Residue of Subset Principle

[Martha McGinnis]

> Natural solution would be some kind of Hierarchy of Features
>
>               IncludingSpeaker > ParticipantsOnly
>
> The idea of such an Hierarchy is briefly mentioned in literature from
> time
> to time, but is there some kind of consensus forming about this issue?
> Alternatively, is there evidence AGAINST Hierarchy of Features?
>
> Also, if present, would such Hierarchy be universal or language specific?


There is indeed some literature on this.  Harley & Ritter have a 2002
paper in Language proposing a universal feature geometry.  I later
published a proposal (also in Language) to fine-tune their treatment of
person in particular, by which the underlying geometry is universal, but
features are activated by contrasts in the input (Noyer's 1992
dissertation has a different approach involving the activation of
filters).  I worked out a specific analysis of the type you describe in a
chapter of a 2008 Oxford volume called "Phi-Theory", edited by Adger,
Harbour & Béjar.

The logic is straightforward, as you point out: Since the feature
[speaker] depends on [participant] in the geometry, [speaker] is more
specific than [participant], so by Paninian disjunctivity, a vocabulary
item with [speaker] only ranks higher than an item with [participant]
only.

However, systems with an inclusive/exclusive 1 pl distinction also have
the feature [addressee], and in such systems a vocabulary item can be
specified just for [addressee] instead.

Examples: English has we <-> [speaker, group], I <-> [speaker], you <->
[participant], they <-> [group], etc.  (I've argued that 3-way person
systems like English actually lack the [addressee] feature altogether.) 
But Ojibwa prefixal agreement has k- <-> [addressee], n- <->
[participant], and w- <-> elsewhere, which means that inclusives have 2nd
person k-, not 1st person n-.

It's hard to get a sense of "consensus".  Harley & Ritter's proposal has
been quite influential, partly in laying the groundwork for debate (e.g.
my work and that of Cowper and Hall, Nevins, and others), and partly in
giving people something to assume (e.g. Béjar's fascinating dissertation).
 I wouldn't say a full consensus has emerged about the details of the
geometry, but I would hazard a guess that most people working in this area
from a generative perspective do assume some sort of geometry. See also
Bobaljik's relevant recent paper on morphological universals, I believe
written for a special issue of The Linguistic Review.

Cheers,
Martha