Tone in Cherokee Nouns draft of May 29, 2005 Tone and Pitch Accent in Cherokee Nouns Keith Johnson University of California, Berkeley Abstract. Tone in Cherokee nouns is described in this paper as a hybrid system of metrical pitch accent and lexical tone. Four speakers of Oklahoma Cherokee spoke nouns in a frame dialog that put each of thirty-seven nouns in citation form, in a declarative sentence, a question, and with contrastive focus. In addition to an optional L*H pitch accent that may occur on the penultimate or antepenultimate syllable only, and three distinctive tones (L fall, H, and H fall) that are restricted to occur only on non final, unaccented syllables, the study also identifies some aspects of Cherokee intonation contours. 1. Introduction. This paper has two aims. First, the paper presents acoustic phonetic pitch traces of Cherokee nouns spoken in citation forms, and in three types of sentential context. These phonetic data supplement a literature on Cherokee tone that contains very limited phonetic data and even more limited data on Cherokee intonation (Lindsey, 1985; Wright, 1996). Second, the paper presents an analysis of the surface phonology of Cherokee tone. This analysis is inspired by, and largely compatible with previous phonological analyses of Cherokee tone (Lindsey & Scancarelli, 1985; Lindsey, 1985; Lindsey, 1987; Wright, 1996) broadly situated within the autosegmental/metrical approach to tone and intonation (Bruce, 1977; Pierrehumbert, 1980; Beckman, 1986). Lindsey (1987) treated Cherokee as a kind of pitch-accent language with accents filled by either [+High] tone or [+slack] vocal folds depending on which mora of a long vowel is considered to be accented (H on the first mora, [+slack] on the second). Wright (1996) on the other hand treated Cherokee as a hybrid tone/accent language with pitch accent marking some syllables and lexically 1 Tone in Cherokee Nouns draft of May 29, 2005 specified tone on others. The analysis presented here builds on Wright’s insight that lexical tone coexists with pitch accent in Cherokee. The suggestion that languages may exhibit hybrid prosodic organization may seem surprising or a little overly complex, so it is useful to bear in mind that several languages have been given this kind of “hybrid” analysis, including Beijing Mandarin Chinese (Chao, 1968; Peng et al., 2005), the Bantu languages Tonga (Goldsmith, 1984) and Kizigua (Kenstowicz, 1989), the English- based Creole Saramaccan (Good, 2004), and Stockholm Swedish and Venlo Dutch (Gussenhoven & Bruce, 1999). In fact, when it comes to the cross-linguistic typology of prosodic systems it is probably better to situate a language along several scales rather than try to categorize it as one of only two or three possible types (Hyman, 2001; Beckman, 1986). Hyman’s (2001, 2005) typological oppositions for tone systems are particularly useful in characterizing Cherokee. Therefore a brief summary, with some liberty of interpretation, is given here. Opposition 1: Free versus culminative. Tone is said to be free when “multiple tones may occur within the same word. Lexical tone, in the canonical case, requires specification of pitch for every syllable in every word. In some languages (usually found in Africa) lexical tones maybe drawn from a relatively small inventory of stationary pitch targets (H or L). In other tone languages (usually found in Asia), lexical tones are drawn from a larger inventory, usually with distinctive pitch contours as well as level tones. Cantonese (Wong et al. 2005) is an example. In this language, tone is paradigmatically contrastive (e.g. Hashimoto (1972) identified lexical tones [55], [35], [33], [23], [22], and [11]) and every syllable, except a few minor grammatical particles, must have a tonal specification. Prominence marking tonal events are culminative if every word has at most one syllable marked for the highest degree of prominence. As this culminative function is realized in discourse it may be used to bring pragmatically important words into the foreground. This is seen in English, where one syllable in most words can be identified as bearing primary stress, and pitch accent may be placed on stressed syllables to mark focus. Opposition 2: Optional vs. Obligatory. It is sometimes said that in English every word has at least one syllable marked for primary stress. This for Hyman (2005) is a definitional property of 2 Tone in Cherokee Nouns draft of May 29, 2005 stress accent systems. Clearly some caveats need to be mentioned if we think of closed-class function words as essentially stressless, but English is also very different from