Vol. 8 (2014), pp. 613–636 http://nflrc.hawaii.edu/ldc/ http://hdl.handle.net/10125/24618 The Study of tone and related phenomena in an Amazonian tone language: Gavião of Rondônia Denny Moore Museu Paraense Emílio Goeldi - Ministério de Ciência, Tecnologia e Innovação Julien Meyer Museu Paraense Emílio Goeldi - Ministério de Ciência, Tecnologia e Innovação Collegium of Lyon, Ecole Normale Supérieure de Lyon This paper describes the methods used to study the tone and some related phenomena of the language of the Gavião of Rondônia, Brazil, which is part of the Mondé branch of the Tupi family. Whistling of words by indigenous informants was discovered to be a very effective method for obtaining phonetic accuracy in tone and length. Methods were devised to map out the system of tone and length. They were subsequently used in the study of other Ama- zonian languages, including Karitiana, Munduruku, Zoró, and Surui of Rondônia, with success. Some notes on tone considerations in orthography are offered, as well as notes on procedures that proved useful in the diachronic study of tone in the Mondé languages. Methods for the study of natural whistled speech used for distance communication are also described, with special attention to the whistled speech of the Gavião, including its use, its efficiency, and the whistling techniques used. The relation between some aspects of Gavião instrumental music and the suprasegmental aspects of the language are also dis- cussed and the methods used to study this are described. Audio and video clips illustrate the phenomena being discussed. 1. INTRODUCTION 1. This paper explains methods used in studying tonal phenomena in the language of the Gavião tribe of the state of Rondônia in western Brazil. Five inter- related subjects are discussed: • achieving phonetic accuracy in the field • orthographic representation and its consequences 1 The authors would like to thank the Gavião speakers who participated in the research reported in the present paper, especially Catarino, João, Josias, Roberto, Sena, Séríìt, and Emílio. The support of the indigenous organizations Metareilá of the Surui (Paiteer) and Panderej of the Gavião is greatly appreciated. The authors would like also to thank the institutions which supported this research, either financially, like the Endangered Language Documentation Program, SOAS (project IPF0136), the Conselho Nacional de Desenvolvimento Científico e Tecnológico - CNPq (Pós Doutorado Junior Fellowship n°159747/2010-4), and the Spinoza project Lexicon and Syntax, coordinated by Pieter Muysken of Radboud University, or by granting the necessary permits for research in Brazil (CNPq International, process No EX032/08-CR and FUNAI, authorization 1879/88). Finally, the members of the Linguistic Division of the Museu Goeldi (MPEG/MCTI) are also thanked for their constant support. Licensed under Creative Commons Attribution Non-Commercial No Derivatives License E-ISSN 1934-5275 The Study of tone and related phenomena in an Amazonian tone language 614 • reconstructing tone/length in a protolanguage • the use of tone/length in natural whistled speech • instrumental music and its relation to language tone/length The first three subjects have been developed by the first author since the middle of the 1970s. The last two have been developed in field studies by the second author since 2009. Some observations about linguistic work and the history of tone study in the region are included to give a concrete picture of what occurs in practice. Some results are presented, though these are too extensive to include in detail. Aside from the study of Gavião the application of the methodology for tone investigation to several other Tupian languages is discussed briefly. 2. LANGUAGES. A forthcoming study using phylogenetic methods (Galucio et. al. submit- ted) basically confirms and extends the internal classification of the Tupi language family that has been generally accepted (Rodrigues 1984/5). Of the ten branches of the family, five are spoken primarily in Rondônia: Arikém, Tupari, Puruborá, Ramarama, and Mon- dé. Others are scattered to the north, east, and south: Munduruku, Juruna, Mawé, Aweti, and Tupi-Guarani. The last three of these obviously form a subfamily, which contains no known tonal language. The two languages of the Muduruku branch are tonal (Picanço 2005). Of the two languages of the Juruna branch, Xipaya (ISO: xiy) has been reported (Rodrigues 1995) as without contrastive tone, while Juruna (ISO: jur) is claimed to be tonal by Fargetti (2007). Of the branches in Rondônia, Karitiana (ISO:ktn) the remaining language of the Arikém branch, is said to have non-contrastive pitch accent (Storto 1999, Storto & Demolin 