Research methods starter

Class starter

Discuss the scenario and use one SCREAM concept to explain your answer.

Causality and Correlation

Students who report more phone notifications tend to score lower on a reading quiz. Does that show the notifications caused the lower scores?

Use in the Study Lab

Reveal model answer

The data show an association. For example, frequent phone use during study time could increase exposure to notifications and leave less time to practise reading. An experiment could assign students to notification and silence conditions while keeping the passage and quiz the same.

Repton School Dubai

IB HL Psychology

Choose where you want to start.

Start hereResearch MethodsLook up definitions, worked examples and IB guidance in the SCREAM knowledge banks. Apply your methodsThe Study LabInterpret data from a fictional study and check your reasoning.

01 Research methods

SCREAM knowledge banks

Search for a term or browse the six SCREAM topics. Use the quick reference for definitions and the in-depth explanations for methods, worked evaluations and IB assessment guidance.

SCREAM · S

Sampling

Evaluate how participants were recruited and how far the findings can be applied.

Jump to in-depth explanations

Population, sample and sampling frame

The target population is the group the research question is about. The sample is the group actually studied. A sampling frame is the list or other source from which the sample is drawn. If the frame excludes some students, a random draw from it still excludes those students.

Random sampling and random allocation

Random sampling uses chance to select participants from a sampling frame. Random allocation uses chance to assign recruited participants to conditions. Allocation can reduce systematic differences between conditions. It cannot fix a biased recruitment process.

Five sampling techniques

Simple random: Draw by chance from a frame so each member has the same chance of selection. Check how complete the frame is and who declines.
Stratified random: Divide the population into relevant subgroups and randomly sample within each. Proportionate sampling matches the subgroup proportions in the population.
Opportunity: Recruit people who are readily available. It is quick, but access determines who can take part.
Self-selected: Invite people to volunteer. Those who respond may differ in interest, time or confidence.
Snowball: Ask participants to refer others. It can help reach a less accessible population, while referrals can overrepresent closely connected networks.

Representativeness and generalisability

Representativeness concerns how well the sample reflects the population on features that matter to the question. Generalisability concerns whether findings extend beyond the sample or setting. A large sample can still be biased. In qualitative work, also ask about transferability.

Worked example and evaluation

The notification study recruited ten volunteers from one class and randomly allocated them to two conditions. Random allocation supports the comparison within the sample. Because volunteers at one school may differ from other students in reading habits or interest, the result has limited reach. If the target were the whole school, a stratified random sample by year group could improve coverage.

In-depth explanations

Coverage, non-response and sampling bias

Coverage bias begins with who is missing from the sampling frame. A list of students in a reading club cannot represent everyone in the sixth form. Non-response bias arises when the people selected who choose to respond differ in relevant ways from those who do not. A survey sent to every student can still underrepresent students who rarely open school messages.

Sampling error is the chance difference between a sample statistic and the corresponding population value. Larger probability samples tend to reduce that variation. Increasing the number of responses through the same biased recruitment route may leave the underlying bias in place.

Evaluate the route into the study. State who was eligible, how they were reached, who took part and which omitted group could change the result. Explain why that difference matters to the outcome.

Read alongside · Pearson pp. 125–128 · Giddens, Sampling Techniques and Research Considerations 1

Worked calculation for stratified sampling

A sixth form has 72 DP1 students and 48 DP2 students. A proportionate stratified sample of 20 would include 12 DP1 students (72 ÷ 120 × 20) and 8 DP2 students (48 ÷ 120 × 20).

Decide why year group matters to the question.
Use a complete list for each year group and draw names at random within each list.
Record who responds, including the number who decline or cannot be reached. Explain how non-response might change the achieved sample.

Recruiting the first 12 DP1 and 8 DP2 volunteers meets those numerical targets, but the selection within each group is voluntary. The quotas alone do not make this stratified random sampling.

Read alongside · Pearson pp. 125–127 · Giddens, Sampling Techniques

Choose a sampling method for the question

For a school-wide estimate of how often students study with notifications enabled, a complete register makes a random or stratified random sample possible. If students are piloting a quiz to find unclear wording, an opportunity sample may be practical because the immediate purpose is to improve the quiz. For a group that is difficult to identify through a register, snowball recruitment may help, while referrals can cluster people with similar experiences.

Evaluate representativeness against the actual target. A large volunteer sample from one school may tell you about its respondents. Generalising to all teenagers in Dubai would require a defensible route from that population to the sample and consideration of differences across schools and settings.

Read alongside · Pearson pp. 125–128 · Giddens, Sampling Techniques

Write a stronger sampling evaluation

Weak. “The sample was small, so it was biased.”

