A long-standing problem in the morphophonology of Chaha, an Ethiopian Semitic language, has been the simultaneous occurrence of labialization (e.g. g ® gW) and palatalization (e.g. z ® Z) as a manifestation of a single suffix (e.g. Leslau 1967, McCarthy 1983, Lieber 1988). For example, a comparison of the two words in (1) shows that only in the impersonal, (1b), the stem-initial consonant is labialized and the stem-final consonant is palatalized.

(1) a. g´n´z-o-m   ‘They aged.’       b. gW´n´Z-m   ‘One aged.’

From a semantic point of view labialization and palatalization in (1b) have the same status as the medial suffix -o of (1a) as both are manifestations of the subject agreement. But the two affixes are asymmetric from a phonological point of view in that labialization and palatalization in (1b) are processes affecting stem-internal segments while the -o in (1a) is a suffix. In this paper we show that this phonological asymmetry is only surface true and should be derived from a symmetric UR. In this view, the UR of the impersonal (1b) is /g´n´z-u-m/ and it has exactly the same paradigm as the personal (1a) but the /u/ decomposes and triggers labialization and palatalization of the preceding consonants. We propose that /u/ includes [labial] and [palatal] elements and that Chaha has a constraint *[labial, palatal], which bars associating [labial] and [palatal] to a single slot. This constraint triggers phonemic decomposition by which [labial] and [palatal] become autonomous so that [labial] yields labialization whereas [palatal] yields palatalization. We also show that given the appropriate context, labialization in Chaha entails palatalization and that simultaneity of labialization and palatalization has nothing to do with the morphological features of the impersonal, i.e. the implicational relation always holds in Chaha.

We extend this analysis to explain the rise of linked sounds in English where we have simultaneously occurring y and u, as in unity [yu: n´tI]. Halle & Mohanan (1985) account for this by invoking a rule of y-Insertion before ˆ (they assume that the u here derives from /ˆ/). However, the y-Insertion analysis does not offer a logical connection between the trigger and the output. But such a connection is mandatory, especially if we want to offer a unified account for the related process of coronal obstruent palatalization before u, e.g. fact vs. factual [fQktSu´l] and before its nonhigh reflex [], e.g. create vs. creature [kri: tS r]. Our proposal is that /u/ includes [labial] and [palatal] elements also in English but the /u/ decomposes and gives rise to the simultaneously occurring sounds and processes.

Our proposal offers an elegant explanation for sound and process simultaneity by using a minimal number of independently needed constraints. In addition, while previous approaches require rule ordering to account for the facts discussed above (e.g. y-Insertion in English must apply prior to palatalization since it is the inserted y which triggers palatalization) the present analysis does not require such ordering since it is /u/ -- not an inserted y -- which triggers palatalization. Besides, we no more have to make the assumption that simultaneous processes/sounds necessarily originate from different sources.