Correlates of intelligence in computer measured aspects of prose vocabulary: word length, diversity, and rarity

Abstract

Computer measured aspects of prose vocabulary as correlates of intelligence are of interest because they offer the potential of assessing intelligence in situations where more direct assessment (e.g. through IQ tests) is either impractically expensive or (as in the case of populations that lived in the past) impossible. This study assessed a word-length measure (average number of letters), two word-diversity measures (ratio of number of different to number of total words, and Yule's Characteristic K, which indicates the repeat rate for words), and a word-rarity measure (proportion of words present on a rare-words list). In the first part of the study, essays of 120 students in Grade 11 and 12 in a private American high-school for whom Cooperative School and College Ability Test (SCAT) scores (which correlate with IQ test scores) were available, were assessed in terms of the vocabulary measures. Only the word-rarity and word-length measures correlated significantly with SCAT scores, and the highest correlations were manifested by the word-rarity measure. In the study's second part, the vocabulary measures were applied to articles selected from American newspapers representing African American (119 articles), general (110 articles), and Jewish-American (109 articles) communities, among which, for whatever reasons, reliable average IQ performance differences have been found. Only the word-rarity measure discriminated in the predicted way among the three sorts of newspapers. Implications for other potential uses of the computerized word-rarity measure for assessing temporal, social, and geographic group differences in intelligence are discussed.