Chapter 45Morphophonology of the Affiliative Suffix and Echo/Copy Vowels

45.1 The short and long affiliative: the vowel-copy principle

The affiliative (possessed) form of a common noun differs from its concept form by a suffix whose vowel echoes the final vowel of the stem. The suffix appears in two lengths: a short form consisting of the copy vowel alone (for vowel-final stems this adds nothing visible to the concept form), and a long form consisting of copy vowel + /h/ + copy vowel again, notated -hV₁ where V₁ is the stem's last vowel Nakagawa (2024: 226–228); Satō (2008: 152–153). The two lengths carry no semantic distinction; the choice between them is governed by prosody and speaker rhythm, and Satō describes the usage split as "unclear" Satō (2008: 152) ‹contested›.

The possessive person-prefix that co-occurs with the affiliative form belongs to the transitive-subject (A) prefix set — identical to the verbal A paradigm — as set out in Chapter 62 (Architecture of the Personal-Affix System: The Four Persons and the S/A/O Paradigms); the full double-marking construction is treated in Chapter 46 (Adnominal Possession: Double Marking and the Appositive Construction). This chapter addresses only the morphophonology of the suffix: its shape, the conditioning environments of its three consonant-final classes, the irregulars, and the diachronic source of the /h/ element.

The conditioning sorts into three groups by the stem's final segment. Vowel-final stems (type ①) produce the most transparent echo: the long form adds -hV₁, the short form adds nothing. Glide-final stems (type ②) take the uniform suffix -e(he). Consonant-final stems other than glides (type ③) belong to one of three lexically listed sub-classes: the default -i(hi), a minority echo-vowel class -V(hV), and a further minority -u(hu) class restricted to stems whose last vowel is /a/ Nakagawa (2024: 226–228).

45.2 Type ①: vowel-final stems

When the concept form ends in a vowel, the short affiliative is identical to the concept form; a bare third-person (zero-prefix) vowel-final noun is therefore ambiguous between concept and affiliative readings and requires context for interpretation Nakagawa (2024: 227); Satō (2008: 152). The long form adds -hV₁, repeating the stem-final vowel:

In each row the suffix vowel is identical to the stem-final vowel: /a/, /u/, /e/, /o/ respectively. Because the short affiliative coincides with the concept form, corpus string-searches cannot separate them; frequency figures for long forms — sapaha 206 tokens in the present corpus — are the only clean affiliative counts for this class. The paradigm itself follows (Nakagawa 2024: 227).

45.3 Type ②: glide-final stems and the monosyllabic subtype

When the concept form ends in the glides /y/ or /w/, the affiliative suffix is uniformly -e(he), regardless of the preceding vowel quality Nakagawa (2024: 227); Satō (2008: 152):

The affiliative hawe ‘voice (possessed form)’ of haw ‘voice’ is attested as both an ordinary possessed noun ('X's voice') and, fully grammaticalized, as one of the four formal-noun evidentials treated in Chapter 44 (Concept Form vs. Affiliative (Possessed) Form). The two uses are morphologically identical; only context separates them Nakagawa (2024: 258–259).

A formally parallel subtype arises with monosyllabic open-syllable (CV) nouns whose concept form ends in /i/ or /u/. These surface as vowel-final but take the type ② suffix -e(he) rather than the type ① echo -hV₁, as if an underlying glide — absent from the concept form — conditions the suffix Nakagawa (2024: 229):

Sakhalin Ainu, which preserves phonemic vowel length, has the corresponding nouns as cii, kuu — long-vowel forms plausibly reflecting an earlier *Ciy/*Cuw sequence and supporting the underlying-glide analysis Nakagawa (2024: 229); Itabashi (2001) ‹speculative›. The affiliative ruwe is by far the most frequent of the four, since its grammaticalization as the recognized-fact evidential places it in clause-final position across a large share of narrative sentences; a clean affiliative count is not available by string-search because ruwe is homophonous with the lexical noun and with a reduced final particle Nakagawa (2024: 258–259).

45.4 Type ③: consonant-final stems

Nouns whose concept form ends in a consonant other than /y/ or /w/ belong to type ③. Three suffix shapes compete, and class membership within type ③ is in large part lexically determined Satō (2008: 153); Nakagawa (2024: 227–228).

45.5 The ③-1 default surfaces in compounds

Sub-type ③-1 is the unmarked pattern. A ③-2 or ③-3 stem that appears as the second element of a compound reverts to the ③-1 suffix, regardless of its own lexical class Nakagawa (2024: 228). The noun at ‘cord, rope’ alone takes the ③-2 long affiliative atu-hu (V₁ = /u/); as the second element of tar-at 'carrying-rope cord' the compound takes tar-ac-i(hi), the ③-1 form Nakagawa (2024: 228). This alternation is the clearest synchronic evidence that ③-1 is phonologically basic and ③-2/③-3 are lexically marked Satō (2008: 153) ‹corpus-suggested›.

The same diagnostic applies to the evidential formal nouns. The noun sir ‘visible appearance, land’ alone gives the short ③-1 affiliative siri, which grammaticalizes as the situational evidential. The formal noun siri blocks the long extension in its evidential function — it is always the bare short form, never *sirihi — suggesting that grammaticalization has lexically frozen the short form in this use Nakagawa (2024: 258–259) ‹corpus-confirmed›.

