Bound person forms in ditransitive clauses revisited
Anna Siewierska* Dik Bakker
University of Lancaster University of Amsterdam
Short title: Bound person forms
Department of Linguistics and English Language
Lancaster LA1 4YT
Department of General Linguistics
University of Amsterdam
1012 VT Amsterdam
In a recent article Gensler (2003) has argued that little can be said about the ordering of bound person markers of the T(heme) and R(ecipient) relative to each other or relative to the verb stem apart from the fact that the outer markers are likely to be the result of a second-level cliticization process. We take issue with this claim and document that quite successful predictions with respect to the ordering of the T and R markers can be made on the basis of morphological alignment. Taking as our point of departure the typology of ditransitive alignment outlined in Haspelmath (2004; 2005), we show that the ordering patterns in which the R is placed closer to the verbal stem than the T are favoured in all relevant alignment types apart from the indirective, which exhibits a preference for positioning the T closer to the verbal stem than the R. These preferences for the ordering of the R and T are argued to relate directly to the frequency of use of the relative person forms and thus are seen as constituting yet another piece of evidence for the usage-based model of grammar being developed within the functional-cognitive typological paradigm (cf. e.g. Barlow & Kemmer 2000; Bybee & Hopper 2001; Tomasello 2003).
Whereas bound markers of tense, aspect and modality display evident cross-linguistic ordering preferences (see Bybee 1985), the ordering of bound person forms, be it relative to the stem or to each other, is much less consistent. Not surprisingly therefore the attempts to account for the order of person forms cross-linguistically proposed to date have met with comparatively little success (see e.g. Givón 1976; Hawkins and Gilligan 1988; Siewierska and Bakker 1996; Bitner and Hale 1996; Cinque 1999). The principles postulated have been able to provide a satisfactory account for at most 60% of the cross-linguistic data. In the main these attempts have centred on the order of bound person forms corresponding to the two arguments of a transitive clause, i.e. the A and P. Our attention, by contrast, will be devoted to the order of the two non-subject arguments of a ditransitive clause, i.e. the T (theme) and the R (recipient).
In a recent paper Gensler (2003) has argued that little can be said about the ordering of the T and R markers relative to each other or relative to the verb stem apart from the fact that the outer markers are likely to be the result of a second-level cliticization process. We will take issue with this claim and document that quite successful predictions with respect to the ordering of the T and R markers can be made on the basis of morphological alignment. Taking as our point of departure the typology of ditransitive alignment outlined in Haspelmath (2004; 2005), we will show that the ordering patterns in which the R is placed closer to the verbal stem than the T are favoured in all relevant alignment types apart from the indirective, which exhibits a preference for positioning the T closer to the verbal stem than the R. These preferences for the ordering of the R and T will be argued to relate directly to the frequency of use of the respective person forms and thus will be seen as constituting yet another piece of evidence for the usage-based model of grammar being developed within the functional-cognitive typological paradigm (cf. e.g. Barlow and Kemmer 2000; Bybee and Hopper 2001; Tomasello 2003).
The article is organized as follows. In section 2 we will briefly present some of the approaches to the order of affixes and in particular person forms that have been developed in the literature and consider to what extent they are applicable to the ordering of the R and T. In section 3 we will take a closer look at the occurrence of bound person forms in ditransitive clauses, concentrating on the factors underlying the use of a bound person form for both the R and T as opposed to just for one or the other. Section 4 will be devoted to the ordering of the R and T relative to each other. After reviewing Gensler’s analysis of the data, we will provide a modification to his predictions based on the ditransitive alignment of the respective forms. We will then test both sets of predictions first on the languages in Gensler’s sample and then on our own extended sample. Some concluding remarks will be offered in section 5.
2 Approaches to the order of bound forms1
The accounts of affix and clitic order currently available fall into two types: synchronic and diachronic. Synchronic analyses have been proposed within both the generative and functional-typological paradigms. The diachronic analyses are essentially functional-cognitive.
Most of the synchronic principles postulated to account for the order of bound forms do not directly encompass person forms. This holds for the various generative analyses based on versions of Baker’s (1985) Mirror Principle as presented in, for example, Pollock (1993), Chomsky (1995) or Cinque (1999) and for the post-syntactic account of affix order offered within an Optimality Theory framework by Trommer (2001). It also pertains to the functional analyses developed within Functional Grammar (Dik 1989; 1997) and Role and Reference Grammar (Foley and Van Valin 1984; Van Valin and La Polla 1997). 2 A synchronic analysis which does embrace person forms, but not necessarily those of the R and T in ditransitive clauses, is that suggested within a Chomskyan Universal Grammar framework by Rice (2000). Rice seeks to account for the order of bound morphemes in terms of scope relations.3 Taking as her point of departure the assumption that scopal relations are structurally represented, Rice adopts the standard phrase-structure model of scope relations according to which a morpheme of greater scope c-commands a morpheme within its scope (e.g. Reinhart 1976, Marantz 1984). In the context of this model, in which subjects c-command objects, Rice suggests that in head-final languages bound subjects (As) should be placed to the right of bound objects (Ps), while in head-initial languages bound subjects (As) should be placed to the left of bound objects (Ps). Under the view that bound subjects and objects are heads and thus suffixes in head-final languages and prefixes in head-initial languages, the predicted orderings are as depicted in (1a) and (1b) respectively.
