A preliminary task, prior to an investigation of ordering relationships among the lexical elements in the verb, is the establishment of the position of the verb stem in Athapaskan languages. Abstracting away from details of the verb, we find that across the language family, the morphemes in the verb come in the order in (1).

1. (1) lexical items – functional items – voice/valence – verb stem

The positioning of the verb stem at the right edge of the verb word is surprising for several reasons. First, it has long been observed that inflection tends to fall outside of derivation (see Anderson 1982, 1988 for detailed discussion). Although the definition of what is inflectional and derivational (for which I use the terms functional and lexical, respectively) in Athapaskan languages is not completely clear (see Kibrik 1995, Rice 1993, Speas 1990 and chapter 16), the Athapaskan verb nevertheless is a striking counterexample to this generalization. Recall the template for Slave provided in chapter 2, repeated here as (2).

1. (2) preverb # quantificational elements # incorporate # object % third person subject % qualifier + subsituation aspect + situation aspect + viewpoint aspect + 1/2 person subject = voice/valence + root + tense/mode/aspect suffix

The items in (2) that are separated by the # boundary symbol are known in the Athapaskan literature as disjunct prefixes.

Athapaskan languages have verbs that are extraordinarily complex, and that pose a challenge to theories of morphosyntactic structure (see, for example, Aronoff 1994, Hargus 1988, Rice 1993, 1998, Speas 1984, 1987, 1990, 1991a,b, Spencer 1991, Travis 1992, for discussion). The verb word is complex in many ways: it is morphologically rich, the surface ordering of morphemes is apparently without reason, discontinuous dependencies are frequent, and blocking effects between morphemes of identical shape but different meaning are abundant.

The goal of this chapter is to outline the structure of an Athapaskan verb as traditionally described and to examine the claim that a template is required to define the ordering of morphemes within the verb.

The Templatic Nature of the Athapaskan Verb

As discussed in chapter 1, the verb in Athapaskan languages is typically described as consisting of a template, or string of fixed order positional classes. The template orders the morphemes, and each morpheme is marked lexically for the position in the template that it occurs in. In addition, phonological boundary types are lexically associated with the different morphemes in order to account for their phonological properties.

A template for Slave ([slevi]), adapted from Rice 1989, is given in (1). See appendix 1 for a list of templates proposed in the literature for a number of languages of the family. Terminology will be clarified throughout the book; I do not attempt to define terms here.

So far, I have made the following observations: morphemes that enter into a scopal relationship with each other exhibit fixed order if their semantic relationship is fixed, if their semantic relationship is variable, their ordering varies according to this relationship, and if there is no semantic relationship between morphemes, their ordering with respect to each other may also vary. Based on these observations, I concluded that morpheme order in the verb of Athapaskan languages is regulated by scope. In this chapter, I would like to examine the morpheme stock of Athapaskan languages to provide further evidence for the claim that morpheme order is regulated not by an arbitrary template but rather by principles of universal grammar.

Predictions

In chapter 3, I laid out two hypotheses that might account for morpheme order in the verb of Athapaskan languages. Under the template hypothesis, morphemes are marked for the position that they appear in, and linear order need not reflect word formation. Thus, the ordering in a template is arbitrary, and, as discussed by Spencer 1991, in many ways quite unconstrained as compared with layered morphology.

Consider the consequences of the template hypothesis for language change. According to this hypothesis, the similarities in morpheme order across the language family are a consequence of common origins: at an early stage of the family, only a single template was available.

In this second part of the book, I have argued for the following points concerning overall ordering of morphemes:

• The verb stem originates within the verb phrase rather than at the right edge of the verb. Under this assumption, the ordering of verbal elements fits the pattern identified in typological studies of languages of the world.

• The voice/valence morphemes are functor predicates that originate within the verb phrase. Under this assumption, the argument-affecting properties of voice/valence are accounted for, as is the fact that they do not occur with independent argument structure.

• The stem raises to the voice/valence markers, and this unit in turn raises to the right edge of the verb phrase. This assumption is required to place the verb and voice/valence markers, a phonological constituent, in the correct position on the surface.

It is in these ways, and these ways only, that the surface ordering of the elements of the verb differs from their more abstract ordering.

I have further argued for the following points about ordering within the lexical items. Within the lexical complex of the verb, the scope hypothesis provides an account of fixed as well as variable orderings. Variable orderings are of three types. First, the orders A-B and B-A can both be interpreted, but have different readings. In this case, the semantic relationship between A and B is not fixed.

