Evidence for direct-access and inferential retrieval in question-answering

“Direct-access” and “inferential” questions were constructed and their properties examined in several converging operations and empirically validated widely used assumptions about direct and inferential retrieval and yielded some information for each.

Abstract

A popular but implicit distinction between direct-access retrieval and inference was put to explicit experimental test. "Direct-access" and "inferential" questions were constructed and their properties examined in several converging operations. Responses to the questions in a four-alternative multiple-choice test were externalized and rated on a direct-access/inference scale. Reaction times to verify singly presented answers were taken for similar questions. The data empirically validated widely used assumptions about direct and inferential retrieval and yielded some information for each: (1) Inference takes longer than direct-access. (2) When direct-access fails, inference is attempted whether or not the question lends itself to inferential processing. (3) Semantic versus phonetic distractors require an additional 500 milliseconds to reject for inferential questions.