What you've got there with that nominaliser sounds a whole dang lot like an
applicative  morpheme, at least if you ignore the concurrent modification
of the verb's syntactic properties. At the least, I could see that turning
into a full-fledged applicative over time - and a very interesting one at
that, since it would probably retain in its argument marking structure some
vestiges of the fact that it was once a nominaliser.

2018/02/21 21:01 "Logan Kearsley" <[log in to unmask]>:

Halkomelem has a nominalizing prefix/clitic /s-/. with obvious
cognates in many other Salish languages, which can attach at three
different syntactic levels all producing the same surface string, thus
producing the possibility for a three-way ambiguity, as demonstrated
by these examples taken from _Syntactic Nominalization in Halkomelem
Salish_ by James J. Thompson:

k'^w s-til-əm-s
a. 'his song'
b. 'what he sang/will sing'
c. 'that/when/because he sings/will sing'

k'^w s-Ɂiłtəl-s
a. 'his food'
b. 'what he ate/will eat'
c. 'that/when/because he eats/will eat'

I have long been puzzled by just what the practical function of
nominalization is in Salish languages, and this paper finally made it
all click for me. Just acknowledging the fact that it is indeed a
highly ambiguous/polysemous process helps enormously in understanding
what's going on, but there's so much more to it as well.

The (a)-interpretations are examples of so-called lexical
nominalization, with /s-/ attaching directly to the root; the
(b)-interpretations correspond to predicate nominalization, with /s-/
attaching to some intermediate level (which is analyzed in more
complexity in the paper, but which might as well be a verb phrase);
and the (c)-interpretations are clausal or propositional
nominalization, with /s-/ attaching at the level of a CP or IP.
Thompson also draws a loose parallel with English /-ing/, with the
three-way variation in forms like "the crowning of the king" / "the
king's crowning" / "crowning the king" being explained by /-ing/
attaching at one of the three different possible syntactic levels
(mapping V to N, VP to NP, and CP to DP) which then require different
amounts and kinds of surrounding scaffolding to attach complements and
produce a full DP. That, I think, really helps in making something
that initially seems super-exotic feel much more familiar, and
providing a more intuitive understanding of why that ambiguity isn't
really such a big deal in practice.

Now, Valaklwuuxa already has a means of nominalizing clauses, invented
before  figured out this whole Salish nominalization thing, so I don't
need to worry about that. Discarding that function of the nominalizer
means that I will instead be conflating things like "crowning the
king" vs. "that the king is/was crowned" with a single form in
Valaklwuuxa, but that really doesn't seem like a problematic bit of

Additionally, the pragmatic difference between (a)-interpretations and
(b)-interpretations seems fairly inconsequential to me, so if I'm
going to import a Salish-style nominalizer into Valaklwuuxa at all, I
can focus strictly on the level of predicate nominalization, and not
really be missing anything. So, the question now is, given that
Valaklwuuxa has no lexical noun/verb distinction, what is the actual
underlying morphosemantic operation that the predicate-level
nominalizer performs, and is that actually useful in Valaklwuuxa?

(The fact that some Salish languages have been seriously claimed to
lack a noun/verb distinction as well, on which claim Valaklwuuxa was
based, suggests that the fact that *they* find the nominalizer useful
should imply that it will be useful in Valaklwuuxa as well--but in
fact the *distribution* of the nominalizer is one major piece of
evidence used to counter-argue that Salish languages *do* have a
significant lexical noun/verb distinction after all, so it may not be
a-priori useful in Valaklwuuxa, which has been explicitly constructed
to make the claim of non-distinction true, and is not subject to
further discoveries that might invalidate it!)

In the real-world situation of Halkomelem as described by Thompson,
there are some complications, largely revolving around agreement
patterns, in determining exactly what kind of syntactic
transformations are going on, but in any case there seems to be a
fairly consistent pattern in the final argument structure that
results: specifically, that a morphosyntactic theme which is not
realized as a core argument of the root is surfaced in subject
position, with the original subject demoted to a possessor.