languages that have unaccented nouns and verbs (Tokyo Japanese, for example). When accent is optional word order or overall pitch range may be used to mark discourse prominence. Opposition 3: Distinctive vs. Demarcative. In syntagmatic structure, tone may be used in word segmentation as a cue for the location of word boundaries in running speech. For example, if stress or pitch accent is restricted to occur on particular syllables (such as the initial, penult, or final) then the presence of pitch accent on a syllable indicates the location of the word boundary. Freely occurring tone, on the other hand is not a cue for word segmentation because tone plays a more paradigmatic role. Words are distinguished from each other by their pitch patterns. This opposition, like the others listed by Hyman is rarely absolute within a language. So, in addition to its syntagmatic functions, prominence-marking in English may also play a limited paradigmatic role, analogous to the role played by tone in a lexical tone language (for example, conTRAST versus CONtrast). Opposition 4. Multi-valued vs. Privative. If tone must take one of two values (H or L) the word prosody can be described as privative. Many Bantu languages fit this description. Prominence marking in English is non-privative because we mark prominent syllables with any one of an inventory of possible pitch accents (*H, *L, *L+H, etc.). Tokyo Japanese on the other hand uses a more privative system, marking prominence with a HL fall to the exclusion of any other pitch pattern. Opposition 5. Equal vs. Subordinative. Within a word, if more than one syllable is prosodically marked, the system is described as subordinative if the relative strength or prominence of the marking is unequal. We see this in the distinction between primary and secondary stress in English. Opposition 6. Pitch vs. Complex phonetic correlates. The phonetic dimensions used to mark prosody vary from language to language. In Cantonese pitch is the main correlate of tone, while in 3 Tone in Cherokee Nouns draft of May 29, 2005 stress accent languages, a combination of suprasegmental properties (pitch, duration, loudness) is commonly used to mark prominence. Segmental properties may also mark prominence. For example, one common way that prominence is marked with segmental, rather than suprasegmental features is by the inventory of segmental contrasts that exists in certain syllables. Thus, is it not rare at all to find that a language has one set of contrasts in open-class words, and another smaller set of contrasts in closed-set words. For example, we find a tendency toward this in English with coronals used is most closed-class words and in affixes. Navajo has the same pattern - coronals in grammatical particles and larger set of contrasts in stems. Laryngeal contrasts show similar restriction to stems or open-class words in some languages. An expanded set of contrasts lends a type of prominence to syllables by giving those positions in words greater phonetic richness relative to a neutral or mono-phonetic pronunciation. These oppositions we could add one more. If tone is specified on all or most syllables we can consider the language to have a dense tone specification, while if only one or few syllables are marked the language has sparse tone specification. As table 1 shows, the difference between a prototypical “lexical stress” language like Cantonese and a prototypical “stress accent” language like English shows up in differences in all seven of the prosodic typology oppositions. However, in addition to providing a description of prototypical cases, this system of classification provides a nice characterization of the prosodic typology of a language like Somali, in which a high tone maybe optionally assigned to final or penultimate syllables (Hyman, 1981). Table 1. Characterizing the prosodic typology of Cantonese, English, and Somali. 4 Tone in Cherokee Nouns draft of May 29, 2005 Opposition Cantonese English Somali 1 free culminative culminative 2 obligatory obligatory optional 3 distinctive demarcative demarcative 4 multi-valued privative privative 5 equal subordinative ?? 