to DeLancey's
Linguistics Home Page
comments welcome: delancey@darkwing.uoregon.edu
Klamath Stem Structure in Genetic and Areal Perspective
Scott DeLancey
University of Oregon
Two grammatical phenomena which are common in New World
languages are the presence in the verb of a series of morphemes
which specify the shape or other classificatory features of the
Theme1 or Instrument argument of the verb, and a
series which indicates the direction or nature of motion or
location. In several California and Oregon languages we find a
particular constellation of features associated with this
structure: the classifying series is prefixed, i.e. they precede
other stems, while the motional/direction series are a distinct
category which follows other elements of the stem. Even in the
analysis of particular languages there is some uncertainty about
whether to treat these two series as affixal or as bound stems;
as Langdon (1987ms) suggests, the question can at best be
answered only for particular stages of particular languages, and
even there the issue is often worth further discussion.
Jacobsen (1980) describes a particularly elaborate system in
Washo, in which both series are found. Unusual features of Washo
are the large number of morphemes in each series, and the fact
that, although both series are bound morphemes, there is a large
number of "bipartite" verb stems consisting only of one morpheme
from each series, with no independent verb stem. The starting
point of this paper is the striking similarity in this respect
between Washo and Klamath. While the Klamath and Washo systems
are dissimilar in points of detail, in its broad outlines
Jacobsen's description of Washo is directly applicable to
Klamath. A similar constellation of features appears to obtain
in Sahaptin and in Atsugewi. Since there is no likelihood of
these parallels among these four languages representing
inheritance from a common ancestor, and since the languages are
geographically contiguous, we can infer an areal phenomenon. For
convenience's sake I will refer to these as "transmontane"
languages, because they are all spoken on the other side of the
mountains from most of the Hokan and Penutian languages.
In this paper I will briefly summarize the constellation of
features which characterizes this areal grouping, and try to show
how Klamath may have come to manifest it. For present purposes I
will adopt Jacobsen's terminology and refer to the initial stem
elements, many of which classify themes or instruments, as
"lexical prefixes" (LP's); the final elements, most of which have
a motion/location sense, Jacobsen labels "dependent verb stems";
since, as we will see, Klamath has other dependent stems which do
not clearly belong to this category, I will adapt Jacobsen's term
and call these "dependent motional stems" (DMS's).
The "transmontane" pattern
The features shared by Klamath and Washo, and, at least to a
considerable extent, Sahaptin and Atsugewi, are as follows.
First, there are distinct LP and DMS categories; the LP's precede
other elements of the stem, and the DMS's follow. Second, though
both LP's and DMS's are bound forms, a LP and a DMS can form an
independent stem, with no independent verb stem necessary, as
Klamath /kselwi/, morphologically |ksv| 'act upon a
living object' (a LP of Barker's position class 4) plus |elwy|
'by the fire, along the edge, into water' (a DMS of class 10).
Thirdly, the number of members of the LP and DMS class is
relatively large, on the order of 50-200 members in each; in
other California languages the numbers are more on the order of
5-20 members in each.2 Correlating with the expanded
size of the categories, at least in Klamath, Sahaptin, and Washo,
is a reduction in the semantic homogeneity of the
morphosyntactically defined categories. In Klamath, for example,
the LP category, while including about 20 each of theme and
instrumental classificatory morphemes, also includes morphemes
with meanings such as 'sit', 'run', 'swim (fish)', 'burn', 'peek,
glance', and 'be striped', which cannot be interpreted as
classificatory. (Some, indeed, would seem to pertain more
closely to the motional category). Likewise, while the large
majority of Barker's list of 129 members of the DMS class code
familiar motional notions like 'up against, to the shore', 'down
from a height', and 'out of a container', there are also a few
morphemes with meanings like 'slitting open (as a sack, bulbous
obj.)', 'dividing, distributing', 'in strips (as meat)', and
'give (sg. obj.)'.
