<1>

Amongst the set of widespread derivational extensions on verbs in Nilotic, there is one prototypically marking an event directed towards some individual or a location, usually referred to as the Dative marker in the study of this language family. The Nilotic family is commonly divided into three branches (following Köhler 1955): Western, Eastern, and Southern Nilotic, and the Dative suffix is attested in all three primary branches of this Nilo-Saharan subgroup. Since Dative marking in Nilotic languages involves the use of cognate morphemes, these distributional facts allow us – in principle – to trace down not only the formal but also the syntactic and semantic history of this verbal marker.

<2>

Below, we shall first have a detailed look at the Dative in one Eastern Nilotic language, Turkana (section 2). The conclusions drawn from this language-internal investigation will then serve as a basis for intragenetic as well as intergenetic comparison of prototypical properties of this type of construction (section 3 and 4, respectively). Minor differences between Nilotic languages with respect to the morphosyntactic status of the Dative correspond to interesting differences between these languages in terms of case-assignment rules and constituency, more specifically the rigidity of constituent order. As further argued on the basis of this cross-linguistic, dynamic comparison, Dative marking on verbs as a valency-changing strategy in Nilotic involves thematic incorporation into the verbal complex of semantic notions that are peripheral to event structures as expressed in basic, non-derived verbs.

<3>

Verbal derivation in Nilotic typically involves suffixation, rather than prefixation, the only derivational prefix being the causative marker.  [2] The Eastern Nilotic language Turkana is rather typical in this respect. Here, the Dative suffix is in paradigmatic contrast with the Ventive and Itive marker (expressing movement towards or away from the deictic centre), whereas each of these may be preceded by a verb extension expressing habituality, or a reduplicated verb root expressing a distributive (pluractional) meaning. Thus:

<4>

The formal realisation of the Dative in Turkana involves a plethora of forms, conditioned by phonology (e.g. vowel harmony and tone), but also by morphology, i.e. different paradigms require slightly different forms for the Dative. Accordingly, the credo "one meaning-one form" rarely applies to the Dative - or other verbal extensions in this language for that matter. (For a description of allomorphic rules, the interested reader is referred to Dimmendaal 1983a:113-119).The following allomorphs occur for the Dative:

<5>

Vowels between parentheses represent word-final vocalic segments which are de-voiced before pause and which are extra short before another word in the same sentence. They always carry a single (high or low) tone, whereas word-final vowels which remain voiced before pause carry a high or low tone followed by a floating tone; see Dimmendaal & Breedveld (1986) for a description.

<6>

The presence of the Dative marker on the verb in Turkana appears to be coindexed prototypically with three distinct types of syntactic complements:

<7>

In his typological survey of this type of semantic modification of event structures as expressed by verbs, Peterson (2007) refers to this verbal marker as the Applicative, following a widespread tradition in the study of Bantu languages. The semantic relations expressed by the Dative marker in Turkana are exactly those assumed by Peterson (2007) to be prototypical for Applicative marking cross-linguistically.

<8>

Turkana has a "Marked Nominative" case system, whereby postverbal subjects of transitive and intransitive verbs take Nominative case (expressed by way of tonal inflection of the noun phrase or pronoun). The corresponding object takes Absolutive case. The Absolutive case is also used for subjects occurring in preverbal (as against postverbal) position, as well as for pronouns and nouns or noun phrases in isolation, i.e. the Absolutive also constitutes the citation form in Turkana. In addition, there is peripheral case-marking, more specifically for Locative and Instrumental case.

<9>

In the first type of Dative construction, involving an additional noun phrase or pronoun with Absolutive case marking, the latter typically appear with a ditransitive verb whereby the additional noun phrase or pronoun precede the primary object, as with the transitive verb -ɪ̀nɔk 'light (something)'.

<10>

Because Turkana uses cross-reference marking on the verb for subjects as well as for objects when the latter refer to first or second person, the syntactic subject and (Dative) object positions need not be filled. Thus, in example (2) above, the syntactic pronouns 's/he' and 'me' may be omitted, at least from a syntactic point of view.

In the second type of Dative construction, the presence of a Dative marker corresponds to the presence of a phrase taking Locative case, which always follows the primary object when the latter occurs.

