Dear WG list,

Tallerman 2009 (Phrase structure vs. dependency: The analysis of Welsh syntactic soft mutation. J. Linguistics 45 (2009), 167–201. <http://journals.cambridge.org/action/displayIssue?jid=LIN&volumeId=45&issueId=01&iid=3533592#>) will be of much interest here for several reasons. Besides being an unusually conspicuous consideration of WG by a non-WGer, it also offers arguments against WG, including arguments in favour of phrases and in favour of wh-traces. I find the arguments in favour of phrases to be unpersuasive, but the evidence for wh-traces strong. 

Tallerman contrasts two formulations of a trigger for soft mutation (SM), an HPSG one (by her and Bob Borsley) and a WG one (by her and Dick). The HPSG one says that when you have two adjacent sister XPs then the second one bears SM if it is a complement. The final WG version is:

"X bears soft mutation if :
(a) X is a FIRST of a valent of a preceding overt head H, and
(b) X is SEPARATED from H by Z."

"First of W" is a horribly clunky way of getting at the notion "first word in the phrase whose root is W". To prove the necessity of the need for this formulation, instead of (a) being just "X is a valent of a preceding overt head H", Tallerman seeks constructions in which the valent has a predependent and the predependent bears SM. The examples she considers clinching are what look to be word-for-word analogues of "am I just as angry as you", where "just" bears SM, and "put Elen the harp two foot away", where "two" bears SM. In the English versions, I would certainly analyse "just" and "two" as the roots of "just as angry as you" and "two foot away", and I gather that Welsh is even more strongly head-initial than English, so I'm not convinced that "First of" was needed here. However, I would not presume to dispute the prevailing belief that there really do exist inflectional phenomena in some languages that affect the left or right edges of phrases, so the issue of whether 
Welsh SM is such a case is a bit moot, if the general case is conceded. 

Returning to the clunkiness of First, for a moment, it would be preferable, I think, to recognize morpho-phonological phrases at the level of Forms: the phrasal form of a head is made by concatenating the lexical form of the head with the phrasal forms of its dependents. This shd be relatively uncontroversial, given that WG has never denied that syntagms larger than the phonological word do exist at the level of Form. So the WG might be reformulated as:

"X's form bears soft mutation if :
(a) X is a valent of a preceding overt head H, and
(b) X is separated from H by another dependent of Z."


Tallerman then discusses a number of cases where the HPSG analysis is superior to the WG:

1. where X but not XP separates head and valent
In "did Megan not sleep", "sleep" does not get SM. For the HPSG analysis, it is claimed that this is because "not", though sister of "sleep", is an X rather than a full XP. She says "An obvious question is whether ddim in (31) might form a constituent with
the following non-finite verb cysgu ‘sleep ’ [...] According to
Borsley & Jones (2005: section 5.2.1), post-subject ddim does not behave like
a pre-modifier of the following verb and does not form a constituent with it." Certainly in English I would see "sleep" as complement of "not", so I'd like to find out what B & J's evidence is for them not forming a constituent.

2. ordinary coordination
When the valent is a coordination, SM turns up only on the first conjunct (i.e. only at the start of the entire coordination. WG predicts that it should turn up on each conjunct. However, this follows not from WG's basis in DG but rather from Dick's insistence on analysing coordination using constituency. In a normal DG analysis of coordination, the correct prediction would be made. 

3. complex coordination and ellipsis
In "What instrument does she play? -- Harp", "harp" does not get SM. Hudson 2007 speculatively suggests that "Harp" is "She plays harp", with "She plays" elliptical. If that analysis is correct, the Welsh order being "Plays she harp", then elliptical material does not trigger SM.
In ‘He went to see Llandaff church in the morning and in the afternoon, Cardiff castle.’ As you would expect, "Cardiff" gets SM. Tallerman assumes that the WG analysis of this would have the second conjunct as "He went to see in the afternoon Cardiff castle", with "he went to see" elliptical. If that is indeed the analysis, then here the elliptical material is involved in triggering SM, which means WG is inconsistent about whether ellipsis does or doesn't trigger SM.
However, WG has never invoked ellipsis for the analysis of complex coordination. Indeed, for English Dick's article on coordination published in the late 80s explicitly denied this, pointing out the position of "both" in 
"He went to see both LLandaff church in the morning and in the afternoon Cardiff castle"
rather than 
"*Both he went to see both LLandaff church in the morning and in the afternoon Cardiff castle"
this latter being what you'd expect if there was ellipsis in the second conjunct.
It's true that WG has come to accept the existence of ellipsis, but the arguments for the existence of ellipsis only apply to the ellipsis of whole phrases, not to just nonphrasal parts of phrases. Hudson 2007 does suggest ellipsis can apply to nonphrasal chunks, as in gapping, but this is a remark made in passing, not a full analysis. To summarize, the sentence fragment "Harp" involves ellipsis only according to a passing suggestion in Hudson 2007, and no WG analysis sees ellipsis in the coordination example. So there is no inconsistency.

4. wh trace
In "Who bought [t] harp?", "harp" gets SM. This really is a massive problem for WG, as far as I can see. In my paper in the Sugayama collection, I did tentatively suggest traces as a possibility for WG, but I subsequently came to think them undesirable. But anyway, my earlier suggestion logically applied to traces of all sorts of movement, yet raising and control don't trigger SM: in "wanted Aled play the harp", "Aled" triggers SM on "play", but in "is Aled wanting [e] play the harp", "play" doesn't get SM.

In summary:
* The existence of phrasal inflection warrants the recognition of phrasal forms.
* WG is misguided in rejecting DG for coordination.
* It seems that wh-traces must be recognized, at least for Welsh. (This one is big news not just for WG, of course.)


--And.