<<

Algonquian grammar myths

Will Oxford University of Manitoba

This paper seeks to clarify several misconceptions that occasion- ally arise with respect to direct-inverse marking, person hierar- chies, obviation, and transitivity in the Algonquian . The paper presents illustrative data from a variety of and is intended to serve as a reference for some of the more complex principles of Algonquian morphosyntax.

1 Introduction

The Algonquian languages are known for their complex and typologically unusual morphosyntax. Despite the existence of a comprehensive descriptive framework (especially Bloomfield 1962 and Goddard 1969), certain misconceptions about Algonquian morphosyntax have arisen both within and outside the Al- gonquianist literature. This paper seeks to clarify several of these “Algonquian grammar myths”, with the more general goal of highlighting what is known about these difficult aspects of Algonquian grammar. I have tried to limit the identification of “myths” to matters of descriptive fact. Many morphosyn- tactic phenomena in Algonquian are controversial, such as whether obviation can be understood as switch- reference (Muehlbauer, 2012), whether unspecified-actor forms are impersonal or passive (Dryer, 1996), whether transitive inanimate theme signs are derivational or inflectional (Piggott, 1989; Goddard, 2007), and whether transitive animate theme signs mark hierarchical alignment, agreement, or viewpoint aspect (Wolfart, 1973; Rhodes, 1976; Bliss, Ritter, & Wiltschko, 2014). Regardless of one’s stance on these issues, it would be unfair to refer to opposing positions as myths. In other cases, however, it is clear that certain descriptive statements are inadequate. This paper focuses on such clear cases, though I acknowledge that my judgments may reflect my own biases, and I may well be labouring under some myths of my own. The paper begins by laying out some of the basic grammatical terminology that is used in Algonquian linguistics (§2). Ten Algonquian grammar myths are then discussed (§3).

2 Algonquian grammatical terminology

Linguists who study Algonquian morphosyntax employ an elaborate and well-developed system of grammatical terminology. The system is essentially that of Bloomfield (1946, 1962), with some additions by Goddard (1969). This section introduces the terminology that will recur in this paper in the description of classes, verb inflection templates, and nominal features. Verb classes. Algonquian verb stems fall into four basic morphological classes, listed in Table 1. Morphologically intransitive stems (AI, II) mark the of their actor argument while morphologically transitive stems (TA, TI) mark the animacy of their undergoer argument. (Since the notion of subjecthood is contentious in Algonquian linguistics, I use the terms “actor” and “undergoer” to refer to verbal arguments throughout this paper, adapted from Bloomfield’s (1946) use of “actor” and “goal”.) See Section 3.8 for examples and further discussion of the verb classes. Verb inflection templates. Nearly all Algonquian languages have a contrast between two formally distinct sets of verb inflection, known as . The   inflection occurs canonically in main clauses and the   inflection occurs canonically in subordinate clauses, although the details of the conditioning are subtle and vary extensively across the family. Both sets of inflection mark mostly the same contrasts, but they do so using quite different morphology. The basic verb inflection

Toronto Working Papers in Linguistics (TWPL), Volume 39 © 2017 Will Oxford W O

Table 1: Algonquian verb stem classes

Class Argument structure AI animate intransitive animate actor II inanimate intransitive inanimate actor TA transitive animate actor + animate undergoer TI transitive inanimate actor + inanimate undergoer templates given for in (1) and (2) are representative of the patterns that are found in most of the languages, although many details vary, additional slots exist, and not all slots are overtly occupied in all forms. The labeling of the slots in these templates follows that in Bloomfield 1946 (prefix, stem, theme sign, mode sign) and Goddard 1969 (central suffix, peripheral suffix).1

(1) Ojibwe TA independent inflection template Example: niwaapamaaaassiwaanaapanek ‘we didn’t see them’ (Nichols, 1980: 290) Theme Central Mode Peripheral Prefix Stem Negative Sign Suffix Sign Suffix ni- waapam -aa -ssiw -aanaa -pan -ek 1- see - - -1 - -3

(2) Ojibwe TA conjunct inflection template Example: waapamaassiwankitwaapan ‘we didn’t see them’ (Nichols, 1980: 317) Theme Central Mode Stem Negative Sign Suffix Sign waapam -aa -ssiw -ankit + -waa -pan see - - -1:3 -3 -

Some authors use numbers rather than names to refer to the inflectional slots, but such systems are entirely dependent on the precise number of slots that are distinguished in a particular linguist’s analysis of a particular and thus do not lend themselves well to cross-family comparisons. Named slots seem preferable to numbered slots in that they allow us to, for example, compare the “central suffix” in and Delaware without needing to translate from one slot numbering system to another. Nominal features. Algonquian nominals may be first person (1), second person (2), animate third person (3), or inanimate third person (0). The use of “3” for animates and “0” for inanimates is conventional. Nominals may be singular () or (); Mi’gmaq and Maliseet-Passamaquody have developed a plural- dual contrast in some contexts. First-person show a contrast between exclusive (1) and inclusive (21). Some transitive forms involve an unspecified or impersonal actor, which is notated as “X”. Third

1Abbreviations: 1→3 or 1:3 = first acts on third; 1>3 = first outranks third; 1 = exclusive first-person plural; 21 = inclusive first-person plural; 3 = animate third person; 0 = inanimate third person; 3′, 0′ = third persons; 3′′ = further obviative; AI = animate intransitive; AI+O = transitivized AI;  = animate;  = direct;  = dubitative;  = exclusive;  = indicative; II = inanimate intransitive;  = inanimate;  = inclusive;  = inverse;  = n-registration;  = negative;  = object;  = obviative; PA = Proto-Algonquian;  = plural;  = possessor;  = preterit;  = proximate; SAP = speech-act participant;  = subordinator;  = ; TA = transitive animate; TI = transitive inanimate; TI1 = TI Class 1;  = theme sign; X = unspecified actor.

2 A   persons are subject to an  contrast that distinguishes more topical  thirds (3, 0) from less topical  thirds (3′, 0′); a second degree of obviation is sometimes posited (3′′). It is often useful to provide a label that identifies the arguments of a transitive form. In this paper a formula such as “1→3” indicates that 1 is acting on 3 (as in (1)–(2) above). This notation is revised from that of Goddard (e.g. 2007), who writes “1—3”. In certain contexts I use the compressed form “1:3”. I reserve the notation “1>3” for the description of hierarchies (= “1 outranks 3”). Most of the inflected nouns and in this paper are given in plural form. This is simply for ex- pository convenience, as the morphology that indexes plurals is more reliably overt than that which indexes singulars and is thus easier for the reader to observe. In most instances a singular form could be substituted without affecting the point being made. In cases where number does play a crucial role in the inflectional pattern, both singular and plural forms will be presented (e.g. §3.2.3, §3.2.6, §3.5).

3 Myths

The ten myths discussed in this paper fall into three groups. The first four myths involve the system of direct-inverse marking and the related person hierarchy: • Myth 1: All TA verb forms are either direct or inverse. • Myth 2: There is an “Algonquian person hierarchy” (in which second outranks first). • Myth 3: Direct and inverse forms are morphologically symmetrical. • Myth 4: All TA conjunct forms contain an overt theme sign. The next three myths involve obviation: • Myth 5: Plural and obviative suffixes are in competition. • Myth 6: Only animate nominals can be obviative. • Myth 7: There is “further obviative” inflection. And the final three myths are related to transitivity: • Myth 8: AI+O verbs are morphologically intransitive. • Myth 9: All TI verbs inflect with the TI theme sign. • Myth 10: The actor argument of a transitive verb must be animate. Direct-inverse marking, person hierarchies, obviation, and transitive-intransitive distinctions are some of the most complex and typologically unusual elements of Algonquian morphosyntax, so it is understandable that certain misconceptions have arisen regarding these topics. It should be noted that most of the myths contain a significant grain of truth, as they are rooted in generalizations that accurately describe the most canonical portions of the morphosyntactic system. In most instances, the problem is that these generalizations, which are accurate as a first approximation, have been incorrectly extended to describe the system as a whole.

3.1 Myth: All TA verb forms are either direct or inverse The inflection of certain transitive animate (TA) verb forms shows a pattern that can be described in terms of a person hierarchy. Consider first the Ojibwe 1↔3 forms in (3) (Nichols, 1980). In both forms, central agreement indexes the first-person argument (ni-…-(i)naan ‘1’) and peripheral agreement indexes the third-person argument (-ik ‘3’), regardless of the thematic roles of the two arguments.

3 W O

(3) a. niwaapamaanaanik b. niwaapamikonaanik ni- waapam -aa -naan -ik ni- waapam -ikw -inaan -ik 1- see - -1 -3 1- see - -1 -3 ‘we see them’ (1→3) ‘they see us’ (3→1)

Based on this pair of forms alone, we could conclude that SAP arguments are always indexed by central agreement and third-person arguments are always indexed by peripheral agreement. The pattern becomes more complex, however, when we consider forms in which both arguments are third-person (one proximate 3, the other obviative 3′), as in (4). Here the central agreement indexes the proximate third person (o-…-waa ‘3’) and the peripheral agreement indexes the obviative third person (-an ‘3′’).

(4) a. owaapamaawaan b. owaapamikowaan o- waapam -aa -waa -an o- waapam -ikw -waa -ak 3- see - -3 -3′ 3- see - -3 -3′ ‘they see the other(s)’ (3→3′) ‘the other(s) see them’ (3′→3)

Since a proximate third person can be indexed by either peripheral agreement (in (3)) or central agreement (in (4)), we cannot posit a rigid correspondence between particular person features and particular agreement slots. Rather, there is a preference for the central agreement to index whichever argument is more central to the discourse: an SAP as opposed to a third person, as in (3), or a proximate as opposed to an obviative, as in (4). In other words, the central agreement indexes the argument that ranks higher on the hierarchy in (5).

(5) SAP > 3 > 3′

This hierarchy captures an additional aspect of the inflection of the forms in (4) and (5): when the higher- ranked argument is the actor, as in (4a) and (5a), the theme sign is realized as the  marker -aa; con- versely, when the higher-ranked argument is the undergoer, as in (4b) and (5b), the theme sign is realized as the  marker -ikw. The theme sign can thus be understood as marking the “direction” of a transitive action with respect to the hierarchy (Hockett, 1966; Wolfart, 1973): the direct theme sign marks the action of a higher-ranked person upon a lower-ranked person; the inverse theme sign marks the opposite. The descriptive apparatus of a person hierarchy and a category of “direction” is necessary in order to account for the TA forms given above. This does not mean, however, that all TA forms can or should be described in these terms. In fact, there are two sets of TA forms for which a direct-inverse description is inappropriate: mixed forms in the conjunct order (§3.1.1) and local forms (§3.1.2). In most of the languages, the description of these forms as “direct” and “inverse” is an unjustified imposition.

3.1.1 Mixed forms in the conjunct order “Mixed forms” are transitive forms in which one argument is an SAP and the other is a third person. Across the entire family, the mixed forms of the independent order display the same direct-inverse pattern that was exemplified for Ojibwe in (3) above: SAP→3 forms contain the direct theme sign (PA *-a·), 3→SAP forms contain the inverse theme sign (PA *-ekw), and the central agreement always indexes the SAP. When we turn to the equivalent forms in the conjunct order, we find a very different pattern. The inflection of the conjunct mixed forms of is shown in Table 2 (Ellis, 1971). Aside from regular sound changes, this paradigm is inherited directly from Proto-Algonquian (see reconstructions in Goddard 2000, 2015) and is thus representative of the pattern that is found in conservative languages across the family. (In order not to presuppose an analysis, I have glossed all theme signs simply as ‘’ here.) If we were to describe this paradigm using the terminology that we arrived at for the independent forms in (3) and (4) above, we would label the SAP→3 forms in the first column as direct and the 3→SAP forms

4 A  

Table 2: Inflection of Moose Cree conjunct mixed forms (Ellis 1971)

“Direct” forms “Inverse” forms 1s→3 -Ø -ak 3→1s -i -t - -1:3 - -3 1p→3 -Ø -akiht 3→1p -i -yamiht - -1:3 - -3:1 2s→3 -Ø -at 3→2s -is -k - -2:3 - -3 21→3 -Ø -ahkw 3→21 -it -ahkw - -21 - -21 2p→3 -Ø -êkw 3→2p -it -âkw - -2 - -3:2 in the second column as inverse, as I have indicated in Table 2. Indeed, many authors have done exactly this (e.g. Ellis, 1971: 90 for Moose Cree; Aubin, 1989: 10 for Algonquin; Quinn, 2015 for ). The problem with applying a direct-inverse description to the TA conjunct paradigm, however, is that neither the theme signs nor the central agreement display the properties that led us to posit a direct-inverse description of the TA independent forms in (3) and (4) above. Consider first the patterning of theme signs. In the TA independent, all SAP→3 forms show the direct theme sign *-a· and all 3→SAP forms show the inverse theme sign *-ekw. In the TA conjunct forms in (5), this is not the case. The “direct” SAP→3 forms show a null theme sign (see §3.4 below) and the “inverse” 3→SAP forms show a pair of theme signs, neither of which is the inverse marker: *-i (Cree -i) when the undergoer is first-person and *-eθ (Cree -is∼it) when the undergoer is second-person. There is no need to invoke a person hierarchy to describe the patterning of these theme signs, as their distribution can be described much more simply in terms of the person of the undergoer: *-i occurs when the undergoer is first-person, *-eθ when it is second-person, and *-Ø when it is (animate) third-person (cf. Rhodes, 1976). This is all that needs to be said about the theme signs in (5). As for the central agreement, in the TA independent it always indexes the argument that ranks highest on the person hierarchy, which, in a mixed form, is the SAP argument. But in the TA conjunct its patterning is chaotic. Two forms in the “direct” column show a central suffix that indexes the higher-ranked SAP argument (-ahkw ‘21’, -êkw ‘2’), as we would expect in a direct-inverse pattern, but the other three forms show portmanteau central suffixes that are dedicated to a particular actor-undergoer combination (-ak ‘1:3’, -akiht ‘1:3’, -at ‘2:3’). In the “inverse” column, one form shows a central suffix that indexes the higher- ranked SAP argument (-ahkw ‘21’), as we would expect in a direct-inverse pattern, but two forms show portmanteau central suffixes (-yamiht ‘3:1’, -âkw ‘3:2’) and two forms show a central suffix that indexes the lower-ranked third-person argument (-t∼k ‘3’), exactly the opposite of what we find in the independent. Central agreement in the TA conjunct is thus clearly not governed by the SAP>3 hierarchy that applies in the independent, as it may index either the SAP, the third person, or both arguments. If neither the theme signs nor the central agreement in the TA conjunct mixed paradigm can be de- scribed in terms of a direct-inverse pattern governed by an SAP > 3 > 3′ hierarchy, what reason is there to describe these forms as “direct” and “inverse”? The only reason is that the hierarchy says they should be direct and inverse, but this is not a valid justification. We posited the hierarchy as a description of cer- tain morphosyntactic patterns (in (3) and (4)). Where those patterns are absent, as in the TA conjunct mixed paradigm, we cannot invoke the hierarchy and its associated terminology, as we would be imposing an analy-

5 W O sis upon the language that is completely unmotivated by the actual data. While there is indeed a direct-inverse pattern in some TA inflectional forms, this pattern does not extend to all TA paradigms.2 In line with this conclusion, it is worth noting that most of the classic Algonquian literature does not describe TA conjunct mixed forms as direct and inverse (the chief exceptions being Ellis 1971 and Wolfart 1973 for Cree). Rather than using labels that refer to an extrinsically-imposed hierarchy, Bloomfield (1962, Menominee), Goddard (1969, Delaware), Voorhis (1974, ), Nichols (1980, Ojibwe), and Costa (2003, Miami-) organize TA forms into paradigms according to which of the four TA theme signs appears in the form. For added neutrality, the theme signs are identified simply by numbers: Theme 1 *-a·, Theme 2 *-ekw, Theme 3 *-i, Theme 4 *-eθ (Bloomfield, 1962). Thus, rather than an “inverse” paradigm, for example, there is a paradigm of forms that contain the theme sign *-ekw, whose membership differs between the independent and the conjunct. The resulting set of TA paradigms is outlined in Table 3; note the variation between the independent and conjunct in the membership of the Theme 2, 3, and 4 paradigms.

