Section B gives you a study you have never seen, 15 marks, and about 40 minutes. You have to discuss it with reference to two or more of the concepts the question specifies.
The instinct is to look for three strengths and three limitations. Resist it. That approach produces a list, and a list stalls in the middle bands however accurate it is.
The question underneath Section B is different. Given how this evidence was produced, how much confidence should I place in the researchers' conclusion, and why?
Everything below is built to answer that.
Diagnose the study before you choose anything
The most common mistake is jumping straight to a memorised evaluation point. Read the stimulus once, then extract six things in order.
| What to extract | Why you need it | |
|---|---|---|
| Aim | What are the researchers trying to find out? | Tells you the intended scope |
| Method | Experiment, survey, interview, observation, or another design? | Determines what inferences are possible |
| Sample | Who, how many, how recruited? | Feeds bias, generalisability and responsibility |
| Measurement | What constructs are claimed, and what was actually measured? | Feeds measurement, bias and perspective |
| Findings | What result was actually obtained? | Stops you discussing method without engaging with evidence |
| Conclusion | What do the researchers claim or recommend? | This is the statement your analysis must calibrate |
Two minutes on this changes the whole answer, because it stops you evaluating a generic study instead of the one in front of you.
Read the conclusion word by word
This is the highest-yield thirty seconds in the section.
Circle the causal verbs: causes, leads to, improves, damages. Underline the population words: adolescents, people, students. Box the broad constructs: memory, wellbeing, relationships.
Many top-band points come from asking whether the design, sample and measurement justify each part of that specific wording.
Then rewrite the conclusion three times: the exact result, the strongest justified inference, and the overclaim. Whatever gap sits between the second and the third is your argument.
Choose concepts by yield, not by comfort
You need at least two. Do not pick the two you revised most thoroughly. Pick the two that let you say the most important things about this particular study.
Usually one concept interrogates the design and another interrogates what the evidence means.
Score each candidate concept on four things. Does the study give you specific evidence for it, or only a vague gesture. Can you trace a clear methodological consequence, or does it stay a label. Does it open a genuinely different angle, or repeat what another concept already said. And does it change the strength, scope or certainty of the claim.
Some pairings tend to be powerful. A questionnaire-based correlation usually rewards measurement and causality. A semi-structured interview rewards perspective and bias. A longitudinal or developmental study rewards change and measurement. Deception or vulnerable participants brings responsibility into play alongside bias.
Two deeply used concepts will outperform four shallow labels.
The paragraph engine
Every analytical paragraph should move through four stages. Learn the movement, not a script.
Locate. Point at a specific feature of the study. "The DV was operationalised as..."
Explain. Say what that feature does, mechanically. "This means that..."
Complicate. Consider the rival explanation or the counterargument. "However... An alternative reading is... This would matter less if..."
Calibrate. State what the evidence now supports. "The researchers can therefore claim... but not..."
A paragraph that stops after Explain has described a feature. A paragraph that reaches Calibrate has judged it. That difference is usually two more sentences rather than more knowledge, and it is what separates analysis that is carried through from analysis that stops early.
The same point at three depths
Take one feature: a study measured wellbeing with a single question.
Weak. The study used a questionnaire to measure wellbeing.
Better. The study measured wellbeing with one question, which is a weakness because one question is not very reliable.
Strong. Wellbeing was operationalised as a single item, which means the construct is being represented by one response on one occasion. Because a single item cannot capture the different dimensions the researchers name in their conclusion, and because it is unusually sensitive to mood on the day, the measure is likely to be unstable. So even if the group difference is real, it is not clear what construct the difference is in. The researchers can claim a difference in reported wellbeing on this item. They cannot claim a difference in wellbeing as they define it.
Count what the third one does. It locates the feature, explains the mechanism, names the specific technical weakness, offers the alternative reading, and calibrates the claim.
The question to ask after every point
Whenever you identify a feature of a study, push one step further.
So what does this change about the evidence? And therefore what should happen to the wording of the conclusion?
If you cannot answer both, the point is still descriptive.
Make two concepts talk to each other
At least once per answer, write a sentence where one concept sharpens the other. This is one of the most reliable ways to lift an answer.
For example: even with strong experimental control, a weak dependent variable means the study may have caused a change in the measured proxy rather than in the broad construct named in the conclusion. That is causality and measurement working together, and neither concept says it alone.
Do not simply mention the second concept in the final sentence. That reads as an afterthought because it is one.
Include one genuine strength
Discussing quality means discussing quality, not only failure.
Strong analysis may praise a design feature, qualify a claim, expose a missing perspective, or show an ethical and methodological trade-off. A control that was actually done well, explained properly, earns full analytical credit: name the control, explain the mechanism, identify the threat it addresses, and state what it does for the causal claim.
An answer that only attacks the study reads as a checklist. An answer that says what the study does establish, and then where it stops, reads as judgement.
The conclusion
A conclusion is not a summary. It is a decision about what the evidence justifies.
Four moves make it work.
| Part | The question it answers |
|---|---|
| Strength | How convincing is the evidence? |
| Scope | For which population, context or behaviour? |
| Certainty | Cause, association, possibility or description? |
| Qualification | What would have to change or be replicated before a stronger claim? |
One conclusion to never write: "Overall the study has strengths and limitations." It says nothing. If your conclusion would make sense to somebody who had not read the study, it is too generic.
Two structural points
Do not write an introduction that restates the study. The examiner has read it. Start with your first analytical point.
And there is no magic number of paragraphs. The mark band is holistic, judged on best fit across the whole response. Four developed paragraphs are not automatically better than three excellent ones, and five concepts are not automatically better than two.
How to practise
Do not write six full answers in a row. That is expensive and it improves you slowly.
Better: take an unseen study, complete the six-point diagnosis, choose two concepts, and write one paragraph that reaches Calibrate. Do that four times in the time one full essay would take. You will correct four habits instead of one.
Then build up. Remove the scaffolds in stages. First with the diagnosis table and concept prompts available. Then with only the concept list. Then with only the stimulus and question, planning in six to eight minutes. Then a full timed 40 minutes with nothing.
Before you finish any practice answer, run this check. Did I pick concepts by yield rather than comfort. Does every paragraph contain a specific detail from the stimulus. Does every paragraph reach Calibrate. Is there at least one sentence where two concepts interact. Did I include one genuine strength. Does my conclusion state both what the study can claim and where it overreaches.
If you want unseen studies with the support removed in stages, plus worked models at each band, that is what the PsychExcel Paper 2 Student Guide is built around.