Whenever linguists discover that I work on Athapaskan languages, I can anticipate the first question that they will ask – however could a child come to learn the order of morphemes in the verb of one of these languages? The order of morphemes seems to be completely without rhyme or reason. Morpheme order is thus a question that is everpresent in one's mind when studying languages of this family. I first began to tackle this problem in 1991, with work on the so-called disjunct prefixes of the Athapaskan verb. A crosslinguistic survey revealed something very striking – little variation existed across the family in terms of the ordering of these elements. I began to feel that I was on the road to an explanation of the ordering of this part of the verb, but the so-called conjunct portion of the verb still left me baffled. One day in the early 1990s Chomsky gave a talk here at the University of Toronto, and I began to get some glimmerings; at least the ordering began to look somewhat less random than it had hitherto seemed. It was after this that I decided that this was a research question that I had to pursue. The quest to come to some personal understanding of morpheme order took me several years, as there was much I had to learn in many different arenas.

I now turn to the aspect system, focusing on the major semantic force of the elements generally known in the Athapaskan literature as mode, conjugation, and aspect. I concentrate almost exclusively on Slave, as it is the language that I am most familiar with. The aspect system provides a great analytic challenge to the researcher interested in morpheme order. I argue that the aspect system of the functional complex of the verb is best viewed as consisting of three major types of items, which mark viewpoint aspect, situation type aspect, and subsituation aspect. A major part of the chapter is devoted to showing that these three components are overtly marked within the verb. The general ordering of these categories is as follows: subsituation aspect < situation aspect < viewpoint aspect. This ordering is, I argue, consistent with the scope hypothesis in that the categories to the right, or higher in the structure, impose conditions on the categories to their left.

Aspect 1: Viewpoint Aspect

Smith 1991, 1997 argues that aspectual meaning is of two major kinds, each contributing its own type of information. Situation type aspect identifies a situation as either an event or a state, and viewpoint aspect conveys a temporal perspective that focuses all or part of the situation (Smith 1997:xiii). In this chapter, I argue that Athapaskan languages have rich overt morphological systems for expressing both situation type aspect and viewpoint aspect. I begin with viewpoint aspect.

In this part of the book, I examine the functional items and their ordering. Whether the ordering of functional items follows from scope has been a point of controversy; see, for example, Anderson 1992, Baker 1992, Halle and Marantz 1994, Lapointe 1996, and Speas 1991a,b for discussion. Baker 1992 discusses this issue, asking: “The question is whether this kind of syntactic motivation can be extended to inflectional morphology … the plausibility of this approach depends on the degree to which the order of inflectional morphemes seems syntactically well-motivated” (102–103). Baker suggests that there are two reasons to pursue a hypothesis according to which the ordering of inflectional elements is principled. First are empirical considerations based on Bybee's research (1985a, 1985b). Bybee, in a crosslinguistic study of inflectional affix ordering, finds universal tendencies that she explains in terms of iconicity, or relevance to the verb. Speas 1991, in a small survey, finds similar results, as do Foley and van Valin 1984. This is one reason to believe that there are principles underlying the ordering of functional items. The second consideration that Baker brings to bear is methodological. A theory in which the ordering of inflectional elements reflects syntactic embedding allows for a restricted interface between syntax and inflection, whereas a theory in which the order of inflectional elements is idiosyncratic and language-particular makes no predictions.

Both subject and object pronominals are found within the verb in all Athapaskan languages. Their ordering is particularly puzzling, especially when subjects are considered. As discussed in chapter 9, subjects occur in two positions in the verb, with first and second person subjects in one slot and third person subjects in another (see below for refinement of this statement). A major distinction is thus made between first and second persons, or speech act participants, and others. Why should this be the case? Objects do not appear to parallel subjects; all objects occur in the same slot. Why should subjects and objects differ in this way? Finally, while objects generally precede third person subjects, this is not always true. Why is this the case? In this chapter I address these questions concerning pronominals. I postpone discussion of the ordering of pronominals with respect to other functional items until chapter 13, focusing here on the ordering of pronominals with respect to each other.

Before beginning, a brief note on the status of these elements is in order. There is lively debate in the Athapaskan literature as to whether pronominals within the verb are pronominal arguments or functional in nature (e.g., Cook 1996, Jelinek and Willie 1996, Potter 1997, Rice and Saxon 1994, Sandoval and Jelinek 1989, Saxon 1986,1989a, b, Speas 1990, 1995, Thompson 1996b, Tuttle 1996, Willie 1991, Willie and Jelinek 1996).

