Trustees of Indiana University Anthropological Linguistics Tone in Navajo Author(s): Joyce M. McDonough Source: Anthropological Linguistics, Vol. 41, No. 4 (Winter, 1999), pp. 503-540 Published by: The Trustees of Indiana University on behalf of Anthropological Linguistics Stable URL: http://www.jstor.org/stable/30028725 . Accessed: 30/04/2013 14:38 Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp . JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected]. . Trustees of Indiana University and Anthropological Linguistics are collaborating with JSTOR to digitize, preserve and extend access to Anthropological Linguistics. http://www.jstor.org This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions Tone in Navajo JOYCEM . McDONOUGH Universityo f Rochester Abstract. Navajo is generally classified as a tone language. Yet an important and large class of verbal morphemes, the morphs of the conjunct domain, are reportedt o be underlyingly toneless. The relationship between the redundancy of the vowels of many of these morphs, their inflectional status, and their tonelessness is unexplored and calls for an explicit investigation before the categorical assumptions behind traditional phonological classification can be applied. This article reports on an instrumental investigation of the tonal system of Navajo and the relationship of these results to its phonology and typology. Navajo is shown to be a tone language insofar as each syllable carries a tonal specification. However, tone specification is very stable, and the tone contours are apparently sensitive to the morphologicalb oundaries in the word. This co-occurrencein dicates that mapping between phonology and phonetics in Athabaskan, as well as the arguably attendant tonal variation in the family, is best accounted for by clearly defined implementation strategies more charac- teristic of laboratoryp honologyt han of pure autosegmental analyses. Very little explicit information has beeng iven on tone in the large body ofAthabaskan grammatical, lexical and text material so farpublished. -Edward Sapir (1925:185) 1. Introduction. In the seventy-odd years since Sapir's remark, explicit studies of tone systems in Athabaskan remain rare. The variety of tonal types reported among these closely related languages, the complexity of their morpho- logies, and the absence of clear autosegmental-type tonal phonologies (such as those found in the Bantu languages) make it difficult to arrive at a phonological classification of these tonal systems. Instrumental analyses of tone will be important in clarifying these issues. The present article is an investigation into the tone system of Navajo based on an instrumental study of the pitch (or fo) contours in verbs. Navajo, spoken in the Navajo Nation in the southwestern United States, is a member of the Southern or Apachean branch of the Athabaskan family, a group of closely related languages that share a complex verbal morphology. The study has two aims. One is to provide an explicit account of the phonetic facts of tone in Navajo as a basis for uncovering its phonological structure and the nature of its tone system. The second, broader aim is to lay a foundation for 503 This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 504 ANTHROPOLOGICALL INGUISTICS 41 NO. 4 discussion of the interaction of pitch and morphologicald omain in Athabaskan, and of the development and typology of tone systems across that family. In tone languages, pitch is used contrastively, together with features like nasality or vowel length, to build vowel prosodies-sets of long and short, high- pitched and low-pitched, nasal and oral vowels. Navajo has such a system. In pure tone languages, pitch has segment-like distribution and lexical items are densely specified for tone. But pitch is independent of segmental material, and tone systems reflect this fact. All phonological and phonetic theories acknow- ledge that constraints operate on the tonal tier as an independent tier. In sys- tems where tonal specification is less dense, tone function tends to be more loosely tied to lexical items and associated instead with larger units such as the phonological phrase. In pitch accent systems, for instance, tone is borne by lexical items, but it is not lexical in function. In such systems, the contrast is between the presence and absence of an "accent"o r mark (Pike 1948; McCawley 1978). A marked item bears a pitch specification; an unmarked one either does not, or gets its specification by virtue of its position in the larger unit. Con- straints then operate on the larger unit, delimiting the number and kind of pitch turning-points within a domain.M eeussen's Law in Bantu, which is a constraint against two adjacent high-tone specifications, is an example of this. In this sense, pitch has moved fromf unctioning as a contrastive unit to functioning as a domain delimiter. In yet other types of language, pitch is even more inde- pendent of the segment and the lexicon; for instance, intonational tunes may develop that can be aligned in systematic ways with lexical items within defined domains. These are intonational systems like that of English. McCawley (1978) has pointed out that these tonal typologies represent points in a continuum. These points or types can be defined by three main factors: the independence of the tonal tier, the function of tone (contrastive or delimiting), and the density of tonal specification. In pure tone systems, pitch is contrastive, the least independent