Explaining alienability splits in the use of overt and zero possessive marking: a source-oriented approach

1 Introduction

A well-known pattern in the encoding of adnominal possession crosslinguistically involves splits in the use of overt and zero marking for alienable and inalienable possession. The notions of alienable and inalienable possession are used here, in a maximally general sense, to refer to a semantic opposition between non-inherent and inherent possessive relations between two entities (e.g., ‘John’s house’, ‘John’s books’ vs. ‘John’s mother’, ‘John’s hand’, or ‘the branch of the tree’), independently of how these relations are encoded in particular languages.^[1] While these two relation types can be treated in the same way in terms of overt and zero marking (both overtly marked, both zero marked, or both overtly or zero marked in different cases), languages also recurrently display splits where overt marking is restricted to alienable possession, or zero marking is restricted to inalienable possession.^[2]

Some of these patterns are illustrated in (1) and (2) below. In the Imonda construction in (1), both alienable possession (‘father’s house’) and inalienable possession (‘my father’) are encoded by the possessive marker -na.

In the Manam constructions in (2), on the other hand, alienable possession (‘the chief’s garden’) is encoded by the possessive marker -?aná, whereas inalienable possession (‘their father’) is zero marked.

Languages, however, do not usually display splits where overt marking is restricted to inalienable possession, or zero marking is restricted to alienable possession.

A classical and very influential explanation for this phenomenon, originally proposed by Haiman (1983, 1985, is in terms of iconicity. Relations of inalienable possession are determined by inherent properties of the possessee, and hence involve higher conceptual contiguity between possessor and possessee than relations of alienable possession. This will be iconically reflected by the relative linear distance between possessor and possessee in the constructions used to encode these two relation types, as well as by the relative degree of morhosyntactic complexity of these constructions. The use of overt marking will be avoided in inalienable possession constructions, both because overt markers are often placed between possessor and possessee, thus increasing the linear distance between the two (as in (1)), and because they increase the morphosyntactic bulk of the construction, even when they are not placed between possessor and possessee (as in (2a)).

Alternatively, the distribution of overt and zero marking across alienable and inalienable possession has been explained in terms of the relative need to disambiguate the possessive relation (Dahl and Koptjevskaja-Tamm 1998, 2001; Haspelmath 2008, 2017; Koptjevskaja-Tamm 1996; Nichols 1988). The semantics of inalienable nouns implies that their referents are in a specific relation with some other referent, so this relation does not need to be overtly specified, and can be zero marked. By contrast, the referents of alienable nouns need not be involved in a relation with other referents, so this relation is more difficult to identify and hence more in need to be specified through overt marking. This explanation involves a principle of economy whereby speakers will use overt marking only when this is really necessary, that is, for meanings that are more in need of disambiguation.^[3]

Both of these explanations imply that alienability splits are motivated by a functional match between the use of overt versus zero possessive marking and general properties of alienable and inalienable possession, either in the sense that this use matches the relative degree of conceptual contiguity between possessor and possessee, or in the sense that it matches the relative need to disambiguate the possessive relation. These matches favor overt marking in alienable possession contexts over inalienable possession ones, and zero marking in inalienable possession contexts over alienable possession ones.

These explanations, however, are generally based on the synchronic distribution of overt and zero marking across alienable and inalienable possession crosslinguistically, not diachronic phenomena that shape this distribution from one language to another over time (though some grammaticalization processes involved in the development of overt marking have been assumed to be triggered by the higher need to disambiguate the possessive relation: Dahl and Koptjevskaja-Tamm 1998; Heine 1997: Chap. 3; Koptjevskaja-Tamm 1996). In this paper, a number of crosslinguistic developmental processes will be discussed that have been shown to recurrently give rise to the possessive markers involved in the alienability splits in question, that is, splits where the use of possessive markers is restricted to alienable possession, or zero marking is restricted to inalienable possession. Such processes, it will be argued, provide no evidence that these splits reflect general properties of alienable and inalienable possession.

As will be shown in Section 2, data from grammaticalization studies and studies of possessive constructions in individual languages worldwide show that the relevant possessive markers are derived from elements originally used for other functions (attributive or appositive demonstratives, various types of appositive nouns, indefinite pronouns, locative elements). These elements evolve into possessive markers as they take on a possessive meaning originally associated with their context of occurrence. This is basically a metonymization process triggered by the co-occurrence of the source element and the possessive meaning, independently of alienability.

