In Skinner's Early Footsteps: Analyzing Verbal Behavior in Large Published Corpora

Abstract

Using data from large databases of words that listeners have rated in terms of how positively or negatively they respond to them, we present a mode of analysis that reveals regularities across several languages. Specifically, a normalized measure of bias in emotional valence shows that the degree to which positively-valenced words predominate in a language is a reliable function of absolute valence intensity. This measure supports the comparison of corpora in which valence has been measured in different ways and can serve as a baseline for evaluating the role of interactive variables. We also suggest a way to apply the measure to characterize the verbal behavior of individuals. Throughout we show how the general topic is amenable to the functional perspective that behavior analysts apply to verbal behavior, which creates an opportunity for behavior analysts to contribute to interdisciplinary inquiries into verbal phenomena.