<11>

The two transitive verbs - ɪ̀n ɔk 'light' and - ɪ ̀r ɛp 'insert' thus both take a Dative suffix, but at the same time they require slightly different case marking as well as syntactic configurations. The additional complement in the case of 'light (for me)' in ex. (2) is to be treated as a kind of object, as shown by the obligatory pronominal object marking on the verb with first and second person ("participants of speech") in Turkana. Compare also a prototypical, highly transitive verb like 'beat':

<12>

The Dative object ('for me') in example (2) above behaves exactly like the object 'me' in sentence (4), in that it is cross-referenced on the verb, thus functioning as a kind of object in either construction. Given the derived nature of the verb form 'light (something) for somebody', and the optional presence of the additional object ('me' in example (2)), the latter may be called a "secondary" object, contrasting with a "primary" object, as with 'me' in sentence (4). Three-place verbs without a Dative extension are rare in Turkana, but do nonetheless exist:

<13>

The verb 'give', which prototypically functions as a three-place verb cross-linguistically, takes a Dative extension in Turkana; the corresponding root form -ɪn- is no longer used as such synchronically in the language:

<14>

The Dative extension may also be added to inherently intransitive verbs, as with the verb -ŋɔ 'be angry':

<15>

A Dative verb may also be combined with passive voice. Passives in Turkana involve impersonal active constructions whereby the subject position is empty, i.e. the latter cannot be filled by a noun phrase or syntactic pronoun.

<16>

Alternatively, the Dative verb may occur with middle voice. Diathesis accordingly does not affect or interact with Dative marking:

<17>

When comparing the two types of Dative-verb constructions, one involving a secondary object and the other one a locative complement, it becomes clear that the former typically refers to human (or at least animate) entities affected by the state of affairs expressed in the verb, i.e. benefactives, recipients, or malefactives (or adversative constructions); the locative-complement type on the other hand typically refers to inanimate entities functioning as a target or ground for some event expressed by a verb. This semantic split is reflected in their distinct mapping onto syntax, and in the distinct position they occupy relative to the primary object. There is ample evidence for such an animacy hierarchy (or "prominence hierarchy") in other parts of Turkana grammar, as argued in Dimmendaal (1983:83-88). The distinct syntactic configurations associated with verbal Dative marking fit in with this more general patterning observable for this language.

Alternative constructions of the type illustrated in the following example suggest that the Dative strategy co-varies with, or alternates with, a construction involving a prepositional phrase. Compare:

<18>

At first sight, then, the presence of the Dative suffix seems to license oblique arguments without the use of a preposition (the proclitic marker à). But the actual system is far more intricate, and far more interesting for that matter, as shown next.

First, there are various lexicalised forms involving Dative verb forms where the corresponding root form is no longer used, i.e. where the peripheral strategy with a prepositional or adverbial phrase does not exist as an alternative. Compare, for example, the verb -ìw-akɪn 'put' which (as in English) requires a locative complement (either a noun phrase marked with Locative case, or an adverb of place) as an obligatory argument; the latter may be omitted when known by the speaker and hearer:

<19>

Second, the distribution of prepositions in Turkana is far more complex, the presence versus absence of a Dative marker being just one conditioning factor. As a matter of fact, it turns out that there are two types of prepositions in this language, and the Dative interacts with only one of these two subcategories. Prepositions derived from nouns, either historically or synchronically, may be added regardless of whether the verb carries a Dative marker or not. These prepositions are used in order to specify the location or search domain for some entity or object. Thus, the prepositional noun tɔɔ ̀ma may be omitted in the following example, or it may be added in order to be more specific about the location of the bread:

<20>

Such prepositional nouns behave like nouns in Turkana, for example in that pronominal modifiers take the possessive form (tɔɔ̀ma kɛŋ 'inside it (lit. its inside)'). Given their inherent locative meaning, such prepositional nouns do not require a locative (gender) prefix, thus tɔɔ ̀ma, not *lɔ̀-tɔɔ̀ma or nà-tɔɔ̀ma). But when expressing a source, they do require such a marker just like regular masculine (rather than feminine) nouns:

<21>

But there are three prepositions in Turkana which are not derived from nouns whose function it is to express grammatical relations with core constituents, and whose presence does in fact interact with Dative marking on the verb: First, the preposition à, which is followed by nouns or noun phrases inflected for Locative case; second, a low-tone preposition kà, a preposition which is always followed by adverbs of place or demonstratives; finally, a preposition (followed by a floating low tone) ka`, which is followed by noun phrases or pronouns expressing a comitative role.

