperature rises from 10 °C to 37 °C, reaching a maximum at 37 °C (optimum temperature). Above 37 °C, reaction rate decreases sharply, falling toward zero by approximately 50 °C.
Explain: Above the optimum temperature, the enzyme begins to denature — the active site permanently changes shape, so the substrate can no longer bind. Enzyme activity falls as more enzyme molecules are denatured.
Practice Q4
A student measures pulse rate before and after exercise. Name two variables that must be controlled, and explain why she should repeat measurements three times.
Answer
Controlled variables (any two): type and intensity of exercise (e.g. same running speed on a treadmill); duration of exercise; time of day measurements are taken; rest period before measurement; age and fitness level of the participant.
Why repeat: Repeating three times and calculating the mean reduces the effect of random error (e.g. slight variation in how rested the student is) and improves the reliability of the results.
Must-know checklist
- I can distinguish observation from inference in any given scenario.
- I can identify IV, DV and at least two specific CVs (with units) in any experiment.
- I can distinguish random error from systematic error and state how to reduce each.
- I can identify an anomaly, give a plausible reason, and calculate the mean without it.
- I can describe a graph trend with specific values, correct vocabulary, and note shape changes.
- I can explain why correlation does not prove causation and identify confounding variables.
- I can evaluate a conclusion against data provided and identify overreach.
- I can suggest specific, justified improvements linked to the weakness they address.
- I know what each command word (state, describe, explain, compare, evaluate) requires.