That turns out to be really useful, because it provides a way to
*relativize* so-called "inherent obliques" (oblique arguments whose
role is specified by the semantics of the verb, analogous to the
situation with English phrasal verbs that require specific
prepositions to accompany them, as opposed to ad-hoc adverbial
complements). Valaklwuuxa only allows relativization of subjects, so
if we need to relativize an oblique argument, we need to find a way to
first move it into subject position, and that is exactly what the
predicate "nominalizer" does! As such, although there are good reasons
for calling it a "nominalizer" in Salish linguistics, the analogous
construction imported into Valaklwuuxa, where it loses its
distributional characteristics and the relation with its other
functions, would perhaps be more accurately called a participle--or
perhaps just a weird kind of grammatical voice transformation (except
for that fact that it can be applied *on top of* other voice

For example, consider the predicate <wukwes> "to offer something". It
starts off intransitive, and we can say something like

wukwes txe swetqe-la / swetqe txe wukwes
offer DEF man-ART / man DEF offer
"The man makes an offer."

Then we can transitivize it (which requires relativizing the subject
and promoting it to matrix predicate position, since no clause can
have more than one non-pronominal core argument):

swetqe txe wukwest txe skweba-la
man DEF offer-TRANS DEF dog-ART

Then we can add the oblique, with or without transitivizing it:

wukwes txe swetqe-la va ta speka-la
offer DEF man-ART OBL INDEF meat-ART
"The man offers a piece of meat."

swetqe txe wukwest txe skweba-la va ta speka-la
man DEF offer-TRANS DEF dog-ART OBL INDEF meat-ART
"The mean offers a piece of meat to the dog."

Now, we can headless create relative clauses filling noun phrases
referring to either the man or the dog (and in fact we already did for
the man):

txe wukwest txe skweba-la va ta speka-la
"He who offers meat to the dog."

txe wukwestsa txe swetqe-la va ta speka-la
"He to whom the man offers meat."

But to get a noun phrase referring to the *meat*, we have to use the
nominalizer (quoted here because I'm not certain what form it will
actually end up taking in Valaklwuuxa):

ta "s"-wukwests txe skweba-la
"That which he offers to the dog." / "That which is his offering to the

And the semantic subject, the man, can be re-introduced as an oblique
co-indexed with the possessive suffix <-s>:

ta "s"-wukwests txe skweba-la va txe swetqe-la
"That which he, the man, offers to the dog." / "That which is the
man's offering to the dog."

For a simpler example, we can consider the intransitive <k'welenb> "to
slow cook, to barbecue". This has an inherent oblique for what is
barbecued, which can be promoted to a core argument by

xe-k'welenbend va txe skelend-la
1SG=barbecue-1EX OBL DEF fish-ART

xe-k'welenbetka txe skelend-la
1SG=barbecue-TRANS-1/3p DEF fish-ART
"I barbecue the fish."

To get a noun phrase meaning "what I barbecue", it is then possible to
either transitivize and then apply inverse voice:

txe k'welenbetsaka

Or just use the "nominalizer" directly:

txe "s"-k'welenbul
DEF NOM-barbecue-1POSS
"That which I barbecue" / "That which is my barbecuing"

Which is a good bit shorter!

This is a slightly exceptional use of possessive morphology, which
usually refers to possession of the absolutive argument of the host
predicate; i.e., if we didn't have an explicit rule for how
nominalization interacts with possessives, we would expect
<"s"-wukwests> to mean "the offering which he owns", and
<"s"-k'wenbul> to mean "my thing which is barbecued". But we have
exactly the same ambiguity in English with "his offering" and "my
barbecue", so I don't think that will be a problem! And if we add
extracted possessive marking on a matrix predicate, as in

skelends txe "s"-k'welenbul
fish-3POSS DEF NOM-barbecue-1POSS
"What I barbecue is his fish." / "My barbecuing is his fish."

then it becomes entirely clear from context that the lower, 1st-person
possessive cannot be a literal possessive (because we already
established that it is *his* fish!), and so must have the
demoted-agent interpretation.

And there's a bunch of other stuff I figured out by reading about
Halkomelem, but that's probably enough for one really long e-mail....