Recognition processes in the multiple-choice test : convergent validation of the constructs of direct access versus inferential/constructive processing



Journal Title

Journal ISSN

Volume Title



The present study investigates the idea that answers to questions in the multiple-choice task are answered via distinct processes. The first process was labeled Direct Access processing and referred to the retrieval of information which has been stored in memory due to specific instances of personal history, such as memorizing facts the night before a test. An alternative selection process was labeled Inferential/Construetive processing and referred to the utilization of conceptual information to construct an answer to a question since no previously stored single fact could serve to answer the question. An example of such a question would be: How many times, at most, will the day February 29 occur in any century? Answering this one question demands knowledge of and correct utilization of the concepts of leap year, centuries, and division. Subjects were given two separate tasks in an attempt to convergently validate these constructs. One task consisted of a reaction time task in which subjects were presented questions (which should initiate one specific retrieval process) followed by a single alternative. Subjects were to respond true or false. It was predicted that direct access question sets would be answered more swiftly than inferential question sets. This held true in 39 of 40 subjects, and the "Process" factor accounted for 21% of the variance in RT's. A second task consisted of a multiple-choice test in which subjects were to describe the process by which they selected an alternative. Independent raters rated these descriptions on a 5-point scale ranging from (1) "clearly direct access" to,(5) "clearly inferential." The factor "process" accounted for over 72% of the variance of the ratings and the direct access sets were rated as lower (closer to (1)) than the inferential sets for 40 of 40 subjects. Various ancillary hypotheses, such as linear RT trends based on type of question and type of false alternative were investigated. These results were then discussed in terms of the potential relevance of these constructs and methodologies to cognition, education, and neuropsychology.