Implicit and explicit attitudes: what is the difference?

Suppose you ask someone how attractive they find a potential employer on a seven-point scale. Their answer is an explicit attitude measure: a directly reported evaluation. You can ask follow-up questions, explore conditions and learn how the person understands the issue.

But answers are influenced by language, self-knowledge, memory, perceived social expectations and the wording of the question. These influences do not mean respondents are lying. They mean self-report needs the same methodological care as any other measurement.

In an indirect procedure, you do not ask only for a direct judgement. Instead, information is inferred from performance on a task. The best-known example is the Implicit Association Test (IAT) introduced by Greenwald, McGhee and Schwartz (1998). Participants categorise stimuli using response keys; their relative performance across different category combinations is analysed.

It is crucial to distinguish an implicit measure - the observed task outcome - from an implicit attitude - the psychological attribute the researcher aims to assess. A measurement procedure does not prove what it measures merely by being labelled implicit. De Houwer and colleagues (2009) argue for examining both the causal role of the intended attribute and the automatic processes responsible for the observed response.

Indirect does not necessarily mean outside awareness. A task effect cannot tell us exactly what someone can or cannot consciously describe.

No. The popular story that questionnaires capture socially desirable claims while indirect tasks reveal the authentic hidden opinion is not supported as a general rule. Disagreements between measures may reflect different constructs, contexts, task effects, measurement error or deliberate evaluation.

Gawronski and Bodenhausen (2006) distinguish associative and propositional processes in an influential account of evaluation and attitude change. Such theories help generate explanations, but cannot by themselves justify diagnosing an individual's secret beliefs.

A standard IAT compares two target categories and evaluation-related attributes using different response-key mappings. Faster or more accurate performance on one pairing can, under suitable assumptions, indicate relatively stronger associations within the design.

An IAT score is not an absolute measure of how much someone likes a particular brand. Its meaning also depends on the comparison category, stimulus set and scoring assumptions.

Reaction times should not be described as directly measuring dopamine, emotions, memories or purchase intentions. They are behavioural data requiring a theoretical interpretation.

The evidence is nuanced and contested. Greenwald and colleagues (2009) reported average predictive associations between IAT outcomes and various criteria in a large meta-analysis, with substantial differences across contexts. Kurdi and colleagues (2019) found associations with intergroup behaviour and contributions from both implicit and explicit measures, alongside considerable heterogeneity. Other research, including Schimmack (2021), raises important questions about construct validity and individual interpretation.

It is therefore not scientifically justified to state that an implicit measure always predicts behaviour better than self-report. The relevant answer depends on the construct, task and criterion. Group-level correlations must not be turned into confident statements about a particular person.

First define what you want to know: attraction, memory, relative association or actual choice. Then select measures that represent those constructs. Do not average heterogeneous metrics as if they were interchangeable.

Control stimulus relevance, language, order effects, training, exclusion rules and statistical uncertainty. Interpret mainly at the level justified by the procedure, and add behavioural or experimental evidence when the question concerns decisions.

Illustrative scenario: respondents say sustainability matters but sometimes choose a cheaper product. This does not prove the self-report was false. Budget, product availability, habits and immediate constraints may all matter. An indirect measure may add evidence in an appropriate design, but cannot replace that contextual analysis.

Indirect measures require transparent purpose, appropriate consent, privacy protection and responsible reporting. They should not be used as unvalidated individual diagnostic tools or as evidence of an applicant's hidden intentions. EEG, eyetracking, IATs and questionnaires measure different phenomena and cannot be combined into a universal mind-reading system.

Explicit measures tell us what people report directly. Indirect measures assess performance patterns that, under appropriate assumptions, may inform researchers about automatic evaluative processes. They can complement each other, but disagreement is not proof that one is false and the other is true.