GENERALS PAPER

Elimination of the obviative

Nattaya Piriyawiboon

Supervisor: Alana Johns

Readers: Michela Ippolito and Elizabeth Cowper

Department of Linguistics, University of Toronto April 9, 2007 1

Elimination of the obviative1 Nattaya Piriyawiboon University of Toronto

Abstract

This paper provides a new analysis of obviation in Algonquian. In Algonquian, animate third persons are commonly divided into the proximate and obviative types. Taking homophony between the inanimate plural suffix and the obviative suffix as a starting point, I argue that positing a separate obviative morpheme is unnecessary and that the obviative morpheme is the inanimate plural morpheme. Following Grafstein (1984) and Dechaine and Wiltschko (2002), obviation serves as a disjoint reference marker and reference tracking. Under a Feature Geometric approach (Harley and Ritter 2002 and Cowper and Hall 2004), I analyze obviation as a syntactic process of feature deletion. Assuming that the arguments in these languages are pronominal (Jelinek 1984, Baker 1996 and Reinholtz 1999), the overt nominals must link to the pronominal arguments by feature matching. Therefore, when two nominals can be linked to only one pronominal, the syntax deletes the person feature of the lower argument so as to eliminate the co- referencing ambiguity. In doing so, the c-commanded nominal loses its person feature and becomes morphologically inanimate. Evidence for this analysis comes from the fact that the obviative person is marked with the inanimate plural morpheme across Algonquian languages and the verb and demonstrative show inanimate person agreement with the obviative person. This study suggests that an inanimate marking on an animate nominal serves to eliminate ambiguous co-referencing between the pronominals and the adjunct nominals.

1 I would like to thank Alana Johns, my supervisor, for her faith in this work and for her countless invaluable advice. My interest in Algonquian stems from her Fields Methods class of 2005. I would also like to thank Elizabeth Cowper for many good ideas that appear in this work and Michela Ippolito, Tanya Slavin, Maria Kyriakaki, Julia Sue and Monica Irimia for their comments on earlier versions. I would also like to thank the audience at WSCLA12 for their feedback especially Elizabeth Ritter, Phil Branigan, Charlotte Reinholtz, Martina Wiltschko and Jeff Muehlbauer. Finally, I would like to thank Terry Spanish for his time and patience in being my language consultant. All errors are my own. 2

I. Introduction

Algonquian languages2 exhibit a morphological distinction between two or more animate third persons participating in the same sentence. One animate third person is unmarked while the rest is marked with an ‘obviative’ suffix. This phenomenon, particular to Algonquian languages, is known as proximate/obviative distinction. Example (1) illustrates obviation in Nishnaabemwin.

(1) John o-gii-waabam-aa-n mkwa-n John 3-past-see AN-dir-obv. bear-obv.3 ‘John saw a bear/bears’ (Grafstein, 1984, p.21) (2) John w-waabnd-aa-n jiimaan John 3-see IN-dir-IN boat ‘John sees a boat’ (Valentine, 2001, p. 311)

In (1), mkwa (bear) bears an obviative suffix -(a)n because it co-occurs with another animate third person John, which is unmarked. The verb is also marked with the same suffix. Notice that when a noun is obviative, its number is neutralized. In (2), none of the referents is marked obviative because the sentence consists of only one animate noun (John) while the other referent ‘boat’ is inanimate. The obviative morpheme has been analyzed under several accounts, for example, it is a discourse marking of focus (Bloomfield, 1962, Hockett, 1966 and Wolfart, 1973), a disjoint reference marker (Grafstein, 1984 and Dechaine and Wiltchko, 2002), a third person span (Aissen, 1997, 2001), and an indirect evidential marking (Muehlbauer, 2006). In this study, I argue that positing a separate category for a third person is unnecessary since the obviative morpheme demonstrates the same characteristics and form as an inanimate plural marking. Under the Feature Geometry Theory (Harley and Ritter, 2002, Cowper and

2 A language family spoken in parts of Canada and the United States. Examples of languages belonging to this family are Cree, Blackfoot, Potawatomi, Menomini, Nishnaabemwin, Passamaquoddy, Innu-Aimun and Fox. 3 The abbreviations used in this paper are: AN = animate; IN = inanimate; sing. = singular; pl. = plural; obv. = obviative; prox. = proximate; ind = indicative; dir = direct Theme; inv = inverse Theme; DAT = dative Case; dem. = demonstrative; 1 = first person; 2 = second person; 3 = third person; incl. = inclusive and excl. = exclusive. 3

Hall, 2004), I analyze obviation as a syntactic process of feature deletion that prevents potential ambiguity in co-referencing between the pronominal arguments and the overt nouns from happening. The data are drawn from Nishnaabemwin, an Algonquian language spoken in Southern Ontario. The objectives of the study are: first, to demonstrate that the ‘obviative’ morpheme is actually the inanimate plural morpheme; second, to explain the process of obviation in Algonquian; and third, to determine the cause of obviative marking in Algonquian. The paper is organized into the following sections. Section 2 outlines the characteristics of Algonquian obviation. Section 3 discusses some previous analyses on this topic of study. Section 4 and 5 deal with the status of obviative as a morpheme and the grammatical gender in this language family. Section 6 presents a new analysis of obviation. The analysis is then applied to some constraints found in the language in Section 7. The results show that eliminating the obviative and analyzing obviation as a syntactic process of feature deletion provides the best account for various phenomena in Algonquian. The paper concludes in Section 8.

II. Obviation in Algonquian

Obviation refers to a process of marking one of the two animate third persons in the same sentence with an obviative suffix. If there are more than two animate third persons, then all are marked except one. In Algonquian, nominals are divided into two grammatical genders: animate and inanimate (see section 5 for more details on grammatical animacy). Only the animate noun is marked obviative4. The obviative person is marked with the suffix /–an/ in Nishnaabemwin. In a transitive clause, obviation is obligatory since there are at least two referents participating in the sentence. The obviative suffix appears on the nominal as well as the head verb. However, the nominals need not be overt. When the nominals are absent, obviation is still present in the verb complex. Example (3) illustrates obviation in a transitive clause consisting of two third persons. Example (4) illustrates a sentence consisting of more than two third persons and example (5) illustrates a sentence with only one overt noun.

4 Reinholtz (personal communication) suggests that both animate and inanimate nouns are marked obviative in Swampy Cree. I will not deal with such a special case in this study. However, I will address this issue in the future research. 4

(3) giw dash kwew-ag w-gii-gnawenm-aa-waa-n niw binoojiiny-an dem. then woman-AN pl 3-past- take care of AN-dir-3pl-obv. dem. child-obv. ‘Those women took care of the child(ren)’ (Valentine, 2001, p. 631)

In (3), kwewag (women) is proximate and binoojinyan (child/children) is obviative. The obviative also marks on the head verb gnawenam (take care of ANIM). The number of an obviative person can be interpreted as singular or plural. The next example illustrates obviation when more than two animate persons participate in a clause.

(4) mii dash gii-biidmawaad niw waabgan-pwaagn-an miinwaa semaa-n and then past-bring conj. 3>3’ dem. clay-pipe-obv. and tobacco-obv. ‘..then he (prox) brought him (obv) that/those clay pipe(s) and also tobacco’ (Valentine, 2001, p. 628)

In (4), there are four animate third persons involved in the sentence. First, the Agent and the Goal are not overt because these two participants have already been introduced in the previous sentence. Only the subject of the sentence is proximate. The rest of the referents (both direct object and indirect object) is obviative. The next example illustrates obviation in a sentence with only one overt noun.

(5) mii dash gii-bkonaad niw waawaashkeshw-an and then past-skin-conj. 3>3’ dem. deer-obv. ‘Then he skinned that/those deer’ (Valentine, 2001, p. 628)

In (5), there is only one overt noun waawaashkesh (deer). However, it is obviative because the Agent is already proximate. This relationship is encoded in the conjunct verb form bkonaad (to skin AN), which indicates that a proximate person acts on an obviative person. From (3) to (5), it seems like the obviative is assigned to the object of the 5

sentence only. However, this is not the case since the obviative person can be the subject of the sentence as well, as illustrated in (6).

