Framing Expectations: The Effects of Transient Degraded Speech Conditions on Competition in Spoken Word Recognition

Speech changes continually in time. Consequently, for listeners to recognize spoken words, they must piece together the incoming message over time. As a listener hears a word like sandwich, they immediately activate multiple candidates from their mental lexicon with similar onsets (sandwich, sandal, santa). They then integrate further auditory input as it arrives, to favor or disfavor these candidates. This competition takes different forms under degraded listening conditions, for example in listeners with cochlear implants. However, it is unclear whether these differences arise from the degraded input itself, or if listeners refine this competition to adapt to poor input. Thus it was investigated how word recognition unfolds in conditions when the target word is clear, but listeners believe they are listening in noise. For the purposes of this study, a new type of noise, referred to as framed noise, was developed, in which a carrier sentence is presented along with background noise (e.g,. now click on the… ), but the target word (…ball) is clear. This was compared to conditions of complete-noise and no noise. Lexical competition was measured using in the Visual World Paradigm, in which listeners matched a spoken word (e.g., sandal) to one of four pictures on the screen (sandal, sandwich, etc.), while fixations to each picture were recorded, revealing participants’ early interpretations. We found listeners in the noise and framed condition waited for further input before fixating on the target and competitors. Listeners also activated competitors longer in the noise condition, but not in the framed. The results indicate that varying degraded auditory input will influence processing strategies more than expectation.