The distribution of individual markers across alienable and inalienable possession is also plausibly a result of factors other than alienability, as will be shown in Section 3. The fact that particular markers are not used for inalienable possession ultimately reflects the distribution of the source construction, which is also not used in inalienable possession contexts, for reasons unrelated to possession. This is confirmed by the fact that, in several languages, possession is or is not overtly marked depending on factors that cut across alienability distinctions and are related to the distribution of the constructions that gave rise to the relevant possessive markers, such as the animacy of the possessor. In addition, possessive markers may develop in inalienable, rather than alienable possession contexts, which goes against the idea that the use of overt marking is favored in the latter contexts as opposed to the former. This phenomenon too can be explained in terms of the distribution of the source constructions that give rise to individual markers.

In this scenario, as will be discussed in Section 4, the attested differences in the crosslinguistic frequency of different alienability splits can be a possibly combined result of several distinct factors, which need not be related to alienability and whose role remains to be investigated. These include the relative crosslinguistic frequency of the source constructions or the developmental processes that can give to possessive markers in alienable and inalienable possession contexts, as well as how often existing possessive markers are extended, retained or lost across these contexts.

These facts, as will be argued in Section 5, call for a new, source-oriented approach to alienability splits and recurrent crosslinguistic patterns in general (as discussed, for example, in Cristofaro 2013, 2017, 2019). In this approach, individual patterns and their relative crosslinguistic frequency are explained in terms of the properties of several distinct source constructions and diachronic phenomena that can give rise to the pattern and shape its crosslinguistic distribution over time, rather than particular synchronic properties of the pattern in themselves.

2 On the diachronic origins of overt marking

If particular alienability splits reflect general properties of alienable and inalienable possession, as assumed in classical explanations, these properties should play a role in the diachronic phenomena responsible for the presence of these splits in individual languages, for example, phenomena that give rise to overt marking and shape its distribution across alienable and inalienable possession. In what follows, a number of recurrent crosslinguistic developmental processes will be discussed that have been shown to give rise to the possessive markers involved in the alienability splits in question, that is, splits where overt marking is restricted to alienable possession, or zero marking is restricted to inalienable possession. These processes, it will be argued, provide no evidence that the relevant markers arise because of principles pertaining to properties of alienable and inalienable possession, such as the relative degree of conceptual contiguity between possessor and possessee, or the relative need to disambiguate the possessive relation.

2.1 Crosslinguistic developmental processes

Possessive markers are known to usually develop from elements originally used for other functions (see, for example, Heine 1997; Heine et al. 1991). This has been shown to also be the case for markers involved in splits where overt marking is restricted to alienable possession, or zero marking is restricted to inalienable possession. A recurrent source for such markers are elements that are originally part of a possessive construction, but do not encode a relation between possessor and possessee. Instead, these elements are used to refer to the possessee, or, in some cases, some other entity involved in the situation being described, and they take on the possessive meaning associated with the construction as a whole as their original meaning is bleached.

Sometimes, the source is a demonstrative element, that is, constructions of the type ‘X that/this Y’, where X and Y are a possessee and a possessor respectively, become ‘the X of Y’. While the exact syntactic structure of the source construction cannot always be reconstructed for individual languages, it is generally assumed that the original function of the demonstrative is to single out the possessee, either by modifying it, e.g., ‘that X (of) Y’, ‘(of) Y, that X’, or by denoting it in an appositive expression, that is, ‘X, that (of) Y’, ‘that (of) Y, X’ (see, for example, Eksell Harning 1980; Pennacchietti 1968 for Arabic and Semitic in general; Schuh 1983, 1990 for Chadic; Hoch 1997: 152 for Ancient Egyptian; and Yap et al. 2010 for Mandarin Chinese).

As illustrated in (3) and (4), for example, a number of Chadic languages display a pattern where possession is overtly marked for alienable nouns and some inalienable nouns ((3a), (4a)), and zero marked for other inalienable nouns ((3b), (4b); these nouns are sometimes quite limited in number). The possessive markers often agree in gender with the possessee, and are ultimately derived from demonstrative roots (Schuh 1977, 1983, 1990).

(3)

a.

k - tk -îi;	mii- tk -ja
goat-POSS.F.SG-2SG	mother-POSS-1.PL.EXCL
‘your goat; our mother’ (cf. -tk ‘this’)
(Schuh 1977: 72, 1990: 603)

b.

bâ-an i
co-wife-1SG
‘my co-wife’
(Schuh 1977: 42)											(Gashua Bade, West Chadic)

(4)

a.

ili	ma	lowoi
horn	POSS.M.SG	boy
“the boy’s horn” (cf. me ‘this’)

b.

il	kimne;	m	Miyim
horn	buffalo	wife	Miyim
‘buffalo’s horn; Miyim’s wife’
(Schuh 1983: 183–184)										(Kanakuru, West Chadic)

In Ancient Egyptian, pronominal possession is either zero marked or marked by the elements py/ty/ny, which agree with the possessee in gender and number and are derived from the demonstrative forms p/t/n. The use of these elements is strongly favored in alienable possession contexts such as the one in (5a), whereas inalienable possession contexts such as the one in (5b) strongly favor zero marking (Gardiner 2017; Kammerzell 2000).