(6) John w-gii-waabm-ig-oon w-gwis-an John 3-past-see AN-inv-obv. 3-son-obv. ‘John’s son saw him’5 (Valentine, 2001, p. 632)

In (6), the subject (his son) is obviative and the object (John) is proximate. The verb is marked with the inverse Theme suffix –ig, instead of –aa (cf. (1)), to indicate that the obviative person is the subject of the sentence (or the Agent of the verb ‘see’). Obviation is not a Case marking since it is only relevant to the animates. It does not indicate any grammatical functions of the referents either due to its presence on both subject and object. Obviation also occurs in a possessive form. An animate noun possessed by another animate noun is obligatorily marked obviative and its number is also neutralized. The examples in (7) illustrate the possessive form of an animate noun. The examples in (8) illustrate the possessive form of an inanimate noun.

(7) a. n-gwiiwzens ‘my boy’ 1-boy b. g-gwiiwzens ‘your boy’ 2-boy c. w-gwiiwzens-an ‘his boy(s)’ 3-boy-obv. d. *w-gwiiwzens ‘his boy’ 3-boy

5 The literal translation is ‘John, he saw him, his son’ where John and the possessor of son are co- referential. 6

(8) a. n-jiimaan ‘my boat’ 1-boat b. g-jiimaan ‘your boat’ 2-boat c. w-jiimaan ‘his boat’ 3-boat (Valentine, 2001, p. 196)

In (7c), gwiiwzens (boy) is possessed by a third person. Therefore, it must be marked obviative. The absence of the obviative marking results in ungrammaticality (7d). In (8c), jiimaan (boat) is possessed by a third person. However, it is not obviative because it is inanimate. The possessive is formed by cliticizing a person agreement morpheme, which represents the possessor, to the head noun. The possessor clitics are the same as the person agreement in the transitive clause: g- for the second person, n- for the first person and w- or o- for the third person, as illustrated in (9).

(9) a. g-noondaw-aa ‘You hear him’ 2-hear AN-dir b. n-noondaw-aa ‘I hear him’ 1-hear AN- dir c. w-noondaw-aa-n ‘He hears him/them (obviative)’ 3-hear AN-dir-obv. (Valentine, 2001, p. 282)

Both transitive and possessive constructions represent a c-commanding domain where one argument c-commands the other. In the possessive construction, the possessor c- commands the possessum. In the transitive construction, the subject c-commands the object. Hence, obviation is sensitive to the syntactic position. Specifically, the c- commanded person (i.e., the possessum and the object) is always obviative and the c- commanding person is always unmarked. In the case of the obviative subject as in (6), 7

the verb has a different Theme marking, which represents the inversion of the syntactic position (see McGinnis 1999 and Bruening 2005 for such analyses).

In sum, we observe that the contexts in which obviation takes place must follow this condition: two animate third persons in a c-commanding domain. The c-commanded argument is almost always obviative. However, the presence of two third persons is not the only requirement for obviation. The obviative person can appear alone in an intransitive clause under the condition that it is labeled obviative in the previous sentence, as shown in (10):

(10) a. ikwew o-gii-baashkizwaa-an moozw-an woman 3-past-shoot-dir-obv. moose-obv. ‘the woman shot a moose’ b. moozw-an gii-bangishin-w-an moose-obv. past-fall-3-obv. ‘the moose fell’ (Grafstein, 1984, p. 34)

In (10a), ikwew (woman) is proximate and moozwan (moose) is obviative. The sentence that follows has an obviative subject although the clause is intransitive. This type of obviative marking is not obligatory. The absence of an obviative morpheme in (10b) does not make the sentence ungrammatical. I assume that this type of obviation is discourse motivated as a way of keeping the label on the referents within the continuation of the story.

In this section, I have presented the environments in which obviation takes place. Obviation is obligatory when two animate third persons are in a c-commanding domain. In the following section, some of the previous analyses on Algonquian obviation are discussed.

8

III. Previous analyses

In this section, different analyses on Algonquian obviation are compared. The obviative has been recognized as a morpheme representing a grammatical category of third persons since the first known document on the Algonquian grammar (Cuoq, 1891). In recent years, many analyses have been proposed to account for this phenomenon, for example Grafstein (1984); Rhodes (1990); Aissen (1997, 2000, 2001); Dryer (1992); Thomason (2003); Bliss and Jesney (2005). In what follows, some of these analyses are discussed.

3.1 Obviative as a discourse marker

In a reference grammar of Nishnaabemwin, obviation is described as discourse marking of prominence of the referents. The definition is given below:

“..within a given clause, only one third person animate referent can be in the foreground at a time; all other third person referents are backgrounded by making them obviative, which is marked by suffixes on animate nouns and by inflection on verbs.” (Valentine, 2001, p. 623)

From this point of view, the unmarked referent is the primary person and the focus of the sentence and the marked referent is the secondary person. This function of obviation has been widely accepted among Algonquian grammarians such as Bloomfield (1962), Hockett (1966) and Wolfart (1973). One might question why inanimate third persons need not be marked to indicate a backgrounded status. It has been argued that there is a hierarchy which represents inherent prominence of each person. Inanimate persons are ranked the lowest; therefore, it is not necessary to obviate them since their status is already lower than animate persons. This person hierarchy is also a mechanism prevalent in determining Algonquian person agreement (Payne 1997). However, stipulating an extra-linguistic mechanism such as the person hierarchy or animacy 9

hierarchy is not an attractive solution since it does not play any roles in other languages and its role in Algonquian is still dubious. Moreover, the notion of ‘focus’ is not independent from obviation, as Grafstein (1984) states “Since no independent definition of ‘focus’ has been proposed, it would appear that the only way we know that a particular third person is in focus is that it is proximate” (Grafstein, 1984, p.44). It is not clear what makes an argument more focused than the other, other than it being unmarked. Furthermore, if obviation marks non-prominence, we cannot explain why the possessor is more prominent than the possessum when the thing possessed is assumed to be the head of an NP. Universally speaking, the focused referent is usually the one that gets a special treatment. For example, it moves to the front of the sentence as in the English cleft construction6 or it bears a stronger stress than the non-focused argument. Marking the non-focused arguments appears to be counter-productive and counterfactual.

3.2 Obviative as a disjoint reference

A non-discourse analysis for obviation is also proposed. According to these accounts, obviation is a mechanism indicating a disjoint reference (e.g. Dunnigan et al. 1978; Grafstein 1984; Dechaine and Wiltchko 2002). Disjoint reference is required when two referents are third persons. In English, the sentence ‘he saw him’ demonstrates disjoint reference while ‘he saw himself’ demonstrates co-reference. The Binding principles requires that a pronominal is free and an anaphor is bound within a c- commanding domain. Grafstein argues that since there is no overt pronominals that indicate disjoint reference (such as he and him) nor anaphoric expressions (such as himself) in Algonquian, the only way the referents are interpreted as disjoint is to marked them with a disjoint marker, the obviative. Therefore, in the sentence w-waabma-aa-n ‘he sees him’, the obviative morpheme indicates that two third persons are disjoint. Without the obviative morpheme, the sentence is ungrammatical because we would not know whether the referents are disjoint or not (*w-waabm-aa). In this study, I follow Grafstein’s analysis in that obviation marks disjoint reference in pronominals; however, I further propose that obviative marking on overt nominals, which was not dealt with in her

6 For example: It was John who came late, or: John, I don’t like him. 10

analysis, serves as binding between the pronominals and the nominals. This analysis is elaborated in Section 6.

3.3 Obviative as a span

Hockett (1966) proposes that obviation is a span for third persons in a discourse. Within the third person span, there is only one prominent person. Aissen (1997, 2001) takes this position and further claims that the Obviation Span is universal. Third persons are ranked on the Obviation Scale in which the proximate is the highest ranked third person and the obviative is the lower ranked third persons. All third persons in the discourse must be assigned an obviation status and, within an obviation span, there can only be one proximate. It is not clear why there should be a separate span for a third person and why third persons should be ranked on a scale. The scale seems to be ad hoc and lacks an explanatory power. If the obviation span is universal, we would expect this span to play a role in other languages. Aissen (1997) claims that obviation is found in languages outside Algonquian such as Tzotzil and Chamorro. Her examples, however, do not illustrate the same type of obviative marking as found in Algonquian as there is no morphological marking on the overt nouns in her examples. As far as obviation is concerned, Algonquian is the only family that has an obligatory distinction among overt nominals.

3.4 Obviative as an indirect evidential marking

Muehlbauer (2006) presents a study on the choice of the proximate and obviative persons with respect to the type of verb they appear with in Plains Cree. He finds that the subject of a psyche verb (such as ‘feel’ and ‘think’) is usually a proximate. An obviative can be the subject of a psyche verb if there is evidential marking or when the event is irrealis. An example of Plains Cree obviation is in (11):

11

(11) Evidential marking (Plains Cree) ee-kii-eetokwee-aya-iteeyihtamiyit conj-prev-dubt-hes-think-obv-3 he [obv] must have thought (Muhlbauer, 2006, p.2)

He concludes that the speaker has no direct knowledge of an obviative’s internal state and that the indirect evidential marking must be used when the speaker speaks about the internal state of an obviative person.

