Optimizing Display Other Formats:      PDF - 318k

Volume 1 Issue 1 (2002)        DOI:10.1349/PS1.1537-0852.A.21

Note: Linguistic Discovery uses Unicode characters to represent phonetic symbols. Please see Optimizing Display for requirements to accurately reproduce this page.

NOTES ON AGREEMENT IN ITELMEN[*]

Jonathan David Bobaljik and Susi Wurmbrand
McGill University

To Ken Hale

Agreement in Itelmen is represented by means of both prefixes and suffixes. While the prefixes reference subjects (of both transitive and intransitive verbs), the suffixal agreement morphemes on a given verb may reference the subject, the object, or an oblique argument, or some combination of these. We propose that the proper characterization of the factors that determine which arguments control suffixal agreement involves a division of labour between morphology and a notion of discourse prominence/salience. In essence, we propose that the suffixal agreement morpheme is an object agreement marker, but the features of the subject are reflected in this position when the object lacks the relevant features (for example, we treat third person as the lack of a person feature), or is absent altogether (thus, intransitive verbs agree twice with their subjects). When a verb occurs with an oblique as well as a direct object, discourse salience will determine which of these non-subject arguments will control object agreement. In addition to providing a description of a complex range of facts from Itelmen, the paper sheds light on the nature of “multiple exponence” and the role of “competition” among affixes for a particular position in the verbal agreement system.

1. Introduction

One of the major goals of morphological analysis is to identify and account for cooccurrence restrictions among morphemes. Once an instance of competition among morphemes or features is identified, it is not obvious a priori what component of the grammar is responsible for the account. In this paper, we examine a complex system of agreement suffixes on verbs and ultimately come to the conclusion that some of the cooccurrence restrictions observed arise from certain interactions between syntax/discourse and morphology, while other restrictions are the result of formal morphological considerations.

The example of competition we will focus on here is taken from the verbal agreement suffixes in Itelmen, a Chukotko-Kamchatkan language spoken on Russia’s Kamchatka peninsula. As will be shown presently, the suffixal agreement morpheme on a given Itelmen verb may cross-reference (i.e., agree with) the subject, the object, or an oblique argument (and sometimes a combination of these). Competition arises when there is more than one potential antecedent or controller for agreement, but where not all features of each controller may find expression on the verb. To take a concrete example (to which we return in detail below), the sentences in (1)-(2) illustrate four forms of the verb zəl- ‘give’. (Here and throughout, the verb is indicated in boldface.) [1]

In (1a), the suffix agrees only with the direct object (underlined in the English paraphrase). The example in (1b) is minimally different, in that the verb shows 1SG agreement with the indirect object, i.e., the dative marked goal. Note in particular that there is no change in case-marking, word order or argument (theta-) structure in these examples.

The pair in (2) involve the same direct and indirect objects; the change in person and number of the subject (and mood) in this case controls the shape of the suffix. Taken together, these examples show that the features of all three arguments (subject, direct and indirect object) are potentially expressible in the agreement suffix, but that at the same time they compete for expression: in no case are the features of all three arguments ever simultaneously expressed.

Our focus in this paper is to identify the factors determining which argument(s) will control agreement and to understand in what way these factors interact. To some degree, our primary goal in this paper is therefore descriptive. However, the generalizations which we bring to light have implications for current morphological theories and for theories of the interaction of morphology and syntax.

In a nutshell, the analysis we will propose is as follows. First, we will argue that the competition between the direct object and the subject (e.g., (1a) vs. (2a-b)) is entirely morphological in nature. That is, the outcome is determined solely by consideration of the feature arrays and by appeal only to morphological principles. Specifically, we propose that the exponence of features of the subject at the suffix position is secondary, arising in two cases: (i) when there is no object (intransitives) or (ii) when the object is deficient in its features (3rd person direct objects). By contrast, the competition between direct and indirect object (as in (1a) vs. (1.b)) is—we suggest—not morphological in nature but rather governed by discourse-pragmatic considerations such as salience.

The paper is organized as follows. In section 2, we provide some brief background about Itelmen and an overview of the general agreement facts. In section 3, we look more carefully at the nature of the agreement suffixes and argue for the morphological treatment of the object-subject competition noted above. We also note implications of our analysis for current theories of morphology. In section 4, we turn to the alternation in agreement between direct objects and indirect objects or oblique arguments. Section 4.2 focuses on a striking difference between Itelmen and the related languages Chukchi and Alutor regarding the role of a person hierarchy in the account. In addition to providing a description of an intricate and complex set of facts from Itelmen, the present paper fits into a broader investigation of the possible bounds of morphology—an inquiry into how much of the paradigmatic aspect of agreement systems is idiosyncratic and how much is systematic. Though we offer this paper as a contribution to a more general understanding of the nature of agreement, we have refrained from working out the details of the analysis in any specific framework. Given the nature of the topic, we feel that it is more beneficial to state as clearly as possible the generalizations that emerge from the data, rather than to force the data into the mould of any particular formalization.

2. Background on Agreement in Itelmen

Before turning to the relevant data and the proposed analysis, it is perhaps useful for the uninitiated to provide a few pieces of background information about Itelmen grammar.

2.1. Itelmen

Itelmen[2] (formerly known also as Kamchadal) is spoken natively by fewer than a hundred people scattered throughout a handful of villages on the west coast of Russia’s Kamchatka peninsula. The language is undoubtedly related to its northern neighbours Chukchi, Koryak, Alutor, and Kerek; together, these five languages constitute the Chukotko-Kamchatkan group (details of the relationship of Itelmen to the others remain to be worked out). Of the three varieties of Itelmen identified by S.P. Krasheninnikov in the 18th century, (Krasheninnikov 1755/1994:137) only the Western variety has survived, itself divided into two major dialect groups, the Northern (spoken in Sedanka and Tigil) and the Southern (spoken in other villages, of which only Kovran and Khairiuzovo remain). Except as noted, the data in this paper is primarily from the Northern dialect; we are not aware of any differences between the dialects that affect the analysis presented here.

There are currently two monograph-length descriptions of Itelmen (both based on the Southern dialect group): Volodin (1976) and Georg & Volodin (1999). Some aspects of Itelmen grammar which will be useful in understanding the examples discussed below include the following:

The basic Itelmen verb consists of a verb stem, various inflectional material, and—most peripherally—agreement affixes: a prefix (sometimes null) and one or more agreement suffixes. See (4):

(4)     agreement.1 — mood — verb stem — aspect — tense — agreement.2

In addition, it should be noted that Itelmen has no passive on the Indo-European model (i.e., demotion of subject to adjunct and concomitant promotion of object to subject). The closest analogue in Itelmen is what we term the “impersonal” construction illustrated in (1). The characteristics of this construction are as follows. The agent (if expressed) bears locative case. The verb takes an invariant prefix n- (realis) or xɐn- (irrealis), regardless of the number of the agent. The sole caseless argument (patient) nevertheless continues to trigger object agreement (not subject agreement as one would expect in a true passive), but the forms of the object agreement affix are not those which are conditioned by a third person subject (where such forms exist, see below). The impersonal construction is frequently—though not obligatorily—used when the agent of the action is third person and the patient (or goal) is first or second person, and is hence sometimes called “inverse”.

As our focus in this article is the feature content of the agreement affixes, and the factors determining their distribution, we turn to these issues now.