6 pitch complex pitch 7 dense sparse sparse The main claim of this paper is that prosody in Cherokee nouns is best described as a hybrid system mixing two types of prosody. One promenence-marking pitch accent per word may occur on the penultimate or antepenultimate syllable, while lexical tone may co-occur before the pitch accented syllable, and in longer words there may be more than one lexical tone. This suggests that in Hyman’s oppositions Cherokee word prosody is culminative, optional, and demarcative; uses multi-valued, pitch, cues without subordination within words; and tone is marked more densely than English and Somali, but less densely than Cantonese. Evidence supporting this analysis will be drawn from pitch contours calculated from recordings of four speakers of Oklahoma Cherokee. Section 2 gives some general background on the Cherokee language, and surveys very briefly prior descriptions of Cherokee tone. The recording and analysis method of the present study are given in section 3. Section 4 presents evidence suggesting that some Cherokee nouns have a L*H pitch accent on the penult or antepenult syllable. This section also gives some very preliminary suggestions regarding phrasal intonation patterns observed in declarative sentences, questions, and focus constructions. In section 5, the three lexical tones (H, L, and HL) are introduced. Pitch traces in this section show their typical pitch patterns and how they interact with pitch accent and phrasal intonation. 5 Tone in Cherokee Nouns draft of May 29, 2005 2. Cherokee. Cherokee is an Iroquoian language (related to Oneida, Seneca, Mohawk, and Onondaga, see Michelson, 1988) spoken in N. Carolina, and Oklahoma. In one of the most important modern reference works on Cherokee, Feeling & Pulte’s (1975) Cherokee-English Dictionary, phonetic tone is written with pitch numbers so that [1] represents the low end of the speaker’s pitch range and [4] represents the high end. Feeling and Pulte identified six tones in Cherokee [1], [2], [3], [4], [23], and [32] and wrote tone on every syllable in the dictionary, except the last syllable of the word. They described nouns as having a falling tone on the final syllable and also noted that the final syllable (which is said to be nasalized) is often deleted in connected speech. In order to make this paper as accessible as possible to Cherokee language specialists and speakers the examples shown in this paper will be written in the romanization used in Feeling and Pulte (1975). This alphabet is a variant of the Americanist phonetic alphabet (Pulum & Ladusaw, 1986), with the following exceptions. The letter “v” is used to write schwa [], and sequences “hl”, “hn”, and “hw” represent voiceless [l], [n] and [w]. For typographical convenience, long vowels are written with two vowel letters [aa] and short vowels with one letter [a]. As is found in many other native American languages, Cherokee has complex morphology, particularly with verbs. This is illustrated very briefly in (1). (1) A few of the forms of the verb “to speak” ga2wo3ni2ha “he is speaking” da2ga2wo3ni2ha “he will speak” da2ga2wo32ni2hi2se23li “he will speak for him” 6 Tone in Cherokee Nouns draft of May 29, 2005 da2ga2wo32ni2si23hli “he will come to speak” Derived nouns (usually from verbs) often have a tone [4] in the penultimate syllable (2). And this common tone pattern may on rare occasions serve to mark a noun even without a deverbal suffix. (2) u1dlv3ga “he’s sick” ju2ni2dlv1gi4?i “hospital” (3) gay2go1gi “lie” gay2go4gi “liar” After Feeling and Pulte’s description, several other researchers have studied Cherokee tone. Lindsey (& Scancarelli, 1985; 1985; & 1987) broke from the “tonal” analysis given by Feeling and Pulte (1975) suggesting that tonal specification in Cherokee is much more sparsely specified than the tonal analysis implies (see also Cook, 1979). In particular, he took tone [2] - the most frequently occuring tone mark in the dictionary to be a default pitch specification that is added to the phonological representation near the end of the phonological derivation. He also made the important observation that syllables transcribed as tone [23] in the dictionary have two realizations. Some syllables written [23] in Feeling and Pulte show quite significant pitch rise, while others do not. The phonetic results reported here support this observation. Lindsey also used the glottal feature [+slack] to account for the falling pitch patterns in the tones identified in this paper as low and high-fall. Finally, Lindsey (1985) concluded that final vowels in Cherokee phonological phrases (usually individual lexical items?) are associated with a H% boundary tone. Wright (1996), building on Lindsey’s autosegmental/metrical analysis, proposed that Cherokee is a hybrid system making use of pitch accent and lexical tone. This contrasts with Lindsey’s attempt to characterize all pitch contrasts in Cherokee using a system of rather dense accent marking (often with two accents per vowel - one per mora), and realization rules in which accents 7 Tone in Cherokee Nouns draft of May 29, 2005 on the first mora of a vowel are realized as [+High] tone and accents on the second mora of the vowel is realized as [+slack] vocal folds. Wright’s analysis recognized distinct lexical tones that play a paradigmatic role in the prosodic system as well as pitch accents that play more of a syntagmatic (culminative, demarcative) role. The present study builds on these prior analyses with more detailed phonetic results that support Wright’s (1996) conclusion that Cherokee is a hybrid system. 