<22>

The two prepositions à and kà (which are in complementary distribution because of the kind of syntactic category that can follow) are obligatory if direction from some source is to be expressed, regardless of whether the verb takes a Dative extension or not. Compare the following alternation:

Examples with the comitative-marking preposition ka`:

<23>

The preposition ka` introduces the Comitative phrase, but it sometimes is accompanied by Dative marking on the verb, as in (19), and sometimes not, as in (18). Also, the two prepositions à and kà are obligatory when expressing Source (direction from), but when a semantic role like Place is expressed, they sometimes are absent when Dative marking occurs on the verb, as in (9), whereas with other verbs carrying a Dative marker the preposition is obligatory, as in the following example:

<24>

Consequently, although there are examples where the use of the Dative marker is indeed in complementary distribution with the use of these prepositions (as in (9) versus (10)), there are also clearcut examples where the preposition occurs in spite of the fact that the verb is expanded with a Dative extension. It would be a gross oversimplification of facts, therefore, if one were to claim that the Dative in Turkana simply involves the syntactic incorporation into the verb phrase of an adpositional form, or claim – in generative terms - that "head movement" in the sense of Baker (1988) is involved. The point is, first, to describe or explain under which conditions these prepositions are obligatory, and, second, which semantic or functional role the Dative extension plays in these syntactic configurations. These two domains, the syntactic and the semantic range of Dative formation and prepositional marking in Turkana, are discussed next.

<25>

Given the fact that certain verbs in Turkana marked for Dative require a preposition, whereas others do not, in spite of the fact that in both cases Place is expressed, it has to be concluded that Dative formation is not a freely generated syntactic phenomenon, but a lexical process. But serving which function?

<26>

First, it turns out that there are differences in information packaging in cases where there is an alternation between a periphrastic construction introduced by the preposition (à / kà or ka` ) and a corresponding construction with a Dative verb where the same preposition is lacking. In the periphrastic construction the prepositional phrase carries special (assertive) focus, presenting information assumed by the speaker not to be present in the mind of the hearer. This information may be new or contrastive to information expressed in preceding discourse, as in (22). In the corresponding construction with the Dative verb this information is presented by the speaker as given information, as in (23).

Additional contrastive examples:

<27>

These facts also put so-called lexicalised Datives, such as the verb 'give (to somebody)' in example (6) in a new perspective. Such verbs for which a corresponding root form without the Dative extension is no longer used apparently express grammaticalised ways of expressing thematic relations. In the case of 'give' the recipient needs to be expressed (as a secondary object), or the latter is known, in which case it may be expressed by a zero anaphor.

<28>

Note that if the implied recipient refers to the speaker(s) ('me', 'us'), Turkana uses a passive (or impersonal active) construction. There may be a cultural reason for this, namely avoiding the use of a directive with this verb by immediately addressing the hearer; instead, an impersonal construction is used, thereby leaving it open who should be the one donating.

<29>

If another constituent is added, e.g. in order to specify a location, the latter needs to be introduced by way of the preposition à in Turkana, as in example (21) above, where information on the position of some transferred figure ('on a plate') represents new information.

<30>

If the lexicalised Dative complement expresses a location, as with 'put down (somewhere)' in examples (11-12), the latter may also be expressed by way of a zero anaphora for the same reason, thereby showing iconically that the location is known from the context in which the verb is used.

<31>

In the following example the verb stem for 'abandon, leave behind' consists of a root -ɛs-, which is no longer used as such, plus an obligatory Dative extension signifying the lexically incorporated location. Because the verb is transitive, a (primary) object may be added.

<32>

As shown again by this example, such locative complements coindexed with the Dative marker on the verb may be expressed by way of zero anaphora. In other words, the actual meaning of the lexicalised Dative verb -ɛs-ɪkɪn is 'abandon somewhere, leave behind somewhere'. The lexicalised verb - i ̀my-ɛkɪn also has a lexically incorporated notion of location. In other words, it means 'let/leave somewhere in a given place or a place known to the hearer':

<33>

If the location needs to be expressed in combination with such verbs with an incorporated semantic notion of location, e.g. because it provides important new information, it again needs to be introduced by way of the preposition à, parallel to the obligatory use of this preposition with verbs with lexically incorporated Recipients, e.g. 'give to somebody' in (21) above.

