Chapter 28Assimilation and Cluster Simplification: Coda /r/, Nasals, and Clusters
Changes involving final /r/ and /n/, nasal place assimilation, particle-specific alternations, and the roles of pauses and sentence boundaries.
28.1 Consonants in connected speech
A consonant at the end of an Ainu syllable can change before the next sound. The best described cases involve final /r/ and /n/, and occur both within words and across word boundaries. Descriptions use the term sandhi for these changes Shiraishi (2022: §§6.1–6.2).
The patterns have different conditions. Some concern articulation, some are restricted to particular grammatical forms, and some distinguish dialectal word shapes. Satō and Nakagawa both caution that many alternations vary by speaker or occurrence Satō (2008: 12–13); Nakagawa (2024: 36–38). A table of possible changes should therefore not be read as an instruction to apply every change in every utterance.
28.2 Changes involving final /r/
The usual patterns include /rt/ becoming [tt], /rc/ becoming [tc], and /rn/ becoming [nn]. Nakagawa illustrates them with the following combinations; the pronunciation column uses his segmental notation Nakagawa (2024: 36):
| Sequence | Underlying combination | Pronunciation | Meaning |
|---|---|---|---|
| /r+t/ | ku=kor tar | kukot tar | my carrying strap |
| /r+c/ | retar cikap | retat cikap | white bird |
| /r+n/ | ku=kor nima | kukon nima | my wooden bowl |
Satō's Chitose examples show the same three environments, and Shibatani describes them in his survey Satō (2008: 12–13); Shibatani (1990: §2.4). Further changes before /s/ or /y/ have narrower distributions. Nakagawa explicitly identifies the pronunciation kukoy yupi from ku=kor yupi “my elder brother” as Chitose Nakagawa (2024: 36).
28.2.1 Final /r/ before another /r/
Before /r/, final /r/ can become [n]: kor rusuy is pronounced kon rusuy “want to have.” This is traditionally described as dissimilation, since the adjacent rhotics become less alike Satō (2008: 13); Nakagawa (2024: 37). Shiraishi also gives kon rametok “his bravery” from kor rametok Shiraishi (2022: §6.1, example 27g).
Shiraishi compares this traditional account with his proposal that /r/-sandhi generally avoids a rhotic followed by another coronal consonant. That proposal groups outcomes otherwise described separately as assimilation and dissimilation Shiraishi (2022: §6.1) ‹contested›.
28.2.2 Recognizing a changed final consonant
‘a white bird’
Nakagawa 2024: 36; Hokkaido (dialect not further specified)
Nakagawa gives the sandhi pronunciation retat cikap. The example uses the underlying forms; English glosses are supplied here.
In this example, the /r/ of the descriptive word changes before the affricate beginning the next word. The resulting [t] belongs to the same syllable that contained /r/. This behavior helps distinguish a final rhotic from a rhotic followed by a full vowel Nakagawa (2024: 38). See Chapter 19 (The Rhotic /r/ and Coda-r) for the related evidence from vowel-like releases and suffix selection.
28.3 Nasal place assimilation
Final /n/ can take the place of articulation of a following consonant. It is realized as [m] before /p/ or /m/, and as [ŋ] before /k/. Nakagawa gives the velar realization in honkor “be pregnant” Nakagawa (2024: 30). Shibatani illustrates the labial pattern with pon menoko “small girl,” pronounced with [m] before the second word Shibatani (1990: §2.4).
These are contextual pronunciations of /n/. A phonemic transcription can retain n where a phonetic transcription records [m] or [ŋ]. This distinction is especially useful when /n/ precedes a form whose initial consonant also changes.
28.4 Final /n/ before continuants
A separate set of changes involves /n/ before /s/, /y/, and /w/. Before /s/ or /y/, /n/ can become [j], written y. The following example is shared by several descriptions Nakagawa (2024: 36); Satō (2008: 13); Shiraishi (2022: §6.1):
‘a small dog’
Nakagawa 2024: 36; Hokkaido (dialect not further specified)
The source gives poy seta as the sandhi pronunciation. The example preserves underlying forms; English glosses are supplied here.
