Aims and hypotheses — An aim states what the investigation seeks to understand; a hypothesis predicts a particular difference or relationship. A useful prediction identifies operational variables and the relevant population or conditions. Asking whether noise affects recall is an aim; predicting fewer specified words recalled under speech than quiet is a directional hypothesis.
Directional predictions — A directional hypothesis predicts the direction of a difference or association, based on an appropriate prior rationale. It must state what higher or lower means for the measured variables. Predicting a negative association is not the same as predicting no association.
Non-directional predictions — A non-directional hypothesis predicts a difference or relationship without specifying its direction. This can be appropriate when a direction is not sufficiently justified. Write the alternative clearly rather than confusing it with the null hypothesis of no relevant effect or association.
Population, sample and frame — The target population is the group to which the researcher wants the findings to apply. The sample contains those actually studied, and a sampling frame lists eligible selection units when one is available. A complete frame and high participation matter as well as the method used to draw names.
Random sampling — A random sample gives eligible members a specified chance of selection through a random procedure. It reduces deliberate selection bias but does not guarantee a perfectly representative small sample. Nonresponse can produce bias even after an initially fair selection.Random sampling and random allocation are distinct. A study can use one without the other; neither guarantees perfect balance in a small sample.
Systematic sampling — Systematic sampling selects every kth eligible entry after a suitable random starting point. Divide the frame size by the required sample size to plan an interval where appropriate. A patterned frame can align with the interval and distort representation.
Stratified sampling — Stratified sampling represents relevant subgroups in planned proportions, normally using random selection within strata. If a population is 60% in one category and a sample of 50 is required, proportional allocation gives 30 from that category. Correct proportions for one characteristic do not guarantee representativeness on every other characteristic.
Opportunity and volunteer sampling — Opportunity sampling recruits accessible eligible people; volunteer sampling recruits those who respond to an invitation. Both are practical but can select distinctive participants. An online invitation is not random sampling merely because many different people could see it.
Participation and generalisation — A selected person may decline or leave, and recruitment conditions can exclude groups unintentionally. State who was approached, who took part and what limits the resulting claim. A large biased sample is not automatically more representative than a smaller well-designed one.
Piloting — A pilot tests whether instructions, materials, timings, categories and recording procedures work as intended. It can reveal ambiguity or an impossible task before the main investigation. It should inform appropriate revision rather than be used to select only the outcome the researcher prefers.
Research conduct and ethics
Demand characteristics — Participants may infer the aim and alter behaviour to cooperate, resist or manage impressions. This can change the meaning of the measured response. Neutral information and carefully designed tasks can reduce the problem without using unjustified deception.
Investigator effects — A researcher's expectations or behaviour can influence instructions, interaction, observation or scoring. Standardisation, independent scoring and appropriate blinding can help. Blinding must specify who lacks what information and why that reduces the proposed influence.
Informed agreement — Consent should be based on understandable information about participation and relevant risks, with appropriate safeguards for age, capacity and context. Agreement from a person in authority does not erase the participant's own welfare and choice. Consent is a process, not simply a signature obtained once.
Withdrawal and pressure — Participants should be able to stop without inappropriate penalty or persuasion. Payment, school power relationships or dependence on the researcher can make a nominal choice feel constrained. Explain how withdrawal would work in the actual procedure rather than adding a vague sentence about a right.
Deception and debriefing — Withholding a hypothesis can sometimes reduce expectancy effects, but deception needs ethical justification and cannot expose people to avoidable harm. Debriefing clarifies the purpose and relevant deception and offers questions or support. It does not retrospectively create fully informed consent or remove every adverse experience.
Protection, privacy and confidentiality — Assess risks of distress, embarrassment, physical harm and identifiable disclosure. Privacy concerns access to personal information or activities; confidentiality concerns how obtained information is handled. Anonymisation reduces identifiability but should not be promised if data or circumstances can still identify a person.
Ethical practical research — Classroom research should use appropriately approved low-risk activities, clear information, genuine choice and suitable data handling. Hypothetical examples in notes are not evidence that a learner completed a practical. A centre must oversee real participation, safeguarding and applicable research standards. The British Psychological Society's Code of Human Research Ethics provides principles including respect, scientific integrity, social responsibility and minimising harm. Its role is to guide justified decisions and review throughout research, not to replace an actual risk assessment with a slogan.
Peer review — Independent experts can examine a study's rationale, design, analysis and interpretation before or after publication. Review can detect weaknesses and improve clarity, but reviewers may miss errors or share assumptions. Publication is evidence of scrutiny, not a guarantee of truth.
