Speed and Reasoning Accuracy Predicted by Problem*Premise Interaction

The classification of the bond between a problem and its premise into interaction classes is a new predictor of extensional reasoning performance. Each question consists of a premise describing a specific situation and a problem statement asking for a probability estimate. To grasp a probabilistic question, a decision maker has to mentally represent the extension of possible outcomes which is the basis for extensional reasoning processes. While many studies have addressed the impact of the type of premise, using the number of mental models to measure its complexity, less is known about how the interaction of premise and problem initiates reasoning processes. The type of problem*premise interaction activates different reasoning strategies, depending on the match of a problem description with the initial mental representation, or the absence of such a match, prompting a scan of the extension, or handling a contradiction once it is detected. The problem*premise interaction and number of mental models have similar significance, yet the magnitude of interaction-class effects is larger. Accuracy is highest for contradictions, followed by matching items, and lowest for items requiring an extensional scan. Total answer time and response times are shortest for contradictions and longest for extensional scan. The problem*premise interaction matters regardless of the premise’s complexity, interpretation and number of models required to represent it. The type of problem*premise interaction emerges as the primary determinant of the difficulty of a question together with the number of mental models needed for its representation.* New predictor of extensional reasoning performance based on mental models * Interaction of problem with premise predicts both response time and accuracy * Each problem-premise interaction class has specific processing requirements * Four classes: match, extensional scan, contradiction, perceived contradictionstatus: submitte