Chapter 52Relational and Spatial (Locative) Nouns and Their Possessed Forms

52.1 The class of relational nouns

A small closed set of nouns in Hokkaido Ainu encodes the spatial or relational position of one entity with respect to another. These nouns — or ‘place/locus’, ka ‘top/surface’, sam ‘side’, tum ‘interior’, corpok ‘underside’, and others catalogued below — are called locative nouns (位置名詞) in the Japanese-medium descriptive tradition Nakagawa (2024: 241, 280, 296–299); Satō (2021); Nakagawa (2022: 478–480).

Three distributional properties define the class. First, the members are obligatorily possessed: a relational noun cannot head a free-standing noun phrase in isolation; it must either be preceded by a noun phrase whose referent supplies the reference object, or it must carry a personal prefix directly to express whose spatial domain is at stake (§7 below; see Chapter 43 (The Alienable/Inalienable Split and Possessive Classification)). Second, Hokkaido Ainu case particles require a place-designating noun (〈場所〉) immediately before them Satō (2008: 34–35); Nakagawa (2024: 296); relational nouns are the main device for building that slot from an ordinary common noun. The canonical pattern is cise or ta 'at the house' (lit. house + place-of + at), and its extension to the other case particles (or un, or wa, or peka) constitutes the primary locative-case system, treated in Chapter 53 (Postpositions, Local Case, Motion Events, Comitative, and Instrumental). Third, a relational noun appears in its possessed (affiliative) form when combined with a host noun: or → oro, ka → kasi (§3). Whether the class is best analysed as a set of inalienably-possessed nouns or as a set of postpositions inflected for possession is the central analytical question of the chapter (§8).

52.2 Inventory and semantics

The table below lists the core members with their concept form, possessed (long) form, spatial meaning, and a representative construction. Possessed forms follow the locative-noun long-form system — the lexically specified suffixes -ke, -V, and -si (§ below) — rather than the common-noun affiliative pattern Nakagawa (2024: 226, 241–246). Nakagawa gives the long forms of or, ka, enka, sam, tum, and corpok directly; samake and the contracted tumke are also frequent in corpus text ‹corpus-confirmed›. The forms given for os, pok, and kotor are not directly verified in the sources consulted ‹speculative›. Nakagawa (2024: 241, 296–299); Tamura (1996: s.v. or, ka, sam, tum, corpok, soy, kotor); Satō (2021)

A few entries call for comment. The pair ka / enka are often treated as co-members of the 'above' domain: ka encodes contact or near-contact with a supporting surface ('on top of, resting on'), while enka encodes vertical distance above without contact ('over, above'). The pair corpok / pok occupies a similar near-duplicate slot in the 'below' domain; the distributional boundary between them is not settled by the sources consulted ‹contested›. The noun kotor appears reliably as a component of the compound tekkotor 'palm of the hand' (lit. hand + surface) Nakagawa (2024: 250), and the pattern suggests a class of vertical-surface spatial nouns; but its use as a free relational noun head (i.e. X kotor ta 'at the face of X') is less thoroughly documented in the sources consulted. The noun so appears in corpus narrative in the construction iwor so ka 'over the top of the hunting ground', where so (upper back area) combines with ka (top) — a stacking of two relational nouns into a single locative expression (see §4 below). Its standalone affiliative form has not been identified in the sources consulted.

Two relational nouns — or and os — also appear in Nakagawa's catalogue of postpositional adverbs (後置副詞), the autonomous adverb class that does not require a 〈場所〉 host Nakagawa (2024: 280–282); Satō (2008: 34–36). This dual membership is a structural fact at the heart of the noun-vs-adposition debate in §8.

52.3 The long-form suffixes -ke, -V, and -si

The possessed forms in the table are built with a small suffix set — -ke, an echo or fixed vowel (-V), and -si — that is lexically specified per noun: Nakagawa's long form (長形) of the locative noun Nakagawa (2024: 241–246). The system is separate from the common-noun affiliative suffixes, and Nakagawa hypothesises that one layer of common-noun possessive morphology was borrowed from it Nakagawa (2024: 226, 247) ‹speculative›. -ke is the most frequent member, and several nouns combine two layers: kasi (ka + -si) is further extendable to kasike(he).

Consonant-final bases insert an echo vowel before -ke: sam-a-ke, tum-u-ke. The contracted shapes without the echo vowel are frequent in text — tumke is much commoner in the corpus than tumuke (aynu tumke wa 'from among the Ainu', amam tumke ta 'among the grain'), and kaske stands beside kasike ‹corpus-confirmed›.