Developed. “The study recruited volunteers from one class. Students who were comfortable with the reading task may have been more willing to join, so the effect could differ in classes with a wider range of reading confidence. Recruiting across several classes would improve coverage, while a larger sample would make the group estimates less sensitive to individual scores.”

The developed answer separates who is represented from how uncertain the estimate is. For qualitative research, use the description of the sample and setting to evaluate transferability.

Read alongside · Pearson pp. 103–104 and 125–128 · Giddens, Research Considerations 1 and 3

SCREAM · C

Causality and Correlation

Decide whether the evidence supports an association or a causal explanation.

Jump to in-depth explanations

Correlation and direction

Correlation describes the direction and strength of a relationship between measured variables. A positive association means higher values tend to occur together. A negative association means higher values on one variable tend to occur with lower values on the other. Inspect the scatterplot for shape and outliers before relying on a coefficient.

What supports a causal inference?

Look for evidence that the proposed cause came first, that the outcome changed with the proposed cause, and that plausible alternatives were addressed. Manipulation, random allocation and standardised conditions can strengthen a causal inference. A finding from one setting still has limits on where it applies.

Reverse direction and third variables

An observed relationship can reflect X influencing Y, Y influencing X, both directions, or a third factor influencing both. Time order can help rule out some explanations. It cannot on its own rule out a third variable.

True experiments and quasi-experiments

In a true independent-groups experiment, researchers manipulate the independent variable and randomly allocate participants to conditions. A quasi-experiment lacks random allocation, for example when comparing existing groups. Pre-existing differences can then offer another explanation for the outcome.

Worked example and evaluation

If students who report more notifications score lower, heavier phone use during study time could account for both more interruptions and less reading practice. In the dashboard experiment, assigning students to hear standardised sounds or read in silence addresses that explanation more directly. Limit the claim to the conditions, participants and outcome studied.

In-depth explanations

Three explanations for the same correlation

Suppose students with more notifications tend to score lower on a reading quiz. The same pattern is compatible with several explanations.

Direct effect: Notifications interrupt reading, which lowers immediate comprehension.
Reverse direction: Students already struggling with the passage may turn to their phones more often and encounter more notifications.
Common cause: Frequent phone use during study time could increase exposure to notifications and reduce reading practice.

A scatterplot can show how the two measures vary together. Timing, comparison groups and controls provide different evidence about the causal explanation.

Read alongside · Popov §1.5 · Pearson pp. 94–95 and 122–123 · Giddens, Correlational Studies

Compare experimental, quasi-experimental and correlational evidence

Randomised experiment: Assign volunteers by chance to a standardised notification sound or silence. The comparison helps isolate the effect of the assigned sound under those conditions.
Quasi-experiment: Compare existing classrooms following different phone policies. Existing differences in teaching, student intake or school routines could explain a difference in scores.
Correlational study: Measure students’ usual notification frequency and their scores without assigning conditions. This can answer whether the measures are associated in that sample.

Evaluate the fit. A controlled experiment can improve confidence about the effect of an assigned sound. A study of usual phone use may better reflect everyday behaviour, while leaving more possible causes open.

Read alongside · Popov §§1.4–1.5 · Pearson pp. 110–112 and 122–123 · Giddens, Experiments and Correlational Studies

Effect size, uncertainty and scope

Separate three questions when judging a causal claim.

Is the comparison fair? Check assignment, timing, standardisation, loss of participants and plausible confounds.
How large and precise is the effect? Consider the size of the difference, the spread of scores and any uncertainty estimate the source reports.
Where does the conclusion apply? Consider the sample, setting, task and follow-up period.

A small randomised study can provide useful causal evidence while giving an imprecise estimate. A large association can still reflect a third variable. Replication in a different group or setting can help test whether a finding holds there.

Read alongside · Pearson pp. 94–101 and 110–112 · Popov §§1.4–1.5 and §1.12

Write a causal evaluation from a study detail

Study detail. A teacher compared quiz results between two intact classes, each taught by a different teacher. One used a no-phone rule.

Evaluation. “The class using the no-phone rule scored higher, but phone policy varied with teacher. Differences in instruction or reading preparation could therefore account for the score difference. Comparing randomly assigned conditions with the same instructions and scoring would strengthen the causal inference.”

The evaluation explains how a specific design choice affects the claim. In an unseen-study answer, name the relevant IB concept and adapt the argument to the evidence in that study.

Read alongside · IB Psychology guide, Paper 2 Section B · Giddens, Experiments and Researcher and Participant Bias

Published study · Wording and eyewitness reports

Loftus and Palmer (1974) varied the verb in a question about filmed collisions. Stronger wording such as “smashed” led to higher speed estimates. In their second experiment, participants who had received that wording were also more likely to report broken glass one week later, although none appeared in the film.

