Words are composed of sound units called phonemes, which are themselves composed of phonological features. Certain featural differences are acoustically less salient and therefore more difficult to perceive than others \citep*{Miller1955}. As languages differ with respect to their phoneme inventory, they also differ in the use of features to distinguish among phonemes, and listeners have difficulty perceiving featural distinctions that are not used in their language \citep{Sebastian-galles2004}. Finally, as languages have different lexicons, they can differ in the relative importance of used features to distinguish words, which might lead listeners to exploit featural information differentially. This project examines the role of phonological features during word recognition. Our aim is to tease apart universal perceptual biases from language-specific ones that emerge from the lexicon.
We will use methods from computational linguistics to propose a linguistically informed formalism to define the functional load of features (i.e., the extent to which they are used to distinguish words in a given language) and apply it to the lexicons of various languages. Focusing on speakers of these languages, we will then conduct cross-linguistic experiments and collect electrophysiological measures (specifically ERPs). As listeners are known to benefit from visual cues during speech processing \citep{McGurk1976}, we will compare two input modalities: auditory and audiovisual.