Three grammatical properties ride on the long form Nakagawa (2024: 241–246). It can head a noun phrase alone, which the Type-1 basic form cannot; it is the form that carries the objective person prefixes (e=samake ta 'beside you'); and it accepts the further possessive extension -he (samakehe). In running text the long form appears almost always in the layered locative frame, followed by a case particle: X samake ta, X tumke ta/wa, and temporal okake ta 'after that, afterwards', a high-frequency narrative connective ‹corpus-confirmed›.

This locational -ke is homophonous with two verbal suffixes — the anticausative -ke and the transitivizing -ke of the lexical valency pairs (Chapter 75 (The Transitivizer -ka and Lexical Valency Pairs)) — but distinct from both: it attaches only to the closed locative-noun class and yields a noun, never a verb.

The string orke carries two readings that have to be sorted contextually, and Tamura's Saru dictionary documents both Tamura (1996: s.v. arikikpa, cikuni, cikosomokur koyaykatanu). In suptom orke 'the place around the middle of the trunk' and utar orkehe it is the ordinary long form of or, the same locative formation described above — the -he extension in the second confirms the analysis. In the paired frame N orke … N orke the same string reads additively: anun orke apa orke nep uwehosi arikikpa 'outsiders and kin alike, all of them vie with each other in working hard'. Corpus tokens divide the same way — tu kewtum orke re kewtum orke 'in two hearts, in three hearts' keeps the locative sense under the numerals ‹corpus-confirmed›.

52.4 Obligatory possession and affiliative forms

Relational nouns belong to the inalienable (obligatorily-possessed) noun class, alongside body parts, kinship terms, and part–whole relational stems. Like those classes, they cannot appear as a syntactic head without an expressed or inferable possessor Nakagawa (2024: 241). The morphology of this possession is the affiliative (所属形): in the construction [NOUN + relational-noun + case-particle], the relational noun takes its possessed form. The phonological mechanism that produces or → oro, ka → kasi, sam → sami, and so on, is the echo-vowel or suffix-vowel process described in Chapter 45 (Morphophonology of the Affiliative Suffix and Echo/Copy Vowels).

A diagnostically important minimal pair is provided by ni ka ta 'onto/on top of the tree' and cikuni kasi un 'toward the treetop' Nakagawa (2024: 298–299). The first uses the concept (bare) form ka before the locative particle ta; the second uses the affiliative kasi before the allative un. The distribution of concept vs affiliative form before different case particles for this noun parallels what Satō Satō (2021) documents for or in Chitose: both forms are grammatical with a following locative noun, but the short-form fusion (or + ta → otta) is the unmarked spoken realization, while oro ta appears in more deliberate or expanded styles Nakagawa (2024: 297) ‹corpus-suggested›.

A community discussion on samampa (nukopoli, aynu-corpora Discord (2023–2026) 2024-11-26) analyses it as sam-am-pa 'lying on the side' — identifying sam 'side' as a productive base entering compound verb formation — and proposes that samama 'to place sideways' = sam-ama (Kayano's Saru dictionary) ‹speculative›. This use of sam as an incorporated nominal base in complex verbs is a point of intersection with the noun-incorporation system (see Chapter 88 (Object (Patient) Noun Incorporation)).

52.5 The layered locative construction

The canonical locative construction in Hokkaido Ainu is a three-member sequence: a reference-object noun phrase, followed by a relational noun in its affiliative or concept form, followed by a case particle. Nakagawa labels the middle position the 〈場所〉 'place' slot — a morphosyntactic position that the locative, allative, ablative, and perlative particles all require Nakagawa (2024: 296); Satō (2008: 34). Relational nouns are the primary filler of that slot for ordinary common nouns.

The particle-specific fusion of or merits particular attention. When or immediately precedes the locative particle ta, the sequence contracts to otta: cise otta 'at the house', nupuri otta 'at/on the mountain'. The long form oro ta remains uncontracted Nakagawa (2024: 297). This fusion does not apply to or un, or wa, or or peka; those remain uncontracted in all varieties.

The construction extends straightforwardly to other relational nouns. The same case-particle set attaches to ka, sam, soy, and the other members of the inventory:

Two relational nouns may stack in sequence when the spatial description requires a two-step reference: an outer reference object, then a spatial noun within it, then a second spatial noun closer to the target. The corpus yields iwor so ka a=ehóyupu wa 'flew over the top of the hunting ground's upper part' (lit. hunting-ground + upper-back + top + 4.A=fly.around + CONJ) ILCAA Ainu materials (1976), where so 'upper-back area' specifies the subregion of iwor 'hunting ground', and ka then positions the protagonist above that subregion ((ILCAA Ainu materials 1976), Kawagami Matsuko, uwepeker 民話8, aa-irc/006 line 52). Such stacking is consistent with the head-initial, modifier-following structure of Ainu nominal possession ‹corpus-suggested›.

