Nasal Harmony in Tupí-Guaraní: A comparative synthesis

Nasal Harmony in Tupı́-Guaranı́: A comparative synthesis Myriam Lapierre & Lev Michael University of California, Berkeley SSLA III; UMass Amherst; October 20, 2018 1 Introduction 1.1 Overview • Languages of the Tupı́-Guaranı́ (TG) family are well-known for exhibiting two closely related phenomena: – allophonic alternations between fully nasal segments (e.g., [m]), complex nasal segments (e.g., [mb]), and in some cases, voiced oral stops (e.g., [b]) – long-distance nasal harmony (LDNH; a.k.a. unbounded or iterative NH) • Despite considerable empirical similarities among TG language with respect to these phenomena, previous analyses of them are largely inconsistent with one another. • Our goal in this talk is to build on previous work unifying TG nasal segmental alternations (Lapierre & Michael 2017) to present a unified analysis of TG LDNH. • Based on an extensive survey of nasal phenomena in TG languages, we argue that TG LDNH systems fall into a small number of broadly similar types. – All TG LDNH systems exhibit R→L NH, triggered by [+nas] segments – And the types of LDNH systems are distinguished by: 1. Whether LDNH is triggered by: (a) nasal vowels (b) both nasal consonants and nasal vowels 2. Whether /m/ exhibits: (a) no oralized allophones (b) only a partially oralized allphone ([mb]) (c) fully and partially oralized allophones ([b, mb]) • As we shall show, the relatively simple typology and analysis sketched out above has previously been obscured by: – Divergent analyses of empirically quite similar allophonic alternations of nasal stops – Divergent analyses of empirically quite similar nasal harmony patterns ∗ e.g., Whether LDNH is analyzed as the result of ‘morpheme-level nasality’ or segmental triggers; ∗ . . . and in the latter case, which segments are considered triggers of LDNH 1 1.2 The Tupı́-Guaranı́ family • The Tupı́-Guaranı́ (TG) family is a well-established subgroup of the the larger Tupian stock, depicted in Figure 1 Figure 1: Internal classification of Tupian family (Galucio et al. 2015) • TG is the largest Tupian subgroup, with over 40 commonly recognized varieties, with a recent phylogenetic classification given in Figure 2. Figure 2: Internal classification of Tupı́-Guaranı́ (Mawé and Awetı́ are outgroup languages; Michael et al. 2014) 2 1.3 Roadmap §2 Empirical facts about LDNH in TG languages §3 Our proposal §4 Previous analyses of LDNH in TG languages §5 Additional support for our analysis §6 Conclusion 2 Long-distance Nasal Harmony in Tupı́-Guaranı́ 2.1 Empirical generalizations • The domain of LDNH in TG languages is the root and its prefixes12 • NH domains exhibit the following three patterns: 1. they are entirely nasal3 (1, 4) 2. they are entirely oral (2, 5) 3. they are split into exactly one nasal span (NS) and one oral span (OS) (3, 6) – When split, the nasal span always occurs to left of the oral span4 Nhandewa Guaranı́ (Costa 2003, 2007) (1) Entirely nasal NHD a. [tũpã] thunder b. [mãmõ] where? c. [mãPẽ] to look d. [ñãnẽ+Rãmõ+j̃] our grandfather (3) Split nasal-oral NHD a. [kũmãnda] bean b. [tSĩ+mẽmb1] my son c. [nẽ+mbaPe] your things d. [mãnduPi] peanut (2) Entirely oral NHD a. [tupa] bed b. [puRuPa] pregnancy c. [tuwitSa] chief d. [pORORO] popcorn 1 But note that not all prefixes form part of the NH domain in all languages. Note that many authors also discuss nasal harmony within the domain of suffixes; however, suffixes do not undergo nasalization in the same way as prefixes and roots. We assume that the nasalization (or lack thereof) of suffixes in TG languages is dependent on a different mechanism than the nasalization of roots and prefixes, likely one dependent on lexically-specified morphophonological properties of the suffixes themselves. For this reason, we leave the question of suffix nasalization to future work. 3 We refer to a NHD or a span within an NHD as ‘nasal’ if all the phonemes in the NHD or span surface with their nasal allophones, and correspondingly, as ‘oral’ if they all surface with their oral allophones. 