<34>

Given the separate role of prepositional nouns (as against pure prepositions), namely that of specifying the search domain, these may be added to constructions introduced by the preposition à or ka. With such constructions the prepositional nouns again behave like regular nouns preceded by these location-marking prepositions:

<35>

This distributional behaviour of the preposition à parallels the obligatory use of this syntactic category with locative phrases in general in the language when combined with a simple (non-derived) verb stem expressing an event:

<36>

The preposition à is also used when movement within a certain area (34) as against movement from or towards some location (35) is involved:

<37>

The Dative extension is not used if the Figure moving towards some location is the subject, rather than the object. This role is performed by the Ventive or Itive markers, which express movement towards or away from the deictic centre. These markers are mutually exclusive also with the Dative extension, i.e. a verb taking a Ventive or Itive marker cannot take a Dative marker (and vice versa).

<38>

Interestingly, however, one verb of motion may be expanded with a Dative extension in order to express movement of the subject in a metaphorical sense, as with the following example, where the same verb is used to express a metaphorical concept like 'settle into' or 'embark upon':

<39>

Also, when the subject and object are coreferential, as with middle voice constructions, Dative extensions are allowed:

<40>

The same verb may also be used as an active (transitive) verb:

<41>

The various semantic intricacies of the derivational system of Dative marking for the Eastern Nilotic language Turkana is typical for the subgroup to which it belongs, the Teso-Turkana cluster. Depending on the event structure expressed by a specific verb, the Dative suffix in

Teso-Turkana expresses the presence of semantic roles like Beneficiary, Malefactive, Recipient, Comitative, or Location. There is an additional shade of meaning attested in Teso-Turkana, but not apparently elsewhere in Nilotic: Intransitive stative verbs in this subgroup may be expanded with a Dative suffix in order to express an affective meaning:

<42>

This semantic extension probably is an innovation of the Teso-Turkana cluster.  [3] The semantic widening may be one reason for the high productivity of Dative marking, which correspondingly has a high frequency in Turkana, much higher it would seem than in Maasai or other Nilotic languages. One such indication comes from a comparison, for example of Bible translations and the frequency with which Datives occur in the translated texts in Turkana as against Maasai.

<43>

There are also interesting differences between Teso-Turkana and other Nilotic languages in the assignment of case in combination with the Dative and in combinatory possibilities between the Dative and other verbal extensions and their relative order, as well as in the order of constituents following the (Dative) verb, as shown next.

<44>

Nilotic languages are usually divided into three major subbranches, following Köhler (1955):

<45>

In the brief survey below, an attempt is made to illustrate Dative constructions with representatives from all major subgroups within each of the three Nilotic branches, starting with the closest relatives of Teso-Turkana, i.e. with Eastern Nilotic languages like Maasai and Bari.

<46>

In an early, highly sophisticated, study of the grammatical structure of a Nilotic language, the description of Maasai by Tucker and Mpaayei (1955), it is pointed out (p.130) that the main sense of the Dative verb is "…action for somebody else…It may also imply action directed to a certain point or goal"; of course, this is exactly what the Dative in Teso-Turkana expresses. The incorporation of direction or location in Maasai is illustrated in the following example:

<47>

As this example shows, the preposition t(ɛ)- is absent when the Dative extension occurs on the verb. This alternation parallels the use versus omission of the preposition à / kà in Teso-Turkana (but only in combination with non-lexicalised Dative verbs in this latter group, as we saw above). In their detailed account, Tucker and Mpaayei (1955) further point out that the Dative object behaves syntactically like an object. As example (48) shows, Maasai treats the recipient as a kind of (secondary) object, i.e. cross-referencing occurs for 'me'.

<48>

Again, this parallels the situation in Teso-Turkana above, where the secondary object is expressed on the verb in the case of first and second person. Also, the secondary object in Maasai precedes the primary object, as in Teso-Turkana.

<49>

However, contrary to Teso-Turkana, Maasai does not have Locative case marking. Consequently, location-marking complements which are coindexed with the verbal Dative marker take the same case form as complements expressing recipients or benefactives, i.e. Absolutive case, in Maasai.

<50>

Compare this configuration to the corresponding translation in Turkana, which uses verb forms that are cognate:

<51>

Note that the locative noun 'on the ground/down' in the Turkana example (50) lacks a Locative gender prefix. This is due to the fact that a number of nouns, including 'ground', 'house', 'mountain', 'well' constitute lexical exceptions (Dimmendaal 1983:348). However, when other nouns were to be used in order to express location, these would take Locative case in Turkana. For example, 'into the bottle' in an example like (51) would be lɔ-cʋ ̀pà (the Absolutive case form being ɛ-cʋ̀pà 'bottle'). In other words, contrary to Teso-Turkana, incorporated Locative phrases in combination with a Dative verb take exactly the same case marking as animate entities functioning as secondary objects in languages like Maasai, i.e. they take Absolutive case, whereas in Teso-Turkana a case split occurs between Absolutive case for Recipients, Benefactives, Malefactives, and Locative case for location-marking complements.