(1) a. head-final: Verb-P-A
b. head-initial A-P-Verb
Although Rice’s discussion of the order of bound person forms does not embrace the R and T, if her analysis were to be extended to the R and T in ditransitive clauses, presumably the R would be seen as having scope over the T, as argued by Polinsky (1998) and Primus (1995). By analogy with the order of the A and P presented in (1), the relative order of the R and T in head-final and head-initial languages might then be expected to be as shown in (2):
(2) a. head-final: Verb-T-R
b. head-initial R-T-Verb
However, neither of the ordering predictions in (1) or (2) can be translated in a straightforward manner onto surface representations since Rice (2000:224-245) allows for the possibility of morpho-phonologically conditioned metathesis, deviations from the predicted order of person forms analysed as lexical categories rather than functional ones and above all the placement of the verb in the “wrong” position. The last point is of particular importance in relation to Athapaskan languages, the order of affixes within which Rice seeks to account for, since these languages are in fact head-final but have prefixes rather than suffixes, counter to the predictions in (1). In sum, the account of the order of bound person forms presented by Rice is too theory-dependent to allow any clear predictions with respect to actual surface orders to be made even for the A and P, let alone for the R and T.
Turning to diachronic explanations of the order of bound forms, the most frequently invoked is the principle of relevance, formulated by Bybee (1985), according to which the placement of an affix is a reflection of the degree to which it exerts a semantic influence on the stem. In other words, affixes which have a greater semantic effect on the stem are expected to be placed closer to the stem than those exerting a smaller effect. In line with this expectation, markers of aspect tend to be placed closer to the verbal stem than those of tense, tense markers tend to occur closer to the stem than modal markers and bound person forms occur still further away from the stem, as depicted in (3).
In the case of the markers of the A and P, since the semantic and syntactic bond between the P and the verb is taken to be closer than that between the A and the verb, the principle of relevance defines a preference for the A to precede the P in prefixal position and the P to precede the A in suffixal position. These are the same ordering preferences as suggested by Rice’s (2000) scope-based analysis though, unlike in the case of Rice, they are not tied to the general initial or final headedness of the language but only the prefixal vs. suffixal location of the bound person markers. Whether the principle of relevance can be extended to the R and T is not absolutely clear. If so, it is likely to predict a preference for the R to be placed closer to the stem than the T (see section 4). In such a case the predicted patterns would be the converse of those potentially following from the scope principle suggested by Rice, i.e. for T > R prefixes as in (4), and for R > T suffixes as in (5).
(4) Abaza (O’Herin 2001:48)
‘She gave it to her.’
(5) Noon (Soukka 2000:202)
‘I present you to her.’
An alternative diachronic explanation for the order of bound morphemes builds on their degree of grammaticalization. Diachronically older forms, i.e. forms that have undergone more development, are expected to occur closer to the stem than younger forms (see e.g. Bybee et al. 1991:33). Since As tend to be more grammaticalized than Ps, this suggests a preference for P > A order among prefixes and A > P order among suffixes. Note that both orders are the very opposite to those predicted by the principle of relevance. Comparable knowledge about the degree of grammaticalization of Rs and Ts is lacking. Therefore we can only speculate. If Rs tend to grammaticalize before Ts, again we would expect the very opposite ordering preferences to those following from the principle of relevance. But if Ts tend to grammaticalize before Rs, then the relevance and diachronic principles make the same prediction and lead us to expect R > T in prefixal and T > R in suffixal position. Another possibility is that the order and degree of grammaticalization of the R and T are not only determined by these two forces, either in isolation or in combination, but that there is a third factor involved. It is precisely this assumption that we will be exploring. But first let us take a closer look at the actual distribution of person forms in ditransitive clauses.
3 Bound person marking in ditransitive clauses
While in the case of monotransitive clauses bound person forms on the verb for both the A and the P are cross-linguistically highly frequent, and may even be seen as the cross-linguistic norm, in the case of ditransitive clauses languages typically have bound person forms for the R or the T but not for both (see e.g. Blansitt 1984; Givón 1984; Gensler 2003). Languages in which bound person forms on the verb are used for the R but not the T appear to be much more common than those in which the converse is the case. Among the 227 languages in our sample for which we could obtain data for ditransitive clauses, 111 (49%) display bound person marking for the R but not the T, as compared to only 56 (25%) with bound person marking for the T but not the R. Another 25 (11%) languages display bound person marking for the R under one set of circumstances and for the T under another. In the Oceanic language Tigak, for example, the two patterns of marking depend on the verb. The verb ‘give’ takes a bound person form for the R while the T occurs as a complement of a preposition. With the verb ‘say’, on the other hand, it is the T that is marked on the verb while the R is attached to a preposition. This is shown in (6).
(6) Tigak (Beaumont 1979:42-43)
3sg:past give-3pl with-3sg
‘He gave it to them.’
b. Ga pising-i su-guk.
3sg:past say-3sg to-1sg
‘He said it to me.’
In Araki, another Oceanic language, the patterns of marking depend on the nature of the T. If the T is inanimate, as is generally the case, the R is bound to the verb while the T occurs as the object of the instrumental/oblique preposition ni/ini or lo. But if the T is human it may take priority over the R with respect to attachment to the verb. In such a case the R is marked by a different preposition, namely sa/isa. Compare (7a) with (7b).
(7) Araki (François 2002:161)
a. Na sile-ko ne-re presin.
I give-2sg obl-some present
‘I feel like giving you a present.’
b. ^ 4
I seq give-2sg to-cst devil
‘I will give you to a devil.’
In the Brazilian language Apurinã the two patterns of marking may be used with the same verb and the same T and R configuration. We see in (8a) that the verb is marked by the third person feminine form -ro corresponding to the T ‘my arrow’, while in (8b) the verbal suffix is the third person masculine -ru corresponding to the R, ‘him’.5
(8) Apurinã (Facundes 2000: 291-292)
a. O-suka-ro uwa-mokaru nu-serepi.
3f-give-3f 3m-goal 1sg-arrow(f)
‘She gave my arrow to him.’
b. Nota suka-ru uwa-mokaru nu-serepi.
