Perception of Indexical Features in Children’s Speech 347First, we can see that all listener groups gave fewer “girl” responses to
stimuli with pre-aspirated [ހp ހt ހk] than they did with plain variants. This
effect was signi¿ cant for both the non-local UK group (chi sq = 6.325, df
= 1, p < .025) and also the Americans (chi sq = 3.86, df = 1, p < .05). Note
that this ¿ nding might appear to go against our predictions based on patterns
observed in speech production. However, the Tyneside listeners were the only
group which did not give a signi¿ cantly lower number of “girl” responses to
pre-aspirated tokens compared with plain tokens. Furthermore, for the pre-
aspirated stimuli the local listeners gave signi¿ cantly more “girl” responses
than both the control groups (compared with non-local UK listeners: chi sq =
4.978, df = 1, p < .05; compared with Americans: chi sq = 4.621, df = 1, p <
.05). There was no difference between the two non-local groups.
- Discussion
As explained in the introduction, this experiment had two main aims: to explore
the range of cues used by listeners in judging a speaker’s sex or gender, and
Figure 14.8 Percentage of “girl” responses to pre-pausal tokens by listener group
and variant.