4 With the exception of the occurrence of blocking segments, such as voiceless obstruents in Type B languages. 2 3 Kaiwá (Cardoso 2009) (6) Split nasal-oral NHD a. [tũkũmbo] rope b. [mãnd1dZu] cotton c. [tũNgusu] flea d. [ĩt1̃NgwaRa] nostril (4) Entirely nasal NHD a. [mĩSĩ] small b. [kũnũmĩ] boy c. [kũñã] woman, female d. [hĩw̃ãPĩ] close, there (5) Entirely oral NHD a. [surubi] tiger catfish b. [seRadu+pe] in the Cerrado c. [adZaka] basket d. [hagwe] hair • In Table 1 we summarize the empirical properties of 18 TG languages, identifying which classes of segments: Column 1: Can appear rightmost in a nasal span (triggers) Column 2: Show alternation between oral and nasal allophones (targets) Column 3: Are observed with a NS to their right and an OS to their left (opaque segments) Column 4: Do not show alternation between oral and nasal allophones, but are observed within a NS (transparent segments) • The first basic distinction is between languages that exhibit nasal harmony (Types A&B), and those that do not (Type C) – In all Type A&B languages, either [mb] or [Ṽ] can be rightmost in a nasal span ∗ This indicates that, in these languages, LDNH is triggered by any [+nas] feature, whether associated with a consonant or vowel. • A.1, A.2, and B languages are distinguished by two characteristics of the LDNH system: 1. Their nasal and voiced stop inventories: – A.1: [m, mb] – A.2: [m, mb, b] – B: [m] 2. The behaviour of voiceless obstruents: – In Type A languages, voiceless obstruents are transparent. – In Type B languages, voiceless obstruents are opaque. 5 There are two dialects of Kawaiweté, one of which presents NH, and the other of which does not. Given the absence of any available data on the dialect of Kawaiweté with nasal harmony, the language has been classified as Type C. We suspect that the dialect of Kawaiweté with NH falls within Type B. 4 A.1 A.1 A.1 A.1 A.2 A.2 A.2 A.2 A.2 A.2 B B C.1 C.1 C.1 C.1 C.2 C.2 Par. Guaranı́ (G & S 1967) Nhandewa (Costa 2003, 2007) Guarayú (Armoye 2009) Tapieté (González 2005) Achê (RöBler 2008) Kaiwá (Cardoso 2009) Mbyá (Thomas 2014) Emérillon (Rose 2008) E.B. Guaranı́ (Daviet 2016) Yuki (Villafañe 2004) Kamayurá (Seki 2000) Guajá (Nascimento 2007) Tapirapé (Leite 1977) Pauserna (Ramirez at al. 2018) Kawaiweté5 (Souza 2004) Araweté (Alves 2008) Tembé (O & S forthcoming) Parakanã (B & P 2017) R-most[+nas] trigger mb, Ṽ [-nas] [+nas] opaque Ø [+nas] [+nas] transparent p, P Ø p, P Ø p, P p, P (L→R) p p, P (R→L) P Ø p, P Ø p, P Ø p, P Ø p, P Ø p, P m, Ṽ [±nas] alternants targets {V, Ṽ}, {w, w̃}, {mb, m} {V, Ṽ}, {w, w̃}, {mb, m} {V, Ṽ}, {w, w̃}, {mb, m} {V, Ṽ}, {w, w̃}, {mb, m} {V, Ṽ}, {w, w̃}, {mb, m, b} {V, Ṽ}, {w, w̃}, {mb, m, b} {V, Ṽ}, {w, w̃}, {mb, m, b} {V, Ṽ}, {w, w̃}, {mb, m, b} {V, Ṽ}, {w, w̃}, {mb, m, b} {V, Ṽ}, {w, w̃}, {mb, m, b} {V, Ṽ}, {w, w̃} p P m, Ṽ {V, Ṽ}, {w, w̃} p P NA Ø NA NA NA Ø NA NA NA Ø NA NA NA Ø NA NA NA {m, mb, bm, bmb} NA NA NA {m, bm} NA NA mb, Ṽ mb, Ṽ mb, Ṽ mb, Ṽ mb, Ṽ mb, Ṽ mb, Ṽ mb, Ṽ mb, Ṽ Table 1: TG LDNH patterns 5 3 Proposal • We argue that LDNH systems across TG languages can be classified according to: 1. Whether or not a language exhibits LDNH – For languages that do exhibit LDNH, the set of possible triggers of LDNH 2. The set of surface allophones of underlying nasal stops /m/ 3.1 LDNH parameters • Languages