(18) Immediate use of intonational cues in a discourse-based visual search task

Intonational prominence signals the informational status of words in discourse. Previous production experiments demonstrated that speakers used L+H* pitch accents more frequently than others (H* or L*) to express the contrastive status of words. Two eye-tracking experiments tested whether distinctive intonational cues have a consistent effect on listeners'comprehension of utterances. Eyemovements were monitored while participants followed pre-recorded audio instructions to decorate Christmas trees. Participants sat in front of a grid with ornaments, and placed the ornaments on the tree following instructions such as ''At the top, hang the blue ball''. Eight sets of targets (balls, angels, eggs, bells, drums, candies, onions and stockings) and four sets of distracters were painted in eleven colors (blue, green, orange, red, gold, silver, brown, gray, yellow, white and purple), and arranged by object type on the grid. Participants' eye movements were with a head-mounted eye-tracker (ASL e5000). Experiment 1 compared fixations for felicitous and infelicitous use of L+H*. Fixation latencies to target cells were shorter when L+H* felicitously marked contrast on the color adjective (L+H* on BLUE in: ''First hang the red ball. -> Next, hang the BLUE ball''), than when it infelicitously marked an immediately repeated noun (L+H* on BALL in: ''Hang the red ball. -> Next, hang the blue BALL.''). In contrast, felicitous contrastive L+H* on the noun (''Hang the blue ball. -> Then, hang the blue DRUM.'') did not facilitate eyemovements to the target as compared to L+H* on the immediately repeated color adjective (''First, hang the blue ball. -> Then, hang the BLUE drum.''). These results suggest that listeners can 'tune' to tonal cues that are relevant to the task (i.e. useful in searching a grid organized by object type). Experiment 2 confirmed the effect of L+H*, using infelicitous instructions without L+H*. Again, eyemovements latencies were shorter when L+H* marked contrast on color (red ball -> BLUE ball vs. blue ball), but not when it marked contrast on the noun (blue ball -> blue DRUM vs. blue drum). 'Garden-path' effects were shown when L+H* was infelicitously placed on the adjective preceding a non-repeated noun (red onion -> GREEN drum). We found early fixations to the preceding target cell (onion), while fixations to the real target (drum) were delayed. The timing of such early fixations suggests immediate use of intonational cues in predicting the upcoming target. This is the first study to demonstrate this anticipatory effect of intonation using eye-tracking technique.