While each of these features occurs in a wide range of
languages with no obvious genetic or areal unity, this particular
constellation of features seems to be characteristic of a
defineable areal grouping, including Hokan and Penutian languages
spoken east of the Sierra/Cascade ranges in Oregon and
northwestern California. This suggests that this stem structure,
shared by Sahaptin, Klamath, Atsugewi, and Washo, is a secondary
areal development which needs to be taken into account in
historical and comparative Penutian and Hokan studies.
In the Hokan and Penutian languages in and surrounding this
area, these features show roughly the following distribution:
Wintu: no LP's, DMS's precede stem
Pomoan: small sets of prefixed LP's and suffixed DMS's
Shasta: small sets of prefixed LP's and suffixed DMS's
Maidu: LP's precede, DMS's follow, fewer than 20 of each;
LP+DMS stems
Atsugewi: LP's precede, DMS's follow, LP+DMS stems; ÷40-50
of each
Washo, Klamath, Sahaptin: LP's precede, DMS's follow, LP+DMS
stems; ÷100-200 of each
Genetic affiliations among these languages remain a controversial
issue. I will assume in the present argument both the Hokan and
the Penutian hypotheses, with Klamath and Sahaptin both Penutian.
While these claims are subject to debate, it is in any case clear
that Washo, Klamath, Sahaptin, and Atsugewi do not form any sort
of genetic unit as opposed to the other languages listed. Thus
the common structural features which appear to characterize them
as a group must have some explanation other than common
inheritance. These four languages do, however, form a
geographically contiguous band along the east side of the
mountains, and there is enough historical evidence of contact
among them to support the hypothesis that they might share some
areal features which distinguish them from other related
languages.
Klamath stem structure
To oversimplify matters for a moment, we can think of the
elements of the Klamath verb stem as of three types: LP's,
DMS's, and independent stems. All three can be seen in the form
/Go:t'aqt-/ 'crawl under head first; wrap the head in a shawl',
which is morphemically |Go:-t'aq-(o)t(n)|, comprising the
instrumental classifying prefix |Go:| 'with head first' (in
Barkers position class 4), the independent stem |t'aq| 'act with
the head' (class 7), and the dependent locational stem |otn|
'on,against, attached to' (class 10).
These classes show the same phonotactic patterning as the
corresponding Washo classes. The prefixes are predominantly of
small phonological profile: a few members, most of which show
other signs of recent entry into the class, are CCVC, but most
consist of one or two consonants, with complex morphophonemic
patterns of syncope, epenthesis, and vowel copying. The
dependent stems are large, often trisyllabic; most have at least
two, and many as many as four, consonants, with many long vowels
and closed syllables (e.g. {aptne:Gi}, {at'amsg}, {c'i:sq'ag},
{oditgo:l}). Jacobsen interprets these differences as indicative
of relative age; this may be dangerous, since, at least in
Klamath, it may be that stress patterns would lead to much more
rapid phonological attrition in some positions in the word than
in others. Nevertheless, a comparison of Klamath and Maiduan
DMS's (below) suggests that the bulkier Klamath forms may be
relatively new, and I will show some internal evidence for the
same conclusion. However, this should not be taken to suggest
that the LP category per se is necessarily an older development
than the DMS category.
In terms of this scheme, a verb stem should be one of four
patterns: independent stem, LP + independent stem, independent
stem + DMS, or LP + independent stem + DMS. In fact matters are
considerably more complicated than this. The first complication
is that, as in the Washo system described by Jacobsen, there are
also many stems consisting only of prefix + dependent stem, i.e.
with no element which can stand on its own as a verb stem: i.e.
/slen-/ 'take a clothlike obj. away', {sle} (class 4)
'act upon a clothlike obj.', {en} (class 10) 'taking,
going'. The second is that there are several morphemes which
seem to be intermediate in their behavior between the independent
stem and one of the other classes, e.g. some members of the
independent stem class which can occur only with a prefix.