In this section, I have discussed some of the previous analyses and shown that obviation is still a much debating issue. There is no consensus as to what function the obviative really plays and some proposals run into a problem when they try to account for it within the Generative Grammar framework. If obviative/proximate distinction is part of language universals, it should be present in other languages. However, few people have discussed this phenomenon outside Algonquian and even when they do, it is still debatable whether it has the same nature as Algonquian obviation (see Aissen (2000) and Dryer (1992) for such analyses).

IV. Eliminating the ‘obviative’

The main hypothesis of this study is that the obviative morpheme is epiphenomenal and that obviation is in fact an inanimate plural marking on an animate nominal. I will demonstrate that the obviative and inanimate plural nominals share at least three characteristics: the form of suffix, number and agreement. First, the obviative suffix and the inanimate plural suffix are homophonous in at least five Algonquian languages, as shown in 12:

12

(12) Comparison of the inanimate and obviative suffixes in four Algonquian languages

1. Proto-Algonquian (Bloomfield, 1946) Number/Animacy Animate Inanimate Obviative Singular -a -i -ali Plural -aki -ali -ahi 2. Fox (Goddard, 1994) Number/Animacy Animate Inanimate Obviative Singular -a -i -ani Plural -aki -ani -ahi 3. Menomini (Bloomfield, 1962) Number/Animacy Animate Inanimate Obviative Singular - - -an Plural -ak -an same as sing. 4. Plains Cree (Dahlstrom, 1991) Number/Animacy Animate Inanimate Obviative Singular - - -a Plural -ak -a same as sing. 5. Nishnaabemwin (Valentine, 2001) Number/Animacy Animate Inanimate Obviative Singular - - -an Plural -ag -an same as sing.

Only the animate nouns are marked with the obviative7; the inanimate nouns are never marked with the obviative. In Algonquian, nouns are divided into two grammatical genders: animate and inanimate. Animate nouns include human, animals, spirits, trees, some plants and some objects e.g. moving machines, personal belongings, card games (Bloomfield, 1962 cited in Goddard, 2001). The rest is inanimate. Gender distinction is not evident when nouns are in singular form but apparent in plural. Algonquian has

7 This statement refers to the obviative –an only. The further obviative –ini, which is also marked on the inanimate noun, will be addressed in Section 7.1. 13

different plural markings for each gender. In Nishnaabemwin, the plural suffix is /–ag/ for animate nouns and /–an/ for inanimate nouns. The examples of the obviative nouns and inanimate plural nouns are illustrated below:

(13) Comparison between the obviative and inanimate plural nouns Singular Plural Obviative Gloss Animate nouns zhiishiib zhiishiib-ag zhiishiib-an duck mnidoo mnidoo-g mnidoo-n spirit pin(y) pin-iig pin-iin8 potato kik(w) kik-oog kik-oon kettle

Inanimate nouns jiimaan jiimaan-an - boat zenbaa zenbaa-n - ribbon, silk bkok(w) bkok-oon - arrow nniinj nninj-iin - my hand (Valentine, 2001, p. 180-184)

As we can see in (13), the obviative ending for animate nouns and the plural ending for inanimate nouns are identical in their forms /-an/ and undergo the same phonological processes; that is, the vowel /a/ drops after a vowel, changes into /o/ after a /w/ and changes into /i/ after a postconsonantal palatal approximant /y/. The obviative form of a noun, according to Valentine (2001), is always determinable from the plural, by changing the final /g/ to /n/. The same can be said about the inanimate suffix –an. The second similarity is found in person clitic and the inanimate suffix. Examples in (14) illustrate person and animacy marking for the obviative and inanimate persons interacting with an animate person.

(14) a. w-waabm-aa-n ‘he (prox) sees him/them (obv)’ 3-see AN -dir-obv

8 Vowel lengthening results from the final approximants of the stem (/w/ and /y/) assimilating to the vowel of the suffix: y + in Æ iin, w + on Æ oon. 14

b. w-waabnd-aa-n ‘he sees it’ 3-see IN-dir-IN c. w-waabm-ig-oon ‘he/they (obv) sees him (prox)’ 3-see AN-inv-obv d. w-waabm-ig-on ‘it sees him’9 3-see AN-inv-IN (Valentine, 2001, p. 287)

All four sentences begin with the person clitic /w-/ and end with /-(oo)n/, which is assumed to agree with the obviative status in (14a) and (14c) and the inanimate status in (14b) and (14d). The final morpheme /-n/ appears on the verb when the inanimate object is both singular and plural (turns to /-on/ after a /g/) and on the verb when the object is obviative. Regardless of whether the final /n/ is an animacy or number agreement, it confirms our hypothesis that obviative and inanimate persons have the same agreement.

The third similarity is found in the demonstrative adjectives and pronouns. The demonstratives used for an obviative person are identical to those used for an inanimate plural person. Examples of demonstrative pronouns in Nishnaabemwin are provided below:

(15) Comparison between the demonstrative pronouns for the animate, inanimate and obviative nouns10 This These That Those Animate maaba gonda wa giwi Inanimate maanda nanda/nonda wi niwi Obviative nanda/nonda niwi

(Valentine, 2001, p. 123-124)

9 An inanimate person cannot ‘see’. This sentence is a controlled sentence specifically constructed for the purpose of comparison. 10 These forms are used in the dialect of Manitoulin Island Odawa. In other dialects, the obviative and inanimate plural demonstratives pronouns are also identical. 15

From table (15) above, we observe that the obviative demonstratives are identical to the inanimate plural ones. An example of the sentence containing an inanimate object and an obviative object is provided in (16). When an inanimate plural noun is the object of the sentence, the verb agrees with its gender and number, as shown in (16a). When an obviative person appears in a sentence, the same verb agreement and demonstrative are used, as shown in (16b).

(16) a. o-gii-ganawaaband-aa-n aniw mitigoons-an 3-past-look at IN-dir-IN pl IN pl dem. bush- IN pl ‘He looked at those bushes (inanimate)’

b. o-gii-bapasidiyeshka-waa-n aniw zhingibeny-an 3-past-kick in the rump repeatedly-dir-obv. dem. helldiver-obv. ‘He gave that/those helldiver(s) (obv.) a couple of powerful kicks in the rump’ (The Dog’s Children, p. 18)

In (16a), the verb agrees with the inanimate object mitigoonsan in number by the suffix – (a)n and with its gender by the final –nd-. The demonstrative used with the inanimate plural noun is aniw (those). In (16b), the verb agrees with the obviative object zingibenyan by the suffix –(a)n and the demonstrative used with the obviative noun is also aniw. In this sentence, the number of the helldiver is unspecified, which can be easily accounted for if we assume that it is marked with a plural suffix. Both examples show striking similarities between the inanimate plural and the obviative persons.

I have shown evidence that the obviative and inanimate plural suffix share at least three characteristics: the form, verb and demonstrative agreement and number. The plural marker and the obviative marker in Nishnaabemwin have always been treated as two distinct morphemes. I argue that they are the same item following the ‘One Form/One Meaning’ principle proposed in Johns (1992) and Monosemy Principle (Cowper 1995). The One Form/One Meaning principle is given below:

16

(17) One Form/One Meaning Principle: Where morphemes are identical or similar in phonological properties, in the unmarked case, they are identical or similar in all lexical properties.

According to the principle in (17) phonologically identical morphemes should be treated as a single lexical entry unless we have evidence of them being distinct. Because the obviative has the same phonological properties as the inanimate plural, I argue that they are the same item. Furthermore, I propose that the obviative morpheme should not be treated as a separate grammatical feature and should be eliminated from Algonquian grammar for the following reasons. First, its occurrence is governed by a context, i.e., two animate third persons in a c-commanding domain. An event never starts off with the subject being obviative unless it is previously assigned the obviative status. Second, there is no personal pronoun representing the obviative person. The proximate and obviative persons are not distinguished by overt pronouns, which suggest that they are not separate categories as usually assumed. Finally, the number of the obviative person is neutralized. It is either unspecified or interpreted as singular or plural. Therefore, I propose that we should eliminate the morpheme ‘obviative’ and treat these two homophonous suffixes as one item: the inanimate plural suffix. Thus, it follows from this assumption that an animate noun can be marked with two plural morphemes: animate plural (-ag) and inanimate plural (-an), each of which encodes different semantics on the noun (see discussion below). The motivation behind inanimate plural marking on the animate noun is discussed in Section 6. Before we discuss about the mechanism of obviation, let us first consider the semantic difference between animate and inanimate nouns in the following section.

