Demonstratives: interpretation thereof

Until a few hours ago, I had assumed that the only difference in the interpretation of demonstrative systems had to do with whether systems employed two-way, or three-way, systems: two-way, like English here/there, this/that, etc vs. three-way like Japanese kore/sore/are etc (or non-standard varieties of English here/there/yon, this/that/thon). However, it turns out that there are interpretive differences between non-standard English and Japanese: whereas selection of this/that/thon is determined only with respect to the object relative to the speaker in NSE, the presence and position of the addressee is relevant as well. For example, in a situation where an unrecognized object is placed 1 metre from an addressee and 3 metres from the speaker, the Japanese speaker may ask:
Sore wa nan desu ka? "What is that?"
However, if the addressee were absent, the Japanese speaker might then ask herself:
Are wa nan desu ka? "What is that (distal)?"
No such contrast arises in NSE, or I suspect Spanish (este/eso/aquello): the more distal form is only used where the distance increases.
This raises the question: how do things work with này/đó/kia?

Demonstratives and Linguistic Interfaces

In writing up the section on use of kinship terms as pronouns in 3rd person contexts, I came across the following phenomenon, which, it seems to me, may have interesting implications for Minimalist conceptions of interfaces.

"Nguyễn Đ. H. (1997:43) mentions a third strategy for speakers of the Saigon dialect, namely, where the demonstrative element ấy is deleted and marked instead by a tone change on the kinship label itself. Nguyễn Đ. H. provides the following examples (observing also that this strategy cannot apply to words that bear inherent high tones such as chú or bác):

(5) a. bà ấy > bả ('she')
b. ông ấy > ổng ('hé)
c. cô ấy > cổ ('she')
d. anh ấy > ảnh ('he')
e. chị ấy > chỉ ('she')

f. thằng ấy > thẳng ('that guy, he')
g. thằng cha ấy > thằng chả ('that bloody guy')

Notice especially the contrast between the last two examples, which suggest that this operation is quite productive: tone-shift applies to the right-edge of the word. This would seem to indicate that the process is not purely lexicalized. On the other hand, it is lexically constrained, since elements bearing inherent high tones cannot be affected. It would also appear to have implications for Minimalist assumptions about the ways in which semantics and phonology can interact outside of narrow syntax (given that it is implausible to suppose that these phonetic properties enter into syntactic computations).

Briefly, then, this appears to be a morphophonological operation that is sensitive to phonological constraints, but which expresses a semantic property. Since the relevant phonetic property (the high tone which is realised as/changed to a low-rising tone when shifted to the left-adjacent segment) would seem to be syntactically inert, the question is how this phonetic change is able to affect interpretation unless semantics and phonology are able to interact independently of the syntactic computation?

Does anyone else think this is really a theoretical problem? or have I missed or misunderstood something obvious?