Researchers:  Dr. Sam Tilsen (Linguistics), Dr. James Sethna (Physics),  Dr. Jeremy Needle (Linguistics)

This project asks how “spontaneous” the emergence of dialects is. There are many reasons why dialectal variation arises within languages, including geographic, historical, and socioeconomic factors. Dialect formation is challenging to study in a controlled way because of this complexity, along with logistical issues in observing speech behavior over time on a large scale.

Each unit of the experiment itself involves groups of around 10 participants who play the game together multiple times a week, for about 3 weeks in a row. In the game, pairs of participants have to share information to navigate a map, and the map is labeled with novel words that might have multiple possible pronunciations. (See example map)

To study this difficult topic, we are developing three large pieces of Polylect:

(1) a custom online video game where participants talk together to accomplish a task;

(2) an audio-processing workflow that takes the many recordings of game conversation and applies different speech recognition and forced-alignment methods to enable phonetic analysis of many thousands of words; 

(3) a quantitative modeling method that draws on statistical physics and other techniques to make sense of the way participants' speech changes over the course of the study as they influence each other.