Search for: All records

Successful communication is thought to require members of a speech community to learn common mappings between words and their referents. But if one person’s concept of CAR is very different from another person’s, successful communication might fail despite the common mappings because different people would mean different things by the same word. Here we investigate the possibility that one source of representational alignment is language itself. We report a series of neural network simulations investigating how representational alignment changes as a function of agents having more or less similar visual experiences (overlap in “visual diet”) and how it changes with exposure to category names. We find that agents with more similar visual experiences have greater representational overlap. However, the presence of category labels not only increases representational overlap, but also greatly reduces the importance of having similar visual experiences. The results suggest that ensuring representational alignment may be one of language’s evolved functions. 

Generic statements like “tigers are striped” and “cars have radios” com- municate information that is, in general, true. However, while the first state- ment is true *in principle*, the second is true only statistically. People are exquisitely sensitive to this principled-vs-statistical distinction. It has been argued that this ability to distinguish between something being true by virtue of it being a category member versus being true because of mere statistical regularity, is a general property of people’s conceptual machinery and cannot itself be learned. We investigate whether the distinction between principled and statistical properties can be learned from language itself. If so, it raises the possibility that language experience can bootstrap core conceptual dis- tinctions and that it is possible to learn sophisticated causal models directly from language. We find that language models are all sensitive to statistical prevalence, but struggle with representing the principled-vs-statistical dis- tinction controlling for prevalence. Until GPT-4, which succeeds. 

Why are some words more frequent than others? Surprisingly, the obvious answers to this seemingly simple question, e.g., that frequent words reflect greater communicative needs, are either wrong or incomplete. We show that a word’s frequency is strongly associated with its position in a semantic association network. More centrally located words are more frequent. But is a word’s centrality in a network merely a reflection of inherent centrality of the word’s meaning? Through cross-linguistic comparisons, we found that differences in the frequency of translation-equivalents are predicted by differences in the word’s network structures in the different languages. Specifically, frequency was linked to how many connections a word had and to its capacity to bridge words that are typically not linked. This hints that a word’s frequency (and with it, its meaning) may change as a function of the word’s association with other words. 

As adults, we continue to learn new word meanings. We can learn new words through ostensive labeling events where a word denotes a clear referent in context, or by having the word explicitly defined for us (Hahn & Gershkoff- Stowe, 2010). However, people also learn word meanings through exposure to how words are used in text (Nagy et al., 1985; Saragi et al., 1978). Here, we examine the relative effectiveness of different ways of learning new word meanings, finding that more ostensive experiences are not necessarily more effective than indirect learning via merely observing how a word is used.