valency and grammatical functions in cherokee
TRANSCRIPT
-
7/28/2019 Valency and Grammatical Functions in Cherokee
1/24
Argument StructureFinal Paper
Hiroto Uchihara
1
Valency and Grammatical Functions in Cherokee
1. Introduction
Cherokee, an Iroquoian language, is a non-configurational, polysynthetic language with an active
case marking system (Dixon 1994: 70-83, Mithun 1991b). Such a language is a challenge to the
theory of argument structure, since it has no clear manifestations of valency or grammatical
functions: NPs are not obligatory within the clause, there is no case marking on NPs, and there are
no syntactic processes which relate to the argument structure of verbs. Thus, it is sometimes claimed
that this kind of languages lack the level of grammatical functions. The aim of this paper is to
address this question, examining the facts from Cherokee.
I will first discuss whether the notion of valency is relevant in Cherokee (section 2). Establishing
that the notion valency is valid in Cherokee, I will further argue that the notion of grammatical
functions is also necessary in Cherokee (section 3).
In order to avoid confusion, I will define the terms I will use in this paper here. I will use the
terms Agent/Patient just to refer to the morphological marking on the verbs: if the argument is
marked by the Agentive series of the pronominal prefixes, or marked as the first argument of the
Transitive series, it is called the Agent. If the argument is marked by the Patientive series of the
pronominal prefixes, or marked as the second argument of the Transitive series of the pronominal
prefixes, it is called the Patient. These categories are traditionally called A/B in the literatures on
-
7/28/2019 Valency and Grammatical Functions in Cherokee
2/24
Argument StructureFinal Paper
Hiroto Uchihara
2
Cherokee (cf. Scancarelli (1987), Montgomery-Anderson (2008), etc.), but I will use the terms
employed for Northern Iroquoian languages for convenience and to avoid confusion.
The termsAgentive (verb)/Patientive (verb) are used for the type of verbs in Cherokee: a verb is
calledAgentive if it takes the Agentive marking except for the Perfective aspect1
(in which it takes
the Patientive marking), and a verb is calledPatientive if the verb takes the Patientive marking in all
the aspects. These labels correspond toA verb/ B verb in Scancarelli (1987)).
Finally, the terms Actor/Undergoerare used to designate the semantic roles of the arguments of
the verbs. The assignment of these semantic roles is based on Van Valin & LaPolla (1992: 152).
2. Valency
2
In this section, I examine whether the notion of valency is relevant in Cherokee. First I will show
that many of the grammatical processes which are helpful for knowing the valency of verbs in other
languages are simply absent or irrelevant in Cherokee. Then, in 2.1 and 2.2, I will try to search for
any grammatical reflexes of valency of verbs in Cherokee.
Previous studies on Cherokee assume that there is a clear distinction between monadic3
1 Historically this aspect corresponds to the Stative aspect in Northern Iroquoian
languages.
2 I use the term valencyto indicate the number of arguments of verbs. I do not use the
term valence in order to avoid confusion with the same term used in HPSG (cf. 3.6.1).
-
7/28/2019 Valency and Grammatical Functions in Cherokee
3/24
Argument StructureFinal Paper
Hiroto Uchihara
3
(intransitive) and dyadic (transitive) verbs in Cherokee. For example, all the verbs in Feeling &
Pulte's (1975) dictionary are enlisted with the information whether the verb is intransitive or
transitive (see also Montgomery-Anderson (2008: 222)).
However, it is actually difficult when one tries to find an unambiguous way of knowing the
valency of a verb in Cherokee. Many of the grammatical categories or processes which help us to
know the valency of verbs in other languages are simply absent or irrelevant in Cherokee. First,
overt NPs are optional, so we cannot tell the number of arguments of the verb just by counting the
number of the overt NPs as in English. Secondly, case-marking on NPs, which may also be helpful
in determining the argument structure, is absent, since Cherokee is a head-marking language. Thirdly,
in languages such as English, only transitive, and not intransitive verbs, can undergo passivization,
and thus it will be a help for knowing the argument structures; but passivization is also absent in
Cherokee. Finally, in some languages, argument increasing processes such as causativization or
applicative alternation tell us the number of arguments of the original verb: for instance, in Japanese,
the causee bears the accusative case when the original verb is intransitive and the dative case when
the original verb is transitive (although there are some exceptions). This process is not helpful in
Cherokee, either, since causee of the causativized verb is always coded as the 'second' argument of
3 I use the term monadicand dyadicto refer to the verbs with single argument and two
arguments, respectively. The reason to use these terms, in stead ofintransitiveand
transitive, is to focus purely on the number of the arguments of verbs.