Economic implications — Psychological research can affect work, education, health, design and public policy, with consequences for costs and productivity. For example, an evaluated approach to reducing avoidable absence may affect both well-being and staffing costs. Benefits, resources, distribution and unintended effects require evidence rather than a claim that every psychological intervention saves money.
Reliability and validity
Reliability — Reliability concerns consistency of measurement under relevant conditions. A consistently miscalibrated measure can be reliable but inaccurate or invalid. Specify whether consistency concerns time, observers or another source rather than saying that a study is reliable because it used numbers.
Test-retest reliability — Administering an appropriate measure again after an interval can assess stability when the underlying construct is expected to be stable. A strong association supports consistency, but memory, practice, genuine change or an unsuitable interval can affect interpretation. Use comparable conditions and justified timing.
Inter-observer reliability — Independent observers code the same material using the same scheme. Agreement can reveal clear categories, while disagreement can identify ambiguities for training and revision. Discussion to settle disagreements is useful for development but should not be misreported as independent agreement.
Improving consistency — Standardise relevant instructions, train observers, define scoring and pilot ambiguous items. Consistency should not be pursued by deleting genuine variation or forcing every participant into an unsuitable category. Retain an audit of changes and their rationale.
Face validity — A measure has face validity when it appears to assess its intended construct to relevant judges. This can improve acceptability but is not strong evidence that it truly captures the construct. A plausible-looking questionnaire can still measure response style more than the claimed trait.
Concurrent validity — Concurrent validity examines agreement with an appropriate established measure of the same construct taken around the same time. The comparison is useful only if the reference is itself relevant and credible. Two measures can agree because they share a bias.
Ecological validity — Ecological validity concerns the relation between the task or finding and the relevant everyday setting. An artificial task can be useful for a narrow mechanism even if broad everyday transfer is uncertain. Evaluate the intended claim, not the building's appearance alone.
Temporal validity — Temporal validity concerns whether a result remains applicable across time as social practices, technologies or circumstances change. A reliable observation from one historical setting may not describe a later one. Repeat or updated evidence helps assess that scope.
Improving validity — Use measures that represent the intended construct, reduce confounds, compare suitable evidence and investigate participants' interpretations. More realism can sometimes reduce control; a stronger design explains the trade-off. No single procedural feature guarantees every form of validity.
Science and reporting
Objectivity and empirical evidence — Objectivity seeks to reduce personal preference in observation and inference; empiricism uses systematic evidence rather than assertion alone. Operational rules and transparent analysis help others assess a claim. Research choices still involve judgement, so transparency is more realistic than claiming a wholly perspective-free observer.
Replication — A replicable procedure is described clearly enough to be repeated appropriately. Replication tests how robust a finding is under specified conditions and may vary method or context. One unsuccessful replication can prompt investigation of statistical power (the chance of detecting an effect of a specified size), procedures and moderators (conditions that change an effect's size or direction) rather than automatically proving fraud.
Falsifiability and prediction — A scientific claim should permit evidence that could count against it. Predicting every outcome after it occurs makes a theory difficult to test. A risky advance prediction and an appropriate design help distinguish genuine testing from reinterpretation after the result.
Theory and hypothesis testing — A theory organises an explanation and generates testable hypotheses; findings can refine rather than simply confirm it once and for all. Competing theories may predict the same result. A strong test identifies an outcome that discriminates between alternatives.
Paradigms and shifts — A paradigm is a shared framework of assumptions, methods and questions within a research community. A paradigm shift changes a substantial framework rather than merely adding one new study. Whether psychology has one or several paradigms is itself open to argument using the meaning of the term.
Abstract and introduction — An abstract concisely summarises the aim, method, main result and conclusion. The introduction develops the rationale, relevant theory and evidence leading to the question or prediction. A reader should understand why the investigation was needed before seeing the outcome.
Method and results — The method describes participants, design, materials and procedure sufficiently for scrutiny and appropriate replication. Results present relevant descriptive and inferential outcomes with transparent units and reporting. Do not hide unexpected findings or present a hypothesis test as the probability that the hypothesis is true.
Discussion and referencing — The discussion interprets findings in relation to the prediction, theory, limitations, alternatives and justified applications. Referencing identifies sources so claims can be traced and credit preserved. A polished report does not substitute for an honest record of what was actually done.
Test yourself
50 questions · Random sets of 10. These quick checks support revision; practise longer explanations and justified judgements too.
Mr Kidman's Classroom