(5)

a.

n y =sn	k -w
POSS.PL=3PL	sorcerer-M.PL
‘their sorcerers’
(Gardiner 2017: 645)

b.

sn=j
brother=1SG
‘my brother’
(Gardiner 2017: 645)								(Ancient Egyptian, Afro-Asiatic)

In Mandarin Chinese, the possessive marker de can be used for both alienable and inalienable possession. As illustrated by the contrast between (6a) and (6b), however, its use is disfavored by particular types of inalienable nouns, such as kin terms (Chappell and Thompson 1992; Li and Thompson 1981: 113–116; Shi and Li 2002). The marker is derived from the demonstrative element di illustrated in (6c) (Shi and Li 2002; Yap et al. 2010).

(6)

a.

xuéxiào	de	jiào
school	POSS	teach
‘school’s teaching staff’
(Li and Thompson 1981: 114)

b.

nǐ	(de)	mèimei
2SG	POSS	younger.sister
‘your younger sister’
(Li and Thompson 1981: 116)

c.

Di	shi	cang	chun	chu
that	be	store	spring	place
‘This is a warm and pleasant place.’ (Mo shan xi ci, c. AD 900)
(Shi and Li 2002: 8)						(Mandarin Chinese, Sinitic)

In other cases, the source of the possessive marker is a noun in an appositive relation with the possessee, which is used to specify some semantic property of the latter. For example, constructions of the type ‘X, food/drink/tree/thing/property Y’, where X and Y are a possessee and a possessor respectively, become ‘the X of Y’. This process has been found to be at the origin of the possessive classifiers attested in a number of Oceanic languages, as illustrated in (7)–(9) below (Lichtenberk 2018, among several others). The use of these classifiers is limited to alienable possession (see the cases of inalienable possession in (7e), (8b), and (9b), where no classifier is used), and individual classifiers generally occur in contexts directly related to the meaning of the source, so that different classifiers are used depending on whether the construction encodes general possession, possession of a food item, a drink item, a vegetable item, and so on, as can be seen from (7a)–(7d). Sometimes, for example, in (7d), the classifier is derived from the noun encoding the possessee, so that this noun is actually originally used in apposition to itself. The same noun may or may not take classifiers depending on whether the referent is alienably or inalienably possessed, as illustrated by the contrast between (8a) and (8b).

(7)

a.

ropi -qi	niu
POSS.CLASS-1SG.POSS	coconut
‘my coconut for drinking’ (cf. ropi ‘drink’)
(Ross 2001: 157)

b.

uma -qi	uri	eteba
POSS.CLASS-1SG	banana	SG
‘my banana plant’ (cf. uma ‘soil’ from Proto-Oceanic * quma ‘garden’)
(Ross 2001: 157)

c.

ane -qi	uri	eteba
POSS.CLASS-1SG	banana	SG
‘my banana to eat’ (cf. ane ‘food’)
(Ross 2001: 157)

d.

ai -qi	ai
POSS.CLASS-1SG	tree
‘my (tall) tree’
(possessive classifier derived from possessed noun, lit. ‘my tree, the tree’)
(Ross 2001: 157)

e.

tama-qi

father-1SG

‘my father’

(Ross 2001: 156)

(Mussau, Oceanic)

(8)

a.

a i -n	j
POSS.CLASS-3SG	bone
‘its bone’ (dog playing with a bone, possessive classifier derived from noun meaning ‘thing’)
(Ozanne-Rivierre 1976: 159)

b.

jeie-n

bone-3SG

‘its bone’ (anatomical)

(Ozanne-Rivierre 1976: 159)

(Iaai, Oceanic)

(9)

a.