-
7/28/2019 Valency and Grammatical Functions in Cherokee
4/24
-
7/28/2019 Valency and Grammatical Functions in Cherokee
5/24
-
7/28/2019 Valency and Grammatical Functions in Cherokee
6/24
Argument StructureFinal Paper
Hiroto Uchihara
6
"He(Agt) sees him(P)." "He(Agt) sees (something unspecified)."
This reflexive affix relates to the orsemantic orsyntactic (if at all) valency, rather than the
morphologicalvalency. For example, the verb in (2) is taking the 3.sg.>3.sg.A pronominal prefix,
a-, which is morphologically homophonous with the monadic 3.sg.A prefix, and undergoes
pronominal prefix alternation; thus it could be considered to be morphologically intransitive (Koenig
& Michelson 2009: 9ff.). Nevertheless, this verb can take the reflexive affix, unlike semantically
monadic verbs.
This reflexive affix also defocuses the 'second' argument of the dyadic Patientive verbs:
(3) Patientive, Dyadic
(3a) u-geeyh-a (3b) u-[a]daa-geeyh-a
3.sg.>3.sg.P-love:PRS-IND 3.sg.P-MID-love:PRS-IND
"He(P) loves her." "He(P) loves (someone unspecified)"
This may show that semantically dyadic Patientive verbs cannot be considered syntactically the
same as semantically monadic Patientive verbs, as have been considered in the previous studies.
Object defocusing construction seems to be a promising test for knowing the argument structure
of verbs in Cherokee. However, productivity of this construction is still not well-known, and further
study is needed to support this claim.
2.3. Summary of section 2
-
7/28/2019 Valency and Grammatical Functions in Cherokee
7/24
Argument StructureFinal Paper
Hiroto Uchihara
7
In this section, we have seen that the notion of valency should be relevant in Cherokee, however
subtle its manifestation may be; it is not that Cherokee does not have the valency distinction, or that
Cherokee has less clear distinction, but just that there are fewer grammatical reflections of this
category in this language.
I will briefly mention the treatment of the 3rd
argument of the verbs in Cherokee (or in Iroquoian
in general). In line with Koenig & Michelson (2009: 17-18), I assume that the 3rd argument of triadic
verbs are not morphologically encoded on the verbs (as opposed to Baker 1996, who claims that
incorporated nouns can morphologically mark morphosyntactic arguments); this is more evident in
Cherokee than in Northern Iroquoian languages since Cherokee is lacking in the productive noun
incorporation process.
3. Grammatical Functions
In section 2, I argued that the notion of valency is relevant in Cherokee, however unclear the
grammatical reflection is. The next question is, whether there is an asymmetry among the arguments
of the verbs in Cherokee, so that the notion of grammatical functions (subject and object) is also
relevant in this language.
It has often been claimed that there is no level of grammatical functions (subject, object, etc.) in
Iroquoian (Mithun 1991a, 2006, Chafe 1994: 150), or in languages with active/ inactive system in
-
7/28/2019 Valency and Grammatical Functions in Cherokee
8/24
Argument StructureFinal Paper
Hiroto Uchihara
8
general (Van Valin & LaPolla 1997: 250ff). However, these claims are often made based on a fragile
ground: they are applying the syntactic tests invented based on the observations on other languages
(such as those in Keenan 1976), which are actually inapplicable or irrelevant to Iroquoian languages.
For example, Mithun (1991a) takes up five criteria frequently employed for determining the
subject and object of a language, and examines the case of Cayuga, a Northern Iroquoian language:
(i) deletion of subject NP in commands; (ii) the omission of coreferent nominals in coordination; (iii)
whether the subject becomes possessor in nominalized clauses; (iv) equivalent NP deletion and
raising in complex constructions; and (v) relativization. Mithun concludes that the notion of subject
does not exist in this language because these criteria do not apply. Similar discussion is found in
Chafe (1994: 150) or Beghelli (1996: 106).
