In a recent article, Harves & Kayne (2012: 120-121) proposed the following universal generalization:
(1) “All languages that have a transitive verb corresponding to need have an overt verb expressing predicative possession […] such that the possessor has nominative case and the possessee is a direct object (with accusative case and no preposition).”
The authors introduce the terms “H-languages” an “B-languages” for classifying languages as regards to their predicative possession constructions. H-languages have a transitive verb meaning ‘to have’ where the possessor is treated as the agent of the verb, while B-languages use a construction with a verb ‘to be, to exist’ and the agent is marked with oblique case.
It appears from their typological survey that a correlation exists between the distribution of H- and B-languages and that of languages having a transitive verb to express the meaning corresponding to English ‘need’ (let us call them N-languages) and of languages that use instead various constructions, such as for instance an impersonal verb with the “needer” marked with an oblique case, or a combination of the possessive construction with a noun meaning ‘to have’ (O-languages).
According to the authors, while there are O-languages which are at the same time B- or H-languages, there are no examples of N-languages that would be at the same time B-languages.
Harves & Kayne (2012: 129-30) discuss one apparent counterexample to (1): Finnish. In this language, ‘need’ is expressed with a transitive verb (2), while in the possessive construction the possessor is in the oblique case. They however argue that Finnish is not a typical B-language, in that the possessee is assigned accusative case (3) rather than nominative as in other B-languages:
(2) Minä tarvitse-n sinu-t.
I.nom need-1sg you-acc
‘I need you.’
(3) Minu-lla on häne-t.
I-adess be.3sg he-acc
‘I have him.’
The authors propose an amendment to (1) in order to account for the Finnish example:
(4) “All languages that have a transitive verb corresponding to ‘need’ are languages that have an accusative-case-assigning verb of possession.”
I will try to argue that even with this amendment, counterexamples to (4) can still be discovered, in particular in the Japhug language (Sino-Tibetan, Sichuan, China). Japhug, as other languages of the Rgyalrong subgroup (see Sun 2003), has ergative alignment on nouns: the agent of an transitive verb is marked with the ergative clitic kɯ, while the only argument of an intransitive verb and the patient of a transitive verbs are unmarked.
(5) [qachɣa nɯ] pɯ-si
fox det aor-die
‘The fox died.’ (The Tiger, 45)
(6) [srɯnmɯ nɯ kɯ] [nɯ lɤ-kɯ-nɯna nɯ] pjɤ-mto
rakshasi det erg distal aor-nmlz:S/A-rest top evd-see
‘The rakshasi saw those who had had a rest.’ (Nyima Vodzer 2.6)
However, the absence of a nominative/accusative alignment on arguments is not an obstacle to testing principles (1) and (4) in this language. Like most polysynthetic languages, Japhug clearly marks transitivity in the verb morphology, by means of various affixes (such as the transitive aorist direct 3>3 prefix a-) or morphological processes (stem alternation in direct sg>3 forms).
Determining whether a language with ergative alignment on NP fits (4), we need to convert (4) to a formulation immune of accusativity bias, using Dixon’s (1994) S (sole argument of an intransitive verb), A (agent of intransitive verb) and O (patient of transitive verb), and also need to take into account the fact that languages can have several constructions expressing possession or ‘need’.
(7) All languages that have at least one construction corresponding to ‘need’ in which ‘need’ is a transitive verb (whose A is the person needing and O the entity needed) are languages that have at least one possessive construction in which the possessed entity has the same morphosyntactic status as the O of a transitive verb.
Let us now examine whether (7) holds for Japhug. In section A, I present the construction meaning ‘need’ and in section B the possessive constructions.
