The practical's purpose and boundaries

This practical uses an experiment within learning and cognition. An experiment compares outcomes across conditions to investigate whether changing one variable produces a difference in another. The guide allows a true or quasi-experiment, with at least five participants for repeated measures or ten participants for independent measures. Those are minimum totals. They don't guarantee a representative sample or convincing evidence.

Class practicals are assessed through Paper 2 Section A. Revise how a method is selected, how it is conducted and how it is evaluated, rather than memorizing a supposedly universal procedure that every student follows. The investigation must stay within learning and cognition. It cannot shift to another context just because an experiment would work there too.

For example, a researcher might ask whether spacing practice sessions affects later recall. This hypothetical investigation illustrates the method; it is not a prescribed practical.

True and quasi-experiments

In a true experiment, the researcher manipulates an independent variable and uses random allocation to determine participants' exposure to conditions. With independent measures, participants are randomly allocated to conditions. With repeated measures, condition order can be randomly allocated or otherwise systematically balanced. Random allocation reduces systematic differences between the people experiencing each condition, but it cannot guarantee perfectly equivalent groups.

A quasi-experiment compares experimental conditions when random allocation to those conditions is not possible, often because the groups already exist. Comparing learners in existing teaching groups may be feasible. However, previous attainment or other differences between the groups could explain a performance difference. The comparison is useful, but the causal conclusion needs more caution.

An independent variable has conditions that are manipulated or compared as the proposed influence on an outcome. The dependent variable is measured as the outcome of that comparison. In the practice-spacing example, spacing the practice sessions would be the independent variable. The dependent variable might be the number of items recalled after a specified delay.

Operationalization specifies precisely how a variable will be manipulated or measured. “Learning improved” isn't precise enough. State what was learned and what counted as a correct response, then specify when performance was measured. These details make the investigation understandable and repeatable.

Choosing an experimental design

In a repeated-measures design, the same participants complete every condition. Each person provides a comparison with their own performance, so individual differences are held relatively constant. Fewer participants are needed too. The drawback is that practice, fatigue or experience of an earlier condition can affect performance in a later one.

An independent-measures design uses different participants in different conditions. It avoids the effects of completing both conditions, though differences between participants may influence the outcome. Random allocation helps address that problem. This design usually needs more participants.

The diagram below separates recruitment from allocation and shows how each design uses the same sample differently. The guide requires at least five people completing both conditions for repeated measures, or ten people distributed across conditions for independent measures—not ten in each condition.

Order effects are changes in performance caused by the sequence in which participants experience experimental conditions. Counterbalancing distributes different condition orders across participants to reduce systematic order effects. Some participants might complete one condition first, while others begin with the alternative. A small sample may not allow perfectly equal numbers in each order. Nor can counterbalancing erase learning that permanently changes later performance.

Choose the design to suit the task. Repeated measures is attractive when participant differences are substantial and exposure to one condition is unlikely to contaminate another. Independent measures is preferable if the first condition would reveal the task's purpose or irreversibly change what a participant knows.

What makes an experiment different from other methods?

An experiment is especially useful for a question about a possible causal influence, provided a safe, workable comparison can be created. Control is its strength. Its weakness is that a controlled task may differ from ordinary learning.

Other methods answer different kinds of questions:

• A correlational study measures the relationship between variables without manipulating them. It can establish an association, but cannot by itself establish which variable causes the other or rule out a third explanation.
• A case study investigates a bounded case intensively, such as one person or group, using detailed evidence. It provides depth, but not the controlled comparison of an experiment.
• An interview gathers participants' verbal accounts through questioning. A structured interview uses predetermined questions; a semi-structured interview combines planned questions with follow-up exploration; a focus group gathers accounts through a moderated group discussion. These methods can explore learners' experiences. Reported experiences, though, are not the same as measured performance under manipulated conditions.
• A survey gathers standardized information from a sample, often through a questionnaire, an instrument containing a set of questions for participants to answer. Surveys can efficiently describe reported habits or attitudes, but they do not manipulate their causes.
• An observation systematically records behaviour. Naturalistic observation records behaviour in its usual setting, while controlled observation uses an arranged setting. Overt observation is known to those observed; covert observation is not. In participant observation, the researcher joins the activity; in non-participant observation, the researcher stays outside it. An arranged observation setting is not automatically an experiment. There still needs to be an experimental comparison.