Barker takes as criterial for membership in either of our
dependent categories the possibility of occurring with an
independent stem. There are, however, a number of stems which
fail this test, but which nevertheless cannot occur
independently; some can occur only with an LP, others only with a
DMS. Thus the inventory of stem-building morpheme classes is
(adapting Barker's labels):
I. Classificatory morphemes (LP's)
II. Stems which require a DMS
III. Free stems (some occur with LP's and DMS's, some do
not)
IV. Stems which require an LP
V. Locative-directive morphemes (DMS's)
Barker notes for several members of my categories II and IV,
which we may call simply dependent stems, that they might well
actually belong to the LP or DMS categories; his classification
reflects only the fact that in his corpus they do not occur with
independent stems.
The existence of bipartite stems, both parts of which are
members of bound categories, poses an analytical and historical
problem. In the case of a construction like |ban-ba:tn-a|
/bamba:ta/ 'wades to the shore, edge', the independent occurrence
of |ban-a| /bana/ 'dives, wades' establishes the verbal nature of
{ban}, and it appears reasonable to then consider {aba:tn}
'to the shore, edge' as a derivational suffix. Similarly,
comparing |?ec'-a| /?ec'a/ 'suckles' with |qbv-ec'-
c'n-a| /qbecc'a/ 'just sucked dry' suggests that {?ec'}
'suck dry, suckle' is the verbal root, with {qbv} 'act
with the mouth' a derivational prefix. However, in the case of a
stem like |ne-ak'a:yi'| /nak'a:y'-/ 'put a flat
object up high', there is no obvious way to identify one of the
morphemes as the verbal root and the other as a derivational
element, since neither can occur alone as a verb stem. The
problem is compounded by the fact that {ne} 'act upon
a flat object' is in the same position class as the
{qbv} of |qbv-ec'-c'n-a|, and
{ak'a:yi'} 'up high' as the {aba:tn} of |ban-
ba:tn-a|, both of which we identified as derivational
affixes on the basis of their occurrence with potentially
independent roots.
Assuming that it is necessary to identify one of the two
morphemes as the locus of the verbal status of the stem, we can
construct an indirect argument based on the fact that, while most
(though not all) of the members of the prefix class can be found
occurring with verb roots that also function as independent
stems, many of the members of the
{aba:tn}/{ak'a:yi'} category do not; they occur
only with members of the prefix class. This suggests that the
DMS's are more like verb roots than are the LP's--or better, in
diachronic terms, that the DMS's probably reflect earlier verb
roots, while the LP's may not.
This suggestion is buttressed by the existence of several
morphemes which seem to be partway along a path of development
from verb root to dependent stem. For example, {p'ac'} 'be
blind, put out an eye, be hidden, out of sight, out (like a
light)', behaves like a class 7 independent root in the simple
stem /p'ac'/ 'eye, water-filled obj. burst', the reduplicated
/p'ap'a:c'/ 'blind', and in stem |kt-p'ac'| /kp'ac'/ 'hit in the
eye' ({kt} 'hit with the fist, kick' is a LP). However, with
vowel-final prefixes this morpheme shows the morphophonemic
behavior characteristic of suffixes: |sle-abc'|
/slapc'/ 'cover with a clothlike obj. so as to hide s.t.'.
Moreover, Barker analyzes one occurrence of {p'ac'} as a
dependent stem; in the form |pniw-bc'| /pniwpc'/ 'blow out a
candle, light', {pniw} 'blow' is a class 7 verb root according to
Barker's analysis, so that {p'ac'} --> |bc'| 'out of sight, out
(like a light)' then must be functioning as a member of the
dependent stem class. Farther along in this direction is
{p'e'q'} '(act) on the face', which like {p'ac'} has allomorphs
which show the morphophonemic behavior characteristic of
suffixes, but unlike {p'ac'} cannot occur without a prefix.
The claim that the dependent motion stems are verbal in
nature, or at least in origin, does not entail any conclusion
about the nature or origin of the lexical prefixes. The noun
classifiying semantics of many of them imply a possible origin as
incorporated nouns, and there is some comparative evidence for a
nominal origin for a few of these: {c'le} 'act upon a
massive shapeless object (such as a piece of meat)', the clearest
case, is obviously related to the noun /c'ole:ks/ 'meat, flesh,
body', and to the cognate Nez Perce form cil akt 'body'.