3. Method. 3.1 Speakers. Four speakers (2 men, 2 women) of Oklahoma Cherokee participated in this study. One speaker (DF) was the lead consultant on this project and participated in each of the recording sessions. The pitch traces shown in this paper are drawn from the speech of the male speakers. These are representative of the women’s patterns, but with better pitch tracking. 3.2 Materials. Word list recorded for this study was composed of examples of the most frequent noun tone patterns found in Feeling and Pulte (1975). The full list of words is given in Appendix 1. The list was composed of six two syllable tone patterns, twelve three syllable patterns, thirteen four syllable patterns, and six examples of tone patterns on five syllable words. These were produced in a script that elicited a citation, declarative, question, and contrastive focus forms, as illustrated in (4). Each word appeared in each context. (4) Part of the script for [jolani] and [kawonu]. jolani “window” jolani jigowahta. “I saw a window” 8 Tone in Cherokee Nouns draft of May 29, 2005 jolani-s digowahta? “Did you see a window?” hla kawonu-dv jigowahta. “No I saw a duck” 3.3 Recordings. A Shure SM58 hand held dynamic microphone was used to record each speaker on a Sony DM1 digital audio tape recorder. Speaker AH was recorded at a picnic table in Honor Heights Park in Muskogee, Oklahoma. Speaker DM was recorded in the community room at the Tahlequah, Oklahoma public library. Speaker EW was recorded in her home near Tahlequah. Speaker DF (Durbin Feeling) was the lead speaker and linguistic consultant for this project. He recorded the script with each one of the other speakers. We used the first of these recordings; the one made in the park. DM, AH, and EW repeated each item in the script after the English gloss was read and after DF read the scripted utterance. That is, three people spoke each phrase of the script, one after the other like this: “window” (linguist), “jolani” (DF), “jolani” (DM, AH, or EW), “I saw a window”, “jolani jigowahta” (DF), “jolani jigowahta” (DM, AH, or EW). 3.4 Analysis. The recordings were transferred to digital audio files and down-sampled to 22.05 kHz samples per second. A research assistant (Tsan Huang - now assistant professor of Linguistics at SUNY Buffalo) marked all of the vowel onsets and offsets in the test words. F0 was calculated at 10 msec intervals in the test words using the algorithm implemented in Entropics XWaves “get_f0” program. F0 was then extracted (using a perl script) into a large data file taking five equally spaced points during each vowel of each test word. Occasionally (less than 5% of the points), the F0 tracking algorithm failed to report an F0 value at one of the five time points. In these cases, the value was estimated by interpolation from neighboring points. Finally, to produce the figures shown in this paper an R script (www.r-project.org) took an 9 Tone in Cherokee Nouns draft of May 29, 2005 average F0 trace for the two male speakers. The average F0 trace was made by taking the average time value and F0 value of each point in the vowels of the test word being plotted. 4. Pitch accent. The analysis of pitch accent here assumes that there are three kinds of nouns in Cherokee. Two kinds have a pitch accent either on the penultimate syllable (5a) or on the antepenultimate syllable (5b). The pitch accent is hypothesized to be the first syllable of a strong-weak (trochaic) foot because, especially in antepenultimate syllables, where the effects of intonation do not obscure it, the pitch pattern of the second syllable of the foot is dependent on the pitch accent. Penultimate accent is written in Feeling and Pulte (1975) as [4], while antepenultimate accent is written [23]. However, as Lindsey (1987) noted, not all syllables given [23] in the dictionary are realized with the L*H pitch accent. The third basic pattern (5c) is the most common pattern, especially in underived nouns, and has no pitch accent. (5) Pitch accent patterns in Cherokee nouns. (a) σ σ σ] Penult accent L*H (b) σ σ σ] Antepenult accent L*H (c) σ σ σ] Unaccented 0 ___________________________ Figures 1, 2 and 3 about here 10