Accounting for Verbal Argument Indexing in Cherokee * Susan Smythe University of Texas at Austin 1. Introduction Cherokee is a polysynthetic, head-marking language; it has no fixed word order for major clausal constituents; and the nominals are completely unmarked for case. Thus, some method of indexing verbal arguments on the head must exist. However, the problem of accounting for verbal argument indexing in Cherokee has not been satisfactorily resolved in the existing literature (Cook 1979; Foley 1975; King 1975; Pulte and Feeling 1975; Scancarelli 1987; Walker 1975). Cherokee allows the occurrence of one pronominal prefix on the verb. However, each pronominal prefix indexes either one or, at most, two arguments, as is the case in most of the Iroquoian languages (see Lounsbury 1978, among others). When the verb is intransitive, its single argument is indexed by the pro1 nominal prefix, as in (1a). When the verb is transitive and both arguments are either first or second person, they are both indexed by the single prefix, as in (1b). However, if the subject is third person or the object is third person and inanimate, only one of the arguments is indexed by the prefix, as in (1c). And, finally, when the verb is ditransitive, the prefix can index only two of the three arguments, as in (1d). (1) a. b. u:ni:lu:ch skwi:st u:ni:-lu:c-h 3PL.B-arrive-PERF many ‘Many people arrived.’ k:kowhtiha k:- kowht-iha 1SG/2SG-see-PRES ‘I see you.’ y:wi people (Scancarelli 1987:176) 2 (Pulte and Feeling 1975:268) * I would like to thank Lisa Green for her guidance and support, Craig Kopris for always popping up with a reference just when I needed it, and the SCIL 11 audience at the University of Texas at Austin for their comments. Any misinterpretations of the data are solely mine. 1 2 See section 2.2.1 for discussion of the A- and B-class single-argument prefixes. Most of my example sentences are from Pulte and Feeling 1975 or Scancarelli 1987. In many instances, I have had to provide either a morpheme breakdown and/or gloss. In some instances, I have simply had to add more explicit information to their morpheme breakdowns and glosses. MIT Working Papers in Linguistics 45, 217−231 Proceedings of the 11th Student Conference in Linguistics, © 2003 Susan Smythe Susan Smythe c. d. hiko:whtiha hi-ko:wht-iha 2SG.A-see-PRES ‘You see it.’ (Pulte and Feeling 1975:268) ani:skaya kaci:ya:kha:ne ki:hli ani:-skaya kaci:y-kha:ne-e:-ha 3PL-man 1SG/3PL.ANI-give-DAT-PRES dog ‘I’m giving the men a dog.’ (Scancarelli 1987:69) The question that I raise here is: how does the morphosyntax index a second argument in cases like (1c) and a third argument in cases such as (1d)? For cases like (1c), I put forth the zero morpheme analysis that has been utilized 3 in numerous other analyses of Native American languages. In doing so, I will argue against Scancarelli’s (1987) analysis of Cherokee, in which she explicitly states that “there is no zero anaphora in Cherokee” (p. 123). For cases like (1d), I will utilize aspects of the Polysynthesis Parameter as laid out in Baker 1996. After discussing Baker’s (1996) analysis of Mohawk in section 2.1, I present the different methods of indexing arguments on Cherokee verbs and my new analysis of pronominal prefixing in section 2.2. In section 2.3, I examine two other potential methods of indexing: the distributive prepronominal prefix and noun incorporation. The conclusion is given in section 3. 2. Analysis: Accounting for Argument Indexing on the Verb 2.1. Review of Baker 1996 In section 2.1.1, I briefly discuss the aspects of Baker’s (1996) Polysynthesis Parameter that are relevant to my analysis of Cherokee. Based on Baker’s analysis of the Mohawk data and the available literature on Cherokee (Cook 1979; Foley 1975; King 1975; Pulte and Feeling 1975; Scancarelli 1987; Walker 1975), I note some similarities and differences between the two languages in section 2.1.2. 