can be classified according to whether or not they exhibit long-distance6 spreading of nasality, which affects all sonorants to the left of the trigger within a phonological word: 1. +NH languages (Type A, B) 2. -NH languages (Type C) • Languages can be categorized according to the set of possible triggers of LDNH: 1. In Type A languages, both nasal consonants /m/ and nasal vowels /Ṽ/ are triggers 2. In Type B languages, only nasal consonants /m/ are triggers 3.2 Nasal consonant allophony • Following Lapierre & Michael (2017), we consider that allophonic alternations between nasal, prenasalized, and fully oral stops [m]∼[mb]∼[b] are best accounted for by positing underlying nasal stops /m/ • /m/ exhibits (partially-)oralized allophones as the result of shielding (Herbert 1986, Stanton 2017) from adjacent oral vowels: (7) a. /m/ → [m] / Ṽ b. /m/ → [mb] / Ṽ V c. /m/ → [b] / V V (All TG languages) (Type A.1, A.2, C.2) (Type A.2) • The following data from Kaiwá (Cardoso 2009; Type A.2), illustrate the allophony: (8) a. /nãmi/ [nãmbi] ‘ear’ b. /oNa/ [oga] ‘house’ 6 Long-distance is here defined as potentially affecting segments beyond those that are immediately adjacent to the trigger. 6 • Languages can be classified according to whether or not they exhibit local7 oralization (LO) of this sort: 3.3 1. +LO languages (Type A.1, A.2, C.2) 2. -LO languages (Type B, C.1) Parameterization of the system types • We argue that the different types of LDNH systems in TG can be accounted for by positing three basic parameters: 1. ±LO: Whether languages exhibit local oralization of nasal consonants /m/ by adjacent oral vowels (i.e., shielding) 2. ±NH: Whether languages exhibit leftward LDNH 3. The order of application of the LO and NH processes • In light of these parameters, the attested types can be classified as follows: 1. Type A: +LO, +NH LDNH triggers = /m, Ṽ/ (a) Type A.1: NH>LO (b) Type A.2: LO>NH /m/ = [m, mb] /m/ = [m, mb, b] 2. Type B: -LO, +NH LDNH triggers = /m, Ṽ/ /m/ = [m] 3. Type C: -NH LDNH triggers = Ø (i.e., no LDNH) (a) Type C.1: -LO, -NH (b) Type C.2: +LO, -NH /m/ = [m] /m/ = [m, mb, bm, bmb] • Ordering of application of LO and NH in Type A languages (+LO, +NH) distinguish between types A.1 and A.2 – In type A.1 languages NH operates before LO ∗ For /Vm/ squences, /V/ always nasalizes: /VmV/ → ṼmV → [ṼmbV] – In Type A.2 languages, LO operates before NH ∗ This allows the oralization rule to bleed the NH rule: /VmV/ → [VbV] • We argue that this analysis is preferable to previous analyses of LDNH in TG – It accounts for the full range of empirical facts – It is well-supported by typological, phonological, and phonetic facts • In the next section, we summarize previous analyses of the facts presented in §2 7 Local is here defined as affecting only immediately adjacent segments. 7 4 Previous analyses of LDNH in TG languages 4.1 Morpheme triggering analyses • Several proposals have been made for a morpheme-triggering analysis of TG LDNH • These either posit underlyingly nasal root morphemes (e.g., Lunt 1973), or a lexicallyencoded nasal melody (e.g., Goldsmith 1976, Piggott 1992) • For reasons of time, we restrict ourselves here to Piggott’s (1992) analysis • According to this analysis, Paraguayan Guaranı́ has a rule that spreads the feature [nasal] leftward, until it encounters either (i) another [nasal] feature, or (ii) a word edge. • Some roots are lexically marked with a floating [nasal] feature – The floating feature maps to the rightmost segment in a word8 , as in Figure 3. Figure 3: Autosegmental derivation of nasal harmony in Paraguayan Guaranı́ ‘tũpã’ god • However, given the presence of morphemes with both nasal and oral spans in