Current Events

The existence of speech errors (e.g., intending to say “coin toss” and saying “toin coss” instead) and tip-of-the-tongue states (where someone knows exactly what they want to say, but cannot find how to say the right word) reveals an important fact about language production: whenever we say a word out loud, our minds need to select and assemble the sounds that make up that word so that we can properly articulate it. This process is called “phonological encoding” and happens quickly and (usually) without our explicit awareness. So how can we learn about what this process is like? I propose that spoken word durations offer us valuable clues into how phonological encoding works. In this talk, I review a pair of foundational patterns from the language production literature: form repetition reduction - where words are produced with shorter durations if they have recently been said or heard - and sequential cueing - where words are produced with longer durations if a word that begins with the same sounds has been recently said or heard - and I propose a theory that explains both effects as emerging organically as artifacts of the phonological encoding process. Specifically, the theory claims that the phonological encoding system learns after every production, and it makes testable predictions about the factors that shape repetition reduction and sequential cueing. I present simulations from a computational implementation of this theory to show that it can correctly account for the classic repetition reduction and sequential cueing effects. Lastly, I present novel predictions made by this theory and present data from 2 behavioral studies supporting one of these predictions: that contrary to previous proposals, repetition reduction and sequential cueing effects arise purely from the sequencing of sounds, and do not rely on lexical (i.e., word) representations.

TBA