TL;DR
This study investigates whether listeners can distinguish between clear and Lombard speech produced by older adults (OAs) in different noisy conditions. The findings suggest that while OAs can produce speech adaptations, these are more readily perceived when both talker and listener are in noise (Lombard) compared to when only the listener is (clear speech), and auditory cues are generally more informative than visual ones.
What Problem Does This Paper Address?
Communicating effectively in noisy environments or with individuals who have hearing impairments often leads speakers to adopt "clear speech" or "Lombard speech" styles. Clear speech is typically an "interlocutor-oriented" adaptation, where the speaker modifies their speech to improve intelligibility for a listener who is having trouble hearing. Lombard speech, on the other hand, is an "environment-oriented" adaptation, where the speaker modifies their speech due to the presence of background noise, even if their hearing is unaffected. While previous research has shown that younger adults can produce clear speech that improves intelligibility, less is known about how older adults (OAs) adapt their speech and whether these adaptations are perceptible to listeners. This paper specifically addresses the question of whether listeners can perceive these speech style changes in OAs, focusing on both auditory and visual cues, and comparing the distinct Lombard and clear speech conditions.
Methods
The study employed a two-part approach: a production study to generate speech stimuli and a perception study to evaluate listener judgments.
Production Study: * Participants: 12 older adults (mean age 71.08) with varying hearing statuses (normal to mild age-related hearing loss) and their younger adult partners. * Task: Participants engaged in a "Diapix" task, a picture-based communication game where they had to find differences. * Conditions: * NORM: Both talkers heard each other normally. * BAB_partner: The younger partner heard the older adult talker in babble noise. This condition was designed to elicit "interlocutor-oriented" clear speech. * BAB_both: Both talkers heard each other in babble noise. This condition was designed to elicit "environment-oriented" Lombard speech. * Recording: Speech and video of the older adult talker were recorded.
Perception Study: * Participants: 32 psychology students. * Stimuli: Short (0.75-4.38s) auditory-only and visual-only clips from the production study. * Experiments: Two experiments were conducted: 1. Clear Speech Perception: Differentiating NORM vs. BAB_partner stimuli. 2. Lombard Perception: Differentiating NORM vs. BAB_both stimuli. * Task: Participants were asked to judge whether the older adult talker was speaking to someone who could hear them well (NORM) or someone who had trouble hearing them (eliciting clear/Lombard speech). * Design: Participants completed both auditory-only and visual-only versions of each experiment, with the order randomized.
Datasets
No new datasets were introduced. The study utilized existing materials:
- DiapixUK: A set of picture materials for eliciting spontaneous speech dialogues, as described in [5].
- Speech Recordings: Newly recorded speech and video data from the 12 older adult talkers and their partners under the three experimental conditions (NORM, BAB_partner, BAB_both).
Key Insights and Ideas
- Perceptual Distinction: Listeners can indeed perceive differences in speech produced by older adults under different noise conditions, suggesting that OAs do make perceptible speech adaptations.
- Lombard vs. Clear Speech Perception: Adaptations in the BAB_both (Lombard) condition were more easily discriminated from normal speech than adaptations in the BAB_partner (clear speech) condition. This implies that the environmental noise-induced Lombard speech is a more robust and perceptually salient adaptation for OAs compared to the interlocutor-oriented clear speech.
- Auditory Dominance: Auditory cues were generally more effective than visual cues for listeners to distinguish between speech styles. This highlights the primary role of the auditory channel in perceiving these speech modifications.
- Variability in OA Production: There was significant variability among older adult talkers in how effectively their speech modifications were perceived. This suggests that individual differences, potentially related to factors like hearing status, play a crucial role.
- Hearing Status Hint: An intriguing preliminary finding suggests that OAs with age-related hearing loss might produce speech modifications that are perceived less effectively, though this requires further investigation.
Key Results
- Auditory-only NORM vs. BAB_partner (Clear Speech): Participants made more errors classifying BAB_partner stimuli as NORM (42.24% errors) than vice-versa (23.22% errors). This indicates that clear speech adaptations were often not perceived, and listeners tended to misclassify them as normal speech. However, the error rate for BAB_partner stimuli was still significantly better than chance (50%), meaning some perception did occur.
- Visual-only NORM vs. BAB_partner (Clear Speech): Similar to auditory, participants made more errors classifying BAB_partner stimuli (47.18% errors) than NORM (31.84% errors). However, the error rate for BAB_partner stimuli was not significantly different from chance (50%), suggesting visual cues alone were insufficient to reliably detect clear speech adaptations.
- Auditory-only NORM vs. BAB_both (Lombard Speech): Participants made far fewer errors in classifying both BAB_both (15.7%) and NORM (12.0%) stimuli. This indicates that Lombard speech adaptations were much more readily perceived than clear speech adaptations.
- Visual-only NORM vs. BAB_both (Lombard Speech): Error rates were higher for BAB_both (37.7%) than NORM (25.6%), suggesting listeners tended to misclassify Lombard speech as normal speech when relying solely on visual cues. However, both error rates were significantly lower than chance.
Limitations and Future Work
- OA Specificity: The study focused solely on older adults. It is unclear if these findings generalize to younger adults, and future work should compare OA and YA productions and perceptions.
- Hearing Status: While a hint was found regarding the impact of age-related hearing loss, this was not a primary focus and requires more rigorous investigation.
- Visual Cues: The limited success of visual cues in perceiving speech modifications suggests further research into specific visual features that might be more informative.
- Listener Factors: The study used young listeners. Investigating how older listeners perceive these adaptations would also be valuable.
- Intelligibility vs. Perception: The study focused on the perception of speech style rather than the intelligibility advantage. Future work could link these two aspects.
Summary updated