Portal de Periódicos da CAPES

Transitive NEED Does Not Imply Transitive HAVE: Response to Harves and Kayne 2012

2014; The MIT Press; Volume: 45; Issue: 1 Linguagem: Inglês

10.1162/ling_a_00151

1530-9150

Anton Antonov, Guillaume Jacques,

Linguistic Variation and Morphology

In a recent article, Harves and Kayne (2012:120–121) propose the following universal:The authors' argument for the validity of (1) is based upon a typological survey of about 50 languages. They introduce the terms H-languages and B-languages in order to classify languages from the viewpoint of their predicative possessive constructions. H-languages have a transitive verb 'to have' whose subject is the possessor and whose object is the possessee. B-languages, on the other hand, use a construction with a copular or existential verb in which the possessor is marked with an oblique case and the possessee is treated as the subject of an intransitive verb.This terminology presents two major problems.First, it neglects the fact that one language can have several competing constructions to express possessive predication. This is for example the case of Latin, which according to table 1 (adapted from Harves and Kayne 2012:126) is an ''H-language,'' whereas the most common predicative possessive construction in this language is of the type normally found in a ''B-language.'' It would thus be more appropriate to talk of B- and H-constructions rather than of B- and H-languages.Second, the classification of possessive predicative constructions into two classes neglects the recent work of typologists on this topic, which distinguishes many more categories (Heine 1997, Stassen 2009, 2011). Stassen (2011), in particular, classifies these constructions into five types (have-possessive, locational possessive, genitive possessive, topic possessive, conjunctional possessive), to which he adds some transitional types.1Furthermore, Harves and Kayne do not seem to take into consideration the fact that in many languages constructions corresponding to the English verb need often have additional (and usually primary) meanings such as 'want', 'have to', and 'lack', and the fact that many languages have distinct constructions for expressing 'need' with inanimate ('I need money') and animate, particularly human, nouns ('I need someone'), as well as with a complement clause ('I need to go'). We will not address these issues in this squib.Table 1 provides a classification of the languages in Harves and Kayne's survey along the following two parameters: languages with an H- or B-construction versus languages with an N-construction ( presence of transitive 'need') or N-construction (absence of transitive 'need'). This table, of course, does not consider the possibility that one language may have several types of constructions for the same meaning.According to Harves and Kayne, while languages that have N-constructions can also have B- or H-constructions, there are no examples of languages with N-constructions that at the same time have only a B-construction for expressing predicative possession.It should be noted that (1) accounts not only for languages with transitive 'have' but also for those that have a quasi-B-type predicative possessive construction in which the possessee receives accusative case. Thus, languages like Finnish that have a transitive verb meaning 'to need' as in (2) and an accusative-case-assigning existential verb as in (3) do not constitute counterexamples to (1) even though they do not have a prototypical transitive verb corresponding to 'have' (Harves and Kayne 2012:129–130).Still, (1) suffers from accusativity bias, and so it seems preferable to us to provide a crosslinguistically neutral formulation using Dixon's (1994) syntactic primitives S (sole argument of an intransitive verb), A (subject of a transitive verb), and O (object of a transitive verb).The reformulation of (1) as (4) also takes into account the fact, already mentioned, that many languages have several competing constructions used to express predicative possession or necessity.Still, counterexamples to both (1) and (4) exist in several unrelated languages: Estonian, Arabic (Moroccan and Algerian varieties), Bantu (isiZulu, isiXhosa, and kiSwahili), Kwa (Likpe and Ewe), and Ayacucho Quechua. For reasons of space, we will not include the Kwa data.Estonian, a language closely related to Finnish, has two parallel constructions corresponding to English need.First, there is a B-type construction involving what seems to be a petrified abstract noun vaja 'need' and the 3rd person singular form of the verb olema 'be, exist', in which the person needing something is marked with the adessive (locative) case, but interestingly enough the entity needed is in the accusative.Still, this is not the only possible way of expressing English need, and indeed, not the most frequent one either since, just like Finnish, Estonian has a transitive verb vajava whose subject (in the nominative) is the person needing something and whose object (in the accusative) is the entity needed. It is thus possible to reformulate (5) and (6) as follows:Thus, Estonian turns out to be a language with an N-type construction, in which case we would expect it to also be an H-type