2.2 The agreement affixes

In Itelmen, as in the other Chukotko-Kamchatkan languages, agreement is expressed by means of obligatory prefix-suffix combinations.[5] The basic pattern is illustrated for transitive verbs in (5) and for intransitive verbs in (6) (fuller paradigms are given below).

Cursory inspection of these examples shows that the prefix clearly indicates the features of the subject (and mood), remaining constant across transitive and intransitive constructions (for example: /t-/[6] indicates a first person singular subject, regardless of the transitivity of the verb).[7] The suffix in these examples indicates the features of the object of the transitive verbs (as in (5)) but the subject of the intransitive verbs (see (6)). Note that as a result, the subject is referenced twice in intransitive verbs—once by the prefix and again by the suffix. This double agreement is a curious and often commented upon aspect of all of the Chukotko-Kamchatkan languages, and has led some previous researchers (to greater or lesser degrees) to describe the Chukotko-Kamchatkan agreement system as manifesting a unique type of “split” ergativity: the prefixes showing a nominative (i.e., vs. accusative) alignment while the suffixes are taken to show an absolutive (vs. ergative) bias (see Nedjalkov 1979, Comrie 1979b, 1981, Spencer 1996, Halle & Hale 1997).[8] In section 3 we will argue that the apparent absolutive alignment is epiphenomenal and does not reflect any deep property of the language; presupposing this argument, we will at this point continue to refer to the prefixes as subject agreement and the suffixes as (primarily) object agreement markers.

The full intransitive agreement paradigm for Itelmen is given in (7). Each of the cells in this table has two forms; in each case, the higher form is the realis (i.e., indicative) mood and the lower form is the irrealis mood. The solid underline indicates the position of the verb stem relative to the prefix and suffix.[9]

The transitive agreement paradigm for Itelmen is far more intricate and is given in two tables. At first blush, the tables seem quite complex, and the reader is invited not to try to absorb all of the details at this point. As we proceed, we will illustrate each of our points by highlighting the relevant affixes in the main text. Thus, we include the tables primarily for the sake of reference, and to give a sense of how the various pieces of inflection fit together.

The table in (8) (=Table 2) gives the paradigm for verbs agreeing with all first and second person objects and with third person obliques. The irrealis/realis distinction is indicated in the same manner as in (8). Since third person direct objects add substantial complexity to the paradigm they are listed separately in Table 3 ((9) below).[10] Note also that Itelmen transitive verbs are divided into two (arbitrary) conjugation classes (see Volodin 1976:205). With the exception of the highlighted forms in Table 3, conjugation class II markers are in parentheses—they alternate with zero in class I. Though interesting in their own right, the class markers are somewhat orthogonal to the main thread of the analysis, and they are not discussed here. See Bobaljik & Wurmbrand (1997) for an overview and Bobaljik (2000) for remarks on the theoretical significance of the class markers.

Having presumably overwhelmed the reader with data, we will now begin to examine various pieces of the paradigm in an attempt to answer the question: what conditions the form of the affixes, in particular the agreement suffixes, in Itelmen?

3. What do the Agreement Affixes Agree with?

In this section, we present our analysis of the factors that govern agreement, especially in the suffix position of the Itelmen verb. First, we will argue that the apparent “absolutive” nature of the suffixes mentioned above is epiphenomenal and that there are no arguments for treating the suffixes in this manner. In section 3.1, we report the observation of Volodin & Vakhtin (1986) that there are no individual affixes which can be said to have an absolutive distribution. We then turn to a more intricate concern, namely, third person direct objects. Exactly where one might have expected the paradigm to be simplest turns out to be where it is most complex. Nevertheless, we will argue that there is order in this part of the paradigm; in brief, we will demonstrate that exactly when the direct object is third person, the features of the subject must be consulted in order to determine the correct form of the agreement suffix. We suggest that this is due to a featural deficiency of third person objects (they have number features only). More importantly, we submit that the assumptions needed to provide an account of the third person direct objects will carry over straightforwardly to provide an account of the apparent double exponence of subject agreement in intransitive constructions.

3.1 Against absolutive agreement

When considering the examples in (5) and (6) it is easy to see the motivation for analyzing the prefixes as reflecting a nominative alignment (transitive and intransitive subjects), while treating the suffixes as markers of absolutive agreement (direct objects and intransitive subjects). Nevertheless, there is a difference between the two groups of agreement affixes, first noted by Volodin & Vakhtin (1986). For the prefixes, the nominative alignment is quite clear. For example, comparing (5a) with (6a), we see that the prefix t- indicates a first person singular subject in the indicative mood, regardless of the transitivity of the verb. Similarly, q- represents a second person subject in what we will call the irrealis mood (Volodin 1976’s “imperative mood” (pobuditel’noe-povelitel’noe naklonenie); its functions include imperative, exhortative, negative future, and potential future). Turning to the suffixes, Volodin & Vakhtin (1986) point out that at the level of the individual affixes, the putative absolutive alignment of the suffixes breaks down. Despite the apparent alignment of the suffix position along absolutive lines (grouping together intransitive subject and transitive object, to the exclusion of transitive subjects), there are no truly absolutive suffixes. For instance, the suffix corresponding to a [1sg] intransitive subject (=“S”) is -kičen, while that corresponding to a [1sg] object (=“O”) is -βum. The unity of transitive subject (=“A”) and intransitive subject prefixes—and the lack of a corresponding unity of intransitive subject and object suffixes—is quite general as can be seen from the more complete paradigm in (10).[11]

(10)

Not only are there no parallels between the shape of the affixes indicating a given person/number combination across S and O functions, to the extent there are parallels to be made across these columns (including the third person object suffixes), these in fact represent a subject-object alignment. For example, the suffix -(kīčen occurs in both the intransitive paradigm and the transitive paradigm (see (9), above). In the intransitive paradigm, this suffix indicates a first person subject. In the transitive paradigm, it does not indicate a first person object, but rather indicates a third person direct object, in the environment of a first person subject. Similarly, the bare suffix -sx (i.e., as opposed to the complex form -sxen) indicating a second person plural subject in the intransitive paradigm occurs in the transitive agreement paradigm indicating a second person plural subject (see especially the form with a third person singular direct object in (9)), again a nominative patterning, rather than an absolutive one. To the extent these observations have any synchronic significance, it is as further support for the observation that the distribution of the individual affixes is best described by the functions of subject versus object rather than absolutive.[12]

That the similarities between the intransitive and transitive paradigms arise most clearly when the object is third person singular is surely no accident. This brings us to the next set of observations which lead us to reject the characterization of the agreement suffixes as “absolutive.”

A second reason not to treat the suffixes as simply displaying absolutive agreement comes from an examination of a broader array of forms. Consider first examples of transitive verbs with third person objects, as in (11), (cf., Table 3, above):

As these examples illustrate, when the direct object is third person, various factors conspire to determine the shape of the agreement suffix. The contrast between (11a) and (11b) shows that the number of the object is relevant (a plural object triggers glottalisation of the final n); but the contrast between (11b) and (11c) shows that the person of the subject is also relevant. The form in (11d) shows that it is not possible to systematically segment the suffixes into two pieces, e.g., treating the -č- in (11a-b) as a first person affix. The agreement suffix for a third person singular object in the context of a second person singular subject in the irrealis mood is the not plausibly segmentable -x(ʷ). The examples in (12) further illustrate the relevance of features other than those of the direct object in determining the shape of the agreement suffix when the direct object is third person (again, see Table 3, above).