Table 3: TA paradigms in conservative languages organized according to theme sign

Forms included in paradigm Paradigm TA independent TA conjunct Forms with Theme 1 *-a·∼Ø SAP→3, 3→3′ SAP→3, 3→3′ Forms with Theme 2 *-ekw 3→SAP, 3′→3 3′→3 Forms with Theme 3 *-i 2→1 2→1, 3→1 Forms with Theme 4 *-eθ 1→2 1→2, 3→2

The classification of paradigms in Table 3 should be the starting point for any theoretical analysis of theme signs, as this classification can be read directly off of the form of the theme sign and thus reflects the data more accurately than does a description that shoehorns all TA forms into “direct” or “inverse” categories.

3.1.2 Local forms “Local forms”, also known as “you-and-me forms” (Goddard, 1967: 67), are transitive forms in which both arguments are SAPs. A pair of Meskwaki independent local forms is shown in (6) (Goddard, 1994). These forms contain the theme signs -i (PA *-i) and -en (PA *-eθ), which differ from the theme signs that appear in independent mixed and non-local forms (*-a· ‘direct’, *-ekw ‘inverse’, as in (3) and (4) above).

(6) a. kewa·pamipena b. kewa·pamenepena ke- wa·pam -i -pena ke- wa·pam -en -epena 2- see - -1 2- see - -1 ‘You see us’ (2→1) ‘We see you’ (1→2)

How are the theme signs in (6) to be described? Note first that in both of the forms in (6), the prefix indexes the second-person argument (ke- ‘2’) rather than the first-person argument. Wolfart (1973) concludes from the parallel pattern in Plains Cree that the person hierarchy includes a 2>1 ranking, which he applies not just to the prefix but to the forms as a whole: 2→1 forms such as (6a) are direct, since the higher-ranked second-person argument is the actor, while 1→2 forms such as (6b) are inverse, since the higher-ranked second-person argument is the undergoer. The theme signs that appear in these forms can then be understood

2Some languages, such as Plains Cree (Dahlstrom, 1989), have extended the direct-inverse pattern, including both the inverse theme sign *-ekw and the SAP>3 agreement preference, to certain TA conjunct mixed forms. In such cases, a direct-inverse description is appropriate—though only for the particular forms that have undergone the change.

6 A   as direct-inverse markers as well: the “direct” form in (6a) contains a “local direct” theme sign -i while the “inverse” form in (6b) contains a “local inverse” theme sign -en. This analysis allows all forms in the TA independent paradigm to be described as either direct or inverse, as summarized in Table 4.

Table 4: Direct-inverse description of TA independent forms (hierarchy 2>1>3>3′)

Configuration Direct forms Inverse forms Local 2→1 direct *-i 1→2 inverse *-eθ Mixed SAP→3 direct *-a· 3→SAP inverse *-ekw Non-local 3→3′ direct *-a· 3′→3 inverse *-ekw

Wolfart’s (1973) characterization of local forms as direct and inverse has been assumed in much sub- sequent work (e.g. Béland, 1978: 47–49; Clarke, 1982: 67–68; Brouillard & Dumont-Anichinapeo, 1987: 95–96; Dahlstrom, 1991: 42–43; Rhodes, 1994: 432; Blain, 1998: 44; Béjar & Rezac, 2009: 53; Wolven- grey, 2011: 81–83; Lochbihler, 2012: 33; Jancewicz, 2013: 195; Bui, 2016: 416). It should be emphasized, however, that the direct-inverse characterization of local forms represents a departure from the classic Al- gonquianist literature. Bloomfield (1962: 141), Hockett (1966: 65), Goddard (1969: 82), Jones (1977: 88), Nichols (1980: 164–165), Pentland (1999: 235), and Valentine (2001: 270) all reserve the terms “direct” and “inverse” for forms in which the theme signs *-a· and *-ekw appear (i.e. mixed and non-local forms) and do not use direct-inverse terminology to describe the local forms. Hockett, in fact, published a paper titled “Di- rection in the Algonquian verb: A correction” in order to address “a long-standing error in H. C. Wolfart’s interpretation of a feature of Algonquian grammar” (Hockett, 1992: 311), namely Wolfart’s characterization of local forms as direct and inverse. What’s wrong with analyzing local forms as direct and inverse? Hockett (1992) does not go into the details, but in an extensive discussion, Macaulay (2009) shows that a direct-inverse analysis of local forms is problematic with respect to both of the criteria that motivated a direct-inverse analysis of mixed and non-local forms in (3) and (4) above, i.e. the patterning of the central agreement and the patterning of the theme sign (see also McGinnis 2005: 712–713 and Chapter 3 of Zúñiga 2006). With respect to the central agreement, note that while the prefix in the examples in (6) above consistently indexes the second-person argument (ke- ‘2’), the central suffix consistently indexes the first-person argument (-(e)pena ‘1’). We thus cannot posit an overall 2>1 hierarchy for TA local forms on the basis of the patterning of the central agreement, because the two components of the central agreement show a split: the prefix is governed by a 2>1 hierarchy while the central suffix is governed by a 1 > 2 hierarchy, and there is no principled reason to decide that one of these hierarchies is more “real” than the other (Macaulay, 2009: 373–374). With respect to the local theme signs, there is both a conceptual argument and an empirical argument against a direct-inverse analysis. The conceptual argument lies in the simple fact that the local theme signs (*-i and *-eθ) are formally distinct from the direct-inverse theme signs that appear in independent mixed and non-local forms (*-a· and *-ekw). Since the local theme signs differ from the clear direct-inverse theme signs, the null hypothesis is that the local theme signs are not direct-inverse theme signs (Macaulay, 2009: 368). The empirical argument lies in the distribution of the theme signs *-i and *-eθ. In the independent, these theme signs occur only in local forms such as those in (6) above, but in the conjuct, they appear not only in local forms, as shown for Meskwaki in (7), but also in mixed forms, as in (8) (Goddard 1994).

(7) a. wa·pamiya·ke b. wa·pamena·ke wa·pam -i -a·k -e wa·pam -en -a·k -e see - -1 - see - -1 - ‘you see us’ (2→1) ‘we see you’ (1→2)

7 W O

(8) a. wa·pamiyameči b. wa·pamena·kwe wa·pam -i -ameč -i wa·pam -en -a·kw -e see - -3:1 - see - -3:2 - ‘s/he sees us’ (3→1) ‘s/he sees you (pl.)’ (3→2)

The occurrence of *-i in 3→1 forms and *-eθ in 3→2 forms is problematic for an analysis in which these theme signs mark “local direct” and “local inverse”, respectively. Consider the 3→1 form, which shows the putative “local direct” theme sign *-i. A local direct analysis of the theme sign is doubly inappropriate here: first, a 3→1 form is not local but rather mixed, so there is no reason for a local theme sign to appear, and second, from the perspective of a 2>1>3 hierarchy, a 3→1 form ought to be, if anything, inverse rather than direct, so there is no reason for a direct theme sign to appear (Macaulay, 2009: 368–369). The direct- inverse analysis of *-i and *-eθ thus fails when we attempt to extend it to the conjunct order. The correct analysis of these theme signs is much simpler: *-i marks a first-person object and *-eθ marks a second-person object (Rhodes, 1976: 189 for the conjunct; Nichols, 1980: 164; McGinnis, 1995: 12, 1999: 9; Brittain, 1999; Pentland, 1999: 235; Valentine, 2001: 270; Goddard, 2015: 372–373). An object-agreement analysis of *-i and *-eθ requires no notion of a direct-inverse hierarchy in local forms and is consistent with the distribution of *-i and *-eθ across all forms, all paradigms, and all languages. We could try to salvage the direct-inverse analysis of *-i and *-eθ by proposing a split analysis: in local forms such as (6) and (7), *-i and *-eθ are direct-inverse markers, while in mixed forms such as (8), which are the most problematic for a direct-inverse analysis, *-i and *-eθ are instead object-agreement markers. A split analysis is conceptually problematic, however, in that it effectively posits four morphemes rather than two: *-i ‘local direct’, *-i ‘first-person object’, *-eθ ‘local inverse’, *-eθ ‘second-person object’. It is then just a coincidence that the two *-i morphemes and the two *-eθ morphemes happen to share the same form. There is also empirical evidence against a split analysis. Certain Ojibwe dialects have replaced the first-person theme sign *-i with a reinforced form -ihši (Rhodes & Todd, 1981: 57). In such dialects, -ihši replaces *-i in all of its contexts: independent local forms, conjunct local forms, and conjunct mixed forms, as shown for Oji-Cree in Table 5 (data from Rogers 1964 and Todd 1970; compare the conservative Southwestern Ojibwe forms from Nichols 1980). The across-the-board replacement of *-i by -ihši indicates that speakers regarded all instances of *-i as the same morpheme. A split analysis of *-i thus must be rejected.

Table 5: Replacement of first-person theme sign *-i by -ihši in Oji-Cree

Form Oji-Cree Southwestern Ojibwe kiwaapamihšinaawaa kiwaapamim ki- waapam -ihši -naawaa ki- waapam -i -mw Independent local 2→1 2- see -1 -2 2- see -1 -2 ‘you (pl.) see me’ ‘you (pl.) see me’ waapamihšiyek waapamiyek waapam -ihši -yek waapam -i -yek Conjunct local 2→1 see -1 -2 see -1 -2 ‘you (pl.) see me’ ‘you (pl.) see me’ waapamihšič waapamit waapam -ihši -č waapam -i -t Conjunct mixed 3→1 see -1 -3 see -1 -3 ‘s/he sees me’ ‘s/he sees me’

8 A  

In summary, the TA local forms, and the theme signs that appear in them, do not warrant a direct- inverse analysis in most of the Algonquian languages.3 The central agreement in local forms in most lan- guages is governed by two opposing hierarchies, 2>1 for the prefix and 1>2 for the central suffix, and the theme signs *-i and *-eθ are more coherently analyzed as first- and second-person objects markers than as local direct and inverse markers. We must conclude that the TA local forms stand outside the direct-inverse contrast. This conclusion is in line with Goddard’s (1974: 319) statement that the local forms follow “a structural principle of their own”.

3.1.3 Summary: TA forms and the direct-inverse contrast In most Algonquian languages, there is no justification for a direct-inverse characterization of TA conjunct mixed forms and TA local forms. The direct-inverse pattern exhibited by TA non-local forms and TA independent mixed forms is a striking and ingenious form of morphosyntactic alignment, but this alignment pattern does not encompass the entire TA paradigm. It is worth remembering that in Bloomfield’s (1962: 141) grammar of Menominee—the most fully-developed statement of his approach to Algonquian linguistics—he sets out seven sets of TA inflections, only two of which are “direct forms” and “inverse forms”.4 The direct-inverse terminology describes a particular type of morphosyntactic pattern and should be invoked only when that pattern is present. As Heath (1998: 83) puts it:

In true “direct/inverse” systems there is a nonzero Inverse morpheme in 3 → {1, 2} combinations which helps differentiate them from direct {1, 2} → 3 counterparts. In some recent papers one notices extensions in the use of these terms, as “inverse” is applied to forms lacking an Inverse morpheme. Such mission creep has destroyed the usefulness of many once valuable linguistic terms…and should be resisted in this case.

Heath does not name any particular languages to which his comment applies, but I suggest that the “mission creep” that he identifies is well-attested in the literature on Algonquian languages, to the extent that it is often simply taken for granted that all TA forms are to be described as direct or inverse. I hope that the preceding discussion has helped to reinforce that this is not in fact the case.

3.2 Myth: There is an “Algonquian person hierarchy” (in which second outranks first) One of the best-known features of Algonquian morphosyntax is the person hierarchy, which is often reported to be 2>1>3 but can in fact be elaborated to the version in (9) (combining Valentine 2001: 268 and Pentland 1999: 235, fn. 14), where X is an impersonal actor, 3 is an animate third person, 0 is an inanimate third person, prime marks obviative, and double prime marks “further obviative” (but see §3.7).

(9) 2 > 1 > X > 3 > 3′ > 3′′ > 0 > 0′

3The only possible exceptions to this statement are the few languages that have extended the inverse theme sign *-ekw to certain local forms: Blackfoot marks 2→1 forms with -ok, the reflex of *-ekw (Frantz, 2009); marks 2→1 forms (but not 2→1 forms) with a theme sign -éiʔê·, which includes -éi, the reflex of *-ekw (Goddard, 2015); and several Ojibwe dialects mark 1→2 forms (but not 1→2 forms) with *-igoo, an extended version of inverse *-ekw that otherwise appears only in the inflection of impersonal forms (Rhodes & Todd, 1981). In these languages, the direct-inverse system may indeed have been extended (if only partially) to the local forms. But these are the exceptions that prove the rule: the existence of inverse marking in local forms in Blackfoot, Arapaho, and some Ojibwe dialects is precisely what makes local forms in these languages different from those of most of the family. 4The full list is as follows: direct forms, inverse forms, thou-me forms, I-thee forms, third person passive forms, first and second person passive forms, and inanimate actor forms.

9 W O

Regardless of how it is expressed, two related aspects of this hierarchy are mythical, or at least misleading: (1) the notion that there is a single hierarchy that governs the morphosyntax of an entire language, and and (2) the notion that second person is universally ranked above first person. Macaulay (2005, 2009) and Zúñiga (2006, 2008) have argued at length that both of these notions are incorrect. Instead of a single global hierarchy, there are in fact multiple hierarchies, each governing a particular inflectional slot. And among these hierarchies, instead of a universal 2>1 ranking, we in fact find both 2>1 and 1>2 rankings. In view of these facts, Macaulay (2009: 364) concludes that “the claim that the Algonquian languages manifest a ranking of 2nd person over 1st is grossly overstated” and Zúñiga (2006: 127) concludes that “[t]he Algonquian person hierarchy 2>1>3 is…at best an oversimplification and at worst an urban legend.” The remainder of this section illustrates the various person/animacy/topicality hierarchies that can be observed throughout the Algonquian inflectional system. Each hierarchy is introduced in terms of the most conservative pattern, which is normally also the pattern that occurs most commonly across the family; any variations that have developed are then noted.