language with a transitive verb used to express predicative possession. Yet this is not the case, as predicative possession is expressed using a B-type construction (''locative'' according to Stassen 2011) almost identical to the one found in Finnish (cf. (3)),2 except for one crucial difference: the possessee stands in the nominative and not the accusative case.As a language with a transitive verb corresponding to English need, but no transitive verb corresponding to English have, Estonian is a counterexample to (4).The most common construction used to convey the meaning 'to need' in Moroccan Arabic is an N-type construction with a verb meaning 'to be lacking' and an object pronoun encoding the person to whom, literally, something is lacking.3But there is also a quite common verb ħtɑ⧗ӡ 'to need', already present in Classical Arabic (ʔiħɑ⧗ӡɑ), related to a very common noun ħɑ⧗ӡɑ, which today means 'thing' but which originally meant 'want, need' (Lane 1863:663–664, Badawi and Haleem 2008:241). The latter meaning is still apparent in this noun's use in an N-type construction (lit. 'to be in need of '), but the related verb usually (though not always) takes the entity needed as its direct object, unlike in Classical and prescriptive Modern Standard Arabic (MSA) where the preposition ʔilɑ⧗ must precede it.4 The following Moroccan Arabic examples are from Caubet 1993:139.5Example (15) is a representative example taken from the Internet that illustrates the present tense use of this verb in modern colloquial Moroccan. 6On the other hand, predicative possession in Moroccan Arabic, as well as in Arabic in general, can only be expressed using a B-type construction quite like the Estonian one, in which the possessor is preceded by a preposition meaning 'at' and followed directly by the possessed entity with or without an intervening copular element, depending on tense (Caubet 1993:355).Interestingly enough, actual MSA usage seems to allow the transitive use of ħta⧗ӡ, as witnessed by the following example:An Internet search for ʔaħta⧗ӡ followed by a couple of common nouns such as waqt 'time' and flus 'money' returned close to a hundred pages of results, showing just how common this construction seems to be.Furthermore, Algerian Arabic (especially as spoken in Algiers), with no transitive verb of predicative possession (like all other Arabic dialects), uses a transitive verb stħaqq 'to need', which has its source in Classical Arabic ʔistaħaqqa meaning 'to deserve, to merit' (Lane 1863:606–607, Badawi and Haleem 2008:224). The latter meaning is still widely attested in all other varieties, but has shifted to that of 'to need' in the Arabic spoken in Algiers. The following sentences illustrate this quite common use with a direct object, the only possible construction available with this verb (Redouane Djamouri, pers. comm.):Moroccan Arabic, Algerian Arabic, and informal MSA thus seem to be another counterexample to the generalization in (4). This counterexample is all the more noteworthy since it illustrates a case where transitive 'need' is the result of natural evolution running counter to prescriptive grammar, which only allows an intransitive construction. Incidentally, Turkish presents the opposite situation: one in which actual usage favors an intransitive construction over the transitive verb (gereksemek 'need') created by prescriptive grammarians, which never really won any currency.7In this section, we cite three Bantu languages, isiZulu, isiXhosa, and kiSwahili, and present data from two of them.First, isiZulu and its closest relative isiXhosa (Nguni, Bantu) offer another example of a typical N-type construction. These two languages have a transitive verb ukudinga corresponding to English need, whose subject is the person needing something and whose object is the entity needed. Although noun phrases in Nguni languages do not receive accusative case marking, they are unambiguously indexed on the verb by way of subject and object concord prefixes. The following examples are from isiZulu:However, contrary to the prediction in (1) or (4), predicative possession in isiZulu (and isiXhosa) is of the B-type (or ''conjunctional''; Stassen 2011). In it, the noun class concord marker corresponding to the possessor (whose presence as an overt noun phrase is optional) is attached directly to the preposition na 'with' with or without an intervening copular element depending on tense (cf. (22c) and (22d)). It is then followed by the possessed entity either directly, in case of indefinite reference (22a), or in case of a referential possessee, redundantly outside of the prepositional phrase after a class concord marker coreferential with the possessee has been attached to the preposition na (22b).8Note that it is impossible to use the object concord prefix in the predicative possession construction, which accounts for the ungrammaticality of the following example:There is thus no doubt that the predicative possessive construction with na 'with' cannot be analyzed as a case of an accusative-case-assigning verb of possession.This situation is by no means unique among Bantu languages. Indeed, kiSwahili for instance also has a transitive verb corresponding to English need.9 Note that agreement