What the design lets you say. The manipulation supports an effect of question wording on the answers given. The initial speed estimate might reflect response expectations. The later report offers further evidence relevant to memory, while a self-report still leaves room for changes in how participants decide to answer. Film-based tasks also differ from witnessing a real collision.

This is an example of evaluating the claim at the level of the actual outcome. Use measurement validity to explain the gap between a reported answer and the underlying memory.

Read alongside · Pearson p. 106 · Loftus and Palmer, original study

SCREAM · R

Research Design

Match the research question to a method and evaluate the choices made within it.

Jump to in-depth explanations

Research questions and hypotheses

A research question identifies what the study seeks to find out. A hypothesis makes a testable prediction about a difference or association. A directional hypothesis predicts which way the result will go. The null hypothesis states that there is no effect or association in the population under the model being tested.

Quantitative and qualitative evidence

Quantitative data represent counts, scores or measurements. Qualitative data describe meanings and experiences through words, observations or other material. Some studies use both, such as a survey with closed ratings and open comments. Choose data that answer the question and analyse each type appropriately.

Experiment and variables

An experiment tests a predicted effect by comparing conditions. The independent variable (IV) is the factor manipulated in a true experiment. The dependent variable (DV) is the measured outcome. A comparison condition gives a baseline. Standardised instructions and procedures help isolate the IV.

Independent groups, repeated measures and matched pairs

Independent groups: Different participants complete each condition. This avoids order effects but makes participant differences relevant.
Repeated measures: The same participants complete each condition. It controls stable individual differences but can produce fatigue, practice or carry-over. Counterbalancing can distribute order effects.
Matched pairs: Match participants on a relevant characteristic and assign each pair across conditions. Matching can reduce selected differences and depends on choosing a meaningful match variable.

Observation

Record behaviour using clear categories or a planned coding schedule. Observation can be naturalistic or controlled, overt or covert, and participant or non-participant. Evaluate reactivity, observer bias, agreement between coders and the ethical implications of how consent and privacy are handled.

Survey or questionnaire

Collect self-report through a standard set of questions. Closed questions are easy to compare. Open questions capture detail in participants’ own words. Check wording, response options, response rates, anonymity and the risk of socially desirable answers.

Interview and focus group

A structured interview uses a fixed set of questions. A semi-structured interview allows planned questions and follow-up. A focus group generates data through facilitated discussion. Compare consistency with depth and consider how the interviewer or other participants may shape what is said.

Correlational study and case study

A correlational study measures variables as they occur and assesses their association. A case study investigates one person, group or situation in depth and can draw on observations, interviews and records. Judge how each method fits the question, and distinguish richness of evidence from breadth of generalisation.

The four IB class practicals

Learning and cognition · Experiment
Human development · Observation
Human relationships · Survey or questionnaire
Health and well-being · Interview

For each practical, know the aim, participants, procedure, data and one justified change. The IA is a research proposal using one of these four methods.

Ethics and researcher responsibility

Apply the IB class-practical guidelines and school procedures. Obtain explicit informed consent, explain the aim and right to withdraw, protect privacy and minimise harm. Debrief appropriately and explain how data will be handled. Identify the specific safeguard the method or situation calls for.

Worked example and evaluation

In the notification experiment, different students take part in each condition. The IV is the sound condition and the DV is the quiz score. Use the same passage, timing, instructions and scoring key. Random allocation reduces systematic group differences, although chance differences can still matter in a small sample.

In-depth explanations

Match the method to the research question

Does a change in X affect Y?: Consider an experiment if manipulation and fair comparison are feasible and ethical.
Do X and Y vary together?: Consider a correlational study. Specify how each variable will be measured.
What do people report, and how often?: Consider a questionnaire. Check coverage and whether the items measure what the question asks.
How do people understand an experience?: Consider an interview or focus group. Use neutral prompts and reflect on whose views may be missing.
What do people do in a setting?: Consider an observation. Define observable categories and test whether observers use them consistently.
What can we learn about a particular case?: Consider a case study. Justify the choice of case and the sources used to interpret it.

Read alongside · Pearson pp. 110–124 · Popov §§1.2 and 1.4–1.7 · Giddens, Research Methods worksheets 1–6

Control, order effects and counterbalancing

Standardisation means delivering the same instructions, timing and scoring across conditions. Keep the treatment difference clear. In the notification example, use the same task and scoring key and vary only whether the standardised sounds play.

In repeated measures, reading the same passage twice could improve the second score through familiarity. Counterbalancing assigns some participants to sound first and others to silence first. This distributes a general order effect. If doing one condition changes later performance in a lasting way, consider independent groups or two comparable passages and balance which passage appears in each condition.