Scancarelli (1987: 118-123), on the other hand, argues that in Cherokee, "subjects and objects
can be said to have syntactic status inasmuch as they are relevant to agreement morphology and
inasmuch as they are not direct representations of semantic roles (ibid.: 118)" (cf. 3.1). However, she
also remarks that there are no "syntactic means of discriminating between transitive subjects and
objects, independent of verb agreement (ibid.: 119)", since subject identification criteria used
crosslinguistically, i.e. (i) word order, (iii) universal quantifier, (iii) position of the adverbs, and (iv)
coreference, do not apply to Cherokee.
These previous studies are just showing that the crosslinguistic subject identification criteria do
-
7/28/2019 Valency and Grammatical Functions in Cherokee
9/24
Argument StructureFinal Paper
Hiroto Uchihara
9
not apply to these languages, rather than that there is no notion of grammatical functions (or that of
subject) in these languages. To see whether there actually exists the notion of grammatical functions,
it is necessary to come up with language-internal criteria.
If we define grammatical functions (subject/ object) as those which are identified through the
Keenan-style subjectivity tests, it is true that these languages lack them, but this is obviously not
what we are looking for. Thus, I tentatively define 'subject' and 'object' as follows: if we can find
grammatical phenomena which target the 'first' arguments of dyadic verbs together with the sole
argument of monadic verbs (the left circles in Figures 1 and 2), or those which target only the
'second' arguments of dyadic verbs (the right circles in Figure 1 and 2), irrespective of the
morphological marking (Figure 1) or the semantic roles of these arguments (Figure 2), we tentatively
call them 'subject' and 'object', respectively:
Figure 1: Level of Morphological Marking Figure 2: Level of Semantic Roles
(Agt=Agent, P=Patient) (A=ACTOR, U=UNDERGOER)
It is true that we cannot immediately equate these notions with 'subject' and 'object' in other
1st
2nd
Dyadic Agt P
P ()
Monadic Agt
P
1st
2nd
Dyadic A U
Monadic A
U
-
7/28/2019 Valency and Grammatical Functions in Cherokee
10/24
Argument StructureFinal Paper
Hiroto Uchihara
10
languages.But at the same time, it is also true that intuitively we are tempted to call these arguments
'subject' and 'object'. We will consider what level exactly it is later in this section.
3.1. Pronominal Prefixes
As we have already seen, Cherokee (transitive) pronominal prefixes encode the Agent and Patient
arguments in a fusional manner. As Scancarelli (1987) points out, "the identification of subject and
object fully determines verb agreement morphology for transitive verbs with first or second person
object (regardless of subject), and for transitive verbs with third person animate objects with first or
second person (or unspecified) subjects (ibid.: 118)". In (4), (a) is an Agentive verb and (b) is a
Patientive verb, but the same pronominal prefix is employed to indicate the 1
st
person singular
'subject' and the 2nd
person singular 'object':
(4) patient = 1st/2
ndperson
(4a) gvv-gowhth-a (4b) gvv-geeyh-a
1.sg.>2.sg.-see:PRS-IND 1.sg.>2.sg.-love:PRS-IND
"I(Agt) am seeing you(P)." "I(Agt) love you(P)."
The fact that (b) is a Patientive verb is evident from the cases where both of the arguments are 3rd
person:
(4') patient 1st/2
ndperson
(4'a) a-gowhth-a (4'b) u-geeyh-a
3.sg.>3.sg.A-see:PRS-IND 3.sg.>3.sg.P-love:PRS-IND
-
7/28/2019 Valency and Grammatical Functions in Cherokee
11/24
Argument StructureFinal Paper
Hiroto Uchihara
11
"He(Agt) sees her(P)." "He(P) loves her."
Since this is applicable only to dyadic verbs, this observation can be explained in terms of the
semantic roles as well: we could state that when the conditions stated above are met, the pronominal
prefixes mark Actor and Undergoer, rather than 'subject' and 'object'. Therefore, this might not be the
best argumentation for showing the existence of grammatical functions in Cherokee, but this fact
does not contradict the existence of grammatical functions, either.
3.2. Nominalization of Imperfective stems.
As Scancarelli (1987: 123) notes, agentive nouns based on the Imperfective stems of verbs with
the nominalizer suffix -i always refer to the 'subject' entity, irrespective of the morphological
marking (Agent/ Patient) or the semantic roles (Actor/ Undergoer) of the arguments of the verb.