marking in this language follows the same principles as in Nguni languages.KiSwahili thus has an N-type construction that according to (4) would entail the existence of a transitive verb of possession; but this is not the case, as predicative possession is expressed as in most other Bantu languages by means of the preposition na 'with', just as in isiZulu and isiXhosa (cf. (22)).As in Nguni languages, the ungrammaticality of the following example shows that we are not dealing with an accusative-case-assigning verb of possession:This does not mean that this kind of construction cannot be reanalyzed and undergo what Stassen (2011) calls an H-drift, thus moving toward a more transitive-like construction. This is what seems to be happening in the Bantu language seTswana, to cite but one example (Creissels 2013).Peruvian (Cuzco, Cajamarca, Huallaga) and Bolivian Quechua are presented by Harves and Kayne (2012) as examples of B-languages without transitive 'need'. While we have been unable to verify this claim, Ayacucho Quechua has a transitive verb whose primary meaning is 'want', but which is the normal (and seemingly only) way to translate English need (Soto Ruiz 2010:102 et passim). Its subject in the (unmarked) nominative is the person needing something, and its object in the accusative is the entity needed.Now this makes Ayacucho Quechua a language with an N-type construction, and we would expect it to also have an H-type construction with a transitive verb used to express predicative possession. Yet this is not the case, as predicative possession is expressed using a Btype construction with the copula verb kay (Zariquiey and Córdova 2008:181), in which the possessor either is encoded in the genitive and cross-referenced on the possessee by the corresponding possessive suffix (see (29)) or else appears in the nominative while the possessee carries a special possessive suffix -yuq meaning 'provided with' (see (30)).The facts presented in this squib offer a challenge to Harves and Kayne's (2012) generalization, as languages from four unrelated language families spoken on three different continents without any Hconstruction turn out to have an N-type construction. The authors' hypothesis is thus unlikely to be valid as an absolute universal.This is not all that surprising, for at least three reasons.First, the sample used by Harves and Kayne (2012) has a clear Eurasian bias. Although some non-Eurasian languages are included (Amharic, Guarani, Mapudungun, Purépecha, Mohawk, and Quechua), they represent only approximately one-eighth of the sample. It is well-known that Eurasia as a whole constitutes a large linguistic area (Dryer 1989:274–275). Therefore, the observed correlation might be a byproduct of a genetically and areally unbalanced sample.Second, the proposed generalization runs counter to Tsunoda's (1985) verb type hierarchy (see table 2). This hierarchy classifies predicates into seven classes based upon the affectedness of the participants. Predicates that are higher on this hierarchy tend to be coded as prototypically transitive verbs, while those on the lower end tend to be expressed as intransitive predicates whereby one of the participants receives oblique case. According to this hierarchy, verbs like have (class 6) rank lower than verbs like need (class 5) and therefore should be less prone to be expressed by transitive predicates within the same language.It should be noted that predicates' behavior in individual languages does not always conform exactly to this hierarchy as first formulated. Indeed, Malchukov (2005) has proposed a more fine-grained two-dimensional classification that discriminates between affectedness of the more subject-like and more object-like participants. Still, the revised version of Tsunoda's hierarchy has not contested the validity of Tsunoda's ordering of verb type classes 5 and 6, which are directly relevant to the present discussion. Thus, the generalization one would be tempted to make on the basis of this hierarchy predicts that if a language has a transitive verb 'to have', it should also have a transitive verb 'to need', exactly the opposite of Harves and Kayne's (2012) proposal.Third, generalizations in lexical typology are made with respect to semantic classes rather than individual items since the latter often show exceptional behavior in individual languages. It is thus unlikely that an exceptionless universal can be established on the basis of translational equivalents of individual items taken from any particular language.Still, Harves and Kayne's (2012) findings seem to challenge Tsunoda's hierarchy and so the crosslinguistic correlation between Ntype constructions and H-type constructions that they posit is certainly worth investigating. As Kayne (2013) points out, the way this generalization has been formulated allows it to be easily tested on other languages. We think, however, that further work on a larger sample is necessary in order to ascertain whether or not it remains statistically significant and can thus be used in studies in comparative syntax.Furthermore, future work on the subject should include a diachronic study in order to ascertain the possible origins of constructions that correspond to English need.