<52>

It is not immediately obvious from a comparison of Maasai with Teso-Turkana, which languages underwent a grammatical reinterpretation in this case. But an intragenetic comparison of case systems in Nilotic clearly shows that the Maasai system is the common norm elsewhere in the Nilotic family, and that the Teso-Turkana languages apparently developed a distinction between Absolutive case (as used for primary and secondary objects) versus Locative case. Historically, the distinction appears to have resulted from the prefixation of a definiteness marker *ni-, which fused with the gender prefix of the following noun (Heine and Vossen 1983). This prefixation apparently only occurred with nouns expressing core functions like subject and (primary or secondary) object, but not with Locative-marking complements.

<53>

Consequently, the cognate forms for Maasai Absolutive gender prefixes are cognate with Locative gender-prefixes in Teso-Turkana. The common historical origin of gender prefixes used in Locative case frames in Teso-Turkana and the Absolute case forms elsewhere in Eastern Nilotic becomes even clear when comparing these with reflexes in Eastern Nilotic languages like Lotuxo, which forms a genetic unit with Maasai called Lotuxo-Maa by Vossen (1982).

<54>

In the Bari group within Eastern Nilotic, gender marking is covert, i.e. only shown on categories modifying the noun. Compare the agreement markers (masculine singular) and (feminine singular) in the following example.

<55>

A case distinction between (core) arguments expressing semantic roles such as recipient and beneficiary on the one side and arguments expressing location (or direction) is also common in languages bordering on the Teso-Turkana area, and belonging to the Didinga-Murle cluster within Surmic (Nilotic’s closest relatives within the Eastern Sudanic branch of Nilo-Saharan). The innovation in Teso-Turkana may thus have come about through areal contact. There is ample lexical and phonological evidence for convergence between Teso-Turkana and Didinga-Murle, as argued in Dimmendaal (1982, 1998, 2005).

<56>

The Dative marker in the more distantly related Eastern Nilotic language Bari is cognate with the Teso-Turkana form, and covers a semantic range comparable to the Dative in Teso-Turkana and Maasai. Nyombe (1987:185) refers to this suffix as the Applicative. Compare:

<57>

Nyombe uses the standard approach of Generative grammar, following Baker (i.e. his 1985 MIT-dissertation; published in a more extensive form as Baker 1988), by arguing that Dative marking on the verb involves "head movement" (incorporation of an "empty preposition"). The latter licenses oblique arguments without the use of a preposition or oblique case. However, at closer inspection the actual distribution of the Dative suffix in Bari also appears to be more complex, as it turns out to be in Teso-Turkana. Whereas prepositions tend to be absent with Dative marking (as constructions of this type are called here) on the verb, they do nevertheless occur, as in Turkana. This becomes evident from other examples in Nyombe (1987). The same morpheme, for example, is also used in order to express direction. Nyombe (1987:38) characterises this latter suffix as a directional marker, although from a formal point of view it is identical to the Dative marker (or Benefactive marker in Nyombe’s terminology). As we saw above for Turkana and Maasai, the incorporation of a locative meaning in combination with this suffix is widespread in these languages. But as illustrated above, prepositions and Dative extensions are not mutually exclusive in Bari either, as the following example shows:

<58>

It therefore is clear that Dative marking and the use of prepositions requires further investigation also for this Eastern Nilotic language.

<59>

The Dative in Southern Nilotic Kalenjin expresses a Recipient, according to Creider & Creider (1989:90). This construction is only possible for third persons, and nouns; when participants (I, we, you (sg/pl)) function as recipients, Nandi and other Kalenjin languages use a Ventive extension. Compare Southern Nilotic Nandi (data from Creider & Creider 1989).

