to DeLancey's
Linguistics Home Page
comments welcome: delancey@darkwing.uoregon.edu
Morphological Parallels between Klamath and Wintu
Scott DeLancey
University of Oregon
Of the Oregon languages which have been suggested as being
affiliated with the California languages labelled Penutian,
Klamath seems on geographical as well as general structural
grounds to be the most promising place to look for evidence of
Penutian relationship.1,2 This paper
records a few suggestive points of morphological similarity
between Klamath and Wintu, its nearest California Penutian
neighbor.3 Wintu data not otherwise attributed are
from Pitkin 1984; Klamath data not otherwise attributed are from
Barker 1964.
We will consider comparisons between the two languages in
several systems: the pronominal paradigms, a special category of
"locative" words with an associated set of special postpositions,
the nominalization paradigm in the verb system, and an associated
marker of definiteness or referentiality. These data are not all
equally compelling. We first present the very striking
similarities in the pronouns, which are of a nature which cannot
be due to chance. The similarities between the "locative"
systems are less obvious, and the differences considerably
greater; nevertheless the similarities are sufficient that they
can contribute to a case for relationship. The possible similar-
ities in the verb system are more subtle, and on their own would
carry little conviction. They are worth considering in the
context of the other morphological similarities, but in any case
we will require some work on internal reconstruction of Klamath,
and internal and comparative reconstruction of Wintuan, before
these can be finally evaluated. The same is true of the t
suffixes in the two languages which I will try to equate in the
final section; although there are clearly intriguing simi-
larities, they require a richer context both of comparative
Klamath-Wintuan and of syntactic analysis of Klamath and of
Wintuan languages before their significance can be finally
determined.
I. Klamath and Wintu pronouns4
The most striking, and most compelling, parallel which I have
observed between the two languages is in the pronominal para-
digms. The two languages share not only the same pronominal
roots, but also contrastive suffixes, object markers, and plural
stem formations. The comparable singular and plural forms are as
follows (note that Barker's /b/ represents unaspirated [p],
contrasting with aspirated /p/ = [ph], and Pitkin's
/p/ is also [p], contrasting with aspirated /ph/ =
[ph]):
SUBJECT CONTRASTIVE OBJECT GENITIVE
1 sg. ni niyo nis net(o)
1 pl. nele-/nite- nele:n
2 sg. mi miyo mis mat(o)
2 pl. male-/mite- male:n
3 sg. pi piyo put putun
3 pl. pite-/puba: puba:n-
WINTU PRONOMINAL PARADIGM (SIMPLIFIED)
SUBJECT CONTRASTIVE OBJECT GENITIVE
1 sg. ni no: nis/no:s gew
1 pl. na:d na:l's/na:ds na:l'm
2 sg. ?i mis mi
2 pl. ?a:d ma:l's/ma:ds ma:l'm
3 sg. bi bas/bo:s/ba mna
3 pl. sa ba:d mna:ls mna:l'm
KLAMATH PRONOMINAL PARADIGM
The Klamath and Wintu 3rd person plural forms are more parallel
than they appear; we can identify the /m/ in the Klamath forms
with the {b} element in the other 3rd person forms, and hence
with the corresponding Wintu forms, on the basis of Gatschet's
(1890:546 ff.) recording of the corresponding Modoc forms as
p'na, p'nalam, etc. The Klamath /m/ thus obviously
represents assimilation of this stop to the following /n/.
We should not overlook as an initial similarity the
existence in both languages of a set of non-contrastive in-
dependent pronouns, which is by no means universal among North
American languages. We can further identify as common to the two
systems the following: a set of pronominal roots {ni} '1st',
{mi} '2nd', {pi} '3rd'; a contrastive suffix {o}, reflected in
the Klamath 1st person contrastive /no:/, and probably in the 3rd
object form /bo:s/, by analogy with the 1st obj. contrastive form
/no:s/; an object marker {-s}, and parallel paired plural stems,
one in -l and one in -t.
It is sometimes suggested that in view of the ubiquity of
1st and 2nd pronominal roots in, respectively, n and
m in North American languages, this resemblance cannot be
taken as indicative of particularly close relationship between
North American languages manifesting it. In this vein it has
been suggested to me that the similarity of the 3rd person roots
is stronger evidence of relationship than that of the 1st and 2nd
person roots. But, of course, such evidence is compelling only
in context; similarities between the third person roots would
carry little weight if the 1st and 2nd person roots were
radically different. The coincidence of the three taken
together, however, must be taken seriously, especially in the
context of the rest of the paradigms.
