OxfordAQA International A-level Psychology (9685): 3.3.3 Research methods 2 – Study Guide

This guide covers section 3.3.3 Research methods 2 in the OxfordAQA International AS and A-level Psychology (9685) specification, Version 3.1 (International A-level exams May/June 2020 onwards). The section sits in Unit 3: Advanced topics and research methods 2. The specification says Unit 3 is a 1 hour 30 minute written exam worth 90 marks and 30% of the International A-level, and that Section C, Research methods 2, carries 30 marks of compulsory questions.

Use it with the revision notes for Research methods 2 and the Research methods 2 practice questions. The 9685 course hub and 9685 checklist show the rest of the course. The specification says you must also show the knowledge from Research methods 1, so keep the Unit 2 study guide close.

Coverage table

Syllabus reference	What you must know or be able to do (from the document)	Where assessed
3.3.3 Methods	Content analysis; case studies	Unit 3, Section C
3.3.3 Scientific processes	Reliability across all methods; test-retest and inter-observer reliability	Unit 3, Section C
3.3.3 Scientific processes	Face, concurrent, predictive and ecological validity; the use of correlation in assessing validity	Unit 3, Section C
3.3.3 Scientific processes	Designing and reporting investigations; abstract, introduction, method, results, discussion, referencing	Unit 3, Section C
3.3.3 Data handling	The purpose of statistical testing; probability and significance; statistical tables and critical values; Type I and Type II errors	Unit 3, Section C
3.3.3 Data handling	Nominal, ordinal and interval data; choosing a test by aim, level of measurement and design	Unit 3, Section C
3.3.3 Data handling	When to use Spearman’s rho, Pearson’s r, Wilcoxon, Mann-Whitney, related t-test, unrelated t-test, Sign test and Chi-squared	Unit 3, Section C

The specification states that you will not carry out statistical tests in the exam. You should be able to comment on the use of named tests and interpret their results. It also says research methods, practical research skills and mathematical skills are assessed in Unit 3.

Content analysis

Content analysis is a way of studying behaviour indirectly, through things people have produced. These include adverts, social media posts, diaries, interview transcripts, films and children’s drawings.

Process.

1. Choose a sample of material (for example, every post on a forum in one month).
2. Create coding categories. These work like behavioural categories in an observation.
3. Read or watch the material and tally each time a category appears. This gives quantitative data.
4. In thematic analysis, the researcher looks for themes that emerge from repeated reading. This gives qualitative data.

Key terms

• Coding: putting units of material into categories.
• Coding unit: what is counted, such as a word, an image or a whole post.

Strengths. The material already exists, so it is cheap and easy to access. Ethical issues are often small when the material is public. Findings can be checked, because another researcher can re-code the same material.

Limitations. Coders interpret meaning, so bias can creep in. The material may be out of context, as you do not know why a post was written. Tallies show how often something appears but not why.

Reliability check. Two coders code the same material separately. Their tallies are compared, which is inter-observer reliability applied to content analysis.

Worked paragraph. A fictional researcher codes 200 food adverts for “health claims” and “taste claims”. Content analysis suits this aim because adverts are permanent and public, so no consent is needed (knowledge, application). Tallying claims converts adverts into frequencies that can be compared (analysis). However, deciding whether “light and fresh” is a health or taste claim needs judgement, so a second coder should code a sample, and agreement should be checked (evaluation).

Case studies

A case study is an in-depth investigation of one person, group, institution or event. It usually uses several methods, such as interviews, observation and psychological tests. Many case studies are longitudinal and follow the case over time.

Henry Molaison (HM). In 1953 the surgeon William Scoville removed parts of HM’s medial temporal lobes on both sides, including much of the hippocampus, to treat severe epilepsy. Scoville and Milner (1957) reported that HM was left with severe amnesia for events after the operation. Later testing found his immediate memory span was normal, and Milner showed he improved at a mirror-drawing task without remembering having practised it.