<60>

Data from Southern Nilotic thus confirm the hypothesis that the presence of the Dative extension has consequences for the thematic structure in a sentence, parallel to the Turkana case discussed above. As pointed out by Creider & Creider (1989), the a-form in the final example occurs when Chebet is the focus of the assertion (or is part of the assertion). In the b-form, Chebet is presented as background information ('What did you do for Chebet?'). This parallels the situation described for Eastern Nilotic above, where Dative marking also corresponds to the presence of some given entity (expressing a Recipient, Beneficiary, or a Location which is either known to the speaker and hearer); alternatively, such predications express thetic statements (in the sense of Sasse 1987). This semantic dimension is also expressed by the Dative in the Keyo lect within Kalenjin, as shown by Mietzner (2009). But as further shown by Mietzner (2009), the same marker (which is referred to as the Applicative by her) is also used in Keyo in order to express direction. Contrary to Eastern Nilotic Turkana, the locative complement takes Absolutive case. And contrary to Eastern Nilotic languages, this noun or noun phrase expressing location can occur immediately after the verb.

<61>

This pattern fits in with observations made by Creider (1989) on the relatively free ("non-configurational") post-verbal constituent in Kalenjin. The constituent immediately following the verb ('water' in the example above) carries assertive focus, i.e. provides important new information in Kalenjin.

<62>

From a syntactic as well as from a semantic point of view, then, there are some interesting differences between Southern Nilotic Nandi (and other Kalenjin languages) and Eastern Nilotic. The Kalenjin languages within Southern Nilotic are verb-initial, as in Turkana or Maasai, but contrary to these latter languages post-verbal constituent order in Kalenjin is free from a syntactic point of view. Nevertheless, the relative order of post-verbal constituents does have consequences for information packaging in a clause, i.e. for thematic structure. Constituent order in languages like Turkana (or Maasai) is far more restrictive, with VSO being the common pattern. VOS order occurs only when objects are pronominal and express a topicalised object. Phrased differently, syntactic order is freely adapted towards the need of focus structure in Southern Nilotic Kalenjin, whereas in Turkana (or Maasai for that matter) the focus system has to adapt to a more rigid constituent order, e.g. by putting focused object constituents in preverbal position (Dimmendaal 1983b).  [4] In spite of the fact that Southern Nilotic Kalenjin and Eastern Nilotic languages like Turkana or Maasai use similar case marking systems for core constituents (Nominative versus Absolutive), and similar cross-referencing systems for pronominal subject and object on the verb as well as cognate valency-changing (derivational) suffixes, they differ in terms of rigidity of constituent order. One reason for this may be the fact that the Southern Nilotic Kalenjin system of verbal derivation is more extensive than the Eastern Nilotic system. One such suffix which appears to have been lost in Eastern Nilotic is the Simulative. As argued next, its former presence may help to explain the complex allomorphy rules for the Dative in Turkana and other Eastern Nilotic languages described above.

<63>

The Dative in Eastern Nilotic languages further involves the presence versus absence of a nasal consonant in specific, morphologically conditioned, contexts, as illustrated above. In Southern Nilotic Kalenjin the Dative marker basically has one form, -ci, whereby the initial consonant is palatalised before a high front vowel historically (-ci < *-ki). This derivational suffix is frequently combined with a simulative marker in Kalenjin languages like Kipsikiis (cf. Toweett 1979):

<64>

This morphosyntactic device left a formal trace (through the presence of a nasal consonant -n) but the Simultaneous marker disappeared as a semantic device in Eastern Nilotic languages, resulting in the morphologization of the alternation in certain paradigms (imperfective against perfective verb forms) in the Eastern Nilotic subgroup within Nilotic. Thus:

(Loss of vowel length is a well-attested innovation of Eastern Nilotic; Dimmendaal 1988.)

<65>

Kalenjin verb derivation is more complex than Eastern Nilotic verb derivation, also because of the possibility of hosting multiple Datives, as in Kipsikiis (data from Toweett 1979):

<66>

Given the double semantic role of the Dative (namely thematic incorporation of a Recipient/ Benefactive/Malefactive or Locative), and post-verbal "scrambling" of constituents linked thematically to the verb, it is now possible to understand structures such as (61). Here, one Dative marker expresses the first role (Recipient), whereas the second Dative marker expresses Direction.

<67>

In Western Nilotic languages Datives are usually expressed through internal morphology (e.g. consonant alternation and vowel change). Historically, this system again appears to be cognate with the suffixation system in Eastern and Southern Nilotic. Andersen (1988) reconstructs a Dative suffix *-VC with a [+ATR] vowel quality in pre-Päri. Note also that in quite a few instances a high front vowel appears in such Dative constructions synchronically in this Western Nilotic language. This makes it extremely likely that the alternation goes back to a common suffix *-ki in Proto-Nilotic. An example from Päri (based on Andersen 1988):

Similar reflexes, again usually involving root-internal vowel and consonant alternations, are attested in Western Nilotic languages like Lango (Noonan 1992).

