October 15: Mits Ota


The Emergence of Sound Repetition Under Pressure for Learning

Mits Ota (work with Aitor San Jose and Kenny Smith)

Tuesday, October 15
11:00am – 12:30pm
S38, 7 George Square

The notion that human languages are shaped by the way they are learned predicts a close match between learning biases and key properties of language, with easier-to-learn features and patterns being relatively common across languages. However, there are counter-examples to this generalisation. In this experimental study, we used one such counter-example in order to explore the idea that whether a learning bias reveals its signature in a linguistic system depends on the amount of learning pressure imposed on the language users, particularly in relation to the amount of pressure for efficient communication. The case in question involves the repetition of phonological elements within a word (e.g., dodo), a pattern that is generally avoided in natural languages despite developmental evidence showing that it is privileged in the context of learning.

 

The experiment was loosely modeled after the cultural-transmission paradigm in Kirby, Tamariz, Cornish, & Smith (2015), and it compared 12 learning-and-transmission groups (or ‘chains’) and 12 communication-only groups (or ‘closed groups’) who learned novel labels for unfamiliar objects. Each chain consisted of 5 sequential pairs of participants who, after learning  the labels, engaged in a communication game using those labels, and then passed them on to the next pair. Each closed group consisted of a single pair of participants who also learned the labels and played the communication game, but repeated the process 5 times among themselves. Additionally, we manipulated the size of the lexicon (12 words vs 18 words) to gauge the impact of pressure for communicative efficiency. The results showed that the amount of consonant repetition in the labels increased over time more quickly in the chains than the closed groups. There was no effect of lexicon size. These outcomes suggest that the potential impact of a learning bias on a linguistic system is conditioned by the degree to which the users are under pressure to learn the exact forms or structures that are transmitted.