3.2.1 Inverse theme sign in independent order: SAP>X>3>3′ >0 Table 6 shows the distribution of theme signs in the TA and TI Class 1 (TI1)5 inflections of the independent order in Proto-Algonquian (based on reconstructions in Goddard 1967, 2007).6 Most of the theme signs can be understood as indexing the undergoer (or object): *-i occurs only with first-person objects, *-eθ with second-person objects, *-a· with animate third-person objects, and *-am with inanimate third- person objects.7 This is not the case, however, for the inverse theme sign *-ekw and its unspecified-actor variant *-eko·, which can be found with objects of all three persons.

Table 6: TA and TI1 theme signs in the Proto-Algonquian independent order

Undergoer Actor 1 2 3 3′ 0 1 — *-eθ *-a· *-a· *-am 2 *-i —*-a· *-a· *-am X *-eko· *-eko· *-a· *-a· *-am 3 *-ekw *-ekw —*-a· *-am 3′ *-ekw *-ekw *-ekw —*-am 0 *-ekw *-ekw *-ekw *-ekw —

*-i 1,*-eθ 2,*-a· /3,*-am .,*-ekw ,*-eko· .

The distribution of the inverse theme sign cannot be tied to particular features of the actor or undergoer: it appears in some but not all forms with third-person actors (e.g. 3→1 but not 3→3′) and in some but not all forms with SAP undergoers (e.g. 3→1 but not 2→1). What matters is the actor-undergoer combination: the inverse theme sign appears whenever the undergoer outranks the actor on the hierarchy in (10). Whenever this is not the case, the theme sign is instead realized as the appropriate object-agreement marker.

5TI verb stems fall into three distinct classes with respect to the inflection that they take: TI Class 1 stems inflect with a theme sign *-am, TI Class 2 stems inflect with a theme sign *-aw, and TI Class 3 stems inflect without a theme sign (see §3.9). When discussing TI inflection, most sources more specifically on TI1 inflection. 6There are no transitive inflections for configurations in which the reference of the actor and undergoer overlaps (e.g. 1→1) or is identical (e.g. 1→1). In the latter case, a derivationally detransitivized stem is used instead. 7The TA animate-object theme sign *-a· has an allomorph *-e· and the TI1 inanimate-object theme sign *-am has allomorphs *-a· and *-e·. For simplicity I have shown only the basic allomorphs in the table.

10 A  

(10) Hierarchy for inverse theme sign, PA independent order SAP>X>3>3′ >0 ←————— 

The PA theme sign pattern in Table 6, and thus the hierarchy in (10), is retained intact in the indepen- dent inflection in nearly all of the languages. The only divergences involve the extension of inverse marking to certain local forms in Southwestern Ojibwe (Nichols, 1980), Arapaho (Goddard, 2015), and Blackfoot (Frantz, 2009), as shown in comparison with the PA pattern in Table 7.

Table 7: Theme signs in TA independent local forms

PA SW Ojibwe Arapaho Blackfoot 2→1 *-i 1 -i 1 -i 1 -oki  2→1 *-i 1 -i 1 -éiʔê· .1 -oki  1→2 *-eθ 2 -in 2 -eθ 2 -o· 1→2 1→2 *-eθ 2 -ikoo . -ê· 1. -o· 1→2

The pattern of extension of inverse marking differs in each language: Southwestern Ojibwe extends inverse marking to 1→2 forms, Arapaho extends it to 2→1 forms, and Blackfoot extends it to 2→1 forms.8 We can model these changes by elaborating the inverse hierarchy, which, in PA, did not include a ranking of first and second person with respect to each other and thus did not trigger inverse marking in local forms. The Blackfoot pattern is the simplest to explain: Blackfoot has ranked first person above second, thus making 2→1 forms inverse. Arapaho has taken a step in this direction, ranking first person plural above other SAPs, thus making 2→1 forms inverse but leaving 2→1 forms with the original first-person object marker. And SW Ojibwe has done exactly the opposite of Arapaho, ranking first person plural below other SAPs, thus making 1→2 forms inverse while leaving 1→2 forms with the original second-person object marker. The hierarchies that result from these changes are shown in (11), simplified by omitting X and 0.

(11) Hierarchies for inverse theme sign, independent order PA SAP>3>3′ SW Ojibwe SAP>1>3>3′ Arapaho 1>SAP>3>3′ Blackfoot 1>2>3>3′ ←———— 

Note that none of the inverse theme sign hierarchies in (11) are equivalent to the 2>1>3 hierarchy that is so often cited for Algonquian. If anything, the trend here is to rank first person above second person, with only SW Ojibwe showing a (partial) ranking of first below second.

8These forms present an additional complication: in each language, the inverse marker that has been extended to the local forms differs in shape from the “ordinary” inverse theme sign, i.e. the reflex of PA *-ekw. The inverse marker in the SW Ojibwe 1→2 form, -ikoo, is in fact the impersonal-actor variant of the inverse theme sign rather than the ordinary Ojibwe inverse theme sign -ikw. The inverse marker in the Arapaho 2→1 form is in fact an amalgam of the ordinary Arapaho inverse theme sign -éi plus a first-person plural marker -ê·. And the inverse marker in the Blackfoot 2→1 forms has the extended shape -oki in contrast to the ordinary Blackfoot inverse theme sign -ok.

11 W O

3.2.2 Inverse theme sign in conjunct order: 3>3′ >0 Table 8 shows the distribution of theme signs in the PA conjunct order (based on reconstructions in Bloomfield 1946 and Goddard 2000, 2015).9 In comparison with the independent (Table 6), the conjunct shows far less inverse marking. In fact, the inverse theme sign appears in only two contexts: when an obviative acts on a proximate (3′→3) and when an inanimate acts on an animate third person (0→3/3′). The shallow hierarchy in (12) is sufficient to capture these contexts. The inverse theme sign appears whenever the undergoer outranks the actor on this hierarchy. Whenever this is not the case—i.e. in the vast majority of conjunct forms—the theme sign is instead realized as the appropriate object-agreement marker.

Table 8: TA and TI1 theme signs in the Proto-Algonquian conjunct order

Undergoer Actor 1 2 3 3′ 0 1 — *-eθ *-a· *-a· *-am 2 *-i —*-a· *-a· *-am X*-i *-eθ *-a· *-a· *-am 3 *-i *-eθ —*-a· *-am 3′ *-i *-eθ *-ekw —*-am 0 *-i *-eθ *-ekw *-ekw —

*-i 1,*-eθ 2,*-a· /3,*-am .,*-ekw 

(12) Hierarchy for inverse theme sign, PA conjunct order 3>3′ >0 ←——— 

Only Kickapoo (Voorhis, 1974), Maliseet- (Sherwood, 1986), and conservative va- rieties of Mi’gmaq (Proulx, 1978) are known to retain the PA conjunct theme sign pattern in its entirety (Goddard, 1969: 85). Most of the languages have made the conjunct pattern more like that of the indepen- dent by extending the inverse theme sign to additional contexts. Meskwaki, for example, has extended the inverse theme sign to 0→SAP forms, while Menominee and all Cree and Ojibwe dialects have extended it to both 0→SAP and X→SAP forms. Certain languages have gone even further than this, extending the inverse theme sign to certain 3→SAP forms as well. Examples are given in Table 9.

Table 9: Theme signs in conjunct 3→SAP forms

PA Plains Cree Delaware Massachusett 3→1 *-i -i -i -i· -əkwi 3→2 * -eθ -it -it -əkw -əkwi 3→1 *-i -i -ikw -əkw -əkwi 3→2 *-eθ -ikw -ikw -əkw -əkwi

In the 3→1 and 3→2 conjunct forms, where PA had the object-agreement theme signs *-i ‘1’ and *-eθ ‘2’, Woods Cree has extended the inverse to the 3→2 form (Starks, 1992), Plains Cree extends it to

9In the conjunct, the theme sign *-a· ‘/3’ has a null allomorph -Ø in certain forms (see §3.4).

12 A   both the 3→2 and 3→1 forms (Wolfart, 1973), Delaware extends it to these forms and also the 3→2 form (Goddard, 1969), and Massachusett extends it further to the 3→1 form (Goddard & Bragdon, 1988), thus making the entire set of conjunct 3→SAP forms inverse, as they are in the independent. The changes in these languages can be modelled by expanding the original 3>3′ >0 inverse hierarchy to rank certain SAPs above third person, as shown in (13). Woods Cree ranks only second-person plural above third, thus making 3→2 forms inverse while leaving the remaining 3→SAP forms with their original object-agreement theme signs. At the other extreme, Massachusett ranks all SAPs above third, thereby converging its conjunct inverse hierarchy with that of the independent.

(13) Hierarchies for inverse theme sign, conjunct order PA 3>3′ >0 Woods Cree 2>3>3′ >0 Plains Cree {1, 2}>3>3′ >0 Delaware {1, 2}>3>3′ >0 Massachusett {1, 2}>3>3′ >0 (same as independent) ←—————— 

Additional complications arise when we try to incorporate the patterning of inanimate-actor forms, impersonal-actor forms, and local forms into the conjunct inverse hierarchies, which I will not consider here. In any case, it is clear from (13) that the conjunct inverse hierarchy is not 2>1>3 in any language.

3.2.3 Independent central suffix: 1>2>1/2>3>3′ >0 The central suffix in the independent order shows the same SAP>3>3′ >0 preference as the indepen- dent inverse theme sign (plus some additional complications to be discussed below). If the two arguments of a transitive form are SAP and 3, the central suffix indexes the SAP regardless of its thematic role, as in the Ojibwe examples in (14) (Nichols, 1980). If the two arguments are 3 and 3′, the central suffix indexes the 3, as in (15). And if the two arguments are 3′ and 0, the central suffix indexes the 3′, as in (16).10

(14) a. niwaapamaanaanik b. niwaapamikonaanik ni- waapam -aa -naan -ik ni- waapam -ikw -inaan -ik 1- see - -1 -3 1- see - -1 -3 ‘we see them’ (1→3) ‘they see us’ (3→1)

(15) a. owaapamaawaan b. owaapamikowaan o- waapam -aa -waa -an o- waapam -ikw -waa -an 3- see - -3 -3′ 3- see - -3 -3′ ‘they see the other(s)’ (3→3′) ‘the other(s) see them’ (3′→3)

(16) a. owaapantaanini b. owaapamikonini o- waapant -aa -in -ini o- waapam -ikw -in -ini 3- see.it - - -3′ 3- see - - -3′ ‘the other(s) see it’ (3′→0) ‘it sees the other(s)’ (0→3′)

The central suffix hierarchy is more elaborate than the inverse theme sign hierarchy, however, in that it also ranks SAPs with respect to one another. First, there is a preference to index plural rather than singular 10The forms in (16) do not contain a peripheral suffix; -ini is the obviative form of the central suffix and the -in that precedes it is a special “n-registration” marker (Rhodes, 1990b: 407) that occurs in transitive forms when the subject or primary object is inanimate.

13 W O

SAPs. If both arguments are SAPs and only one is plural, the central suffix indexes the plural, as in the Meskwaki examples in (17), where only the second-person argument is plural, and (18), where only the first-person argument is plural (Goddard, 1994).

(17) a. kewa·pamipwa b. kewa·pamenepwa ke- wa·pam -i -pwa ke- wa·pam -en -epwa 2- see -1 -2 2- see -2 -2 ‘you (pl) see me’ (2→1) ‘I see you (pl)’ (1→2)

(18) a. kewa·pamenepena b. kewa·pamipena ke- wa·pam -en -epena ke- wa·pam -i -pena 2- see -2 -1 2- see -1 -1 ‘we see you’ (1→2) ‘you see us’ (2→1)

The independent central suffix hierarchy is thus at least 1/2 > 1/2 > 3 > 3′ > 0. One further elaboration of the hierarchy is possible: when both of the SAP arguments are plural, the central suffix indexes the first-person plural argument. This point is demonstrated by the same forms that were presented in (18) to indicate the ranking of 1 over 2. It is, in fact, inconsequential whether the second person in these forms is notionally singular or plural: the plural first person takes precedence in either case (-(e)pena ‘1’). Since the central suffix is the only slot that can express the number of an SAP, the number of the second person in these forms is left unmarked and can be interpreted as either singular or plural. With the addition of the 1>2 ranking, the complete hierarchy for the independent central suffix is as in (19).11

(19) Hierarchy for central suffix, independent order 1>2>1/2>3>3′ >0

The hierarchy in (19) is attested across most of the family. The patterning of SAPs has changed in the few languages that have extended inverse marking to the local forms, i.e. Blackfoot, Arapaho, and some Ojibwe dialects (§3.2.1); see section 2.1 of Cenerini (2017) for a detailed discussion. The only other significant divergence, well-reported in the literature, is that certain dialects of Cree show a 2>1 ranking instead of the usual 1>2 (Déchaine, 1999: 60; Macaulay, 2005, 2009: 371–373 and references therein; McGinnis, 2008: 175; Zúñiga, 2008: 281–282). Consider first the Plains Cree forms in (20) (Wolfart, 1973), which show the same 1>2 ranking as the Meskwaki forms in (18): the central suffix indexes the plural first person (-(i)nân ‘1’) and the number of the second person is left unmarked.

(20) a. kiwâpamitinân b. kiwâpaminân ki- wâpam -it -inân ki- wâpam -i -nân 2- see -2 -1 2- see -1 -1 ‘we see you (sg/pl)’ (1→2) ‘you (sg/pl) see us’ (2→1)

The minimally different Moose Cree forms in (21) (Ellis, 1971) show the opposite 2>1 ranking: here the central suffix indexes the plural second person (-(i)nâwâw ‘2’) and the number of the first person is left unmarked. The independent central suffix hierarchy for dialects with this pattern is shown in (22).

11No ranking can be ascertained between 1 and 2 because distinct 1 and 2 central suffixes do not exist. The central suffixes that occur in Meskwaki independent local forms are -(e)pena ‘1’, -(e)pwa ‘2’, and -(e) ‘1/2’. (The -e- in parentheses in these suffixes appears only when preceded by a consonant.)

14 A  

(21) a. kiwâpamitinâwâw b. kiwâpaminâwâw ki- wâpam -it -inâwâw ki- wâpam -i -nâwâw 2- see -2 -2 2- see -1 -2 ‘I/we see you (pl)’ (1→2) ‘you (pl) see me/us’ (2→1)

(22) Hierarchy for central suffix, Moose Cree independent order 2>1>1/2>3>3′ >0

Neither of the independent central suffix hierarchies has the form 2>1>3. The Moose Cree hierarchy in (22) comes closest with its ranking of 2 over 1, but first and second singular remain unranked with respect to each other. And it is much more common for 1 to be ranked over 2, as in (19) above, in direct contradiction of the notion that second person always takes precedence over first in Algonquian.