In independent groups, random allocation helps distribute participant differences. A small sample can still have chance imbalances, so state that limit when interpreting results.

Read alongside · Popov §1.4 · Pearson pp. 110–112 · Giddens, Experiments

Observation categories, records and observer agreement

Define behaviour so two observers can recognise it consistently. For example, record “eyes leave the passage for at least two seconds” rather than “loses concentration.” A brief pilot can uncover unclear categories and help observers use the same rules.

Event sampling: Record each occurrence of a behaviour. This suits discrete events such as requests for help.
Time sampling: Record behaviour at set intervals or within set windows. State the rule and consider events that might occur between checks.

Compare independent records for the same observation periods. High agreement supports consistent coding. It does not establish that looking away is a valid measure of lost attention.

Read alongside · Pearson pp. 112–118 · Giddens, Observations and Research Considerations 1 · IB guide, Human development practical

Improve surveys, interviews and focus groups

Survey. Ask one thing per item. Replace “How distracted and stressed were you?” with two separate questions. Anchor responses to a period such as “during yesterday’s homework” and pilot the scale for unclear or missing options.

Interview. Replace the leading question “How did notifications disrupt you?” with “What happened when you heard a notification?” A neutral follow-up such as “Can you describe an example?” gives room for experiences the researcher did not anticipate.

Focus group. Discussion can reveal agreement and disagreement. A dominant voice can also shape what quieter participants say. Use neutral facilitation and be clear about the limits of confidentiality in a group.

Read alongside · Pearson pp. 119–122 · Popov §§1.6–1.7 · Giddens, Surveys and Questionnaires and Interviews

Case studies, triangulation and transfer

A case study can combine interviews, observation and records to examine a person, group or event in depth. A study of how one school introduced a phone policy might compare student accounts, observed routines and records of policy implementation.

Triangulation means comparing evidence across sources or methods. Agreement can strengthen an interpretation. Disagreement is also informative. Student accounts might describe fewer interruptions in class while observations show more device use between lessons. Explain what each source captures and where the claim needs qualifying.

Give enough detail about the school and how evidence was collected for readers to judge whether the findings might be relevant elsewhere.

Read alongside · Pearson pp. 107–109 and 123–124 · Popov §1.6 · Giddens, Case Studies and Research Considerations 3

From class practicals to the IB research proposal

For each of the four practicals, keep a record of the question, method, sample, procedure, data and ethical safeguards. Be ready to explain why a choice was made, compare it with another method and design an alternative way to investigate the same topic.

For the IA research proposal, specify a population of interest and a question worth investigating for that group. Justify a method from the four class practicals, explain who you would recruit and how, and make the procedure and proposed data collection concrete. Show how ethics shapes those choices.

A class practical provides experience you can draw on for Paper 2. The IA assesses the proposed investigation and the reasoning behind it.

Read alongside · IB Psychology guide, Paper 2 and Internal Assessment · IB Teacher Support Material · Pearson pp. 327–332 and 338 onward · Popov Chapter 7

Ethics choices that change the design

Consent and withdrawal: Explain the aim, what participation involves, and the right to stop without pressure. The IB guide requires explicit informed consent using a form and written parent or guardian consent for participants under 16. Follow school approval and any additional requirements before recruitment.
Privacy: Collect only what the question requires. Use a code when data still need to be linked to a participant for withdrawal. Explain who can access the data and when identifiers will be removed.
Protection from harm: IB class practicals must avoid harm, distress and discomfort. Conformity and obedience studies are excluded. Choose an alternative where a sensitive topic, intrusive records or a power imbalance would create risk.
Debriefing: Explain the activity and address questions afterwards. Tell participants what happens to their data.

Evaluate an ethical decision by naming who could be affected, what the risk is and what the safeguard changes.

Read alongside · IB Psychology guide, Class practicals and ethical guidelines · IB Teacher Support Material · Popov §1.8

Published study · What the case of H.M. can establish

Scoville and Milner (1957) documented severe difficulty forming new lasting declarative memories after H.M.’s surgery for epilepsy. Their detailed assessment described a pattern of impairment alongside abilities that were relatively preserved. The comparison between those abilities made the case informative for theories of memory.

Evaluate the inference. A rare clinical case gives access to a pattern difficult to investigate experimentally. Surgery involved more than a single isolated brain structure, and the clinical history is specific to the patient. The case supports a role for medial temporal structures in new memory formation. Broader or more anatomically precise claims need converging evidence.

Distinguish the clinical decision from an experimental manipulation. Researchers studied the consequences of treatment.

