Postlabial raising and paradigmatic leveling in A’ingae: A diachronic study from the field Maksymilian Dąbkowski *

. This paper discusses and analyzes the variation between ai and ɨi in A’ingae (or Cofán, an Amazonian isolate, ISO 639-3: con ) by comparing the data reported in Borman’s (1976) dictionary with contemporary productions. In Borman (1976), ai does not generally appear after labial consonants; the distribution of ɨi is not restricted. In some modern productions, postlabial ai is allowed when the diphthong crosses a morpheme boundary ( a + i ). I propose that Borman’s (1976) distribution of ai and ɨi is a consequence of a diachronic change of ai to ɨi after labial consonants (* ai > ɨi / B _). The contemporary distribution reflects paradigm leveling and contact-induced replacement: Borman’s (1976) ɨi corresponds to contemporary ai if a is present in another related form. In novel productively-formed words, the availability of postlabial raising is speaker-specific. The proposed sound change of postlabial raising (* ai > ɨi / B _) is unusual and lacks obvious phonetic motivation. I speculate that postlabial raising reflects postlabial rounding (* ai > * ui / B _) opacified by subsequent unconditioned unrounding and centralizing of the back round vowel (* u > ɨ ).

postlabial raising reflects phonetically natural postlabial rounding (*ai > *ui / B _) opacified by subsequent unconditioned unrounding and centralizing of the back round vowel (*u > ɨ).Thus, I present a study that combines internal reconstruction with a comparison between a relatively recent language description and contemporary fieldwork data to understand the trajectory of language change and make sense of variation in a fieldwork context.
The rest of the paper is structured as followed.Section 2 gives background on the language and its speakers.Section 3 describes and analyzes the diachronic relationship between the diphthongs ai and ɨi.Section 4 concludes.
2. Language background.A'ingae (or Cofán, ISO 639-3: con) is an endangered and highly underdocumented Amazonian language isolate spoken by ca.1,500 Cofán people in the province of Sucumbíos, Ecuador and the department of Putumayo, Colombia.In recent history, the Cofán have experienced severe economic, ecological, and political pressures.Notwithstanding, the Cofán attitudes towards their language and heritage remain uniformly positive (Dąbkowski 2021a).
The history of the Cofán can be traced back to the Eastern Andean Cordilleras, where they used to live around the 16th century.They have since then descended into the Amazon Basin (Lucitante 2019).The typological profile of A'ingae reflects this history of the Cofán migration, as the language both retains typically Andean features and shows Amazonian innovations (AnderBois et al. 2019).A'ingae is robustly spoken in most Cofán communities, especially in Ecuador.There is limited bilingualism with Kichwa and robust bilingualism with Spanish (Dąbkowski 2021a).
Previous work on the phonetics and phonology of A'ingae includes Borman's (1962) phonological sketch, Repetti Ludlow et al.'s (2019) phonetic study, Fischer & Hengeveld's (in press) grammatical sketch, Sanker & AnderBois's (2021) internal reconstruction of nasality, and Dąbkowski's (2021bDąbkowski's ( , 2023) ) phonological analyses of metrical stress and the glottal stop.A short dictionary compiled by Borman (1976) will serve as a basis for comparison between A'ingae from 50∼70 years ago and contemporary A'ingae.The phonemic inventory of A'ingae (given in Table 1) is moderately large, totaling twentyseven consonants and five vowels, which can form eleven diphthongs.A'ingae vowels can be either oral or nasal; diphthongs are either fully oral or fully nasal.Diphthongs are generally rare in the language,1 which means that many consonant-diphthong sequences are unattested.However, I will argue that some of these gaps are not accidental.
All the contemporary data were collected by the author remotely in the spring of 2022 and reflect the judgments of three native speakers from the province of Sucumbíos, Ecuador.

Description and analysis.
In this section, I compare and analyze the realizations of words that contain ai or ɨi as reported by Borman (1976) and as produced by three contemporary native speakers.First, I discuss the distribution of the two diphthongs in native roots.Second, I address the adaptations of borrowings.Lastly, I look at morphological complex forms.Borman (1976;henceforth B76) is the most complete A'ingae dictionary to date.It was authored by Marlytte "Bub" Borman, a missionary SIL linguist.He and his wife Roberta "Bob" Borman worked in the Cofán communities since 1954 (Hugo Lucitante, p.c.).Thus, any systematic differences between the Borman (1976) and contemporary A'ingae provide evidence for language change in the past 50∼70 years.The Bormans worked predominantly in the community of Dureno, Sucumbíos (Hugo Lucitante, p.c.).Thus, I assume that the data reported by Borman (1976) reflect the Dureno language variety.