V. Difference between animate and inanimate nouns

In this section, I would like to draw attention to the semantic difference between two grammatical genders in Algonquian. Examples in (13), page 13, demonstrate that the grammatical animacy of a noun does not necessarily correspond to its semantic animacy as we perceive in reality. Some objects such as ‘kettle’ are always grammatically 17

animate but some plants such as ‘strawberry’ are always grammatically inanimate while other plants are animate. Therefore, animacy can only be predictable to a certain degree. In a study on Algonquian grammatical gender, Goddard (2001) shows that gender can be reversed in the discourse. In narrative, an inanimate noun shifts to animate when it speaks or acts like a living thing. The following excerpt from Menominee illustrates gender shift.

(18) a. aanow keh kekeenuanenemuaw paahpenotamek eneh weenekan in vain of course I-forbade-you toy with-2/INAN that-IN skull ‘In vain I forbade you to abuse that skull [INAN]’ b. enewen-peh pes-nekaamit enoh weenekan then-QUOT AN.SG. came singing that-AN skull ..then that Skull-Being [AN] sang as it came..’ (Goddard, 2001, p. 202-203)

Gender shift is reflected in the verbal inflection and demonstrative. In (17a), weenekan (skull) is inanimate, as we can see from the use of eneh (that-INAN) and the verb final – ta- which indicates that the object is inanimate. Then, in the following sentence, it becomes animate because it sings. The shift is reflected in the use of enoh (that-AN). Similarly, animate nouns can also shift gender to inanimate. The examples below come from Fox (Goddard, 2001). The animate form represents the older usage.

(19) Animate Inanimate manetoowa ‘spirit, god’ Æ manetoowi ‘spiritual, sacred’ aamoowa ‘bee’ Æ aamoowi ‘honey’ O ishkodekaan ‘fire-steel Æ O ishkodekaan ‘lighter’ (Goddard, 2001, p. 210-212)

In (19), the animate forms represent the origin and the inanimate forms represent the derived. This type of gender shift is lexicalized. Fox is an excellent source of data for gender shift due to the distinctive system of gender marking for singular nouns. Another 18 use of the inanimate gender is collective. In Fox, a number of nouns have two genders: inanimate when having a collective meaning and animate when referring to an individual. Examples from Fox are provided in (20).

(20) Animate Inanimate shooniyaha ‘a coin, a bill’ shooniyahi ‘silver, money’ owiyasa ‘a piece of meat’ owiyasi ‘meat’ owinenwa ‘a piece of fat’ owinenwi ‘fat’ anakeehkwa ‘a piece of bark’ anakeehkwi ‘bark’ (Goddard, 2001, p. 210-212)

Another form of semantic contrast is specific versus generic meaning. Some pronominals are inanimate when referring to the generic or the ordinary, and animate when referring to the special or the unusual.

(21) Animate Inanimate asenya ‘stone used in sweatlodge’ aseni ‘stone’ apehkweeshimoona ‘pillow’ apehkweeshimooni ‘head support’ ahpishimoona ‘mattress’ ahpishimooni ‘thing for lying on’ mehtekwa ‘tree that is inviolate’ mehtekwi ‘stick, tree’ (Goddard, 2001, p. 214)

Gender shift in Fox demonstrate that where there is a gender contrast, there is also a semantic contrast. The semantic differences are summarized in Table 1:

19

Table 1: Semantic differences between animate and inanimate nouns11 Semantic Animate Inanimate Examples Specificity special generic ahpishimoona AN ‘mattress’ ahpishimooni IN ‘thing for lying on’ Inherent number individual collective owiyasa AN ‘a piece of meat’ owiyasi IN ‘meat’ Volition powerful powerless enoh weenekan AN ‘that skull-being’ eneh weenekan IN ‘that skull’ Relation source product aamoowa AN ‘bee’ aamoowi IN ‘honey’

We can conclude from the examples of gender shift that the semantics of a nominal is affected when the gender is shifted. If we assume that an animate noun becomes inanimate in the case of obviation, it follows that the semantic of that noun is affected by the gender shift as well. Specifically, the animate noun in the ‘obviative’ status should be considered less specific and less powerful than the other animate noun in the same sentence. This claim is supported by the observation made by Muehlbauer (2006) that it is unusual to report the internal state of the obviative person. In terms of number, it is shown that nouns in animate gender refer to individual referents while those in inanimate gender refer to collective referents (cf. 20). Therefore, we can also deduce that each gender has different inherent number. Specifically, an animate noun has singular as its default number and an inanimate noun has plural as its default number.12

How are animate and inanimate nouns distinguished in terms of features? Nominals with different genders should possess distinctive features. One way is to posit the features [±animate]. In a recent theory of features, bivalent features are rejected. In this paper, I adopt the Feature Geometry theory proposed in Harley and Ritter (2002) and Cowper and Hall (2004). Under this approach, the presence of a complex feature- geometric dependency structure entails the presence of all its possible substructures. The

11 These differences concern nouns that have two genders only. They are by no means including nouns that have a fixed gender. 12 I would like to thank Johns and Ritter (personal communication) for this idea. 20

Feature Geometries for [person] and [number] in Algonquian, adopted from Bejar (2003), are illustrated in (22). The π node represents the person feature. The first person is specified with the features [π[part]] (a person who is a participant) and the second person is specified with the features [π[part[addr]]] (a person who is a participant and an addressee). The animate third person is the default interpretation of the π feature (a person who is not a participant). Since the second person is the most specified person, it is the most marked. The number feature is represented by the # node with singular as the default interpretation and the feature [pl] signifies plural.

(22) Feature Geometry of Algonquian pronouns 3rd 1st sg 2nd 1st excl pl Inclusive π # π # π # π # π # (pl) Part Part (pl) part pl part part pl Addr Addr

According to the geometry in (22), the second person is the most featurally complex. Therefore, this explains why the second person is preferred as agreement controller over other persons: it is the most specific person (see detail in 6.2). The feature complexity in pronouns explains some of the properties found in the language that cannot be captured under the Person Hierarchy such as the difference between inclusive and exclusive first persons. How is animacy distinction captured under the present Feature Geometry analysis? Cowper (personal communication) suggests that inanimate and animate nouns are distinguished by the presence or absence of the person node.13 Therefore, the inanimate noun lacks a person node, hence not a person, and only has the number node whereas the animate noun has both person and number nodes. The number specification is also different between the two genders. The animate noun has singular as its default number and the inanimate noun has plural as its default number. The geometries in (23) illustrate animacy distinction.

13 In Harley and Ritter (2002)’s feature geometry, the animacy distinction is represented by the features [animate] and [inanimate] which are the subfeatures of the Class node under the Individuation node (number). 21

(23) a. Animate third person R π # (plural) b. Inanimate third person R # (singular)

The lack of the feature [plural] refers to singularity in animate nouns while the lack of [singular] refers to plurality in inanimate nouns. Evidence for the lack of a person feature in an inanimate person comes from the fact that inanimate third persons never show person agreement with the verb and the verb only agrees with their number.

This section deals with the grammatical gender in Algonquian and its features. I have shown that gender shift affects the semantics of a noun, for example a noun in the animate form is more specific than the inanimate form. Animacy distinction is captured by the presence or absence of the person node. Animate nouns contain a person and a number node whereas inanimate nouns only contain a number node and are defaulted to have the plural interpretation. In the following section, obviation is analyzed as the process of feature deletion in order to avoid co-referencing ambiguity.

VI. Obviation: a new analysis

This section presents an analysis of obviation. I propose that obviation occurs in a language where the argument structure is restricted to pronominals (Jelinek 1984, Baker 1996 and Reinholtz 1999) and where the overt nominals are not marked for morphological Case. When two animate nominals co-occur in the same clause, both of them are the potential antecedents of the pronominal because both have the same amount of the person feature which leads to an interpretation ambiguity. To avoid this ambiguity, the syntax deletes the person feature of the person to be linked to the lower argument so 22 that it is distinct from the higher one in order to prevent mismatched linking. Section 6.1 presents the argument structures of Algonquian. Section 6.2 discusses about person agreement in Algonquian. The process of feature deletion in the case of obviation is presented in section 6.3 and section 6.4 discusses about the plural feature.