2005). The one remaining language of the Ramarama branch, Karo (ISO:arr), is analyzed as having pitch accent (Gabas 1999). The Tupari branch and the Puruborá branch have not been reported as tonal. We will be mainly concerned with the languages and dialects of the Mondé family, all of which have contrastive tone and length. Of the Aruá (ISO: arx), who have been in contact since the 1930s, only five speak their language. There are two semi-speakers of Salamãy (also called Mondé, ISO: mnd), who have also been in contact since the 1930s. The other dialects and languages, Gavião (ISO:gvo), Zoró (no ISO), Cinta Larga (ISO: cin), and Surui (Paiter, ISO: mdz)) are vigorous. The language of the Surui is not mutually intelligible with the others. The speech of four tribes of Group B, Aruá, Gavião, Zoró, and Cinta Larga, is mutually intelligible, but that of the Salamãy is significantly different from these, though closer to Group B than to Surui. The internal classification of the branch (Moore 2005) is given below (Figure 1). FIgurE 1. Internal classification of the Mondé branch of the Tupi family Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 615 A different internal classification has recently been offered by Anonby (2012) but it suffers from severe problems2 in the data used. At the time of the first author’s original research, 1975-8, the Amazonian languages were relatively unknown. The five branches of the Tupi family in the state of Rondônia in western Brazil had little study. The region was a frontier zone and the Zoró and Cinta Larga tribes were largely still out of contact with national society. The Surui (Paíteer) had come into contact only in 1969 and were in a state of hostility with the Zoró. Of the numer- ous Tupi languages it was thought at that time that only Munduruku was tonal. 3. DETERMINING PHONETIC ACCURACY–INITIAL DIAGNOSIS AND WHISTLING. One of the first stops in the first author’s field research was a visit to the SIL missionar- ies who were working with the language of the Surui. One of them remarked that, oddly, a two-syllable word could be stressed on both syllables. Asked if that was not high tone instead, the missionary replied that, no, some people had had that idea, but an investigation by SIL consultants had concluded that it was not. As field research began among the Gavião one of them remarked one evening that it was hard to read in their language: there were many words that were the same but somewhat different. Fears that the language was tonal were confirmed after three days of elicitation. Four different two-syllable words were found which had the same consonants and vowels. And the linguist, who is devoid of musical talent, could not hear what the differences between the words were. If the contrast were in stress, then one would expect only two contrasts: one word stressed on the first syllable and one stressed on the second syllable. But there were four words, which would be consistent with tone contrasts. After one or two weeks of futile struggle to hear the tone, the linguist was enlightened by his consultant, who dandled his cap and, with a meaningful glance, whistled six syllables. When asked what that was he answered, “My new cap.” Then he whistled it again, clear and pure.3 (The pitch is indicated by a line above the words.) (1) ǫó-dáàt sérék kôro ‘My new cap.’ 1s-head cloth new   2 For example, many of the Gavião forms in the article are obviously in the idiosyncratic orthography of the New Tribes mission. The h they use to indicate length (see below) is imagined by Anonby to be a real voiceless fricative and he postulates that its absence in Zoró and Cinta Larga transcrip- tions is an indicator that this fricative was lost in those languages. Since the missionary transcription of Surui also uses h to mark length this is interpreted as showing a close relation between Surui and Gavião. Oddly enough, ‘shotgun’ is considered by Anonby to be a cognate between Surui and Gavião, though it could not have been an item of Proto-Mondé culture. 3 The symbols c and j denote palatal affricates, y the palatal glide, and s and z dental affricates. The voiced bilabial fricative is indicated by v and the glottal stop by an apostrophe. The tone/length sym- bols are rather complicated and are illustrated in section 4. Nasalization is indicated by the ogonek beneath the vowel. Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 616 Within an hour the basic outline of the tone system seemed to be short and long syllables with height contrasts of high, low, and rising, though there was some question about possible mid tone, as in the word for ‘cloth’ above. Later it was seen that there were rising-falling long tones and also long falling tones. The four contrasting words were clearly combinations of the short tones: (2) magap magáp mâgap mâgáp fat egg 1s+fat 1s+egg One simple test for checking whether speakers were perceiving tone or perceiving stress involved a minimal pair, ‘fat’ and ‘egg’, given above. When the two syllables were pronounced with low tones but with stress on the second syllable, consultants judged the word to be ‘fat’, giving importance to the tone and not the stress. The local New Tribes missionary, who had been among the Gavião for nine years, was initially doubtful that the language was tonal, but then had to agree that it was and adopted whistling to investigate the suprasegmentals. He was the first to observe that high tones were downstepped to mids after certain long tones, but made the error of thinking that this occurred after all long tones and that tones which did not trigger downstep must, perforce, be short (in spite of many counter examples). Elsewhere the SIL missionary with the Cinta Larga had already discovered that their language was tonal. However he had difficulty in his phonological analysis and left Brazil after being removed from the Cinta Larga by Brazilian authorities. As the news of tonality spread, the missionaries with the Surui revised their views and decided that the language was tonal, with a simple inventory of long and short syllables and two height contrasts. More recent studies (Meer 1985, Guerra 2004) basically agree. 4. SUMMARY OF BASIC ANALYSIS OF GAVIÃO TONE/LENGTH. The initial research on the Gavião language was carried out with a very low budget in a remote area of Brazil in the middle of the 1970s, before portable computers. The principal method used for obtaining phonetic accuracy--whistling--was a necessity at that time and place. It succeeded, and, having obtained phonetic accuracy it was possible to accumulate the facts of the Gavião system of tone/length, though a better understanding of the system had to wait for later autosegmental analysis. An analysis of Gavião tone and length and its motivation are presented in Moore 1999 and is too long to be repeated here. But the basic points can be summarized briefly. The summary will also serve to explain the transcription used here, which admitedly is adapted to the Brazilian keyboard and needs revision, in collaboration with the Gavião. The short tones (tones attached to one short vowel) are low, high, and short rising, with the latter (rather marginal) arising from LH sequences. Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 617 (3) V V́ V̂ L H L H       These have long counterparts which are constant, without fluctuations conditioned by what followed. (4) V V̀4 V́ V́ V V́ L H L H       But there are also two problematic long syllables. One of these falls when nothing follows and is high non-finally, causing any following high tones to downstep to mid. The other is a long syllable with a rising-falling tone when nothing follows and a rising tone non-finally, causing any following high tones to downstep to mid. One problem was to account for the final vs. non-final changes. Another was to account for the asymmetry: why in non-final position the tone of the second type of syllable is rising instead of low. The solution offered is that these two syllable types have a floating low tone after them. (5) V́V̀ V̂ V̀ H L L L     If nothing follows, the floating low attaches to the preceding V, resulting in a low end point. If something does follow, the floating low tone attaches to it and does not cause a fall. If what follows the floating low is a high tone, then, by a general rule, the sequence H LH becomes a mid tone, accounting for the downstep. The floating low tone also explains why the asymmetrical rising tone occurs: by a general rule a low tone long syllable rises before a following low tone, in a dissimilation consistent with the Obligatory Contour Principle. Examples, spoken and whistled: 4 The grave accent over the second V distinguishes long low-tone vowels from sequences of two short low-tone vowels. Long low-tone vowels can be distributed in two syllables, for example, volò ‘come’. The grave accent is also used on the second vowel of the alternating tones which produce downstep. The generalization is that it marks syllables whose end point shows alternations condi- tioned by what follows. Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 618 ką́ày ‘burning’ Long Falling ‘burning’, whistled ką̂ày ‘old’ Long Rising-Falling ‘old’, whistled ką́ày mága ‘there is burning’ Long High, Mid, Low ‘there is burning’, whistled ką̂ày mága ‘there is old’ Long Rising, Mid, Low ‘there is old’, whistled There are no simple constraints on tone sequences. For example, in a three-syllable word all combinations can occur: LLL, HHH, HLH, etc. However there are significant canonical patterns related to the morphology. Methodologically, as study of the language progressed, knowledge of the canonical forms helped identify tone and length, since only a restricted selection appears in certain environments. For example, verb stems have level tone, with a long syllable or a reduplicated syllable finally. (6) tígí ‘knock down’ kátaà ‘cut’ są́á ‘sew’ sonò ‘rot’ Adjective stems almost always end in long syllables which trigger downstep. (7) sǫ̂òt ‘rotten’ ką́ày ‘burning’ patíì ‘heavy’ Bisyllabic unpossessed nature words have a lengthened penultimate syllable if the last syllable is long. (8) iitiì ‘deer’ kaasáàl ‘macaw’ vaakôòy ‘vulture’ There are four pronominal clitic classes in Gavião, and the tone of these is correlated with the tone of the first syllable of the stem. A general word-structure contraint is that short tones cannot follow long tones unless the short tones are on affixes. Observed tone changes were noted in a slip file and the accumulated examples were examined periodically. Systematic testing of tone interactions verified hypotheses about tone rules. It is important to achieve accuracy since even a few errors are sufficient to mask a pattern. Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 619 5. WHISTLE METHODS FOR DETAILED STUDY. Two methods were used to determine the tone of each new word encountered. One was to place a standard word with known tones before or after the target word. One useful standard word in Gavião was the demon- strative má ‘other’, which was high and short and could appear before nouns, adjectives, transitive verbs, and auxiliaries. Another useful standard word was the particle terè ‘true’, whose tones were always low when nothing followed and which could occur after nouns, adjectives, and verbs. Since tone is relative this method is useful for determining the num- ber of height levels. In isolation it is difficult to distinguish HL from ML or HM. To see if there are more than two levels it is necessary to produce a number of combinations of three syllables and see how many tones are necessary to account for the patterns seen, control- ling for any possible processes of raising or lowering, which can be detected by observing the same word/morpheme with higher or lower tone in different environments. The second method was to devise a set of standard words to include all known tones, and, in the case of bisyllablic words, all combinations of these tones. Longer standard words can be constructed by combining shorter ones. The consultant was asked to whistle the target word and then the standard word that was suspected of being the same. This method requires a more versatile consultant. It was usually easy to hear if the two whistles were identical or not and often consultants would say that they were or were not, after be- coming accustomed to the task. When transcribing texts each word was individually transcribed in isolation using whistling and then the whole sentence was whistled by the consultant as the linguist looked at the transcription to see if it matched the tone line being whistled. The linguist then whis- tled the whole sentence himself to see if the consultant agreed. Of course, transcriptions were checked from time to time to see if the consultant was consistent from one week to the next. Transcriptions were also checked with other consultants for independent verification. Years later, the whistle-based pitch transcription of recorded words was compared to the pitch in instrumental analysis of the same recorded words. There was strong agreement. The general applicability of whistling as a means of hearing tone accurately depends on the details of how it is carried out. Points that may be helpful are given below. 1. Explanation can make the work more intelligible for the speakers and increase the prestige of their language. For the Gavião, it was explained that Portu- guese had contrasts of stress, for example, sábia, sabia, and sabiá, which were different in intensity, whereas Gavião had something similar but with tone contrasts. It was explained that Chinese, Japanese, and many African languages were tonal, and that tone was natural and widespread, though the European languages usually were not tonal. For that reason the local Jálâày (non-indigenous people) had difficulty in hearing Gavião tone, though it was easy for the Gavião. One had to study linguistics and listen carefully to what the Gavião were telling him to do intelligent work with their tone language. 2. The tone is clearer if the whistling airflow is out and not in, and slow enough to be perceived and imitated readily. 