52.6 The central role of or

Among the relational nouns, or occupies a singular position: its meaning has bleached to the point where it functions as a general locative placeholder for any noun requiring the 〈場所〉 slot, not just for entities that constitute physical places. Satō's dedicated Chitose study Satō (2021) traces the drift of or from a concrete relational noun 'the place of X' toward a grammaticalizing locative morpheme that renders any nominal as a place-designating argument.

The bleaching is visible in combinations where the host noun is not inherently a physical location. In tan KIKAY or ta a=ewkoytak hi an yakun 'if we had a conversation with this machine' (lit. this machine's-place at), Nakagawa (2024: 284) (ex. 373, Saru; 田村 1984: 50), the host noun KIKAY (a recording machine) is an instrument, not a spatial locus; or ta constructs the 〈場所〉 argument that the particle ta requires, mapping the instrumental participant onto a locative frame. The Chitose corpus adds many parallel instances where abstract events or instruments appear in the host-noun position Satō (2021).

Whether this drift constitutes the grammaticalization of or into a general locative case-marker, or whether it reflects a productive but still compositional use of the relational noun, is a point on which Satō Satō (2021) and Nakagawa Nakagawa (2024: 241, 296–299) take partially different positions ‹contested›: Nakagawa treats the construction as productive composition (noun + spatial-noun + particle), while Satō's Chitose evidence shows or extending into contexts where the 'place of X' semantics is increasingly opaque, suggesting at least the beginning of a grammaticalization cline. The issue connects directly to the noun-vs-adposition debate in §8, because an advanced-stage grammaticalized or would be functionally indistinguishable from a locative postposition.

52.7 First-class and second-class locative nouns

Nakagawa Nakagawa (2024: 241) distinguishes two sub-classes of locative noun in Hokkaido Ainu, a distinction also partly reflected in Satō Satō (2008: 34–36). The first-class locative nouns are those whose primary function is to supply the 〈場所〉 slot: they follow a host noun and directly host a case particle. or, ka, sam, soy, and corpok are the clearest members. The second-class members are those whose primary occurrence is in compound nouns and verb bases rather than in the free-standing [NOUN + relational-noun + case-particle] frame. kotor 'surface' is the paradigm case: it appears reliably in the lexicalized compound tekkotor 'palm of the hand' (tek ‘hand’ + kotor) Nakagawa (2024: 250) and has left traces in the archaic directional forms, but it is not a productive first-class head in the contemporary spoken register. Similarly, tum 'interior/midst' appears in place-name formation (proposed: o-tum-nay 'head-interior-river' for the toponym Otomonai; nukopoli, aynu-corpora Discord (2023–2026) 2026-01-22) ‹speculative› and in compound verbs (tumpu 'storehouse' = tum 'interior' + pu; antitwilight, aynu-corpora Discord (2023–2026) 2024-11-16) ‹speculative›, but its distributional range in the first-class frame remains less thoroughly documented than that of or or ka.

The class boundary is not sharp, and intermediate members exist. In the postpositional adverb system Nakagawa (2024: 280–281), os(i) 'after' appears as an autonomous postpositional adverb that does not require a 〈場所〉 host. This dual membership — relational noun in one construction, postpositional adverb in another — is also observed for or itself: Nakagawa lists it in both noun paradigms and in the broader locative system as a particle-combining locative element. The overlap between the two classes is structural evidence for the noun-vs-adposition debate (§8).

An archaic directional suffix -na is attested in frozen forms: setunna 'from the front' and kotunna 'from behind', analysed as setur-na and kotor-na respectively (antitwilight, aynu-corpora Discord (2023–2026) 2026-06-09); compare also pokna 'below', which may preserve the same suffix on the relational noun pok. This suffix is not a productive case particle in contemporary Hokkaido Ainu, and no primary-source grammar in the read set treats it as such ‹speculative›. Its existence as a fossil form is consistent with the hypothesis that the relational-noun class is older than the current case-particle system and that some members grammaticalized into the particle system at an earlier stage.