ā	mwà	t ɛ -n
the	house	POSS.CLASS-his
‘his house’ (cf. tɛ ‘property, goods’)
(Moyse-Faurie and Ozanne-Rivierre 1983: 119)

b.

pūnı̄-n
head-his
‘his head’
(Moyse-Faurie and Ozanne-Rivierre 1983: 118)			(Cèmuhî, Oceanic)

DeLancey (1986) assumes that a similar process gave led to the use of the marker -gu for alienable possession in Newari. This marker, he submits, is derived from an element meaning ‘thing’, so that the source construction was ‘Y’s thing, X’. The marker (usually described as a nominalizer) may or may not used with the same nouns depending on whether they are alienably or inalienably possessed, as illustrated by the contrast between (10a) and (10b).^[4]

(10)

a.

ram-ya- gu	tasbir
Ram-GEN-NOMLZ	picture
‘Ram’s picture’
(a picture belonging to Ram; originally, ‘Ram’s thing, a picture’)
(DeLancey 1986: 6)

b.

ram-ya	tasbir
Ram-GEN	picture
‘a picture of Ram’
(DeLancey 1986: 7)										(Newari, Tibeto-Burman)

In yet other cases, the possessive marker is derived from an indefinite form. This pattern has been described for languages where particular referents can be alienably or inalienably possessed, and overt marking is used in the former case but not in the latter (Thompson 1996; Young and Morgan 1980). In Navajo, for example, the construction in (11a) involves alienable possession, ‘my milk (from the store)’. In this case, as can be seen from (11b), possession is encoded through a marker derived from the indefinite form ’a-, so that the meaning of the construction is, literally, ‘my something(’s)/somebody(’s) milk’. This reflects the fact that, in such cases, there actually are three entities involved, the possessor, the possessee, and the unknown source of the latter, which is encoded by the indefinite form (Young and Morgan 1980: 7). When the referent in question is inalienably possessed, as in (11c), ‘my milk (from my breast)’, on the other hand, the possessor and the source of the possessee overlap, and the marker is not used.

(11)

a.

she- ′a -be

1SG-3INDEF-milk

‘my milk’ (from a secondary source, as milk purchased at the store)

(Young and Morgan 1980: 7)

b.

′a -be

3INDEF-milk

‘something’s milk’

(Young and Morgan 1980: 7)

c.

shi-be
1SG-milk
‘my milk’ (from my own breasts)
(Young and Morgan 1980: 28)								(Navajo, Na-Dene)

Another recurrent source for the possessive marker are elements that encode a relation between two entities, but not one of possession. In such cases, the elements in question develop a possessive meaning as a result of context-driven inferences. A case in point are constructions that describe the location of some entity X in relation to some other entity Y, of the type ‘X at Y’, ‘X at the place/home (of) Y’. The relevant constructions can be reinterpreted as possessive ones, plausibly because an entity described in terms of its location with respect to some other entity can be inferred to belong to the latter (Claudi and Heine 1989; Heine 1997: chap. 2). As a result, a locative element within the construction becomes a possessive marker, ‘the X of Y’. This is illustrated in (12)–(14) below for Ngiti, Acholi, and Kabyie. In these languages, the relevant markers are limited to alienable possession, and they may or may not be used with the same noun depending on whether the referent is alienably or inalienably possessed. This contrast is illustrated by (13a) and (13b), as well as by (14a) and (14b).^[5]

(12)

a.

àba	bhà	dzalí-nga
father	POSS	courtyard-NOMLZ
‘my father’s courtyard’ (cf. ibhà ‘at home’)
(Kutsch Lojenga 1994: 154)

b.

ts -du
hand-1SG.INAL.POSS
‘my hand’
(Kutsch Lojenga 1994: 202)									Ngiti (Nilo-Saharan)

(13)

a.

wiìc	pā	lát n
head	POSS	child
‘the head of the child’ (e.g., a sheep’s head belonging to the child, cf. pā ‘house’)
(Claudi and Heine 1989: 5; Crazzolara 1955: 47)

b.

wiìc	lát n
head	child
‘the child’s (own) head’
(Claudi and Heine 1989: 5; Crazzolara 1955: 47) (Acholi, Nilo-Saharan)

(14)

a.

kólú	tɛ́	píya
blacksmith	POSS	children
‘the blacksmith’s children’ (typically those living in his compound but not his own, cf. tɛ́ ‘home’)
(Heine et al. 1991: 148)

b.

kólú	píya
blacksmith	children
‘the blacksmith’s (own) children’
(Heine et al. 1991: 148)							(Kabyie, Niger-Congo)

3 Explaining the distribution of individual markers

If the development of particular possessive markers is indeed independent of the properties of alienable and inalienable possession, as argued in the previous section, the question arises as to why the distribution of these markers is constrained by alienability, in the sense that they are restricted to alienable possession, or are not used for some cases of inalienable possession. The literature on the origins of these markers is based on a general (if often implicit) assumption that this is a result of a diachronic scenario where they develop in alienable possession contexts and are either not extended or only partially extended to inalienable possession ones (rather than being initially used and then lost in the latter contexts: see, for example, Claudi and Heine 1989; Creissels 2001; Koptjevskaja-Tamm 1996; Schuh 1983). In this case, explanations for the distribution of individual markers should ultimately account for why they develop in alienable, rather than inalienable possession contexts. In what follows, it will be shown that this is a direct reflection of the distribution of the source construction, which is usually not used in inalienable possession contexts due to various types of incompatibilities unrelated to possession.

