Ground-truth validation of the "earbuds method" for measuring acoustic nasalance.

Abstract

Among the variety of methodological tools available for quantifying the degree of nasalization, nasometry has remained a popular choice for both academic and clinical research on speech due to its relatively low cost, portability, ease of use, and interpretability. In its simplest form, nasometry involves two microphones that independently capture the acoustic radiation from the mouth and nose. Recently, a method was proposed by Stewart and Kohlberger [(2017). Lang. Doc. Conserv. 11, 49-80] of using earbuds to achieve a similar result. Although the method has been widely used since, most notably in the context of linguistic fieldwork, it has yet to be tested against ground-truth measurements of acoustic nasalance from a commercially established device; the current study attempts to provide this ground-truth comparison. A wide variety of characteristics of the relationship between nasalance measurements created using acoustic data co-registered from a commercial nasometer, silicone earbuds, and flat earbuds are thoroughly examined. The results suggest that the "earbuds method" may provide reliable estimates of nasality when assessed at a global level (i.e., overall averages, relatively high/low nasalance scores coinciding with the presence/absence of nasalization), but that the accuracy of these estimates may decline at more fine-grained levels in different ways that are outlined in the discussion.