Others are more problematic: {sle} 'act upon a
clothlike object', for example, is clearly related to the root
*sla 'mat', but this occurs both in the noun /slays/ 'mat' and in
the verb root /slan-/ 'spread out a mat', and there is no obvious
way to establish one or the other as the etymologically primary
sense. (There is a considerable number of such roots in Klamath,
which Barker consistently treats synchronically as basically
verbal. There is a reasonable case for this synchronic analysis,
although I suspect that a decent case could also be made for
treating such roots as fundamentally neutral as to syntactic
category. In any case, the historical problem remains). There
are also a handful of prefixes with apparent Sahaptian cognates,
suggesting that these are of at least Proto-Klamath-Sahaptian
provenience, making it impossible, in the present state of our
knowledge, to identify origins for them in independent roots.
Klamath and Maiduan
Let us first compare Klamath with Maiduan, a Penutian language
which shows only traces of this transmontane typological pattern.
Maiduan has both prefixed LP and suffixed DMS categories, exactly
like Klamath, and their positional behavior is similar to that of
Klamath. However, the number of members in each category is
quite small; there are fewer than twenty DMS's; while no source
gives an exhaustive list of the LP's, the number seems to be of
the same order. (Dixon lists 17 prefixes, but notes that there
are probably others to be identified). Klamath has five or ten
times as many of each.
For the most part the Maiduan LP's are indeed classifying
morphemes. The morphophonemic variation of the Maiduan prefixes
described by Ultan (1967) is strikingly reminiscent of that in
Klamath, and at least two of the Maiduan prefixes could be linked
with plausible Klamath cognates. Compare Maiduan {wO} 'with a
stick-like instrument' and Klamath {w} 'act with a long
instrument, fall'; there is even a bimorphemic construction with
this element which is apparently cognate: Maidu
w¢o-kot-dau 'to chop off the end of a log' (Dixon
1911:697) and Klamath |w-gatt'-| /wgatt'-/ 'chop in two, chop
down', consisting of {w} (4) 'act with a long instrument',
{gatt'} (7) 'break, cut in two, chop down' (Barker 1963:426).
Also suggestive is the semantic similarity of Maiduan {jO}
'[act with] pointed instrument' (Ultan), and Klamath {y'o:}
'shoot (a weapon) pl. times'. The semantic equation is improved
by recourse to Dixon's gloss:
Actions performed with the end of a long
thing, endways, or in a direction parallel to
the length of the thing
a sense which would obviously include the normal use of an arrow;
and by examples like y a-bak-dau 'to knock bark off tree
by stroke with arrow or bullet' and y‚-dek-ton 'to shoot
through anything, and pierce'.
The dependent motional stems (Dixon's "suffixes expressing
direction of motion", Shipley's Motion-Location Auxiliaries,
Ultan's Locative Directionals) correspond to their Klamath
counterparts much as do the prefixes. Though only a tenth as
numerous, they have the same position in the verb complex and the
same central semantic function; compare the parallel formations:
Maidu: /wel‚-doj/ 'to run up (for example, a hill)', {wele}
(IS) 'run'; {doJ} (DMS) 'up'
Klamath: |wle-qwe:L-| 'few, four-legged animal run down
a hill', consisting of {wle} (IS, 7S-v) 'run (few;
four-legged animal), {(e)qwe:L} (DMS, 10sv) 'down the
hill, out of a tree, down a slope'
There are 16 of these in Maidu, all C, CV, or CVC. All have
clearly motional/locational meaning. They occur suffixed in verb
themes, and also combine with the neutral motional formant
{?y..'} to produce motion verb stems (see below).