2.1.1 The Polysynthesis Parameter The goal of Baker’s 1996 Polysynthesis Parameter is two-fold: to account for polysynthetic languages using the Principles and Parameters (P&P) framework and, thus, to fit polysynthetic languages into the mold of Universal Grammar. To achieve these goals, Baker introduces the Polysynthesis Parameter, which he also calls the Morphological Visibility Condition (MVC). The MVC is given below in (2). 3 See, e.g., Baker 1996 for Mohawk; Dayley 1985 for Tzutujil Maya; MacKay 1999 for Misantla Totonac; Merlan 1976 for Huauhtla Nahuatl; to name just a few. 218 Argument Indexing in Cherokee (2) The Morphological Visibility Condition (MVC) A phrase X is visible for theta role assignment from a head Y only if it is coindexed with a morpheme in the word containing Y via: (i) an agreement relationship, or (ii) a movement relationship (p. 17) A by-product of Baker’s interest in proving that the P&P framework can be used to analyze polysynthetic languages is the theoretical position that polysynthetic languages are configurational after all. Though he takes the position that overt nominals occur in adjunct positions rather than argument positions, he does not maintain that pronominal morphemes are arguments. Rather he argues that pro and movement traces appear in argument positions and are coindexed with pronominal morphology on Mohawk verbs. My purpose in this paper is not to argue for or against Baker’s Polysynthesis Parameter or to debate the configurationality of polysynthetic languages. I am less interested in his theoretical framework and more interested in his fundamental analysis of how arguments, whether phonologically null or overt, are indexed on the verb. Baker explicitly states that “the key idea of the Polysynthesis Parameter is that agreement morphemes and incorporated noun roots are part of the same system” (p. 19), the former via (i) and the latter via (ii) of the MVC. In order to make sure that the MVC applies to all Mohawk clauses, Baker posits the existence of a null or zero morpheme that is coindexed with all third person neuter arguments. Thus, he argues that a zero morpheme indexes third person neuter subjects (3a), objects (3b), and incorporated noun roots (3c). (3) a. Tsi yo-knór-u wa'-(Ø)wak-óhare-'s-e' akwatyá'tawi that NsO-rain-STAT FACT-(NsS)/1sO-body-wash-BEN-PUNC my-shirt ‘The rain washed my shirt for me.’ b. Ra(Ø)-núhwe'-s MsS/(NpO)-like-HAB ‘He likes cars.’ c. ne NE (Baker 1996:194) 4 ká'sere' car t-a-híy-(Ø)u-' CIS-FACT-1sA/MsO-(Ns).give-PUNC ‘I gave it to him.’ (Baker 1996:21) (Baker 1996:204) 2.1.2 Similarities and differences between Mohawk and Cherokee Mohawk and Cherokee are both members of the Iroquoian language family. All verbs in both Mohawk and Cherokee are marked by a single pronominal prefix to show the person, number, and gender features of both the subject and the ob4 I have inserted the Ø’s into these examples. 219 Susan Smythe ject of the clause (Baker 1996; Cook 1979; King 1975; Pulte and Feeling 1975; Scancarelli 1987). However, while Mohawk’s gender system is made up of masculine, feminine, zoic, and neuter (Baker 1996), Cherokee’s gender system is comprised of human, non-human animate, and inanimate (Scancarelli 1987). According to Lounsbury (1978), Proto-Iroquoian also had a system of pronominal prefixing, and, according to Cook (1979), the Proto-Iroquoian gender system had four genders, like that of Mohawk. Furthermore, both Cherokee and Mohawk have portmanteau transitive prefixes that index two arguments, but that can no longer be broken down synchronically into separate morphemes (Baker 1996; Pulte and Feeling 1975). The pronominal prefixing systems in both languages are further complicated by the complex morphophonemic changes that occur when various morphemes are combined with the verb root and with each other (Baker 1996; Cook 1979; King 1975; Pulte and Feeling 1975; Scancarelli 1987). Finally, many transitive verbs in both languages can be made ditransitive by the inclusion of a dative or benefactive suffix. In both Cherokee and Mohawk, the arguments of a ditransitive verb usually bear the theta roles of agent, theme, and goal/source; however, in both languages, the agreement prefixes on ditransitive verbs never agree with the theme (Baker 1996; Cook 1979; Scancarelli 1987). There are many other similarities between the two languages, but most are not relevant to this analysis. However, it is worth noting one major difference between the two languages. Baker notes that Mohawk does not have 5 infinitives. Cherokee, on the other hand, does (Scancarelli 1987) and, thus, it violates Baker’s MVC (1996:25). However, since I am not arguing for or against Baker’s framework, I will not address this problem. 