Paper 3 question 2 gives you a source with numbers in it and asks you to analyse the findings and state a conclusion linked to the claim.
The trap is that most of the work is not mathematical. You are not being asked to calculate anything. You are being asked to read what the numbers say, and then check whether the researchers' conclusion goes further than their own data supports.
That second half is where the marks are.
The two layers of numbers
Descriptive statistics describe the sample in front of you. Means, medians, percentages, standard deviations. They tell you what happened.
Inferential statistics ask how unusual that pattern would be if there were really no effect in the population. They are what lets researchers say something beyond the people they tested.
Question 2 sources often report both. Knowing which layer you are looking at stops you making claims from one that only the other could support.
Reading the middle and the spread together
A mean on its own is half a description.
The mean is the arithmetic average, and it is pulled by extreme values. The median is the middle score, and it is not. The mode is the most frequent value.
The standard deviation describes the typical spread of scores around the mean. A larger SD means more variability. Two groups can have nearly identical means and completely different spreads, which means they differ in consistency rather than in typical performance.
Whenever a source gives you a mean with no spread, that gap is worth a sentence. Whenever it gives you both, compare them, because a large SD relative to the difference between groups tells you the groups overlap heavily.
Two effects worth naming when you see them. A ceiling effect is when most participants score at or near the maximum, so real differences are hidden and the task was too easy. A floor effect is the reverse.
Reading a results table
Tables look intimidating and are usually simpler than they appear. Work through them in a fixed order.
Read the caption and work out what each row and column actually is. Find the sample size, and check whether it is given per group. Find the descriptive statistics, usually means with standard deviations. Find any test result and its associated numbers. Then, before you interpret anything, ask what is missing.
What is missing is often the point. No sample size per group. No spread. No effect size. No comparison condition.
Findings and conclusions are not the same thing
This is the distinction question 2 is really testing.
A finding is what the data shows. A conclusion is what the researchers say it means. There is always a gap between them, and sometimes the gap is too wide.
Language sits on a ladder, and the rung a researcher chooses should match the design they used.
| Rung | Language | Justified when |
|---|---|---|
| Description | scored higher, differed, was associated with | Any data at all |
| Association | is related to, predicts, is linked with | Measured variables, nothing manipulated |
| Tentative cause | may contribute to, is consistent with a causal role | A design that partly rules out alternatives |
| Cause | causes, leads to, improves, damages | Manipulation, control, and plausible alternatives ruled out |
When a source reports a correlation and concludes that one thing improves another, the researchers have climbed two rungs without the evidence to stand on them. Naming that precisely is worth more than a general remark about causation.
Claim surgery
Before you write anything, read the researchers' conclusion word by word.
Circle the causal verbs: causes, leads to, improves, damages. Underline the population words: adolescents, students, people. Box the broad constructs: memory, wellbeing, engagement.
Then check each part against the design. Does this method license that verb? Does this sample license that population? Does this measure license that construct?
Most top-band points in question 2 come from that comparison, and it takes under a minute.
Then rewrite the conclusion at three levels: the exact result, the strongest justified inference, and the overclaim. The gap between the second and the third is your answer.
How to write the answer
Question 2 is worth 6 marks and wants two things: analysis of the findings, and a conclusion linked to the claim.
State what the data actually shows, using the numbers specifically rather than saying the groups differed. Then say what that supports, and where the source's own conclusion goes beyond it.
A pattern that works: the finding in precise terms, the limitation on what it can support, and the conclusion restated at the size the evidence carries.
Do not accept a source's own conclusion just because it is printed. Researchers overreach, and spotting it is worth real marks.
Two habits worth building now
Always ask what is not shown. Missing spread, missing sample size, missing comparison group, missing effect size. Absence is evidence about how much confidence a finding deserves.
Always separate size from significance. A result can meet a statistical threshold and still be too small to matter, and a real effect can be missed entirely by a small study. Those are two different questions, and the vocabulary for them is covered in the article on p-values, significance and correlation.
One thing SL students should know
Data analysis is studied at both levels, but it is only formally assessed at HL through Paper 3. No Paper 2 question at either level asks you to calculate or formally interpret statistics.
That does not make it irrelevant at SL. Paper 2 Section B hands you an unseen study, and being able to read whatever it reports is what makes your evaluation specific rather than generic.
If you want the full notation decoder, worked results tables and graded answers, that is what the PsychExcel Paper 3 Complete Guide is built for.