The productions I report come from three native speakers which I identify as JXM, RGQ, and SIA.All three speakers are male.JXM and SIA both come from the Ecuadorian community of Dureno (where the Bormans used to work), which controls for dialectal variation.JXM is 36 y.o. and SIA is 23.RGQ comes from the Ecuadorian community of Dovuno, Sucumbíos and is 34 y.o.
First, I consider the distribution of ɨi and ai in native A'ingae roots.The distribution of the high fronting diphthong ɨi in native roots is not restricted with respect to the major place of articulation. 2This is to say, the diphthong can appear after velars (1a-b), coronals (1c-g), and labials (1h-j).In the data sets below, the first column gives the root as reported by Borman (1976) 3 along with its meaning.The other three columns give the realizations produced by contemporary speakers.
To obtain the contemporary judgments, I asked each consultant whether each word can be realized with ai or with ɨi.For example, for 'catfish' reported by Borman (1976) as kʰɨiʋo (1a), I asked if the word can be realized as kʰɨiʋo and kʰaiʋo.For each realization identified as correct, I requested that it be repeated out loud.If both realizations were identified as correct, I asked if they preferred one realization over the other.If only one realization was identified as correct, I asked if other native speakers could use the incorrect pronunciation.For no items in (1) were any differences found between Borman's (1976) and the contemporary productions.
( To account for this distributional gap, I propose that the diphthong ai underwent raising to ɨi after labial consonants (3), resulting in a conditioned merger of ai and ɨi.The capital letter B abbreviates {f, pʰ, p, ᵐb, m, ʋ}.(Although I refer to postlabial raising here as a "sound change", I will propose below that it represents a telescoping of two changes: *ai > *ui / B _ and *u > ɨ.) postlabial raising (4).In the data set below, the first column gives the donor language form and the name of the language (parenthesized).The following notation is used in reporting the consultants' judgments.No superscript indicates a given speaker's only or preferred realization.A superscripted at sign ( @ ) indicates that the speaker identified the realization as incorrect, dispreferred, or absent from their idiolect, but recognized that other native speakers could use it.An asterisk ( * ) indicates that the speaker identified the realization as archaic.A superscripted question mark ( ? ) indicates that the speaker was inconsistent (they provided different judgments on different occasions).Realizations rejected as categorically wrong and identified as nonexistent within the entire speech community are not given.Items absents from a consultant's idiolect are represented with an em dash ( -).Contemporary judgments which differ from realizations reported in Borman (1962) are additionally marked with a ::::::::::::::: wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline wavy underline.
All three speakers recognize both rosaʋɨita and rosaʋaita.JXM prefers rosaʋaita and SIA prefers rosaʋɨita.RGQ vacillates between the two realizations.I propose that rosaʋɨita is the older realization and rosaʋaita shows a recent replacement of ʋɨita with ʋaita motivated by greater phonological similarity to the source language.This is plausible given that the compound is relatively morphologically transparent and many Cofán people are moderately bilingual with Kichwa (Dąbkowski 2021a).Remarkably, RGQ explicitly notes that "the elders would always say rosaʋɨita," lending further credibility to this scenario.
According to Borman (1976), the Spanish paitʃe 'paiche (a fish species)' is borrowed as both pɨitsɨ and paitsɨ.The former shows postlabial raising; the latter is more faithful to the source language.All three consultants identify paitsɨ as the correct form; none recognize pɨitsɨ.I propose that paitsɨ won over pɨitsɨ due to a pressure to reflect the pronunciation of the source language more accurately.This scenario, again, is feasible because of a high degree of bilingualism with Spanish among the A'ingae speakers (Dąbkowski 2021a).Now, I consider the realizations of /Ba+i/ at morpheme boundaries to see whether postlabial raising is only a historical change or if it has been learned as an active phonological rule.To do so, I investigate the Ba-final roots followed by i-initial suffixes.