In (12a), the complex suffix -sxiʔn [2pl>3pl] references features of both the subject and (direct) object. In (12b), only the features of the transitive subject [2pl] are expressed by the suffix -sx. And finally, in (12c-d) we have a minimal pair in which mood affects the choice of agreement suffix with a [2sg] subject and [3sg] object.

Abstracting away from the complication posed by class II marking when the subject is second person singular (see section 2.2 and footnote 12), we may extract the agreement suffixes for third person singular objects from Table 3 (and examples (11)-12)) Doing so yields (13). Note that the one robust generalization over these forms is that the form of the suffix expresses the number of the object (i.e. by the glottalization of “n”, or a glottalized “n”—written ʔn—the regular nominal plural throughout the language) but is otherwise conditioned by the (person, number) features of the subject.

(13)

The distribution of the agreement suffixes for third person direct objects adds another challenge to the notion that the agreement suffixes follow an absolutive alignment. The absolutive function by definition includes the subject only when the verb is intransitive, but these examples show that the expression of the features of the subject in the suffix position is not restricted to intransitive subjects. The features of the transitive subject also condition the suffix agreement exactly when the direct object is a third person. Once again, we find evidence that the fundamental alignment of the Itelmen agreement morphology (like Itelmen syntax) follows a subject versus object orientation, and the absolutive pattern in the suffixes turns out to be only apparent.

3.2 Analysis

Our interpretation of the facts just discussed is that they arise from the combination of two considerations. First, we stipulate that the suffix (i.e. object) agreement position is obligatory in Itelmen finite verbs.[13] Second, we assume that third person direct objects lack person features, and are characterized only by a singular vs. plural distinction.[14] These two considerations yield a conflict in transitive clauses where the direct object is third person. As the object (by assumption) lacks person features, the first requirement must be satisfied by features from elsewhere. We suggest that this is exactly what the forms in (11)-(13) illustrate; the number features of a plural direct object find expression in the glottalized nasal, but the person features in (13) are contributed by the subject argument.

On the surface, this leads to apparent multiple exponence. The features of the subject are simultaneously expressed by a prefix and by a suffix. On our analysis, the primary exponent of the subject features is always the prefix. The appearance of subject features conditioning the form of the suffixes in (11)-(13) is secondary, in the sense that the features are copied from a higher position. This approach to the third person direct objects predicts (all else being equal) that whenever there are no person features of a direct object to fill in the obligatory suffix slot, the features of the subject will be copied in this way. This approach leads straightforwardly to an account of the apparent “absolutive” bias of the suffix agreement system. Intransitive clauses by definition lack (direct) objects; hence, the features of the subject must be copied to fill in the suffix agreement slot. As opposed to the forms in (13), though, in the intransitive suffixes (7) even the number features expressed are those of the subject as there is no object to supply them. For example, -kičeʔn (with a final, glottalized n) in intransitives marks the plurality of the first person subject, while the same form in the (class II) transitive paradigm marks the plurality of the object, indicating of the subject only that it is first person.

Before closing this section, we note that there are in principle sources other than the subject for providing features to the object slot when the object is deficient. For example, an indirect object could provide features. Mark Baker (personal communication) has suggested that the description of direct-indirect object agreement competition in Nahuatl may be looked at in this way. According to the brief discussion in Launey (1979: 172-174), if the direct and indirect object of a ditransitive verb are both definite, only one of the two may have its person features expressed by a verbal prefix. However, the plural feature of a third person direct object may be expressed, even if the person features expressed are those of the indirect object (see also the discussion around (31), below). In addition it should be noted that copying is not the only possibility for satisfying the obligatory features requirement assumed here. Languages could also resolve the conflict by supplying default features, for example. We have no principled account to offer for why in Itelmen the subject is privileged in the feature copying (or sharing) process, and we hope the topic will provide a starting point for subsequent cross-linguistic exploration.

In sum, what we have attempted to show here is that regardless of the details of the account, it is clear that any analysis of the transitive agreement suffixes must make reference to the features of the subject just in case the direct object is third person (i.e., lacks person features). Whatever mechanism is employed to do so (e.g., feature copying/sharing as proposed in Bobaljik & Wurmbrand 1997, contextual allomorphy, as in Bobaljik 2000, or syntactic agreement as in Adger, Béjar & Harbour 2001), that mechanism can plausibly be extended to the intransitive cases as well. Importantly, the fact that the features of the transitive subject are necessary for determining the shape of the agreement suffix in some cases provides additional prima facie evidence against treating the suffixes as absolutive agreement.

4. Oblique agreement in Itelmen

In the previous sections, we have been concerned with showing that the object agreement suffixes are not a part of an absolutive agreement system, but are rather primarily (direct) object agreement markers. To this point, we have avoided an additional complexity, mentioned in the introduction and indicated in the two rightmost columns of Table 1, namely, the complementarity between the direct object agreement suffixes and those agreeing with an indirect object or other dative element (expressed or otherwise), such as goals, benefactives, and possessors. Some examples of agreement with such obliques are given in (14) (underlining in the paraphrase indicates the element the verbal suffix agrees with):

In (14a), the suffix -nen [3sg.oblique] agrees with an (implicit) goal or possessor (the former would bear dative case, if expressed overtly). In (14b), the suffix -um [1sg] agrees with the overt dative-marked indirect (goal) object. In (14c), the suffix -nen [3sg.oblique] agrees apparently with a locative-marked indirect (source) object.[15] Note that (14b) contrasts minimally with (1a), repeated here as (15):

The only overt difference between (14b) and (15) is in the choice of agreement suffix. As indicated by the English paraphrases, the two sentences are (nearly) synonymous and in particular have the same ditransitive argument structure. This can be seen clearly in the forms of the overt pronouns: regardless of which argument triggers agreement, the caseless [2sg] pronoun is interpreted as the direct object (patient) and the dative-marked [1sg] pronoun is interpreted as the indirect object (goal). Since the argument structure is reflected only in the pronouns, the verbs are in principle ambiguous: if the pronouns are omitted—which the language freely permits—a sentence consisting only of the verb in (15) could mean ‘will he give (someone/something) to you?’ or ‘will he give you to (someone/something)?’[16]

Example (14c) is like (14a-b) and (15) in that the suffix -nen expresses the features of only one of the verb’s two internal arguments. In this case, -nen references a [3sg] indirect object, but makes no reference to the features of the direct object, which must be inferred from the lexical NP or from context. Again, the same argument structure combination could be expressed by means of a different suffix on the verb, referencing only the features of the direct object (see (16)-(17), below). One difference between (14d) and the previous examples is that there are discrete suffixes for third person indirect objects (as opposed to third person direct objects), while first or second person object suffixes do not distinguish between direct and indirect objects. In this respect, the inventory of Itelmen object agreement suffixes is reminiscent of object clitics in, e.g., French or Italian. Third person clitics in French have distinct forms for direct object (le [m.sg], la [f.sg], les [pl]) versus indirect (dative) object (lui [sg], leur [pl]), while first and second person clitics are the same for both direct and indirect objects (me [1sg], te [2sg], nous [1pl], vous [2pl]).