Guaranı́, the author must posit the existence of a different [nasal] feature – This [nasal] feature is segmentally specified for nasal stops, as in Figure 4. Figure 4: Autosegmental derivation of nasal harmony in Paraguayan Guaranı́ ‘kũmãnda’ bean 8 The floating [+N] feature maps to the rightmost segment with a Spontaneous Voicing (SV; roughly equivalent to sonorant) node. Because voiceless obstruents are not sonorants, they do not undergo nasalization, as they do not have an SV node on which the [+N] feature can dock. 8 • [mb] segments are derived by positing a post-oralization rule for nasal consonants • Piggott formalizes this as a Voice Fusion rule: SV-nodes are fused within a syllable; the features of the right node (i.e. the nucleus or head of the syllable) dominate • This accounts for the observation that [mb] segments are always at the boundary between a nasal and an oral span in split morphemes, as in Figure 5 Figure 5: Autosegmental derivation of post-oralization in Paraguayan Guaranı́ ‘kũmãnda’ bean • The need to posit two separate processes of LDNH in the same language seems like an unnecessary analytical complication – As shown in Section 2, all of the data can be accounted for with a single generalization: ∗ The rightmost nasal segment spreads its nasality leftward. • Though it requires some complications, the Autosegmental treatment of NH can account for most of the empirical facts in Type A.1 languages9 • However, the occurrence of voiced oral stops [b] in Type A.2 languages cannot be accounted for with a morpheme-level analysis – [b] and [mb] are in complementary distribution in Type A.2 languages – [b] is an anti-triggering segment: ∗ /m/ is a phonemically nasal stop and is thus expected to trigger NH ∗ /m/ loses its [nasal] feature when it occurs between two phonemically oral vowels (/m/ → [b] / V V) ∗ Thus [b] is no longer a trigger, which bleeds the application of LDNH 9 But note that no morpheme-level analysis has succeeded in accounting for the full range of variation in suffixes 9 • There is no way to account for this bleeding process in Type A.2 languages – Voiced oral stops [b] and pre-nasalized stops [mb] are represented with the same features ∗ This is because they are considered to be in free variation, where voiced stops pre-nasalize to facilitate voicing ∗ However, this is not true in Type A.2 languages ∗ Thus, the morpheme-level analysis cannot account for the complementary distribution of [mb, b] – There is further no way to represent rightward preoralization, from an oral vowel to a following nasal consonant ∗ Even if such a stipulation were implemented, there would be no way to prevent the delinked (i.e. floating) [nasal] feature from attaching to a potential host to its left 4.2 Segment triggering analyses • Thomas’ (2014) analysis of Mbyá (Type A.2) exemplifies a segment-triggering analysis with postnasalization • Prenasalized stops /mb/ are triggers of harmony • Voiceless obstruents are transparent • Voiced stops and sonorants are targets • Harmony spreads leftward across an entire root and all of its prefixes (9) a. b. c. d. /tSe+mbaRaka/ → [tSẽmbaRaka] /mbandiPo/ → [mãndiPo] /tataend1/ → [tãtãẽnd1] /dZaRõ/ → [ñãR̃õ] my guitar yuka lantern angry • In Thomas’ analysis, nasal harmony is modeled as a (bidirectional) constraint that forces adjacent syllable nuclei to agree in nasality – Directionality of spreading is constrained from R→L by giving preference to the [+/-nas] value of a stressed vowel over that of unstressed vowels ∗ Root-final vowels always bear stress – /mb/ is post-nasalized when it occurs before a nasal vowel (/mb/ → [m] / 10 Ṽ) • The segment triggering analysis presented here is superior to the morpheme