Poster D54, Monday, March 26, 8:00-10:00 am, Exhibit Hall C
The unbearable lightness of meaning: Linking adjective informativity and flexibility
Sarah Solomon1, Sharon L Thompson-Schill1; 1University of Pennsylvania
The informational content of individual words is flexible across contexts. Even simple adjectives vary in their informativeness when paired with different nouns: “dark” is more informative in “dark paint” than it is in “dark charcoal.” The challenge is thus to propose a theory of conceptual structure and combination that is consistent with these flexibility effects. Entropy is a measure from information theory that reflects the information content of a signal; here we use it to reflect the informativity of an adjective with respect to a specific noun, and test whether this construct can predict adjective flexibility. We paired the scalar adjectives of “light” and “dark” with 45 object noun concepts which spanned the conceptual brightness range (e.g., “snow”, “fur”, “charcoal”), and collected darkness probability and entropy values for each concept. Explicit darkness value judgments were made by adjusting a slider above a visual bar that ranged from white to black. Some subjects reported the darkness value for each of the 45 unmodified concepts (e.g., “snow”); others reported the darkness value for 45 modified concepts (e.g., “dark snow”, “light fur”). For each concept, we calculated the “dark” and “light” effect by calculating the extent to which darkness was modulated by each modifier, and calculated the general adjective effect for each concept. We found that entropy did predict the general adjective effects, suggesting that adjective informativity determines the amount of property modulation involved in the comprehension of adjective-noun combinations. This quantitative model facilitates the study of the neural underpinnings of conceptual flexibility.
Topic Area: LANGUAGE: Semantic