Given that the two languages both have non-contrastive
pronominal forms which are not bound into the verb, the fact that
both have contrastive formations is not surprising; but the fact
that both appear to have the same contrastive morpheme is
noteworthy. The distribution of the {o} element in the two
paradigms is not absolutely identical. Note that in Klamath the
3rd singular {bi} form is contrastive; the noncontrastive 3rd
singular function is carried by the demonstratives. Moreover,
Barker's data contain no 2nd person contrastive form.5
It is striking, however, that in both languages we not only find
the {o} element, but also find it restricted to singular forms.
The next shared feature is the objective suffix {s}. The
resemblance of the objective forms is of particular interest
because, while {s} is the regular object suffix in Klamath, it
does not occur in Wintu outside of the pronominal paradigm. (The
regular Wintu object suffix is {um}, the resemblance of which to
the Klamath and Sahaptian genitive markers deserves further
investigation; kin terms and 3rd person pronouns take a |-t|
object suffix).6
The most significant, because the most inexplicable, resem-
blance between the paradigms is the parallelism of the two plural
stems, one in -l and one in -t. In Wintu this alternation occurs
only in the 1st and 2nd plural forms; in Klamath the actual
alternation is likewise restricted to 1st and 2nd person,
although it appears possible that in the 3rd person plural forms
the subject (contrastive) form may be constructed on the -t stem,
and the oblique forms on the -l stem. (Gatschet lists only -l
stem oblique forms for 1st and 2nd person, but for 3rd person
plural object lists, along with p'nals or p'nalas,
a form pa:ntch, presumably his recording of /pa:nds/,
which could be the corresponding -t stem. He also list a form
p'na:s, which lacks either plural formative). With the
possible exception of the Klamath 3rd person forms, the
alternation has no apparent synchronic significance in either
language. Note that in Klamath only the -t stem occurs in
subject forms, and only the -l stem in genitives; with the object
suffix the two stems are in free variation.6 In Wintu
likewise only the -l stem occurs in genitive forms; in the
paradigms given by Pitkin there is no separate objective form for
the plural pronominals.
We must also take note of several other differences among
the paradigms. To begin with, I have omitted from the Wintu
paradigm a dual series and a set of reflexives with no parallel
in Klamath, as well as an inclusive form and an instrumental
series which, unlike the object forms, are formed regularly by
the addition of the normal instrumental suffix to the oblique
stem. From the transparent regularity of the dual, reflexive,
and instrumental series in Wintu we can infer that they are
recent developments within Wintuan, rather than old elements lost
in Klamath.
Included in the data given here are several Klamath forms
which do not match with anything in Wintu:7 the 2nd
sg. and pl. subject {?i} and {?a:d}, 3rd pl. subject {sa}, and
1st sg. possessive {gew}. These do not seem likely to admit of
any comparative interpretation, but they are also precisely the
forms which do not fit regularly within their own
paradigm.8
II. Special locative constructions
Klamath has a set of 28 morphemes of generally locational or
directional sense, which are distinguished as a class by their
occurrence with a set of locative suffixes which do not occur
with members of other categories (Barker 1964:278-80, 285-7).
The suffixes which occur with these "Locatives" are ({dal'} and
{y'e:n'} occur with members of some other syntactic categories):
{t} {t'a} 'in, on, at'
{t'i:t} 'in, on, at' (Barker analyzes this as {t}
plus a morpheme {'i:t}
occurring only with {t})
{na} 'around, in a general area'
{dal'} 'toward'
{y'e:n'} "nonnominative theme formant"
Wintu has a similarly defined set of eleven "uninflected
words of directional meaning" , and a set of locative suffixes
which occur only with these "directionals" and two other
morphemes, /tu/ 'ahead' and /xun/ 'toward' which do not occur as
free forms (Pitkin 1984:262-3). The suffixes are:
{ti} 'at, in'
{da} 'from, of'
{dal} 'at, in the direction of'
{el} 'toward, in'
While there do not seem to be any obvious (or even plau-
sible) resemblances between semantically corresponding members of
the Klamath Locative and Wintu directional stem catego-
ries,9 there is an obvious parallel in the existence
in each language of such a category, defined by its own unique
locative morphology. The parallel gains interest when we note
the similarities in the sets of special locative suffixes which
define this category:
Klamath Wintu
{t}, {t'a}, {t'i:t} {ti} 'at, in'
'in, on, at'
{dal'} 'toward' {dal} 'at, in the direction
of'
{el} 'toward'
These resemblances are certainly striking, but in themselves too
few to count as very much evidence for anything. However, their
evidentiary value is considerably increased by their
morphosyntactic parallelism: if to the obvious phonological and
semantic congruence we add the fact that in each language these
morphemes constitute a special category of locative suffixes
which occur only with a small set of locative/directional
morphemes, the degree of arbitrary resemblance between the two
sets begins to be more than should be casually attributed to
chance. Note, moreover, that the restricted distribution of
these morphemes in both languages is evidence of antiquity; they
appear to be old locative constructions retained in a semi-frozen
lexical subsystem, and replaced elsewhere by newer formations.