Phineas Gage. In 1848 an iron rod passed through Gage’s skull in a railway construction accident. He survived. His physician, John Harlow, reported changes in his personality. Later historians have argued that many popular details of the case were exaggerated.

Strengths.

• Rich, detailed data, often both qualitative and quantitative.
• Allow study of rare cases that could never be created ethically, such as brain damage.
• A single case can challenge a theory. HM’s normal short-term span alongside lost long-term learning supported separate memory stores.

Limitations.

• One case may not generalise to others.
• The researcher may become close to the participant, which risks bias in interpretation.
• Data about the past (retrospective data) rely on memory or old records, as in the Gage case.
• Cases cannot be repeated, so findings are hard to replicate.

Worked paragraph. HM’s case supports the view that the hippocampus is needed to form new long-term memories (knowledge). The surgery was precisely recorded and he was tested over decades (application). However, HM had epilepsy before the operation, so his brain may not have been typical (analysis). This means findings from one person must be checked against other patients and brain imaging before being generalised (evaluation).

Reliability

Reliability means consistency. A reliable measure gives the same result if the study or test is repeated in the same way.

Test-retest. Give the same test or questionnaire to the same people on two occasions, with a gap long enough that they cannot recall their answers but short enough that the trait has not changed. Correlate the two sets of scores.

Inter-observer reliability. Two or more observers record the same behaviour separately, using the same categories and time sampling. Their tallies are correlated.

Interpreting the correlation. A strong positive correlation shows good reliability. A rule of thumb used in many textbooks is +0.80 or above. The specification does not give a figure, so your teacher will confirm what is expected.

Improving reliability.

• Questionnaires: rewrite ambiguous items, or replace open questions with closed ones.
• Observations: operationalise categories more clearly and train observers together.
• Experiments: standardise procedures and instructions.
• Content analysis: tighten coding definitions and pilot them.

Validity

Validity means accuracy. A valid measure measures what it claims to measure, and valid findings can be generalised beyond the study.

Type	Meaning	How it is assessed
Face validity	The measure looks, on the surface, as if it measures what it should	An expert inspects the items
Concurrent validity	Scores agree with an established measure of the same thing	Give both measures; correlate the scores
Predictive validity	Scores predict a later behaviour or outcome	Correlate scores with the later outcome
Ecological validity	Findings generalise to real-life settings	Compare with field studies; judge how natural the task and setting were

Correlation in validity. For concurrent and predictive validity, a strong positive correlation between the new measure and the comparison supports validity.

Worked paragraph. A fictional company designs a 20-item test of nursing aptitude. To check concurrent validity, applicants could also complete an established aptitude test, and the two sets of scores could be correlated (knowledge, application). To check predictive validity, scores could be correlated with supervisors’ ratings after a year in post (application). A strong positive correlation in each case would suggest the test measures aptitude (analysis). A weak link with job ratings might reflect training received after selection rather than a poor test, so one correlation is not proof (evaluation).

Designing and reporting investigations

Design choices come from Research methods 1: aim, hypothesis, sampling, design, variables, controls and ethics. Research methods 2 adds how the study is written up.

Section	What it contains
Abstract	A short summary of the aim, hypothesis, method, results and conclusion, so a reader can decide whether to read on
Introduction	Background theory and past research, leading to the aim and hypotheses
Method	Design, participants and sampling, materials, procedure and ethics, in enough detail to replicate the study
Results	Descriptive statistics, graphs and tables, and the inferential test with its outcome
Discussion	What the results mean, links to past research, limitations, improvements and real-world applications
Referencing	Full details of every source cited

A reference usually lists author(s), year, title, journal, volume and page numbers. For example: Scoville, W. B. and Milner, B. (1957). Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery and Psychiatry, 20, 11-21. Your teacher will confirm the format they want.

Probability, significance and inferential tests

Purpose. Descriptive statistics show a difference or correlation in the sample. An inferential test tells you how likely it is that this result happened by chance. That decides whether you can accept the alternative hypothesis.

Null hypothesis. This states there is no difference or no correlation. The test tells you whether to reject it.