Read alongside · Giddens, Case Studies and Research Considerations 1 · Scoville and Milner, original report

SCREAM · E

Explanations

Consider alternative explanations, including confounding variables, and ask what would make an interpretation more credible.

Jump to in-depth explanations

Extraneous variables and confounds

An extraneous variable is a factor outside the intended study variables that could influence the outcome. It becomes a confound when it varies systematically with the condition and could explain the result. Name what differs, how it could affect the outcome and which direction of bias is plausible.

Participant bias and demand characteristics

Participants may adjust behaviour when they guess the study aim, know they are observed or want to give an acceptable answer. Standard instructions, neutral wording and private responses can help. Choose safeguards that fit the method and the consent requirements.

Researcher bias and reflexivity

Researchers’ expectations can shape recruitment, instructions, observations, scoring and interpretation. Options include using a scoring rubric, keeping scorers unaware of group assignment when appropriate, and documenting decisions. Reflexivity involves examining how researchers’ perspectives and roles influence the research.

Credibility and transferability

Credibility concerns whether the interpretation is convincing and grounded in the qualitative evidence. Check that themes are supported by data, interpretations account for contradictory examples and the analysis process is transparent. Transferability asks whether readers have enough detail about the participants and setting to judge relevance elsewhere.

Worked example and evaluation

If the notification group alone works in a noisier room, room noise could reduce comprehension separately from the planned sounds. That makes noise a plausible confound. Use equivalent testing conditions and vary only the planned sounds. If scorers expect lower marks in that group, score from anonymised responses with a fixed answer key.

In-depth explanations

Diagnose a threat and choose a matching control

Participant differences: If one condition begins with stronger readers, prior ability could account for higher scores. Random allocation, matching on a relevant measure or a suitable repeated-measures design can address this in different ways.
Situational differences: If the sound group works at a busier time of day, ambient noise or fatigue could influence its scores. Standardise the setting, timing and instructions or balance them across conditions.
Order and carry-over: If everyone takes the silent condition first, familiarity could favour the sound condition. Counterbalance the order or choose independent groups if the first condition changes later performance in a lasting way.
Scoring differences: If scorers know the conditions and expect a result, borderline answers may be marked differently. Use a scoring key and mask group labels where practical.

For each threat, identify the study detail, the process that might affect the outcome and the control that addresses it.

Read alongside · Pearson pp. 103–107 and 110–112 · Giddens, Experiments and Research Considerations 2

How bias can enter at different stages

Recruitment. The way a study is advertised may attract students with particularly strong views about phones. Collection. An interviewer’s follow-up may invite participants to describe disruption more often than helpful experiences. Analysis. A researcher may attend to statements that fit their expectation while overlooking contrary evidence. Reporting. A conclusion may present one pattern without explaining how common or consistent it was.

Match safeguards to the risk. Neutral recruitment language addresses who responds. A common interview guide supports consistency. Keeping a record of coding decisions and looking for cases that challenge an interpretation helps readers examine the analysis.

Read alongside · Pearson pp. 104–109 · Giddens, Research Considerations 2 and Suggested Answers

Reflexivity and qualitative credibility in practice

A teacher interviewing their own students expects phones to distract them. Students may tell the teacher what they think the teacher wants to hear, and the teacher may ask more follow-up questions about problems than benefits.

A reflexive approach could record that expectation before interviews, use a neutral interviewer where possible, compare how follow-up questions were used and look for accounts of students who found alerts helpful or easy to ignore. Explain how these decisions affected the interpretation.

Credibility improves when the reader can see how conclusions follow from the data. Transferability depends on enough contextual description to judge whether the finding could be relevant elsewhere. Describe the participants, setting and circumstances that matter to the finding.

Read alongside · Pearson pp. 105–109 · Giddens, Transferability and Credibility and Researcher and Participant Bias

Worked Paper 3 research-considerations response

Illustrative source. A school interviewed students who volunteered after a policy change. All interviews were conducted by the person who introduced the policy. The report gives only comments in favour of the change.

Developed response on credibility. The interviewer’s role may make students reluctant to criticise the policy. Using a neutral interviewer with consistent, open prompts could make disagreement easier to express. The report should also show how opposing or mixed accounts were handled. That would let readers judge whether the themes reflect the range of responses.

When the question asks about transferability, move from the specific setting to a reasoned comparison with the proposed new setting. Similar policies alone may be insufficient if student access to devices or teaching routines differ.

Read alongside · IB Psychology guide, Paper 3 Question 3 · IB specimen Paper 3 · Giddens, Transferability and Credibility

SCREAM · A

Analysis of Data

Read the display, describe the pattern, cite evidence and qualify the conclusion.