(5) and (6) are examples of Agentive verbs. (5) is a monadic verb, and (6) is a dyadic verb. Forms
in (a) represent the Imperfective forms of the verbs, and those in (b) are the agentive nouns based on
the verbs in (a):
(5) Agentive, Monadic
(5a) ga-wniisg-vvi (5b) ga-woniisg-i
3.sg.A-speak:IMPF-ASS 3.sg.A-speak:IMPF-NOM
"He(Agt) was speaking" "speaker"
(6) Agentive, Dyadic
(6a) ga-noosgsg-vvi (6b) ga-noosgii
sg-i
-
7/28/2019 Valency and Grammatical Functions in Cherokee
12/24
Argument StructureFinal Paper
Hiroto Uchihara
12
3.sg.>3.sg.A-steal:IMPF-ASS 3.sg.>3.sg.A-steal:IMPF-NOM
"He(Agt) was stealing it(P)." "stealer"
In the examples above, the agentive nouns in (b) refer to the entity who 'speaks' and 'steals'. From
these cases alone, we are not sure whether the agentive nouns above refer to the Agent argument or
the 'subject' argument of the verbs. If we look at Patientive verbs, however, the picture becomes
clearer ((7) is a monadic and (8) is a dyadic verb):
(7) Patientive, Monadic
(7a) u-hdlvvg-vvi (7b) uu-hdlvvg-i
3.sg.P-be.sick:IMPF-ASS 3.sg.P-be.sick:IMPF-NOM
"He(P) was sick." "patient"
(8) Patientive, Dyadic6
(8a) u-[a]dg-vvi (8b) uu-[a]deeg-i
3.sg.>3.sg.P-throw:IMPF-ASS 3.sg.>3.sg.P-throw:IMPF-NOM
"He(P) was throwing it." "pitcher"
In these cases, the agentive nouns refer to the Patient arguments, i.e. the one who is 'sick', and the
one who 'throws'. Table 1 summarize the situation observed above:
Table 1
marking valency Argument referred to by the agentive noun
(5) Agentive Monadic Agent/Actor/'subject'
(6) Agentive Dyadic Agent/Actor/'subject'
(7) Patientive Monadic Patient/Undergoer/'subject'
6 It would be better to have an example of a prototypical Patientive verb, where the first
argument of the verb is less agentive. I was not able to find such an example in my data,
but if I can elicit such a data, I would expect it to behave in the same way.
-
7/28/2019 Valency and Grammatical Functions in Cherokee
13/24
Argument StructureFinal Paper
Hiroto Uchihara
13
(8) Patientive Dyadic Patient/Actor/'subject'
In order to make a generalization as to which argument of the verb the agentive nouns refer to, the
level of morphology (Agent/ Patient) or the level of semantic roles (Actor/ Undergoer) are not
sufficient. If we take the level of morphology, for instance, the generalization would be something
like "agentive nouns based on Agentive verbs refer to the Agent arguments, while those based on
Patientive verbs refer to the Patient arguments". If we introduce the level of grammatical functions,
we can simply state that the agentive nouns refer to the 'subject' of the verbs.
One might suggest that the crucial point is animacy here: it is the animate argument which the
resulting agentive noun refers to. This may explain many of the cases, but it does not explain the
cases when both of the arguments of a dyadic verbs are animate (e.g. -h- 'kill'): the agentive noun
still refers to the 'subject' argument (i.e. killer), and not to the one who was killed.
3.3. Reflexive affix adaa(d)-
As we have seen in 2.2, reflexive affix adaad(d)- is added to the verb bases to defocus the 'object'
of dyadic verbs (Montgomery-Anderson 2008: 365). This reflexive affix always defocuses the
'object' of the verbs, irrespective of whether the verb is morphologically Agentive or Patientive. In
the case of an Agentive verb (9), what is defocused in (9b) is the one who is 'seen', which is
morphologically Patient:
-
7/28/2019 Valency and Grammatical Functions in Cherokee
14/24
Argument StructureFinal Paper
Hiroto Uchihara
14
(9) Agentive, Dyadic
(9a) a-gowhth-a (9b) a-[a]daa-gowhth-a3.sg.>3.sg.P-see:PRS-IND 3.sg.A-MID-see:PRS-IND
"He(Agt) sees him(P)." "He(Agt) sees (something unspecified)."
