Methodology reference

What is projective response analysis?

Projective response analysis is a structured method for recording what a candidate says about an ambiguous stimulus, assigning that response to defined classes, and relating the resulting pattern to pre-specified constructs. Obscura treats those stages as separate evidence operations. An observation is not yet a classification, and a classification is not a conclusion about a person.

Observation records the response

Observation is the narrowest stage. It preserves the candidate's words, the stimulus presented, the order of presentation, and response timing. It can also record an absence of response or an answer that cannot be reliably transcribed. The observer does not decide whether a response is insightful, typical, or relevant to a role. The purpose is to retain the event before a scoring rule is applied.

This separation follows the modular logic described by Lievens and Sackett's The effects of predictor method factors on selection outcomes in Journal of Applied Psychology (2017), which treats selection procedures as combinations of measurement factors rather than as indivisible impressions. Obscura's public research bibliography identifies that paper and the other sources used in its methodology record.

Classification applies a defined vocabulary

Classification converts an observed response into one or more response classes. A class may describe content, form, latency band, or the presence of a defined feature. The classification rule must be specified before it is used. Two analysts should therefore be able to inspect the same response and identify the same applicable classes, even if they later disagree about what those classes should predict.

Classification is narrower than interpretation. Calling a response a bilateral figure, for example, is a coding decision. It does not establish a candidate's risk tolerance, interpersonal style, or likely performance. Those claims require separate construct definitions and validation evidence. The distinction prevents a descriptive label from acquiring more meaning than the scoring model has earned.

Interpretation connects patterns to constructs

Interpretation examines a pattern of classified responses against a construct model. The model states which features are combined, which are weighted, and what outcome criterion is relevant. It also sets boundaries around what the result cannot establish. A percentile is a comparison within a defined reference population; it is not a diagnosis and does not describe a candidate's complete capability.

Ryan and Ployhart's review, A century of selection (2014), describes the long history of combining selection methods and the need to examine how a procedure relates to the criterion it claims to predict. That principle applies directly here: interpretation must remain tied to the role-relevant criterion and the validation sample.

Why the separation is operationally useful

Keeping the stages distinct makes review possible. A client can ask whether a response was captured accurately, whether the classification rule was applied consistently, or whether the construct inference is supported by validation data. These are different questions with different remedies. A transcription problem is corrected at observation. An ambiguous coding rule is revised at classification. Weak criterion evidence requires a limitation on interpretation or a new validation study.

Obscura's method therefore presents a response pattern as a scored assessment signal, not as a complete account of a candidate. The record is useful when the chain from response to decision remains inspectable.

Last updated: 28 May 2026. Related record: what limits an assessment's use?