of text, and the most densely specified. Intonational systems are at the other end of the spectrum; pitch is not lexically contrastive, it is largely independent of the text, it often serves as a domain delimiter (Jun 1993), and (other factors being equal) it is less densely specified across an utterance in such languages than it is in other types. There are many examples of systems where pitch is used in two independent ways. In intonation systems like English, pitch is used to define stress and, independent of stress, as the building block of the intonational tunes that are superimposed on the stress system. It is less clear how the points in the typo- logical continuum might interact within a system. It is difficult to see how intonational tunes could be superimposed on a tone language with its densely specified tonal tier, how a pitch accent system and a tone system might interact, or what an intermediate type might be like. I will argue, however, that all of these possibilities occur in Athabaskan. The richness of Athabaskan morphology provides a fertile ground for exactly these intermediate tonal types, and this This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 1999 JOYCEM . 505 MCDONOUGH accounts for the range of types, the dearth of tonal phonology, and the typo- logically odd tonal systems found in the family. In fact, I suggest that any tone language with a rich inflectional domain will be inclined to develop systems like the ones found in Athabaskan because of the properties common to all inflec- tional or function morphemes, such as severely reduced contrasts and the tendency for function morphemes not to bear stress. In a language like English, function words are dispersed throughout a domain; in Athabaskan, as in other morphologicallyc omplex languages, there are domains that consist of strings of inflectional morphemes, and these domains impose tonal constraints that shape tone function, density, and independence. Sections 2 and 3 of the article provide an overview of the morphological design of Navajo (and Athabaskan in general) and of tone in Navajo. Section 4 reports on an instrumental study of the pitch contours in Navajo verbs. This is followedb y discussions of Navajo tonal phonology and of the implications of the results of the analysis for the structure and design of tonal systems in Navajo and Athabaskan. 2. An overview of Athabaskan morphology. A concise and motivated model of verbal morphology is essential to any investigation of the Athabaskan sound system, because morphophonemics tructure is a construct of morphologi- cal models and assumptions about morphophonemics tructure underlie analyses of sound systems. Here I describe a working model of the verbal complex adapted to the purposes of this article. There are several, often conflicting models of the morphologicals tructure of the Athabaskan verbal complex( Morice1 932; Sapir and Hoijer 1967; Kari 1976, 1989; Young and Morgan 1980, 1987; McDonough1 990, 1998; Hargus 1995). For many purposes, the abstract slot-and-filler or position-class template of Young and Morgan (1987) is assumed. Important to the present topic are the three well-established morphologicald omains within the Athabaskan verbal complex: the proclitic "disjunct"( D) and inflectional "conjunct"( I) domains and the verb stem (V) (Sapir and Hoijer 1967; Kari 1976; Young and Morgan 1987). The boundary between the disjunct domain and the conjunct domain (the D-I boun- dary) is marked with a crosshatch (#), in keeping with traditional Athabaskan practice. The boundary between the inflectional domain and the stem domain (the I-V boundary) is marked with a double slash (//). (1) hasdlbis ha#s8-411//bs D I V up#SPERF1:S -CL//handle.globular.object:PERF 'I drove it up.' (Young and Morgan 1987:429) Underlying morphophonemic structure within the domains is a matter of debate, but tangential to the present discussion. In this article, I assume a This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 506 ANTHROPOLOGILCINAGL UISTICS 41 NO.4 version of the morphologicalt emplate of Young and Morgan (1987), except for positions VI and VIII in this template. Here I assume that these two positions are occupied by a portmanteau morph of subject and number agreement, marked for conjugation, which I call the "mode-subject"m orph.' In (1) above, this morph is a perfective from the s- (or "si-")c onjugation, 'SPERF:1S',m eaning the first person singular form of the "si" perfective (see also Kari 1989). The conjugationsm ark "mode"t:h e Navajo tense-aspect system of conjugations that mark the imperfective and the perfective, the optative and the future. Young and Morgan list the conjugations in full (1987:200-201, 206-50). The port- manteau morphs of mode-subject are taken directly from their lists. The disjunct domain (D) contains a group of morphs with cliticlike proper- ties and primarily adverbial content (although pronominal constructions occur among them). The domain is optional in the sense that the morphemes in this domain are optional, though there are subcategorization constraints that may require proclitics.2 I refer the reader to the literature on the well-established domain boundaryb etween the disjunct and conjunct domains (Sapir and Hoijer 1967; Kari 1976; Young and Morgan 1980, 1987). The conjunct domain (I) is marked as containing six position classes and three subclasses in Young and Morgan's template (1987). For our purposes, some simplificationi s in order.S ome of these morphemes