The productivity of ‘unnatural’ labial palatalization in Xhosa

Xhosa (Bantu, South Africa) has a pattern of labial palatalization. When the passive suffix /-w-/ is added to a stem ending in a labial, the labial becomes palatal (uku-lum-a → uku-luɲ-w-a). Two main types of analysis have been proposed for this alternation: (i) the ‘phonological analysis’, which states that the alternation is part of speakers' synchronic phonological grammar, and (ii) the ‘lexical analysis’, which states that speakers learn the palatalized passive forms as part of their lexical knowledge. To distinguish between the two hypotheses, we conducted a wug test in which speakers of Xhosa were shown nonce verbs and asked to provide their corresponding passive form. Since the phonological analysis requires there to be a synchronic phonological rule, speakers should passivize nonce forms in the conditioning environment. Under the lexical analysis, however, speakers are predicted not to palatalize nonce forms, since the real-word palatalized passives they produce are simply stored in the lexicon. Our results show a great deal of inter-speaker variation. While some speakers produced palatalized nonce forms nearly 100% of the time, other speakers failed to produce any palatalized nonce forms at all. We argue that labial palatalization may be analyzed in different ways by different speakers