triggering analysis in that it can account for the full range of LDNH fact without having to posit two separate LDNH processes • However, given that such a system does not include a LO rule of the type in (7), it cannot account for the occurrence of fully oral voiced stops [b] in Type A.2 languages – Crucially, it does not derive the observation that fully voiced stops [b] are in complementary distribution with fully nasal and pre-nasalized stops [m, mb] 4.3 Interim summary • We have summarized the major types of analyses proposed to account for the empirical facts of nasality in TG languages • We argued that our proposed analysis is preferable to previous analyses – It accounts for the full range of empirical facts – It is well-supported by typological, phonological, and phonetic facts • In the following section, we provide additional support for the bleeding analysis based on an additional broad typological generalization 5 Additional support for the bleeding analysis • The bleeding of the NH rule in Type A.2 is crucially dependent on the pre-oralization of /m/ (10a) • Post-oralization of /m/ is insufficient on its own to bleed the NH process (10b) – This is a case of doubly-triggered anti-harmony 10 (10) In a Type A.2 language (e.g., Kaiwá) a. /VpVmV/ → VpVbV → [VpVbV] complete oralization b. /VpṼmV/ → VpṼmbV → [ṼpṼmbV] post-oralization • Perhaps the reason why preoralization inhibits the leftward spreading of the [nasal] feature is because it operates in the opposite direction – Both LDNH and post-oralization are processes that operate R→L – Pre-oralization, on the other hand, operates L→R • There appears to be a more pervasive typological link between the presence of pre-oralization [bm] in a language and the absence of LDNH 10 See Lionnet 2017 for a case of doubly-triggered rounding harmony. 11 • In Table 1, Type C languages (those without LDNH) fall into two types: – Type C.1 languages show no allophonic alternations for nasal stops (/m/ → [m]) – Type C.2 languages show allophonic alternation between fully nasal and partially oralized nasal consonants [m, mb] (as do Type A languages) ∗ Type C.2, unlike Type A, present preoralized nasal stops [bm] • It appears that the presence of preoralized nasal stops is linked to the loss of LDNH as a productive phonological process – No language with LDNH in Table 1 shows pre-oralized allophones – All of the languages with pre-oralized allophones in Table 1 lack LDNH – The innovation of pre-oralization may be a diachronic pathway to the loss of LDNH • We take the presence of Type C.2 languages as a further argument in favour of the proposal to order LO and NH with respect to one another – In Type A.2 languages, NH fails to occur in precisely those cases where pre-oralization takes place 6 Conclusion • TG languages exhibit typologically very similar processes of LDNH – Based on our typological survey, TG languages exhibit one of three kinds of LDNH systems • Previous analyses cannot account for the full range of similarities and differences in NH harmony systems across the TG language family – Our analysis can account for all of the typological facts • We argue that an analysis proposing segmental triggers of LDNH is preferrable to one positing morpheme-level nasality • We further claim that allophonic variation among oral and nasal stops [m, mb, b] is best accounted for by assuming underlying nasal stops /m/ • Finally, we argue that the full range of variation among LDNH systems in the family can be derived according to the following parameters: 1. Whether or not a language exhibits LO (±LO) 2. Whether or not a language exhibits LDNH (±NH) 3. In +LO, +NH languages, how LO and NH are ordered with respect