Measurement is a research process that assigns values or categories to characteristics according to specified rules. In psychology, producing a score isn't enough. We need to ask whether it gives a defensible account of the psychological characteristic we want to understand.

Prospective and retrospective approaches; longitudinal and cross-sectional designs

A prospective approach begins at a starting point and collects evidence about experiences or outcomes that follow. Researchers might, for example, record participants’ current sleep habits, then monitor their attention over the following months. Collecting information as events unfold can reduce reliance on later recollection. Following participants does take time, though.

With a retrospective approach, researchers examine past experiences or events through recollections or existing records. They might ask participants about their sleep during a previous examination period. This gives access to past experiences without waiting for them to occur, but people’s recollections may be incomplete or reconstructed. Existing records can help, though they may lack the measurements the researcher now needs.

A longitudinal research design measures the same participants on more than one occasion over time. By contrast, a cross-sectional research design compares different groups at a single period in time. The figure separates the timing of an approach from the arrangement of a design. These are two decisions, not interchangeable labels.

Longitudinal research can reveal change within individuals, but participants may leave the research. Repeated exposure to a task may also affect their later performance. Cross-sectional research is usually quicker. A difference between age groups, however, does not necessarily show how an individual changes with age: those groups may have grown up under different circumstances.

Keep the distinction precise. Prospective versus retrospective concerns the direction in which evidence is collected relative to events. Longitudinal versus cross-sectional concerns whether researchers measure the same people repeatedly or compare different groups. For example, a longitudinal investigation can use historical records containing repeated measurements.

Interpretive approaches: content analysis

An interpretive approach examines the meanings people give to experiences, actions or communication. It is useful when a numerical score alone would leave out what an experience means to the participant.

Content analysis is an analysis technique that systematically categorizes material such as interview transcripts, written accounts or recorded communication. Researchers first decide which material to include. They then identify relevant categories and apply them to the material. Depending on their purpose, they may count occurrences, interpret meanings or do both.

Imagine analysing accounts of a difficult transition. Statements about avoiding others might go into one category; statements about seeking companionship might go into another. Counting the categories shows how often they occur. Reading the surrounding passages helps explain what they mean. Frequency is not automatically importance: an experience mentioned rarely may still be central to an account.

In practice, specify the unit of analysis—a word, sentence or complete response—and set out clear category rules. Check whether different researchers apply those rules consistently. Keep enough context to avoid distorting a participant’s meaning. The categories themselves form part of the measurement process; they are not a neutral window onto the material.

Brain imaging

Brain imaging is a group of techniques that produces representations of brain structure or activity-related processes. Structural images show anatomical features. Functional images show signals associated with activity during a task or state.

Ask what the technique actually detects. Some functional techniques measure changes in blood oxygenation, rather than recording thoughts or the firing of individual neurons directly. The resulting image needs interpretation. An activity-related signal is not a photograph of a psychological process.

To investigate attention, a researcher must connect the imaging signal to a clearly specified task and an appropriate comparison condition. Movement during scanning can affect the measurement, as can differences between participants. The technique’s ability to distinguish locations or changes over time also matters. An association between a brain signal and task performance does not, by itself, establish that the measured region causes the behaviour.

Statistical significance and type I and type II errors

Statistical significance is a property of a statistical result that meets a predetermined criterion for being inconsistent with a specified null hypothesis. A null hypothesis is a statistical claim specifying no effect, no difference or no association of the kind being tested.

A p-value is the probability of obtaining a result at least as extreme as the observed result, calculated assuming that the null hypothesis and statistical model are correct. A significance level is a decision threshold chosen to judge whether a result is statistically significant. Both quantities are dimensionless.

For example, with a threshold of 0.05, a p-value at or below that threshold meets the chosen significance criterion. It does not mean there is a 95% probability that the research explanation is correct. Nor does statistical significance tell us whether a difference is large or psychologically important, or whether the factor being investigated caused it.

A type I error occurs when a statistical decision incorrectly rejects a true null hypothesis. A type II error occurs when a statistical decision incorrectly fails to reject a false null hypothesis. Put plainly, the first detects an effect that isn't really there. The second misses an effect that is there.

A stricter significance criterion generally reduces the risk of a type I error. With other conditions unchanged, though, it can increase the risk of a type II error. Small samples and inconsistent measurements can make genuine effects harder to detect. A result that is not statistically significant does not therefore prove that no effect exists.

Self-reported and empirical data

Self-reported data are data participants supply about their own experiences, characteristics or behaviour. These reports give access to experiences that may not be directly observable, such as perceived worry. Responses can, however, be affected by memory or by how participants interpret the wording. Attempts to present oneself favourably can affect them too.

Empirical data are data obtained through observation or experience rather than reasoning alone. Examples include behavioural records and instrument readings. Self-reports can also form part of empirical research, so “self-reported” and “empirical” are not mutually exclusive opposites.

A useful comparison is between what people report and what other measurements record. Someone’s reported concentration may differ from the number of omissions they make on a task because the measures capture different aspects of performance. That disagreement does not automatically make one measure worthless. It prompts us to examine what each actually measures.

Constructs, variables and operationalization

A construct is a theoretical concept used to represent a psychological characteristic that cannot be observed directly as a single object. A variable is a characteristic that can take different values or categories across people, situations or occasions. One or more measurable variables can represent a construct, but the construct is not identical to any single score.

Operationalization is the process of defining a variable through the specific procedures used to manipulate or measure it. Researchers could, for example, operationalize persistence as the time a participant continues attempting an optional, difficult task before choosing to stop. The diagram traces the chain from the psychological idea to the recorded observation.

The measure is precise. It is not, however, the whole of persistence. Time spent might also reflect enjoyment of the task or the time available to the participant. Misunderstanding the instructions could affect it too. A useful operational definition states the procedure clearly enough for someone to repeat it and explains why the resulting observation represents the intended characteristic.

For a practical investigation, specify the task and instructions, along with the recording rule and conditions. If you record time, state the unit and exactly when timing starts and stops. These details turn an appealing idea into a measurement another researcher could reproduce.

Choice of research method

A research method is a systematic procedure for collecting evidence to address a research question. Choose the method to fit the question and the kind of measurement required—not simply because it is the easiest procedure to organize.

An experiment lets a researcher manipulate a factor and compare outcomes under controlled conditions. Observation records behaviour in a specified setting. Through interviews, participants can explain experiences in depth; questionnaires can collect standardized responses from a larger group. These methods produce different evidence and have different limitations.

Measuring how often someone interrupts, for example, requires a clear behavioural recording rule. Understanding what interrupting means to that person may require an interview. Neither method answers both questions equally well. Combining methods can broaden the evidence, but only when each contributes something relevant.

When choosing a method, consider access to participants and the time available. Also consider ethical acceptability, the setting and whether the measurement suits the people involved. Then judge what the evidence supports. Even a carefully recorded score needs a clear connection to the research question and a cautious interpretation.