Jump to in-depth explanations

Descriptive statistics and distributions

The mean uses every score and is sensitive to extremes. The median is the middle value and can be informative with skewed data. The mode is the most frequent value. The range is maximum minus minimum. Standard deviation describes spread around the mean. Check centre, spread, overlap, skew and outliers before comparing groups.

Choose and read a data display

Frequency table: Counts how often each category, score or interval occurs.
Bar graph: Compares categories or group summaries. Group averages can hide spread and overlap.
Histogram: Shows the distribution of quantitative data grouped into adjacent intervals.
Line graph: Shows change across an ordered scale, such as time.
Box plot: Shows the median and middle half of the data. Check what the whiskers and outlier markers represent.
Scatterplot: Shows paired quantitative observations. Check direction, form, strength and unusual points.

Correlation coefficients and statistical inference

A correlation coefficient summarises the direction and strength of an association. For Pearson’s r, values range from −1 to +1 and a value near zero indicates little linear association. Inferential results help evaluate what a sample can tell us beyond the observed data. A p-value describes how unusual the observed result, or a more extreme one, would be under the null hypothesis and model assumptions. Interpret it alongside effect size, uncertainty and the study design.

Thematic analysis

Read the data closely, code relevant passages, group codes into candidate themes, review them against the data and support each theme with examples. A code labels one feature of a response. A theme captures a wider pattern relevant to the research question. Use contrasting responses to refine the interpretation.

Worked example and evaluation

The notification group averaged 7.4 and the silence group averaged 9.2 out of 12. The mean difference is 1.8 points. Scores overlap at 8 and 9, and only five students took part in each condition. Report the observed pattern and qualify the precision and reach of the finding.

Applying data analysis in HL Paper 3

Read the axes, units, source and legend before interpreting a figure. Cite a specific pattern or value and explain the significance of a limitation. Build separate skills in graph interpretation, data analysis, research considerations and synthesis across sources. See the Study Lab for practice.

In-depth explanations

How to read and evaluate a graph

Orient. Identify the variables, units, groups, legend and scale. Check whether the values are scores, percentages, counts or summaries.
Describe. State the direction or group difference and cite two relevant values where useful. Check for spread, overlap, unusual points and changes across the scale.
Interpret. Explain what the pattern means for the study question, within the limits of the method.
Qualify. Name a feature that limits interpretation and explain its consequence. An axis that starts well above zero can make a modest difference in bar heights look much larger. Missing group sizes make it harder to judge how much evidence supports each summary.

For a box plot, locate the median and quartiles and read the legend for the whisker convention. Overlapping boxes describe the displayed distributions and do not settle whether a difference is statistically significant.

Read alongside · Pearson pp. 86–93 · Popov §1.11 · Giddens, Visualising Data · IB specimen Paper 3

Worked quantitative comparison

Illustrative comprehension scores out of 12
Condition	Scores	Mean	Median	Range	Sample SD
Notifications	6, 7, 7, 8, 9	7.4	7	3	1.14
Silence	8, 9, 9, 10, 10	9.2	9	2	0.84

The silence mean was 1.8 points higher. The higher median supports the same pattern, and the sample standard deviations suggest slightly less spread in silence. The groups still overlap at 8 and 9, so the higher group average does not describe every individual.

Conclusion. In this fictional sample, scores tended to be higher in silence. With five students per condition, one or two scores could materially change the summaries. The descriptive results alone give no inferential test or uncertainty interval.

Read alongside · Pearson pp. 86–97 · Popov §§1.10–1.11 · Giddens, Visualising Data · Repton Study Lab fictional dataset

Skew, outliers and the choice of summary

A right-skewed distribution has a longer tail toward higher values. A left-skewed distribution has a longer tail toward lower values. For reading times of 20, 21, 21, 23 and 50 seconds, the mean is 27 seconds and the median is 21 seconds. The 50-second value pulls the mean up, so the median better describes the centre of the main cluster in this example.

Investigate an unusual value before deciding how to handle it. It might reflect a recording error, a different procedure or a genuine response. Record the reason for any exclusion. Where it matters, compare results with and without the value so readers can see how sensitive the conclusion is.

Read alongside · Pearson pp. 87–93 and 95–97 · Popov §§1.10–1.11 · Giddens, Visualising Data

Interpret statistical significance and uncertainty

Descriptive statistics summarise the observed sample. Inferential statistics use a model and assumptions to assess what the data suggest beyond the observed scores. Read the comparison, effect estimate, uncertainty and assumptions together.

Illustrative report. A separate study estimates a 2-point difference and gives a 95% confidence interval of 0.2 to 3.8 points. The point estimate is 2. The interval spans a small to a substantially larger difference. Under the stated method, the values shown are compatible with the data to varying degrees, and the interval excludes zero.

