Poster #49 - Children draw on a concept’s central form to understand ambiguous sentences

• In Event: Poster Session 03

Abstract

A distinctive component of human cognition is our capacity to combine concepts together (Chomsky, 1957; Pinker, 1994). Yet even simple combinations produce a structure that is ambiguous with respect to its precise meaning. Consider a sentence like “The café is efficient.” The human mind encodes a vast amount of information as part of our broad, generic knowledge of cafés—properties like having electricity, wifi, music, outdoor seating, places to work, windows, restrooms, along with many others (Gelman, 2003; Leslie, 2012). But to which of these properties does “efficient” apply? Considering each interpretation would be computationally costly for the organism. And yet, humans rapidly generate and interpret combinations like these all the time during everyday life. Understanding how humans solve this problem is crucial to understanding the emergence of complex thought. Here, we propose that from early in language development an important bias exists which helps children narrow this vast hypothesis space.

Prior research demonstrates that adults narrow this hypothesis space by prioritizing interpretations in terms of the central form of each concept (Haward, 2022). The central form represents a subset of a concept’s properties in a privileged manner (e.g., representing serving hot drinks for the concept café, called a form property). For example, form properties are conceived of as permanent parts of the kind (e.g., “cafés will serve hot drinks forever”) and exhibit psychological certainty (e.g., “I’m certain the café will serve hot drinks”). Even when form properties (e.g., serving hot drinks) and broader ‘generic knowledge’ properties (e.g., having restrooms, electricity, etc.) are both seen as highly generalizable to instances of the kind, adults still overwhelmingly favor interpretations in terms of form properties. Adults favor these ‘form interpretations’ for concepts from any domain, and even for novel concepts.

We present data that this bias is early emerging, present by at least age four, and so may be crucial to language learning. On each trial, preschoolers (n=149, M=5.48 years, SD=.82) heard a character, Ali Alien, produce an ambiguous sentence about an instance of a kind (e.g., “The kitchen is new”; see Fig. 1). Two friends, Sarah and Tim, also heard this sentence, and each made a guess about what Ali was saying. One guessed an interpretation that involved a form property (e.g., “The kitchen’s oven is new”) while the other guessed an interpretation that involved a generic property (e.g., “The kitchen’s clock is new”). The child was then asked which character made the best guess. Stimulus checks confirmed that both properties licensed a true generic, and a posttest confirmed that the children had encoded these form properties as form properties. Strikingly, children interpreted sentences in terms of form properties (61%) more than generic properties (39%; p<.001), including by age four (n=50; 64% vs 36%; p<.001). Given the early emergence of this bias, we discuss whether it is required for the emergence of complex thought and preview ongoing work investigating whether children prioritize form properties in this way for novel concepts.

Authors

• Paul Haward, University of California - Berkeley
  Presenting Author

• Mahesh Srinivasan, University of California - Berkeley