Oblique Agreement (as we will call this) is important to an understanding of Itelmen agreement for a number of reasons. For one, oblique agreement adds yet another prima facie counter-example to the notion of suffixal agreement as absolutive. There is no evidence (that we are aware of) to support a Dative-Shift (i.e., raising to object or 3 –› 2 advancement) analysis of examples in which the Dative shows agreement (see the discussion of (14b)-(15) above). Another characteristic of oblique agreement that is relevant in the present context is its apparent optionality. In section 3.1 we discussed an interaction between the features of the subject and those of the direct object which may be described as a form of competition for expression. The competition is resolved deterministically by grammatical principles: if the object has features to express (i.e., if the direct object is not third person), then the features of the object ‘win out’ over those of the subject. The features of the subject are expressed in the suffix position only when the object has no features to contribute (including as a special case, intransitives, where there is no object). For oblique agreement, the existence of minimal pairs such as those in (14b)-(15) shows that a different explanation must be sought, and it is to this that we now turn.

4.1 Goals and discourse salience

To begin with, we will restrict the discussion to ditransitive verbs, i.e., verbs with two internal arguments, a direct object bearing no overt case marking and a dative marked indirect goal argument. Other instances of oblique agreement will be discussed in section 4.3.

The examples in (14) indicate the possibility of the suffix agreeing with an argument other than the direct object (or subject, via the mechanism sketched above). When we examine sentences in isolation, we find many cases of minimal pairs where agreement with either the direct or the indirect object is apparently grammatical. One such minimal pair we have seen above in (14b-c), others—with only third person internal arguments—are given in (16)-(17). Recall from the discussion of (14d) that Itelmen has distinct third person indirect object suffixes (sg.: -nen, pl.: -peʔnen).

Restricting the discussion for the moment to cases like (16) and (17) with two third person internal arguments, speakers consulted accept both the (a) and (b) forms, and offer similar minimal pairs as translations of Russian prompts. Adding contextual information, however, the variation disappears to a large degree. When we set up a topic in the preceding sentence, agreement is almost exclusively with this argument. This is illustrated in (18)-(19).

Speakers were presented with pairs of Russian sentences as in (18a) and (19a) and asked to translate these into Itelmen. In (18), the question ‘Where is my knife?’ establishes a context with ‘my knife’ as the topic (or most salient argument) for the next sentence. The continuation, ‘Who did I give it to?’ has a ditransitive verb ‘give’ which in principle could agree with either internal argument in Itelmen (see, e.g., (14b,c)). Interestingly, in this and similar contexts, speakers regularly prefer direct object agreement, i.e. with the (null) pronominal corresponding to the topic ‘knife’ (see 6b). In (19) on the other hand, the first sentence sets up ‘my brother’ as a topic. Again, the verb in the continuation ‘What did I give to him?’ potentially allows agreement with either object. In contrast to (18b), however, agreement in (19b) is with the indirect object ənna-nk ‘to-him’, which is the topic of the discourse fragment. While the particular examples above involve wh-questions, this is not a factor at play in the pattern; the agreement alternation is replicated when neither argument is an interrogative pronoun, as in (20).

Note also that the presence or absence of overt pronominals in the clauses of interest is orthogonal to the determination of agreement. As already mentioned, Itelmen freely allows pro-drop (i.e., null arguments); more to the point, the presence or absence of overt pronouns or NPs does not correlate in any way with the choice of agreement suffix. This can be seen already in (18) and (19) where the topic pronoun is omitted only in the former.

The relevance of the discourse topic to agreement is very clear in the data when both internal arguments are third person. When only one argument is third person, however, and the other first or second person, the situation is less clear. Consider the following examples:

The topic is set up in the same way in (18) and (21). In both examples, ‘the knife’ is the logical direct object of the verb ‘give’ in the continuation. While in (18b) the verb agrees with this direct object, in (21b) agreement is preferred with the second person indirect object. Likewise, comparing (19) and (22), the third person indirect objects in each case correspond to the previously set up topics, but in (22) the verb agrees instead with the second person direct object. Thus, in both examples (21b) and (22) the verb fails to agree with the topic set up in the preceding sentence. This pattern is systematic in Itelmen; the relevance of topic prominence seems to disappear when one of the internal arguments is not third person.

In the next sub-section we will suggest a partial account of the effect of first and second person arguments on the relationship between prominence and agreement. The account relies on the investigation of the effects of a person hierarchy on agreement in the related languages Alutor and Chukchi. While the person hierarchy applies deterministically in these languages, in Itelmen the person hierarchy and the prominence hierarchy interact, leading to the type of pattern seen in (21) and (22).

4.2 Person Prominence

Alutor and Chukchi, as mentioned in the introduction, have agreement systems quite similar to Itelmen, with most prefixes and suffixes clearly cognate. As described by Comrie (1979a) and Mel'chuk (1988), one crucial factor in determining which of a ditransitive verb’s internal arguments governs suffix agreement in these languages is the person hierarchy in (23a). Consider first the agreement system in Alutor. In his description of the morphosyntax of agreement in this language, Mel'chuk (1988) reports that only one verb—jəl-ekki ‘to give/give as a wife’—allows agreement with an indirect object. Moreover, he observes that whether the verb agrees with the direct or indirect object is determined as follows:

The conditions in (23) yield the following pattern for Alutor: If there is a first person object—direct or indirect—then it will trigger suffix agreement. In the absence of a first person object, a second person object—again, direct or indirect—will trigger suffix agreement. Finally, only if there is no first or second person object, then the verb will agree with the third person direct object. According to Comrie (1979a) and Mel'čuk (1988), the same is true for Chukchi except that in this language the direct and indirect object cannot both be first or second person; at least one internal argument has to be third person. Forms from Chukchi illustrating (23a) are given in (24) [some glosses and paraphrases added or altered].[18]

Examples from Alutor illustrating the same point are given in (25)[19]

The following examples from Chukchi and Alutor respectively illustrate(23b), where plural agreement is with the direct object, not the indirect:

Finally, as noted, only Alutor allows both internal arguments to be participants in the speech act (1 or 2 person), displaying the effects of the hierarchy 1>2, as shown in (27).

In (24)-(25) and (27), the verb has to agree with the indirect object since in each case, it outranks the direct object (2>3 and 1>2) according to the person hierarchy in (23). In (26) on the other hand, there are no first or second person arguments, and so the verb obligatorily agrees with the direct object and not with the indirect object, as evidenced by the plural marking in (26b) and the absence thereof in (26a).

An important issue that arises at this point is whether the effects of this person hierarchy are morphological or extra-morphological (e.g., syntactic or discourse-related). As we see, the person hierarchy in Alutor and Chukchi uniquely determines agreement in these cases.[20] As we have seen, Itelmen differs from these languages sharply in this regard. First, when there are two third person internal arguments, it is discourse prominence (not grammatical function or thematic role) that determines which argument controls agreement, and second, when one argument is a speech act participant (first or second person), the discourse saliency effects seem to still play a role in Itelmen, at least to the extent that person alone does not rigidly determine the outcome of the competition to control agreement. Thus in (15), we find agreement with a second person argument despite the presence of a first person potential controller. Similarly, in (28) the verb agrees with a third person object even in the presence of a second person oblique.