Athapaskan languages are often thought of as the ultimate challenge by linguists interested in issues of morphosyntax, and linguists working on these languages are alternately admired and pitied. The languages have notoriously complex verb morphology, with the verb typically described as consisting of a stem and a number of prefixes, both inflectional and derivational, whose ordering is unpredictable and must be stipulated through the use of position class morphology, or a template. In addition, phonological patterning in the verb is typically also considered to be unpredictable, and some type of boundary information is built into the template. It often appears as if any generalization that one draws about morphosyntax is falsified by the verb of some Athapaskan language. As a result, the bulk of work on Athapaskan languages has taken as its primary concern aspects of verb morphology. This book represents yet another contribution to that area. It concerns a topic that has garnered much attention in Athapaskan languages, the ordering of morphemes within the verb. My contribution, as I discuss in this chapter, is to question the notion of a template as a word formation device. Instead, I propose that morpheme ordering is to a large degree regulated by principles of scope.

Consider first some of the oddities exhibited by a verb of the description in the last paragraph. First, template morphology is highly marked in languages of the world (see, for example, Myers 1987, Rice 1991, 1993, Speas 1984, 1987, 1990, 1991a, b, and, from a somewhat different perspective, Baker 1988, 1996).

The chapters in this part of the book examine the lexical items that form part of the verb in an Athapaskan language (chapter 6). In addition, this part includes discussion of the position of the verb stem with respect to other lexical items (chapter 5) and the status of the voice/valence markers (chapter 7). It is in this domain of the verb, I argue, that we see the greatest discrepancy between surface ordering and what I claim the underlying ordering to be; semantic factors are a primary determinant in the ordering, but syntactic factors enter in as well. I argue in chapter 7 that the voice/valence elements, like the verb stems, are predicative in nature. I further argue in chapters 5 and 7 that the predicative elements of the verb originate outside of their surface positions and move to them. Finally, I argue in chapter 6 that the ordering of the preverbs, quantificational elements, and incorporates is not unexpected, but is a direct consequence of scopal ordering, with elements of greater scope appearing to the right of those within their scope. In the rest of this chapter, I introduce the reader to the lexical elements of the verb.

Definitions

Recall from chapter 2 that the surface ordering of elements within the verb word of an Athapaskan language is often expressed as a template. The Slave template is repeated in (1); see appendix 1 for templates in a number of languages.

In this chapter I examine the ordering of the lexical items discussed in chapter 4.1 argue that evidence exists, both within individual languages and across the family, for the following generalizations:

1. (i) Elements in a fixed scopal relationship occur in a fixed order with respect to each other.

2. (ii) Elements in which the scopal relationship can be reversed occur in variable order, with interpretation related to order.

3. (iii) Elements that do not enter into a scopal relationship with each other may occur in different orders, both within a particular language and across the family.

As a result, elements that are related by scope have a predictable ordering, whereas ones that do not enter into a scopal relationship must have their ordering stipulated for the particular language.

On the Ordering of the Lexical Items

In the Athapaskan literature, statements are often made to the effect that ordering is relatively fixed within the functional (conjunct) complex across the family, but that when the lexical (disjunct) complex is examined, ordering varies. For instance, Cook 1989:194–195, in a discussion of the numbering of verb prefix positions, states that “pan-Athapaskan prefix order may be established at least for the conjunct prefixes where the differences in prefix categories are primarily in the disjunct prefixes.” Kari 1989:449 makes this point explicitly: “the innermost prefixes [functional – KR] are more directly comparable across the languages than are the outermost prefixes [lexical – KR].”

I have argued that the verb in Athapaskan languages illustrates a close mirroring between function and structure, with the largest disruption in this functionally determined ordering coming in the surface position of the verb stem and voice/valence. In this chapter I review the types of cases that argue for the scope principle and consider consequences of this principle for the lexicon.

Recall the predictions of the scope hypothesis from Part I.

1. (1) a. The ordering of elements is fixed when the scopal relationship between them is fixed.

2. b. The ordering of elements may be variable when there is no scopal relationship between those elements.

3. c. The ordering of elements is variable when the scopal relationship between them is not fixed.

First consider the claim that ordering is fixed when the scopal relation between elements is fixed. This has been shown in several ways. For example, the relationship among participant subject, viewpoint aspect, situation aspect, and object agreement/number is fixed across the family, and I have argued that this ordering is predicted by the scope hypothesis. As another example, preverbs occur in a fixed position with respect to quantifiers and incorporates, again as predicted by the scopal ordering hypothesis.

The second claim is that ordering may be variable when no scopal relationship exists between elements. We have seen several cases of variable ordering. Noun class markers do not interact and occur in variable order across the family.