I give-3m 3m-goal 1sg-arrow(f)
‘I gave my arrow to him.’
What determines the two patterns of marking is not clear. In the Chukotko-Kamchatkan language Itelmen, which also displays verbal person marking either with R or the T, the choice of person form is determined by topichood (Bobaljik and Wurmbrand 2002). Significantly, this difference in person encoding on the verb is not accompanied by any change in order, case marking or argument structure as shown in (9).
(9) Itelmen (Bobaljik and Wurmbrand 2002:2)
a. ^ əl-aƗ-in kza kəna-nk.
father-loc iprs-give-fut-2sg you me-dat
‘Will father give you to me?’
b. Isx-enk n-zəl-aƗ-um kza kəna-nk.
father-loc iprs-give-fut-1sg you me-dat
‘Will father give you to me?’
In contrast to bound person marking of just the R, or just the T or one or the other depending on factors such as those mentioned above, the marking of both T and R is rare. It occurs in only 35 (15%) languages of the relevant 227 in our sample. Furthermore, even in these 35 languages the marking of both the T and R by bound person forms is typically not obligatory. Most commonly, bound forms for the T and R are in complementary distribution with lexical NPs or independent person forms, as is the case for instance in the Bantu languages. Compare (10a) and (10b).
(10) Kinyarwanda (Gary and Keenan 1977:91, 92)
‘John sent it to her.’
John 3sg-past-send-appl-asp Mary letter
‘John sent the letter to Mary.’
Alternatively the bound person form for the R may be obligatory, while that for the T may be subject to various restrictions. Arguably the most common of these is that the T must be third person. This has been termed by Haspelmath (2004) the Ditransitive Person Role Constraint (DPRC). Such a constraint is manifest in, for example, Chickasaw, French, Modern Greek, Monumbo, Delaware and Southern Tiwa. Additional constraints are also found. For example, according to Andrews (1975:42-43) in Classical Nahuatl the T is marked by a bound person form on the verb in addition to the R only if the R is first or second person and the T is third person plural, as shown in (11).
(11) Classical Nahuatl (Andrews 1975:42-3)
‘I gave them to you.’
In the Muskogean language Chickasaw the T receives overt marking by means of the form pit- only if human, as in (12).
(12) Chickasaw (Munro and Gordon 1982:110)
Catherine-subj Larry Bonnie-nonsubj
‘Catherine sent Larry to Bonnie.’
In the Australian language Ngiyambaa, in turn, the T must be definite. Observe the lack of a bound form for the T in (13a) as compared to (13b).
(13) Ngiyambaa (Donaldson 1980:131-32)
‘You gave him fish.’
‘I gave it to him.’
While typically it is the T that fails to be marked on the verb if the relevant restrictions are not met, in some languages the affected argument is the R. For example, in the Austronesian language Muna both the T and R are marked on the verb (rather than just the R) only when the T is third person singular as in (14a). When the T is third person plural, it is attached to the verb while the R is encoded as the object of the preposition ne, as in (14b).
(14) Muna (van den Berg 1989:67)
‘She bought it for me.’
‘She bought them for me.’
3sg-buy-3pl loc 1sg
‘She bought them for me.’
A similar situation may be observed in the Mexican language Yatzachi el Bajo Zapotec in which, however, apart from the subject only third person non-subject arguments may be bound to the verb. While both the T and R may be bound to the verb as in (15a), the inanimate T may be attached to the verb in preference to an animate R, as in (15b).6
(15) Yatzachi el Bajo Zapotec (Marlett 1985:104)
compl-give-3resp-3fam (R)-3anim (T)
‘He gave it to him.’
b. Gw-nezXw-aa-n llebo.
‘I will give it to him.’
Finally, there are also languages in which the verb may display verbal person marking of both the T and R or of either of the two. This is the case in Literary Arabic. According to Retsö (1987:228) person marking of the verb of both the R and T as in (16a) is the norm only with the verb ‘give’. This construction co-exists with two alternative ones. In the first the R is marked on the verb while the T is realized as a suffix attached to the pronominal base iyya, as in (16b). In the second, the T is marked on the verb and the R is suffixed to the preposition la, as in (16c).
(16) Literary Arabic (Retsö 1987 :228, Gensler 2003:203)
‘He gave it/him to me.’
b. 'A'ta-hu 'iyya-ya.
‘He gave me to him.’
c. ^ (T)
‘I gave it to you.’
Significantly the construction with both the R and T on the verb is an option only when the R is first or second person and the T is third person. Other Arabic dialects tend to favour either of the other two constructions.
The discussion above has made it quite clear that the presence of two bound person forms in ditransitive clauses corresponding to the R and the T is cross-linguistically rather rare. This may be taken to be to a large extent a reflection of the rarity in discourse of ditransitive clauses with two pronominal non-subject arguments of any type.7 The dominant pattern of ditransitive clauses seems to be for the T to be a lexical NP and the R a pronominal one. Hence the preference for bound person marking of the R as opposed to the T, which too has been documented above. Another factor contributing to the rarity of languages displaying ditransitive clauses with both the R and T marked by person forms on the verb is the previously mentioned Ditransitive Person Role Constraint coupled with the presence of third person bound non-subject forms only for humans, as is the case for instance in the New Guinea language I’saka (Donohue and Roque 2004:62) or alternatively the lack of overt forms for third person P or T altogether, as is the case in, for example, Seri (Marlett 1990:514). Although constructions with bound person forms for both the R and T are uncommon, the order of the two when they do co-occur needs to be accounted for. It is to this that we now turn.
