ABSTRACT
However, the spreading of [voice] from the rightmost obstruent to a non-adjacent obstruent can only proceed if sonorants do not have a specification for [voice] in the underlying representation. This is shown in (3), which shows that the assimilation goes right through an intervening sonorant consonant. The subsequent specification of the feature [voice] in sonorant segments is achieved by a default rule: Russian does not contrast voiced and voiceless sonorants.