A. Japhug constructions expressing the meaning ‘to need’
To express the meaning corresponding to English ‘need’ in Japhug, three constructions are attested. First, one may use the intransitive stative verb ʁzi ‘need’. With this verb, the needing entity is an adjunct, marked with the genitive ɣɯ, and the needed entity is the S of the verb with zero marking:
(8) ki ɯʑo ɣɯ ɲɯ-ʁzi
this he gen const-need
‘He needs this.’ (elicitation)
Second, one can use ra ‘have to’, which can be translated as ‘need’ in some cases:
(9) aʑɯɣ ɯ-ku nɯ ma mɤ-ra
1sg:gen 3sg.poss-head top apart.from neg-npst:need
‘I only need (want) its head.’ (The Flood, 6)
If Japhug only had these constructions for expressing the meaning corresponding to English ‘need’, we would have to consider it as an O-language. However, another construction involving the same verbal root is also attested: naʁzi ‘to need’, a transitive verb whose A (marked with ergative) is the needing entity and whose O is the needed entity:
(10) nɯnɯ ndɤre, wuma pɯ-naʁzi-t-a ko ri
this conj very pst.ipf-need-pst-1sg assertive but
‘This, although I needed it.’ (Gesar, 65)
This verb is the regular tropative form of ʁzi. The tropative nɤ-/na- is a derivation that creates a transitive verb out of a stative intransitive one, and whose O corresponds to the S of the original verb. Although similar to causative from the point of view of valency change, the semantics of this derivation is quite different: for a stative verb meaning X, the meaning of the transitive verb is ‘to consider as / to find X’ (Jacques 2012).
Japhug can be considered as both an O- and an N-language, but according to (7), the existence of a verb such as naʁzi would predict that a transitive verb meaning ‘have’, or a peculiar construction where the possessee is treated as the O of a transitive verb, exists in Japhug.
B. Japhug predicative possession constructions
Two predicative possessive constructions are found in Japhug. First, we find a B-type construction, with the possessor marked with the genitive, and the possessed entity is the S. The existential verbs tu and ɣɤʑu ‘be there’ can be used in this construction:
(11) ɯ-mɲaʁ χsɯm pjɤ-tu
3sg.poss-eye three evd.pst-be.there
‘He had three eyes.’ (Gesar, 64)
Another (considerably rarer) possessive construction exists in the language, with the intransitive verb aro ‘to possess’:
(12) aʑo tɤ-rte aro-a
I indef-hat npst:have-1sg
‘I have a hat.’
Aro is however a morphologically intransitive verb. The possessor (S) agrees with aro in person and number as can be seen in (12), and cannot be used with the ergative kɯ. Aro is not the only such verb with intransitive morphology but compatible with two NPs. The critical question to determine whether the existence of this verb aro threatens generalizations (4) or (7) is the status of this external NP: should it be analyzed as an argument, or as mere adjunct?
Case marking here is not informative, as the external NP is unmarked, like the S and the O (in this head-marking language, even locative NPs can appear without any postposition or case marker). Harves and Kayne could argue that the construction in (12) is of the same category as the Finnish construction in (3), and that this case is already accounted for by generalization (4). However, there is evidence showing that the external NP here is an adjunct and not an argument.
In Japhug, as in the related Tshobdun language (see Sun’s (2006) study of relativization), the core arguments S, A and O can be relativized by means of the nominalizing prefixes kɯ- and kɤ-.
The nominalized form kɯ-ɤro however can only mean ‘the one who possesses’, never ‘the thing which is possessed’, as in the following example:
(13) iʑo kɯrɯ tɕheme mɤ-kɯ-ɤro maka me tɕe
we Tibetan woman neg-nmlz:S-possess at.all npst:not.to.be.there conj
‘Among us, Tibetan women, there is no one who does not possess (such a belt).’
It is therefore clear that the external NP cannot be analyzed as having the same morphosyntactic status as the O of a transitive verb, and Japhug is therefore a counterexample to (7).
One could propose the following reformulation of (4), instead of (7):
(14) All languages that only have constructions corresponding to ‘need’ in which ‘need’ is a transitive verb (whose A is the person needing and O the entity needed) are languages that have at least one possessive construction in which the possessed entity has the same morphosyntactic status as the O of a transitive verb.
With such a formulation, Japhug would not constitute a counterexample anymore, since there are constructions expressing the meaning ‘need’ in this language which are not based on a transitive verb. However, it would also make it more difficult to test Harves and Kayne’s generalization.
I conclude that while Harves and Kayne’s generalization deserves to be tested against a larger sample of languages, it is likely that other (and perhaps clearer) counterexamples will be found.