4 The order of the R and T
4.1 The three basic patterns
In principle the order of bound forms may be considered in terms of linear precedence, i.e. from left to right, or in terms of closeness to the stem. Any account based on linear precedence would be rather difficult to reconcile with the principles of grammaticalization. Therefore we will be discussing the order of the R and T in terms of their respective closeness to the verbal stem. This gives us three basic patterns: (a) the T being closer to the stem than the R; (b) the R being closer to the stem than the T; and (c) the T and R being on opposite sides of the stem.8
Each of the above patterns is attested in two variants. In languages in which the T is placed closer to the verbal stem than the R, the R may precede the T in prefixal position, as in (17) from the Papuan language Ekari, or follow the T in suffixal position, as in (18) from Kashmiri.
(17) Ekari (Doble 1987:84)
‘Carry him for us.’
(18) Kashmiri (Wali and Koul 1997:156, 88)
‘He gave me a pen.’
And analogously in the case of languages in which the R is closer to the verbal stem than the T. An example of the T preceding the R in prefixal position was given earlier in (4) from Abaza. And an illustration of the R preceding the T in suffixal position was provided in (5) from the Nilo-Saharan language Noon. As for the placement of the R and T on opposite sides of the stem,
as one would expect from the ordering of lexical objects which typically occur on one side of the verb, languages in which the T and R differ in terms of prefixal and suffixal position are uncommon. An example of this pattern in which the T is a prefix and the R is a suffix is given in (19) from the Papuan language Yimas and another in which the R is a prefix and the T is a suffix is provided in (20) from Agarabi, a Trans-New Guinea language from the Eastern Highlands.
(19) Yimas (Foley 1991:212)
‘They two showed it to him.’
(20) Agarabi (Whitehead 1981:45)
John book 1(R)-give-neut-3(T)-3
‘John gave me a book.’
While in most languages only one of the above patterns of the placement of the R and T forms is found, in some there is variation conditioned either by tense, aspect or mood or person/number (e.g. French, Mudang, Yimas). In various Indo-European languages such as Albanian, French, Macedonian and Modern Greek, for example, the R proclitic normally precedes the T proclitic, as shown in (21a) on the basis of Modern Greek. However in the case of verbs in the imperative or participial form, enclitics rather than proclitics are used. Furthermore, specifically in Modern Greek with monosyllabic imperatives, while the order in (21b) with the R preceding the T is preferred, the placement of the T clitic closer to the verbal stem than the R enclitic as in (21c) is also a possibility.
(21) Modern Greek (Lascaratou 1994:73)
‘He sent me it.’
b. Δo-se= mu=to.
‘Give me it!’
c. Δo-se= to=mu.
‘Give me it!’
In French in turn the R > T order of proclitics obtains only when the R is first or second person. With third person Rs the order is T > R. Compare (22a) with (22b).
(22) French (Simpson and Withgott 1986:163)
‘He gives it to me.’
b. Il le=lui=donne.
‘He gives it to him.’
4.2 Gensler’s analysis
Using a cross-linguistic convenience sample, Gensler (2003) identified 32 languages in which both the R and T could, under some set of circumstances, be marked on the verb.9 Among these languages no preferences for the order of the R and T could be discerned. In terms of closeness to the stem, 17 languages displayed orders in which the T was closer to the stem than the R, 19 languages had orders in which the R was closer to the stem than the T and 4 languages required or allowed for the two to occur on opposite sides of the verb. In one language, Kiowa, the forms of the R and T were fused in preverbal position. And yet in another language, Delaware, the R was fused with the verb while the T was located postverbally. The distribution of these patterns in prefixal vs. suffixal position among the languages in Gensler’s sample is shown in (23).10
(23) a. R-T-Verb 9 lgs
b. Verb-T-R 8 lgs
c. T-R-Verb 7 lgs
d. Verb-R-T 12 lgs
e. T-Verb-R 3 lgs
f. R-Verb-T 1 lg
These data led Gensler to suggest that the order of the R and T is random and cannot be accounted for in any other than diachronic terms. Gensler argued that given the principles of grammaticalization, we may expect the person forms further removed from the stem, be they R forms or T forms, to be diachronically younger than the inner forms. The outer forms may be seen to be the product of a second-level cliticization
process. As such they may be assumed to be heavier segmentally, i.e. to involve more segments than the inner forms and/or to be internally complex, i.e. reveal traces of internal complexity such as the vestiges of an applicative marker as well as a person marker. In short Gensler’s analysis suggests that if the R is heavier than the T it should be placed further away from the stem than the T, if the T is heavier than the R, the T should be placed further away from the stem than the R and if both are equally heavy or light either order is possible.
Gensler’s analysis of the person forms of the R and T among the 32 languages in his sample enabled him to identify 14 languages in which the forms of the R and T could be seen to differ with respect to heaviness. In two of these languages the R and T were positioned on opposite sides of the verb. With these two languages disregarded, his sample thus yielded 12 languages for the testing of his hypothesis about the ordering of the R and T relative to each other. His predictions, i.e. that the heavier T or R would be located further away from the stem than the lighter R or T were borne out by 8 of the 12 languages. The four exceptional languages were all Afro-Asiatic, namely Akkadian, Berber, Egyptian and Hausa. Gensler argues that the inner placement of a heavier R in Afro-Asiatic languages represents a single case historically and is not due to a second stage cliticization process, which would counter his analysis, but rather a process of morphemic replacement. He hypothesizes that the heavy inner R was at the relevant proto-stage a light clitic, more or less as in example (16) cited earlier from Literary Arabic, and was subsequently replaced “in situ by heavy forms involving language specific augments”.
The success rate of Gensler’s word order predictions is 67% (8 out of 12), if interpreted relative to the number of languages in which the R and T differ in heaviness. But it drops to just over a quarter (28%; 8 out of 29) if measured relative to the total number of languages with bound person marking for both the T and R located on the same side of the verb in his sample. This figure of 28% can be improved on considerably if several additional factors are taken into account.