These distinctions help with method selection. They do not alter the guide's requirement to use an experiment for this practical.

Control and conclusions about causality

An extraneous variable is any variable other than the independent variable that could influence the dependent variable. When an extraneous variable varies systematically with the independent variable and offers an alternative explanation for the result, it is a confounding variable.

Suppose the differently spaced practice conditions also receive different total amounts of practice. Any difference in recall could reflect practice amount rather than spacing. Naming the problem isn't enough. The procedure should keep total practice comparable across conditions.

The causal diagram shows why a difference between conditions does not automatically demonstrate that the intended independent variable caused the outcome. An alternative influence can follow the same condition pattern and affect the dependent variable.

Standardization means using consistent procedures across participants and conditions, apart from the intended independent-variable difference. Keep instructions, materials and timing consistent where appropriate, along with the testing environment and scoring rules. Random allocation addresses participant differences; standardization addresses procedural differences. Neither can replace the other.

Control strengthens a causal explanation when the independent-variable difference comes before the outcome and plausible alternatives are reduced. A quasi-experiment usually leaves more uncertainty about pre-existing group differences. Even with a true experiment, the conclusion must stay limited to what was manipulated and measured. One result does not explain learning in every situation.

Sampling: who takes part?

A population is the group of people to which a research question or conclusion refers. The sample is the subset of that population included in the study. Recruitment determines who enters; allocation determines which condition recruited participants experience. Random allocation is not random sampling.

The guide's sampling techniques differ in their practical strengths and limitations:

• Random sampling gives every member of the defined population an equal chance of selection. This recruitment technique reduces selection bias, but requires access to a suitable population list. Selected people may also decline.
• Stratified sampling divides the population into relevant subgroups and recruits participants from each, commonly in proportion to their population representation. It improves representation of the chosen characteristics. However, it requires accurate subgroup information and does not control every relevant difference.
• Opportunity sampling recruits available, accessible participants. It is practical for a small investigation, but may overrepresent one age group, educational setting or level of motivation.
• Self-selected sampling recruits people who volunteer in response to an invitation. It is convenient and supports voluntary participation. Volunteers, however, may differ from non-volunteers in interest, confidence or available time.
• Snowball sampling recruits through existing participants, who help identify further potential participants. It can reach less accessible populations, but connected participants may share characteristics and produce a narrow sample.

A small convenience sample can make an experimental comparison feasible without supporting a broad population claim. Meeting the minimum sample requirement does not remove sampling bias.

Ethical responsibility changes the procedure

Ethical considerations affect what can be manipulated and who can participate, as well as how the task is conducted. A learning task should avoid unnecessary stress, humiliation or interference with participants' education. Taking part should not affect grades or access to teaching.

Informed consent is voluntary agreement to participate, based on adequate information about the research and its foreseeable implications. Give participants understandable information about what taking part involves. Younger participants may need additional permissions. When someone in authority recruits participants, they must avoid pressure to agree.

The right to withdraw entitles participants to stop taking part without penalty. Explain how withdrawal works, including any limits on removing data once they have been fully anonymized.

Confidentiality protects identifiable participant information from unauthorized disclosure. Anonymity means there is no identifying information linking a participant to their responses. Coded scores are not necessarily anonymous: the researcher may still hold a list connecting codes to names.

Deception involves withholding or misrepresenting information about a study. Revealing every detail may change behaviour, but that does not automatically justify deception. Any limited disclosure needs ethical justification and appropriate safeguards. Debriefing is an explanation of the study's purpose and procedures after participation, which also addresses questions or concerns. It cannot make an otherwise harmful procedure acceptable.