3.2.4 Independent person prefix: 2>1>3(>3′ >0) The person prefix on TA independent verb forms is governed by a 2 > 1 > 3 hierarchy. If the two arguments are first and second person, the prefix indexes the second person, as in the Meskwaki examples in (23) (ke- ‘2’). If the arguments are an SAP and a third person, the prefix indexes the SAP, as in (24) (ne- ‘1’). These generalizations hold regardless of the thematic roles and number features of the two arguments.

(23) a. kewa·pamipena b. kewa·pamenepena ke- wa·pam -i -pena ke- wa·pam -en -epena 2- see -1 -1 2- see -2 -1 ‘you see us’ (2→1) ‘we see you’ (1→2)

(24) a. newa·pama·pena b. newa·pamekona·na ne- wa·pam -a· -pena ne- wa·pam -ekw -ena·n -a 1- see - -1 1- see - -1 -3 ‘we see him/her/them’ (1→3) ‘s/he sees us’ (3→1)

In Meskwaki-Kickapoo, Miami-Illinois, Arapaho, most Cree dialects, and most Menominee paradigms, the first- and second-person prefixes illustrated in (23) and (24), from PA *ne- ‘1’ and *ke- ‘2’, are the only person prefixes that appear on independent verbs. In these languages, independent verb forms with only third-person arguments simply do not take a prefix, as illustrated for Meskwaki in (25). The 2>1>3 hier- archy exemplified by the data in (23) and (24) is thus all there is to say about the prefix in these languages.

(25) a. wa·pame·waki b. wa·pameko·ki wa·pam -e· -w -aki wa·pam -ekw -w -aki see - -3 -3 see - -3 -3 ‘they see the other(s)’ (3→3′) ‘the other(s) see them’ (3′→3)

In other languages, however, independent transitive verbs also show a third-person prefix, from PA *we- ‘3’. This is the case in , Ojibwe-, , and the Eastern languages (except for contemporary Mi’gmaq, which has lost the independent inflection entirely), as well as the archaic preterit in- flections of Plains Cree (Wolfart, 1973: 43) and the Menominee negative order (Bloomfield, 1962: 169–170). In languages with a third-person prefix, the prefix hierarchy can be expanded to include a ranking of prox- imate, obviative, and inanimate third persons, although this ranking must be deduced indirectly by regard- ing the prefix and central suffix as a unit. Consider, for example, the Southwestern Ojibwe forms in (26) (Nichols, 1980), in which both arguments are animate third persons: one proximate, the other obviative.

15 W O

Which of these third persons does the prefix o- ‘3’ index? A clue comes from the central suffix, which indexes the proximate third person in these forms (-waa ‘3’). Except in local forms, the prefix and central suffix always act together as a unitary -  (Goddard, 1969: 104) that indexes a single argument (e.g. ne-…-pena ‘1’ in (24a)). In the forms in (26), the central-participant marker is o-…-waa ‘3’, and since the plural suffix -waa must be indexing the proximate argument (as the obviative argument is not plural), we must conclude that the prefix o- indexes the proximate argument as well.

(26) a. owaapamaawaan b. owaapamikowaan o- waapam -aa -waa -an o- waapam -ikw -waa -an 3- see - -3 -3′ 3- see - -3 -3′ ‘they see the other(s)’ (3→3′) ‘the other(s) see them’ (3′→3)

The same logic applies to forms in which the two arguments are animate and inanimate third persons, as in (27). Here the central-participant marker o-…-ini ‘3′’ contains the obviative suffix -ini, which indexes the animate obviative argument; the prefix o- thus must be indexing the animate argument as well.

(27) a. owaapantaanini b. owaapamikonini o- waapant -aa -in -ini o- waapam -ikw -in -ini 3- see.it - - -3′ 3- see - - -3′ ‘the other(s) see it’ (3′→0) ‘it sees the other(s)’ (0→3′)

We thus have evidence, albeit indirect, that in languages such as Ojibwe, the person prefix prefers to index animate proximates over animate obviatives, and to index animate third persons over inanimate third persons. This gives the prefix hierarchy in (28b) (compare the simpler hierarchy in (28a), which applies in languages that lack a third-person verbal prefix).

(28) Hierarchy for person prefix in independent transitive forms a. Languages without a third-person prefix: 2>1>3 b. Languages with a third-person prefix: 2>1>3>3′ >0

The prefix hierarchy does exhibit the 2>1>3 ranking that is sometimes attributed to Algonquian as a whole. It is better to regard this hierarchy as a property of the prefix slot in particular, however, since, as we have seen, other slots are governed by different hierarchies. This is the approach taken by Bloomfield (1946: 95), who describes the 2>1>3 “preference” (his word; Bloomfield did not use the term “hierarchy”) as a property of the prefix, not as anything more general. Similarly, Hockett (1966: 61) builds the 2>1 ranking into the meaning of the prefix itself by glossing second-person *ke- as “addressee involved” and first-person *ne- as “addressee excluded but speaker involved”. These glosses ensure that all forms involving first and second persons will show the second-person prefix. As in Bloomfield’s description, nothing in Hockett’s analysis reaches beyond the prefixes themselves. Claims that the 2>1 ranking is a more general property of Algonquian grammar—or even Algonquian social cognition (Cyr, 1996)—came later.

3.2.5 Inclusive first-person plural: 2>1 Algonquian languages distinguish exclusive (1) and inclusive (21) first-person plurals, as illustrated for Miami-Illinois personal pronouns, nominal possessor agreement, independent intransitive subject agree- ment, and conjunct object agreement in (29)–(32) (Costa, 2003). Second-person plural forms are also shown for comparison. Morphological slots that will be discussed below are shown in bold.

16 A  

(29) a. niiloona b. kiiloona c. kiilwa n- iilaw -ena k- iilaw -ena k- iilaw -wa 1-  -1 2-  -1 2-  -2 ‘we.’ (1) ‘we.’ (21) ‘you.’ (2)

(30) a. nisekohsena b. kisekohsena c. kisekohsawa ni- sekohs -ena ki- sekohs -ena ki- sekohs -awa 1- aunt -1 2- aunt -1 2- aunt -2 ‘our (excl) aunt’ (1) ‘our (incl) aunt’ (21) ‘your (pl) aunt’ (2)

(31) a. nimpaahpimena b. kipaahpimena c. kipaahpimwa ni- paahpi -mena ki- paahpi -mena ki- paahpi -mwa 1- play -1 2- play -1 2- play -2 ‘we (excl) play’ (1) ‘we (incl) play’ (21) ‘you (pl) play’ (2)

(32) a. waapamiaminci b. waapamelankwi c. waapamelaakwi waapam -i -amint -i waapam -el -ankw -i waapam -el -aakw -i see -1 -3:1 - see -2 -21 - see -2 -3:2 - ‘s/he looks at us (excl)’ ‘s/he looks at us (incl)’ ‘s/he looks at you (pl)’ (3→1) (3→21) (3→2)

What is noteworthy in the above forms is that whenever an inclusive is indexed by pure person- marking morphology, it is always treated as a second person rather than a first person. In (29b), (30b), and (31b), the inclusive is indexed by the second-person prefix ki- rather than by the first-person prefix ni-, while in (32b), the inclusive is indexed by the second-person object theme sign -el rather than by the first-person object theme sign -i. The treatment of the inclusive as a second person can be understood as a hierarchy effect (Macaulay, 2009: 365–366): if we consider the underlying representation of the inclusive to contain both first-person and second-person features, then both the person prefix and the object theme sign must be governed by a 2 > 1 hierarchy that causes them to realize the second-person feature of the inclusive rather than its first-person feature. The idea of a 2 > 1 hierarchy for the prefix is nothing new (§3.2.4), but it is interesting that the realization of the object-agreement theme signs is governed by this hierarchy as well.12 The 2 > 1 hierarchy in the treatment of the inclusive is stable across most of the family. The one exception that I am aware of, reported by Macaulay (2009), is from Mi’kmaq, which, according to Fidelholtz (1968), has both a conservative possession construction in which the inclusive is indexed by a second-person prefix and an innovative possession construction in which the inclusive is instead indexed by a first-person prefix (see Macaulay 2009: 374–375 for data). Macaulay concludes that the Mi’gmaq data “appear to reflect a transitional stage, in that both hierarchies for choice of possessive prefix can be found in the speech of (at least some) speakers: 2>1…and 1>2” (Macaulay, 2009: 375).

12We might imagine an alternative approach to the treatment of the inclusive that does not make reference to a hierarchy. It could be the case that the inclusive is fundamentally represented as a second person, with its first-person features being somehow secondary. The treatment of the inclusive as a second person by the prefixes and theme signs in (29)–(32) would then follow without the need to posit a 2 > 1 hierarchy. What would not follow, however, is the fact that the suffixal agreement in (29b), (30b), and (31b) treats the inclusive as a first person rather than a second person (-ena, -mena ‘1’). The generalization appears to be that when the inclusive is indexed by pure person- marking morphology, such as the prefix or the theme sign, it is treated as a second person, but when it is indexed by morphology that marks number in addition to person, such as the central suffix, it may be treated as a first person.

17 W O

3.2.6 Conjunct central suffix: no unified hierarchy The patterning of the TA conjunct central suffix differs significantly from that of its independent coun- terpart and may be the most chaotic component of the Algonquian agreement system. This section describes the patterning of the TA conjunct central suffix in Kickapoo (Voorhis, 1967), one of the few languages that preserves the Proto-Algonquian pattern intact (Goddard, 1969: 85). While glimpses of hierarchy effects arise in parts of the system, the overall pattern cannot be described in terms of a hierarchy alone. For ease of exposition, I will divide the TA conjunct forms into several sub-paradigms, starting with the non-local forms in Table 10, in which one argument is a proximate third person and the other is obviative (3′), inanimate (0), or impersonal (X). The central suffix in these forms shows a familiar pattern: it consis- tently indexes the proximate third person regardless of its thematic role. This is consistent with the 3>3′ >0 hierarchy that is attested elsewhere in the conjunct order, governing the inverse theme sign (§3.2.2).

Table 10: When no SAPs are involved, index the proximate 3

Form Suffix Arguments indexed 3→3′ -t 3 proximate 3 3→3′ -waat 3 proximate 3 3′/0→3 -t 3 proximate 3 3′/0→3 -waat 3 proximate 3 X→3 -t 3 proximate 3

The sub-paradigm in Table 11 consists of local forms in which one of the two SAP arguments is plural. Here there is also a familiar pattern: when the plural SAP is first person, it is indexed to the exclusion of the second person. This is consistent with the 1>2 ranking that governs the central suffix in the independent order (§3.2.3). When the plural SAP is second person, however, there is variation: when the second person is the actor (2→1), the central suffix indexes it, but when the second person is the undergoer (1→2) the central suffix shows a special portmanteau form that indexes both arguments simultaneously (-ako ‘1:2’, distinct from both -aan ‘1’ and -eekw ‘2’; cf. Dahlstrom 1991: 44 for the Plains Cree cognate -akok). This is our first example of a pattern that cannot be captured by a person/number hierarchy alone: to describe the second-person plural forms in Table 11, we must refer not only to the features of the arguments, but also to their thematic roles (-eekw ‘2’ when 2 is the actor, -ako ‘1:2’ when 2 is the undergoer).

Table 11: When both arguments are SAP and one is plural, index the plural (1>2) or both

Form Suffix Arguments indexed 2→1 -aak 1 first-person plural 1→2 -aak 1 first-person plural 2→1 -eekw 2 plural 1→2 -ako 1:2 both (cf. -aan 1, -eekw 2)

The sub-paradigm in Table 12 contains forms in which an SAP acts on a third person. Here the central suffix either indexes the SAP (if it is 21 or 2) or indexes both arguments simultaneously using a special portmanteau form (if the SAP is 1, 2, or 1). The patterning of the central suffix in these forms is compatible with an SAP > 3 hierarchy in the sense that there are no forms in which 3 is indexed to the exclusion of the SAP, but the hierarchy alone does not explain why some forms show agreement only with the SAP while others show agreement with both SAP and 3.

18 A  

Table 12: When SAP acts on 3, index either SAP or both arguments

Form Suffix Arguments indexed 1→3 -ak 1:3 both (cf. -aan 1, -t 3) 2→3 -at 2:3 both (cf. -an 2, -t 3) 1→3 -aket 1:3 both (cf. -aak 1, -t 3) 21→3 -akw 21 SAP 2→3 -eekw 2 SAP

The sub-paradigm in Table 13 contains mixed forms in which the undergoer is a plural SAP. As in the previous sub-paradigm, the central suffix either indexes the SAP (if it is 21) or indexes both arguments using a special portmanteau form (if the SAP is 1 or 2). Here, again, the pattern is weakly consistent with an SAP>3 hierarchy in that the SAP is always indexed by the central suffix, but the hierarchy alone cannot explain why the non-SAP is sometimes indexed as well.

Table 13: When a non-SAP acts on a plural SAP, index either SAP or both arguments

Form Suffix Arguments indexed 3→1 -amet 3:1 both (cf. -aak 1, -t 3) X/0→1 -amek X/0:1 both (cf. -aak 1, -k X/0) 3/X/0→21 -akw 21 SAP 3/X/0→2 -aakw 3:2 both (cf. -eekw 2, -t 3, -k X/0)

The final sub-paradigm, in Table 14, contains local and mixed forms in which the undergoer is a singular SAP. In such forms, the central suffix always indexes the actor (cf. Bhatia, Kusmer, & Vostrikova, 2016 for Ojibwe). The resulting patterns are strikingly non-hierarchical: the local forms 1→2 and 2→1 are asymmetrical, with -aan ‘1’ appearing in the former and -an ‘2’ in the latter, and the mixed forms show agreement for the third-person actor rather than the SAP undergoer, contravening the SAP>3 ranking that usually prevails. A person/number hierarchy brings no insight to the patterns here.

Table 14: When the undergoer is a singular SAP, index the actor

Form Suffix Arguments indexed 1→2 -aan 1 actor 2→1 -an 2 actor 3→1 -t 3 actor 3→2 -k13 3 actor 3→1/2 -k X actor 0→1/2 -k 0 actor

The patterning of the central suffix in the Kickapoo TA conjunct, which reflects the state of affairs in Proto-Algonquian, cannot be described solely in terms of person/number hierarchies. This statement remains true in most of the daughter languages, although often to a less striking extent than in Kickapoo, as many of the languages have restructured portions of the TA conjunct paradigm along direct-inverse lines (§3.2.2). 13The animate third-person central suffix -t has the allomorph -k after a consonant. This is why the shape of the suffix differs in the 3→1 and 3→2 forms: the suffix follows -i ‘1’ in the former and -eh ‘2’ in the latter.

