Purposeful listening in challenging conditions: A study of prediction during consecutive interpreting in noise

Abstract

Prediction is often used during language comprehension. However, studies of prediction have tended to focus on L1 listeners in quiet conditions. Thus, it is unclear how listeners predict outside the laboratory and in specific communicative settings. Here, we report two eye-tracking studies which used a visual-world paradigm to investigate whether prediction during a consecutive interpreting task differs from prediction during a listening task in L2 listeners, and whether L2 listeners are able to predict in the noisy conditions that might be associated with this communicative setting. In a first study, thirty-six Dutch-English bilinguals either just listened to, or else listened to and then consecutively interpreted, predictable sentences presented on speech-shaped sound. In a second study, another thirty-six Dutch-English bilinguals carried out the same tasks in clear speech. Our results suggest that L2 listeners predict the meaning of upcoming words in noisy conditions. However, we did not find that predictive eye movements depended on task, nor that L2 listeners predicted upcoming word form. We also did not find a difference in predictive patterns when we compared our two studies. Thus, L2 listeners predict in noisy circumstances, supporting theories which posit that prediction regularly takes place in comprehension, but we did not find evidence that a subsequent production task or noise affects semantic prediction.

Fig 1. Example of the three conditions of the visual scene.

Example for the sentence "Bob proposed and gave her a ring that had cost him half his monthly wage". The critical image appears in the top left-hand corner in the, from left, Target (ring), Competitor (ribbon) and Unrelated (letter) condition.

Fig 2. Experimental procedure for an experimental trial.

From left to right: 1., Drift correct. 2., A blank screen is shown and the sentence begins. 3., 1000ms before predictable word onset the visual array is shown (here in the Target condition). 4., The sentence finishes and a blank screen is shown. In the consecutive task participants have 12000 to complete their interpretation. 5., The comprehension question appears.

Fig 3. Fixation proportions on target, competitor and unrelated objects in the Noisy Speech study.

The listening task is shown on the left, and the consecutive task on the right. Open circles along the top represent bins during which there was a significant difference between fixation proportions on target and unrelated objects. Filled circles represent bins during which there was a significant difference between fixations proportions on competitor and unrelated objects. Transparent thick lines are error bars representing standard errors.

Fig 4. Fixation proportions for the filler items in the in the Noisy Speech study.

The listening task is shown on the left and the consecutive task on the right. Fixation proportions on target, competitor and unrelated objects when images were presented with the filler sentences. Black dots along the top represent bins during which there was a significant difference between competitor and unrelated conditions. Transparent thick lines are error bars representing standard errors.

Fig 5. Graph showing fixation proportions on target, competitor and unrelated objects in the Clear Speech study.

The listening task is shown on the left, and the consecutive task on the right. Open circles along the top represent bins during which there was a significant difference between fixation proportions on target and unrelated objects. Filled circles represent bins during which there was a significant difference between fixations proportions on competitor and unrelated objects. Transparent thick lines are error bars representing standard errors.

Fig 6. Fixation proportions for the filler items in the in the Clear Speech study.

The listening task is shown on the left and the consecutive task on the right. Open circles along the top represent bins during which there was a significant difference between target and unrelated conditions. Transparent thick lines are error bars representing standard errors.

Fig 7. Graph showing fixation proportions for all experimental trials across both Noisy and Clear speech studies.

Open circles along the top represent bins during which there was a significant difference between target and unrelated conditions. Black dots along the top represent bins during which there was a significant difference between competitor and unrelated conditions. Transparent thick lines are error bars representing standard errors.

Fig 8. Graph showing fixation proportions for all filler trials across both Noisy and Clear speech studies.

Fixation proportions on target, competitor and unrelated objects when images were presented with the filler sentences. Black dots along the top represent bins during which there was a significant difference between competitor and unrelated conditions. Transparent thick lines are error bars representing standard errors.

Fig 9. The (lack of) relationship between training in consecutive interpreting and the extent of prediction.

The y-axis shows the number of months of consecutive interpreting, and the x-axis shows the difference in fixation proportions between the Unrelated and the English competitor (left) and Target (right) conditions.

Fig 10. The relationship between participants’ Lextale result and the extent of semantic prediction.

The y-axis is the LexTale result, expressed as a percent, and the x-axis is the difference in fixation proportions between target and unrelated objects in the period from -550ms before word onset until word onset. Results from the Clear Speech study on the left, and from the Noisy Speech study on the right.