When a study reports p = .03 and set a .05 threshold in advance, the result meets that threshold under the test assumptions. It does not mean there is a 3% probability that the null hypothesis is true. A result above the threshold can be inconclusive, especially when estimates are imprecise. Judge practical importance in the units of the task.

Read alongside · Pearson pp. 97–101 · Popov §1.12 · IB Psychology guide, Data analysis and interpretation

Interpret a correlation coefficient in context

Suppose a study reports Pearson’s r = −.55 between notifications per hour and quiz score. The negative sign means higher notification rates tend to accompany lower scores. The magnitude describes linear association in that sample. It does not mean a 55% score drop.

Look at the scatterplot before interpreting the coefficient. A curved pattern can have a coefficient near zero even when the variables are related. One unusual observation can also change the coefficient. Restricted values, for example sampling only students with very similar phone use, can change the observed association.

A conclusion should identify the variables, direction and sample and then qualify the causal claim. Return to the possible explanations for a correlation when interpreting the pattern.

Read alongside · Pearson pp. 93–95 and 100–101 · Popov §§1.5 and 1.11 · Giddens, Correlational Studies

Worked thematic analysis from quotation to theme

Fictional comments from the notification study
Response	Possible code	Role in interpretation
“After the sound I went back to the start of the paragraph.”	Rereading to recover place	Shows disruption to reading.
“I kept thinking about whether the message was urgent.”	Anticipating a message	Shows attention shifting away from the passage.
“I noticed the sound, but I carried on reading.”	Maintaining focus	Challenges a claim that all interruptions caused disruption.

A possible theme is recovering or maintaining focus after an interruption. The first two excerpts describe different paths away from the passage. The third suggests a boundary to the theme. Review the full set of responses before deciding whether to keep one theme or separate it into patterns.

A quotation illustrates the interpretation. Its presence alone does not show how common an experience was. Document coding decisions and show how the theme changed when you considered divergent responses.

Read alongside · Pearson pp. 101–102 · Popov §1.13 · Giddens, Thematic Analysis and Suggested Answers

What each HL Paper 3 question is asking you to do

Question 1 · Graph interpretation · 3 marks: Explain one issue that limits the interpretation of the named source. Name the feature and explain what it makes harder to judge.
Question 2 · Data analysis · 6 marks: Analyse the source specified in the question and state a conclusion explicitly linked to its findings. Select the relevant comparisons and support them with evidence.
Question 3 · Research considerations · 6 marks: Discuss the named qualitative consideration, such as credibility, bias or transferability, in relation to the specified source.
Question 4 · Synthesis · 15 marks: Evaluate the given claim by drawing together relevant sources and your understanding of the HL extension. Explain where evidence converges, where it differs and how design affects the weight you give it.

In the specimen on technology and mental health, an experiment, a focus-group source and a correlational source offer different kinds of evidence. When drawing them together, keep each study’s question, measure and design in view.

Read alongside · IB Psychology guide, Paper 3 · IB specimen Paper 3 and markscheme · Pearson pp. 333–337

SCREAM · M

Measurement

Check how a psychological construct was operationalised, how consistently it was measured and what the scores mean.

Jump to in-depth explanations

Construct and operational definition

A construct is an idea such as attention, motivation or anxiety. An operational definition states exactly how a variable is manipulated, observed or scored. Different operational definitions can support different interpretations of the same construct.

Reliability

Reliability concerns consistency. Test–retest reliability asks whether scores are similar on repeated measurement when the construct is stable. Inter-rater reliability asks how consistently independent raters score the same material. Internal consistency asks whether items designed to measure the same construct tend to work together.

Validity of measures and conclusions

Construct validity: How well the evidence supports interpreting the score as a measure of the intended construct.
Content validity: Whether the items adequately cover the intended content or behaviours.
Face validity: Whether the measure appears appropriate on inspection.
Internal validity: How well the design supports the intended causal interpretation.
External validity: How well findings apply across people, situations or times. Ecological validity concerns relevance to real-world settings or behaviour.

Types of measurement

Nominal data identify categories, such as notification type. Ordinal data have a rank order, such as a single rating from strongly disagree to strongly agree. Interval data have equal steps without a meaningful zero. Ratio data have equal steps and a meaningful zero, such as time spent reading. Match the summary and display to the data collected.

Worked example and evaluation

The dashboard quiz gives one point per correct response across 12 items. A standard scoring key can make marking consistent. To support the intended interpretation, check that the items sample the passage fairly and require comprehension. An immediate quiz gives limited evidence about long-term retention or sustained attention.

In-depth explanations

Operationalise attention, motivation or comprehension

Start with the construct and the claim you want to make. “Attention” could mean sustained focus, the ability to ignore a distraction or the ability to return to a task. Each definition points to different observations.