3.2 Some further distributional evidence

Some further facts about the crosslinguistic distribution of possessive markers also support the idea that this distribution depends on the distribution of the source constructions that give rise to individual markers, rather than general properties of alienable and inalienable possession.

In several cases, the use of specific markers is influenced by semantic factors that cut across alienability distinctions. This results in patterns where the same type of possession (either alienable or inalienable) is either overtly marked or not in different contexts. Additionally, some instances of inalienable possession are overtly marked, while certain cases of alienable possession are not. These patterns contrast with both the assumption that overt marking depends on general properties of alienable and inalienable possession and the assumption that overt marking inherently favors one possession type over the other.

The first assumption implies that all instances of alienable or inalienable possession should be treated uniformly in terms of overt marking. The second assumption implies that if overt marking is employed for inalienable possession at all, it should be employed in all cases of alienable possession as well. However, these patterns find their explanation in the distribution of the source constructions that give rise to the relevant possessive markers.

In Old French, for example, the possessive marker de was used in contexts where the possessor was inanimate (Palm 1977). Inalienable possession with inanimate possessors was overtly marked, as illustrated in (15a), whereas inalienable and alienable possession with animate possessors were not, as illustrated in (15b) and (15c).

(15)

a.

ele	vint	au	mur	del	castel
she	came	to.the	wall	of.the	castle
‘She came to the wall of the castle’
(Aucassin et Nicolette XVI)										(Old French, Romance)
(Palm 1977: 108)

b.

Les	os	Eumon	et	Agolant
the	bones	Eumon	and	Agolant
‘The bones of Eumon and Agolant’
(La chanson d’Aspremont, 4391)
(Palm 1977: 43)

c.

La	maison	l’	emperor
the	house	the	emperor
‘The house of the emperor’
(Le roman de la rose, 1038)
(Palm 1977: 54)										(Old French, Romance)

This distribution is consistent with the origin of the possessive marker, as de originally encoded motion away from some point of origin (‘from’, ‘away from’, ‘out of’: Heine 1997, among others), and points of origin are usually inanimate entities.

In Mandinka (see also Creissels, this volume), possession may or may not be overtly marked for different nouns involving the same type of relation between possessor and possessee, for example, as illustrated in (16), ‘slave’ versus ‘master’, ‘wife’ versus ‘husband’, or ‘pupil’ versus ‘teacher’.

(16)

a.

à	la	jò o
3	POSS	slave
‘his slave’
(Creissels 2001: 446)

b.

à	màario
3	master
‘his master’
(Creissels 2001: 446)

c.

à	la	mùsoo
3	POSS	wife
‘his wife’
(Creissels 2001: 446)

d.

à	kèe
3	husband
‘her husband’
(Creissels 2001: 446)

e.

à	la	kàrandi o
3	POSS	pupil
‘his pupil’
(Creissels 2001: 446)

f.

À	kàrammoo
3	teacher
‘his teacher’
(Creissels 2001: 446)											(Mandinka, Niger-Congo)

This pattern is explained by the fact that the possessive postposition, la, was originally a locative postposition (ultimately derived from a non meaning ‘mouth’, ‘opening’, ‘edge’). The postposition is used for possessees that can be described as being located at the possessor’s place, for example, slaves, wives, or pupils, as these live at the master’s, husband’s or teacher’s place respectively, but not masters, husbands, or teachers, as these do not usually live at the slave’s, wife’s, or pupil’s place (Creissels 2001; Grégoire 1984).

In Ngiyambaa, kinship relations are marked by the dative affix -gu, also used for alienable possession, goals, and purpose. Other cases of inalienable possession, on the other hand, are zero marked (Donaldson 1980). This is consistent with the fact that, in kinship relations, the possessee can be conceptualized as performing a specific function for the possessor, for example, being a parent, child, or sibling to someone. This is, in fact, paralleled by English expressions such as ‘mother/brother to John’, ‘secretary to the minister’, or ‘queen to Henry the Eight’.

