First, it is natural to expect vowels to be phonetically more similar in singing in
comparison with speech. That was generally shown for European academic singing (e.g., Sundberg, 1987: 117) as well as for traditional singing (Ross, 1992). This rule,however, is valid only partly for our case. It is obviously correct if applied to Aukštaičiai samples (Fig. 4; top): the Aukštaičiai vowel system covers considerably less space in singing compared to speech. For Dzūkai samples, however, this is not the case (Fig. 4; bottom). We may make the conclusion that the pronounced difference between spoken vowels remains large in singing (Dzūkai). Relatively (slightly) ‘covered’ spoken vowels tend to be even more covered in singing (Aukštaičiai). Thus, singing is likely to be a kind of ‘exaggerated’ speech in terms of covering. In simple words, the Dzūkai ‘sing as they speak’ (uncovered), and the Aukštaičiai ‘sing as they exaggeratedly speak’ (covered).
http://www.musicstudies.org/JIMS2008/articles/Ambrazevicius_...