In the case of a Patientive verb (10), what is defocused in the form with the reflexive affix (10b) is
the one who is 'loved', which is morphologically not marked:
(10) Patientive, Dyadic
(10a) u-geeyh-a (10b) u-[a]daa-geeyh-a
3.sg.>3.sg.P-love:PRS-IND 3.sg.P-MID-love:PRS-IND
"He(P) loves her." "He(P) loves (someone unspecified)"
Table 2 summarizes the observation above:
Table 2
marking Defocused argument
(9) Agentive Patient/Undergoer/'object'
(10) Patientive (no marking)/Undergoer/'object'
In order to make a generalization as to which argument of the verb is defocused, neither the level
of morphology nor the level of semantic roles work (we cannot generalize that the reflexive affix
defocuses the Undergoer arguments, since Undergoer arguments of monadic verbs are not defocused
by this affix). By introducing the level of grammatical functions, we can generalize that the reflexive
affix defocuses the 'object' arguments.
3.4. Causativization
-
7/28/2019 Valency and Grammatical Functions in Cherokee
15/24
Argument StructureFinal Paper
Hiroto Uchihara
15
In Cherokee, the causative suffix attaches to the verb base to form causative verbs. When verbs
are causativized, the 'subject' arguments of the original verbs become the 'causee', irrespective of the
valency or the morphological marking. The 'object' arguments of the original dyadic verbs will not
be marked on the pronominal prefix of the causative verb7
(this is because the pronominal prefixes
in Iroquoian languages can mark up to only two arguments at most).
(11) and (12) are examples of Agentive verbs. (11) is a monadic verb and (12) is a dyadic verb. (a)
forms show the original forms and (b) forms show the causative forms of the verbs:
(11) Agentive, Monadic
(11a) a-[a]leehwisdh-a (11b) a-[a]leehwisd-hdh-a
3.sg.A-stop:PRS-IND 3.sg.>3.sg.A-stop-CAUS:PRS-IND
"He(Agt) is stopping." "He(Agt) is stopping him(P)."
(12) Agentive, Dyadic
(12a) ji-gowhtih-a (12b) jii-gowht-vhdh-a
1.sg..>3.sg.inan.A-see:PRS-IND 1.sg.>3.sg.an.-see-CAUS:PRS-IND
"I(Agt) am seeing it(P)." "I(Agt) am showing to him(P)."
In both cases, the original 'subject' arguments become the 'causee'. In the case of the dyadic verb
(12), the original object is no longer marked in the causativized form in (12b). This is clear from the
fact that the morphological marking in (12b) for the 'object' has changed from inanimate (inan.) in
7 This situation is quite different from the typological tendency to preserve the case
marking of original O (Comrie 1981: 167-177). This may be because Cherokee is a
primary/ secondary object language (Dryer 1987).
-
7/28/2019 Valency and Grammatical Functions in Cherokee
16/24
Argument StructureFinal Paper
Hiroto Uchihara
16
(12a) to animate (an.).
Next, let us see the situation for a Patientive verb (13):
(13) Patientive, Monadic
(13a) u-hnalvvg-a (13b) ga-hnalvv-sdih-a
3.sg.P-become.angry:PRS-IND 3.sg.>3.sg.A-become.angry-CAUS:PRS-IND
"He(P) is becoming angry." "He(Agt) is making him(P) angry."
Here, we see that the original 'subject' in (13a), which is morphologically the Patient, becomes the
'causee' in the causativized verb (13b).
Table 3 summarizes the observation above, and again shows that the level of morphological
marking (Agent/ Patient) or that of semantic roles (Actor/ Undergoer) are not sufficient to state
which argument of the verb becomes the causee of the causativized verb.
Table 3
marking valency Causee
(11) Agentive monadic Agent/Actor/'subject'
(12) Agentive dyadic Agent/Actor/'subject'
(13) Patientive monadic Patient/Undergoer/'subject'
As in the case of Agentive Nouns (3.1), one might suggest what is crucial here is the animacy of
the argument: it is the animate argument in the original verb which becomes the causee with
causativization. I do not have any example of a dyadic verb with two animate arguments which
undergo causativization, so I cannot give a counterexample to this position, but my prediction is that
what is at issue here is not the animacy.
