Masked speech perception: the effect of age and language background
Most communication in everyday life takes place in less than ideal acoustic conditions. The presence of background noise or competing voices affects some populations more adversely than others. For example, Goossens et al. (2017, Hear. Res. 344:109) found that, although normally hearing older adults’ self-reported hearing abilities do not differ from younger and middle aged adults’, their speech reception thresholds (SRT) are poorer. The researchers found age (rather than hearing impairment) to explain a significant part of this decline in performance. Here, we expand their study to younger and non-native listeners.
To measure self-assessed speech intelligibility in everyday situations, the Speech, Spatial and Qualities of Hearing (SSQ; Noble et al., (2013), Int. J. Audiol. 52(6):409) questionnaire was used. To measure SRTs, we ran an adaptive coordinate response measure task [CRM; adapted from Bolia et al. (2000), JASA 107(2):1065] for simple sentences (e.g. “show the dog where the red six is”) presented in (a) a speech-shaped noise masker (SSN), and (b) a 3-talker-babble masker (BABB). Participants were (i) native English speakers (8-85 years, N=114, 60 female) divided into six age groups [young children (CH-Y), older children (CH-O), young adults (YA), middle aged adults (MA), younger older adults (OA-Y) and older adults (OA-O)], and (ii) non-native speakers of English (N=19, 18 female, mean age 24.89). All had normal audiometric thresholds up to at least 4 kHz or self-reported normal hearing (for L2 speakers) and anyone over 65 was screened for cognitive impairment.
The results showed similar self-assessed speech intelligibility scores in the SSQ for all L1 groups. On the contrary, L2 listeners significantly differed from the L1 groups, reporting lower abilities in everyday speech understanding (p<.05). In the CRM, the effect of masker type was significant (p<.001) with better SRTs for SSN (M=-5.32 dB) than for BABB (M=-1.68 dB). The interaction between noise type and group was also significant (p<0.042): for BABB, but not for SSN, better thresholds were obtained for YAs compared to all other age groups apart from MAs (p<.05). Conversely, L2 listeners differed from all L1 groups in SSN (all p<.001), and from YAs and MAs in BABB (p<.010).
In line with Goossens et al. (2017), we found that when background noise is more cognitively demanding, there is a larger decline in speech perception in OAs. Additionally, we also found children (8-16) to show reduced speech perception abilities in this masker type. Overall, non-native speakers' performance was most affected by both types of masker.