There are two relevant i-initial suffixes: the periodic -ite PRD and the instrumental -iʔkʰɨ INS. 8 The periodic -ite PRD appears in the conventional names of seasons of the year, but it can also be used to productively derive new periods of time.The traditional season names which can be found in Borman (1976) are given in (5).The first column gives the root and its meaning.The second column gives the season name and its time span as reported by Borman (1976).(sĩ)mɨ ĩta -c.waita (Kichwa) (ɾosa)ʋɨita 'marigold' @ (ɾosa)ʋɨita, * ?(ɾosa)ʋɨita, (ɾosa) When the root ends in a, but not Ba, the environment for postlabial raising is not satisfied, so the season name (unsurprisingly) shows no postlabial raising (5a).When the root ends in Ba, Borman (1976) reports postlabial raising for all the season names (5b-f).The three contemporary consultants report forms with as well as without postlabial raising.They show variation in which forms they accept and/or prefer, but most of them recognize both forms of each season name as available for at least some speakers.
I propose that the forms with postlabial raising (ɨi) are older, whereas the forms without postlabial raising (ai) are analogical innovations (na : naite :: tʃaɾapa : tʃaɾapaite).Note that the season names are conventional and not entirely predictable from the meaning of the root.Nevertheless, paradigm leveling can take place because native speakers are aware of the morphological relationship between the root and the season name.
Both JXM and RGQ generally prefer forms with ai, even though they recognize that forms with ɨi are present in the speech community.Remarkably, RGQ explicitly identifies sãfɨ ĩte, tʃaɾapɨite, and omɨite as forms that could be used by the elders.SIA generally prefers the forms with ɨi.This makes SIA apparently the most conservative speaker, despite being the youngest of the three.The interim timeline of the developments discussed up to this point is given in Figure 1.The conventionalized season names are sufficiently morphologically transparent for some speakers to level the postlabial ɨi to ai.Nevertheless, the season names are semantically noncompositional and have to be learned on a case-by-case basis.Thus, the presence or absence of postlabial raising in (5) might show lexically-specific effects.
To test whether postlabial raising is a productive phonological process, I asked about the pronunciation of neologisms derived with the periodic suffix -ite PRD (6).In its productive usage, -ite PRD derives time period nouns and adverbials from nouns and verbs.E.g. fetʰa means 'open' and fetʰaite can mean either 'opening season' or 'in the opening season' (6a-i).Some of the season names derived in this way are absurd, e.g.tʃinaite 'daughter-in-law season' (6a-iii).Since they were almost certainly never heard before, these productions must reflect productive phonology. (

5) CONVENTIONALIZED FORMS DERIVED WITH -ite PRD
The three consultants show three different patterns when it comes to novel formations with the periodic -ite PRD.JXM only judges forms with ai as grammatical and does not recognize any variation in the speech community (6).Thus, JXM did not generalize postlabial raising as a productive phonological rule.
SIA does not allow for the raising of morphologically-derived ai to ɨi if the sequence arises after a non-labial consonant (6a).However, when the sequence arises after a labial, SIA allows for, but does not require, raising.Some forms with ɨi are identified as dispreferred or used by others (6b-iii,iv) but most of them are equally available for SIA.I speculate that the presence of forms where both ai and ɨi appear in the speech community (4-5) has led SIA to acquire postlabial raising as an optional but productive phonological rule (7).Note that since SIA does not raise ai to ɨi after labials unconditionally (2, 4-5), the rule must reference the morpheme boundary.
Finally, RGQ shows the greatest variation in his judgments.He always prefers forms with ai.Depending on the word, forms with ɨi are identified as impossible, possible but dispreferred, used by others, or archaic.Notably, RGQ notes that in some words the underlying sequence /ai/ may be realized as [ɨi], even when it is not preceded by labial consonant.This suggests that RGQ generalized the raising of ai to ɨi beyond its original conditioning environment (8).Given that RGQ does not raise ai to ɨi unconditionally (2, 4-5), but only allows for it in a derived environment, the acquired rule must reference the morpheme boundary.JXM and SIA's judgments reported for -ite PRD are largely replicated with -iʔkʰɨ INS.JXM never recognizes form with ɨi (9); SIA allows for raising postlabially (9b), but not otherwise (9a).A notable difference is that SIA identifies the ite-forms with postlabial raising as good, but he identifies many iʔkʰɨ-forms with postlabial raising as dispreferred or used only by others.
