Three experiments are conducted to examine cognitive representations of hypothesis sets in the testing of multiple competing hypotheses. In all conditions, the noncomplementarity of opinion revisions remain a strong phenomenon.