Before /w/, [ww] is also documented: Satō gives aw wentarap from an wentarap Satō (2008: 13). Shiraishi groups these triggers as continuant consonants and analyzes the change in terms of their articulatory properties Shiraishi (2022: §6.1). The analysis is his account of the shared pattern.
The particle wa has another pattern: it can become ma after /m/ or /n/, and pa after /p/. Thus an wa can surface as an ma, pronounced [amma] Satō (2008: 13). Nakagawa restricts this alternation to grammatical uses of wa and notes individual variation Nakagawa (2024: 37). It is not a general change of every initial /w/ after a nasal.
28.4.1 Pauses and sentence boundaries
Shiraishi's study of Saru recordings found a difference between the two sandhi systems. /n/-sandhi did not extend across a sentence boundary in the material examined, whereas /r/-sandhi did. An actual pause could block either process Shiraishi (2022: §6.2). Sentence boundaries and pauses therefore have distinct roles in this account.
‘‘Say it again. Say it again.’’
Kayano 1998a: 74, via Shiraishi 2022 §6.2, example 28c; Saru
Kayano's example as reproduced by Shiraishi. The final /n/ of the first yan does not undergo sandhi before the following ye, despite the reported absence of an intervening pause.
Shiraishi interprets the difference through distinct domains for the phonological rules. The observation is tied to his recording sample; its frequency and dialect coverage should not be expanded beyond that evidence Shiraishi (2022: §6.2).
28.5 Written forms and lexicalized outcomes
Printed texts handle sandhi in several ways. Nakagawa notes that changes inside a word are often represented in the spelling, while changes across word boundaries are often left unwritten. He also illustrates underscore conventions for indicating the affected boundary Nakagawa (2024: 38).
For example, retar plus -cir appears as retatcir “swan.” Across words, a source may write either ritur ta or the changed form ritut ta “on the way” Nakagawa (2024: 38). Absence of a changed consonant in the spelling does not establish its absence from the pronunciation.
28.6 Eastern dialect correspondences
Some regional word forms reflect consonant assimilation within the lexical item. Nakagawa and Fukazawa contrast widespread hotke “lie down” with hokke in Obihiro, Kushiro, and Bihoro. Their comparison also includes Bihoro tutte beside tupte “move something” elsewhere Nakagawa & Fukazawa (2022: §3.3, table 5).
Nakagawa treats such established dialectal word shapes separately from productive connected-speech alternations. The presence of /kk/ in a Kushiro word does not license a general /tk/ → /kk/ rule across adjacent words Nakagawa (2024: 36–37). Broader regional patterns are discussed in Chapter 160 (Hokkaido Dialect Variation in Sound and Grammar).
28.7 Conditions to keep distinct
| Pattern | Relevant condition | Scope of the description |
|---|---|---|
| /r/-sandhi | Following consonant; pause and dialect | Can extend across sentence boundaries in the cited Saru recordings. |
| /n/-sandhi | Following continuant; sentence boundary and pause | Did not extend across sentence boundaries in that recording sample. |
| Nasal place assimilation | Following consonant's articulation | Phonetic realization can differ from the phonemic spelling. |
| Changes of wa | Preceding consonant and grammatical identity | Lexically restricted, with speaker variation. |
| Eastern cluster correspondences | Dialect and lexical item | Established word forms do not entail a productive boundary rule. |
These distinctions follow the source accounts summarized above Nakagawa (2024: 30, 36–38); Shiraishi (2022: §§6.1–6.2). Vowel-contact phenomena are treated separately in Chapter 29 (Glide Epenthesis and Vowel-Hiatus Resolution).
References cited in this chapter
Kayano (1998a) ·Nakagawa (2024) ·Nakagawa & Fukazawa (2022) ·Satō (2008) ·Shibatani (1990) ·Shiraishi (2022)