6.1 The argument structures in Algonquian

The constituents in Algonquian have the following characteristics: word order is relatively free, overt nominals can be omitted; and DPs are discontinuous. I will not discuss discontinuous DPs at this time since it is out of the scope of the present study (see Reinholtz (1999) for discussion). Examples in (24) illustrate free word ordering in Nishnaabemwin. Examples in (25) illustrate omission of nominals.

(24) ‘John sees the bear’ a. John o-waabm-aa-n mkwa-n SVO John 3-see AN-dir-IN bear-IN pl b. o-waabm-aa-n mkwa-n John VOS 3-see AN-dir- IN bear- IN pl John c. mkwa-n o-waabm-aa-n John OVS bear- IN pl 3-see AN-dir- IN John d. John mkwa-n o-waabm-aa-n SOV John bear- IN pl 3-see AN-dir- IN

From (24), we see that the nominal constituents appear quite freely with respect to positions within the verb complex.14 Once the nominals are introduced, they do not need to reappear in the following sentence, as illustrated in (25):

14 Junker (2004) reports that in East Cree the most natural order is verb-obviative-proximate, as in (23b); however, the order obviative-proximate-verb (bear (o)-John (p)- he (p) sees him (o)) is ungrammatical. She proposes that this is because the preverbal position is reserved for the topicalized argument. 23

(25) Gii-mnozw-aa-n giiwenh go niwi wa giigoonyan, gii-mwaad Past-cook-dir-IN reportedly INpl dem ANsg dem fish-IN pl past-eat ‘That fellow cooked the fish, and ate it’ (Valentine, 2001, p. 620)

The sentence (25) consists of two referents: a human (wa) and a fish (giigoonyan). In the conjunct sentence, both nominals disappear because the verb inflection gii-mwaad ‘he ate it’ links the animate person to the subject position and the inanimate person to the object position (the linking or co-referencing between the pronominals and overt nominals is elaborated below). To account for the argument structure in Algonquian and other polysynthetic languages, I assume the Pronominal Argument Hypothesis (Jelinek, 1984)15, which accounts for the behaviours of the so-called non-configurational languages (Hale, 1983). PAH is summarized in (26).

(26) The Pronominal Argument Hypothesis (PAH) a. In language X, pronominal elements (clitics or agreement) must be related to each argument position in the clause. b. In the presence of a pronominal element related to an argument position Y, an overt NP Z cannot occupy position Y at S-Structure. Z can only be an adjunct to some constituent from which it binds Y. c. A clitic pronoun may be coindexed with a nominal, providing the L-case of the nominal and the G-case of the clitic pronoun are compatible (assigning a distinct index to each clitic).

PAH states that in some languages, the argument position is restricted to pronominals. The arguments in these languages are DPs that only contain phi-features (for example, person and number).16 Overt nominals are adjoined to the clause and bind the

15 See also Baker (1996) for a similar analysis on Mohawk. 16 According to Baker (1996), this is due to the fact that the verb in polysynthetic languages does not assign Case to its argument. The Case Filter (Chomsky, 1981) bars an NP to which a Case is not assigned from appearing in the argument positions. 24

pronominal arguments by Case matching. Since nominals appear outside the argument position, they can appear freely as an adjunct. In this section, I will focus on the linking between the nominals and the pronominals rather than agreement because I assume that obviation plays an important role in argument co-referencing rather than in agreement. The linking between the pronominal argument and the overt nominals is called co-referencing. PAH assumes that a nominal co-refers with a pronominal if they have a compatible Case. The languages studied in Jelinek (1984) and Baker (1996) are Warlpiri and Mohawk respectively. In these languages, overt nominals have Case marking, as shown below:

(27) Warlpiri (Jelinek, 1984, p. 57) Ngajulu-rlu ka-rna-rla-jinta karli-ki 1-ERG pres-1sg NOM-3 DAT-3DAT boomerang-DAT warri-rni ngarrka-ku see-non past man-DAT ‘I’m looking for a boomerang for the man’

In (27), the nominals, boomerang and man, which appear in the adjuncts are linked to the pronominal arguments (rla (3 DAT) and jinta (3 DAT)) by Case matching. However, in Algonquian, overt nouns are not marked for Case. Therefore, co-referencing between the nominals and the pronominal arguments cannot be established by Case matching. Following Grafstein (1984), I assume that in Algonquian, co-referencing is established by feature matching between the overt nominals and the pronominal arguments. To illustrate this point, consider the following sentences17 in Nishnaabemwin:

(28) a. n-gii-waabm-aa mkwa ‘I saw a bear’ 1-past-see AN-dir bear b. n-gii-waabm-ig mkwa ‘A bear saw me’ 1-past-see AN-inv bear

17 These two sentences are constructed by the author. 25

In (28), the verb shows a first person agreement (n-). The overt nominal mkwa ‘bear’ bears no Case marker. However, the nominal ‘bear’ is not linked to the first person morpheme n- because the features between the pro and the nominal do not match: n- has the features [person] and [participant] but mkwa only has [person] feature. Of course, its meaning should play an important role too as we do not expect anyone to link ‘bear’ to the speaker. However, the syntax only sees the features of the pronominals and evaluates whether they match with the features of the nominal or not. Therefore, feature matching plays an important role in assigning a theta role to an overt nominal. These examples illustrate that an overt nominal binds with the argument that has matching person features even though it is not marked for Case. Grafstein (1984) accounts for feature matching between a nominal and a pronominal as the principle of Argument Association, stated below:

(29) Argument Association Principle (Grafstein, 1984, p. 160) Associate a noun form with an argument position if the features of that noun form match the features attributed to that argument position.

There is one more principle in the Argument Association proposed by Grafstein which is omitted in the present study. This principle posits that when the features of two or more noun forms match the features attributed to the same argument position, the position of the nominals determines co-referencing. Specifically, the first nominal is linked to the first morpheme. This is where my analysis parts with the principle in (29). As shown in (24), overt nominals appear rather freely. Thus, the argument association cannot rely on word order. Therefore, I propose a revision of the Argument Association Principle in (30).

(30) Argument Association Principle (revised) In a pronominal argument language, the arguments are associated to the nominals by Case matching if the nominals are marked with Case; otherwise, they are associated by matching the phi-features.

26

I assume that the principles in (30) apply to all pronominal argument languages. The pronominal argument is a parameter active in the languages where the argument positions are restricted to pronominals. The nominals are assigned a theta role by matching their phi features with the pronominals. The linking can be done in two ways: first, by Case matching if nominals are marked with Case; or second, by feature matching if they are Case-less. Algonquian languages do not mark Case on the nominals. Therefore, the phi-features of the nominals are evaluated to match those of the pronominals in order to be licensed to appear. To illustrate this principle, consider a transitive clause with overt nominals. Example (2), repeated below, illustrates the linking between the pronominals and the overt nominals:

(2) w-waabnd-aa-n jiimaan John ‘he sees it, boat, John’

3- see IN-dir-IN boat John

In (2), the third person proclitic w- has one person feature and the suffix –n represents inanimate agreement, thus has no person feature. The clitic w- is thus linked to John which also has one person feature and jiimaan is linked to the pronominal –n because both contain only one number node. John is assigned the Agent role by the Theme suffix –aa. The sentence is interpretable as ‘John is the one who sees the boat’. In sum, I have proposed that Algonquian arguments are restricted to pronominals. Overt nominals appear in the adjunct and must be bound with the pronominals by feature matching in order to be interpretable. Next, person agreement in Algonquian is discussed.