III. Some notes on the Klamath and Wintu verb
At first glance a suggestion of a genetic connection between
Klamath and Wintu would seem to find no support whatever in the
verb systems of the two languages. While they do share a number
of non-universal syntactic and semantic categories, none seem so
striking as to require any explanation beyond chance and shared
areal features, and structurally the system described by Barker
and that described by Pitkin seem on superficial inspection to
have little in common beyond a taste for suffixation. However,
through internal reconstruction and further analysis, the Klamath
system can be simplified to a point where it is much more
directly comparable to that of Wintu, and at that point
similarities can be found between the exponents of some
comparable categories in the two languages which may well repay
further investigation.
Of the numerous suffixal position classes in Barker's analysis
of the Klamath verb, several of those closest to the stem
(Barker's classes 17-19 and 21) can be shown to be later
developments of morphemes originating as locative "suf-
fixes",10 and thus ignored for comparative purposes.
Another class (Barker's 20) consists of morphemes whose
phonological bulk and semantic specificity imply fairly recent
incorporation into the verb. In order to compare the remaining
elements of the verb complex, it is necessary to add one
dimension to Barker's analysis, and distinguish subsystems of
suffixes, rather than simply position classes. Thus, for
example, Barker places in a single class (22) the morphemes
{astg} 'tried but failed to', {ang} 'polite imperative
singular', and {w} 'past nominalizer', and in the next class (23)
{a} 'indicative', {i} 'sg. imperative', {at} 'pl.
imperative', and {y} 'noun formant'. This presentation obscures
the fact that {astg} cannot be followed by {i} or
{at}, {ang} can be followed only by {at} or
nothing, while {w} can only, and must, be followed by {y}.
While I do not yet have a complete reanalysis of the Klamath
verb to present, it seems at present that Barker's classes 22-25
can be reorganized into a system roughly like the
following:11
INDICATIVE SYSTEM: astg} a
Wi:
at
IMPERATIVE SYSTEM: i
(ang () at)
ek
n'a
SUBORDINATE SYSTEM: ank
tko
(dgi) wk
-----------------------------------------------------------
NOMINALIZATION SYSTEM: k' y s t
w i
A Tentative Reconstruction of the Klamath Verb
The nominalization system will be discussed below; rough glosses
for the other morphemes are: {astg} 'tried but failed
to', {Wi:} 'almost, used to', {a} 'indicative', {at} 'can, ought
to', {ang} 'polite imperative', {i} 'sg. imperative',
{at} 'pl. imperative', {ek} '1st sg. hortative', {n'a}
'1st pl. hortative', {ank} 'temporal anteriority', {dk(o)}
'(resulting) state', {dgi} (usually + {wk} or {gi} 'be' + {wk})
'complement to verbs of persuasion or coercion', {wk} 'cause,
purpose, or goal'.
All Wintu verb forms are based on one of three stems, which
Pitkin calls Indicative, Imperative, and Nominal. The Indicative
stem formant, in form and allomorphy, is reminiscent of the
Klamath indicative suffix {a}. The Imperative formation is not
obviously similar to the Klamath imperative system; there are,
however, reasons to think that some or all of the Klamath
imperative suffixes may be relatively recent developments. Of
particular interest here is the Wintu Nominal stem. It is formed
by a morpheme whose basic shape is {i}, which ordinarily must be
followed by one of the two "nominal aspect" morphemes {s} or {t}.
Compare the Wintu and Klamath nominal systems, recalling that
just as Klamath {y} must be followed by {s}, which may then
optionally be followed by {t} or {i}, so Wintu {i} must be
followed by either {s} or {t}:
Wintu: t
i
s
Klamath: k' t
y s
w i
This minor formal similarity is, in itself, not compelling,
although if we were sure on other grounds (such as the similarity
of the pronominal paradigms, for example) that the languages were
related, it would seem likely prima facie that the corresponding
parts of these paradigms are cognate. Obviously this argument
will be strengthened if the functions of the corresponding
i/y, s, and t morphemes are found to be
somehow relatable.