Though we don’t yet have a statistical count of the frequency of examples that violate the person hierarchy, our impression is that examples such as (28)—agreement with a third person over a first or second person—are quite rare, at least with ditransitive verbs. It is even more difficult to make convincing generalizations about examples where both internal arguments are first and second person, since contexts for such sentences are themselves quite uncommon.[21]

Summing up, we have purposely left the notion topic or prominence rather vague in the discussion. It is evident in (18)-(19) that an argument introduced into the discourse in one sentence constitutes some form of topic—i.e., old information—in the next. However, it is not necessarily the case that only arguments introduced in this way will constitute old information. For instance, it is plausible that first and second person arguments are inherently at least as prominent as third person discourse topics (and thus the person hierarchy might ultimately be reduced to a more carefully refined notion of topic or prominent argument). Nonetheless, it may be that exceptional violations of the person hierarchy arise from other strategies that can increase the prominence of a third person argument even to the point where its prominence surpasses that of a second person argument. For the purposes of this paper, we have kept the two prominence factors distinct, in part because in the related languages only the person hierarchy seems to play a role.

Ultimately, the important observation is that only one of a ditransitive verb’s two internal arguments is ever visible in the morphology for the determination of suffixal agreement. This differs from the cases with third person direct objects in which features of both the object and the subject were relevant to determining the form of the agreement suffix. We have shown in addition that the correct account of the indirect versus direct object competition appears not to be morphological in nature, but that extra-morphological considerations conspire to select one of two internal arguments as prominent for the purposes of agreement. We now turn to examples of oblique agreement with other verb classes and will then turn to a minor refinement of the characterization of the direct versus indirect object competition.

4.3 The extent of oblique agreement

The preceding section discussed agreement with oblique (dative) arguments from the perspective of verbs with multiple internal arguments (i.e., ditransitive verbs). Our conclusion was that, in contrast to the purely morphological effects (section 3), the apparent optionality regarding the choice of whether or not to agree with an oblique argument is a matter of two (presumably connected) discourse-related hierarchies: one which we have loosely called “topic-prominence” and the other a person hierarchy. Itelmen differs from its neighbors in that it allows oblique agreement with an apparently wide range of verb classes (although there is significant speaker variation about which we have nothing to say). Examples of oblique agreement with verbs other than ditransitive ‘give’ include lexical causatives of transitive verbs, such as ɬintɬi- ‘to put X (clothing) on Y’ (<ɬin- CAUSATIVE and t(ɸ)ɬ(i)- ‘to put on, to wear, to bring/carry’) in (16b), above, and verbs with an optional goal argument such as əntxla- ‘to bring’ and ɬa-/la- ‘to say, tell’:

Labile (i.e., alternating) verbs also seem to allow oblique agreement, sometimes with a shift in meaning. For example, the verb əlčqu- ‘to see’, normally taking transitive agreement, can occur with oblique agreement with the sense ‘to look at, to watch’:

In the remainder of this section—which represents ongoing work—we would like to turn briefly to cases of oblique agreement with intransitive verb stems. Examples of oblique agreement with intransitive stems differ from the examples discussed above in three ways, which we note below. Given the differences, we leave open in this paper the question of whether or not our analysis should extend to these cases, however, in section 4.4 we briefly discuss one way in which the theory could be refined to accommodate the intransitive cases.

The first point where intransitive verbs differ from the transitive and ditransitive stems above is that intransitive verbs appear to allow (oblique) agreement with the possessor of the subject (as noted by Bogoras 1922 and Volodin 1976, 1984). Example (14a) might be interpreted in this way, as may those in (31).[22]

Volodin (1976:260-1) gives 12 examples similar to the ones in (31), with a range of intransitive verb classes, of which we repeat two here:

It is of course difficult to conclusively determine whether these constitute agreement with a possessor directly (perhaps via some form of possessor raising), or whether these involve agreement with an implicit dative “affectee” argument (e.g., benefactee or malefactee) on the model of, e.g., Hebrew, or the English construction represented by ‘My child was crying on me all night’ (see Borer & Grodzinsky 1986, Landau 1999 for some discussion of this issue with respect to dative clitics in Hebrew; see also Takehisa 2001 for general discussion of possessor raising versus affectee). Note that even if the examples in (31)-(32) involve agreement with possessors, oblique agreement with intransitive verbs is certainly not limited to possessors, as shown by (33):

A second point of difference between the manifestations of oblique agreement with transitive and intransitive stems concerns the range of permissible participants. Regarding oblique agreement with transitive verbs as discussed throughout this paper, the oblique argument controlling agreement may be any person (1, 2 or 3) and any number (contra earlier descriptions, see fn. 16). Nevertheless, in cases of oblique agreement with intransitive stems, what we have found to date is that agreement is only possible with third person possessors/oblique arguments. Consultants (in Sedanka at least) consistently reject examples such as Volodin’s (32a) with a second person plural possessor. The following examples are representative.

Example (34a) shows that either intransitive agreement or (apparent) possessor agreement with a third person plural is possible with the verb čača- ‘cry’, but (34b) shows that a second person plural subject cannot control oblique agreement in such cases.

While there is clearly fertile ground for continued investigation of the conditions on oblique agreement in Itelmen intransitives, the two points just considered do not seem (to us) to bear on the analysis presented in this paper. There is a third aspect of Itelmen intransitive oblique agreement which does, perhaps, necessitate a refinement of the account we have presented above, if indeed these cases of agreement are to be subsumed under the theory we have developed. We examine this in the next section.

4.4 A possible refinement: oblique, if prominent.

Our treatment of the apparent optionality of oblique agreement with (di-)transitive verbs may yield the impression that it is simply the “most prominent” argument which is selected as the controller of agreement. To the extent that intransitive verbs, such as kʼoɬ- ‘come, arrive’, allow agreement with an oblique, such a characterization might have undesirable consequences when taken together with our suggestions regarding suffix subject agreement in intransitives. For concreteness, consider again (14a), repeated here.

Recall that we have treated the agreement with an intransitive subject at the suffix position as parasitic or secondary. Extending the analysis of the third person direct objects, we claimed that the agreement with the subject in the suffix position was the result of a copying or sharing mechanism, invoked to satisfy the requirement that the suffix have features. This feature copying/sharing is necessary when the direct object has no person features, or when there is no direct object (as in the case of intransitives). Now, intransitive verbs quite generally are able to occur (displaying regular intransitive agreement) with datives or with a possessed subject. An example was given in (34a), repeated here.

Given that this verb is in principle capable of bearing oblique agreement, and given moreover that it has no direct object, it would seem that the presence of a dative element (or possessor) should make that element the most prominent non-subject member of the clause. (Recall that the topicality of the subject played no role in the analysis of section 3.1.) Accordingly, an analysis strictly along hierarchical lines (i.e., “the most prominent internal argument”) would lead one to always expect agreement with the dative (or possessor) in such cases, since in the absence of another internal argument, the dative would always be “most prominent”. Rather, what such examples force is an account under which the Oblique agreement facts are stated as favoring the dative, if it is prominent (in whatever sense turns out to be relevant), with the “normal” agreement pattern obtaining otherwise.

5. Summary and Conclusion

In this paper, we have identified and attempted to account for a set of cooccurrence restrictions among the agreement suffixes in Itelmen. The main empirical observations are:

We have argued that there are distinct phenomena at work in Itelmen, and that the burden of the account of each of these observations is divided among different modules of grammar.