4.3 A refinement
In his analysis of the order of bound R and T forms, Gensler by and large ignored how these forms relate to those used for the transitive P. We in turn hold that the patterns of identification of the R and T with the P play a crucial role in how the R and T are ordered relative to each other. Following Haspelmath (2004) we will refer to the patterns of identification of the R, T and P as ditransitive alignment. By analogy with monotransitive alignment, the major patterns of ditransitive alignment are: indirective, secundative, neutral and tripartite. In indirective alignment, which is the ditransitive counterpart of accusative, the T is identified with the P in contrast to the R. In secundative alignment, the ditransitive counterpart of the ergative, it is the R which is identified with the P while the T is distinct. Neutral alignment obtains when the P, T and R are all treated the same, and tripartite alignment when each is treated differently. The identification of alignment may be achieved on the basis of morpho-phonological or syntactic criteria. We will adopt a morpho-phonological criterion which entails morpho-phonological identity of the bound person forms of the P and T in indirective alignment, the P and R in secundative alignment and the P, T and R in neutral alignment.11 Given the morpho-phonological criterion we will also need to recognize a fifth minor alignment pattern in which the forms for the R and T are the same but distinct from those used for the P. We will refer to this minor alignment pattern with the same term as used for the corresponding monotransitive pattern, i.e. double oblique.
Our basic rationale for viewing patterns of morpho-phonological alignment as relevant for the determination of the order of bound R and T person forms comes from considerations of the effects on grammatical structure of token frequency. In usage-based models of grammar (c.f. Barlow and Kemmer 2000; Bybee and Hopper 2001; Croft 2001) it is widely recognized that more frequent forms are likely to undergo reduction and cliticization earlier than less frequent forms. Since languages tend to have considerably fewer ditransitive verbs than transitive ones and ditransitive clauses occur in discourse less often than transitive ones, it seems pretty reasonable to assume that person forms which occur in both transitive and ditransitive clauses are likely to cliticize earlier than those that are specific to ditransitive clauses alone. In indirective alignment the relevant forms are those of the P and T. Accordingly, we may expect the T forms to grammaticalize earlier than the forms of the R, which in the normal course of events after the subsequent grammaticalization of the R should result in either of the orders in (24).
(24) a. R-T-Verb
In secundative alignment, by contrast, it is the R which is the same as the P. Consequently, the forms of the P and R are more common than those of the T. Accordingly, we may expect the R forms to grammaticalize in preference to those of the T. In fact the forms of the T may not undergo reduction and cliticization at all. But even if they do at some later stage, the resulting orders should be as in (25).
(25) a. T-R-Verb
Needless to say, in the case of neutral alignment, it cannot be the higher token frequency of one set of person forms as opposed to another which determines the order of the R and T, since the forms used for the P, T and R are all the same. Nonetheless, frequency may be seen to play a significant role, though not token frequency per se but rather the frequency of the use of pronouns, particularly person ones, in R as opposed to T roles. That the R is much more frequently pronominal than the T has already been commented on. Recall from section 3 that there are many more languages in which it is the R which is manifested by a person form bound to the verb than in which it is the T. In fact, virtually all the instances that we are aware of in which the T takes precedence over the R in terms of bound person marking on the verb involve constructions in which the R is adpositionally marked, as in (26) from the Mayan language Mam where this is the only pattern, or as in Tigak (6), Araki (7) or Muna (14) where the marking of the T is just one of the existing possibilities.12
(26) Mam (England 1983:183)
^ recpast-emph 3sg(T)-dir-3pl(A)-give-dir 1pl-to
‘They gave it to us.’
Recall also that within languages the R is much more often pronominal than the T. In English there is even a well-known constraint, restricted to double object as opposed to prepositional constructions, against clauses with a pronominal T and non-pronominal R; in other words if the T is pronominal so must the R be. Note the contrasts in (27).
(27) a. I gave him the sweets.
Another piece of evidence in support of the predilection for Rs as opposed to Ts to be pronominal is that in languages which exhibit suppletion of the verbal stem dependent on person, it is always the person of the R and not of the T which is involved (Comrie 2000). This suggests that in languages manifesting neutral alignment of the bound person forms of the P, R and T we may expect the form placed closer to the verbal stem in ditransitive clauses to be interpreted as the R rather then the T. In other words we may expect a preference for the orderings in (25) as opposed to those in (24). This also holds for instances of tripartite alignment. If both the R and T are phonologically distinct from each other and from the P, the primacy of the R over the T with respect to person marking and pronominality suggests that either only the R should cliticize to the verb or at least it should do so earlier than the T, again yielding the orders in (25). And a similar argument can be advanced for positioning the R closer to the stem than the T in double oblique alignment. In sum, our considerations of the frequency of use of specific sets of person forms as reflected by their ditransitive alignment coupled with the evident preference for Rs as opposed to Ts to be pronominal suggests that the location of Rs closer to the verbal stem than Ts should be favoured in all alignment types but for the indirective.
Our predictions with respect to the ordering of bound person forms of the R and T in ditransitive clauses fair considerably better with respect to Gensler’s sample than his own.13 They fully account for the ordering found in 17 of the languages in his sample and make partially correct predictions, i.e. for at least one of the occurring orders, for another four languages. Our success rate for the sample as a whole (after the exclusion of five languages, four with fused markers or markers placed on opposite sides of the verb, namely and Mundang) is thus 63% (fully correct) and 78% (partially correct), as compared to his 29%.
The difference between Gensler’s account of the order of the R and T and ours essentially boils down to the cases where the forms of the R and T do not differ in heaviness. Gensler makes no predictions in such instances while we predict a preference for the R to be positioned closer to the stem than the T.