are obligatory, such as the mode, person, and subject morphemes of the mode-subjectm orph.3B ecause these morphemes appear in a portmanteau mode-subject morph, this morph is always present in this domain in the verb. This morph always constitutes the penultimate syllable in the verb.4 Except for this morph, the syllables of the conjunct domain are overwhelmingly consonant-vowel (CV), and the vowel is i. Thus, outside the obligatorym orphemes,v owel quality and syllable shape of the morphs of the I domain are predictable.5 In addition, in this article, the "classifier"m orph, actually a type of valence marker, is assigned to the I domain, in keeping with Young and Morgan's practice. The classifiers have an unusual prosodic status in the language. They often do not appear in the surface forms, although they have a series of sys- tematic effects on surrounding consonants. The reader is referred to the extensive literature on these consonant mutations, which fall under the rubrics of classifier phonology and the "d-effect"a lternations (Sapir and Hoijer 1967; Howren 1971; Young and Morgan 1987; Bennet 1987).6 The third domain is the verb stem domain (V). In Navajo, this is the final syllable in the word.7 Lexical items from this category, like the mode-subject morphs, may have long vowels and syllable codas. It is an interesting fact that this domain is the only place in the word where the full set of vocalic and con- sonantal contrasts occur (McDonough and Ladefoged 1996). In this simplified overview, there are two obligatory morphs in the verb word: the portmanteau mode-subject morph and the verb stem. Every verb in Athabaskan has morphemes from at least two domains: conjunct (mode-subject) This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 1999 JOYCEM . MCDONOUGH 507 and the verb stem. Thus verbs are minimally bisyllabic. Some examples of verbal constructionsa re providedb elow. The disjunct (D), conjunct (I), and verb stem (V) domains are marked. (2) honillid ho-nf/l/id 3S-NPERF:1/3S//appear/come:PERF I V 'I/he appeared/came.('Y ounga ndM organ1 987:76) (3) yishcha (y)ish//cha OIMP:1S//cry:IMP I V 'I cry.'( Younga ndM organ1 987:779) The form in (1) has morphemes from all three domains; while that in (2) contains two conjunct morphs and the verb stem, but no disjunct morphemes. The form in (3) is a minimal verb, with the two obligatory morphs, the mode- subject portmanteau ish (0-imperfective, first person singular) and the imperfect form of the verb stem (-cha 'cry'). An onset requirement in Navajo accounts for the initial consonant in yishcha. In summary: * There are three domains in the Navajo verb: the proclitic disjunct (D) and inflectional conjunct (I) domains and the verb stem (V). * The minimal verb is two syllables long, with morphemes from two domains: the mode-subject conjunct morphs (I) and the verb stem (V). * The vowels of the conjunct morphemes, except for mode-subject, are over- whelming i, and the syllable structure of material in this domain is CV. 3. Tone in Navajo. 3.1. Tone typology. The Athabaskan languages are generally classified as tone languages, though there is considerable variation in the tonal types re- ported among them (Golla 1970; Krauss 1964, 1978; Tuttle 1991, 1998; Hargus and Tuttle 1995). However, comprehensive descriptive studies of Athabaskan tonal phenomena are scarce. For instance, one of the earliest grammars of Athabaskan, Morice's (1932) grammar of Carrier,n ever mentions tone (cf. Pike 1986; Story 1989; McDonough1 989). Young and Morgan (1980, 1987, 1992) give tone little attention. Even phonological studies such as Rice (1989) on Slavey, oriented to particular representational issues in phonology, or Cooke (1971) on Sarcee, do not purport to lay out the tonal design of the language. There is almost no work on Athabaskan tone comparablei n scope or intent to work done by Michelson (1988) or Chafe (1977) on tone and intonation in Iroquoian, for instance.8 This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 508 ANTHROPOLOGICLAILN GUISTICS 41 NO. 4 There are several possible reasons for this neglect. One reason is simply that Athabaskan languages are not comprehensivelyd ocumented. Another relates to Sapir's (1925) comment that tone is not "phoneticallyc omplex"i n Athabaskan, by which he means that there are very few systematic tonal alternations in Navajo. Alternations are the source of phonological data in traditional genera- tive analyses. A third reason relates to the morphological complexity of Atha- baskan. Tone is more closely tied to and varies less independently of mor- phophonology (another of Sapir's observations) than is generally assumed in autosegmental analyses. To put this observation another way: the tonal system is not free of the morphology. An example is the alternation between the two forms of the second person singular subject marker for the imperfective conjugation, ni- and H-tone. This alternation is conditionedb y a morphological environment (see Kari 1976:40). Navajo is described as having two tone levels, high and low; low is the unmarked or default tone. Young and Morgan (1987:xiii) claim that each sylla- ble carries a tone, the telltale mark of a tone language. However, complications arise in applying this classification to Navajo. In pure tone languages, tone is lexically contrastive. Although tone is contrastive in Navajo verb stems (the V domain), it does not bear a similar functional load in the inflectional domain (I), where tonal distribution is