3. If possible, it is better to eventually have consultants whistle tone without articulating consonants. The Gavião consultants were asked to “just whistle Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 620 the tone, not the letters.” Since whistled pitch is determined by tongue height, articulating consonants will alter the tone. In the case of the Cinta Larga the missionary linguist concluded that tone was always rising before palatal consonants, but this was an artifact which disappeared when only the tone was whistled. For some consultants it is unnatural to stop articulating consonants when individual words are being whistled, since the consonants give more clues as to what the words are in whistled speech and individual words lack clues from context. 4. As collaborators, native Amazonian women do not whistle and older men have difficulty in understanding the point of whistling individual words. 5. The whistling generally should not be done in front of others, since most people don’t understand the purpose of whistling to determine tone and find it strange. If the consultant is accustomed to using whistled speech it is highly unnatural for him to whistle to a group in face-to-face contact. 6. A recording of someone speaking words in a tone language and then whistling them is useful as a model for what is needed. Before playing the recording, explain, “This is a really good recording. The speaker did it so well that even a white linguist could hear the tone and began to understand it.” Then whistle along with the recording a bit and say, “I can’t hear the tone well when he speaks, but it is clear to me when he whistles. Let me try a few words in your language. This word X, is it like this (whistle a guess) or like this (whistle other guess)? Hmm, you do it and maybe I can get it. Wow! That’s great. Now I can hear it. Let’s try this other word that has been making me curious.” 6. WHISTLING METHODOLOGY APPLIED TO OTHER LANGUAGES. The analysis of the Gavião tone system was sufficient for the first author’s doctoral dissertation, which focused on syntax. The later autosegmental analysis produced neat rules which accounted for the observed tone changes. Priority in recent years has been documenting and describ- ing the tone and length in the other dialects and languages of the Mondé branch of languag- es. Whistled speech and the relation between Gavião tone and music is being investigated by the second author in collaboration with the first. The whistling methodology was later used with other languages and was taught to stu- dent trainees of the Museu Goeldi. Further experience, with other Amazonian languages, has been positive, though the applicability in other world regions and other tone typologies is another question. The language of the Karitiana, which belongs to the Arikém branch of the Tupi family, turned out to have predictable tone with two levels and long and short syllables. Collabo- rators could easily whistle tone even though it was not contrastive. The Karitiana say that they whistle to each other when hunting. The tonal system of the Munduruku was investigated by Gessiane Picanço while still an undergraduate, using whistling. Within three days she had come to the conclusion that there were only two contrastive levels, with laryngealized vowels always having low tone. The accepted analysis at that time (Braun & Crofts 1965), which had prevailed for thirty Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 621 years, was that Munduruku contained four “accents,” one of which was laryingealization. A generation of Brazilian students had transcribed this imaginary system without noticing that it did not correspond to reality. Later the two-tone analysis was shown to be consistent with instrumental pitch analysis and the phonology was described (Picanço 2005). Whistling in the Makurap (ISO: mpu) language of the Tupari branch of the Tupi fam- ily was initially difficult to obtain. The younger men said that they did not know how to whistle their language. But in conversation about the Gavião whistled speech in the forest one Makurap who was somewhat older stated that his people did that too, and whistled several animal names to illustrate. When this was applauded he produced more words and eventually a long lexical list spoken twice and whistled twice. That data is waiting to be analyzed. Elderly Tupari (ISO: tpr) can still use whistled speech though this does not seem to be in common use any more. Snatches of this have been recorded but not yet analyzed. The Zoró, a tribe who are neighbors of the Gavião and whose speech is highly similar to that of the Gavião, whistled a lexical list with little difficulty and observed that in certain environments the tone changed. Like the Gavião, the Zoró communicate frequently by whistling. Their stereotype of Gavião speech is that it is full of long, looping tones. For reasons that are still not entirely understood, whistling yielded more problematical results among the Surui of Rondônia (Paíteer), where whistled speech by young people is more limited and infrequent. Individuals whistled the same words differently, and the same person changed from one day to the next. With practice and training some younger speakers became consistent. 