We have achieved a similar level of success with respect to a somewhat larger sample of 44 languages displaying bound person marking of both T and R, at least under some set of circumstances. This sample, presented in Appendix 2, consists of Gensler’s sample without Mundang and four languages which are irrelevant for this exercise, viz. Basque, Kiowa, Delaware and Ungarinjin, and enriched by an additional 17 languages. The distribution of ordering patterns reflecting our predictions relative to alignment among the languages in this sample is shown in (28). The alignments specified are those involving a third person T. Languages with an asterisk have more than one order.
(28) Orders of the R and T in line with our predictions
R-T-Verb: *Albanian, *Bulgarian, Ekari, *French (R=1 or 2), Southern Tiwa, *Modern Greek, Sumerian
Verb-T-R: Amele, Anggor, Colloquial Arabic (B), *French , Gooniyandi, Kashmiri, Kate, Manam, Selepet
T-R-Verb: Yimas14 (R= 1 or 2)
Verb-R-T: Colloquial Arabic (A)
T-R-Verb: Chinook, Kinyarwanda, Lakhota, Moshi, Nkore-Kiga,
Verb-R-T: Ashaninca, Classical Arabic, Diola Fogny, Doyayo, Koromfe, Nama, Noon, *Wolof, Zapotec
Verb-R-T: Hausa, Monumbo, Ngiyamba
The distribution of orderings which counter our predictions relative to alignment is presented in (29).
(29) Orders of the R and T countering our predictions
T-R-Verb: Abaza, Abkhaz, *French (R=3)
Verb-R-T: *Albanian, Akkadian, Au, Berber, *Bulgarian, Egyptian, *Modern Greek
R-T-Verb: Slave, *Wolof, Classical Nahuatl (T=3pl)
Interestingly enough, our overall success rate with respect to fully correct predictions is virtually the same as Gensler’s, namely 68%. However, whereas Gensler’s analysis embraced only 12 of the 32 languages in his sample, ours covers all the languages in our sample. Among these 44 languages the order of the R and T fully corresponds to our predictions in 30 cases, partially reflects our predictions in 6 and runs counter to our predictions in 8 cases. Thus under a favourable interpretation 36 out of 44 languages (82%) exhibit an ordering pattern of the R and T in line with our predictions.
On closer inspection we see that most of the exceptions to our predictions with respect to the ordering of the R and T involve indirective alignment. Some of the exceptional orders occur in languages in which they co-exist with patterns which conform to our predictions. The languages in question are all Indo-European, namely Albanian, Bulgarian, French and Modern Greek. In Albanian, Bulgarian and Modern Greek, the order in which the R is placed closer to the verbal stem than the T is clearly the minority pattern. Of the exceptional languages that display only a single ordering of the R and T, three are the same Afro-Asiatic languages that constituted an exception to Gensler’s second-stage cliticization account, namely Akkadian, Berber and Egyptian. Another two are the Northwest Caucasian Abaza and Abkhaz. It is of interest to note that in both languages the morpho-phonological form of the R is distinct from the P and T essentially only in the third person. The forms used for third person Rs are, however, homophonous with those used for the A. Thus if what underlies the placement of the R or T closer to the stem is token frequency, the positioning of an R which is homophonous to an A closer to the stem than the T which is homophonous to a P would be fully in line with the dictates of frequency, since As are much more often pronominal than Ps or Rs, let alone Ts (cf. Du Bois 1987). The last language in which the order of the R and T in indirective alignment counters our expectations is Au a Papuan language of the Toricelli family. Thus in all of the 26 cases found with person forms manifesting indirective alignment, 62% (16 out of 26) are in line with the expectations, which is only marginally better than chance.
The other alignment types consistently reflect a preference for positioning the R closer to the stem than the T. Of the three exceptional orders in neutral alignment, the one found in Classical Nahuatl could in fact be disregarded, as the presence of T marking in addition to R marking is very restricted. Recall that such marking occurs when the T is third person plural and the R is a first or third person. Alternatively the T must be indefinite. The case of Slave is also somewhat suspect. All the examples of two bound non-subject person forms on the verb provided in Rice’s (1989:627, 775) grammar involve either a T which is unspecified rather than one corresponding to a definite referent or oblique objects rather than clear Rs. The latter typically include an incorporated adposition and consequently the alignment is not really neutral. This leaves Wolof in which the order with the T closer to the stem than the R is the minority pattern.
5 Concluding remarks
We have seen that the placement of the R and T relative to the verbal stem is in fact somewhat more predictable than has been recently suggested. The factor that plays a strong predictive role is ditransitive alignment. In all alignment types but for the indirective type, there is a strong preference for locating the R closer to the verbal stem than the T. In indirective alignment, by contrast, it is the T which is more commonly located closer to the verbal stem than the R. This ordering is not, however, as strongly preferred as we would have expected. There are a considerable number of languages which exhibit indirective alignment but require or allow for the placement of the R in an inner location. The question arises why this should be the case.
The explanation that we can offer is one based on competing motivations. We have argued that since in languages with indirective alignment the person forms used for the T are more frequent than those of the R, by virtue of the former being the same as those of the P, the T forms may be expected to cliticize earlier than the R forms. Nonetheless, as we have also argued, in ditransitive clauses it is the R rather than the T which tends to be pronominal. Thus while the forms used for a third person T are more frequent overall than those used for the R, in ditransitive clauses the converse is the case. Accordingly, the local frequency of the R forms may override the overall higher token frequency of the T forms and result in the earlier grammaticalization of the R than the T. In the case of the other alignment types there is no such conflict between token frequency of actual person forms and frequency of pronominal usage within the ditransitive construction. The person forms used to express the R will always be more frequent than those of the T irrespective of whether they are or are not homophonous with those of the P due to the strong tendency for Rs but not Ts to be human and pronominal.
