to DeLancey's
Linguistics Home Page
comments welcome: delancey@darkwing.uoregon.edu
SSILA Conference, San Diego DRAFT: please do not
January 6, 1996 cite without author's
permission
Argument Structure of Klamath Bipartite Stems
Scott DeLancey
University of Oregon
This paper is intended as an empirical contribution to the theory
of argument structure (by whatever name). Any theory of argument
structure must be empirically grounded in linguistic fact, and my
primary purpose here is to present some facts about Klamath.
Unfortunately, though there are one or two good speakers of
Klamath still living, it's not really possible at this point to
collect paradigmatic data elucidating subtle points of
grammaticality. Still, what it remains possible to do given the
materials provided for us by Gatschet, Stern, and Barker is more
than I can present in 15 minutes. I want to talk here
specifically about the most intriguing aspect of verbal structure
in Klamath. This is a feature found in a number of languages in
the western U.S. (I will be talking about this areal pattern at
BLS this winter), which Len Talmy has written about in Atsugewi,
and which Bill Jacobsen, writing about Washo, has christened the
"bipartite stem" construction.
Subject and object in Klamath
I can't spare time to tell you much about subject and object in
Klamath (for a good beginning see Rude's paper in the HaasFest
volume), except to say that there definitely are distinctly
syntactic subject and object relations, and that they are often
not clearly marked. Constituent order in Klamath is quite free,
and cannot distinguish subject from object. There is no
indexation of arguments in the verb. Case marking will sometimes
distinguish subject and object, but not always. There can easily
be two nouns in a clause not marked for case, with no overt cue
to which is which. In such cases we have only the semantics of
the clause to rely on. This is, of course, unsatisfactory, and I
have tried as much as possible in this paper to illustrate my
claims with examples where case marking is explicit;
unfortunately, this is not always possible with the materials at
hand.
Bipartite stems
The most striking fact about Klamath verbs is that most verb
stems are bimorphemic. For example:
1) /ksawal-/ 'living object is, goes on top'
/ksew-/ 'living object is, goes in water or a
flat place'
/lawal-/ 'round object is, goes on top'
/lew-/ 'round object is, goes in water or a
flat place'
/howwal-/ 'run, jump on top'
/howw-/ 'run, jump into water or a flat place'
2) /nqew'-/ 'break in two with a round instrument'
/wqew'-/ 'break in two with a long instrument'
/mp'ak'-/ 'smash to pieces with a round
instrument'
/wp'ak'-/ 'smash to pieces with a long instrument'
Barker, in his Grammar and Dictionary, presents a position class
analysis of bipartite stems, recognizing two classes of
components: classificatory stems and locative-directive
suffixes, of which the initial and final elements of our examples
are instances. For reasons which should be obvious soon I adopt
another terminological suggestion of Jacobsen's, and refer to
Barker's "classificatory stems" as "lexical prefixes". Barker's
analysis suggests a simple formula for bipartite stems:
3) LP + LDS
In fact, though--as suggested by my arrangement of examples in
(1-2)--there are (at least? XX) three different categories of
bipartite stem, corresponding to three more-or-less distinct
types of LP (one of which is in no sense "classificatory").
These can be roughly distinguished by their combinatorial
properties, though there is some overlap and fuzziness XXX
there. They can, however, with a handful of (lexical XX??)
exceptions, be quite neatly distinguished by differences in
argument structure.
A. Change-of-state verbs
The first combinatorial criterion for distinguishing among
bipartite stems has to do with the second element of the stem.
Certain LP's combine primarily or only with LDS's (ex. set 2),
while another set combine primarily or only with what I will call
change-of-state stems (ex. set 1). These are distinguished from
LDS's in part semantically, but XXX that they can occur
unprefixed, while LDS's are bound morphemes which must have a
preceding LP or (rarely) other verb stem.
As simple, monomorphemic stems, the COS stems are
intransitive, and describe a change of state in their single
argument:
4) sciqtGis-ti: boq'e:wis ?a gatt'-atk
bridle-PART leather DEC break-STAT
'The leather of the bridle is broken.'