In the small sample of four roots which do not have a labial consonant before the final a, RGQ only recognizes forms with ai, and considers raising to be impossible (9a).In roots that have a labial consonant before a final vowel, RGQ considers raising to be largely impossible.In a few of those roots, raising is possible but dispreferred, used by others, or archaic (9b).Thus, the general availability of raising for RGQ is considerably lower in iʔkʰɨ-forms than in ite-forms.Since the periodic -ite PRD is a derivational morpheme and -iʔkʰɨ is an inflectional morpheme, I speculate that the difference may be attributed to morphological boundary strength.
In interim summary, I observed that (i) the diphthong ai is unattested after labial consonants in native A'ingae roots, (ii) some borrowings and conventionalized season names adapt postlabial ai as ɨi, and (iii) in novel morphologically complex words ai may sometimes optionally rise to ɨi.To account for these facts, I proposed that A'ingae had postlabial raising as a regular sound change.Some borrowings and conventionalized season names have Bai instead of the expected Bɨi due to ongoing language contact-induced replacement and paradigm leveling.The availability of both Bai 8 The other morpheme which begins with i is the instrumental case -iʔkʰɨ INS.Like the English with, the A'ingae instrumental also has the comitative function.Thus, the instrumental -iʔkʰɨ INS attaches freely to any animate or inanimate noun (9).and Bɨi in some lexical items catalyzed for some speakers the acquisition of (postlabial) raising as an optional morphologically-conditioned phonological rule.Nevertheless, postlabial raising lacks an obvious phonetic motivation.This is not a problem for the synchronic phonological rules in (7-8) since there is ample evidence that phonetically unnatural phonological rules can be acquired (Bach & Harms 1972;Kiparsky 1973;Hyman 2001).However, a phonetically unnatural process is not an expected sound change.
To address this issue, I speculate that A'ingae postlabial raising came about as two subsequent sound changes.First, postlabial *ai underwent rounding to *ui (10i).Second, the round back vowel *u underwent unconditioned unrounding and centralization to ɨ (10ii).(This idea was first suggested to me by Chelsea Sanker, p.c.) Two facts lend credibility to this scenario.First, the pressure to round postlabial vowels is independently attested in A'ingae.The diphthong ae can be rounded to oe after labials (11).This rounding is optional and may be partial (i.e. intermediate realizations such as a̹ e are possible).Postlabial rounding can apply within roots (11a) and across morpheme boundaries (11b).
Second, the origin of the Cofán people can be traced back to the Andean Cordilleras (Lucitante 2019).In the inventories of Andean languages, the vowel u is commonly attested (Figure 2), but ɨ is rare (Figure 3).Thus, reconstructing the vowel inventory of *a, e, i, o, u for the precolonial A'ingae is consistent with what one might expect given the known geography of the language at that time.The vowel ɨ, on the other hand, is common in Amazonian languages (Figure 3).Thus, the centralization and unrounding of *u to ɨ (10ii) is a plausible contact-induced shift.
The proposed timeline of changes is summarized in Figure 4.Note that this timeline implies that the original postlabial change, the postlabial rounding, must have happened over 400 years ago, a long time before any replacement and leveling of ɨi to ai.

Conclusion.
In conclusion, I investigated the diachronic relationship between the two A'ingae diphthongs ai and ɨi.I observed that ai does not appear after labial consonants, and that some borrowings and conventionalized forms show a shift of ai to ɨi.To account for these facts, I proposed that postlabial raising of ai to ɨi took place in A'ingae as a regular sound change.There are forms in contemporary A'ingae where ai appears after labial consonants.I argued that these are cases of paradigmatic leveling and replacement driven by language contact.
Native speakers differ when it comes to the application of postlabial raising to new productive formations: JXM does not allow for postlabial raising, SIA generalizes it as an optional but 9 productive phonological rule, and RGQ appears to generalize the raising beyond its original conditioning environment.Finally, I speculated about the postlabial raising of ai to ɨi as a regular sound change given its phonetic unnaturalness.I suggested that postlabial raising took place in two steps, with postlabial rounding (*ai > *ui / B _) followed by u-centralization (*u > ɨ).These two changes, I argued, are consistent with the phonological profile of the language and its known geographic history.