19 W O

Nevertheless, most of the languages do retain at least some of the non-hierarchical patterns from the old system, such as the occurrence of third-person central agreement in 3→1 conjunct forms, which continues to be found across nearly the entire family. The TA conjunct thus shows us that hierarchies alone are not the full story on the patterning of Algonquian agreement. (See Bhatia et al. 2016 and Xu 2016 for theoretical accounts that aim to capture the patterns discussed above in a unified way.)

3.2.7 Independent peripheral suffix: 3>3′ >3′′ (indirectly) For simplicity, this section considers the patterning of the peripheral suffix only in independent in- dicative transitive forms in which both arguments are third persons (direct 3→3′, inverse 3′→3). Here the languages fall into three major groups. In Delaware, Massachusett, Western , Mahican, and Proto- Algonquian (Goddard, 2007: 211), the peripheral suffix indexes the lower-ranked third person if it is defi- nite. This is shown for Munsee Delaware in (33) (Goddard, 1969), where the peripheral suffix indexes the lower-ranked definite obviative 3′ (-al ‘3′’) rather than the higher-ranked proximate 3. If the lower-ranked third person is instead indefinite, the peripheral suffix indexes the higher-ranked third person regardless of its definiteness. This is shown in (34), where the peripheral suffix indexes the higher-ranked proximate 3 (-ak ‘3’) while the lower-ranked indefinite obviative 3′ is left unindexed. (For clarity, I translate definite and indefinite obviatives as ‘.’ and ‘.’ here.)

(33) a. mi·la·wá·wal b. mi·lko·wá·wal w- mi·l -a· -wa·w -al w- mi·l -əkw -wa·w -al 3- give - -3 -3′ 3- give - -3 -3′ ‘they. give to .’ (3→3′) ‘. gives to them.’ (3′→3)

(34) a. lo·sé·wak b. lo·só·ko·k lo·sw -e· -w -ak lo·sw -əkw -w -ak burn - -3 -3 burn - -3 -3 ‘they. burn .’ (3→3′) ‘. burns them.’ (3′→3)

In Ojibwe-Potawatomi, Shawnee, Maliseet-Passamaquoddy, and Cheyenne, the peripheral suffix in- dexes the lower-ranked third person regardless of the definiteness of the arguments. This is illustrated for Southwestern Ojibwe in (35) (Nichols, 1980), where the peripheral suffix indexes the lower-ranked obviative 3′ rather than the higher-ranked proximate 3.

(35) a. owaapamaawaan b. owaapamigowaan o- waapam -aa -waa -an o- waapam -igw -waa -an 3- see - -3 -3′ 3- see - -3 -3′ ‘they. see ’ (3→3′) ‘ sees them.’ (3′→3)

In Meskwaki-Kickapoo, Cree-Innu, Miami-Illinois, Menominee, and Arapaho, the peripheral suffix indexes the higher-ranked third person regardless of the definiteness of the arguments. If the arguments are proximate 3 and obviative 3′, the peripheral suffix indexes the proximate 3 while the obviative 3′ is left unindexed, as shown for Plains Cree in (36) (Wolfart, 1973). If both arguments are obviative 3′, the peripheral suffix indexes the one that is more central to the discourse (i.e. 3′ rather than 3′′), as in (37).

(36) a. wâpamêwak b. wâpamikwak wâpam -ê -w -ak wâpam -ikw -w -ak see - -3 -3 see - -3 -3 ‘they. see ’ (3→3′) ‘ sees them.’ (3′→3)

20 A  

(37) a. wâpamêyiwa b. wâpamikoyiwa wâpam -ê -iyi -w -a wâpam -ikw -iyi -w -a see - - -3 -3′ see - - -3 -3′ ‘ sees  ’ (3′→3′′) ‘  sees ’ (3′′→3′) (e.g. ‘John’s motheri sees heri dog’) (e.g. ‘John’s mother’si dog sees heri’)

The rules that determine the outcome of peripheral agreement in independent indicative TA non- local forms are summarized in (38), where “lower-ranked” and “higher-ranked” refer to the positions of the arguments on the hierarchy 3>3′ >3′′.

(38) Peripheral agreement in TA non-local forms a. Delaware: index lower-ranked argument if definite; if not, index higher-ranked argument b. Ojibwe: index lower-ranked argument c. Cree: index higher-ranked argument

All three rules make reference to a 3 > 3′ > 3′′ hierarchy, although the hierarchy is accessed differently in each rule; it is not simply a matter of always indexing the highest-ranked argument as is typically the case with Algonquian hierarchies. The Delaware rule is itself a hierarchy, as agreement with a lower-ranked definite third is prioritized over agreement with a higher-ranked third regardless of the thematic roles of the arguments. The Delaware rule has been reconstructed for Proto-Algonquian (Goddard, 1967, 1974, 2007); Ojibwe and Cree have simplified the original pattern by retaining only one of the rule’s two clauses.

3.2.8 Summary: Algonquian hierarchies Rather than being tied to particular thematic roles, Algonquian agreement markers are typically gov- erned by hierarchies that establish a ranking of particular nominal features (person, number, obviation, ani- macy). No marker shows a hierarchy that contradicts the ranking SAP>3>3′ >0, so this, if anything, could be considered as the global Algonquian person hierarchy. There is little benefit, however, to adopting a one- size-fits-all approach to hierarchy effects in Algonquian, because the patterning of the agreement markers differs both within and across languages in various ways, such as:

• Which features are included in the hierarchy (see e.g. the absence of SAPs from the conjunct inverse theme sign hierarchy in conservative languages; §3.2.2) • How SAPs are ranked with respect to each other (e.g. 2>1 for the person prefix but 1>2>1/2 for the central suffix; §3.2.3–3.2.4) • The degree to which the hierarchy governs the pattern (see e.g. the conjunct central suffix, which is only partially governed by person/number hierarchies; §3.2.6) • The manner in which the hierarchy is accessed (see e.g. the independent non-local peripheral suffix, which indexes the higher-ranked third in Cree and the lower-ranked third in Ojibwe; §3.2.7)

The notion of a single global Algonquian person hierarchy thus provides only the roughest approximation of the patterning of the agreement markers, and the notion of an invariant 2>1 ranking is simply false.

3.3 Myth: Direct and inverse forms are morphologically symmetrical It is often said that the inflection of corresponding direct and inverse verb forms differs only in the choice of theme sign (e.g. Goddard, 2007: 232). For example, the two Ojibwe forms in (39) (Nichols, 1980) show the same first-person central agreement (ni-…-(i)naan ‘1’) and third-person peripheral agreement (-ik ‘3’) despite the fact that the first person is the actor in (39a) and the undergoer in (39b). The only

21 W O inflectional difference between the two forms is that the 1→3 form in (39a) shows the direct theme sign -aa while the 3→1 form in (39b) shows the inverse theme sign -ikw.

(39) a. niwaapamaanaanik b. niwaapamikonaanik ni- waapam -aa -naan -ik ni- waapam -ikw -inaan -ik 1- see - -1 -3 1- see - -1 -3 ‘we see them’ (1→3) ‘they see us’ (3→1)

It is true that corresponding direct and inverse forms often show this type of symmetry. However, there are also numerous instances in which direct and inverse forms are not inflectionally symmetrical. The Meskwaki counterparts of the Ojibwe forms in (39) are an example (Goddard, 1994). As in Ojibwe, the Meskwaki 1→3 form in (40a) shows the direct theme sign -a· while the 3→1 form in (40b) shows the inverse theme sign -ekw. But there are two further differences: the form of the first-person plural central suffix differs between -pena in (40a) and -ena·n in (40b), and the third-person peripheral suffix -aki ‘3’ appears in (40b) but is absent from (40a), leaving the number of the third-person argument ambiguous.

(40) a. newa·pama·pena b. newa·pamekona·naki ne- wa·pam -a· -pena ne- wa·pam -ekw -ena·n -aki 1- see - -1 1- see - -1 -3 ‘we see her/him/them’ (1→3) ‘they see us’ (3→1)

Asymmetries similar to that in (40) are also found in Kickapoo (Voorhis, 1974), Shawnee (Andrews, 1994), Potawatomi (Hockett, 1966), and Miami-Illinois (Costa, 2003). Several Eastern languages show a related but more complex asymmetry, illustrated for Unami in (41) (data and terminology from Goddard 1969). In the direct, the Unami equivalent to the Ojibwe and Meskwaki 1→3 forms above is a pair of forms: the  1→3 form in (41a) occurs when the third-person undergoer is definite while the  1→3 form in (41b) occurs when the third-person undergoer is indefinite. The objective and absolute forms differ in the presence or absence of a peripheral suffix (-ak ‘3’) and in the form of the first-person plural central suffix (-wəna·n vs. -hməna·). The objective-absolute contrast disappears, however, in the inverse: regardless of the definiteness of the third-person argument, the only option is to use the objective inflection in (41c).14

(41) a. nəmi·la·wəná·nak b. nlo·sá·həna c. nəmi·lko·ná·nak n- mi·l -a· -wəna·n -ak n- lo·sw -a· -hməna· n- mi·l -əkw -wəna·n -ak 1- give - -1 -3 1- burn - -1 1- give - -1 -3 ‘we give to them’ (1→3) ‘we burn one/some’ (1→3) ‘they give to us’ (3→1)

The patterning of agreement in 1→3 and 3→1 forms in Unami, Meskwaki, and Ojibwe can be described as follows. In Unami, the direct (1→3) shows an objective-absolute contrast while the inverse (3→1) is exclusively objective—a pattern that has been reconstructed for Proto-Algonquian (Goddard, 1967, 2007). The inverse form remains objective in Ojibwe and Meskwaki, but in the direct, both languages have neutralized the original objective-absolute contrast: Ojibwe retains only the objective 1→3 form (cognate with the Unami 1→3 form in (41a)) while Meskwaki retains only the absolute 1→3 form (cognate with the Unami 1→3 form in (41b)). This description is summarized in Table 15, which shows the central and peripheral suffixes that appear in each of the inflectional forms discussed above. Both Unami and Meskwaki show a direct-inverse asymmetry: in Unami, the direct differs from the inverse in that absolute and objective forms are both available, while in Meskwaki, where the direct objective

14An objective-absolute contrast does exist when the third person is inanimate (i.e. 0→1), but when the third person is animate (i.e. 3→1), as in all examples in this section, the objective form is the only option (Goddard, 1969: 174).

22 A  

Table 15: Inflection in independent direct 1→3 and inverse 3→1 forms

Language Direct objective form Direct absolute form Inverse (objective) form PA *-wena·n-aki *-hmena *-wena·n-aki Unami -wəna·n-ak -hməna· -wəna·n-ak Ojibwe -inaan-ik —— -inaan-ik Meskwaki —— -pena -ena·n-aki form has been lost, the direct differs from the inverse in that only an absolute form is available. In Ojibwe, on the other hand, the loss of the direct absolute form has created a symmetrical pattern in which the direct and inverse are both objective. The perfect symmetry of direct and inverse forms in Ojibwe is thus in fact an innovation. Direct-inverse asymmetries of the sort displayed by Meskwaki are not a one-off oddity; rather, they are reflections of the original constitution of the independent agreement pattern.

3.4 Myth: All TA conjunct forms contain an overt theme sign Table 16 shows the independent and conjunct inflection of TA direct forms in Southwestern Ojibwe (Nichols, 1980). There is a striking difference between the independent and conjunct paradigms: the direct theme sign -aa appears throughout the independent paradigm, but in the conjunct it can be observed only in the non-local forms (3→3′). None of the conjunct mixed forms (SAP→3) shows a sequence -aa, nor any other sequence that could be identified with a TA theme sign. On the surface, at least, it seems that these forms lack a theme sign altogether.

Table 16: Inflection of Ojibwe TA direct forms

Independent Conjunct 3→3′ o-…-aa-n …-aa-t 3→3′ o-…-aa-waa-n …-aa-waat 1→3 ni-…-aa …-ak 2→3 ki-…-aa …-at 1→3 ni-…-aa-naan …-ankit 21→3 ki-…-aa-naan …-ank 2→3 ki-…-aa-waa …-ek

The pattern in the conjunct inflections in Table 16 goes back to Proto-Algonquian (see reconstructions in Goddard 2000, 2015) and can be observed in many of the languages. There are at least three logically possible analyses of the inflectional structure of an Ojibwe 2→3 conjunct form such as waapamat ‘you (sg) see him/her’. We could take the theme sign to be completely absent, as in (42a). We could take the initial short a of the inflectional ending to be a shortened version of the direct theme sign -aa, as in (42b). Or we could take the direct theme sign to be present in a null allomorph, as in (42c).

(42) a. waapam -at b. waapam -a -t c. waapam -Ø -at see -1:3 see - -3 see - -1:3

The analysis in (42a) can be discarded on the grounds that the role of the theme sign in TA inflection is too fundamental for it to simply be omitted from certain forms. The analysis in (42b) was proposed for the equivalent forms in Shawnee (Andrews, 1994: 198) and comes up relatively frequently in unpublished

23 W O work, but it, too, must be rejected, for several reasons. First, there is no plausible motivation for the direct theme sign -aa to shorten to -a word-medially; word-final position is the only environment in which long vowels are shortened in PA and phonologically conservative languages (Bloomfield, 1946: 93). Second, the morphological material that is left behind when we segment out the putative theme sign -a cannot be coherently analyzed: the 1→3 and 2→3 endings would be segmented as -a-k and -a-t, with only the -k and -t suffixes left to distinguish the first- and second-person actors, but -k and -t are in fact third-person suffixes (see e.g. -t in the 3→3′ form in Table 16) and thus cannot mark the required distinction. Third, the 21→3 ending would be segmented as -a-nk, with a putative 21 central ending -nk, but we can see from other paradigms that it is in fact -ank as a whole that is the 21 central ending; the a thus cannot be passed off as a direct theme sign in this form. Finally, the 2→3 form has the ending -ek, which lacks the segment a altogether and thus requires a different analysis in any case. The correct analysis is that in (42c): the TA conjunct forms that appear to lack a theme sign on the surface actually contain a null allomorph of the direct theme sign (Goddard, 1969: 87). The allomorphy of the direct theme sign is phonologically conditioned: the null allomorph appears whenever the theme sign is followed by a vowel. (Note that in Table 16, the forms that appear to lack a theme sign are precisely those in which the theme sign would be followed by a vowel-initial suffix.) We can restore the overt allomorph of the theme sign in such forms if we add a following consonant-initial suffix such as negative -ssiw, as shown in Table 17 (cf. Rhodes, 1976: 176–177; Goddard, 2006: 189; Quinn, 2006: 259–260; Costa, 2003: 342). The occurrence of the direct theme sign -aa throughout the negative paradigm confirms the validity of positing a null direct theme sign in the corresponding affirmative forms. The occurrence of a-initial central suffixes such as -at ‘2:3’ along with the direct theme sign -aa in the negative forms confirms that the a in the central suffixes is entirely independent from the direct theme sign -aa (contra the analysis in (42b)).

