Many languages have different adpossessive (= adnominal possessive) constructions for inalienable possessed nouns (= body-part or kinship nouns) and all other nouns. For example, Maltese has id Pietru ‘Pietru’s hand’ with no marker when a body-part is possessed, but il-ktieb ta’ Pietru [the-book of Pietru] with a possessive preposition when an alienable noun is possessed.
That there are highly general cross-linguistic trends in these constructions has been known at the latest since Nichols (1988), and even earlier Haiman (1983) proposed an explanation that has been widely adopted: According to Haiman, the zero-coding of inalienable adpossession is due to iconicity of distance: In body-part and kinship relationships, the possessor and possessed are conceptually very close, and this is reflected in the coding. By contrast, in alienable relationships, there is greater conceptual distance, which corresponds to formal distance (a possessive marker occurs between adpossessor and possessed noun).
This seemed a good explanation to me for a long time, but after reading the chapters in Chappell & McGregor’s (1996) edited volume, I started having doubts. The authors there describe a number of cases where there is a marker that does not occur between the two nouns, but in peripheral position (e.g. Koyukon se-tel-e’ [1SG-sock-POSSD] ‘my socks’) – what these share with the Maltese type is that there is longer coding in the alienable construction, but it does not lead to greater distance. There are also languages where the possessor forms are longer, without there being a clear marker that occurs between both elements (e.g. Juhoan mì tjù ‘my house’, vs. m bá ‘my father’). And perhaps most strikingly, some languages have a special marker for lack of possession: For example, in Yucatec Maya, there is an antipossessive marker -tsil that must be used when a kinship term is used without a possessor (e.g. in tàatah ‘my father’ vs. tatah-tsil- ‘father’).
Thus, I wrote a paper (first conceived of in 2006, for a plenary talk at the Syntax of the World’s Languages in Lancaster, organized by Anna Siewierska) where I propose an explanation in terms of frequency-induced predictability. My paper’s title is “Explaining alienability contrasts in adpossessive constructions: Predictability vs. iconicity” (Haspelmath 2017).
The gist of the proposed explanation is:
Adpossessive constructions have the less coding, the more expected/predictable the presence of an adpossessor is.
The universal claim can be formulated in the most general terms on the basis of a scale of possessor prominence:
The possessor-prominence scale:
inalienable nouns – alienable nouns – super-alienable nouns
The universal claim goes as follows:
Universal of adpossessive coding:
The further to the left on the possessor-prominence scale a noun is located, the shorter (and hence more bound) is the marking of the possessive construction, and the longer (and hence less bound) is the marking of the nonpossessed occurrence. Conversely, the further to the right on the possessor-prominence scale a noun is located, the longer is the marking of the possessive construction, and the shorter the marking of the nonpossessed occurrence.
By the time I finally wrote up this paper (which took a while to be published, also because it was rejected after revision by the first journal I sent it to), I had thought that the “predictability vs. iconicity” in the title was a not much more than a strawman, because Haiman’s (1983) proposal was by now almost 35 years old. However, it kept being cited (e.g. by Aikhenvald 2013: 8-9, in a prominent volume on the typology of possessive constructions), so I kept the “iconicity” contrast (which I had already argued against in an earlier 2008 paper).
Now quite recently, the journal Cognitive Linguistics published a paper by Simon Devylder, claiming that “Diagrammatic iconicity explains asymmetries in Paamese possessive constructions“.
So is iconicity the correct explanation after all? Devylder makes this claim on the basis of a detailed study of Paamese (an Oceanic language of Vanuatu). He finds that in Paamese, a notion of “functional distance“ is useful for explaining why the short (markerless, or „direct“) construction is used with consanguineal kin terms (mother, sister, grandson, etc.), but not with affinal kin terms (wife, brother-in-law, etc.). In Paamese society, people have a much closer relationship with their consanguineal kin, and this seems to be reflected in the grammatical coding. For body parts, there is a similar contrast between external, i.e. fully experienced body parts (head, nose, armpit), and internal body parts (heart, kidney, bladder) – the latter require the use of the longer (“indirect“) possessive construction. Again, this could be because of iconic matching between linguistic distance and “embodied experience of distance“.