With an initial LP, drawn from a specific subset of the LP's
which corresponds to what Americanist linguists since Sapir have
called "instrumental prefixes", they are transitive:
5) he: cik ?ins q'ay s?ab-i:-wapk, coy hon
if CONT 2nd/1st NEG tell-BEN-FUT then that
mi-s ni ye=qew'i-wapk.
2nd-OBJ I w.foot=break.in.two-FUT
'If you won't tell me, then I'll stomp you in two.'
[This is Old Grizzly speaking to Awl, threatening to
step on him and break him in half.]
The subject argument of such a transitive stem is
unproblematically an Agent; the object argument, as well as the
subject of the intransitive construction, I will call a Theme,
which I think is the correct role for the undergoer of a change
of state.
Note that the relation between the argument structure of the
intransitive and transitive corresponds precisely to that of
labile or "ergative" verbs of the break class in English and many
other languages. The analysis here is in some ways less
problematic than in a language like English, where the same verb
stem occurs both intransitively and transitively. In Klamath we
can argue that the change-of-state stem, the second element of
these bipartite stems, brings with it a Theme argument slot, and
that the LP then adds an Agent slot. How this is done is not
completely clear. At first blush it seems that these LP's
classify an Instrument. While it would not work in traditional
Case Grammar, a case can be made that Instruments in fact occur
only together with Agents, in which case an indirect mechanism
could be proposed by which the lexicalization in the verb stem of
an Instrument necessarily implies an Agent. At best this would
be cumbersome, however, and, for various reasons, I am not
comfortable with the idea of Instrument arguments being lexically
required by verbs. There is reason to think that these prefixes
lexicalize a type of action rather than a type of object. For
example, Barker glosses the prefix {kt} as 'hit with the fist,
kick', but among the stems formed with this element we find
several which do not contain that semantic element, for example:
6) kten- 'throw (a spear, dart)'
kteqwe:L- 'slide downhill (as a log, avalanche)'
ktoc'i:p'- 'slide out of a tubular object (as lead
out of a pencil)'
All of these, as well as the more common 'hit, kick' senses, can
be subsumed under a broader sense of 'force applied by a long
object (such as an arm or leg) in the axial dimension'.
If this really works (this issue isn't closed, but I'm going
to leave it here) then an intransitive form like /qew'-/
lexicalizes a process leading to a change of state; a transitive
form like /wqew'-/ then adds to this lexical reference to an
action on the part of an Agent, thus describing a transitive
event. This rather directly expresses the logic of Chafe's
(1970) categorization of transitive verbs like break as Process +
Action.
B. CLS + LDS stems
The change-of-state system is a minor variation on a pattern
familiar from other languages, and quite likely universal. There
is another, larger class of bimorphemic stems which are more
exotic. These consist of a LP and a LDS. Look at the following
series:
7)
on top in water underneath
living object: ksawal- ksew- ksodi:l-
round object: lawal- lew- lodi:l-
long object: ?awal- ?ew- ?odi:l-
Stems of this type consist of a classifying initial element
referring to a category of object and a second element describing
motion, location or path on the part of that object. These stems
are indifferently intransitive or transitive, according to
context; thus most of Barker's glosses for full verb forms have
the proviso "intransitive also", e.g.
8) /ksew-/ 'puts a living obj. into water, flat place.
(intr. also)'
Some examples from text illustrate the labile transitivity of
Classifying LP + LDS stems:
9) coy honk ga?as ks=aa:q-damn-a
now NARR thus living.obj=on.lap-over.and.over-DEC
'Now he was lying around in her lap this way.' (1963a:70 )
10) ?e:, ci cik honk bonwa, do:m, ks=ew-ank
Eh PRT PRT NARR drink much living.obj=in.water-PART
?ambo-tdat bi honk
water-LOC he DEM
"?e...," he said, drank, very much, himself sitting in the
water. (1963a:86-7)
11) cakl'a-dat ks=iGog-a sa ?aysis-as
storage.basket-LOC living.obj.=put.into-INDIC 3pl Aisis- OBJ
'They put Aisis into a large basket.'