<68>

Dative functions include the expression of Malefactive as an additional shade of meaning, in Lango:

<69>

Whereas the Dative mainly involves internal morphology (i.e. consonant and vowel alternation in the root) in languages like Päri and Lango, there are a number of Western Nilotic languages where the Dative has disappeared altogether. The reason why Western Nilotic languages like Luo lost the Dative appears to be phonological in nature: The loss of gemination as a phonological phenomenon in this language resulted in neutralisation of formal distinctions in examples such as the root form as against the Dative form of 'send' above. Instead, Luo uses a periphrastic (prepositional) strategy in order to express a beneficiary role:

<70>

Interestingly, pronominal complements introducing a Beneficiary or a Recipient role in Luo encliticise onto the verb, i.e. headward migration results in the re-emergence of bound morphemes in this Western Nilotic language. This development again may be due to areal influence from neighbouring Bantu languages; there is lexical and grammatical evidence that Luo converged towards neighbouring Bantu languages referred to as the Suba group (Dimmendaal 2001). This development strongly suggests that languages may go through a cycle, whereby the same semantic notions are incorporated into the verbal meaning again.

<71>

Nilotic languages using the latter valency-changing device with verbs always have a Dative marker. But the inverse system, instrumental marking without Dative marking, does not seem to be attested. Accordingly, the Dative appears to be more stable as a verbal derivational marker than more peripheral role markers such as the instrumental suffix. As a head-marking strategy Dative-marking probably predates Nilotic. Cognate forms are found elsewhere in Eastern Sudanic, for example in the closely related Surmic languages (compare Randal 1998 on Tennet), and possibly beyond (cf. Heine 1990).

<72>

In his study of Dative constructions in Austronesian, Donahue (2001) observes that peripheral constituents such as prepositional phrases or adverbs may be peripheral in one sense, but they may become part of the core information structure of a verb, without necessarily becoming a direct object.  [5] It is in this latter type of context that Austronesian languages apparently tend to use a Dative marker. And it is exactly in this kind of configuration that Turkana and other Nilotic languages use a verbal Dative marker. In semantic terms, the promoted referent involved is either an animate entity (mostly human) affected positively or negatively by a verbal event, or a location. Such a referent must be identifiable to the addressee; more specifically, following the terminology of Chafe (1987), the cognitive status of the referent(s) in the discourse must be active (i.e. it is the current focus of consciousness), accessible (i.e. it is textually, situationally or inferentially available) or inactive (involving the hearer’s long-term memory). The Dative marker thus helps to tailor information structure in these languages, more specifically the pragmatic state of referents in sentences in the minds of speech participants. Zero marking referents are highly accessible. Consequently, the frequent use of zero anaphora in combination with Datives in Turkana and elsewhere in Nilotic, either in the case of secondary objects or with locative complements, is to be expected. The syntactic split between secondary object and locative positions appears to be a reflection of a more general property of Turkana and other members of the Teso-Turkana cluster, the operation of a Prominence Hierarchy. This hierarchy affects cross-reference marking on verbs, constituent order amongst others (Dimmendaal 1983:83-88).

<73>

In terms of information packaging, Dative formation in Turkana is part of a system of verbal-derivational marking, which also includes ventive and itive marking. (For a description of Ventive and Itive locative marking along these lines, see Dimmendaal 2003); Turkana has lost instrumental marking on main verbs, a derivational property still attested in closely related languages like Toposa or Maasai. The presence of any one of these extensions in combination with a verb affects the thematic structure of a clause. When Dative marking on the verb occurs, for example, the corresponding secondary object or locative complement is part of the pragmatic presupposition (or the focus of assertion) of the clause. Lexicalised Dative constructions in Turkana constitute grammaticalised or conventionalised strategies for information packaging (Dimmendaal 2003). On the other hand, Dative constructions in Turkana which are derived from a verb root which is still used as such express less expected or unexpected event structures, involving the transfer of a Figure along a Path (in the sense of Talmy 1985) either physically or metaphorically, and affecting either an animate (most often a human) Recipient etc., or, alternatively, some Location or Ground whose cognitive status in the discourse is not active.