to one another 12 Acknowledgements We would like to thank the audience of the 8th Conference on Indigenous Languages of Latin America (CILLA VIII), held at the University of Texas at Austin in October 2017, for comments and discussion on a related talk. We would also like to thank Florian Lionnet, Nicholas Rolle, Darya Kavitskaya, Larry Hyman, Sharon Inkelas, and the members of the UC Berkeley Phonetics and Phonology Forum for stimulating discussions on the data and analysis presented in this paper. All remaining errors are our own. References Armoye Urarepia, Celso. 2009. Análisis de la lengua guarayo (tesina). Ms. Baraúna, Fabı́ola and Gessiane Picanço. 2017. Tendências na implementação fonética de consoantes nasais em Tupı́-Guaranı́. LIAMES 17(1): 143-157. Cardoso, Valéria Faria. 2009. Sistematização da Fonologia Kaiowá: Nasalização e/ou Oralização. Sı́nteses(14), pp. 31-72. Costa, Consuelo de Paiva Godinho. 2007. Apyngwa rupigwa: nasalização em NhandewaGuarani. PhD dissertation, Universidade Estadual de Campinas. Cunha, Péricles. 1987. Análise fonêmica preliminar da lı́ngua Guajá. Master’s thesis, Universidade de Campinas, Brazil. Daviet, Windy. 2016. Observations sociolinguistiques et analyse de la phonologie du dialecte ava du guaranı́ bolivien. Master 2 Thesis, Université Lumière Lyon 2, France. Dooley, Robert. 1984. Nasalização na lı́ngua guarani. Estudos sobre lı́nguas Tupi do Brasil. Série Linguı́stica 11: 7-35. González, Hebe. 2005. A grammar of Tapieté. PhD dissertation, University of Pittsburgh. 43-54. Galucio, Ana Vilacy, Sérgio Meira, Joshua Birchall, Denny Moore, and Nilson Gabas Júnior. 2015. Genealogical relations and lexical distances within the Tupian linguistic family. Boletim do Museu Paraense Emı́lio Goeldi(10)-2, 229-274. Gregores, Emma and Suárez, Jorge. 1967. A Description of Colloquial Guaranı́. The Hague: Mouton & Co. Herbert, Robert. 1986. Language universals, markedness theory, and natural phonetic processes. New York: Mouton de Gruyter. Lapierre, Myriam, and Lev Michael. 2017. Nasal segments in Tupı́-Guaranı́: A comparative synthesis. Talk presented at the 8th Conference on Indigenous Languages of Latin America (CILLA VIII), University of Texas at Austin. Lionnet, Florian. 2017. A theory of subfeatural representations: the case of rounding harmony in Laal. In Phonology 34.3, pp. 523–564. Leite, Yonne de Freitas. 1977. Aspectos da fonologia e morfofonologia tapirapé. Master’s thesis, Universidade Federal do Rio de Janeiro, Brazil. Michael, Lev, Natalia Chousou-Polydouri, Keith Bartolomei, Erin Donnelly, Sérgio Meira, Vivian Wauters, and Zachary O’Hagan. 2015. A Bayesian Phylogenetic Classification of Tupı́Guaranı́. LIAMES. Lı́nguas Indı́genas Americanas 15(2): 193-221. Nascimento, Ana Paula Lion Mamede. 2008. Estudo Fonético e Fonológico da Lı́ngua Guajá. MA thesis, Universidade de Brası́lia. 13 Orjuela, Lorena, and Sérgio Meira. ms. Tembé (Tenetehara). Padua, Alexandre Jorge. 2007. Contribuição para a fonologia da lı́ngua apiaká (tupı́guaranı́). MA thesis, Universidade de Brası́lia. Pereira, A. A. 2009. Estudo Morfossintático do asurinı́ do Xingú. PhD dissertation, Universidade Estadual de Campinas, Brazil. Rose, Françoise. 2011. Grammaire de l’Émérillon teko, une langue tupi-guarani de Guyane Française. Langues et Sociétés d’Amérique traditionnelle(10). Leuven: Peeters. Rose, Sharon & Walker, Rachel. 2011. Harmony Systems. In J. Goldsmith, J. Riggle, & A. C. L. Yu (eds.), The Handbook of Phonological Theory, Second Edition. Berlin: Blackwell Publishing Ltd. Rößler, Eva-Maria. 2008. Aspectos da gramática achê: descrição e reflexão sobre uma hipótese de contato. MA thesis, Universidade Estadual de Campinas. Thomas, Guillaume. 