7. TONE AND ORTHOGRAPHIC QUESTIONS. Given the challenges of analyzing a tone system and the time investment required, it is not surprising that many educators and many linguists discourage speakers of tone languages from indicating the tone in the orthography. The missionary with the Gavião at least tried to write the tone and length, in spite of his limited formal training. Unfortunately, his precipitous analysis created problems for the indigenous community when it was incorporated into the writing system. In this system the long syllables which do not provoke downstep are marked as short and those that do trigger downstep are marked with an h. As a result words that are clearly different are written as the same. For example, the following three words are all written as aka.5 (9) aka ‘kill’ a-aka ‘kill himself’ aa-kaà ‘goes’ kill 3c-kill 3c-go When the Gavião writing practices were surveyed the results were interesting. They showed that many of the speakers were hearing the long syllables which were to be written as short and were trying to indicate them in some impromptu manner. a. About one third wrote the long vowels as short, as taught, all aka. 5 The symbol 3c- denotes a cross-referencing or coreferential third person prefix. Language Documentation & conservation voL. 8, 2014 The Study of tone and related phenomena in an Amazonian tone language 622 b. About one third wrote the long vowels with the h, though these do not provoke downstep: aka, ahka, ahkah. c. About one third wrote the long vowels with two vowels: aka, aaka, aakaa. One obviously brilliant young Gavião teacher perceived the above problems but sought advice from the missionary, who would be reluctant to recognize any difficulties. It will now be a difficult task to make the orthography more adequate, given the mission’s influence and the degree of confusion already installed. The Zoró requested linguistic assistance from the JOCUM (JOvens Com Uma Missão- -Youth With a Mission) mission, but the tone transcription was faulty. Then, following the orientation of a short-term linguistic consultant from Rio de Janeiro, they eliminated tone and length from the writing system, making it defective for documentation, since the pronunciation cannot be recovered from the writing. The writing of the Surui language was sampled in 2011 (Nevins & Moore 2011). Methodologically, this is a useful procedure for initiating orthographic revision. Writing samples, using the same list of words, were taken from various members of the indigenous community and then compared, with the result then shown to the group. When there are inconsistencies in the writing the community is motivated to understand and correct the situation. The samples showed that after thirty years and four linguists the transcription of words was completely mixed, with great variety, which is troubling to the community, who would like to write down their traditional culture before it erodes. The Surui are also under pressure from short-term linguistic consultants to not indicate tone and length. 8. DIACHRONIC TONE IN THE MONDÉ LANGUAGES. In early attempts at reconstructing the tone and length of Proto-Mondé it appeared that reconstruction would be hopeless, given the multitude of correspondences. However, after correcting errors in published data through careful fieldwork, including Zoró data in the correspondences, and investigating some of the synchronic tonal alternation rules of Zoró and Surui, the correspondences became more apparent. As an example, the data below deal with correspondences to the Gavião long tone which rises and falls in final position and rises non-finally, causing downstep in immediately following high tones. This tone is analyzed (Moore 2009) in Gavião as an underlying long low tone followed by a floating low tone. In the following cognates the corresponding tone in Zoró is long low and in Surui it is long high. (10) English Gavião Zoró Surui armadillo maazôòy vązooy valóóy-aa wild turkey vaakôòy vakooy vakóóy-a: rotten sǫ̂òt sǫot i-ʃǫ́ót But in the next set of cognates the same Gavião long tone corresponds to long high tones in Zoró and to long low tones in Surui. Language Documentation & conservation voL. 8, 2014