Breton mutation: linear or autosegmental?

Much of the recent work in phonology has focused on the development and application of autosegmental and metrical theories to suprasegmental phenomena, e.g., stress, tone, syllable structure. The new notions of phonetic structure allowed by these theories, as well as the new formal devices to account for them, have provided insight in an area where standard generative theory provided little. Furthermore, a new line of research opened up when it was discovered that the tools of suprasegmental research also provide insight into certain segmental processes. For example, Goldsmith (1979) showed that the change of s --> h in Buenos Aires Spanish is best analyzed as the loss of supralaryngeal features, but with retention of the laryngeal ones. Within an autosegmental framework this is realized by assuming that supralaryngeal features are on a separate tier from the laryngeal one. A further extension to segmental processes was made by Halle and Vergnaud (1981) who argued that most harmony processes are better described in either autosegmental or metrical terms, depending on the type of harmony. (By harmony processes they intend the familiar ones such as vowel harmony as well as such assimilatory processes as voicing agreement and palatal agreement.