Introduction
Hearing protection is frequently avoided in noisy environments like workplaces or concerts due to concerns about communication interference. Standard passive earplugs often attenuate high frequencies more than low frequencies, creating a "muffled" sound that makes distinguishing consonants difficult. While expensive active hearing protectors can mitigate this, they are often cost-prohibitive.
This study investigates whether low-cost "musician" earplugs, designed for a flatter frequency response, offer a viable alternative. Crucially, it explores the "Lombard effect"—the natural tendency for speakers to change their voice in noise—and how the addition of visual cues (lip-reading) acts as a "great equalizer" in speech perception.
Methodology
The researchers conducted a controlled experiment with 60 Dutch-speaking participants in an ecologically valid setting. Unlike studies using pre-recorded audio, this experiment employed a "live loop" where participants acted as both speakers and listeners. This captured the effects of hearing protection on both speech production and perception.
Participants were tested under six conditions:
- Four types of pre-molded musician earplugs (HPD 1-4)
- One conventional foam earplug (HPD 5)
- A control condition with no protection (6)
Each condition was evaluated in two modes: audio-only and audio-visual (via a webcam feed). The testing environment simulated a realistic social or moderate industrial setting with background music at 81.2 dB(A). Speech intelligibility was measured using standardized Dutch monosyllabic word sets (NVA and Brugse lists) and a 5-point subjective ease-of-use scale.
Results and Analysis
Attenuation Profiles
The performance of the various hearing protection devices (HPDs) was closely linked to their attenuation characteristics. Devices that performed worst in audio-only conditions, specifically the foam earplug (HPD 5) and certain musician plugs (HPD 2), exhibited the highest attenuation in the 500 Hz to 4000 Hz "speech zone."
Speech Perception Performance
In audio-only scenarios, the foam earplug and two musician earplugs (HPD 1 and 2) resulted in significantly lower word recognition scores compared to the no-protection control ($p < 0.05$). However, the introduction of visual cues significantly altered the results. When participants could see the speaker's lips, the performance gap between the different earplugs and the control condition narrowed significantly. In the audio-visual condition, the effect of the earplug type became only marginally significant ($p = 0.03$), demonstrating that visual information can almost entirely compensate for acoustic degradation.
Subjective Experience and the Lombard Effect
Despite quantitative data showing similar word scores for some musician plugs and foam plugs, participants consistently rated the standard foam earplug as the most difficult to use. This suggests that perceived ease of use is influenced by factors beyond raw intelligibility, such as comfort and perceived sound quality.
Furthermore, the study suggests that earplugs disrupt the "Lombard feedback loop." High attenuation may prevent speakers from accurately gauging background noise, leading them to speak too softly for their listeners because they perceive their own voice as being louder than it actually is relative to the environment.
Discussion and Limitations
The study highlights that lip-reading is an essential component of communication when using hearing protection. While these findings are promising for moderate noise environments (around 80 dB), the authors note that musician earplugs may not provide sufficient protection in high-intensity environments (100+ dB) where high-attenuation foam plugs are mandatory.
Future research will focus on the specific ways certain earplug designs might hinder a speaker's ability to adapt their voice to background noise, particularly focusing on HPD 1, which performed unexpectedly poorly despite a seemingly favorable attenuation profile. This suggests that the interaction between speech production and perception is more complex than simple frequency response curves suggest.
Summary updated