The frequency of error in syllogism solving suggests that not all subj
ects are using logic. The atmosphere and matching hypotheses suggest w
hat they might be doing instead but predict some of the same (correct
and incorrect) responses. Reexamination of the data supporting the atm
osphere hypothesis (Sells, 1936) shows that the procedure employed was
unsatisfactory and that the results obtained support the matching hyp
othesis as well as they support the atmosphere hypothesis. It is argue
d on theoretical grounds that the matching hypothesis should be prefer
red. An experiment is reported in which subjects (N = 71) were require
d to draw conclusions from syllogistic premises and to construct premi
ses from which given conclusions followed. It is shown that subjects m
ay be divided into three groups: (n = 16) consisting of subjects who u
sed logic and made few errors; (n = 25) of subjects whose correct and
incorrect responses were in accordance with the matching hypothesis; a
nd (n = 30) of subjects who were not matching but trying to do logic a
nd not doing it well.