2.2. Pronominal prefixes Single-argument prefixes are used in two situations. First, they are used when the verb is intransitive and, thus, requires only one argument. This situation is discussed in section 2.2.1. Second, single-argument prefixes are used when the verb is transitive or ditransitive, thus requiring two or three arguments, respectively, and all arguments are third person. I discuss the transitive data in section 2.2.2 and the ditransitive data in section 2.2.3. 2.2.1 Intransitive verbs The use of single-argument prefixes on intransitive verbs is somewhat straightforward in that the person and number of the argument determine the person and number of the prefix. However, the selection of single-argument prefixes is complicated by the fact that there are two classes of prefixes. Scancarelli (1987) calls these the A- and B-classes, and she notes that, although the A-class corresponds to subjective or agentive arguments and the B-class corresponds to objective or patient arguments, these correspondences are very rough and do not always hold. 5 See example (12), section 2.2.2 220 Argument Indexing in Cherokee In the Cherokee example in (1a) above, the intransitive verb arrive subcategorizes for a single argument many people. The prefix indexes the third person plural features of the argument many people. In Mohawk, like in Cherokee, the pronominal prefix indexes the single argument of an intransitive verb, as in (4). (4) wa-ha-[a]táw-' FACT-MsA-swim-PUNC ‘He swam.’ (Baker 1996:195) 2.2.2 Transitive verbs According to Scancarelli’s (1987) analysis of Cherokee transitive verbs, the choice of prefix is straightforward when the pronominal prefix morphology on the verb indexes two arguments and these two arguments are a combination of first and/or second persons. This is shown above in (1b), where the portmanteau prefix indexes both the first person singular subject and the second person singular object features. The choice of prefix is also straightforward when the object argument is third person animate, as in (5). The pronominal prefix hi:- is the portmanteau meaning ‘second person singular subject acting on third person singular animate object’. Thus, in (5), the single prefix indexes both arguments, as in (1b). (5) hi:ko:whtiha hi:-ko:wht-iha 2SG/3SG.ANI-see-PRES ‘You see him.’ (Pulte and Feeling 1975:268) Compare example (5) with example (1c), repeated below in (6a). The only phonological difference between (5) and (6a) is the long vowel in the prefix of (5). The only semantic difference is in the animacy of the third person object, which is animate in (5) and inanimate in (6a). Comparison of the transitive verb in (6a) and the intransitive verb in (6b) shows that there is no difference at all in the prefixes. In (6a), hi- apparently indexes both the subject and the object; while in (6b), the same prefix indexes only the subject. I would like to suggest that there is actually a phonologically null morpheme indexing the third person inanimate object in (6a). The revised morpheme breakdown is given in (6c). In this example, there is no need to specify that the third person morpheme is singular; the reason why is explained in section 2.3.2. (6) a. hiko:whtiha hi-ko:wht-iha 2SG/3SG.INANI-see-PRES ‘You see it.’ (Pulte and Feeling 1975:268) 221 Susan Smythe b. c. hiwo:niha hi-wo:n-iha 2SG.A-speak-PRES ‘You are speaking.’ (Pulte and Feeling 1975:250) hiko:whtiha hi-Ø-ko:wht-iha 2SG-3INANI-see-PRES ‘You see it.’ Similarly, in Mohawk, the pronominal prefix indexing a subject and a neuter object on a transitive verb (7a) is the same as the one indexing a subject on an intransitive verb (7b). (7) a. b. ye-núhwe'-s FsA/(NsO)-like-HAB ‘She likes it.’ ye-atáw-s FsA-swim-HAB ‘She swims.’ (Baker 1996:204) (Baker 1996:204) Comparison of (8a) and (8b) indicates that something similar is happening in Cherokee when the subject is a third person. Example (8a) shows that an intransitive verb with an absolutive, first person singular subject is indexed with the B-class, single-argument prefix aki-. In example (8b), the same prefix is used on a transitive verb to indicate both the third person singular subject and the first person singular object. Based on the evidence provided by (6), I believe that there is also a zero morpheme in (8b); here, the zero morpheme is indexing a third person subject. The new morpheme breakdown is given in (8c). (8) a. b. 222 akiyo:siha aki-yo:s-iha 1SG.B-hungry-PRES ‘I’m hungry.’ akiko:whtiha aki-ko:wht-iha 1SG.B-see-PRES ‘He sees me.’ (Pulte and Feeling 1975:276) (Pulte and Feeling 1975:268) Argument Indexing in Cherokee c. akiko:whtiha Ø-aki-ko:wht-iha 3SG-1SG.B-see-PRES ‘He sees me.’ Again, a similar process is at work in Mohawk. The pronominal prefix in (9a) is used to index both the subject and the object on the transitive verb, and the same prefix is used in (9b) to index the subject of an intransitive verb. (9) a. b. yako-ya'takéhnha-s (NsA)/FsO-help-HAB ‘It helps her.’ yako-yót[]-e' FsO-work-HAB ‘She works.’ (Baker 1996:204) (Baker 1996:204) Even though Scancarelli claims that there are no zero anaphora in Cherokee, she, nevertheless, writes that the single-argument A- and B-prefixes are used on transitive verbs when (i) the A-prefix indexes an agent that occurs with a third person singular inanimate patient and (ii) the B-prefix indexes a patient that occurs with a third person singular agent (1987:37). This statement implies that there is no particular morphology that indexes, in the first case, a third person singular inanimate patient or, in the second case, a third person singular agent. Baker (1996) uses the word portmanteau to describe the transitive pronominal prefixes. This choice of word implies that what appears to be a single morpheme today was, at an earlier stage in the language, actually two distinct morphemes that fused together. If there were indeed two distinct morphemes, one indexing the subject and the other indexing the object, it would be very easy to imagine that a third person argument could be indexed on the verb with a zero morpheme. And this is exactly what Baker posits for Mohawk; he argues that a zero morpheme indexes both third person neuter subjects and objects (see example (3) in section 2.1.1). Similarly, I argue that, in Cherokee, both the third person subject/agent and the third person inanimate object/theme are indexed on the verb by a zero morpheme. However, I must stipulate that there is one exception to this situation. All Cherokee verbs are required to have a phonologically non-null pronominal prefix; this includes instances in which both arguments are third person, as in (10). I suggest that the morpheme u:- in (10) is nothing more than a slot-filler that simply occupies the required pronominal prefix slot. 223 Susan Smythe (10) u:l:khwti u:-l:khwt-hi 3SG.B-like-PRES ‘He likes it/him.’ (Scancarelli 1987:78) The question of which of the two arguments this filler morpheme refers to can be settled by the Cherokee hierarchy of semantic roles, given in (11). (11) Semantic Role Hierarchy 6 7 Agent > Effector > Locative > Theme > Patient (Scancarelli 1987:116) According to Scancarelli, “the NP argument bearing the most highly ranked semantic role will be the subject of the verb; the NP argument bearing the next most highly ranked semantic role will be the object of the verb” (1987:116). I would like to take this notion a step further by suggesting that, in a sentence such as (10) above, in which both subject and object are third person animate, the argument bearing the most highly ranked semantic role is coindexed with the phonologically non-null morpheme and that the argument with the less highly ranked semantic role is realized as a zero morpheme. Thus, the reanalysis of (10) is given below in (12). (12) u:l:khwti u:-Ø-l:khwt-hi 3SG.B(SUBJ)-3SG(OBJ)-like-PRES ‘He likes it/him.’ Further evidence for a zero morpheme is found in verbs appearing in the infinitival aspect. Infinitives, which normally do not take agreement morphology, always do in Cherokee, as shown below in (13). (13) atawo:sti a-atawo:-sti 8 3SG.A-swim-INF ‘to swim’ (Scancarelli 1987:85) In the above example, an A-class, single-argument prefix indexing a singular third person appears in the pronominal prefix slot of the infinitival verb. As I mentioned above, this slot must always contain a phonologically non-null prefix, even if it is a dummy place holder. 6 7 8 An effector can be an instrument or a non-controlling agent (Scancarelli 1987). A locative can be an owner, a recipient, a benefactive, or a source (Scancarelli 1987). This particular verb always takes A-class prefixes (Scancarelli 1987). 224 Argument Indexing in Cherokee 2.2.3 Ditransitive verbs In the previous section, I discussed the transitive data in which the pronominal prefixes index both a subject and an object of a verb. However, if the verb has three arguments, which two of the three arguments will be indexed on the verb? Clearly, if the semantic role hierarchy given in (11) could be used above to determine which of the two third person arguments could be indexed by the pronominal prefix on a transitive verb, it could also be used to determine which two arguments can be indexed on a ditransitive verb. In example (14), the gloss contains three arguments, the agent I, the patient it, and the effector him. (14) ci:ko:wthhtiha ci:y-kowth-ht-iha 1SG/3SG.ANI-see-CAUS-PRES ‘I’m showing it to him.’ (Scancarelli 1987:68) These three arguments, according to the hierarchy in (11), are ranked as follows: agent/1SG > effector/3SG.ANI > patient/3SG.INANI. Only the two most highly ranked arguments may be indexed by the phonologically non-null pronominal prefix, as indicated in the gloss. Under past analyses, the third argument remains unaccounted for in the verbal complex. Scancarelli explicitly states that “if a verb has a dative or in9 strumental argument as well as a patient, the dative or instrumental is the object for purposes of [verbal] agreement and the 10patient is not marked on the verb, except that if the patient is plural” (1987:69). However, since major constituent word order is completely free in Cherokee and since the language is completely agglutinating otherwise, I feel that there must be some method for indexing all three arguments on the verb. Like Cherokee, Mohawk allows one pronominal prefix to occur on the verb, and this prefix can index up to two verbal arguments. When a verb has three arguments, it is always the theme that gets left out of the prefix. Baker (1996) argues that the theme is indeed indexed on the verb, but that it is incorporated rather than prefixed, as in (15a) below. He even suggests that the theme can be a phonologically null morpheme to explain the instances when no incorporated elements are visible on the verb root, as in (15b) below. Since only themes can be incorporated into a verb, “this analysis also explains the important distributional fact that whenever there are fewer argument factors than NP arguments in a given clause, it is always the theme that fails to trigger agreement [in the pronominal prefix]” (p. 207). 9 Scancarelli alternately calls this argument instrumental and causative. 10 I discuss the plural exceptions in section 2.3.2. 225 Susan Smythe (15) a. b. wa-hiy-a'shár-u-' FACT-1sA/MsO-knife-give-PUNC ‘I gave a/the knife to him.’ t-a-híy-(Ø)u-' CIS-FACT-1sA/MsO-(Ns).give-PUNC ‘I gave it to him.’ (Baker 1996:205) (Baker 1996:204) I argued above that, in Cherokee, a third person subject/agent or a third person object (which is usually a theme or patient) will be indexed on a transitive verb by a zero morpheme provided that the other argument is either first or second person. Now I suggest that the third and semantically lowest ranked argument of a ditransitive verb will also be indexed on the verb by a zero morpheme. However, rather than occurring as a null pronominal prefix, it will occur as a null noun root incorporated into the verb following Baker’s (1996) analysis of Mohawk. In (14) above, it is indexed on the verb as -Ø-. Further evidence for this analysis is given in section 2.3.1. The revised morpheme breakdown for (14) is given in (16) below. (16) ci:ko:wthhtiha ci:y-Ø.kowth-ht-iha 1SG/3SG.ANI-3SG.INANI.see-CAUS-PRES ‘I’m showing it to him.’ Before moving on, I want to discuss the Cherokee dative and instrumental/causative derivational suffixes mentioned above. Example (14)/(16) contains the causative (CAUS) suffix. According to Scancarelli, this suffix always appears on a verb that has a non-controlling agent (effector) object, and “the verb is interpreted as having a non-volitional causee” (1987:116). The sentence translation in this example contains the word show, but the verb is glossed as see in the morpheme breakdown. A more appropriate translation would be, ‘I make him (or cause him to) see it’. This translation better illustrates that the effector argument is a non-controlling agent, i.e., an agent not acting of his own volition. Of all the three-argument Cherokee sentences that I have found, a few contain the causative derivational suffix, but the majority contain the dative derivational suffix. Therefore, I would like to take this opportunity to discuss it. According to Scancarelli (1987), a verb needs a dative suffix if it is indexed with a locative (owner, recipient, benefactive, or source) argument, as in examples (17a) and (17b) below. (17) 226 a. skhiwasisi ski-Ø.wa-hsi-si 2SG/1SG-3SG.INANI.buy-DAT-PUNC ‘(You) buy it from me!’ (Scancarelli 1987:68) Argument Indexing in Cherokee b. Aya Ca:ni k:y:ne:l:i k:y-Ø.n-e:l-:i I John 1SG/2SG-3SG.ANI.hit-DAT-PERF ‘I hit John for you.’ (Scancarelli 1987:68) In (17a), the agent and the source are indexed by the pronominal prefix and the zero noun root patient is incorporated into the verb. In (17b), the agent and the benefactive are indexed by the pronominal prefix and the zero noun root patient is incorporated into the verb. These two examples demonstrate that the animacy of the patient is irrelevant in a dative sentence. Although all of the Cherokee grammars (Cook 1979; King 1975; Pulte and Feeling 1975; Scancarelli 1987) maintain that derivational suffixes may cooccur, I have found no clear examples that contain both the causative and dative suffixes. 2.3. Other methods of indexing agreement features There are two other methods by which certain arguments can be indexed on Cherokee verbs. In section 2.3.1, I discuss more evidence that theme arguments are indexed through incorporation; and, in section 2.3.2, I look at instances in which themes are indexed via the distributive prepronominal prefix. 2.3.1 The classificatory verbs Baker (1996) argues that, in Mohawk, patient objects, which cannot be indexed by the pronominal prefix system for reasons similar to those I have given for Cherokee, are indexed through noun incorporation. In section 2.2.3, I suggested that the same strategy is at work in Cherokee. Although Mohawk and Cherokee are both members of the Iroquoian language family, they have been separated for at least 3,500 years (Scancarelli 1987:10). They still have similar pronominal prefixing systems, but they differ in their levels of noun incorporation productivity. Whereas Mohawk still has a productive system of noun incorporation, Cherokee’s system of noun incorporation is almost completely lexicalized (Mithun 1984). Although the classificatory verbs are unproductive lexical items in Cherokee, they still provide evidence that Baker’s (1996) theory holds for Cherokee as well as for Mohawk. In example (18) below, the agent I and the