Table 17: Null and overt direct theme signs in Ojibwe TA conjunct forms

Conjunct affirmative Conjunct negative (-ssiw ‘’) 3→3′ …-aa-t …-aa-ssi-k 3→3′ …-aa-waat …-aa-ssi-kwaa 1→3 …-Ø-ak …-aa-ssiw-ak 2→3 …-Ø-at …-aa-ssiw-at 1→3 …-Ø-ankit …-aa-ssiw-ankit 21→3 …-Ø-ank …-aa-ssiw-ank 2→3 …-Ø-ek …-aa-ssiw-ek

The pattern in Table 17 can be observed in several of the languages, as shown in Table 18 for the Miami-Illinois negative (Costa, 2003), the Meskwaki dubitative (Goddard, 1994), the Southwestern Ojibwe negative (Nichols, 1980), the Delaware negative (Goddard, 1969), and the Moose Cree dubitative (Ellis, 1971). In each language, the affirmative form shows the null allomorph of the direct theme sign due to the following vowel-initial suffix -ak ‘1:3’ while the corresponding negative or dubitative form shows the overt allomorph of the theme sign due to the following consonant-initial negative or dubitative suffix. The myth that emerges from this discussion lies in the not-uncommonly-seen analysis in (42b) above, in which the a of the central suffix is incorrectly segmented as a direct theme sign. The impetus for this anal- ysis is presumably the idea that all TA conjunct forms must contain an overt theme sign. Though reasonable, this idea is incorrect. It is possible for the direct theme sign to be null.

24 A  

Table 18: Restoration of overt theme sign in 1sg→3 conjunct forms

Language Affirmative Negative/Dubitative Miami-Illinois -Ø -ak -i -aa -hsiiw -ak -i - -1:3 - - - -1:3 - Meskwaki -Ø -ak -i -a· -w -ak -e·ni - -1:3 - - - -1:3 - Ojibwe -Ø -ak -aa -ssiw -ak - -1:3 - - -1:3 Delaware -Ø -ak -a· -w -ak - -1:3 - - -1:3 Moose Cree -Ø -ak -â -w -ak -wê - -1:3 - - -1:3 -

3.5 Myth: Plural and obviative suffixes are in competition In several Algonquian languages, an animate noun can be marked as either plural or obviative but not both. This is the case in Cree-Innu, most Ojibwe dialects, Potawatomi, Menominee, Delaware, Massachusett, and Cheyenne (Bliss & Oxford, 2017) and is illustrated for Plains Cree in (43) (Wolfart, 1973). The plural form in (43b) must be interpreted as proximate rather than obviative, and the obviative form in (43c) cannot be marked for number and is thus ambiguous between singular and plural.

(43) a. 3 sîsîp ‘duck’ b. 3 sîsîp-ak ‘ducks’ -ak . c. 3′ sîsîp-a ‘duck(s) (obviative)’ -a . One possible analysis of this pattern is as follows. The suffix -ak marks ‘animate plural’ and the suffix -a marks ‘animate obviative’. If an animate noun is notionally both plural and obviative, the plural marker -ak and the obviative marker -a compete for insertion in the suffix slot, and it is the obviative marker that wins the competition. This analysis implies that there is some sort of preference for marking obviation over plurality, or at least some sort of preferential ranking of the -a and -ak suffixes. A different, more conservative pattern is found in Meskwaki-Kickapoo, Shawnee, Miami-Illinois, Maliseet-Passamaquoddy, Penobscot, and Arapaho. In these languages, animate nouns can be marked for number whether they are proximate or obviative, as shown for Miami-Illinois (Costa, 2003) in (44).

(44) a. 3 alaankw-a ‘star’ -a .. b. 3 alaankw-aki ‘stars’ -aki .. c. 3′ alaankw-ali ‘star (obviative)’ -ali .. d. 3′ alaankw-ahi ‘stars (obviative)’ -ahi .. In the Miami-Illinois system, there is no need to posit competition between nominal suffixes: each suffix is specified for a full set of nominal features (animacy, obviative, number), so for any animate noun, there is only one suffix that has the correct combination of features. The structure of the Miami-Illinois system sheds light on that of the Cree system. The conclusion that the Cree suffix -ak marks just ‘animate plural’ may have been hasty. From a paradigmatic perspective, the null suffix that appears on proximate singulars and the -ak suffix that appears on proximate plurals both stand in contrast with the obviative suffix -a. Since they contrast with an obviative suffix, -Ø and -ak are most accurately described as proximate singular and plural suffixes, not just unmarked singular and plural suffixes. This revised analysis is shown in (45).

25 W O

(45) a. 3 sîsîp-Ø ‘duck’ -Ø .. b. 3 sîsîp-ak ‘ducks’ -ak .. c. 3′ sîsîp-a ‘duck(s) (obviative)’ -a . (/)

Under the revised analysis in (45), the notion of competition has dissolved. If a Cree noun is notionally both obviative and plural, there is no competition between -ak and -a, because -ak marks proximate and is thus out of the running. The only marker that can be inserted is obviative -a. The number-neutrality of this marker, rather than arising from competition between -a and -ak, is simply a matter of paradigmatic syncretism—one of a number of such syncretisms that are attested across the family (Bliss & Oxford, 2017). The analysis of -ak as ‘animate proximate plural’ is preferable to its analysis as ‘animate plural’ in two ways. The specification of -ak as proximate more accurately reflects the contrastive structure of the nominal suffix paradigm, in which both -Ø and -ak contrast with obviative -a. And the resulting analysis is simpler in that it does not force us to posit some sort of hierarchy to determine the outcome of the competition between -ak and -a, because under this analysis there is no competition in the first place. The upshot of this discussion is that the nominal suffix should not be regarded as a “plural suffix” or an “obviative suffix”, with potential competition between the two. Rather, it is a suffix that fusionally marks all the features of the noun (animacy, number, obviation). Underspecification effects such as the number-neutrality of Cree obviatives are best understood as instances of paradigmatic syncretism rather than consequences of competition between two morphemes for realization in a single slot.

3.6 Myth: Only animate nominals can be obviative In most Algonquian languages, the inflection of animate nouns shows a proximate-obviative contrast but that of inanimate nouns does not. This is shown for Plains Cree in (46) (Wolfart, 1973): an animate noun such as sîsîp ‘duck’ can be marked as proximate singular, proximate plural, or obviative, but an inanimate noun such as maskisin ‘shoe’ can only be marked as singular or plural.

(46) a. 3 sîsîp-Ø ‘duck’ -Ø .. b. 3 sîsîp-ak ‘ducks’ -ak .. c. 3′ sîsîp-a ‘duck(s) (obviative)’ -a . (/) d. 0 maskisin-Ø ‘book’ -Ø . e. 0 maskisin-a ‘books’ -a . Wolfart (1978) shows, however, that inanimate nouns “participate in obviation covertly” despite not being marked for obviation. Consider the Plains Cree sentences in (47), which involve the inanimate noun cîmân ‘canoe’ (Wolfart, 1978: 259). The sentences differ in that cîmân co-occurs with an SAP in (47a) and with a third person in (47b). If cîmân were animate, it would be proximate in (47a) (as it is the only third person in the clause) and obviative in (47b) (as it contrasts with the proximate third person okimâw ‘chief’). While the form of cîmân itself is the same in both sentences, the verb that modifies it (misâ- ‘be big’) differs: an obviative agreement suffix -iyi is added in (47b), exactly where we would expect cîmân to be obviative.

(47) a. Niwâpahtên cîmân ê-misâk. ni- wâpaht -ê -n cîmân ê- misâ -k 1- see - -1 canoe - be.big -0 ‘I see a canoe that is big.’ (1→0) b. Okimâw wâpahtam cîmân ê-misâyik. okimâw wâpaht -am -w cîmân ê- misâ -iyi -k chief see - -3 canoe - be.big - -0 ‘The chief sees a canoe that is big.’ (3→0′)

26 A  

We may conclude, with Hockett (1966: 60–62), Wolfart (1973: 29, 1978: 259), Rhodes (1976: 199), Goddard (1984: 274), and Drapeau (2014: 337), that inanimates are subject to the proximate-obviative con- trast just as animates are. The difference is that obviation of animate nominals is marked by both noun inflection and verb agreement while obviation of inanimate nominals is marked only by verb agreement. This conclusion may not extend to the entire family, but it does extend to all languages whose verbs have inanimate obviative agreement forms. As Bliss and Oxford (2017) discuss, such languages are numerous:

[I]nanimate obviative verb agreement is widely attested, occurring in Cree (Wolfart 1973), Ojibwe (Rhodes 1976), Meskwaki (Goddard 1994), Cheyenne (Goddard 2000), and early records of Delaware (Goddard [1969]) and Arapaho (Cowell et al. 2014); it has also been reconstructed for Proto-Algonquian (Bloomfield, 1946: 94; Pentland, 1996: 349; Goddard, 2000: 98).

Some dialects of Cree-Innu and Ojibwe have made the system more transparent by developing overt obvia- tive inflection for inanimate nouns (see discussion in Bliss and Oxford 2017), but regardless of the existence of such inflection, the patterning of verb agreement indicates that the proximate-obviative contrast is under- lyingly present on inanimates in all of the languages listed above.

3.7 Myth: There is “further obviative” inflection It is sometimes said that Algonquian languages distinguish not one but two degrees obviation: the obviative (3′) and the further obviative (3′′) (e.g. Hockett, 1966). The putative “further obviative” category arises in recursively possessed nominals such as the Plains Cree example in (48) (Wolfart, 1978: 257), where the possessed noun ‘chief’s son’ functions as the possessor of ‘horse’. Since any noun that is possessed by a third person must be obviative, it is unsurprising that we find the obviative suffix -a on both ‘son’ and ‘horse’. What is interesting, however, is that on the recursively possessed noun ‘horse’, obviative -a is preceded by an additional obviative suffix -iyi.15 In Hockett’s analysis, this -iyi suffix marks a further degree of obviation: if the obviative can be described as “fourth person” (3′), then the further obviative is “fifth person” (3′′).

(48) okimâw okosisa otêmiyiwa okimâw o- kosis -a o- têm -iyi -a chief 3- son -3′ 3- horse - -3′ ‘the chief’s son’s horse’

Wolfart (1978) shows, however, that the “further obviative” analysis is incorrect. If two degrees of obviation truly did exist, we would expect the further obviative to appear not only in possessor obviation contexts such as (48), but also in contexts in which both of the arguments of a transitive verb are obviative. In a sentence such as ‘the chief sees his son killing a fish’, for example, where ‘chief’ is proximate, ‘son’ and ‘fish’ will both be obviative. We would thus expect ‘fish’ to be marked as further obviative in order to distinguish it from its obviative clausemate ‘son’. But, as Wolfart observes, this is not the case: as shown in (49), ‘son’ and ‘fish’ are both marked only with the obviative suffix -a; the putative “further obviative” suffix -iyi does not appear on ‘fish’.

(49) Okimâw wâpamêw okosisa kinosêwa ê-nipahâyit. okimâw wâpam -ê -w [o- kosis -a kinosêw -a ê- nipah -â -iyi -t] chief see - -3 [3- son - fish - - kill - - -3] ‘The chief sees his son killing a fish.’ (Wolfart, 1978: 260)

15I gloss -iyi as  rather than 3′ because there is no contrast between 3, 3′, and 0 in the slot that hosts obviative -iyi, unlike in the slot that hosts obviative -a, which can also host proximate (3) and inanimate (0) suffixes.

27 W O

The contention that obviative inflection distinguishes two degrees of obviation is thus not supported by the data, and a different analysis of the -iyi suffix in (48) is needed. Wolfart (1978) shows that a more parsimonious analysis is available: -iyi in (48) is simply a marker of agreement with an obviative possessor. Analyzing -iyi as obviative possessor agreement explains why it appears in possessor obviation contexts such as (48) but not in clausemate obviation contexts such as (49). Indeed, this is the analysis set out for Proto-Algonquian by Bloomfield (1946: 96), who states simply that “[a]n obv. possessor is indicated by -eliw”. The analysis of -iyi as obviative possessor agreement is supported by the fact that -iyi also appears on verbs as obviative subject agreement, as shown in (50b) (cf. the proximate subject in (50a)).

(50) a. Iskwêw nikamôw. iskwêw nikamo -w woman sing -3 ‘The woman () sings.’ (Wolfart, 1978: 256) b. Okimâw okosisa nikamoyiwa. okimâw o- kosis -a nikamo -iyi -w -a chief 3- son -3′ sing - -3 -3′ ‘The chief’s son () sings.’ (Wolfart, 1978: 260)

Given that nominal possessor agreement and verbal subject agreement show strong formal parallels, as illus- trated for a first-person plural possessor and subject in (51a) and (51b) respectively (Wolfart, 1973), the dual function of -iyi in marking obviative possessors and obviative subjects is exactly what we would expect.

(51) a. nimaskisininân b. ninipânân ni- maskisin -inân ni- nipâ -nân 1- shoe -1 1- sleep -1 ‘our shoe’ ‘we sleep’

Since nominal -iyi is the only inflectional marker that has been claimed to explicitly mark the further obviative category, the analysis of -iyi as obviative possessor agreement removes the putative further obviative from the set of features that can be realized by inflectional morphology. There is one context in which the notion of a “further obviative” remains useful, though as a property of discourse rather than morphosyntax. Normally, when both arguments of a TA verb are third person, the more topical one is proximate and the less topical one is obviative. Direct verb inflection then indicates that the proximate acts on the obviative, as in (52a). However, it can also be the case that both arguments are obviative, as in (52b). Here the direct verb inflection seems to indicate that the more discourse-central obviative (‘the chief’s son’) acts on the more discourse-peripheral obviative (‘the chief’s son’s horse’).16

(52) a. Okimâw wâpamêw okosisa. okimâw wâpam -ê -w o- kosis -a chief see - -3 3- son -3′ ‘The chief () sees his son ().’ (3→3′) (Wolfart, 1978: 260) b. Okimâw okosisa otêmiyiwa wâpamêyiwa. [okimâw o- kosis -a] [o- têm -iyi -a] wâpam -ê -iyi -w -a [chief 3- son -3′] [3- horse -. -3′] see - - -3 -3′ ‘The chief’s son () sees his own horse ().’ (3′→3′′) (Wolfart, 1978: 261)

16See Rhodes 1990a:111–112 for further discussion of two-obviative sentences in Ojibwe.

28 A  

In examples such as (52b) it is convenient to refer to ‘son’ as “nearer obviative” (3′) and ‘horse’ as “further obviative” (3′′), thus allowing the verb inflection to be described as a direct 3′→3′′ form. Notations such as this are often used in TA inflectional paradigms (e.g. Nichols, 1980: 289; Goddard, 1994: 190). It must be remembered, however, that these notations refer only to the discourse roles of the two obviative nominals (such as ‘son’ and ‘horse’ above). As far as the morphosyntax is concerned, both nominals are simply obviative: they both show the obviative suffix -a and the verb shows the obviative subject agreement marker -iyi. In discourse terms, the sentence can be understood as involving a further obviative, but in morphosyntactic terms, the sentence contains no inflection that specifically marks “further obviative”.