These are directly comparable, both structurally and
semantically, with the Klamath dependent stems. The major
differences between the two classes are the larger individual
size, and much larger inventory, of Klamath DMS's. There is good
reason to suppose, however, that many of the Klamath DMS's are
etymologically composite, i.e. that they reflect lexicalizations
of earlier sequences of DMS's. As a particularly interesting
example, consider the following Klamath dependent stems:
{aba:ni} 'to the limit, end of a place'
{aba:tn} 'up against, to the shore, leaning
against'
{aba:yi'} 'diagonally against, up against'
All have a common element /ba:/, which might perhaps be assigned
the recurrent semantic element 'to, against'; this would then be
comparable to the Maidu MLA {paj} 'against' (with Klamath
{aba:yi'} directly comparable). This implies that some
Klamath DMS's are etymologically bimorphemic (see below); thus it
is worth noting that sequences of DMS's can occur also in
Maiduan, as /?yc'¢no-/, which is {?y} 'motion formant (see
below), 'go over (a mountain)' ({c'o} 'up over the edge', {no}
'along, etc.').
But the most important point of similarity between the Maidu
and Klamath locationals is the fact that a free stem can be
formed of a prefix and a dependent stem, as Konkow |b¢- do:n|
'throw (a ball, rock, etc.) up', consisting of ({bO} 'rock-like
instrument', {doJ} (LD) 'up, upward'). Indeed, there is one
prefix, {?y}, which is described as having no intrinsic semantic
content, but functioning only to allow the dependent stems to act
as motional. This is structurally (though apparently not
phonologically) comparable with the Klamath LP {gv},
with essentially the same function; compare Maidu /?yn¢/ 'go,
walk' ({no} 'along, general motion without implication of
direction or attitude') with Klamath /gen-/ 'go' ({gv}
'move of one's own volition', {en} 'action away, taking
away, going'.
The expansion of the dependent stem classes in Klamath
It is not clear what significance to attach for comparative
purposes to the structural parallels between Maiduan and Klamath.
On the one hand, Maiduan languages are not so geographically
isolated from our transmontane group as to preclude the
possibility of shared structural features reflecting marginal
participation in the areal development. In this connection we
should note the lack of any such parallels in the nieghboring
Wintuan languages. On the other hand, the parallels are indeed
striking, and the parallel between Maidu w¢o-kot-dau and
Klamath /wgatt'-/ (above) certainly suggests that the position of
the LP's may be common inheritance.
In any case, the claim that to the extent that Klamath is
more similar in this system to Atsugewi or Washo than to Maiduan
this represents common areal development amounts to the
hypothesis that Klamath was once much more like Maiduan in this
respect, and that its present state reflects a major expansion of
the two dependent stem classes. In this section I will present
some evidence for a relatively recent origin of a number of
members of each class, and in the process suggest some of the
routes by which the classes may have grown. We have already seen
that the synchronic line between independent and dependent
category is not firm; as Jacobsen notes for Washo, we have good
reason to believe that both the LP and the DMS categories are fed
by recruits from the independent stem class.
A few LP's are clearly related to independent nominal or
verbal roots, and can be assumed to represent secondary
assimilation of these roots into the dependent stem set:
{c'le} 'act upon a massive shapeless object
(such as a piece of meat)', cp. /c'ole:ks/ 'meat, flesh,
body'
{del} 'look'; assuming an original instrumental
sense 'with the face', cp. {del?n} 'face'
{sle} 'act upon a clothlike object' apparently <
*sla 'mat', cp. /slays/ 'mat', /slan-/ 'spread out a mat'
The semantics of several other LP's suggest a verbal origin:
{spon} 'lead someone, guide, take by the hand'
{twv} 'twist, bore'
{swe} 'look, look like'
{gv} 'go, move of one's own volition'
{pv} 'pull', {sbv}, {spi} 'drag'
Moreover, as with the dependent stems, there are two
morphemes whose behavior straddles the prefix and independent
position classes: {sdig} 'smell' and {nqen} 'shout', both of
which occur as independent verb stems, but also as prefixes in
construction with other members of the independent class. Here
the syntax confirms the inference from the semantics that these
must indeed be verbal rather than nominal in nature, and
presumably in origin.
