Smolensky (1993, 1995, 1997) proposes that two independent constraints - C₁ and C₂ - can be conjoined to form a conjunctive constraint C₁&C₂ through an operation called local conjunction. The conjoined constraint C₁&C₂ does not replace its member constraints C₁ or C₂. Rather, it forms a separate constraint, which is ranked separately from its member constraints. A conjunctive constraint C₁&C₂ is violated if and only if C₁ and C₂ are both violated. Though local conjunction has enjoyed some success in analyses of phenomena such as synchronic chain shifts (Kirchner 1996), dissimilation and OCP-related phenomena (Alderete 1995; Ito^ and Mester 1998), and stress (Hewitt and Crowhurst 1995; Crowhurst and Hewitt 1997), the number of the documented cases that call for local conjunction remains small. The number of cases involving local conjunction of faithfulness constraints is even smaller, with the only case we know of being Kirchner’s study of chain effects. The scarcity of phenomena that call for local conjunction raises the question of the need for local conjunction. This paper provides an additional argument for local conjunction by demonstrating that Kikuyu consonant mutation calls for local conjunction of faithfulness constraints.

Kikuyu exhibits a productive process of root-initial consonant mutation triggered by the affixation of a prefix made up of a placeless nasal, represented as /N-/. Two types of root-initial consonants participate in this process: a) voiceless plosives and b) voiced fricatives. Under the /N-/ prefixation, root-initial voiceless plosives undergo voicing (1) while voiced fricatives are hardened into stops (2). In addition to voicing and hardening, /N-/ assimilates to the place specification of root-initial consonants. With either voiceless stops or voiced fricatives, the outputs of the /N-/ prefixation are identical; the prefixal nasal is fused with the root-initial consonants, emerging as prenasalized voiced stops.

1. /ko-N-tom-a/ à [koo-ⁿdom-a] ‘to send me’
2. /ko-N-reh-a/ à [koo-ⁿdeh-a] ‘to pay me’
3. /ko-N-T E E c-a/ à [koo-T E E c-a] ‘to stab me’

Unlike affixation to roots with voiceless stops or voiced fricatives, the prefixation of /N-/ to roots with initial voiceless fricatives does not yield a prenasalized voiced stop. What appears instead is the elision of /N-/ with the voiceless fricative remaining intact in (3).

The challenge that confronts an optimal-theoretic (OT) analysis lies in the handling of the /N-/ deletion and non-mutational output in (3). In Kikuyu, root-initial voiceless plosives must undergo voicing in order to emerge as prenasalized voiced stops while voiced fricatives must undergo hardening to surface as prenasalized stops. Whatever constraints are involved, they must allow voicing and hardening to account for the mutational outputs in (1) and (2). If postnasal voicing and hardening are allowed, what is it that prevents a voiceless fricative from undergoing both voicing and hardening to emerge as a prenasalized voiced stop? Therein lies the challenge that the Kikuyu data present for an OT analysis. We show that the account of the non-mutational outputs calls for a conjunctive constraint that combines two faithfulness constraints - Ident (voi) and Ident (cont) into Ident(voi)&Ident(cont) . In order for a voiceless stop to become a voiced stop, it must violate Ident (voi), while a fricative must incur a violation of Ident (cont) to emerge as a stop. But in order for a voiceless fricative to emerge as a voiced stop, this segment must violate Ident (voi) and Ident (cont), something that Ident(voi)&Ident(cont) is designed to block. Ident(voi)&Ident(cont) makes it possible to account for the generalization that while input and output segments may differ in either voice or continuancy, they may not differ in both feature specifications in Kikuyu. Under this view, Ident(voi)&Ident(cont) restricts the degree to which an output segment may deviate from its counterpart in the input. That is, it limits the degree of abstractness, enhancing the learnability of the system.

We demonstrate that an OT analysis relying on local conjunction is superior to a rule-based serial analysis of consonant mutation in three key respects. First, the OT analysis reveals the connection underlying the deletion of /N-/ and the lack of root-initial mutation in (3). Second, the OT analysis captures the functional unity of mutation and the /N-/ deletion. Under the OT analysis, the deletion of /N-/ and the mutation of the root-initial consonants may seem different, but their goals are the same, captured by identical constraints in the analysis. Lastly, but most importantly, the OT view makes sense from the viewpoint of learnability. With local conjunction, the deletion of /N-/ and the lack of mutation in (3) now emerge as a consequence of the system imposing restrictions on the degree of abstractness. In contrast, a rule-based serial analysis will have to rely on opaque rule ordering – that is, counter-feeding – to account for (3), a move that is clearly not enhancing the learnability of the system.