benefactive the men are indexed on the verb by the pronominal prefix kaci:y-. The dative prefix indicates that the oblique argument indexed on the verb by the pronominal prefix is the indirect object rather than the direct object. The verb is the lexicalized classificatory verb meaning to give a living thing to someone. According to Rosen (1989), classifier noun incorporation does not change the transitivity of the verb because the incorporated noun root does not satisfy an argument of the verb. However, the incorporated noun root can index a verbal argument, and, in (18), the incorporated noun root living thing indexes the patient argument dog on the verb. Furthermore, since the incorporated noun root does not satisfy the verb’s patient argument, that nominal argument may appear in the sentence with the verb, as seen in (18), where ki:hli ‘dog’ co-occurs with the verb. Thus, when the 227 Susan Smythe verb is a classificatory ditransitive verb, there is no need to posit a phonologically null incorporated noun root. (18) ani:skaya kaci:ya:kha:ne ki:hli ani:-skaya kaci:y-kha:ne-e:-ha 3PL-man 1SG/3PL.ANI-living.N.give-DAT-PRES dog ‘I’m giving the men a dog.’ (Scancarelli 1987:69) Another example of patient noun incorporation is given in (19). Here, the classificatory verb incorporates the noun root meaning to give a solid object to someone. (19) s:kth kaci:ne:l:i kaci:y-ne-e:l-:i apple 1SG/3PL.ANI-solid.N.give-DAT-PERF ‘I gave them an apple.’ (Scancarelli 1987:68) Use of the classificatory verbs is not limited to three-argument verbs, as shown in (20). Here the classificatory verb meaning to fetch a solid object has only an agent and a patient argument. However, since this is a case of a third person subject acting on a third person object, only the agent is indexed by the pronominal prefix; the patient is indexed by the incorporated noun root. (20) sa:kwu hwiski u:tlista wu:ki:s u:-tlista w-u:-ki:s-:i one whiskey 3SG-container TRANS-3SG.B-solid.N.fetch-PERF ‘He went and got one of his whiskey bottles.’ (Scancarelli 1987:187) 2.3.2 The distributive prepronominal prefix In section 2.1.3, I quoted Scancarelli, saying that, when a verb has a dative or causative argument and a patient argument, the latter will not be marked on the verb, except when the patient is plural (1987:69). The distributive prepronominal prefix te:- has two possible functions. First, it can mark plural action of the verb, as in (21), and, second, it can mark the plurality of third person inanimate 11 and animate non-human objects, as shown in (22) and (23), respectively. In (21), the sole argument of the verb is singular, not plural; therefore, the distributive prefix indicates that the verbal action is plural. 11 I have found no examples in which the distributive prefix is used to index the plurality of a human patient. 228 Argument Indexing in Cherokee (21) ta:suhwiska te:-a-asuhwi-ska DIST-3SG.B-bark-PRES ‘It’s barking.’ (Scancarelli 1987:221) Examples (22a) and (22b) illustrate that the distributive prefix marks plurality of third person animate non-human objects. Comparison of (22a) with (22b) shows that the distributive prefix indicates plurality of the patient and not the benefactive/locative argument. In (22a), the benefactive argument man is singular, but 12 the patient argument dogs is plural. When the benefactive is pluralized in (22b), the pronominal prefix changes to accommodate this feature, but the distributive still indicates the plurality of the patient dogs. Compare these two examples with (18) above, in which the benefactive is plural and the patient is singular. (22) a. askaya te:ci:ya:kha:ne ki:hli a-skaya te:-ci:y-kha:ne-e:-ha 3SG-man DIST-1SG/3SG.ANI-living.N.give-DAT-PRES dog ‘I’m giving the man dogs.’ (Scancarelli 1987:69) b. ani:skaya te:kaci:ya:kha:ne ki:hli ani:-skaya te:-kaci:y-kha:ne-e:-ha 3PL-man DIST-1SG/3PL.ANI-living.N.give-DAT-PRES dog ‘I’m giving the men dogs.’ (Scancarelli 1987:69) (23) is an example of the distributive prefix marking the plurality of a third person inanimate patient, in this case flowers. The patient is indexed on the verb by means of the classificatory verb. The agent is indexed on the verb by the pronominal prefix, and the benefactive is indexed by the null pronominal prefix. (23) Ca:n tu:n:nele Me:li te:-u:-Ø-nne-e:l-e:-i DIST-3SG.B-3SG-flexible.N.give- Mary John DAT-ANDATIVE-PERF ‘John gave Mary flowers.’ ani:cil:ski ani:-cil:ski 3pl-flower (Scancarelli 1987:88) Because the plurality of a patient is marked on the verb separately from its person and animacy features, it could be said that patients are doubly marked on the verb. 12 Remember that many non-human nouns in Cherokee are not inflected at all; ki:hli ‘dog’ is one of them. 229 Susan Smythe The distributive prepronominal prefix is not limited to use with ditransitive verbs. In example (24), the distributive marks plurality of the patient on a transitive verb. (24) te: hiko:whtiha te:-hi-Ø-ko:wht-iha DIST-2SG-3INANI-see-PRES ‘You see them (inanimate).’ (Pulte and Feeling 1975:268) 3. Conclusion In conclusion, I have shown that it is possible to account for all of the verbal argument indexing in Cherokee by positing the existence of a zero-morpheme. By comparing Cherokee data to Baker’s (1996) analysis of Mohawk, I have shown that it is plausible (i) that both third person agents/subjects and third person themes/objects are indexed on the verb by a zero morpheme when the other argument is either first or second person; (ii) that when both arguments are third person, the argument bearing the most highly ranked theta role is marked on the verb by a dummy morpheme because every Cherokee verb must have a prefix in its pronominal prefix slot; and (iii) that themes/objects, which cannot be indexed on ditransitive verbs by the pronominal prefix system, are indexed through the remnants of a once productive system of noun incorporation. Future research into this topic could go in two very different directions. First there is need to reexamine Baker’s (1996) MVC in light of the Cherokee data regarding infinitivals and the frozen, non-productive classifier and compound noun incorporation forms. The second direction is diachronic in nature; the possibility that Proto-Iroquoian had two pronominal prefix slots on the verb template needs to be explored. References Baker, Mark C. 1996. The Polysynthesis Parameter. New York: Oxford University Press. Cook, William Hinton. 1979. A Grammar of North Carolina Cherokee. Doctoral dissertation, Yale University. Dayley Jon P. 1985. Tzutujil Grammar. Berkeley: University of California Press. Foley, Lawrence Paul. 1975. Phonological Variation in Western Cherokee. Doctoral dissertation, University of Texas at Austin. King, Duane Harold. 1975. A Grammar and Dictionary of the Cherokee Language. Doctoral dissertation, University of Georgia. Lounsbury, Floyd G. 1978. Iroquoian languages. In B.G. Trigger (ed.), Volume 15: Northeast of The Handbook of North American Indian Languages, William C. Sturtevant (ed.), Washington, D.C.: Smithsonian Institution, pp. 334–343. MacKay, Carolyn J. 1999. A Grammar of Misantla Totonac. Salt Lake City: The University of Utah Press. Merlan, Francesca. 1976. Noun incorporation and discourse reference in modern Nahuatl. International Journal of American Linguistics 42(3): 177–191. Mithun, Marianne. 1984. The evolution of noun incorporation. Language 60: 847–894. 230 Argument Indexing in Cherokee Pulte, William and Durbin Feeling. 1975. Outline of Cherokee grammar. In Durbin Feeling, Cherokee-English Dictionary, William Pulte (ed), Tahlequah: Cherokee Nation of Oklahoma, pp. 235–355. Rosen, Sara Thomas. 1989. Two types of noun incorporation: A lexical analysis. Language 65: 294–317. Scancarelli, Janine. 1987. Grammatical Relations and Verb Agreement in Cherokee. Doctoral dissertation, UCLA. Walker, Willard. 1975. Cherokee. In J. M. Crawford (ed.), Studies in Southeastern Indian Languages, Athens: University of Georgia Press, pp. 189–236. Department of Linguistics University of Texas at Austin Austin, TX 78712 ssmythe@mail.utexas.edu 231