Comparing vowel category response surfaces over age-varying maximal vowel spaces within and across language communities

We investigate vowel category perception within and across languages by proffering a statistical methodology for creating vowel category response surfaces over maximal vowel spaces based on the responses of subjects from five different language communities to vowel stimuli generated by an age-varying articulatory synthesizer. The methodology is based on an additive modeling approach to surface regression within the general smoothing spline approach to statistical modeling. We also put forward a simple method for the comparison of surfaces and demonstrate its basic utility by comparing response surfaces derived from Greek and Japanese subjects. We discuss the results of the comparison with attention to the potential of the approach to reveal meaningful differences between and within the vowel systems of different language communities.