Vowels are often nasalized in [VN] sequences. Cohn (1993a) argued that
contextually nasalized vowels in English are unspecified for nasality,
and are phonetically nasalized through coarticulation. In contrast,
Solé (1995) argued that contextual nasalization is so extensive in
English that nasalized vowels must be phonologically [+NASAL]. This
disagreement reflects a broader problem: how do we determine the
phonological specification of segments with phonetically intermediate
values for nasality (or other properties)?

We contribute to this debate with data from A'ingae, a language spoken
in the Ecuadorian and Colombian Amazon. A'ingae has a /V Ṽ/ contrast,
along with left-to-right, phonological nasal spreading. We present
aerodynamic data showing that A'ingae also has extensive, but partial
nasalization in [VN, VⁿD]. This partial nasalization is phonetically
distinct from contrastive /Ṽ/, and from nasal vowels derived by
spreading. We argue that these results are inconsistent with treating
partially nasalized vowels as [+NASAL], or as specified for 'weak'
nasalization. Instead, partial nasalization in A'ingae reflects
underspecification.