Probability (p). Psychologists usually use p ≤ 0.05: there is 5% or less probability that the result is due to chance. A stricter level, p ≤ 0.01, is used when the risk of a wrong conclusion matters more, such as a study of a new treatment.

Critical values. Each test produces a calculated value. You compare it with a critical value in a statistical table. To find the right critical value you need:

• the significance level (usually 0.05);
• whether the hypothesis is directional (one-tailed) or non-directional (two-tailed);
• the number of participants (N) or degrees of freedom (df).

For most tests the calculated value must be equal to or greater than the critical value. For Mann-Whitney, Wilcoxon and the Sign test it must be equal to or less than the critical value. A memory aid: tests with an “r” in the name (Spearman’s rho, Pearson’s r, unrelated t-test, related t-test, Chi-squared) need a calculated value that is greater. Each table usually prints this rule underneath.

Type I error. Rejecting the null hypothesis when it is true: a false positive. More likely if the level is too lenient, such as 0.10.

Type II error. Accepting the null hypothesis when it is false: a false negative. More likely if the level is too strict, such as 0.01. Using 0.05 balances the two risks.

Levels of measurement

• Nominal: data in separate categories; each person counts in one category only. Example: whether each driver stopped or did not stop at an amber light.
• Ordinal: data that can be put in order, but the gaps between points are not equal. Example: ratings of mood from 1 to 10. Ordinal data are converted to ranks for testing.
• Interval: data on a scale with equal, standard units. Example: time in seconds, or temperature.

Interval data are the most precise. They can be turned into ordinal data by ranking, but not the other way round.

Choosing a statistical test

The specification names three factors: the research aim, the level of measurement and the experimental design.

Aim and design	Nominal	Ordinal	Interval
Difference; unrelated (independent groups)	Chi-squared	Mann-Whitney	Unrelated t-test
Difference; related (repeated measures or matched pairs)	Sign test	Wilcoxon	Related t-test
Correlation or association	Chi-squared	Spearman’s rho	Pearson’s r

Textbooks usually add that the t-tests and Pearson’s r (parametric tests) need interval data that are roughly normally distributed, with similar spread in each condition. Your teacher will confirm how much of this the exam expects.

Worked example. A fictional researcher tests whether people who revise in silence recall more than people who revise with music. Two separate groups of 10 rate their recall confidence on a 1-10 scale. Mann-Whitney gives U = 18. Mann-Whitney is right: the aim is a difference, the design is independent groups, and ratings are ordinal. The hypothesis is directional, so the one-tailed critical value at p ≤ 0.05 for n1 = n2 = 10 is 27. U must be equal to or less than this. 18 ≤ 27, so the result is significant and the null hypothesis is rejected. The one-tailed critical value at p ≤ 0.01 is 19, so the result is also significant at the stricter level, which reduces the risk of a Type I error.

Common errors

• Saying “the results are significant” without comparing calculated and critical values.
• Using the “greater than” rule for Wilcoxon, Mann-Whitney or the Sign test.
• Choosing a related test for independent groups. Matched pairs count as related.
• Calling rating scales interval data. Ratings are ordinal.
• Confusing reliability (consistency) with validity (accuracy).
• Describing concurrent validity with no correlation against an established measure.
• Writing that a Type I error is “a mistake in the method”. It is a wrong decision about the null hypothesis.
• Treating a case study as useless because N = 1; say what it can and cannot show.
• Putting conclusions in the results section, or new data in the discussion.

Where next

Test yourself with the Research methods 2 practice questions. For Unit 2 methods, use the revision notes for Unit 2 and Unit 2 practice questions. The other Unit 3 sections are covered in the Psychology of sleep study guide. For exam technique, read the 9685 exam preparation guide.

Official syllabus

OxfordAQA International AS and A-level Psychology (9685) specification, Version 3.1 (International A-level exams May/June 2020 onwards), section 3.3.3 Research methods 2; Unit 3 assessment details in section 2.3.