Trait-names: A psycho-lexical study.

Abstract

An experiment was arranged whereby all three judges independently classified into the four selected columns 300 representative words, drawn from the total list according to a principle of representative distribution. The results of this study may be expressed in percentages of the total number of terms that each pair of judges assigned to identically the same columns. Taking only the instances where all three judges agree, we find 141 words or 47 per cent of the list, whereas 6.25 per cent represents the chance expectation. Examining this average agreement more closely we next determine the peculiarities of each individual judge when his placements are compared with those of the other two judges. This analysis calls attention to perhaps the principal source of unreliability, namely the tendency of each judge to have a mental set of "leniency" favoring the inclusion of marginal or doubtful terms in one column rather than another. Four outside judges selected 130 of the 300 terms as strange and unfamiliar to them. The agreement of the three editors for this group of terms averaged only 45 per cent, as against 47 per cent for the total list, and 48 per cent for the remaining 170 more familiar terms. Apparently established usage and familiarity enhances but slightly the reliability of the placement.