(17)

a.

bura:y- gu	ba:ba:
child-DAT	father
‘child’s father’
(Donaldson 1980: 230)

b.

adhu	giyanhddha-nha	idji-la:	winar- gu -dhi	miri-dji
I	fear-PRES	this.CIRC-EST	woman-DAT-CIRC	dog-CIRC
‘I am frightened of this woman’s dog’
(Donaldson 1980: 107)

c.

ani-la:	mayi	wa a un
that-EST	person	waaun
‘that person’s waaun (spirit)’
(Donaldson 1980: 231)								(Ngiyambaa, Australian)

Some languages also display patterns where the use of particular markers is restricted to inalienable possession, whereas alienable possession is encoded by different markers. In Faroese, for example, the dative preposition til ‘to’ is mostly used for kinship relations, whereas alienable possession is encoded by the locative preposition hja ‘at’ (Stolz et al. 2008: 231).

(18)

a.

mamma	til	Kjartan
mum	to	Kjartan
‘Kjartan’s mum’
(Stolz et al. 2008: 223)

b.

eg	havi	[…]	gamla	gandastavin	hjá	Charlie
I	have:1SG		old:ACC	wand:DEF	at	Charlie
‘I have got Charlie’s old wand’
(Stolz et al. 2008: 222–223)							(Faroese, North Germanic)

From a synchronic perspective, such patterns are entirely consistent with the hypothesis that there are general principles that favor the use of overt marking for alienable rather than inalienable possession. This hypothesis allows for overt marking to be used for inalienable possession as well, as long as it is also used for alienable possession. Diachronically, however, one possible scenario for the emergence of these patterns (to be investigated in each particular case) is one where distinct developmental processes give rise to possessive markers in alienable and inalienable possession contexts, respectively. If there were some general preference for the use of overt marking for alienable possession, one would expect new possessive markers to originate in alienable possession contexts and possibly be extended to inalienable possession ones, rather than originating in the latter contexts.^[8] The fact that some marker develops in inalienable possession contexts, however, may be a result of the distribution of the source construction. For example, as observed earlier in regard to Ngiyambaa, the use of dative markers for kinship relations is consistent with the semantic nature of these relations, so it is to be expected that dative markers may be used for these relations even if they are not used for alienable possession, as in Faroese.

4 Accounting for the crosslinguistic frequency of different alienability splits

The idea that the distribution of possessive markers across alienable and inalienable possession reflects the distribution of the source construction provides no specific explanation for the attested differences in the crosslinguistic frequency of various alienability splits. As described earlier, splits where overt marking is limited to alienable possession or zero marking is confined to inalienable possession are significantly more frequent than ones where overt marking is restricted to inalienable possession or zero marking is restricted to alienable possession. These frequency differences are traditionally regarded as evidence for principles favoring the former splits because they comply with general properties of alienable and inalienable possession.

Diachronically, however, the crosslinguistic frequency of different alienability splits will be a result of the frequency of those splits arising in particular languages, or being retained across different generations of speakers once they are in place in a language. As illustrated in Figure 1, alienability splits can arise as zero marking is replaced by overt marking either in alienable possession contexts but not in inalienable possession ones (diagram (i)), or vice versa (diagram (ii)). This may be a result of the development of a new possessive marker, or the extension of an existing one (for example, in a situation where alienable and inalienable possession are both zero or overtly marked in different cases, some marker may be extended to the cases of alienable possession that are zero marked, leading to zero marking becoming restricted to inalienable possession). Another possible developmental path for alienability splits is the replacement of overt marking with zero marking either in inalienable possession contexts but not in alienable possession ones (diagram (iii)), or vice versa (diagram (iv)), as a result of the loss of existing possessive markers.

Figure 1:

The development of alienability splits (Ø = zero marking; X = overt marking).

Existing splits will be retained or lost depending on whether overt marking or zero marking continues to be used in their respective contexts, as illustrated in Figure 2. Overt marking will be lost when existing possessive markers are replaced by zero marking, either in alienable possession contexts (diagram (i)) or in inalienable possession contexts (diagram (ii)). Zero marking will be lost as a result of the development of a possessive marker or the extension of an existing one, either in inalienable possession contexts (diagram (iii)) or in alienable possession contexts (diagram (iv)).

Figure 2:

The loss of alienability splits (Ø = zero marking; X = overt marking).

These facts mean that any differences in the crosslinguistic frequency of different alienability splits will be a result of how frequently possessive markers develop in alienable and inalienable possession contexts, or are retained, lost or extended across these contexts. The relative crosslinguistic frequency of these phenonena is logically independent of the properties of alienable and inalienable possession, and should therefore be investigated in its own right.

