Date of Award

Spring 1981

Project Type


Program or Major


Degree Name

Doctor of Philosophy


Scientific reasoning has become a topic of recent psychological research. Studies have focused on Karl Popper's idea that scientists should try to falsify, or disconfirm, their hypotheses instead of verifying them. Results indicate that both scientists and college students prefer to use confirmatory logic on simple tasks that model scientific reasoning. The only attempt to instruct subjects to use disconfirmatory reasoning failed (Mynatt, Doherty & Tweney, 1977) though subjects that falsified on their own initiative were more successful than subjects who tried to confirm.

All the studies of scientific reasoning have focused on individuals. But major advances in science are often made by groups, e.g., the research teams that discovered the structure of DNA and developed the atomic bomb. Experimental studies of group problem-solving have compared the performance of interacting groups with that of concocted groups composed of an equal number of individuals working separately. When there is a single, right answer to a problem, interacting groups perform about as well as the best member of each equally large concocted group, but better than the average person working alone.

This thesis synthesizes the literatures on scientific reasoning and group problem-solving by combining their two major variables in a single study. Communication was manipulated by running subjects in groups of four and either telling them to interact or to work separately. Strategy was manipulated by instructing subjects to follow either disconfirmatory or confirmatory approaches to the task, which was based on New Eleusis, a card-game designed to model the "search for truth." Each group had to solve the same four increasingly difficult Eleusis problems.

The overall design was a 2 (interacting vs. noninteracting) x 2 (disconfirmatory vs. confirmatory) x 4 (the Eleusis problems) split plot. Analyses-of-variance were conducted on the number of correct solutions and the time-to-solution achieved by groups in each condition.

Even though a manipulation-check revealed that disconfirmatory groups did try to follow their suggested strategy, there were no significant differences in the performances of confirmatory and disconfirmatory groups. This result replicates Mynatt et al.'s (1977) earlier research.

Interacting groups performed no better than the best member of each non-interacting group, where the best is defined as the person who solved each rule in the least time. Interacting groups also took significantly more time. But interacting groups did solve a significantly higher percentage of problems (80%) than all non-interacting individuals combined (33%). These results replicate earlier research on group problem-solving (Steiner, 1972).

A follow-up study, the same task and interacting groups, revealed that disconfirmatory instructions produce superior performance when subjects have maximum freedom to design their own experiments. When the range of possible experiments is limited, confirmatory groups may serendipitously disconfirm their hypotheses.

A discussion of the implications of these results for science and suggestions for future research were included in the thesis.