1 first person
2 second person
3 third person
CST construct suffix
FAM familiar (honorific)
OBL oblique preposition
RECPAST recent past
RESP respect (honorific)
SEQ sequential aspect
* Correspondence should be sent to the first author.
1. Our use of the term bound form here encompasses both affixes and clitics. For some discussion of the problems arising in distinguishing the two, see Siewierska (2004:24-34) and the references cited there. Although the location of the latter may be syntactically rather than morphologically determined, the verb is virtually always one of the categories to which person markers (of arguments) may attach.
2. Both Baker’s (1988) Mirror Principle and the analyses suggested within Functional Grammar (Dik 1989; 1997) and Role and Reference Grammar (Foley and van Valin 1984; Van Valin and La Polla 1997) are indebted to Bybee’s (1985) Principle of Relevance, discussed below.
3. In many generative accounts of bound forms the affix vs. clitic distinction is crucial as the proposed analyses may involve only the former or alternatively only the latter, however defined. Rice (2000) includes both types of forms under her analysis.
4. It is important to note that the language does have overt person markers for the third person objects.
5. It needs to be mentioned that Facundes (2000:192) questions whether the verb ‘give’ in Apurinã should be considered a ditransitive as opposed to a transitive verb.
6. Marlett (1985:107) states that in another of the Zapotecan languages, Isthmus Zapotec, only third person inanimate markers may be attached to the verb. This is cross-linguistically quite exceptional.
7. Relevant statistical data for English are presented in Siewierska and Hollmann (2006).
8. Another possible pattern is hierarchically determined order, where the order of the R and T depends on which is higher on the person hierarchy. According to Wise (1986:585) this is the basic principle determining the order of the R and T in the Andean language Nomatsiguenga, the first and second person always being placed before the third irrespective of which is the R and which the T. Unfortunately Wise does not provide examples involving both a third person T and R.
9. Gensler speaks of 31 languages but when one counts the two types of Colloquial Arabic, the correct figure is 32.
10. Some languages have more than one pattern. Therefore the numbers do not add up to 32 (languages) but to 42 (patterns).
11. As in the case of monotransitive alignments, there may be splits dependent on person and number. The most common ones involve neuter alignment for the first and second person and indirective for the third. Such is the case in Abaza, Abkhaz, Albanian and French, for example. In Tarifit Berber (McClelland 2000:2-21) the alignment is indirective both in the second and third person singular and neutral in the first person singular and in all persons in the plural.
12. When the R is not adpositionally marked, the encoding of the T by a person marker on the verb in preference to the R is generally due to topicality, animacy or definiteness restrictions.
13. We did not manage to find data for one of the languages in Gensler’s sample, namely Mundang. We therefore had to exclude it from our considerations.
14. This pattern is difficult to classify in terms of alignment since the 1st and 2nd person forms exhibit secundative alignment and the third person forms indirective. If both the T and R are third person the R forms occur as suffixes rather than as prefixes.
Andrews, J. R. (1975) Introduction to Classical Nahuatl. Texas: University of
Baker, M. (1985) The mirror principle and morpho-syntactic explanation.
^ 16: 537-576.
Barlow, M. and S. Kemmer (2000) (eds.) Usage-based models of language.
Stanford: CSIL Publications.
Beaumont, C. H. (1979) The Tigak language of New Ireland. Pacific Linguistics,
B-58. Canberra: ANU.
Bitner, M. and K. Hale (1996) Ergativity: toward a theory of heterogeneous class.
Linguistic Inquiry 27.4: 531-604.
Blansitt, E. L. (1984) Dechticaetiative and dative. In F. Plank (ed.), 127-150.
Bobaljik, J. and S. Wurmbrand (2002) Notes on agreement in Itelmen.
Linguistic Discovery 1. http://journals.dartmouth.edu/cgi-bin/webobjects/journals.woa/2/xmlpage/1/archive
Bybee, J. (1985) Morphology. Amsterdam: John Benjamins.
Bybee, J., W. Pagliuca and R. D. Perkins (1991) Back to the future. In E. Traugott
and B. Heine (eds.) ^ Vol. 2. Amsterdam:
John Benjamins. 17-59.
Bybee, J. and P. Hopper (eds.) (2001) Frequency and the emergence of linguistic
structure. Amsterdam: John Benjamins.
Chomsky, N. (1995) The minimalist program. Cambridge (Mass.): MIT Press.
Cinque, G. (1999) ^ .
Oxford: Oxford University Press.
Comrie, B. (2001) Recipient person suppletion in the verb “give”. Paper given at the
4th International Meeting of the Association of Linguistic Typology, Santa
Barbara 19-23 July 2001.
Croft, W. A. C. (2001) ^ . Oxford: Oxford University
Dik, S. C. (1989) The Theory of Functional Grammar. Part 1. The clause.
Dik, S. C. (1997) The Theory of Functional Grammar. Part 1: The structure of the
clause. Berlin: Mouton de Gruyter.
Doble, M. (1987) A description of some features of Ekari language structure.
Oceanic Linguistics 26: 55-113.
Donaldson, T. (1980) Ngiyambaa: the language of the Wangaaybuwan.
Cambridge: Cambridge University Press.
Donohue, M. and L. San Roque (2004) I'saka. A sketch grammar of a language of
North-Central New Guinea. Canberra: Pacific Linguistics. The Australian
Du Bois, J. W. (1987) The discourse basis of ergativity. Language 63.4: 805-855.
England, N. (1983) A grammar of Mam, a Mayan language. Austin: University
of Texas Press.
Facundes, S. da Silva (2000) The language of the Apurinã people of Brazil.
Ph.D dissertation, University of New York at Buffalo.
