Mapping semantic space: property norms and semantic richness

Emiko J. Muraki,David M. Sidhu,Penny M. Pexman

DOI: https://doi.org/10.1007/s10339-019-00933-y

2019-09-24

Cognitive Processing

Abstract:In semantic property listing tasks, participants list many features for some concepts and fewer for others. This variability in number of features (NoF) has been used in previous research as a measure of a concept's semantic richness, and such studies have shown that in lexical-semantic tasks responses tend to be facilitated for words with high NoF compared to those for words with low NoF, even when many other relevant factors are controlled (Pexman et al. in Psychon Bull Rev 9:542–549, 2002; Mem Cogn 31:842–855, 2003; Psychon Bull Rev 15:161–167, 2008; Goh et al. in Front Psychol, 2016.  https://doi.org/10.3389/fpsyg.2016.00976). Furthermore, shared features (i.e., features that are shared by multiple words) appear to facilitate responses in lexical-semantic tasks to a greater degree than distinctive features (Devereux et al. in Cogn Sci 40:325–350, 2016; Grondin et al. in J Mem Lang 60:1–19, 2009). This previous work was limited, however, to relatively small sets of words, typically those extracted from the McRae norms (McRae et al. in Behav Res Methods 37(4):547–559, 2005). New property listing norms provide the opportunity to extract NoF values for many more items (Buchanan et al. in Behav Res Methods 51:1849–1863, 2019). The purpose of the present study was to test whether NoF effects generalize to this larger item set, and to explore how NoF is related to other measures of semantic richness, including subjective ratings of concreteness, imageability, body–object interaction, sensory experience, valence, arousal, and age of acquisition, as well as more objective measures like semantic diversity, number of associates, and lexical centrality. Using the new Buchanan norms, we found significant NoF effects in lexical decision (is it a word or a nonword?) and semantic decision (is it concrete or abstract?) tasks. We also found significant effects of words' number of shared (less distinctive) features in each task. Further, factor analyses of all semantic richness measures showed a distinct factor structure, suggesting that there are clusters of semantic richness dimensions that seem to correspond to more embodied semantic dimensions and more distributional semantic dimensions. Results are interpreted as evidence that semantic representation is multimodal and multidimensional, and provide new insights about the structure of semantic space.

psychology, experimental