(PP1329) Is Effortful Guessing a Part of Listening Effort? Data Processing in a Dual-Task Paradigm

Should guessing during speech recognition count as “listening effort?” Although a common listening behavior, excessive guessing could confound the effects under investigation. We re-analyzed data from two previous studies that utilized a dual-task paradigm to examine how reverberation affects listening effort. We aim to compare and contrast two different data analysis methods (i.e. including all response times versus including response times only when word recognition is correct). Preliminary analysis indicated that including all response times was a more sensitive approach to measure changes in listening effort. Thus, we believe responses while ‘guessing’ should be included in measures of effort.

Summary: 

Rationale

Behavioral measurements of listening effort are becoming increasingly popular in research laboratories, with potential for clinical use in the future. Yet, it is unanswered, and largely ignored, whether we should include all response times during testing, or only the ones where a participant correctly recognized the speech? Including all response times better applies to real-life listening situations, as guessing is a common behavior when speech understanding becomes challenging. However, excessive guesses when listening becomes challenging due to fatigue or frustration can contaminate the dataset . Although studies in our laboratory routinely use all responses, many other investigators report using only responses from correct responses, and others do not even report their choice. It is not clear which measure is most sensitive to changes in listening effort, nor is it clear if both strategies will lead to similar conclusions. In other words, does including the responses during trials where participants incorrectly guess about the speech affect our interpretation of listening effort?

Design

Data obtained using the above-mentioned two different methods (all responses, correct responses) collected previously (Picou, Gordon, & Ricketts, 2016; Huang, Ricketts, Hornsby, & Picou, in progress) were re-analyzed to investigate whether this difference in data analysis would change the outcome of a study. Both studies examined the effect of reverberation on listening effort in adults with normal hearing, specifically, under moderate and high reverberation. The response times for each study were analyzed separately, using linear mixed effects models. For each study, initial models included main effects of background noise level and reverberation times and a random factor of participant. In addition, a fixed factor of response time type (all included, only included for correct responses) was included for both sets. Significant main effects and interactions were carried forward to parsimonious models, which were used for pairwise comparison testing, adjusting for family-wise error rate.

Results and Conclusions

Preliminary results indicate that two methods yielded different pattern of results. When all response time data were included, increases in reverberation significantly increased listening effort. However, when only the response times with correct word recognition were included, this effect was nonsignificant. In addition, average response times were longer in all conditions when we included all responses compared to using the alternative method, even in the hardest condition. If excessive guessing prevailed, when we included all response times, we should be seeing the opposite result. Therefore, including all response times seems to be the better option. However, including both methods in studies might be valuable. Specifically, the numeric difference between results obtained using the two methods would be the amount of effort the listeners exerted when guessing and incorrectly recognizing the word. This value might reflect listeners’ motivation in adverse situations.

Learner Objectives

Participants will be able to compare and contrast two different methods of handling response time data in a dual task paradigm. For clinicians that are planning to implement behavioral measures of listening effort in clinical testing, the current study may provide some guidance on response time interpretation.