Belief, Incongruity, and Syllogistic Reasoning
Lawrence T. Frase · Psychological Reports · 1966
In a recent srudy (3), words which had been rated on the evaluarive scales of the Semantic Differential were combined to produce sentences varying in three levels of incongruiry (5). Error and rime scores on validiry judgments of syllogisms were influenced systematically when these sentences were used in syllogistic conclusions. The questions explored in this study dealt with the nature of the relationship between beliefs in conclusions and the amount of incongruency of the sentence used in the conclusion. One week after taking the syllogistic reasoning test the 64 Ss from the original study rated rhe 48 syllogistic conclusions on a series of probabilistic scales designed to measure belief (2). Evaluative scales were used as filler items and scale directions were alternated. Half of the 48 syllogistic conclusions were universal (e.g., All mothers are evil) and the other half were particular (e.g., Some mothers are evil). The mean incongruity rating of each conclusion was correlated with its average belief raring. The Pearson correlations were -.52 (p < .01) and -.66 (p < .01) for the universal and particular conclusions, respectively. These results support Osgood and Tannenbaum's notion that the more incongruous an assertion is, the less its credibility. In the original study (3) the effect of highly incongruous conclusions upon error scores was less than the effect of moderate incongruity, suggesting a function akin to Osgood and Tannenbaum's incredulity factor in syllogistic reasoning. The original srudy showed there were more errors on particular as opposed to universal conclusions. The point biserial correlation between belief and quantifier (universalparticular) was computed across all 48 conclusions (r~,!, = .91, p < .01). Evidently, incongruity predicts belief in assertions between individual words, but neither incongruity nor belief predicts error or time scores (correlations were essentially zero for universal and particular conclusions separately). Qualifiers, on the other hand, are a powerful variable in manipulating Ss' belie!s in assertions, their errors and time scores. The present results indicate that, although attitudes (evaluative ratings) may be manipulated by asserting relations between selected attitude objects (I), both beliefs in those assertions and reasoning behaviors (which are independent of belief to some extent) can easily be manipulated by quantifying assertions. In terms of a mediarion model, belief in an assertion is contingent upon the incompatibilities of the words associated (qualifiers), and even more critically, in the number of relations which are asserted within a given hierarchy (quantifiers). Osgood (4) has suggested the differential effecrs of qualifiers which this study reveals.