In particular, one possible diachronic scenario that may explain the observed differences in the crosslinguistic frequency of different alienability splits is one where possessive markers develop more frequently in alienable possession contexts compared to inalienable possession ones. This results in zero marking being more frequently replaced by new possessive markers in the former contexts than in the latter. To achieve this, two possibilities exist: either the source constructions giving rise to possessive markers are more frequently employed in alienable possession contexts than in inalienable possession ones, or the developmental processes whereby these constructions actually give rise to such markers are more frequent in the former contexts than in the latter.

As discussed in Section 3, whether a source construction is used in alienable or inalienable possession contexts can be linked to properties of that construction that are independent of possession. In such cases, alienability will not influence the crosslinguistic distribution of individual constructions across these contexts. Similarly, if specific developmental processes are unrelated to alienability, as argued in Section 2, this factor won’t affect how these processes are distributed across alienable and inalienable possession contexts.

Consequently, understanding the relative crosslinguistic distribution of particular source constructions or developmental processes across alienable and inalienable possession contexts should rely on an analysis of the properties of those constructions and processes themselves, rather than general properties of alienable or inalienable possession.

As detailed earlier, for example, in several cases where a possessive marker is not used for inalienable possession, this is explained by the fact that the marker is derived from an appositive or a locative construction relatively incompatible with inalienable possession contexts, of the type illustrated in (7)–(10) and (12)–(14) above. Other types of appositive or locative constructions, however, could in principle be used in such contexts, e.g., ‘the mother, the parent of John’, ‘the arm, the body part of John’, ‘the heart in John’, or ‘the cover on the book’. These constructions could give rise to possessive markers used for, or possibly restricted to inalienable possession.

If possessive markers consistently develop from the former type of appositive or locative constructions but not from the latter, this can be because the former type of constructions are more frequent crosslinguistically, or because they more frequently give rise to possessive markers. Yet, the relative crosslinguistic frequency of different locative or appositive constructions and that of the developmental processes in which they are involved need not be related to alienability, so these phenomena should be investigated independently.^[9]

Another possible diachronic scenario that may explain why specific alienability splits are significantly more common than others crosslinguistically involves changes in the distribution of existing possessive markers, either through extension or loss. For instance, it’s possible that existing markers are more frequently extended from inalienable to alienable possession contexts than the reverse, resulting in zero marking being more frequently replaced by these markers in alienable possession contexts than in inalienable possession ones. Alternatively, existing markers might be more commonly lost in inalienable possession contexts than in alienable possession ones, leading to zero marking emerging more frequently in the former than the latter.

The extension, retention and loss of existing markers are much less investigated than the development of new ones, and they are arguably independent of the origins of individual markers. It is possible, then, that these phenomena are related to general properties of alienable and inalienable possession, such as the relative degree of conceptual contiguity between possessor and possessee, the relative need to disambiguate the relation between the two, or the relative degree of entrenchment of an existing possessive construction, as mentioned in Section 3.1 in regard to extension.

These properties, for instance, could favor the extension of an existing marker to alienable possession contexts, or its loss in inalienable possession ones. This resembles the concept of natural selection in biological evolution: the distribution of genetic traits in a population is influenced by a preference for certain traits, driven by their adaptiveness to the environment, regardless of the processes that initially gave rise to those traits. Nevertheless, this idea should be thoroughly investigated through an examination of actual instances of extension, retention, and loss of existing possessive markers crosslinguistically.

The two scenarios just outlined have not really been investigated so far, so there is no evidence about which of the two is actually the case. The data discussed in this paper show that possessive markers recurrently develop in alienable possession contexts and fail to be extended to at least some inalienable possession ones. This, however, does not tell anything about the relative crosslinguistic frequency of these phenomena compared to the development of possessive markers in inalienable possession contexts, the extension of existing markers from these contexts to alienable possession ones, or the loss of existing markers in both types of contexts.

Ultimately, differences in the crosslinguistic frequency of different alienability splits cannot be taken as evidence for principles that favor or disfavor these splits because of general properties of alienable and inalienable possession. These differences will be a possibly combined of several different phenomena that need not be related to alienability, namely the differential development, extension, retention or loss of possessive markers across alienable and inalienable possession contexts. These phenomena remain largely unexplored, and their respective role should be disentangled in order to account for the relative crosslinguistic frequency of individual splits.

