Lexical bias and prosodic cues: An eye-tracking study of compound/phrase disambiguation

Abstract

Despite compounds and phrases exhibiting distinct prosodic cues (e.g., pitch and duration), adults fail to use these cues in certain contexts. In two eye-tracking experiments, I examined English speakers' attention to prosody in adjective-noun string disambiguation to explore a lexical bias previously reported (e.g., 'hot dog' is interpreted as the food, regardless of prosody.) In experiment 1, 24 participants were presented with pictures of the compound and phrasal representations for possible compounds (20 known, 10 novel) accompanied by an audio presentation of either a phrasal or compound production. Replicating previous findings, participants sometimes ignored prosody and exhibited biases towards compound pictures for common compounds and towards phrasal pictures for novel compounds. The fixation data reveals similar patterns in which participants appeared to not consider the alternative picture, even when it matched available prosodic cues. In experiment 2, the phrasal/compound pictures were decoupled and placed with unrelated distractors and participants were asked whether either picture matched the audio presentation of a phrasal or compound production. This experiment helps to determine whether the exhibited lexical bias is an artifact of experimental designs using minimal pairs or whether prosody is still ignored when frequency and novelty are not directly competing.