1.1 - Experiments and causal claims

Manipulation and causal reasoning

An experiment tests whether changing one variable produces a change in another.

The independent variable (IV) is the condition varied by the researcher in a true experiment. The dependent variable (DV) is the measured outcome. A researcher might compare learning a word list in silence with learning it while quiet instrumental music plays. The IV is the sound condition; the DV could be the number of words correctly recalled after two minutes. Naming “memory” alone is less precise than stating how memory is measured.

Other factors can also affect recall. These extraneous variables include the difficulty of the lists, the learning time and participants' previous knowledge. Keeping the lists equally difficult and giving everyone the same learning time makes the comparison fairer. If the music group always gets a harder list, list difficulty varies systematically with the IV: it is a confounding variable, providing an alternative explanation of the result.

Where feasible, randomly allocating people to conditions helps distribute individual differences without systematically favouring one condition. This differs from randomly selecting a sample from a population. The details of allocation and experimental designs are developed later in the course.

Manipulation, control and a fair comparison support a causal inference: a claim that the IV contributed to the DV difference. They do not make an experiment automatically flawless. A poorly controlled experiment may not isolate its proposed cause, and a well-controlled result in one task may not generalise to every setting or person.

Laboratory experiments

A laboratory experiment manipulates the IV in conditions arranged to give the researcher substantial control.

“Laboratory” describes the controlled research situation, not a requirement for specialist equipment. A quiet room with fixed instructions, identical computers and controlled task timings can serve this purpose. In the recall example, headphones can deliver the same music volume to each participant, while the screen presents words for the same duration.

This control improves internal validity when plausible alternative causes are reduced. Standardised materials and procedures also make replication, repeating the study to check its findings, easier. Consistent procedures support reliability; they do not guarantee that every replication will obtain identical scores.

Control can come at a cost. Remembering isolated words in a room may differ from revising meaningful notes at home. Ecological validity concerns whether the task and setting reflect behaviour in everyday contexts; artificiality can limit that particular generalisation. It does not show that every laboratory finding is useless. A simple task may isolate an important process even if further research is needed outside the laboratory.

Participants may also work out the aim from cues in the procedure. These demand characteristics can change behaviour: someone who thinks the researcher expects music to help might make an unusual effort in that condition. Neutral instructions and avoiding unnecessary cues can reduce this risk. Ethical information and the right to withdraw still matter; hiding the aim is not automatically justified.

Field experiments

A field experiment manipulates the IV in an everyday setting where the behaviour of interest normally occurs.

For example, researchers could vary whether a library sign uses a polite request or a direct instruction, then measure how many visitors lower their voices. The sign wording is deliberately changed; merely watching library visitors without changing a condition would be an observation, not a field experiment.

The setting and activity may be more natural than an arranged laboratory task. Visitors who do not know the precise research aim may show fewer demand characteristics. These features can improve the relevance of the findings to that library behaviour. Awareness is a separate issue: a field experiment can be overt, and a natural setting alone does not guarantee natural behaviour.

There is usually less control over outside events. One sign might be displayed during examinations, when the library is unusually crowded, and the other during a quiet week. Crowding, rather than wording, could affect noise. Using comparable times and locations reduces this problem, although it cannot control everything. Recreating the same setting exactly is also harder, so replication may be less straightforward.

Consent, privacy and withdrawal need careful planning when people encounter a manipulation in an everyday setting. Researchers must consider risk and whether observation or recording is acceptable; “real life” is not an ethical exemption. The choice between field and laboratory methods depends on whether the immediate priority is tight control or evidence about behaviour in its usual context.

Quasi experiments

A quasi experiment compares conditions defined by a naturally occurring difference rather than an IV manipulated by the researcher.

Researchers might compare adults with an existing hearing impairment with adults without one on the same visual-memory task. Hearing status is the naturally occurring IV and visual-memory score is the DV. Researchers cannot randomly assign participants to acquire an impairment. They can still standardise the task and match relevant characteristics, such as age, where appropriate.

This method allows investigation of differences that would be impossible or unethical to create. It is useful when the psychological question concerns a real characteristic or experience. However, the groups may differ in other ways: education, previous training or age might affect memory too. Because the researcher has not created and randomly allocated the group difference, a memory difference cannot confidently be attributed to hearing status alone. Matching known factors improves the comparison but does not remove unknown differences.

Setting and manipulation are separate dimensions. A quasi experiment can take place in a highly controlled laboratory or in an everyday setting. Calling a study “quasi” does not mean it is uncontrolled in every respect; it identifies the origin of its IV. Likewise, a laboratory location does not convert a naturally occurring IV into a manipulated one.