52.8 Non-third-person possession

The possessor of a relational noun is third person (zero-marked) when the reference object is supplied by the preceding host noun phrase. But relational nouns can also be directly possessed by speech-act participants via the standard personal prefixes treated in Chapter 62 (Architecture of the Personal-Affix System: The Four Persons and the S/A/O Paradigms) and Chapter 46 (Adnominal Possession: Double Marking and the Appositive Construction). In that case, the relational noun itself — not the host noun — bears the person prefix:

The full person paradigm for or follows the ordinary nominal possession pattern: ku=or 'my place', e=or 'your place', or (zero) 'his/her/its/their place', ci=or (1PL.EXCL) 'our place', a=or (4th person) 'the protagonist's/their general place' Nakagawa (2024: 241, 296). The same paradigm applies to all other first-class relational nouns: ku=ka ta 'on top of me', e=sam ta 'beside you' ‹consensus›. These forms appear in contexts where the locative relation targets a person rather than a physical landmark, and they are attested in body-part locatives and relational idioms (see Chapter 47 (External Possession and Possessor Raising) for how this intersects with external possession).

A productive derived construction involves the verb eykaun 'to surpass/be superior to', analysed by nukopoli (aynu-corpora Discord (2023–2026) 2024-11-24) as ka un 'be at the top of (someone)' with an applicative or transitivizing suffix — effectively the relational noun ka with allative un grammaticalized as a comparative marker ‹speculative›. Whether this represents a synchronically transparent construction or a frozen derivation is not settled by the sources consulted.

52.9 Noun or adposition?

The central analytical question is whether relational nouns constitute a subclass of nouns or a subclass of postpositions. The dispute turns on which grammatical properties are weighted most heavily, and different descriptive traditions have come down on different sides ‹contested›.

Evidence for the noun analysis. Relational nouns take person affixes in exactly the same slots and with exactly the same forms as other inalienably-possessed nouns (ku=or, ku=sam, ku=ka parallel ku=sapa 'my head', ku=tek 'my hand'); they appear in the possessed (affiliative) form under possession, following the same morphophonological rules as the inalienable class (Nakagawa (2024: 241); Nakagawa (2022: 478–480)). They enter compound-noun and incorporated-noun formation as ordinary nominal bases (tekkotor, tumpu), which is diagnostic of noun status in Ainu (see Chapter 88 (Object (Patient) Noun Incorporation)). Nakagawa Nakagawa (2024: 241, 280) treats them as a noun subclass throughout. The same analysis underlies Ijäs's pedagogical treatment Ijäs (2023), which presents the locative construction as a possessed-noun-plus-particle sequence.

Evidence for the adposition analysis. Functionally, relational nouns occupy the same distributional slot as adpositions: they appear between a nominal complement and a case particle, and they encode the same spatial relations that languages with postpositions typically encode with a postpositional head. Several members of the class — notably os(i) 'after' — appear in Nakagawa's catalogue of postpositional adverbs (後置副詞) Nakagawa (2024: 280–282); Satō (2008: 34–36). Satō's Chitose evidence for or Satō (2021) shows the form extending beyond referential 'place of X' into a general locative-marking function, which is the expected trajectory of a grammaticalizing postposition, not of a lexical noun. The pedagogical grammar by Ijäs Ijäs (2023) places some members of the class in its postposition section alongside case particles — a pragmatic descriptive choice that reflects the adpositional function these forms serve for learners.

The most defensible synchronic position is that the class is in transition: the morphological behaviour (person affixes, affiliative alternation, incorporation) identifies the members as nouns, while their functional distribution (obligatory position between complement and particle, encoding spatial relations, extension to non-spatial contexts) pushes in the adposition direction. A parallel situation is found in several languages of Mesoamerica, where the same class of inalienably-possessed nominal forms — often glossed as 'relational nouns' in the descriptive literature — occupies this intermediate position between noun and adposition (Bugaeva 2022). In those typological terms, Hokkaido Ainu relational nouns are best labelled relational nouns rather than postpositions or adpositions: they are nominal in morphology and adpositional in function ‹contested›.

Whether first-class members like or and ka are more advanced on a grammaticalization cline toward postpositions than second-class members like kotor is a plausible hypothesis supported by Satō's work on or Satō (2021) but not yet generalized across the full inventory ‹speculative›. The syntactic behavior of relational-noun phrases in nonconfigurational and pronominal-argument contexts is taken up in Chapter 94 (Pro-Drop, (Non)configurationality, and the Pronominal-Argument Debate) and Chapter 95 (Oblique and Adjunct Syntax; Verbless and Presentational Clauses).