3.8 Myth: AI+O verbs are morphologically intransitive This “myth” is in fact a valid statement for some Algonquian languages, but it is not valid as a de- scription of the family as a whole. As background, note that Algonquian verb stems are formed minimally from the combination of an acategorial root and a verbalizing suffix known as a  (Bloomfield, 1946; Goddard, 1990), which marks the animacy of one of the verb’s arguments. Morphologically intransitive finals mark the animacy of the actor argument (AI = animate intransitive, II = inanimate intransitive) while morphologically transitive finals mark the animacy of the undergoer argument (TA = transitive animate, TI = transitive inanimate). Proto-Algonquian stems containing each type of final are exemplified in Table 19.

Table 19: PA verb stems with root *wa·p- ‘light, vision’ (Bloomfield, 1946)

Stem Final *wa·pi- ‘to look on’ (animate actor) -i AI ‘do/be (as animate)’ *wa·pan- ‘to dawn’ (inanimate actor) -an II ‘do/be (as inanimate)’ *wa·pam- ‘to look at someone’ -am TA ‘do to animate’ *wa·pant- ‘to look at something’ -ant TI ‘do to inanimate’

The morphological transitivity of a stem typically lines up with its syntactic transitivity, but this is not always the case. There is a class of “AI+O” verbs (animate intransitive plus object; Goddard 1969) that have intransitive AI finals but nevertheless select both an actor argument and an undergoer argument (see discussions in Bloomfield 1962: 47; Goddard 1969: 37–38; O’Meara 1990: 96–102; Rhodes 1990b; Dahlstrom 2013). These verbs are often described as being morphologically intransitive but syntactically transitive. The validity of this description, however, depends on what is meant by “morphologically intran- sitive”. The morphological transitivity of an Algonquian verb form can be assessed on the basis of at least the three criteria in Table 20. Table 20: Criteria for morphological transitivity

Criterion Morphologically transitive form Morphologically intransitive form Final has TA/TI final has AI/II final Theme sign inflection includes theme sign inflection lacks theme sign Agreement agreement indexes one or two arguments agreement indexes one argument

With respect to the first two criteria, AI+O verbs are indeed morphologically intransitive. They show the same AI finals as do ordinary AI stems, as illustrated for several Plains Cree AI finals in Table 21 (from Tollan and Oxford to appear; data from Bloomfield 1946, Wolfart 1973, Hewson 1993, Dahlstrom 2013). And their inflection lacks a theme sign, as shown by the examples in (53) and (54) further below. However,

29 W O

Table 21: AI finals in Plains Cree AI and AI+O stems

Final AI stem AI+O stem -i ‘do’ nîmi- ‘dance’ kimoti- ‘steal , be a thief’ -isi ‘be’ sêkisi- ‘be scared’ wanikiskisi- ‘forget , be forgetful’ -ihkê ‘make’ sôniyâhkê- ‘make money’ kikâpôhkê- ‘enhance one’s soup with ’ -simo ‘lie’ kawisimo- ‘lie down’ aspisimo- ‘use  for a pillow, bed’ -kwâso ‘sew’ kîsikwâso- ‘finish one’s sewing’ kaskikwâso- ‘sew , sew’ with respect to the third criterion, agreement, AI+O verbs do qualify as morphologically transitive in some languages. Ojibwe AI+O verbs agree with both of their arguments, as exemplified for the Nishnaabemwin dialect in (53a) (Valentine, 2001: 244). Here there is central agreement indexing the first-person actor (ni- …-naan ‘1’) and peripheral agreement indexing the third-person undergoer (-ig ‘3’), just as in the TA form in (53b). (Note, however, that the AI+O form differs from the TA form in that it lacks a theme sign.)

(53) a. nmiigwenaanig b. nwaabmaanaanig ni- miigwe -naan -ig ni- waabam -aa -naan -ig 1- give.away -1 -3 1- see - -1 -3 ‘we give them away’ (AI+O 1→3) ‘we see them’ (TA 1→3)

The double agreement in (53a) is clearly a transitive morphological pattern, so we cannot say that Ojibwe AI+O verbs are entirely morphologically intransitive. The most we can say is that they are morphologically intransitive in some respects (AI final, absence of theme sign) but not all (agreement). In addition to languages with the Ojibwe agreement pattern, there are also languages whose AI+O verbs do not show double agreement, such as Plains Cree. The Cree AI+O form in (54a) shows agreement only with the first-person actor whereas the TA form in (54b) shows agreement with both arguments. The agreement morphology in the AI+O form is thus clearly less transitive than that in the TA form.

(54) a. nikimotinân b. niwâpamânânak ni- kimoti -nân ni- wâpam -â -nân -ak 1- steal -1 1- see - -1 -3 ‘we steal him/her/it/them’ (AI+O 1→3/0) ‘we see them’ (TA 1→3)

In languages that pattern like Cree, AI+O verbs are morphologically intransitive with respect to all three criteria (AI final, no theme sign, no double agreement). In such languages it is entirely valid to state that AI+O verbs are morphologically intransitive and syntactically transitive. In languages of the Ojibwe type this statement must be qualified to acknowledge the transitivity of the agreement morphology.

3.9 Myth: All TI verbs inflect with the TI theme sign All transitive animate (TA) verb stems follow the same overall inflectional pattern, taking the theme signs *-i ‘1’, *-eθ ‘2’, *-a·∼-e· ‘/3’, and *-ekw ‘’ as appropriate (see §3.1, §3.2.1, §3.2.2). The inflection of transitive inanimate (TI) verb stems is less uniform. There are three inflectional classes of TI stems in Proto-Algonquian: TI Class 1 stems inflect with a theme sign *-am (with allomorphs *-a· and *-e·), TI Class 2 stems inflect with a theme sign *-aw (with allomorphs *-a· and *-o·), and TI Class 3 stems

30 A   inflect without a theme sign (Bloomfield, 1962; Goddard, 2015: 378). Examples of each TI stem class are given for Proto-Algonquian in Table 22, along with the corresponding TA stems (Bloomfield, 1946).17

Table 22: TI classes and corresponding TA verbs (Proto-Algonquian)

TI class TI theme sign Inflected TI verb Corresponding TA verb TI1 *-am∼*-a·∼*-e· *wa·pantamwa *wa·pame·wa wa·pant -am -w -a wa·pam -e· -w -a see. -1 -3 -3 see. - -3 -3 ‘s/he sees it’ ‘s/he sees him/her’ TI2 *-aw∼*-a·∼*-o· *ešihta·wa *ešihe·wa ešiht -a· -w -a eših -e· -w -a thus. -2 -3 -3 thus. - -3 -3 ‘s/he makes it so’ ‘s/he makes him/her so’ TI3 no theme sign *na·twa *na·θe·wa na·t -w -a na·θ -e· -w -a fetch. -3 -3 fetch. - -3 -3 ‘s/he fetches it’ ‘s/he fetches him/her’

The TI theme signs index the inanimate object (Goddard, 2007: 240), just as the non-inverse TA theme signs index the person of the animate object (*-i ‘1’, *-eθ ‘2’, *-a·∼-e· ‘/3’). The distinction between the TI1 and TI2 theme signs *-am and *-aw is purely formal and carries no interpretive significance, much like the thematic vowels of Romance verbs. It is also worth noting that the information conveyed by the TI theme signs (i.e. ‘inanimate object’) is entirely redundant, as the TI stem itself indicates that the object is inanimate. (For discussion of the implications of this redundancy, see Piggott, 1989; Pentland, 1999: 236; and Goddard, 2007: 239–40.) Since the TI theme signs are informationally redundant, the absence of a theme sign from the inflection of TI3 verbs does not affect the meaning of the inflected verb form. The description of TI forms given above follows that in Bloomfield’s (1962) grammar of Menominee, which refined his earlier work on Proto-Algonquian (Bloomfield, 1946) and Ojibwe (Bloomfield, 1958). In these earlier works, Bloomfield regarded only TI Class 1 verbs as true TI verbs; he posited a separate class of “pseudo-transitive” verbs that lumped together TI2 verbs, TI3 verbs, and AI+O verbs (i.e. animate intransitive verbs that take an object; §3.8 above). Bloomfield’s earlier analysis is inadequate in two respects. First, the placement of TI2 and AI+O verbs in the same class misses an inflectional distinction: TI2 verbs inflect with a theme sign while AI+O verbs do not (see §3.8). Second, the placement of TI2 and TI3 verbs in the same class as AI+O verbs misses a distinction in selectional properties: TI verbs take an inanimate object while AI+O verbs can take an object of either gender (Pentland, 1999: 231; Goddard, 2015: 406). Despite Bloomfield’s (1962) revision of his approach to TI verbs, his earlier analysis “continues to reverberate in the Algonquianist literature”, as Goddard (2015: 406) puts it. It is often assumed that the TI1 theme sign *-am∼*-a·∼*-e· is the only TI theme sign, and that a verb is a TI verb only if it inflects with this theme sign. As a result of this assumption, TI2 and TI3 verbs are often described as members of the AI+O class rather than as TI verbs. It is, of course, possible that some languages could indeed have reanalyzed TI2 and TI3 verbs as belonging to the AI+O class, but in the absence of evidence for such a reanalysis, the most conservative approach is to recognize separate classes of TI1, TI2, and TI3 verbs, as in Bloomfield (1962), with the latter two classes being distinct from the class of AI+O verbs (Goddard, 1974: 319).

17The stems in Table 22 are from Bloomfield (1946), but the analysis of these stems as TI1, TI2, and TI3 follows Bloomfield’s (1962) grammar of Menominee, which refines the descriptive framework set out in the 1946 sketch.

31 W O

3.10 Myth: The actor argument of a transitive verb must be animate Transitive animate (TA) verbs are defined by the presence of a TA final, which marks the verb as taking an animate undergoer argument. It is sometimes suggested that the actor must be animate as well. It has recently been written, for example, that in both (Arppe, Junker, & Torkornoo, 2017: 52) and Cheyenne (Despić, Hamilton, & Murray, 2017: 168), TA verbs are defined as having an “animate sub- ject and animate object”. In Blackfoot, this is true: inanimate TA actors are disallowed (Frantz, 2009). Inanimate-actor TA forms are scantily attested in Arapaho as well (Cowell & Moss, 2008). In all other lan- guages, however, it is clear that a TA actor may be either animate or inanimate. Paradigms of TA inanimate- actor forms can be found in most comprehensive grammars of Algonquian languages. The contrast between animate-actor and inanimate-actor TA forms is illustrated for Munsee Delaware in (55) (Goddard, 1969).

(55) a. nlo·só·kw b. nlo·só·kwən n- lo·sw -əkw n- lo·sw -əkw -ən 1- burn - 1- burn - - ‘s/he burns me’ (3→1) ‘it burns me’ (0→1)

TA inanimate-actor forms can be difficult to elicit from speakers. To illustrate some contexts in which these forms may appear, textual examples from Delaware and Massachusett are given in (56) and (57).18

(56) Delaware TA inanimate-actor examples (Goddard, 1969: 159) a. nto·khika·k·ó·ne·n ni kəmant·o·wá·k·an ‘your power wakes us up’ b. wəla·te·namowá·k·an nəməšhika·khóməna ‘gladness comes over us’ c. ntəmšó·kwə̆ ne·n yə paxkší·kan ‘this knife cut us’ (57) Massachusett TA inanimate-actor examples (Goddard & Bragdon, 1988: 521) a. wenuhkunkq ‘they (inan.) are round about him’ b. nootammantahikkummin ‘it troubles us’ c. ogquehchippánukquog ‘they (inan.) wet them’ d. nuttinne wauwaonukqunash ‘they (inan.) bear witness of me’ e. wunishkhukqunash ‘they (inan.) defile him’

The foregoing discussion applies to transitive animate (TA) verbs, which take an animate undergoer. For transitive inanimate (TI) verbs, which take an inanimate undergoer, the empirical picture is much less clear. At the very least we can say that inanimate actors are far less common with TI verbs than they are with TA verbs. In Ojibwe, inanimate actors are banned from TI forms entirely (Rhodes, 1990b: 402–403). In at least two languages, however, TI inanimate-actor forms—or a functional equivalent—are attested. In Moose Cree (Ellis, 1971: 86) and Meskwaki (Goddard, 1994: 190), the addition of a suffix -makan/-mikat following the TI theme sign creates a form that takes an inanimate actor, as illustrated in (58) for Moose Cree and (59) for Meskwaki (cf. the ordinary TI animate-actor forms in the (a) examples).

(58) a. wâpahtam b. wâpahtamômakan c. wâpahtamômakanwa wâpaht-am-w wâpaht-amô-makan-w wâpaht-amô-makan-w-a see.it - -0 see.it - - -0 see.it - - -0 -0 ‘s/he sees it’ (3→0) ‘it sees it’ (0→0) ‘they (inan.) see it’ (0→0)

18Since the purpose of these examples is just to illustrate the types of contexts in which TA inanimate-actor forms may be used, morpheme segmentations are not provided.

32 A  

(59) a. wa·patamwa b. wa·patamo·mikatwi wa·pat -am -w -a wa·pat -amo· -mikat -w -i see.it - -3 -3 see.it - - -0 -0 ‘s/he sees it’ (3→0) ‘it sees it’ (0→0)

The ‘it sees it’ forms in (58)–(59) serve the function that would be expected of TI inanimate-actor inflection, but there is evidence that these forms do not in fact belong to the TI inflectional paradigm. The -makan/-mikat marker that characterizes these forms is actually a derivational final that ordinarily creates inanimate intransitive (II) verb stems in which “inanimate subjects have taken on the properties of animates” (Thomason, 2003: 36). This derivational function is illustrated for Meskwaki in (60): -mikat derives the II stem in (60b) from the animate intransitive (AI) stem in (60a) (data from Thomason 2003: 36).

(60) a. menwa·čimowa b. menwa·čimo·mikatwi menwa·čimo -w -a menwa·čimo -mikat -w -i narrate.well -3 -3 narrate.well - -0 -0 ‘s/he tells good stories’ (AI 3) ‘it tells good stories’ (II 0)

Since -makan/-mikat is an II final, the ostensible TI inanimate-actor forms in (58) and (59) actually seem to involve derived II verb stems rather than true TI stems, as concluded by Thomason (2003: 113, fn. 28) and implied by Ellis (1971: 86). That is, in a Meskwaki form such as wa·patamo·mikatwi ‘it sees it’, inflection is added not to the TI stem wa·pat- ‘see something’ but rather to the derived II stem wa·patamo·mikat- ‘see something’—the difference being that the TI stem takes an animate actor while the II stem takes an inanimate actor. Derived II stems such as wa·patamo·mikat- ‘see something’ differ from typical II stems, however, in that they take an undergoer argument in addition to an actor argument (Thomason, 2003: 113), and could perhaps thus be regarded as “II+O” verbs by analogy with the better-known AI+O verbs (§3.8). In summary, it is incorrect to state that Algonquian transitive verbs must have animate actors, as most of the languages allow the actor of a TA verb to be either animate or inanimate (with Blackfoot as a notable exception). For TI verbs, however, the limited available evidence indicates that the actor is indeed required to be animate. Ojibwe TI verbs are ungrammatical with inanimate actors, and Moose Cree and Meskwaki require the use of a derived II+O stem rather than a true TI stem in inanimate-actor contexts.