Devylder compares his explanation with mine (in terms of frequency) and thinks that the frequency explanation cannot explain the Paamese situation. This seems to be true, at least for the two types of kinship terms: A noun such as ’daughter-in-law’ is not significantly less often possessed than a noun such as ’daughter’ (Devylder, p. 24). For external vs. internal body parts, I am not so sure – it could be, for example, that kidneys and lungs are typically talked about in contexts where the animal from which they were removed is no longer thought about (“Today we’ll have kidney for dinner“), so that they are often non-possessed.
But it is important to note that my explanation in terms of frequency-based predictability is an explanation of a universal tendency. I am not claiming that it should have an effect on every language, and that it explains everything in all the languages that show its effects. This may be a frustrating aspect of my work from the point of view of the fieldworker: It does not make claims that could be refuted by individual languages (though of course detailed language-particular work often provides a lot of food for thought).
Thus, in my view, Devylder has not provided an argument for iconicity as a general explanation, because his claims are restricted to Paamese. We simply do not know whether consanguineal kin terms and affinal kin terms are generally treated differently in the world’s languages. This may be, and if it is true that affinal kin terms show a general tendency to be coded like alienable nouns, even though they are not less often possessed, then this would be an argument that the frequency-based explanation is not sufficient. (For body-part terms, the same applies, though here it will be much harder to come up with frequency counts that are representative, because internal body parts are used rarely and may not occur often enough in most texts that reflect traditional cultures, i.e. cultures that did not have the possibilities of modern medicine.)
It could be that the iconicity explanation is correct for Paamese, but I am generally skeptical that highly general explanations are suitable for language-particular phenomena. While Paamese kinship organization may be very specific as well, the idea of „functional distance“ is not so specific, and it lets us expect to find very similar patterns in many other languages. It’s not so clear that this is the case. But I also cannot exclude it, so further comparative grammatical research, and further cross-linguistic corpus research, is clearly needed. Iconicity and predictability could both be motivating factors, and it could be that we will eventually be able to show that they are both at work. I am looking forward to further work that addresses this issue.
Aikhenvald, Aleksandra Y. 2013. Possession and ownership: A cross-linguistic perspective. In Alexandra Y. Aikhenvald & R. M. W. Dixon (eds.), Possession and ownership: A cross-linguistic typology, 1–64. Oxford: Oxford University Press.
Chappell, Hilary & William McGregory (eds.). 1996. The grammar of inalienability: A typological perspective on body part terms and the part-whole relation. Berlin: Mouton de Gruyter.
Devylder, Simon. 2018. Diagrammatic iconicity explains asymmetries in Paamese possessive constructions. Cognitive Linguistics 29(2). 313–348. doi:10.1515/cog-2017-0058.
Haiman, John. 1983. Iconic and economic motivation. Language 59(4). 781–819.
Haspelmath, Martin. 2008. Frequency vs. iconicity in explaining grammatical asymmetries. Cognitive Linguistics 19(1). 1–33.
Haspelmath, Martin. 2017. Explaining alienability contrasts in adpossessive constructions: Predictability vs. iconicity. Zeitschrift für Sprachwissenschaft 36(2). 193–231. doi:10.1515/zfs-2017-0009.
Nichols, Johanna. 1988. On alienable and inalienable possession. In William Shipley (ed.), In honor of Mary Haas: From the Haas Festival Conference on Native American linguistics, 475–521. Berlin: Mouton de Gruyter.
If evidence from individual languages does not falsify the frequency explanation, what (if anything) would? What evidence could someone give you to make you think that it is not correct?
Universal frequency trends make predictions about universal coding trends. If there is no universal trend, the phenomenon may well be a simple historical accident. Language-particular frequencies may predict speaker behaviour, but they cannot predict language systems, which are alsways the result of a long historical trajectory.
Yes, I agree! This is definitely worth following up on.
Indeed, “we simply do not know whether consanguineal kin terms and affinal kin terms are generally treated differently in the world’s languages.” But we do observe the same difference in, say, Siwi Berber: some consanguineal kin take the “inalienable” construction (eg əṃṃe-s “his mother”, wəltma-s “his sister”), but no affinal kin do (eg taṭəggalt nn-əs “his female in-law”, aʕdil nn-əs “his wife’s sister’s husband”). Finding the same phenomenon in two such widely separated languages suggests that it is unlikely to be highly language-particular, and that it would be worth checking this systematically.