The experiments reported here provided some data on the accuracy with which listeners rate the age of speakers, and some knowledge about the relative contributions of underlying factors such as F0, the formant frequencies, and the 'verbal maturation' of children. The speech material consisted of excerpts from interviews with boys and girls in the age groups 5, 7, 9, and 11 years. Utterances which contained verbal cues to age were eliminated after presenting written versions to a panel of judges. The retained utterances were LPC-analyzed and resynthesized with modifications, to obtain (1) natural speech, (2) whispered speech, (3) speech of 9 year olds with F0, formants, and speech rate modified as at an age of 5, 7, 9, and 11 years, and (4) whispered versions of (3). The listeners agreed in their age ratings with an SD that increased from 1.3 at 5 to 1.8 at 11 years. The differences in SD between the four types of speech were small, but there was a bias toward 8 to 9 years in the modified versions. Multidimensional regression analysis showed the perceptual weight of the factors to decrease slightly in the order (1) verbal maturation, (2) formants (and rate), (3) F0. As for the distinction boy/girl, there was a slight improvement in correctness with age and a general bias to classify the younger children as girls and the older ones as boys.
1994. vol. 3: 1171-1174 p.