skewed and low tone is common enough to be con- sidered the norm.9 As Ladd (1996:148) points out, there are no purely phono- logical distinctions in structure among tone, accent, and intonation systems. On the phonological level, all of these systems are alike in being comprised of a series of tonal specifications, such as H or L. The phonology operates on these specifications. There is no phonological relationship between the function of a feature like tone and the phonological patterns it participates in. The distinc- tions among these various tonal types emerge from the distinct functional loads that tone can carry in a grammar( Beckman 1986:44).1oI nsofar as function plays a role in determining tonal typology, the differences in the functional load of pitch between stems and the I domain in Navajo may have resulted in the development of distinct tonal patterns in the two domains. I suggest that this is the case in Athabaskan in general, and that this is the origin of the tonal varia- tion among the Athabaskan languages. As discussed in the previous section, verbs have a minimum of two syllables and always have at least one morpheme from the domain of redundant tone specification, the I domain (see examples (1)-(3)). Since contrastive pitch and tonal density are important to determining whether a language is a "tone lan- guage," clarifying the relationship between redundancies and tonal specifica- tion is essential to our understanding of the tonal phonology and typological classification of the Navajo verb. By extension, the investigation will also eluci- date the nature of tonal typology across Athabaskan. In the next section I will point out some of the salient properties of tone in the three domains of the Navajo verb. This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 1999 JOYCEM . 509 MCDONOUGH 3.2. Tone and morphology. Most descriptions concur that there is little tonal phonology in Athabaskan. Commenting on Sarcee, a Northern Athabaskan language with three contrastive tone levels, Sapir noted that "Phonetically the [tonal] complexity is only moderate" (Sapir 1925:189). On the other hand, he notes, the interaction of tone and morphology is less straightforward: Tone is not a matter entirely of the inherent pitch of a given word or element nor is it entirely a matter of grammaticals ymbolism.B oth types of tone function are inextricably woven in Sarcee in a system of considerable complexity, from a morphologicals tandpoint. [Sapir 1925:189] Sapir (1925), Cook (1971), and others have recognized three types of tone in Athabaskan, which Sapir named "inherent tone," "grammatical tone," and "varying tone." These types correspond to lexical tone, tone that marks gram- matical or inflectional categories such as aspect, and alternating tone. The first two types are relatively straightforward; it is of some interest that the third type, "varyingt one," is assigned the same status as these first two types. In Navajo, lexical tone appears in noun and verb stems and in the clitic groups (the disjunct and enclitic domains). Examples of disjunct proclitics that are specified for tone are: na-'about, around', specified as L, and ch'i 'around, out', specified as H. The verb roots also carry tonal specification. The root bggs "globular object" carries L in its basic form, while the root chaid 'act with the fingers or arms' carries H. (See Young and Morgan's root dictionary [1987] for full sets of the alternate forms of roots.)11 Young and Morgan list the imperfective forms of stems as basic. Surface realizations, called "verbs tems," are specifiedf or aspect.12T he imperfective and perfective stems of the two preceding roots are bags (imperfective, specified for this root as L), b~@is( perfective, specified as H), and child (imperfective, specified for this root as H) and chid (perfective, specified as L). These are examples of grammatical tone. Thus the tone type found in verb stems is both lexical ("inherent")a nd grammatical.13 Syntagmatic tone appears on the mode-subject morphemes of the conjunct domain. Examples are the third person singular forms yicha 'he's crying' (im- perfective) and yicha 'he cried' (perfective).14T he high tone alternation on yi marks the distinction between the imperfective and perfective (cf. Kari 1976). Use of tone syntagmatically is widespread in Athabaskan (Story 1989; Mc- Donough 1989 on Carrier;R ice 1989 on Slavey). The characteristics of pitch in Navajo stems may be summarized as follows: The lexical entry of each verb stem (the root) specifies either a high or low tone, although there are few minimal pairs of any kind among these roots. The distribution of high and low tone is lexically contrastive in roots, and H and L tones are about equally represented. This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 510 ANTHROPOLOGLIINCGALU ISTICS 41 NO4. * The surface forms of verb stems are marked for aspect. Aspect is expressed in large part by tonal differences. * Except for some frozen forms, there are no contour tones in stems. The tone bearing unit is the syllable. * There are no tone interactions between I-domain roots and morphemes. Some tonal alternations with enclitics take place under limited conditions (see section 5). The characteristicso f pitch in the inflectional or "conjunct"( I) domain in Navajo verbs are as follows: * The distribution of high and low tone is asymmetric in the I domain; low tone is the default specification. This unexplained asymmetry is accom- panied by a reduction