Assessment methods and what each can and cannot show

Notions clés

Self-report

Questionnaires and rating scales are efficient and give access to inner experience, but they depend on insight, memory and willingness to disclose. Global retrospective reports are open to recall bias; ecological momentary assessment (EMA) samples current experience repeatedly in daily life to reduce it.

Multi-informant reports

Parents, teachers, partners and supervisors see behavior in different settings, so their reports agree only modestly. Across 341 studies of children and adolescents, cross-informant correlations averaged .28, and a discrepancy can reflect real differences between contexts rather than an error by one informant.

Direct observation

Observing behavior where it occurs, with defined codes and recording rules, shows the conditions before and after a behavior without relying on anyone's recall. It is costly and samples only a limited period and setting.

Interviews, from unstructured to structured

Unstructured interviews are flexible but vary across interviewers; structured interviews fix the questions and scoring; semi-structured interviews add follow-up probes. In one meta-analysis, diagnoses from routine clinical evaluations and standardized interviews agreed only modestly (mean kappa .27).

Performance measures

Ability tests, work samples and simulations show what a person can do under standard conditions, which neither self-report nor informants can show. In personnel selection, work samples sit alongside structured interviews, which ranked highest in validity in a 2022 re-analysis.

Group and organizational methods

Surveys, focus groups, records such as turnover and multisource ratings describe teams and organizations. Self-ratings of job performance correlated only .22 with supervisor ratings in a meta-analysis and were more lenient, so several rating sources are combined.

Exemple commenté

A hotel chain asks a psychologist to assess the guest service skills of the front desk staff at its eight hotels so that it can plan training. The operations director proposes simply asking each employee to rate their own skills.

1. Self-ratings are quick, but they tend to run more lenient than supervisor ratings and agree with them only modestly, so they cannot stand alone.
2. A structured behavioral interview about recent guest problems, scored on anchored scales, adds consistent evidence of how each employee says they handle service situations.
3. A work sample, such as a standard role-played complaint scored by trained raters, shows what each employee actually does under the same conditions.
4. Supervisor ratings and records such as guest complaints add the workplace view; where the sources agree and disagree points to specific training needs.

Combine methods whose errors differ, so that each covers what another misses.