4 Conclusion

This paper has attempted to shine a light on various obscure corners of the Algonquian morphosyn- tactic system. In many cases, a generalization that accurately describes the most readily available data (e.g. the SAP > 3 > 3′ > 0 hierarchy that governs inverse marking in the independent order) breaks down when confronted with data from lesser-known parts of the system (e.g. the absence of an SAP>3 ranking in the conjunct order). Theoretical analyses can benefit from taking into account the full empirical picture as set out in the most rigorous descriptive works such as Bloomfield (1962) and Goddard (1969).

References

Andrews, K. R. (1994). Shawnee grammar (Unpublished doctoral dissertation). University of South Car- olina. Arppe, A., Junker, M.-O., & Torkornoo, D. (2017). Converting a comprehensive lexical database into a computational model: The case of East Cree verb inflection. In Proceedings of the 2nd Workshop on the Use of Computational Methods in the Study of Endangered languages (pp. 52–56). Association for Computational Linguistics.

33 W O

Aubin, G. F. (1989). Some verb paradigms in Golden Lake Algonquin: II. In Papers of the 20th Algonquian Conference (pp. 1–16). : Carleton University. Béjar, S., & Rezac, M. (2009). Cyclic Agree. Linguistic Inquiry, 40, 35–73. Béland, J.-P. (1978). Atikamekw morphology and lexicon (Unpublished doctoral dissertation). University of California, Berkeley. Bhatia, S., Kusmer, L. P., & Vostrikova, E. (2016). Indirect interaction of person and number in Ojibwe. (Presented at WSCLA 21, UQAM, Montreal, April 2) Blain, E. M. (1998). The role of hierarchies and alignment in direct/inverse. In D. H. Pentland (Ed.), Papers of the 29th Algonquian Conference (pp. 53–56). Winnipeg: University of Manitoba. Bliss, H., & Oxford, W. (2017). Patterns of syncretism in nominal paradigms: A pan-Algonquian perspec- tive. In M. Macaulay & M. Noodin (Eds.), Papers of the 46th Algonquian Conference. East Lansing: State University Press. Bliss, H., Ritter, E., & Wiltschko, M. (2014). A comparative analysis of theme marking in Blackfoot and Nishnaabemwin. In J. R. Valentine & M. Macaulay (Eds.), Papers of the 42nd Algonquian Conference (pp. 10–33). Albany: SUNY Press. Bloomfield, L. (1946). Algonquian. In H. Hoijer (Ed.), Linguistic structures of Native America (pp. 85–129). New York: Viking Fund Publications in Anthropology. Bloomfield, L. (1958). Eastern Ojibwa: Grammatical sketch, texts and word list. Ann Arbor: University of Michigan Press. Bloomfield, L. (1962). The Menomini language. New Haven: Yale University Press. Brittain, J. (1999). A reanalysis of transitive animate theme signs as object agreement: Evidence from Western Naskapi. In D. H. Pentland (Ed.), Papers of the 30th Algonquian Conference (pp. 34–46). Winnipeg: University of Manitoba. Brouillard, E., & Dumont-Anichinapeo, M. (1987). Grammaire algonquine respectant les règles langagières et coutumières propres aux communautés du Lac Simon ainsi que du Grand Lac Victoria. Lac Simon, QC: Centre culturel Amikwan. Bui, T. (2016). Menominee agreement: Two probes for two hierarchies. In Proceedings of ConSOLE XIV (pp. 414–427). Cenerini, C. (2017). The prominence of speech-act participants: A study of diachronic change in Algonquian person agreement. (PhD generals paper, University of Manitoba) Clarke, S. (1982). North-West River (Sheshātshīt) Montagnais: A grammatical sketch. Canadian Ethnology Service Paper no. 80. Ottawa: National Museums of . Costa, D. J. (2003). The Miami-Illinois language. Lincoln, NE: University of Nebraska Press. Cowell, A., Alonzo Moss, S., & C’Hair, W. J. (2014). Arapaho stories, songs, and prayers: A bilingual anthology. Norman, OK: University of Press. Cowell, A., & Moss, A., Sr. (2008). The . Boulder, CO: University Press of Colorado. Cyr, D. (1996). La lune est notre grand-mère: la notion de vérité comme principe assimilateur. Recherches amérindiennes au Québec, 26, 43–51. Dahlstrom, A. (1989). Morphological change in Plains Cree verb inflection. Folia LInguistica Historica, 9(2), 59–71. Dahlstrom, A. (1991). Plains Cree morphosyntax. New York: Garland. Dahlstrom, A. (2013). Argument structure of quirky Algonquian verbs. In T. H. King & V. de Paiva (Eds.), From quirky case to representing space: Papers in honor of Annie Zaenen (pp. 61–71). Stanford: CSLI. Déchaine, R.-M. (1999). What Algonquian morphology is really like: Hockett revisited. In L. Bar-el, R.-M. Déchaine, & C. Reinholtz (Eds.), Papers from the Workshop on Structure and Constituency in Native American Languages (pp. 25–72). MIT Occasional Papers in Linguistics 17. Despić, M., Hamilton, M. D., & Murray, S. E. (2017). Multiple and cyclic agree: Person/number marking in

34 A  

Cheyenne. In A. Kaplan, A. Kaplan, M. K. McCarvel, & E. J. Rubin (Eds.), Proceedings of WCCFL 34 (pp. 167–176). Somerville, MA: Cascadilla. Drapeau, L. (2014). Grammaire de la langue innue. Québec: Presses de l’Université du Québec. Dryer, M. S. (1996). Passive vs. indefinite actor construction in Plains Cree. In D. H. Pentland (Ed.), Papers of the 27th Algonquian Conference (pp. 54–79). Winnipeg: University of Manitoba. Ellis, C. D. (1971). Cree verb paradigms. International Journal of American Linguistics, 37, 76–95. Fidelholtz, J. L. (1968). Micmac morphophonemics (Unpublished doctoral dissertation). MIT. Frantz, D. G. (2009). Blackfoot grammar (Second ed.). Toronto: University of Toronto Press. Goddard, I. (1967). The Algonquian independent indicative. In A. D. DeBlois (Ed.), Contributions to anthropology: Linguistics I (Algonquian) (pp. 66–106). National Museum of Canada Bulletin 214, Ottawa. Goddard, I. (1969). Delaware verbal morphology: A descriptive and comparative study (Unpublished doctoral dissertation). Harvard. (Published 1979, Garland, New York) Goddard, I. (1974). Remarks on the Algonquian independent indicative. International Journal of American Linguistics, 40, 317–327. Goddard, I. (1984). The obviative in Fox narrative discourse. In W. Cowan (Ed.), Papers of the 15th Algonquian Conference (pp. 273–286). Ottawa: Carleton University. Goddard, I. (1990). Primary and secondary stem derivation in Algonquian. International Journal of Amer- ican Linguistics, 56, 449–83. Goddard, I. (1994). ’s Fox lexicon: Critical edition. Winnipeg: Algonquian and Iroquoian Linguistics Memoir 12. Goddard, I. (2000). The historical origins of Cheyenne inflections. In J. D. Nichols (Ed.), Papers of the 31st Algonquian Conference (pp. 77–129). Winnipeg: University of Manitoba. Goddard, I. (2006). The Proto-Algonquian negative and its descendants. In H. C. Wolfart (Ed.), Actes du trente-septième Congrès des Algonquinistes (pp. 161–208). Winnipeg: University of Manitoba. Goddard, I. (2007). Reconstruction and history of the independent indicative. In H. C. Wolfart (Ed.), Papers of the 38th Algonquian Conference (pp. 207–271). Winnipeg: University of Manitoba. Goddard, I. (2015). Arapaho historical morphology. Anthropological Linguistics, 57, 345–411. Goddard, I., & Bragdon, K. J. (1988). Native writings in Massachusett. American Philosophical Society Memoir 185, Philadelphia. Heath, J. (1998). Pragmatic skewing. International Journal of American Linguistics, 64, 83–104. Hewson, J. (1993). A computer-generated dictionary of Proto-Algonquian. Ottawa: Canadian Museum of Civilization. Hockett, C. F. (1966). What Algonquian is really like. International Journal of American Linguistics, 32, 59–73. Hockett, C. F. (1992). Direction in the Algonquian verb: A correction. Anthropological Linguistics, 34, 311-315. Jancewicz, W. J. (2013). Grammar enhanced biliteracy: structures for facilitating read- ing in Naskapi (Unpublished master’s thesis). University of North Dakota. Jones, D. (1977). A basic Algonquin grammar. Maniwaki: River Desert Band Council. Lochbihler, B. (2012). Aspects of argument licensing (Unpublished doctoral dissertation). McGill, Montreal, QC. Macaulay, M. (2005). On the 2>1 prominence hierarchy of Algonquian. In Proceedings of WIGL 2005 (pp. 1–24). Macaulay, M. (2009). On prominence hierarchies: Evidence from Algonquian. Linguistic Typology, 13, 357–389. McGinnis, M. (1995). Word-internal : Evidence from Ojibwa. In P. Koskinen (Ed.), Proceedings of the 1995 CLA Annual Conference (pp. 337–347).

35 W O

McGinnis, M. (1999). Is there syntactic inversion in Ojibwa? In Papers from the Workshop on Structure and Constituency in Native American Languages (pp. 101–118). Cambridge, MA: MITOPL. McGinnis, M. (2005). On asymmetries in person and number. Language, 81, 699–718. McGinnis, M. (2008). Phi-feature competition in morphology and syntax. In D. Harbour, D. Adger, & S. Béjar (Eds.), Phi theory: Phi-features across modules and interfaces (pp. 155–184). Oxford: Ox- ford University Press. Muehlbauer, J. (2012). The relation of switch-reference, animacy, and obviation in Plains Cree. International Journal of American Linguistics, 78, 203–238. Nichols, J. D. (1980). Ojibwe morphology (Unpublished doctoral dissertation). Harvard. O’Meara, J. (1990). Delaware stem morphology (Unpublished doctoral dissertation). McGill. Pentland, D. H. (1996). Verbs of precipitation in Algonquian. In J. D. Nichols & A. C. Ogg (Eds.), Nikotwâsik iskwâhtêm, pâskihtêpayih! : Studies in honour of H.C. Wolfart (pp. 337–353). Winnipeg: Algonquian and Iroquoian Linguistics. Pentland, D. H. (1999). The morphology of the Algonquian independent order. In D. H. Pentland (Ed.), Papers of the 30th Algonquian Conference (pp. 222–266). Winnipeg: University of Manitoba. Piggott, G. L. (1989). Argument structure and the morphology of the Ojibwa verb. In D. B. Gerdts & K. Michelson (Eds.), Theoretical perspectives on Native American languages (pp. 176–208). Albany: SUNY Press. Proulx, P. (1978). Micmac inflection (Unpublished doctoral dissertation). Cornell. Quinn, C. (2006). Referential-access dependency in Penobscot (Unpublished doctoral dissertation). Har- vard. Quinn, C. (2015). Recovering the Penobscot TA conjunct: Documentary, analytical, and revitalization considerations. (Paper presented at the 47th Algonquian Conference, Winnipeg, October 24) Rhodes, R. A. (1976). The morphosyntax of the Central Ojibwa verb (Unpublished doctoral dissertation). Michigan. Rhodes, R. A. (1990a). Obviation, inversion, and topic rank in Ojibwa. In Proceedings of BLS 16, supplement (pp. 101–115). Rhodes, R. A. (1990b). Ojibwa secondary objects. In K. Dziwirek, P. Farrell, & E. Mejías-Bikandi (Eds.), Grammatical relations: A cross theoretical perspective (pp. 401–414). Stanford: CSLI. Rhodes, R. A. (1994). Agency, inversion, and thematic alignment in Ojibwe. In Proceedings of BLS 20 (pp. 431–446). Rhodes, R. A., & Todd, E. M. (1981). Subarctic Algonquian languages. In J. Helm (Ed.), Handbook of North American Indians, volume 6: Subarctic (pp. 52–66). Washington, DC: Smithsonian. Rogers, J. H. (1964). Survey of Round Lake Ojibwa phonology and morphology. In Contributions to anthropology, 1961–62, part II (pp. 92–154). National Museum of Canada Bulletin 194, Ottawa. Sherwood, D. F. (1986). Maliseet-Passamaquoddy verb morphology. Canadian Ethnology Service Paper no. 105. Ottawa: National Museums of Canada. Starks, D. (1992). Aspects of Woods Cree syntax (Unpublished doctoral dissertation). University of Mani- toba. Thomason, L. (2003). The proximate and obviative contrast in Meskwaki (Unpublished doctoral disserta- tion). University of Texas at Austin. Todd, E. M. (1970). A grammar of the Ojibwa language: The Severn dialect (Unpublished doctoral disser- tation). University of North Carolina, Chapel Hill. Tollan, R., & Oxford, W. (to appear). Voice-less unergatives: Evidence from Algonquian. In Proceedings of WCCFL 35. Somerville, MA: Cascadilla. Valentine, J. R. (2001). Nishnaabemwin reference grammar. Toronto: University of Toronto Press. Voorhis, P. H. (1967). Kickapoo grammar (Unpublished doctoral dissertation). Yale University, New Haven, CT.

36 A  

Voorhis, P. H. (1974). Introduction to the Kickapoo language. Bloomington: University. Wolfart, H. C. (1973). Plains Cree: A grammatical study. Transactions of the American Philosophical Society, New series, Volume 63, Part 5. Philadelphia. Wolfart, H. C. (1978). How many obviatives: Sense and reference in a Cree verb paradigm. In E.-D. Cook & J. D. Kaye (Eds.), Linguistic structures of Native Canada (pp. 255–272). Vancouver: UBC Press. Wolvengrey, A. (2011). Semantic and pragmatic functions in Plains Cree syntax (Unpublished doctoral dissertation). University of Amsterdam. Xu, Y. (2016). Feature competition in Algonquian agreement (Unpublished master’s thesis). University of Manitoba. Zúñiga, F. (2006). Deixis and alignment: Inverse systems in Indigenous languages of the Americas. Ams- terdam/Philadelphia: Benjamins. Zúñiga, F. (2008). How many hierarchies, really? Evidence from several Algonquian languages. In M. Richards & A. L. Malchukov (Eds.), Scales (pp. 277–294). Universität Leipzig.

37