Immediate comprehension: Score correct answers on a 12-item passage quiz administered immediately after reading. A suitable set of items needs to cover the passage and test understanding.
Sustained attention: Record performance across several time windows on a task designed to require continued attention. One total score could hide a drop over time.
Motivation: Measure how long students persist when given the option to stop, or use a justified self-report measure. Persistence can also reflect skill, task difficulty or how much time students have.

Specify what is recorded, when it is recorded and how it is scored. Explain which aspects of the construct the measure captures.

Read alongside · Pearson pp. 102–104 · Popov §§1.3 and 6.5 · Giddens, Reliability, Validity and Generalizability

Choose a reliability check and interpret it

Test–retest: Use the same measure at two appropriate times when the construct is expected to be stable. Very short intervals may produce memory effects. Long intervals may allow real change.
Inter-rater reliability: Have two people independently score the same behaviours or responses using shared rules. Compare their ratings and investigate disagreements.
Internal consistency: For a multi-item scale intended to capture one construct, examine whether items behave consistently together. High consistency alone cannot show that the scale measures the intended construct.

A worked check for the reading quiz would give two scorers the same sample of open responses and a rubric. If they disagree mainly on partial answers, make the partial-credit rules clearer and check agreement again.

Read alongside · Pearson pp. 102–103 · Popov §1.3 · Giddens, Reliability, Validity and Generalizability

Diagnose which kind of validity is at stake

Content validity: A comprehension quiz asks only about the first paragraph. Review item coverage against the entire passage and the kinds of understanding it should measure.
Construct validity: A reading score may reflect English vocabulary or prior knowledge as well as comprehension. Look for evidence that the intended interpretation holds across relevant groups and tasks.
Face validity: A quiz appears to assess understanding. That impression can be useful for acceptability, but stronger evidence comes from how the items and scores behave.
Internal validity: Noise differs by condition, so the sound manipulation may not explain the result. Improve control of the design.
External validity: A brief classroom task may not reflect extended independent reading. Compare the task and sample with the setting where you want to apply the result.

In an evaluation, identify whether you are questioning the measure, the causal comparison or the scope of the conclusion.

Read alongside · Pearson pp. 102–104 · Popov §1.3 · Giddens, Reliability, Validity and Generalizability

Compare measurement across cultures and settings

Imagine comparing how willing students in two settings are to ask for help. The same questionnaire item may bring to mind help from a teacher in one setting and help from friends in another. Translated wording can also differ in tone, and groups may use a rating scale differently.

Clarify who the question refers to, pilot the wording with students in both settings and ask what they think the item means. Where appropriate, use language review and compare the survey with interviews or observations. Report what the checks support and where comparisons remain uncertain.

These checks help distinguish a difference in the construct from a difference in how it was measured. If the two interpretations remain plausible, qualify the comparison.

Read alongside · Pearson pp. 102–104 · Popov §§1.3 and 6.6 · Giddens, Research Considerations 1 and 3

SCREAM is a revision mnemonic. In an IB assessment, use the concepts named in the question and consider researcher responsibility where relevant.

Apply SCREAM in the Study Lab

02 The Study Lab

Do notifications affect reading comprehension?

Apply SCREAM to a fictional study. Read the setup, inspect the scores and decide what the results support.

Study setup

Ten students volunteered from one class. Researchers randomly assigned five to hear three standardised notification sounds while reading and five to read in silence. Everyone had the same passage, reading time and 12-item quiz immediately afterwards.

Illustrative quiz scores out of 12
Condition	Scores	Mean
Notifications	6, 7, 7, 8, 9	7.4
Silence	8, 9, 9, 10, 10	9.2

Compare the means visually

Mean quiz scores out of 12

Notifications7.47.4

Silence9.29.2

What can you see in the raw scores that the means alone do not show?

Try a stretch question and reveal the model response

Question · Suppose the notification group took the quiz in a noisier room. Explain how this could affect the conclusion and suggest one design improvement.

Model response · Room noise would be a potential confound because it varies with the notification condition and could lower comprehension scores independently. Test both groups in the same quiet room under equivalent conditions, varying only the notification sounds.

Try thematic analysis

Three fictional students describe interruptions while reading. Give each response a short code, then look for a theme and a response that complicates it.

“After the sound I went back to the start of the paragraph.”
“I kept thinking about whether the message was urgent.”
“I noticed the sound, but I carried on reading.”

Reveal one possible interpretation

Possible codes are rereading, anticipating a message and staying on task. The first two suggest a theme of attention shifting away from the passage. The third shows why the theme should allow for differences in how students respond.