A considerable number of DMS's manifest partial similarities
which in some cases must reflect etymological connections,
although the precise developments involved remain to be
elucidated. Good examples are:
{aptne:Gi} 'on top of a full load'
cp. {aba:tn} 'up against', {oyGi} 'up, above, over'
{k'acwe:g} 'in a tight place, corner; stuck' and {ak'akwe:g}
'stuck in a hole, tight place'
cp. {ak'c'wy} 'in a narrow place, corner ...', {iwy'G} 'into
a container, sack, receptacle'
{owedg} 'taking out of a socket'
cp. {odg} 'out of a container'
{akyamn} 'around, embracing, surrounding a central object'
cp. {oyamn} 'around, carrying or holding an object
aimlessly'
{oGamn} 'up, upstream'
cp. {amni} 'up, upstream, uphill'
In each case we can infer that the longer form reflects an
original sequence of DMS's; recall that such sequences are
attested both in Klamath and in Maiduan.
Barker (1964:146) notes the irregular allomorphy of {oyGi}
(Class 10 DMS) 'up, above, over', which includes an irregular
reduplicated intensive form which reflects an allomorph |oye:Gi|.
Two pairs of DMS's seem to be related by an accentual shift,
which probably reflects this same old intensifying construction:
{elwy} 'by the fire, along the edge, into water'
{eli:w} 'on the very edge'
{aLn} 'alongside'
{oli:n} 'off the edge, side, overboard, on the edge'
There are three DMS's which from their meaning and phonological
shape appear to represent such intensive forms, though they lack
corresponding non-intensives: {ote:g} 'deep into', {aba:ni} 'to
the limit, end of a place', {aqye:tn} 'right beside'.
As an intriguing final example, note the evident
relationship between the DMS {ksl'} (Class 10 DMS) 'on a fire'
and the LP's {k'lv} and {sl'v} (Class 4
LP), both 'act upon fire'. We cannot at present untangle the
history of this set, but it is clear that either the DMS or the
LP's must be a secondary development in that role; the most
likely hypothesis is that both are secondary specializations of
the same originally nominal or verbal material.
Notes
1) In the sense of the "thematic relations" approach to case
roles pioneered by Gruber (1976). The Theme is the argument in a
clause which is represented as being in or undergoing a change of
location or state.
2) A fuller areal perspective on the distribution of the
various features of this complex must take note of the fact that
the neighboring Numic languages show the most ambitious
development of the instrumental prefix class of any of the Uto-
Aztecan languages (Nichols 1974).
References
Aoki, Haruo. 1963. On Sahaptian-Klamath linguistic affiliations.
IJAL 28:172-82.
Barker, M.A.R. 1963. Klamath dictionary. UCPIL 31.
. 1964. Klamath grammar. UCPIL 32.
Dixon, Roland. 1911. Maidu. Handbook of American Indian
Languages, ed. by F. Boas, pp. 679-734. Washington: BAE.
Jacobsen, William. 1980. Washo bipartite verb stems. American
Indian and Indoeuropean studies: Papers in honor of Madison
S. Beeler, ed. by K. Klar et. al., 85-99.
Langdon, Margaret. 1987ms. Morphosyntax and problems of
reconstruction in Yuman and Hokan.
. 1988. More on stem structure in Hokan languages.
presented at the 1988 Hokan-Penutian Workshop.
Nichols, Michael. 1974. Northern Paiute historical grammar.
Ph.D. dissertation, University of California, Berkeley.
Pitkin, Harvey. 1984. Wintu grammar. UCPIL 94.
Shipley, William. 1963. Maidu texts and dictionary. UCPIL 33.
. 1963. Maidu grammar. UCPIL 41.
Talmy, Leonard. 1972. Semantic structures in Enlgish and
Atsugewi. Ph.D. dissertation, University of California,
Berkeley.
Ultan, Russell. 1967. Konkow grammar. Ph.D. dissertation,
University of California, Berkeley.