Two Indices Are Better than One: Building on Robbins, Nolan, and Chen (2017)

Abstract

When informants from a given culture are asked to list items from a specified semantic domain, their lists provide two indicators of each term’s prominence or salience: its frequency of appearance across lists and its mean position within the lists that include it. Smith et al. (1995), Sutrop (2001), and most recently Robbins et al. (2017) have defined salience measures that combine these two sources of information. We argue that although frequency and mean position are correlated, the association between them is not perfect and not linear. Plotting them as separate axes of a scatterplot can be informative and complementary to combining them in a single salience measure. We illustrate this with scatterplots for three semantic domains: color terms, animal names, and body parts.