12) hon-c-ipg-abli hokt honkt ks=oyamn-atk
fly-along-hither-back DEM DEM.OBJ living.obj=carrying-STAT
'He flew back carrying him' (1963a:90-1)
Thus while the COS stems do not pose the same analytic problem as
their English equivalents, these bipartite stems--which do not
have any direct counterpart in English--do. But they are
transparently easy to describe in terms of a simple set of case
roles--the two parts of the verb lexicalize, respectively, a
Theme and a Loc.
When the clause has a distinct NP corresponding to the path
or location indicated by the LDS, this is marked as locative:
13) coy honk na:nok Ge:s cewam'c- am
then DEM.OBJ all ipos Old Antelope-GEN
?i=Gog- a mna-tant y'agi- dat
pl.=put.in.container-INDIC 3sPOSS-OBL.LOC burden.basket-LOC
'Then [she] put all Antelope's ipos into her basket.'
14) s?as?abam'c qtan- a ks=elwy-ank loloqs-dat
Old.Grizzly sleep-IND living.obj.=by.fire-HAVING fire-LOC
'Old Grizzly slept, lying by the fire.'
C. Compound stems
There is a third, less cohesive, category of bipartite stem,
exemplified by the likes of
/howwal-/ 'run, jump on top'
/howw-/ 'run, jump into water or a flat place'
from ex. (1). These have a LDS as the second element, but the LP
is drawn from a motley residual set of morphemes which do not
show the characteristic semantic or syntactic behavior of either
the Instrumental or Theme-classifying subcategories. The bulk of
these have meanings related to manner of motion or position, e.g.
15) {cv} 'sg. sit, slide'
{tgv} 'sg. stand'
{lvlw} 'pl. stand'
{gv} 'go, move of one's own volition'
{hod} 'sg. run, jump';
Bipartite stems formed on one of these are typically
intransitive, with no alternate transitive use:
16) he: mi-s sli-wapk, Gleg-atk gintak,
if 2nd-OBJ shoot-FUT sg.die-STAT even
?i hot-tgal-b@l-ank sli-wapk honk-s.
2sg jump-get.up-back-having shoot-FUT DEM-OBJ
"If he shoots you, even if you're dead, you'll jump back up
and shoot him."
17) Glewy-ank g-oWaskc-a l@meys-?as gelwipc-ok
quit-having go-away-IND Thunder-OBJ visit-PURP
c'asga:y'a:k-s ha-ks-akc'wi-tk.
Little.Weasel-OBJ REFL-living.obj-narrow.place-STAT
'Leaving there, he went off to visit the Thunders, with
Little Weasel in his pocket.'
The relation between the subject and the LDS is the same as with
the Theme-classifying stems, i.e. the verb predicates of its
subject motion or location defined by the LDS. I have a few
examples of stems of this type used transitively:
18) coy mi-s ni g-alamn-bag-wapk.
then 2nd-OBJ I go-behind-CIS-FUT
'And I'll come following you.'
Presumably the transitivity here is an unsystematic consequence
of the lexical semantics associated with |g-alamn| 'go-behind'.
Note, however, that this is not directly comparable with the
systematic transitive use of bipartite stems consisting of a
classifying LP + LDS. In the regular pattern the subject of the
intransitive and the object of the transitive are the same Theme
argument; in (18) the agentive Theme argument is subject even in
the transitive reading.
A handful of other LP's combine with LDS's to form
intrinsically transitive stems; the transitivity is usually
directly predictable from the intrisically transitive semantics
of the LP.
D. Stem categories and argument structure
Thus the inventory of systematic bipartite stem types is as
follows:
19) Instrumental LP + COS stem Agent, Theme
Classifying LP + LDS (Agent), Theme
oblique LOC
Motion LP + LDS Agent = Theme
oblique LOC
For these regular patterns the argument structure can be derived
compositionally;
20) A COS stem has a single NP argument slot, for a Theme.
LDS's specify a LOC, and thus necessarily have a NP argument
slot for a Theme.
Instrumental LP's and the motional miscellaneous LP's name
an activity, and thus provide an Agent slot.
Classifying LP's refer to a Theme, but have to be in
construction with a second element which provides a
syntactic slot for it, i.e. they do not appear to take
arguments of their own.