of segmental phonological contrasts in this domain. * Phonological alternations affecting pitch occur, but are restricted to pro- gressive spread under limited conditions across the disjunct-conjunct boundary (see section 5). Tone in the disjunct domain has the following properties: * Tone is lexically contrastive (e.g., na-'about' versus n- 'upward'). * H and L tones are more evenly represented among morphemes. * There is progressive spread of H tone to the conjunct domain. To address questions raised by these characteristics, an explicit study of Navajo tone was undertaken. In the study, two particular questions were asked. The first concerns the density of tonal specification in the Navajo verb and relates to the issues of typological classification. In McCawley'st ypology (1978), tone languages are densely specified, with at least one tone per tone-bearing unit. This tonal density is directlyr elated to the function of tone as a contrastive unit. Certain distributional properties are characteristic of tone when tone does not have a contrastive function: (i) tone is less densely specified on the phono- logical tier, and (ii) tone is associated to a domain larger than a lexical item, resulting in distributional asymmetries between tones. The morphemes of the inflectional domain in Navajo are characterized as being underlyingly toneless (i.e., tone is predictable and not contrastive), and the distribution of H and L in this domain is asymmetric. Thus we may expect that this domain will have a phonetically less densely specified tonal tier and that tone will carry a distinct functional load here. For instance, surface tone contours in this domain may be due to extrapolation between the H targets pro- vided by morphemes marked with lexical H tone, or, if there are none, between the tones in the disjunct domain and the stem domains with no L targets. This is an empirical issue, and one that is investigated in this article. Relevant to this issue is the fact that these domains are comprised of functional morphemes, which are often associated with limited phoneme sets and reduced contrasts. The second question is an extension of the first and addresses the inter- This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions 1999 JOYCEM . 511 MCDONOUGH action of pitch with the morphology of the Navajo verb. This question involves the relationship between function and pitch as well as the particulars of Atha- baskan tone. Delimitation and culmination are properties classically associated with stress and accent, and not with tone systems (Trubetzkoy 1969). Yet the internal morphological boundaries of the Navajo verb are reflected in the im- plementation of low tone, as I will demonstrate below. In conjunction with tonal asymmetry and the redundancy of low tone, these facts point to a tonal system in the inflectional domain in Navajo distinct from that in other domains. This, I suggest, is the probable source of tonal variation in the language family. A third issue, only partially addressedh ere, is that of tonal universals. Some candidates for these are declination, boundary tones, and tone-intonation inter- actions. Since this is the first study of its kind on Navajo, it cannot purport to address these matters fully, especially in view of the fact that there has to date been little theoretical work dealing with the tonal systems of morphologically complex languages. Some particular questions are elucidated, however. No straightforwardd eclination or downstep is apparent in Navajo. Nor can we give a simple account of boundary tones; if they exist at all in Navajo, they are not apparent in the data. These issues are addressed in section 6. Since no previous work of this kind has been done on Navajo, a controlled set of data was used, restricting the study to an instrumental investigation of the pitch contours of the Navajo verb in a prepared list of citation verb forms. These constraints are discussed in section 4. 4. A instrumental study of pitch in Navajo. The primary focus of this article is a study of the local fo (pitch) targets and the intonation contours of speakers' utterances, collectedf romr ecordedr ecitations of a prepared word list. From these contours,i nformationa bout phonological specification and the tonal structure of Navajo can be inferred. This is discussed in section 5. The nature of the recitation and the word list used in the recitation is discussed below in section 4.2. 4.1. Model. We will assume a "target and interpolation model" of tone, follow- ing those proposedb y Girding and Bruce (1981), Girding (1984), Lindau (1984), Beckman (1986) and Beckman and Pierrehumbert (1986), Laniran (1993), and others. In this model, the pitch contours of an utterance are the result of the implementation of phonological tone specification by algorithms that assign or alter fo values (the targets) and interpolate transitional pitch values between those targets.15 Lexical items with tonal specifications are submitted to the lexical phonology,w here word formationp rocesses and phonological constraints modify underlying representations, including tonal specifications, resulting in a phonological tone pattern. We will assume that the output of this process is a word, a verb form marked for tone. Thus the lexical output for the verb form yidzis 'I drag it' would have H and L tones associated with the first and second This content downloaded from 128.151.229.209 on Tue, 30 Apr 2013 14:38:26 PM All use subject to JSTOR Terms and Conditions