Argumentation based on causativization might be problematic, since causativization in Cherokee
-
7/28/2019 Valency and Grammatical Functions in Cherokee
17/24
Argument StructureFinal Paper
Hiroto Uchihara
17
(and Iroquoian in general) is rather lexical with some idiosyncrasies, rather than a syntactic
productive process. Moreover, the same affix is employed for introducing an instrumental argument
to the verb, which may behave differently in terms of argument realization.
3.5. Benefactive
Cherokee verbs can add a beneficiary argument by attaching the Dative suffix to the verb. In the
case of monadic verbs (cf. 14), the valency increases by one argument. In the case of dyadic verbs,
one of the original arguments of the verb is deprived of the argument status and the beneficiary
argument becomes the new argument of the benefactive verb (15, 16). Here, we will focus on the
argument which is not affected by this benefactivization.
(14) is a case of an Agentive monadic verb. Here, the Dative suffix introduces the Patient
argument (a person for whom the act of speaking is conducted), and the original Agent argument
(speaker) remains the Agent argument in (b):
(14) Agentive, Monadic
(14a) ga-wnih-a (14b) ga-wniihis-h-a
3.sg.A-speak:PRS-IND 3.sg.>3.sg.A-speak-DAT:PRS-IND
"He(Agt) is speaking." "He(Agt) is speaking for him(P)."
Next, (15) is a case of an Agentive dyadic verb:
(15) Agentive, Dyadic
-
7/28/2019 Valency and Grammatical Functions in Cherokee
18/24
Argument StructureFinal Paper
Hiroto Uchihara
18
(15a) g-agwiyh-a (15b) jiiy-agwiyvv-eeh-a
1.sg.>3.sg.inan.A-pay:PRS-IND 1.sg.>3.sg.an.A-pay-DAT:PRS-IND
"I(Agt) am paying it(P)." "I(Agt) am paying him(P)."
Here, the verb replaces the theme argument (the thing which is paid, i.e. "money") with the
beneficiary argument (the one to whom the payment is made) by attaching the Dative suffix. This is
evident from the fact that in (15a) the pronominal prefix shows that the Patient argument is
inanimate (g-), while in (15b) it shows that the Patient argument is animate (jiiy-). The Agent
argument is not affected by benefactivization and remains as the 'payer'.
So far, it is not clear whether the argument which is not affected by the benefactivization is the
Agent argument or the 'subject'. If we look at the Patientive verbs, the situation becomes clearer:
(16) Patientive, Dyadic
(16a) u-hwasg-a (16b) a-hwahs-eh-a
3.sg.>3.sg.P-buy:PRS-IND 3.sg.>3.sg.A8-buy-DAT:PRS-IND
"He(P) is buying it." "He(Agt) is buying from him(P)."
Here, the benefactive verb replaces the original theme argument (the thing which is bought) with
the source argument (the one who he is buying from), while the original Patient argument ('buyer') is
not affected by the benefactivization and remains the 'buyer'. The observation above can be
summarized as below:
Table 4
marking valency unaffected argument affected argument
8
The pronominal prefix changes from Agent to Patient series with benefactivization.
-
7/28/2019 Valency and Grammatical Functions in Cherokee
19/24
Argument StructureFinal Paper
Hiroto Uchihara
19
(14) Agentive Monadic Agent/Actor/'subject'
(15) Agentive Dyadic Agent/Actor/'subject' Patient/Undergoer/'object'
(16) Patientive Dyadic Patient/Actor/'subject' ()/Undergoer/'object'
In order to specify the argument which is not affected by benefactivization, we should refer to the
level of 'subject', not to the morphological level of Agent/ Patient argument. Since I cannot find an
example of Patient monadic verb in my data, we cannot convincingly conclude that it cannot be
stated in terms of the level of semantic role (Actor/Undergoer), either, but I expect them to behave in
the same way as other verbs.
3.6. Summary for section 3
I have presented five grammatical phenomena which require the level independent either from
that of semantic roles or morphological marking, and intuitively, we are tempted to call the
arguments in that level 'subject' and 'object'. However, as we have noted in the introduction to this
section, we are not still sure whether we can equate this level with the level of grammatical functions
which are identified through subjectivity tests frequently used in other languages (cf. Keenan 1976).