Ethical constraints may require a less intrusive task or a quasi-experimental comparison instead of manipulation. That changes the method; it is not merely an administrative detail.

Validity and reliability

Validity is the extent to which a measure or research conclusion supports its intended interpretation. Each form asks a different question:

• Internal validity concerns whether the outcome difference can credibly be attributed to the independent variable rather than an alternative explanation.
• External validity concerns whether findings apply beyond the people, tasks and settings studied.
• Content validity concerns whether a measure adequately covers the relevant aspects of what it is intended to measure.
• Face validity concerns whether a measure appears, on inspection, to measure what it claims to measure.
• Construct validity concerns whether a measure genuinely represents the psychological concept it is intended to assess.

A recall score, for example, may look like a sensible memory measure and therefore have face validity. Yet performance could depend heavily on reading ability, complicating construct validity. A narrow task cannot automatically represent every aspect of learning.

Reliability is the consistency of a measurement or procedure under comparable conditions. Clear instructions and explicit scoring criteria help keep measurement consistent. Reliability does not guarantee validity. A task can consistently measure something other than the intended concept.

Control and realism can also involve a trade-off. A tightly standardized task may strengthen internal validity while making its application to everyday studying less secure. Useful evaluation identifies what the control gains and what the setting limits.

Bias, reflexivity and credibility

Researcher bias is a systematic influence of the researcher's expectations or assumptions on research decisions or outcomes. It may affect instructions or scoring, or shape the interpretation of an ambiguous result. Participant bias is a systematic influence of participants' expectations, motives or response tendencies on their behaviour in the study. They may try to confirm what they think the researcher wants. They might also work unusually hard because they feel evaluated.

Reflexivity involves critically examining how the researcher's position, assumptions and actions influence the research. Announcing an intention to be objective does little to check unconscious bias. A researcher can document expectations before collecting data and inspect whether the instructions favour one condition. Explicit scoring rules help too. Where feasible, another person can check ambiguous scoring without knowing the condition.

Credibility is the extent to which a research account and its conclusions are trustworthy in light of the evidence and procedures. For an experiment, transparent recruitment and a clearly described manipulation strengthen credibility. So do consistent measurement and an accurate record of procedural changes. Conclusions should acknowledge alternative explanations. Do not discard unexpected findings simply because they conflict with expectations.

Generalizing and transferring findings

Generalizability is the extent to which findings can be applied beyond the studied sample or setting. Consider sample size and recruitment method, as well as participant characteristics. Task realism and the similarity of the target population to the sample also matter.

Transferability is the extent to which findings can be meaningfully applied to another context or group on the basis of relevant similarities. To judge transfer, enough procedural and contextual detail is needed to compare the original study with the new setting. A plausible-looking result does not, by itself, justify either a broad population claim or a context-specific application.

A brief task with volunteer students may not apply equally to younger learners or those from different language backgrounds. It may not apply equally to extended classroom learning either. A careful account describes these boundaries and identifies which similarities would support application elsewhere. Replication with other samples or settings can provide further evidence, but do not claim it unless it has actually occurred.

What a clear practical write-up contains

A good account connects the research question to the method instead of listing disconnected procedural facts. Identify the learning-and-cognition question and explain why a true or quasi-experiment suits it. State the independent and dependent variables in operational terms.

Then describe recruitment and participant numbers. Set out the design, allocation or condition order, and materials. Include standardized instructions, scoring and ethical safeguards. Another researcher should be able to understand what participants experienced without having to fill in important gaps.

Keep the findings and conclusion tied to the recorded outcome. In the discussion, explain how control and sampling affect the conclusion's strength and reach, alongside measurement, bias and ethics. Each proposed improvement should address an identified limitation: changing recruitment tackles a sampling problem; changing scoring tackles a measurement problem. Simply increasing the sample does not repair a confound.