In section 3, we argued that the proper account of the observation in (35ii) lies squarely in the morphology. When the verb does not agree with an oblique (either because there is none, or because the oblique is, by hypothesis, not sufficiently prominent) then the resulting agreement pattern is uniquely determined for each given combination of subject and object features. If the object is first or second person, then the person and number of the object alone determine the shape of the agreement suffix. If there is no direct object, then the person and number of the subject determine the shape of the agreement suffix. Finally (and central to the understanding of the pattern), when the object is third person, the number features of the object act in consort with the features of the subject to determine the agreement pattern at the suffix position. We take this deterministic nature of the core agreement patterns (unlike those of the oblique versus core alternation) to be important evidence that the two observations in (35) reflect the workings of different parts of grammar.

In section 4, we presented evidence that determination of whether a given verb will agree with an oblique or a core argument requires appeal to contextual or discourse factors, and we have loosely identified an interaction between “topic prominence” and a person hierarchy as the relevant factors. One observation which we take to be important is that Itelmen differs from Chukchi and Alutor in that the variation in agreement between an oblique and a core argument in Itelmen is determined strictly neither by the person hierarchy nor by thematic role or grammatical function. Thus in Itelmen, but apparently not in Chukchi, we find the apparent optionality (i.e., from a morphosyntactic perspective) of minimal pairs like (1) in which the syntax can be held constant, yet the morphology can vary.

Though this paper is still preliminary in many respects, the generalizations presented here have certain implications for current theories of morphology. In particular, our analysis addresses an issue which is typically left open in position-class approaches to morphology (both traditional and more theoretically-inclined, such as Inkelas 1993). While the identification of ordered “slots” allows a statement of which morphemes are in complementary distribution, such templatic or position class analyses typically do not provide insight into why, in any given context, one particular morpheme wins out over another of the same class. For instance, a position class analysis of Itelmen (such as Volodin 1976) recognizes that direct and indirect object suffixes are in complementary distribution and that there is a range of allomorphs of putative third person direct object suffixes. However, we have gone a step further, attempting to elucidate the reasons for the choice of one morpheme over another to fill a given position.

Our analysis furthermore sheds light on the nature of multiple exponence in agreement systems, i.e., instances in which a single feature appears to be expressed by more than one morpheme. We have argued that the most concise account of Itelmen agreement morphology takes the expression of subject features in the suffixes to be secondary, in essence, the reflex of a mechanism which copies features from their primary locus of exponence (in the case of Itelmen, the prefixes) to provide features for an obligatory position which would otherwise be unfilled.

We have for the most part refrained from committing ourselves to any specific framework. Doing so would have required making numerous arbitrary assumptions, the more so if one is to integrate the remainder of the inflectional morphology not discussed here. Needless to say, this will be an important undertaking, however, we felt that at this stage it was more useful to focus on bringing to light the relevant generalizations. As a general point of methodology, we believe that the generalizations which surface in our analysis must be compared with generalizations about other complex agreement systems before one begins to draw any firm conclusions about morphological and morphosyntactic universals.

6. References

Adger, David, Susana Béjar & Daniel Harbour. 2001. Allomorphy: adjacency and agree. Paper presented at GLOW 24, Universidade de Minho, Braga, Portugal. April 2001.

Bittner, Maria & Ken Hale. 1996. Ergativity: Toward a theory of a heterogeneous class. Linguistic Inquiry 27. 531-604.

Bobaljik, Jonathan David. 1998. Pseudo-ergativity in Chukotko-Kamchatkan agreement systems. Lea Nash (ed.) Ergativity: Recherches Linguistiques de Vincennes 27. 21-44.

---. 2000. The ins and outs of contextual allomorphy. University of Maryland Working Papers in Linguistics, ed. by Kleanthes K. Grohmann & Caro Struijke, 10. 35-71.

---. (To appear). Review of Georg & Volodin (1999). Anthropological Linguistics.

Bobaljik, Jonathan David & Susi Wurmbrand. 1997. Preliminary notes on agreement in Itelmen. PF: Papers at the Interface: MIT Working Papers in Linguistics, ed. by Benjamin Bruening, Yoonjung Kang & Martha McGinnis, 30. 395-423.

---. 2001. Seven prefix-suffix asymmetries in Itel’men. To appear in: Proceedings of the 37th Chicago Linguistics Society.

Bogoras, Waldemar. 1922. Chukchi. In Franz Boas (ed.). Handbook of American Indian languages, 631-903. Government Printing Office, Washington.

Borer, Hagit & Yosef Grodzinsky. 1986. Syntactic cliticization and lexical cliticization: the case of Hebrew dative clitics.The syntax of pronominal clitics, Hagit Borer, ed., 175-217. (=Syntax and Semantics 19). Orlando: Academic Press.

Comrie, Bernard. 1979a. The animacy hierarchy in Chukchee. The elements: a parasession on linguistic units and levels, ed. by Paul R. Clyne, William F. Hanks & Carol L. Hofbauer , 322-329. Chicago: Chicago Linguistic Society.

---. 1979b. Degrees of ergativity: Some Chukchee evidence. Ergativity: Towards a theory of grammatical relations, ed. by Frans Plank, 219-240. London/New York: Academic Press.

---. 1980. Inverse verb forms in Siberia: Evidence from Chukchee, Koryak, and Kamchadal. Folia Linguistica Historica 1/1. 61-74. doi:10.1515/flih.1980.1.1.61

---. 1981. The languages of the Soviet Union. Cambridge: Cambridge University Press.

---. 1983. The genetic affiliation of Kamchadal-Some Morphological Evidence. Papers in Linguistics 16/3-4.109-120. doi:10.1080/08351818309370598

Dixon, Robert M. W. 1994. Ergativity. Cambridge: Cambridge University Press.

Dunn, Michael. 1999. A grammar of Chukchi. Ph.D. Dissertation, Australian National University .

Fortescue, Michael. 1997. Eskimo influence on the formation of the Chukotkan ergative clause. Studies in Language 21/2: 369-409. doi:10.1075/sl.21.2.05for

Georg, Stefan & Alexander P. Volodin. 1999. Die itelmenische Sprache: Grammatik und Texte. Harassowitz Verlag, Wiesbaden. (=Tunguso Sibirica 5)

Halle, Morris & Ken Hale. 1997. Chukchi transitive and antipassive constructions. Ms MIT, Cambridge, Mass.

Inkelas, Sharon. 1993. Nimboran position class morphology. Natural Language and Linguistic Theory 11. 559-624. doi:10.1007/bf00993014

Krasheninnikov, Stepan Petrovich. 1755/1994. Opisanie zemli Kamchatki, vol. II. Sankt Peterburg: Imperatorskaja akademija nauk. Facsimile edition (1994) Petropavlovsk-Kamchatskij: Kamshat.

Landau, Idan. 1999. Possessor raising and the structure of VP. Lingua 107/1. 1-37. doi:10.1016/s0024-3841(98)00025-4

Launey, Michel. 1979. Introduction à la langue et à la littérature azteques, tome 1: grammaire. Paris: L’Harmattan.

Mel'chuk, Igor A. 1973. Model’ sprjazhenija v aljutorskom jazyke I. (=Predvaritel’nye publikacii, problemnaja gruppa po eksperimental’noj i prikladnoj lingvistike (preprints) 45) Moscow: Institut russkogo jazyka AN SSSR.

---. 1988. Dependency syntax: Theory and practice. New York: State University of New York Press.