It is obvious that defining grammatical functions through those syntactic subjectivity tests is not
what we want (Dixon 1994: 127-130,Langacker 2008: 364-365). It is thus necessary to define the
notion of grammatical functions independent from Keenan-style subjectivity tests. To attempt to
achieve this purpose, I will look at the possible treatments of these notions in two current linguistic
-
7/28/2019 Valency and Grammatical Functions in Cherokee
20/24
Argument StructureFinal Paper
Hiroto Uchihara
20
frameworks, HPSG (3.6.1) and Cognitive Grammar (3.6.2).
3.6.1. HPSG
One possibility is that the 'subject' and 'object' in Cherokee that we have been discussing so far in
this section might correspond to the 'first' and 'second' arguments in the level of
ARGUMENT-STRUCTURE in the HPSG framework (Sag et al. 2003: 205, Koenig & Michelson
2009: 4ff.), not to the 'subject' and 'complement' in the level of VALENCE (Sag et al. 2003: 62)9.
If this is the case, my observations above are in line with Koenig & Michelson's (2009: 8, 24)
claim, namely that what is invariant across languages like Oneida, a Northern Iroquoian language,
and those like English, is the level of ARGUMENT-STRUCTURE, not the level of VALENCE (their
Argument Structure Hypothesis (ibid.: 8)): Iroquoian and English share the asymmetry in the level of
ARGUMENT-STRUCTURE, but not in the level of VALENCE, since the former simply lacks the
level of VALENCE.
Moreover, if we assume that the grammatical functions identified through Keenan-style
subjectivity tests refer to the level of VALENCE, we can explain the inapplicability of these tests to
Iroquoian languages: these tests are not applicable since these languages simply lack this level.
9
This suggestion was first made by Dr. Karin Michelson.
-
7/28/2019 Valency and Grammatical Functions in Cherokee
21/24
Argument StructureFinal Paper
Hiroto Uchihara
21
3.6.2. Cognitive Grammar
Another way to deal with this issue may be a semantic-based approach, such as Dixon (1994:
Ch.5) or Langacker (2008: 363ff.). Here, I will consider the possibility that 'subject' and 'object' in
Cherokee discussed so far might correspond to 'trajector' and 'landmark' in Langacker's (2008:
363ff.) Cognitive Grammar.
Langacker is also unsatisfied with the trend to identify the subject/ object as those which are
determined through Keenan-style subjectivity tests. To quote his remarks: "the grammatical behavior
used to identify subject and object do not serve to characterize these notions but are merely
symptomatic of their conceptual import (Langacker 2008: 364)", or "it is simply presupposed that
subjects are identified by the sorts of grammatical behaviour typical of subjects in familiar European
languages (ibid.: 365)".
Langacker, instead, claims that the subject and object relations "are grammatical manifestations
of trajector/landmark alignment" (ibid.: 365), the subject being "a nominal that codes the trajector of
a profiled relationship", and the object being "one that codes the landmark" (Langacker defines the
'trajector' and 'landmark' as the primary and secondary focal prominence in a profiled relationship
(ibid.: 70ff, 365)).
Langacker's definition of 'subject' and 'object' is independent of the Keenan-style subjectivity
tests, and may be the closest to the notions we have seen in this paper. However, it is hard to know
-
7/28/2019 Valency and Grammatical Functions in Cherokee
22/24
Argument StructureFinal Paper
Hiroto Uchihara
22
whether it is exactly the same as what Langacker has in mind, partly because of the somewhat
subjective (?) nature of Cognitive Grammar. Moreover, based on his analysis on Seneca, another
Iroquoian language, he seems to be thinking that the morphological Agent is his 'subject' in
Iroquoian (ibid.: 381-382)10
, which is a quite different view from ours.
4. Conclusion
As we have seen so far, equivalents of valency or grammatical functions actually do seem to be
found in Cherokee, however subtle their grammatical manifestations are. This is to the contrary to
the claims made by previous studies.
The Cherokee data examined above warns us two following points. First, we have seen that
inapplicability or irrelevance of some syntactic tests have often been mistaken as non-existence of
those categories in the language by some linguists: if one defines some grammatical categories in
terms of whether the language passes crosslinguistic syntactic tests, - and obviously that is not what
we want-, one cannot discuss the relevance of that category in a language where those syntactic tests
are not applicable for an independent reason. To discuss the existence of a grammatical category, we
should not totally rely on those crosslinguistic syntactic tests, but should look for language-internal
10 He considers the absolutive arguments as the 'subjects' in ergative languages
(Langacker 2008: 374-375).