Some functional parallels can be adduced which bolster the
case somewhat, although the argument cannot be pursued very far
without a more thorough understanding of both Klamath and Wintu
syntax than I possess. Wintu {i} and Klamath {y} are roughly
parallel in having no particular function beyond creating a
nominal stem. Wintu {s} and {t} are the exponents of the famous
system of "nominal aspect" described by Lee (1942, 1944).
Without going into the intricacies of the system described by Lee
and the reanalyzed system described by Pitkin, we may note simply
that the formation reflected in the |s| allomorph of Pitkin's
generic morpheme {s} appears historically to be a simple
nominalizer (cf. Pitkin 1984, Whistler 1980:256), and thus
parallel to Klamath {s}. The {t} 'particular aspect' category
and its Klamath parallels participate in both the noun and verb
systems, and thus deserve a separate section.
IV. The definite {t} suffixes
The |t| which participates in the Klamath nominalization system
is analyzed by Barker as the same morpheme as a {t} which he
glosses 'referential', which occurs also with various other
nominals. In most of its uses it indicates "something previously
referred to" (Barker 1964:173); in this function it occurs freely
with demonstratives, and otherwise only with time expressions.
This rather peculiar restriction suggests that the current
distribution of the morpheme in Klamath may be a relic of an
earlier more productive construction, although other hypotheses
are possible.
With verb stems {t} occurs only with a preceding {s}, and the
distinction between nominalizations in {-s} and in {-s-t} marks
the distinction between complement clauses with subject the same
as ({-s}) and different from ({-s-t}) that of the main clause.
Thus we find {-s-t} in examples like:
domna ?a-n honks swi:-s-t
hear DEC-1st DEM/OBJ singing
'I hear him singing.'
The {t} morpheme in Wintu marks the "particular" aspect of
nominals. I will not attempt to summarize here the discussions
by Lee and Pitkin of the semantics and morphology of this
category, beyond noting that it is associated with, among other
things, referentiality and definiteness,12 and that
|t| is one of two exponents of the category, the other being |h|.
The |t| allomorph13 occurs regularly with pronominals
and nominalized verbs, i.e. in a distribution reminiscent of
Klamath 'referential' {t}.
Thus in form, distribution, and at least rough function, the
Klamath 'referential' and the Wintu 'particular' {t} show
considerable similarity. Again, it seems that, while the
resemblances are not sufficient to themselves make a case for
relationship, they are of a nature that in the context of other
evidence for relationship makes them worth pursuing further.
Notes
1) Work reported in this paper was supported in part by
Grant # RO-21292-86 from the National Endowment for the
Humanities.
2) This is not, of course, to say that there is no use in
searching further afield. It is clear at this point that there
is a genetic relationship between Klamath and Sahaptian, from
which it follows that if it is worthwhile to look for Klamath-
California connections, it is worthwhile to look for Sahaptian-
California connections. In the case of other putatively Penutian
languages, so little comparative work has been done (though
recently see Buckley 1988) that there is no substantive basis for
predicting what the outcome of such work would be.
3) This locution should not be taken to imply that there
is any reason at this point to think that the relationship
between Klamath and Wintu is particularly close within Penutian.
4) The data in this section will be published separately
in IJAL.
5) The Klamath data suggest that the contrastive is no
longer a transparent regular formation in the language. This
probably correlates with the fact that the non-contrastive forms
have a strong tendency to cliticize, normally in sentence-second
position. The language as recorded thus seems to be tending
toward a Sahaptin-like pronominal system, with second position
clitic non-contrastive pronouns, and only contrastive independent
pronouns. (There is no apparent difference in the systems
recorded by Barker and by Gatschet, so it is probably not the
case that the decay of the contrastive formation represents a
late phenomenon associated with the moribund state of the
language when Barker recorded it).
6) Howard Berman and Victor Golla, in personal communica-
tions, have both pointed out an -s/-se object suffix in Cos-
tanoan. Note however that there is some evidence for an old
vocalic object marker in some California languages (Berman
1983:402). This latter may be reflected in the Klamath {a}
object suffix used with adjectives and kin terms (Barker
1964:235, 266), which seems to be a conservative retention of an
older pattern, and which may be equatable with the Sahaptian
subject suffix {a} which occurs only with kin terms.
7) Barker 1964:240: "No semantic distinction is found
between |na:ds| and |na:l's|, but the latter is more frequent";
the same is true for the 2nd and 3rd person alternations.