5 Concluding remarks

Classical explanations of alienability splits in the use of overt and zero possessive marking imply that this use reflects principles of optimization of grammatical structure, either in the sense that it iconically matches the relative degree of conceptual contiguity between possessor and possessee, or in the sense that it economically matches the relative need to disambiguate the possessive relation. These principles are assumed to be responsible both for the distribution of overt and zero marking across alienable and inalienable possession in particular languages, and for the relative crosslinguistic frequency of different alienability splits.

Diachronic evidence about the origins of the relevant alienability splits shows, however, that the possessive markers involved in these splits recurrently arise through the reinterpretation of pre-existing elements originally performing other functions in various types of source constructions. These constructions are all associated with the notion of possession, either inherently or contextually, so they recurrently give rise to possessive markers from one language to another.

The development of such markers, however, is a result of metonymization processes triggered by the contextual co-occurrence of the source element and the possessive meaning, independently of the assumed optimization principles. The distribution of individual markers across alienable and inalienable possession ultimately reflects the distribution of the source construction, also independently of such principles. When particular markers are not used for inalienable possession, the source construction is one not used in inalienable possession contexts, for reasons unrelated to possession. When the source construction is used in particular alienable or inalienable possession contexts, the resulting possessive markers are used in those contexts, irrespective of whether possession is zero or overtly marked elsewhere in the language.

This is in line with standard accounts of grammaticalization processes and processes of constructional reinterpretation in general, in that the development and initial distribution of the constructions resulting from such processes are usually assumed to be driven by the properties of particular source constructions and their contexts of use. Similar source constructions and developmental processes will recurrently give rise to constructions with similar distributions crosslinguistically, but this is independent of the synchronic properties of the resulting grammatical patterns (see, for example, Bybee 2006; Bybee et al. 1994).

This analysis makes it possible to account for a number of synchronic patterns that go against the assumed optimization principles, for example, patterns where the same possession type (either alienable or inalienable) is not always treated in the same way in terms of overt or zero marking, ones where overt marking is used for inalienable possession while zero marking is used for at least some cases of alienable possession, or ones where possessive markers may have originated in inalienable, rather than alienable possession contexts.

This analysis also provides an explanation for the notorious difficulties in defining crosslinguistically valid grammatical categories of alienable and inalienable possession (Chappell and McGregor 1996; Nichols 1988, among many others). Possessive markers are not used for the same range of alienable or inalienable nouns crosslinguistically, which makes it difficult to define some general category of alienable or inalienable possession that could account for the synchronic distribution of such markers in different languages. This is naturally explained by the fact that individual markers continue the distribution of the source, and are gradually extended to more and more contexts over time. Markers with different distributions will be derived from different sources, or will be at different stages of extension from one possession context to another.

In this analysis, the explanation of the emergence of particular alienability splits is distinct from the explanation of why some splits are significantly more frequent than others crosslinguistically. Particular splits can be a result of the properties of multiple source constructions and their contexts of use, rather than general properties of alienable and inalienable possession. The attested differences in the crosslinguistic frequency of different splits will reflect differences in the frequency of multiple source constructions, multiple developmental processes whereby these constructions give rise to possessive markers, or multiple processes of extension, retention or loss of these markers once they are in place in a language. The latter differences may or may not be related to alienability, and they remain to be investigated.

In accordance with prior diachronically oriented typological research (Bybee 2008; Claudi and Heine 1989; Creissels 2001; Aristar 1991), these facts call for a novel, source-oriented approach both to alienability splits in the use of overt and zero possessive marking and recurrent crosslinguistic patterns in general (for detailed discussion, see Cristofaro 2013, 2017, 2019; Schmidtke-Bode and Grossmann 2019). This approach is similar to Evolutionary Phonology and other diachronically oriented approaches to crosslinguistic phonological patterns (Blevins 2004; Ohala 1993, 2003). In language typology, recurrent crosslinguistic patterns are generally assumed to be a result of recurrent diachronic phenomena that give rise to the pattern and shape its crosslinguistic distribution (Bybee 1988, 2008, 2009; Givón 1975, 1979; Greenberg 1969, 1978). Yet, explanations for individual patterns and their relative crosslinguistic frequency are usually framed in terms of synchronic properties of these patterns, such as whether they comply with some assumed principle of optimization of grammatical structure. This implies that the development, retention or loss of particular patterns crosslinguistically will ultimately depend on their synchronic properties. These phenomena, however, can be a result of several distinct diachronic factors, often unrelated to the synchronic properties of the pattern. Explanations of individual patterns and their crosslinguistic frequency should then be informed by an understanding of these factors, rather than particular synchronic properties of the pattern in themselves.