45.6 True irregulars

Two nouns have affiliative forms that none of the productive rules derives. par ‘mouth’ yields paro(ho), not the expected ③-1 *parihi. Nakagawa traces this to the word's etymology: par is a compound of pa 'upper part' and or 'inside, place', so the ending -o(ho) is the locative long-form of or rather than an affiliative suffix proper Nakagawa (2024: 230). The second irregular, mor ‘temple-hair’ → moru(hu), is attributed to a morphological re-analysis of an earlier stem moru Nakagawa (2024: 230) ‹speculative›.

Kinship terms add further formal idiosyncrasies: address-form vs reference-form pairs where the two stems take different affiliative classes (e.g., reference yup ③-1 vs the address form yupo ①), and frozen dialectal variants. These are treated in Chapter 48 (Kinship and Honorific Possession).

45.7 Lexical conditioning: what is predictable and what must be memorized

Types ① and ② are phonologically predictable from the concept form's final segment: any vowel-final stem belongs to ①; any /y/- or /w/-final stem (surface or underlying) belongs to ②. Within type ③ the situation is different. Satō states that "there are many exceptions … each must be memorised individually" Satō (2008: 153), and Nakagawa 2024 concurs that ③-2 and ③-3 are lexically listed. For V₁ = /a/ stems the task is hardest: a given noun may belong to ③-1, ③-2, or ③-3 with no phonological predictor. For V₁ = /i/ the situation is simpler — all /i/-stem consonant-final nouns take ③-1 Nakagawa (2024: 227–228).

The pedagogical grammar of Ijäs (2023) presents the consonant-final class under the label 'belonging form' and treats -ihi as the default suffix, consistent with the ③-1 analysis. Ijäs does not distinguish ③-2 from ③-3 as labeled sub-classes; body-part forms with non-default suffixes are presented as items to be memorised. For the obligatorily-possessed nouns that constitute the core affiliative class, the lexical class assignment is part of each noun's dictionary entry.

45.8 Diachrony: the source of -hV and the vowel-copy analysis

Chiri 1952 described the vowel quality of the affiliative suffix as an instance of vowel harmony spreading a feature from the stem Chiri (1952). The present description reframes this as vowel copy: the suffix vowel is a morphologically triggered copy of the stem's last vowel, not an autosegmental spread. The distinction is empirically relevant: in type ③-2 the copy operates across an intervening consonant (tek → -e), where a strictly local harmony rule would require extra stipulation Nakagawa (2024: 226–228) ‹contested›.

The origin of the ③-2 echo-vowel class is addressed separately by Nakagawa 2024: 228, 247–248, who proposes that it derives from the locative long-form suffix productive on relational nouns. The relational noun sut ‘behind; grandmother’ takes the locative long form sutu; the same pattern recurs in the compound tap-sut 'shoulder' → tap-sut-u(hu), suggesting that the ③-2 suffix in such nouns was originally a locative ending that later spread into the common-noun affiliative paradigm. Nakagawa describes the connection as "intimately related" Nakagawa (2024: 228, 247–248), and notes that the ③-3 pattern (V₁ = /a/ → /u/) "cannot be explained historically" Nakagawa (2024: 228) ‹speculative›. Neither proposal has been independently replicated. See Chapter 52 (Relational and Spatial (Locative) Nouns and Their Possessed Forms) for the relational-noun long-form paradigm.

The /h/ in the long-form suffix (-hV₁) has a proposed diachronic correlate. Sakhalin Ainu, which preserves phonemic vowel length, has long vowels in nouns corresponding to Hokkaido affiliatives with -hV; Itabashi 2001 argues that proto-Ainu coda consonants were lost in Hokkaido and partially compensated by pitch accent, while their trace survives in Sakhalin long vowels Itabashi (2001). Under this account the /h/ in sapa-ha, siki-hi, etc. reflects a historical boundary consonant or coda that became /h/ rather than zero in the possessive environment Nakagawa (2024: 229); (Vovin 1993) ‹speculative›.

45.9 Sakhalin contrast and the -ike/-hike variant

Sakhalin Ainu uses the same vowel-copy mechanism but with the extended long form as the more consistent default: where Hokkaido freely alternates short hawe and long hawehe, Sakhalin typically retains the extended shape. A community analysis of Sakhalin data observes that the stem san yields sanihi via san + insertion /i/ + -hi, and that makan similarly yields makanihi; the analysis further notes a Sakhalin variant -hike in a subset of forms (sani ike alternating with sanihi), where -ke may represent an additional nominal element appended to the affiliative aynu-corpora Discord (2023–2026) (nukopoli, 2024-12-16) ‹speculative›. The distributional and historical relationship between Sakhalin -hike and Hokkaido -hi has not been formally investigated in the accessible sources ‹speculative›.

A distinct element -ike appears in a small set of Hokkaido compound nouns where Chiri glosses it as 'person (who is X)' Kindaichi & Chiri (1936). The eastern-dialect form onneke(he) 'parent/mother (Tokachi, Nemuro)' has been analysed as onne 'old' + -ike, with the compound then additionally taking the affiliative suffix (onneke-he), showing that -ike here is a stem-forming element and the affiliative suffix is separate aynu-corpora Discord (2023–2026) (antitwilight, 2024-11-15) ‹speculative›. The affiliation of this -ike to the affiliative paradigm, if any, is etymological rather than synchronic and is not a productive pattern in Hokkaido Ainu.