6.2 Person agreement

In order to account for obviation, we need to understand the morphology of the Algonquian agreement system. Recall that Algonquian transitive verbs agree with one person and that person is realized as a person proclitic. That argument can be subject or object. This type of agreement is non-canonical (Corbett 2005) because the agreement 27 controller is not determined by the grammatical function. Consider the following example from Nishnaabemwin:

(31) a. g-noondaw-i ‘You hear me’ 2-hear-direct b. g-noondoo-n ‘I hear you’ 2-hear-inverse

Verbs always agree with the second person no matter which role it plays. The verb ‘noondaw’ in (31a) and (31b) agrees with the second person by the prefix g-. How do we know that the other argument is the first person? We know from the direct/inverse Theme suffixes –i and –in, which are portmanteau morphemes. The direct Theme suffix -i indicates that the first person is the object of the sentence. The inverse Theme suffix -in indicates that the first person is the subject. We see that Algonquian agreement does not follow the grammatical functions of the arguments. Table (2) illustrates the controller of person agreement in a transitive construction:

Table 2: Person agreement in a transitive construction

AGENT THEME CONTROLLER First person Second person Second person Second person First person Second person First / Second person Third person First / Second person Animate third person First / Second person First / Second person First / Second / Animate Inanimate First / Second / Animate Inanimate Animate Animate

This agreement asymmetry has led many linguists to assume that there exists a hierarchy in terms of person and animacy and posit a Person Hierarchy and Animacy Hierarchy to account for agreement asymmetry found in Algonquian languages. The Person Hierarchy in Algonquian is presented below: 28

(32) Algonquian Person Hierarchy 2 > 1 > 3 > 3’ > 0 According to the Person Hierarchy, all animate persons (1, 2, 3) outrank the inanimate person (0) and among the animate persons, the second person (2) is ranked the highest followed by the first (1), the third person proximate (3) and the third person obviative (3’). Notice that according to this hierarchy, the obviative third person is treated as a separate category and is sometimes referred to as a fourth person. This hierarchy has been treated as the foundation of the Algonquian agreement system by many, including Aissen (2001), Bliss (2005) and Mccaulay (2005). Nonetheless, many linguists have questioned its status in Universal Grammar (see Johns (2001) and Bruening (2005) for such analyses). I reject such a hierarchy for the following reasons. First, there is no reason why persons should be ranked. The hierarchy is merely an observation of the asymmetry in verbal agreement in Algonquian. Second, it does not offer any predictions or have implications for other languages. Third, the criteria of ranking are not obvious. What makes the second person higher ranked than others? Finally, we cannot assume that the ‘obviative person’ is a true person when it does not occur in other languages. A theory has been proposed to account for the context-sensitive agreement in Algonquian without stipulating a hierarchy. This theory is called ‘Φ-syntax’ (Bejar 2003). Under this theory, parametrical differences in agreement systems are caused by different feature specifications in the agreement probe in each language. Following Harley & Ritter (2002) and Cowper & Hall (2004), Bejar analyzes pronouns as having asymmetrical feature complexities. The feature specifications of each person in Algonquian proposed by Bejar (2003) are as follows:

(33) Person Feature specifications Second person [person [participant [addressee]]], [number (group)] First person [person [participant]], [number (group)] First person inclusive [person[participant [addressee, speaker]]], [number[group]] Third person [person], [number (group)] Inanimate person [number (group)] 29

In Algonquian, the second person is the most specified as it is the person who participates in the discourse and who is the addressee. The first person is the person who participates in the discourse but is not the addressee. The animate third person is the default interpretation of the feature [person]. The inanimate third person is not a person; therefore only has a number feature. When we compare the feature specification of the feature [person] to the Person Hierarchy in (32), we can see the asymmetry in feature complexities. [π] is the symbol for [person].

(34) Person Hierarchy: 2 > 1 > 3 > 0 Feature specifications: π π π participant participant addressee

When the second and first persons are in the sentence, the second person is the controller because it is the most specified argument. When first and third persons are participating in the sentence, the first person is the controller because it is more specified than the third person. The complexities in terms of Φ-features play an important role in determining the agreement controller in Algonquian. We have seen that DPs with more complex person features control the person agreement in these languages. What happens when both DPs are equally specified? When two animate third persons co-occur, both of them are equally specified for the [person] feature. Bejar suggests that w- is the default morpheme when the feature specification cannot determine the agreement controller. For example, in w-waabmaa-n ‘he sees him’, the clitic w- arises to represent third person agreement by default. To conclude, the non-canonical controller of person agreement is determined by a pre-specified probe. This probe is pre-specified so that only one argument can agree with the verb at a time. This argument must be the most specified person in the domain. If the probe does not find a more specific argument, as in the case of two animate third persons in a sentence, the default agreement appears. Because we have already established that the nominals must be linked to the pronominal in the argument positions, how would two 30

animate persons be linked to only one pronominal? I suggest that obviation takes place to resolve this co-referencing ambiguity. Next, an analysis of obviation in Algonquian is proposed.

6.3 Feature deletion in obviation

It has been shown that Algonquian overt nominals must be linked to a pronominal argument by feature matching. Once a co-referential relation is established, nominals can be omitted. Thus, co-referencing between nominals and pronominals is an instantiation of reference tracking. When two animate nominals co-occur, one must be linked to the person proclitic w-. Following previous analyses on obviation in terms of topicality, I assume that the subject is the topic of the sentence by default and thus must link to the person proclitic. One nominal is assigned the feature [topic]. This feature will allow it to be linked to the subject position while the other nominal lacking the feature [topic] must delete its [person] feature in order to be featurally distinct from the other nominal. The process of feature deletion is demonstrated in (35):

(35) Feature deletion at S-structure

Example: w-waabm-aa-n mkwan John ‘hei sees himj, bearj, Johni’

Step 1: At D-structure, insert verb stem [waabi] which has two arguments ‘John’ and ‘mkwa’. ‘John’ has one person feature and so does ‘mkwa’.

Step 2: At S-Structure, the stem is merged with a pro, the person feature of pro checks the uninterpretable person feature in V and the verb agrees with its animacy yielding object animacy agreement. VP V [person] pro [person] waabi + am Æ [waabam] ‘see animate’

31

Step 3: Merge the other pro at the Spec VoiceP. VoiceP

pro[per] VoiceP Voice VP

V pro[per] [waabam]

Step 4: Feature evaluation The probe evaluates the person features of each argument. Both arguments have equal specification of person features. The clitic w- results as default agreement.

Step 5: Merging the overt nominals Nominals are merged in the non-argument positions. The link between the nominals and pronominals are established by feature matching. Both nominals can be bound to w- because both have matching person feature with the pronominal w-.

Step 6: Feature deletion The nominal with the feature [topic] is linked to the pronominal w- so they must have matching features, i.e. both must have the feature [person]. The nominal without the feature [topic] deletes its person feature in order to be distinct from the feature in pro and hence, does not co-refer with it. CP

CP Johni [per] [topic]

CP mkwaj [per]

C IP I VoiceP

proi [per] VoiceP w- Voice VP

V pro [per] waabam

32

Step 7: Spell-out The nominal deleting the person feature has a bare [number] node and has the same feature specification as an inanimate noun. Without the feature [singular], the nominal has plural interpretation. Thus, number agreement for an inanimate person –an is attached to the nominal and the verb agrees with the inanimate status by the suffix –n.

Step 8: Co-referencing

w-waabm-aa-n mkwa-n John ‘he sees it/them, bear(s), John’

3-see AN-dir-IN bear-IN pl John

mkwan is linked to the inanimate morpheme –n and John is linked to the person clitic w-. No matter which order the nominals appear in, they will always have the same interpretation, which explains why Algonquian has free word order. The nominals can also be omitted once the co-referencing has been established. As long as the same nominals are the topic of discussion, their features must be maintained. In (35), the Theme suffix –aa indicates that the pronominal w- is Agent and the pronominal –n is Theme. When they switch roles, the Theme suffix –ig indicates that the pronominal w- is Theme and the pronominal n- is Agent. This inversion of role is illustrated in (14c), repeated below:

(14) c. w-waabm-ig-oon (mkwa-n) (John) ‘the bear sees John’ 3-see AN- inv- IN pl bear-IN pl John

John is still linked to w- even though the roles are reversed. The co-referencing between the two nominals and the pronominal still holds. One might question why the stem final is that of animate (waabam ‘see AN’) instead of that of inanimate (waabndang ‘see IN’). The reason is because the feature deletion takes place after the object-verb animacy agreement. The co-referencing is done after the verb agrees with the object animacy.

33

Our assumption that obviation is the resolution for co-referencing ambiguity is supported by the fact that an inanimate nominal is not allowed in the subject position if the object position is filled with another inanimate. In Algonquian, a transitive clause with two inanimate persons is simply missing (for example, The rock hits the ground). An inanimate Agent must be expressed as an instrumental morpheme instead of a subject (for example The sail moves because of wind where wind is incorporated in the verb). Two inanimates cannot co-occur because both of them have matching features with the same inanimate pro, which contains only a number node. Thus, co-referencing cannot be established and the syntax cannot fix this problem by deleting the person feature any further since inanimate third persons do not have a person feature to eliminate. This is why two inanimate third persons cannot be both subject and object of the same sentence.

6.4 Why plural?

In this section, I will deal with the issue of why an obviative person must be marked with a plural marker. One possibility is that the plural morpheme /-an/ gives different interpretations to different types of nouns.18 To illustrate, we see the same discrepancy in English determiner ‘some’. When modifying an uncountable noun, ‘some’ signifies ‘a certain amount of’ that noun. In contrast, when modifying a countable noun, ‘some’ signifies ‘an indeterminate individual or object’. This contrast is shown below:

(36) a. I bought some sugar. b. Some guy called.