Moll, Tatjana A. 1960. Ocherk fonetiki i morfologii sedankinskogo dialekta itel’menskogo jazyka Uchenye zapiski, Leningradskij gosudarstvennyj pedagogicheskij institut im. A. I. Gercena, kafedra jazykov narodov krajnego severa 167.193-222.

Murasugi, Kumiko. 1994. A constraint on the feature specification of Agr. The morphology-syntax connection: MIT Working Papers in Linguistics, ed. by Heidi Harley & Colin Phillips, 22.131-152.

Nedjalkov, Vladimir P. 1979. Degrees of ergativity in Chukchee. Ergativity: Towards a theory of grammatical relations, ed. by Frans Plank, 241-262. New York/London: Academic Press.

Noyer, Rolf. 1997. Features, positions and affixes in autonomous morphological structure. New York: Garland Press.

Phillips, Colin. 1993. Conditions on agreement in Yimas. Papers on case and agreement I: MIT Working Papers in Linguistics, ed. by Jonathan David Bobaljik & Colin Phillips, 18.173-213.

Skorik, Piotr Ja. 1977. Grammatika chukotskogo jazyka, chast’ II: glagol, narechie, sluzhebnye slova. Leningrad: Nauka.

Spencer, Andrew. 1996. Agreement morphology in Chukotkan. Essex Research Reports in Linguistics 10.1-34.

Takehisa, Tomokazu. 2001. The distribution of agent and affectee in natural languages. Ph.D. Comprehensive Evaluation Paper. McGill, Montreal.

Volodin, Aleksandr P. 1976. Itel’menskij jazyk. Leningrad: Nauka.

---. 1984. Struktura predlozhenija v itel’menskom jazyke. Struktura predlozhenija v jazykax razlichnyx tipov: Paleo-aziatskie jazyki, ed. by V. Z. Panfilov & P. Ja. Skorik, 240-270. Leningrad: Nauka.

Volodin, Aleksandr P. & Nikolai B. Vakhtin, 1986. Ergativnost' i mexanizm glagol’nogo soglasovanija (opyt tipologii). Paleo-aziatskie jazyki , ed. by Piotr Ja. Skorik, 111-133. Leningrad: Nauka.

Worth, Dean Stoddard. 1961. Kamchadal texts collected by W. Jochelson (=Janua Linguarum - Series Maior II). The Hague:Mouton.

Authors' contact information:

Jonathan David Bobaljik
jonathan.bobaljik@mcgill.ca
Susi Wurmbrand
susanne.wurmbrand@mcgill.ca

McGill University
Department of Linguistics
1085 Dr. Penfield
Montréal QC H3A 1A7
CANADA

[*]We are most deeply indebted to the speakers of Itelmen who shared their time, hospitality, and knowledge with us, especially (but not only): T.N. Bragina†, N.I. Chatkina, T.E. Guterova, A.D. Ivashova, G. Makhanova, L.E. Pravdoshchina, S.E. Prichin, A.E. Shamuraeva†, E.E. Silina†, V.I. Uspenskaja, I.I. Yaganov† and N.S. Yaganova (Northern dialect), V.P. Krasnoyarev†, G.N. Lastochkina†, V.V. Prichin†, N.V. Prichina, T.P. Sadovnikova†, M. P. Slabodchikova, E.P and V.D.† Zaporotskaja, G.D. and N.Z. Zaporotskij, and D.N. Zhirkov† (Southern dialects). We also acknowledge a deep intellectual debt to the pioneering work by A.P. Volodin on Itelmen grammar.
A preliminary version of this paper appeared in the MIT Working Papers in Linguistics as Bobaljik & Wurmbrand (1997). The work reported on here represents a part of a larger, on-going project. Many people have commented on the larger project; for comments specific to this paper we wish to thank Mark Baker, Alec Marantz, Ken Hale, Morris Halle, David Pesetsky, the reviewers for Linguistic Discovery and audiences at McGill, Berkeley, MIT and SCIL 11 at the University of Texas at Austin. We would also like to thank Rachel Turk for her assistance with the research for section 4.3. This project has been partially supported at various stages by funding from the National Council for Soviet and East European Research (grant to D. Koester 1993-94), the Milton Fund of Harvard University (1995-96), the Faculty of Graduate Studies and Research, McGill University, Fonds pour la Formation de Chercheurs et l’Aide à la Recherche #00-NC-2043 and the Social Sciences and Humanities Research Council #410-99-0902. Errors of fact, interpretation and anything else can still be blamed on the first author.

[1] Except as noted, the data in this paper were collected by the first author on two extended trips to Kamchatka (1993-94, Spring 1996) and by both authors on two subsequent trips (1996, 2001). Examples are for the most part from the Northern (Sedanka) dialect of (Western) Itelmen, though there are no differences between the dialects which affect the analysis presented here (see Moll 1960 for discussion). For some important background information on Itelmen, see section 2.1 below.
The following abbreviations are used in this paper: ABSolutive, ACCusative, ADJective, AGReement, ASPect, AUXiliary, DATive, DIMinutive, DISTRIButive, EMPHatic, ERGative, FUTure, HABITual, HORTative, IMPRS=impersonal, INFinitive, IRRealis mood, LOCative, NEGative suffix, NOMinative, OBJect, OBLique, PL=plural, POSSessive, PRESent, SG=singular, SUBject, TNS=tense, VOCative, II=class II conjugation.
Discrete morphemes are separated by hyphens in the examples and the glosses; a dot in the glosses separates distinct features expressed on a single portmanteau morpheme. Where a gloss has the form, e.g., 2PL>3SG it indicates a portmanteau for second person plural subject acting on third person plural direct object.
Examples are broadly transcribed (i.e., various allophonic alternations such as automatic palatalization of /ɬ, l, n/ before /č/ are not indicated so as to increase readability). We have used IPA except /č/ (=IP [tʃ], also for typographic reasons, underdots have been omitted on s, z, which in Itelmen are apical, post-alveolar fricatives. Note also, that sequences written as a glottal stop plus nasal or "ʔl" correspond to a single, glottalized segment. Note finally that certain uncertainties in the transcription (e.g., vowel quality) have been regularized where they do not bear on the segmentation or the analysis here.

[2] Also written itel’men, from Russian, which is in turn from the ethnonym itənmən (Volodin 1976:5).

[3] This distinguishes Itelmen from the other Chukotko-Kamchatkan languages, which all have (partial) Ergative-Accusative case systems. The other languages also have noun incorporation, which Itelmen lacks.

[4] On some phonological asymmetries between inflectional prefixes and suffixes, see Bobaljik & Wurmbrand (2001).

[5] As noted by Comrie (1983), the intricate parallels and cognates in this agreement system constitute one of the strongest arguments for a genetic Chukotko-Kamchatkan grouping.

[6] Prefixes t- and q- have ejective allomorphs before vowels and (possibly) sonorant-initial stems.

[7] This is not true of third person plural subjects, see below.

[8] A nominative-accusative system (syntax, agreement or case) groups transitive and intransitive subjects together, treating objects as the marked category. An ergative-absolutive system groups objects and intransitive subjects together, treating transitive subjects as the marked category (see Dixon 1994 for a survey). While some theories (notably Bittner & Hale 1996) hold that nominative and absolutive are identified at some level of abstraction, we continue to use the descriptive terms, whereby ‘nominative’ and ‘absolutive’ refer to distinct ways of grouping grammatical functions.

