Prosodic clues in language recognition: how much information do listeners need to identify Maori and English?

As part of an investigation into the mutual influence of Maori and New Zealand English phonologies into the 21st century, we report on three experiments designed to test whether the languages can be identified from purely prosodic cues. In the first experiment, listeners with varying degrees of exposure to the Maori language heard 15-second stretches of low-pass filtered (400Hz) natural speech from three speaker groups (Historical Elders, Present-day Elders and Young speakers). Listeners were able to identify English and Maori significantly above chance level. In the second experiment, listeners were able to identify the three groups of speakers when the stimuli were filtered to parallel a land-line telephone and to a lesser extent when the stimuli were low-pass filtered. Those listeners with greater exposure to Maori performed better. The third experiment used synthetic stimuli that retained only the pitch and/or loudness information from the original speech. Listeners were able to identify the two languages above chance level when pitch information was included.