The different interpretations are also expressed in the stress pattern. An indeterminate ‘some’ bears stronger stress than a quantifier ‘some’. By the same token, if –an indicates plurality on an inanimate noun, it should have a different interpretation when appears with an animate noun. I will posit that –an signifies plurality when appears with a nominal that only has the number node and the default number is plural. On the other

18 This is suggested by Johns (personal communication). 34

hand, –an signifies indeterminate number when appears with a nominal whose default number is singular. This is formulated below:

(37) Plural vs. Indeterminate a. Inanimate plural noun R number Ex: mitikoons-an ‘sticks’ stick-IN pl

b. Animate plural noun R person number plural Ex: kwew-ag ‘women’ woman-AN pl

c. Animate obviative noun (animate lacking the person feature) R person number Ex: kwew-an ‘woman/women’ woman-IN pl

Let us assume that –ag is attached to the nominal that has the [number[plural]] specification and –an is attached to the nominal that only has the [number] specification.19 In (37c), the animate noun only has one number node left because the syntax deletes its person feature to resolve co-referencing conflict. –an is attached to a bare number node. However, an animate noun has singular as its default number. Therefore, the nominal is semantically singular but morphologically plural, which yields

19 In other words, -ag is attached to a more complex number and –an is simply attached to a less complex number (-an is the default morpheme for [number]). Hence, -ag is more marked than –an. An inanimate singular noun will have the feature [singular] under the number node, preventing –an from appearing. 35

uninterpretable or ambiguous number. This is why an obviative person can be interpreted as both singular and plural. Plural seems to have double functions in many languages in that it can inherit extra semantics on top of number. In English, the plural pronoun ‘they’ can be used referring to an individual whose identity and gender are unknown as in ‘A person cannot help their birth’. Another example of a plural pronoun with singular interpretation in English is the pluralis majestatis ‘we’, indicating the superiority of the speaker but not necessarily the number of the speaker. French and other Indo-European and Slavic languages have T-V distinction where the second plural pronoun is used to refer to a singular addressee in the formal context e.g., the French tu and vous, Italian tu and voi, and the Russian ты and вы. Sauerland et al. (2005) propose that plural is semantically unmarked. Specifically, singular refers to the cardinality of one whereas plural does not necessarily mean the cardinality greater than one. Consider the following examples (Sauerland et al., 2005, p.1).

(38) a. You are welcome to bring your children. b. You are welcome to bring your child.

(38a) is used in the situation where the speaker is unsure of the number of the guest’s children whereas the speaker can state (38b) only if he is certain that the guest only has one child. These examples indicate that plural is unspecified for number, hence, the unmarked feature of number.20

In sum, I propose that obviation is a syntactic process of feature deletion to facilitate co-referencing between the nominals and the pronominals. The link between the two is established by feature matching. When there are more than one nominal that can be linked to the pronominal, the syntax deletes the person feature of all nominals except one which has the feature [topic]. Obviation is thus a syntactic process that eliminates co-referencing ambiguity in a language where the arguments are restricted to

20 A similar analysis is proposed in Harbour (2004). 36

pronominals and the nominals are Case-less. This analysis is applied to certain phenomena in Algonquian in the following section.

VII. Applying the analysis

In this section, I apply the proposed analysis to some phenomena observed in Algonquian: the possessive construction and the Possessor Constraint (Rhodes, 1993).

7.1 Obviative in the possessive construction

How does our analysis apply to obviation in the possessive form? In English, a sentence like ‘John shows Tom his car’ has two interpretations: John shows Tom John’s car or John shows Tom Tom’s car. This ambiguity is caused by binding conflict when two antecedents can be co-referential with the pronominal possessor. In Algonquian, such ambiguity is eliminated by different marking on the possessum. Namely, when the possessor is co-referential with the subject of the sentence, the possessum is marked obviative (or inanimate plural). When the possessor is co-referential with a person in a non-subject position, the possessum is marked further obviative. Examples below illustrate the possessive form of third person possessors:

(39) John w-gii-waabm-aa-n w-gwis-an John 3-past-see AN-dir-IN 3-son-IN pl

‘Johni saw hisi son’

In (39), the possessor of wgwisan is John. The possessum has the ending -an. When the possessor is not co-referential with the subject of the sentence, the possessum is marked with –ini as illustrated in (40):

(40) John w-gii-waabm-aa-ni w-gwis-ini John 3-past-see AN-dir-ini 3-son-ini

‘Johni saw hisj son’ 37

In (40), the possessor of wgwisini does not co-refer with the subject of the sentence but with a person appearing outside the clause. This is shown by the ending –ini, which is called ‘further obviative’ in the Algonquian grammar. –ini also appears on an inanimate noun. In (41a), the possessum is already inanimate thus the possessum need not be obviative. The possessor of boat is John. In (41b), boat belongs to someone else; thus the head noun jiimaan is marked with –ini.

(41) a. John w-gii-waabnd-aa-n w-jiimaan John 3-past-see IN-dir-IN 3-boat

‘Johni saw hisi boat’ b. John w-gii-waabnd-aa-ni w-jiimaan-ini John 3-past-see IN-dir-ini 3-boat-ini

‘Johni saw hisj boat’

Example (41b) illustrates that the further obviative is also marked on an inanimate noun. Our analysis states that in a transitive clause, two nominals must be featurally distinct so that they can be linked to the appropriate pronominals. This distinction is made in syntax by feature deletion. In the possessive form, co-referencing ambiguity is in the pronominal possessor which can be linked to any third persons in the sentence. To avoid this ambiguity, the possessum is marked with an inanimate plural suffix as in (39) if the possessor pronominal is linked to the person linked to the c-commanding pro in the sentence. When the possessor pronominal is linked to other persons, the possessum is marked with –ini, as in (40) and (41b). I will leave to further research to determine the nature of –ini. For now, we only observe that –ini dissociates the possessor from the subject of the sentence and associates it with the person in the non-subject position, which may or may not be obviative. When the possessor is linked to an obviative person, the possessum is marked with –ini. (42) illustrates co-referencing between the possessor and the arguments of the same clause.

38

(42) a. John w-miin-aa-n w-nimsh-an Tom-an 3-give AN-dir-IN 3-dog-IN pl Tom-IN pl

‘Johni gives Tomj hisi dog’

b. John w-miin-aa-n w-nimsh-ini Tom-an 3-give AN-dir-IN 3-dog-ini Tom-IN pl

‘Johni gives Tomj hisj dog’

To indicate that the possessor of the dog is the subject of the sentence, John, the possessum is marked inanimate plural because it is c-commanded by the pro linked to the subject. On the other hand, Tom, who is linked to the object pronominal of the verb ‘give’ is also inanimate because it is linked to a c-commanded position. When Tom is the possessor, the possessum is marked with a suffix indicating that the nominal is c- commanded by the pro linked to a c-commanded argument. It is the lowest in the c- commanding relation; and therefore, I suggest that –ini marks on the lowest nominal to indicate that that nominal belongs to the inanimate person instead of the animate person.

7.2 The Possessor Constraint

Rhodes (1993) observes that a sentence cannot contain only the obviative persons without a proximate. This constraint is referred to as “The Possessor Constraint”. Consider the examples in (37):

(43) a. John w-gii-waabmaa-n w-gwis-an John 3-past-see AN-dir-obv 3-son-obv

‘Johni saw hisi son’ b. John w-gii-waabm-ig-oon w-gwis-an John 3-past-see AN-inv-obv 3-son-obv

‘Johni’s son saw himi’

39

c. *w-gwis-an w-gii-waabm-aa-n Johnan 3-son-obv 3-past-see AN-dir-obv John-obv

‘Johni’s son saw Johni’ (literally: his son saw John) (Valentine, 2001, p. 647)

In (43a), ‘John’ is the subject and ‘his son’ is the object. The proximate is assigned to ‘John’ and the obviative to ‘his son’. The sentence is grammatical. In (43b), ‘his son’ is the subject and ‘John’ is the object. ‘John’ is proximate and ‘his son’ is obviative. The sentence is marked with an inverse Theme –ig to show that the obviative person is the subject of the sentence. The sentence is grammatical. In (43c), ‘his son’ is the subject and ‘John’ is the object. However, both ‘his son’ and ‘John’ are obviative and the verb is marked with the direct Theme. This sentence is ungrammatical because both arguments are obviative. Rhodes proposes that if the subject is possessed by the object and both the possessor and possessum are obviative, the clause is ungrammatical. This constraint can be easily captured under our analysis. The sentence (43c) is ungrammatical for two reasons. First, none of the nominals can be linked to the pro w- because both nominals lack the person feature. Second, if the possessor of son is John, and John is obviative, son must be marked with –ini and not –an. Analyzing the Possessor Constraint under our analysis gives a better understanding for why it is impossible to have two obviative persons without a proximate in a sentence. Two obviative persons in a sentence result in binding conflict between the nominals and the pronominals and between the possessor and its antecedent.