Foley, W. A. (1991) ^ . Stanford: Stanford
Foley, W. A. and R. D. Van Valin (1984) Functional syntax and universal grammar.
Cambridge: Cambridge University Press.
François, A. (2002) Araki. A disappearing language of Vanatu. Pacific Linguistics
522 Canberra: The Australian National University.
Gary, J. and E. Keenan (1977) On collapsing grammatical relations in
universal grammar. In P. Cole and J. Sadock (eds.), Syntax and semantics 8. Grammatical relations, New York: Academic Press. 83-120.
Gensler, O. (2003) Object ordering in verbs marking two pronominal objects: non-
explanation and explanation. Linguistic Typology 7.2: 187-231.
Givón, T. (1976) Topic, pronoun and grammatical agreement. In C. Li (ed.)
Subject and topic. New York: Academic Press. 151-188.
Givón, T. (1984) Direct object and data shifting: semantic and pragmatic case.
In F. Plank (ed.) Objects. London: Academic Press. 151-182.
Hawkins , J. A. and G. Gilligan (1988) Prefixing and suffixing universals in relation
to basic word order. Lingua 74: 219-259.
Haspelmath, M. (2004) Explaining the ditransitive person-role constraint: a usage
based account. Constructions 2 [free online journal,
Haspelmath, M. (2005) Ditransitive constructions: the verb give. In: M. Haspelmath,
M. Dryer, D. Gil and B. Comrie (eds.) ^
structures. Oxford: Oxford University Press. 426-429.
Lascaratou, C. (1994) Performance principles in word order variation. Athens:
Parousia Journal Monograph Series 29.
McClelland, C. W. (2000) The interrelations of syntax, narrative structure, and
prosody in a Berber language. Lewiston: The Edwin Mellen Press.
Marantz, A. (1984) On the nature of grammatical relations. Cambridge (Mass.):
Marlett, S. A. (1985) Some aspects of Zapotecan clausal syntax. Working papers of
the Summer Institute of Linguistics University of North Dakota 29: 83-154.
Marlett, S. A. (1990) Person and number inflection in Seri. International Journal of
American Linguistics 56: 503-541.
Munro, P. and L. Gordon (1982) Syntactic relations in Western Muskogean: a
typological perspective. Language 56: 81-115.
O’Herin, B. (2001) Abaza applicative. Language 77.3: 477-493.
Plank, F. (ed.) (1984) Objects. London: Academic Press.
Polinsky, M. (1998) A non-syntactic account of some asymmetries in the double
object construction. In J-P. Koenig (ed.) Conceptual structure and language: bridging the gap. Stanford: CSLI. 403-423.
Pollock, J-Y. (1993) Notes on clause structure. Amiens: Université de Picardie.
Primus, B. (1995) Cases and thematic roles. Ergative, accusative, active.
Habilitation dissertation. University of Munich.
Reinhart, T. (1976) The syntactic domain of anaphora. Doctoral dissertation. MIT.
Retsö, J. (1987) Copula and double pronominal objects in some Semitic languages.
^ 137: 219-245.
Rice, K. (1989) A grammar of Slave. Berlin: Mouton de Gruyter.
Rice, K. (2000) Morpheme order and semantic scope. Cambridge: Cambridge
Siewierska, A. and D. Bakker (1996) The distribution of subject and object
agreement and word order type. Studies in Language 20.1: 115-161.
Siewierska, A. and W. Hollmann (2007) Ditransitive clauses in English with
special reference to Lancashire dialect. In M. Hannay and G.J. Steen (eds.)
Structural-functional studies in English grammar. Amsterdam: John Benjamins.
Simpson, J. and M. Withgott (1986) Pronominal clitic clusters and templates. In
H. Borer (ed.) ^ 19 The Syntax of Pronominal Clitics.
New York: Academic Press. 149-174.
Soukka, M. (2000) A descriptive grammar of Noon. Munich: Lincom Europa.
Tomasello, M. (2003) Constructing language. A usage-based theory of language
acquisition. Cambridge (Mass.): Harvard University Press.
Trommer, J. (2001) A hybrid account of affix order. ^ 37: 469-480.
Van den Berg, R. (1989) A grammar of the Muna language. Dordrecht: Foris.
Van Valin, R. D. and R. J. LaPolla (1997) Syntax. Cambridge: Cambridge
Wali, K. and O. N. Koul (1997) Kashmiri: a cognitive-descriptive Grammar. London:
Whitehead, C. R. (1981) Subject, object and indirect object: towards a typology of
Papuan languages. ^ 13: 32-63.
Wise, M. R. (1986) Grammatical categories of Preandine Arawakan languages of
Peru. In: Desmond C. Derbyshire and G. K. Pullum (eds.) Handbook of
Amazonian Languages, vol. 1. Berlin: Mouton de Gruyter. 567-642
Gensler’ s (2003) Sample N=32
Abkhaz, Akkadian, Albanian, Amele, Classical Arabic, Colloquial Arabic A, Colloquial Arabic B, Basque, Berber, Chinook, Delaware, Egyptian, French, Gooniyandi, Hausa, Kamberra, Kashmiri, Kate, Kiowa, Manam, Monumbo, Mundang, Classical Nahuatl, Nama, Nkore-Kiga, Slave, Somali, Southern Tiwa, Sumerian, Ungarinjin, Wolof, Yimas
Our extended sample (N = 44) consisting of the languages in Appendix 1 without Basque, Delaware, Kiowa, Mundang and Ungarinjin, plus the following additional 17 languages:
Abaza, Anggor, Ashaninca, Au, Bulgarian, Diola Fogny, Doyayo, Ekari, Kinyarwanda, Koromfe, Lakhota, Modern Greek, Moshi, Ngiyamba, Noon, Selepet, Zapotec.