[9]Note that the effects of various phonological processes are obscured by the presentation here. Also, the forms in the tables are from the Northern dialect (though the differences do not affect the analysis in any way that we can detect, see below). Certain forms also show inter- and intra-speaker variation. For example, the first person intransitive suffix (in both dialects) alternates (apparently freely) between -kičen/-kičeʔn and -k (sg. and pl.). The string /ičen/ is not deletable in the transitive paradigm, nor is the apparent optionality of this portion of the suffix a recent phenomenon (e.g., attributable to language attrition); both forms occur throughout the texts collected by Waldemar Jochelson (Vladimir Iochel’son) in the early 1900s (Worth 1961). Moreover, Skorik (1977:.20, fn.11) notes that the first person plural intransitive suffix is also optional in Chukchi (see also Dunn 1999:80) Why this should be is a mystery. The form for [3pl.OBLIQUE] also shows a good deal of variation: Volodin (1976) gives -(ki)pəen, -(ki)pəeʔn, and -(ki)pneʔn; we also have at least -(ki)peneʔn and -(ki)neʔn attested in our notes.

For the record, the differences between the dialects in the agreement morphology are noted here: The [1pl] realis prefix in the Southern dialect group is n- rather than nt- (for transitives and intransitives alike). In both dialects, the prefix xɐn- may be used for [1pl.irr] in place of mə- for transitive and intransitives alike. Note that the initial /x/ in the this prefix (in all its uses) can be omitted in the Northern dialect. In the transitive paradigm, the form of the first person object suffixes is also different in the Southern dialect groups: -mə [1sg], -məŋ [1pl].

[10] Both Volodin (1976) and Moll (1960) suggest -(β)ŏmnen as a portmanteau for first person objects with third person subjects. Based on our material, we prefer to treat this as a sequence of first person object suffix -βum followed by an independent third person subject (en)clitic. This clitic is attested in our material as =nin, =nen or =in/=iʔn.While its exact distribution remains unclear, we have found that this clitic (in contrast to agreement suffixes) can be omitted, and need not be adjacent to the verb.

[11] This is true of Chukchi to a lesser degree, see Dunn (1999:180). For discussion of the apparent exceptions, see Bobaljik (1998). Note that Fortescue (1997) reaches a similar conclusion about the Chukotko-Kamchatkan agreement paradigms from a diachronic perspective.

[12] In addition to the suppression of the class marking, there are two ways in which the data in the table deviate from the general characterization given in the text. First, the suffix -sx occurs with all [2PL] subjects—the only instance of true suffixal agreement with a transitive subject. Note, though, that -sx occurs peripheral to the other suffixal agreement morphology, e.g., q-tøli-xk-um-sx [2:SU-bring-CLASSII-1S:OB-2P:SU] ‘bring me!’, and is strange within C-K morphology in other ways as well. Second, the prefix for a [3PL] subject varies for transitivity (n-~). See Bobaljik (1998) for a comparison to Chukchi.

[13] This may be expressed, for example, as a “morphological EPP” effect as proposed in Phillips (1993) for Yimas, or as a requirement of the Autonomous Morphological Structure of Itelmen in the framework proposed by Noyer (1997).

[14] It is not immediately obvious that third person lacks person features throughout the language. While the third person intransitive subject suffixes appear to reflect number only, the indirect object forms and the 3>3 allomorphs of the Class II markers appear to require reference to third person. In terms of agreement, Murasugi (1994) has observed that it is a general pattern of languages that object agreement makes fewer distinctions than subject agreement, and that direct objects are poorer than indirect objects. This is an area which clearly requires further cross-linguistic research.

[15] It is not clear whether agreement is directly with the locative marked source or whether this is more accurately described as agreement with a covert dative affectee or possessor, ‘e.g., I took my friend’s mushrooms from him.’ See section 4.3 below.

[16] Bogoras (1922) and Volodin (1976) (see also Georg & Volodin 1999:145) claim that indirect object agreement is possible only with third person (singular and plural) and second person plural arguments. However, agreement with first and second person indirect objects is robustly attested in our material, from both dialects and from many speakers (examples (14c) and (21b) are representative of a larger range of data in our corpus). We leave the reason for the discrepancy between our corpus and those of Bogoras and Volodin as a mystery for the present time.

[17] In Itelmen, negation in the present and past tense are expressed by means of a compound construction in which the lexical verb takes a negative participial form (varying only for transitivity) and the inflectional morphology is expressed on an auxiliary verb. The auxiliary may be freely omitted.

[18] In particular it should be noted that our gloss of the prefix ne- in these examples as subject agreement departs from Dunn (1999) where the prefix is held to be an inverse marker (pp. 181ff). See Comrie (1980) for general discussion, and Bobaljik (2000:52ff) for explication of the analysis assumed here (though for the present argument, nothing actually hinges on this point).

[19] An important caveat is in order regarding the (a) example. We have added a prefix to the example that is not reported in Mel'chuk (1988:295). Available descriptions indicate that the form should have a prefix and I. Mel'chuk (personal communication, 11/1998) confirms that the absence of the prefix in this particular form is anomalous among the data he has collected. We have continued to include this example as it forms illustrative minimal and near minimal pairs with other examples given here; in addition, the presence or absence of the prefix is not relevant to the issue at hand, what is important is which argument controls agreement.

[20] That the person hierarchy is more morphologized in Chukchi, Koryak and Alutor is also shown in the domain of inverse constructions. Thus, in certain configurations in which the agent is lower on the hierarchy than the patient in a transitive construction, special inverse (or “spurious” antipassive) morphology is required (see Comrie 1979a, 1980, Spencer 1996, Halle & Hale 1997, Dunn 1999). Note that the inverse construction in these languages in certain contexts is marked by antipassive (i.e., detransitivizing) morphological changes on the verb, but syntax (as evidenced by case marking) is not affected; this makes the construction reminiscent of deponent constructions in, e.g., Latin and Greek.
Itelmen also differs from its related languages in the extent to which the person hierarchy is involved in forcing inverse constructions (including those discussed in the previous footnote, and the impersonal construction in Itelmen). First, while inverse/spurious antipassive morphology is obligatory with certain person/number subject/object combinations in Alutor and Chukchi, in Itelmen, despite perhaps a tendency to find impersonal constructions when a third person acts on a first person, this is by no means obligatory. Second, while the inverse construction does not affect case marking in Alutor and Chukchi, in Itelmen there is a change of subject case in the impersonal construction, as mentioned in section 2.1 above.

[21] There may be a difference between interactions of first and second person arguments on the one hand, and between interactions of these with a third person on the other. We have a handful of examples such as (27) showing agreement with a third person in the context of a first or second person. Example (14) is the one reliably clear example of agreement with second person taking precedence over agreement with a first person that we have double-checked. One reason for the apparent rarity of such examples may be the paucity of relevant contexts (to elicit these, we set up contexts in which, for example, an adult and a child converse).

[22] Caution is in order with these examples, as the apparent pluralizing glottalization in the suffixes is not expected from our perspective. The (b) example occurs twice consecutively in a narrative, and the context favors an affectee reading (see below) for the singular possessor of the belt. The protagonist (who is possessed of enormous speed) runs home from hunting ducks for food, but the speed of his passing tears the bodies of the ducks away from his belt and he is left foodless.