8) I have omitted only non-matching categories, such as
the Wintu inclusive; where the two languages share a category but
show non-comparable forms the forms are nevertheless given. I
have not included in the discussion the Klamath reduced clitic
forms; most of these are transparently relatable to the fuller
Klamath forms discussed here.
9) Victor Golla (p.c.) notes a few forms from Yokuts,
Takelma, and Chinookan which suggest that the Klamath {ni} /
{gew} alternation may be an old inheritance; this would imply
that the Wintu 1st genitive form might be a later analogical
regularization. Gatschet (1890:547) suggests that 2nd person
{?i} and {?a:d} were innovated to avoid homophony between the
singular *mi and the 2nd sg. genitive, and between the plural
*ma:t and an evidential particle {mat} which, like the pronouns,
tends to occur in 2nd position. There may be an element of truth
in this, but it is at least oversimple; note that this account
requires, to begin with, an explanation for how 2nd person *mi,
presumably originally subjective, acquired its genitive function.
10) For example, while the names of the cardinal directions
belong to the relevant category in both languages, there are no
apparent resemblances in the forms. However, it is worth noting
that some at least of the Klamath directional terms appear to be
of no great antiquity: /lobi:t-dal'/ 'east, toward the east' is
literally 'toward the front' (Barker 1963:220), referring to the
orientation of a house.
11) Klamath has a bipartite stem construction very similar
to that described by Jacobsen (1980) for Washo, with over a
hundred primarily locative or motional "suffixes"; it is these
that are the source for a significant part of the modern verb
morphology. This system, which is not shared with Wintu, may be
a trans-montane areal phenomenon, as very similar systems are
found in Sahaptian and Atsugewi, while similarly elaborated
systems are lacking in Wintu and Pomo, and in general appear not
to be found west of the mountains.
12) Barker also includes in class 23 the future suffix
{wabg}, but in Gatschet's data this occurs regularly followed by
the indicative {a}, and must thus be relegated to an earlier
position class. I am hopeful that further analysis of Gatschet's
data will make it possible to assign {wabg} to its historically
proper position.
White (1983:5) notes that of the non-nominalizing members of
category 23, only {wabg} can occur non-finally, that is, followed
by members of classes 24 and 25. Thus elimination of {wabg} from
class 23 obviates the need for separate classes 24 and 25 outside
of the nominalization system.
13) While noting that the synchronic function of the
particular aspect cannot be described simply as an analogue to
definiteness in western European languages, Pitkin suggests that
the {t} suffix may have developed from "a topic-
alizing/foregrounding suffix *{t} with something like the force
of a definite article" (1984:203). (Cp. the similar hypothesis
of Whistler (1980)
14) |t| and |h| are treated, explicitly by Pitkin and
implicitly by others, as allomorphs of a single particular aspect
morpheme {t}. While this is certainly the appropriate synchronic
analysis, it seems clear that the two morphs have different
REFERENCES
Barker, M.A.R. 1963. Klamath Dictionary, UCPL 31. Berkeley and
Los Angeles: University of California Press.
Barker, M.A.R. 1964. Klamath Grammar, UCPL 32. Berkeley and Los
Angeles: University of California Press.
Berman, Howard. 1983. Some California Penutian morphological
elements. IJAL 49:400-12.
Buckley, Eugene. 1988. Coast Oregon Penutian: A lexical
comparison. Proceedings of the Third Pacific Linguistics
Conference. Eugene, OR: Dept. of Linguistics, University of
Oregon.
Gatschet, Albert S. 1890. The Klamath Indians of Southwestern
Oregon, Part I. Contributions to North American Ethnology, vol.
II. Washington: Government Printing Office.
Jacobsen, William. 1980. Washo bipartite stems. K. Klar, M.
Langdon, and S. Silver, eds., American Indian and Indoeuropean
Studies: Papers in Honor of Madison S. Beeler, pp. 85-99. New
York: Mouton.
Lee, Dorothy Demetracopoulou. 1942. Noun categories in Wintu.
Zeitschrift fr vergleichende Sprachforschung 6.197-210.
Lee, Dorothy Demetracopoulou. 1944. Categories of the generic and
particular in Wintu:. AA, n.s. 46:362-9.
Pitkin, Harvey. 1984. Wintu Grammar, UCPL 94. Berkeley and Los
Angeles: University of California Press.
Whistler, Kenneth. 1980. Proto-Wintun Kin Classification: A Case
Study in Reconstruction of a Complex Semantic System. Ph.D.
dissertation, University of California, Berkeley.
White, Robin. 1973. Klamath Phonology. Studies in Linguistics
and Language Learning 12. Seattle: Dept. of Linguistics,
University of Washington.