Age estimation in foreign-accented speech

Abstract

Previous research has shown that listeners are generally fairly accurate in estimating people’s age from their speech (e.g., Moyse, 2014). However, most age estimation studies have been based on American English, and very few have explored cross-language effects. While Braun and Cerrato (1999) found no effect of language when presenting listeners with German and Italian voices, Nagao and Kewley-Port (2005) conducted an age estimation study with stimuli presented in English and Japanese and found that listeners were more accurate at estimating age in the familiar language. The current study sets out to test whether such a sociolinguistic effect may extend to foreign-accented speech.
In our perception experiment 28 English first language participants listened to randomly presented audio-stimuli and were asked to estimate the speakers’ age. The audio stimuli were 40 clips of 20 English English speakers and 20 Japanese first language speakers reading the ‘Please call Stella’ passage in English, retrieved from the Speech Accent Archive (Weinberger, 2015). The age range of speakers was from 18 to 70s. With the audio stimuli available, the best attempt was made to get a balanced age distribution and also age-match the speakers in the two language groups. The participants’ age estimate and reaction times were recorded.
Our statistical analysis shows that Japanese first language speakers were estimated to be younger than their English age-matched counterparts with no age by first language interaction. Additionally, participants were significantly slower in reacting to clips recorded by Japanese speakers. These results suggest that listeners may not only differ in how they estimate the age of speakers in different languages but also in how they estimate the age of speakers with different accents. These findings have theoretical implications as they highlight that even such a seemingly universal phenomenon as age may be expressed or perceived differently by speakers of different languages. The practical implications include our need for awareness of such differences when age estimation is used for decision-making puposes in real life (e.g., forensics).