Untitled

112 Bogus clusters, syllabic consonants and vowel syncope. . .the [tl] clusters are not possible branching onsets or coda-onset sequences,the only logical solution left is to admit that they are spurious or bogusclusters and as such separated by an empty nuclear position. This is actuallythe stand taken by, among others, Harris (1994), Gussmann(2002), and the Government Phonology tradition.Admittedly, the Strict CV model has nothing more to offer in the discussionconcerning bogus clusters. Both Government Phonology and theStrict CV approach represent such clusters in a similar fashion. The onlydifference between the two theories lies in the fact that, while in theformer the representation of bogus clusters is the only possible solution,in the latter this is a consequence of the regular syllabification procedures.Recall that in the Strict CV approach there are no codas or branchingconstituents at all, hence any consonant cluster is separated by anucleus. From the above it follows that in the Strict CV model bogusclusters lose their peculiarity. The most serious shortcoming of the representationpostulated by both theories, however, is that we still cannotexplain why bogus clusters have a limited distribution in that theynever occur word-initially. In other words, if bogus clusters are in facttwo onsets separated by an empty governed nucleus, they should be possibleword-initially as well as word-medially. There is nothing in eitherof the two theories which would forbid the existence of the empty nucleusbetween the first two consonants of the word. Note that if we claimedthat the word-internal bogus clusters contract Infrasegmental Government,we would face exactly the same problem, that is, why IG cannothold word-initially. The conclusion we arrive at, therefore, is thatboth GP and the Strict CV model predict word-initial bogus clusters whichare not admitted in English. Table (7) offers some more illustration ofbogus clusters.(7)a[tl]ascu[tl]asski[dn]eychu[tn]eyA[tl]antican[tl]erme[dl]eya[Tl]etebu[tl]ermo[tl]eymau[dl]inbe[dl]amHarris (1994) points out that even though such obstruent plus sonorantsequences are of the rising-sonority profile, they are not allowed toco-occur in a branching onset. In fact, they are not truly adjacent and assuch constitute bogus clusters. To sum up, in GP bogus clusters cannotbe represented as either a branching onset or a coda-onset sequence.They wind up as two distinct onsets separated by the empty governednuclear position (8).