An Empirical Study of Abstract Concepts

Felix Hill, Anna Korhonen, Christian Bentz, A Quantitative Empirical Analysis of the Abstract/Concrete Distinction, Cognitive Science 38 (2014) 162–177.

Abstract: This study presents original evidence that abstract and concrete concepts are organized and rep- resented differently in the mind, based on analyses of thousands of concepts in publicly available data sets and computational resources. First, we show that abstract and concrete concepts have differing patterns of association with other concepts. Second, we test recent hypotheses that abstract concepts are organized according to association, whereas concrete concepts are organized according to (semantic) similarity. Third, we present evidence suggesting that concrete representations are more strongly feature-based than abstract concepts. We argue that degree of feature-based structure may fundamentally determine concreteness, and we discuss implications for cognitive and computational models of meaning.

From the concluding remarks (173-174):

Instead of a strong feature-based structure, abstract representations encode a pattern of relations with other concepts (both abstract and concrete). We hypothesize that the degree of feature- based structure is the fundamental cognitive correlate of what is intuitively understood as concreteness.

By this account, computing the similarity of two concrete concepts would involve a (asymmetric) feature comparison of the sort described by Tversky. In contrast, computing the similarity of abstract concepts would require a (more symmetric) comparison of rela- tional predicates such as analogy (Gentner & Markman, 1997; Markman & Gentner, 1993). Because of their representational structure, the feature-based operation would be simple and intuitive for concrete concepts, so that similar objects (of close taxonomic cat- egories) come to be associated. On the other hand, for abstract concepts, perhaps because structure mapping is more complex or demanding, the items that come to be associated are instead those that fill neighboring positions in the relational structure specified by that concept (such as arguments of verbs or prepositions). Intuitively, this would result in a larger set of associates than for concrete concepts, as confirmed by Finding 1. Moreover, such associates would not in general be similar, as supported by Finding 2. [...]

Linguists and psychologists have long sought theories that exhaustively capture the empirical facts of conceptual meaning. Approaches that fundamentally reflect association, such as semantic networks and distributional models, struggle to account for the reality of categories or prototypes. On the other hand, certain concepts evade satisfactory charac- terization within the framework of prototypes and features, the concept game being a prime example, as Wittgenstein (1953) famously noted. The differences between abstract and concrete concepts highlighted in this and other recent work might indicate why a general theory of concepts has proved so elusive. Perhaps we have been guilty of trying to find a single solution to two different problems.

I note that they say nothing about conceptual metaphor theory.