This paper offers an analysis of the much-studied phenomenon of "soft mutation", or lenition of initial consonants, in Welsh. The purpose of the paper is to show that the relevant facts can be parsimoniously accounted for without the use of devices like syntactic movement or empty categories with phrasal status. Instead, the phenomenon can be captured in a purely monstratal, derivationally monotonic theory of grammar, using only the resources of lexical specification of local phonological features, and rules of little more expressive power than context-free grammar.
Zwicky 1984:389 points out that the very diverse contexts or "triggers" for this mutation (which include certain determiners, prepositions, verbs, particles, etc.) seem to observe the following very natural constraint:
(1) The trigger determining a rule feature for a morphophonemic rule must be adjacent to the affected word and c-command it.
Among such triggers, the effect of the tensed verb can be illustrated by the following minimal pairs, adapted from Ball andMiller 1992 and Tallerman 1990, 2006, 2009, in which the mutation from cath ("cat") to gath is in all cases the relevant alternation:
(2) a. Gwelodd y ddynes gath.
saw the woman cat
"The woman saw a cat"
b. Gwelodd cath y ddynes.
saw cat the woman
"A cat saw the woman."
(3) a. Pwy welodd gath?
who saw cat
"Who saw a cat?"
b. Pwy welodd cath?
who saw cat
"Whom did a cat see?"
(4) a. Gwelodd gath.
saw cat
"He/she saw a cat"
b. Gwelodd cath.
saw cat
"A cat saw."
(5) a. Mae cath ddu yn yr ardd.
is cat black in the garden
"There is a black cat in the garden."
b. Mae yn yr ardd gath ddu.
is in the garden cat black
"There is in the garden a black cat."
Zwicky (1984) and Roberts (2005) follow traditional descriptive accounts in interpreting these data in terms of a nominative/accusative case-marking function for softmutation in these specific contextsHowever, Tallerman points out that such a role appears inconsistent with the alternation seen in (5a,b), where the black cat in each example seems to fulfill the same thematic role. Moreover, as noted earlier, soft mutation does not apply to objects of non-finite verbs, except when, like (5b) they too are separated from the said verb by some further argument in a nonstandard position:
(6) a. Roedd y ddynes yn gweld cath ddu ar deledu
was the woman ing- see cat black on television
"The woman was seeing a black cat on television"
b. Roedd y ddynes yn gweld ar deledu gath ddu
was the woman ing- see on television cat black
"The woman was seeing on television a black cat"
Tallerman (2006, 2009) proposes that soft mutation is triggered by a linearly preceding phrasal category, such as NP or PP, and not by a mere verb. The trigger constraint must accordingly be changed (because under Tulliver’s assumption of binary branching the putative trigger is non-commanding).
This assumption accounts for all of the facts (2) to (6). However, it does so only at the cost of assuming that, in the subject question (3a), a wh-trace with NP phrasal status in VSO subject position intervenes and triggers soft mutation in the object. (In (3b), no object wh-trace in VSO position intervene between verb and subject, so mutation is not triggered on the latter.) Similarly, a phonologically empty pronoun subject (pro) with the same phrasal status in the same position has the same effect in (4a), but is absent in (4b). Tallerman suggests that the phenomenon of soft mutation provides evidence for the linguistic and psychological reality of such empty categories.
The present paper shows that these data can be accounted for under the following simpler assumptions: that soft mutation (rather than non-mutation) is the default for arguments other than first arguments; that the exceptions considered here correspond to the notion "first argument of the verb", rather than case as such; and that the morphonemic rule concerned is subject to Zwicky's principle (1). The analysis is developed using combinatory categorial grammar (CCG), which is attractive for the purpose because it is strongly lexicalized, supports "surface compositional" semantics, and directly expresses the notion "first argument". No further assumptions other than those standard in CCG are required, and in fact the above assumptions are compatible with a wide range of grammatical theories.
However, under those same assumptions, excluding phonologically-active and syntactically-realized empty categories as a residue of relativization and subjectdrop is a forced move. Hence, the phenomenon of Soft Mutation in Welsh may be taken as evidence against the necessity of such constructs in syntactic theory, rather than evidence for it.