-
7/28/2019 Valency and Grammatical Functions in Cherokee
23/24
Argument StructureFinal Paper
Hiroto Uchihara
23
evidence which is independent from those syntactic tests.
Secondly, what we learn from the facts about Cherokee is that subtle grammatical manifestations
do not necessarily mean non-existence of that category in that language: although Cherokee shows
slightest grammatical reflexes of valency or grammatical relations, these categories do exist, and
without these categories, description of Cherokee grammar is impossible.
A further problem concerns whether the 'valency' or 'grammatical functions' we have seen above
are actually exactly the same category as in other languages where those crosslinguistic syntactic
tests are applicable. I have tentatively suggested that 'valency' and 'grammatical functions' discussed
above belong to the level of ARGUMENT-STRUCTURE, and not VALENCE, in the HPSG
framework. Further descriptive and theoretical study is needed to confirm this hypothesis.
[REFERENCES]
Baker, Mark. 1991. On some subject/ object non-asymmetries in Mohawk. Natural Language and
Linguistic Theory 9: 537-76.
Baker, M.C. 1996.The polysynthesis parameter. New York: Oxford University Press.
Beghelli, Filippo. 1996. Cherokee Clause Structure. In Pamela Munro ed., Cherokee Papers from
UCLA. Los Angeles: University of California. 105-114.
Chafe, Wallace. 1994.Discourse, consciousness, and time: the flow and displacement of conscious
experience in speaking and writing. Chicago: Chicago University Press.
Comrie, Bernard. 1981.Language Universals and Linguistic Typology. U. of Chicago Press.
Dixon, R.M.W. 1994.Ergativity. Cambridge: Cambridge University Press.
Dryer, Matthew. 1987. On primary objects, secondary objects and antidatives. Language 62:
808-845.
Feeling, Durbin. 1975. Cherokee-English dictionary. William Pulte ed. Tahlequah, OK: Cherokee
Nation of Oklahoma.
-
7/28/2019 Valency and Grammatical Functions in Cherokee
24/24
Argument StructureFinal Paper
Hiroto Uchihara
24
Keenan, Edward. 1976. Towards a universal definition of subject. In Charles Li ed. Subject &
Topic. New York: Academic Press.
Koenig, J.P. & Karin Michelson. 2009ms. Invariance in Argument Realization: the case ofIroquoian.
Langacker, Ronald. 2008. Cognitive Grammar: A Basic Introduction. Oxford University Press.
Levin, Beth and Malka Rappaport Hovav. 2005.Argument Realization. CUP.
Mithun, Marianne. 1991a. The Role of Motivation in the Emergence of Grammatical Categories: the
Grammaticization of Subjects. In Heine & Traugott eds. Approaches to grammaticalization.
Amsterdam: Benjamins 2: 159-84.
Mithun, Marianne. 1991b. Active/ Agentive case marking and its motivations. Language 60(3):
510-546.
Mithun, Marianne.2006. Voice without Subjects, Objects, or Obliques: Manipulating argument
structure in agent/ patient systems (Mohawk). In Tsunoda Tasaku & Kageyama Taro eds., Voice
and Grammatical Relations: In honor of Masayoshi Shibatani. Amsterdam: John Benjiamins.
195-218.
Montgomery-Anderson, Brad. 2008.A reference grammar of Oklahoma Cherokee. Ph.D. diss., U. of
Kansas.
Pulte, William & Durbin Feeling. 1975. Outline of Cherokee Grammar. in Durbin Feeling. 1975.
Cherokee-English dictionary. William Pulte ed. Tahlequah, OK: Cherokee Nation of Oklahoma.
235-355.
Sag, Ivan, Thomas Wasow, and Emily Bender. 2003. Syntactic Theory: A Formal Introduction, 2nd
Edition. Center for the Study of Language and Information - Lecture Notes
Scancarelli, Janine. 1987. Grammatical relations and verb agreement in Cherokee. Ph.D. diss., U. of
California. Los Angeles.
Van Valin, Robert & Randy LaPolla. 1997. Syntax. Cambridge: Cambridge University Press.