In this section, I have applied the proposed analysis to two phenomena observed in Algonquian. First, obviation in the possessive form is explained as the process of co- referencing the possessor with the argument in the sentence. If the possessor is linked to the subject of the sentence, the possessum deletes its person feature because it is c- commanded by the subject, which has a person feature. If the possessor is linked to a non-argument person, the possessum is marked with the suffix –ini. Second, the Possessive Constraint which states that the possessor cannot be linked to the obviative 40

person is explained by the fact that there cannot be two inanimate persons in the sentence due to binding conflict.

VIII. Conclusion

In this paper, I have shown that the morpheme ‘obviative’ is not a separate grammatical category as usually categorized as a type of third persons. This morpheme is argued to be the inanimate plural marker. Evidence comes from the fact that an obviative person is ambiguous in number and that it triggers the same agreement as the inanimate person. Algonquian has syntactic obviation because it has pronominal arguments; nouns are divided into animate and inanimate; overt nominals have no Case marking; and co-referencing between the pronominals and the nominals are done via feature matching. Obviation is shown to be the process of feature deletion whereby the syntax reduces the co-referencing ambiguity by deleting the person feature from one of the nominals so that it does not match the feature of the pronominal, which must be linked to the topic of the sentence. The implications this study provide for the future research is that we should find obviation in a language where arguments are pronominal and overt nominals are not marked for morphological Case. Our analysis predicts that a language where the overt nominals are Case-less but are not morphologically distinguished does not exist.

41

References

Aissen, Judith. 2001. The obviation hierarchy and morphosyntactic markedness. Linguistica atlantica 23: 1-34. Aissen, Judith. 2000. Yi and Bi: Proximate and Obviative in Navajo. Papers in Honor of Ken Hale. Endangered and Less Familiar Languages Working Papers 1, ed. by Andrew Carnie, Eloise Jelinek, MaryAnn Willie. Cambridge, MA: MIT Press. Aissen, Judith. 1997. On the Syntax of Obviation. Language 73: 705-750. Baker, Mark. 1996. The Polysynthesis Parameter. Oxford University Press. Bejar, Susana. 2003. Phi-syntax: a theory of agreement. Doctoral dissertation, University of Toronto. Bliss, Heather and Karen Jesney. 2005. Resolving hierarchy conflict: Local obviation in Blackfoot. Calgary Papers in Linguistics 26, pp. 92-116. Bloomfield, Leonard. 1962. The Menomini Language. Yale University Press. Bloomfield, Leonard. 1946. "Algonquian." Linguistic Structures of Native America. ed. Harry Hoijer et al., pp. 85-129. New York: Viking Fund. Bruening, Benjamin. 2005. The Algonquian Inverse is Syntactic: Binding in Passamaquoddy. Unpublished manuscript. Cowper, Elizabeth, and Daniel Currie Hall. 2004. The pieces of π. Paper presented at the Annual Meeting of the Canadian Linguistic Association, Winnipeg, May 2004. Cowper, Elizabeth. 1995. English Participle Constructions. Canadian Journal of Linguistics vol 40, no. 1, pp. 1-38. Cuoq, J.A. 1981. Grammaire de la langue algonquine. Proceedings and Transactions of the Royal Society of Canada, 1st series, Vol. 9: pp. 85-114. Dahlstrom, Amy. 1991. Plains Cree Morpho-Syntax. Garland Publication. Dechaine, Rose-Marie. 1999. What Algonquian morphology is really like: Hockett revisited. MIT Occasional Papers in Linguistics 17:25-72. Dechaine, R.M & M. Wiltschko. 2002. Decomposing pronoun. Linguistic Inquiry 33: 409-442. 42

Dryer, Matthew. 1992. A comparison of the obviation systems of Kutenai and Algonquian. Papers of the Algonquian Conference 23: 119-163. Dunnigan, T., P. O’Malley and L. Schwartz. 1978. A Functional Analysis of the Algonquian Obviative. Minnesota Papers in Linguistics and Philosophy of Language 5: 7-22. Goddard, Ives. 2001. Grammatical Gender in Algonquian. In Papers of the Thirty-third Algonquian Conference. Wolfart, H.C. ed. 33: 195-231. Goddard, Ives. 1994. Leonard Bloomfield’s Fox Lexicon. Algonquian and Iroquoian Linguistics. Grafstein, Ann. 1984. Argument Structure and the Syntax of a Non-configurational Language. Doctoral dissertation. McGill University. Hale, Kenneth. 1983. Warlpiri and the grammar of non-configurational languages. Natural Language and Linguistic Theory, 1: 5-47. Harbour, Daniel. 2003. The Kiowa Case for Feature Insertion. Natural Language and Linguistic Theory, Vol. 21 No. 3: 543-578. Harley, Heidi, and Elizabeth Ritter. 2002. Person and number in pronouns: a feature- geometric analysis. Language 78:482-526. Hockett, Charles. 1966. What Algonquian is really like. IJAL 32: 59-73. Jelinek, Eloise. 1984. Empty Categories, Case and Configurationality, Natural Language and Linguistic Theory, 2:1, pp. 39-76. Johns, Alana. 2001. An inclination towards accusative. Linguistica Atlantica 23:127-144. Johns, Alana. 1992. Deriving Ergativity. Linguistic Inquiry 23, Vol 1: 57 – 87. Junker, Marie-Odile. 2004. Focus, obviation and word order in East Cree. Lingua 114: 345-365. McGinnis, Martha. 1999. Is there syntactic inversion in Ojibwa? In Proceedings of the Workshop on Structure and Constituency in Native American Languages, ed. Leora Bar-El, Rose-Marie Dechaine & Charlotte Reinholz. MIT Occasional Papers in Linguistcs 17, 101-118. Muhlbauer, Jeff. 2006. How does it feel to be an obviative? Obviation and speaker knowledge in Plains Cree. Paper presented at the 38th Algonquian Conference. University of British Columbia. October 27-29, 2006. 43

Payne, Thomas. 1997. Describing Morphosyntax: a Guide for Field Linguists. Cambridge University Press. Reinholtz, Charlotte. 1999. On the Characterization of Discontinuous Constituents: Evidence from Swampy Cree. International Journal of American Linguistics, Vol. 65, No. 2, pp. 201-227. Rhodes, Richard. 1993. The Possessor Constraint. Paper presented at the 25th Algonquian Conference, Ottawa, Canada. Rhodes, Richard. 1990. Obviation, inversion and topic rank in Ojibwa. Proceedings of the Berkeley Linguistics Society 16: 101-115. Rhodes, Richard. 1976. The morphosyntax of the Central Ojibwa verb. Doctoral dissertation, University of Michigan. Ritter, Elizabeth. 1993. Where's Gender? Linguistic inquiry 24:44, 795-803, MIT Press. Ritter, E. & S. Rosen. 2005. Agreement without A-positions: another look at Algonquian. Linguistic inquiry 36:44, 648-660, MIT Press. Sauerland, U., J. Andersen & K. Yatsushiro. 2005. The plural is semantically unmarked. In S. Kepwer and M. Reis (ed.), Linguistic Evidence. Mouton de Gruyter, Berlin. Sauerland, Uli. 2005. On the semantic markedness of phi-features. Proceedings of the Phi-Workshop, McGill University, Montreal, Canada. Thomason, Lucy Grey. 2003. The Proximate and Obviative Contrast in Meskwaki. Doctoral dissertation, University of Texas at Austin. Valentine, J. Randolph. 2001. Nishnaabemwin Reference Grammar. University of Toronto Press. Williams, Angeline. 1991. The dog's children : Anishinaabe texts told by Angeline Williams. edited and translated by Leonard Bloomfield ; newly edited and with a glossary by John D. Nichols. University of Manitoba Press. Wolfart, H.C. 1973. Plains Cree: A Grammatical Study. Transactions of the American Philosophical Society. New Series – Vol. 63, part 5.