References

Andersen, Torben 1988a

'Ergativity in Päri: A Nilotic OVS language.' Lingua 75:289-324

Andersen, Torben 1988b

'Consonant alternation in the verbal morphology of Päri.' Afrika und Übersee 71:63-113

Baker, Mark 1988

Incorporation: A Theory of Grammatical Functioning Changing. Chicago: University of Chicago Press

Chafe, Wallace L. 1987

'Cognitive constraints on information flow.' In: Russell S. Tomlin (ed.), Coherence and Grounding in Discourse, pp.21-52. Amsterdam and Philadelphia: John Benjamins

Creider, Chet A. 1989

The Syntax of the Nilotic Languages: Themes and Variations. Berlin: Dietrich Reimer

Creider, Chet A. and Jane Tapsubei 1989

A Grammar of Nandi. Hamburg: Helmut Buske

Dimmendaal, Gerrit J. 1982

'Contacts between Eastern Nilotic and Surma groups: linguistic evidence.' In: John Mack and P. Robertshaw (eds.), Culture History in the Southern Sudan, pp.101-10. Nairobi: B.I.E.A.

Dimmendaal, Gerrit J. 1983a

'Turkana as a verb-initial language.' Journal of African Languages and Linguistics 5,1:17-44

Dimmendaal, Gerrit J. 1983b

The Turkana Language. Dordrecht: Foris Publications

Dimmendaal, Gerrit J. 1988

'The lexical reconstruction of Proto‑Nilotic: A first reconnaissance.' Afrikanistische Arbeitspapiere 16:5-67

Dimmendaal, Gerrit J. 2001

'Language shift and morphological convergence in the Nilotic area.' Sprache und Geschichte in Afrika 16/17:83-124

Dimmendaal, Gerrit J. 2003

'Locatives as core constituents.' In: Erin Shay and Uwe Seibert (eds.), Motion, Direction and Location in Languages: In Honor of Zygmunt Frajzyngier, pp.91-109. Amsterdam and Philadelphia: John Benjamins

Dimmendaal, Gerrit J. and Anneke Breedveld 1986

Tonal influence on vocalic quality. In: Koen Bogers et al. (eds.), The Phonological Representation of Suprasegmentals, pp.1-33. Dordrecht: Foris Publications

Donohue, Mark 2001

'Coding choices in argument structure: Austronesian applicatives in texts.' Studies in Language 25,2:217-254

Heine, Bernd 1990

'The Dative in Ik and Kanuri.' In: William Croft et al. (eds.), Studies in Typology and Diachrony, pp.129-149. Amsterdam and Philadelphia: John Benjamins

Köhler, Oswin 1955

Geschichte der Erforschung der nilotischen Sprachen. Afrika und Übersee, Beiheft 28

Langacker, Ronald W. 1987

Foundations of Cognitive Grammar Vol 1. Theoretical Prerequisites. Stanford: Stanford University Press

Mietzner, Angelika 2009

Applikativ und „boundary crossing“ im Bantu und im Nilotischen. Unpublished manuscript

Noonan, Michael 1992

A Grammar of Lango. Berlin: Mouton de Gruyter

Nyombe, Bureng G.V. 1987

Argument-bearing Affixes in Bari. Ph.D. dissertation, CUNY

O’Herin, Brian 2001

'Abaza applicatives.' Language 77,3:477-493

Peterson, David A. 2007

Applicative Constructions. Oxford: Oxford University Press

Sasse, Hans-Jürgen 1987

'The thetic/categorical distinction revisited.' Linguistics 25:511-580

Talmy, Leonard 1985

Lexicalization patterns: semantic structure in lexical forms. In: Joseph H. Greenberg (ed.), Universals of Human Language: Vol.4 Syntax, pp.625-649. Stanford: Stanford University Press

Toweett, Taaitta 1979

Kalenjin Linguistics. Nairobi: Kenya Literature Bureau

Tucker, Archibald N. and J. Tompo Ole Mpaayei 1955

A Maasai Grammar with Vocabulary. London: Longmans, Green and Co.

Van Valin Jr., Robert D. and Randy J. LaPolla 1997

Syntax: Structure, Meaning, and Function. Cambridge: Cambridge University Press

Vossen, Rainer 1982

The Eastern Nilotes: Linguistic and Historical Reconstructions. Berlin: Dietrich Reimer

____________________________

* The present paper emerged from a research project Nr. HE 574/31-1 on Participant Marking in African lan guages, financed by the DFG (German Research Foundation). I would like to thank the DFG for making this project possible. Special gratitude is expressed to Angi Mietzner and Helma Pasch for their comments on an earlier draft of this paper, and to Marisa Weiner for the editorial work.