2014. A split analysis of nasal harmony in Mbyá. Revista Lingüı́stica (10)-2, pp. 75-104. Seki, Lucy. 2000. Gramática do kamaiurá. Editora da Universidade Estadual de Campinas. Souza, Patrı́cia de Oliveira Borges. 2004. Estudos de aspectos da lı́gua kaiabi (tupı́). MA thesis, Universidade Estadual de Campinas. Solano, Eliete de Jesus Bararúa. 2009. Descrição gramatical da lı́ngua araweté. Master’s thesis, Universidade de Brası́lia, Brazil. Stanton, Juliet. 2017. Constraints on the Distribution of Nasal-Stop Sequences: An Argument for Contrast. Cambridge, Massachusetts: MIT Dissertation. Vasconcelos, Eduardo Alves. 2008. Aspectos fonológicos da lı́ngua xetá. MA thesis, Universidade de Brası́lia. Villafañe, Lucrecia. 2004. Gramática yukı́: Lengua tupı́-guaranı́ de Bolivia. Tucamán, Argentina: Ediciones del Rectorado Universidad Nacional de Tucamán. 13-22. Appendix 1: More segment triggering analyses Post-oralizing analyses • Cardoso’s (2009) analysis of Kaiwá (Type A.2) exemplifies a postoralization analysis with segmental triggers • Stressed and unstressed nasal vowels /Ṽ/ are harmony triggers – Voiceless obstruents are transparent – Sonorants (nasal stops, glides, and vowels) are targets • LDNH spreads leftward across an entire root and all of its prefixes • In this analysis, LDNH is modeled as a rule of a leftward spreading [+nas] feature – Oralization of nasal stops is modeled as a bidirectional rule of [-nas] spreading, triggered by [-nas] vowels 14 – Oralization is crucially ordered before LDNH in order to account for the bleeding effect of fully voiced oral stops [b] • Nasal stops /m/ are pre- and postoralized when they occur adjacent to an oral vowel (11) (11) a. /m/ → [mb] / Ṽ V b. /m/ → [b] / V V (12) a. b. c. d. /surumi/ → [surubi] /tũNusu/ → [tũNgusu] /nãmi/ → [nãmbi] /SiRĩnõ/ → [SĩRĩnõ] catfish flea ear hummingbird Spontaneous pre-nasalization analyses • Spontaneous prenasalization analyses of NH include Daviet’s (2016) analysis of Eastern Bolivian Guaranı́ and Rose’s (2008) analysis of Emérillon (both Type A.2 languages) • SPN analyses assume that surface nasal consonants [m, mb] are allophones of underlying voiced oral obstruents /b/, as in (13) (13) a. /b/ → [mb] / V V11 b. /b/ → [m] / Ṽ c. /b/ → [mb ∼ b] / # V • Under this analysis, there are two kinds of triggers of LDNH: 1. [mb], which arise through rule (13a) 2. root-final nasal stops [m], which arise through rule (14)12 (14) [-cont] → [+nas] / # (15) /tab1dZ/ → [tãm1̃ñ] ‘grandfather’ • This analysis relies on a crucial feeding order of the above-presented rules – Since the segmental inventory does not include any phonemically nasal stops, rules (13) and (14) must apply first – The surface nasal stops [mb, m] then trigger leftward LDNH 11 Rose (2008:435) observes that this rule operates optionally morpheme-initially, i.e. when preceded by a heteromorphemic vowel. 12 Rose (2008) adds a rule (14) nasalizing voiced stem-final non-continuants, exemplified in (15), to account for the existence stem-final nasal consonants, since the rules given in (13) do not account for such segments. 15 (16) a. /sibo/ → [sĩmbo] b. /tudaga/ → [tũnãNga] rope (Emérillon) father-in-law (East. Bol. Guaranı́) • In Lapierre & Michael (2017), we argued that the spontaneous nasalization mechanism is phonologically implausible – Prenasalization of voiced stops between two oral vowels (/b/ → [mb] / V not phonetically motivated13 13 V) is While pre-nasalization of voiced stops sometimes occurs to facilitate voicing (in these cases, termed venting), this is not, in fact, the motivation for the occurrence of [mb] segments in TG, and in particular not when they occur at the boundary between a NS and an OS. 16