The Syntax of the Reiterated Numeral Construction1

advertisement

The Reiterated Numeral Construction

*

Jason Kandybowicz

Rutgers University

1.0 Introduction

Numerals in Nupe 1 are able to function as two kinds of modifiers. While in certain syntactic environments they can modify nominal phrases, as in English, in others they serve as adverbial modifiers. Consider the following sentences.

(1) a. Bàági gútá ba nakàn.

men three cut meat

‘Three men cut the meat.’

b. Musa ba nakàn gútá.

Musa cut meat three.

‘Musa cut three pieces of meat.’ c. Musa ba nakàn gútá-gútá.

Musa cut meat three-three

‘Musa cut the meat in threes.’

(e.g. ‘Musa cut the (plurality of) meat in batches of three.’)

This paper is concerned primarily with the syntactic issues raised by numeral constructions such as those in (1c.) above, an under-researched phenomenon virtually undiscussed in the literature 2 . It will be argued that in this construction the numeral and its copy form a constituent which is right-adjoined to the VP. I will refer to this numeral complex as the Reiterated Numeral ConstructionRNC hereafter. After sufficient discussion of Nupe’s inventory of adverbs and adverbials, the claim that RNCs admit of the category label adverb will be justified and defended. The paper concludes with an investigation into the licensing constraints governing RNCs.

2.0 RNC Constituency

In light of the fact that the presence of a copied numeral gives rise to a difference in meaning between (1b.) and

(1c.), inquiry into the distinction between such numeral constructions as / gútá / and / gútá-gútá / seems well warranted. Given the adjectival interpretation of the numeral in (1b.) and the seemingly adverbial counterpart interpretation of the numerals in (1c.), two possibilities regarding the nature of these numeral arrays become evident.

The first possibility is that gútá

1

and gútá

2 are morphologically distinct. As such, the proposal here is that the first occurrence of gútá right adjoins to the NP housing nakàn as an adjectival adjunct in both (1b.) and (1c.), while the

1

copied gútá adjoins as an adjunct to the VP comprising [ ba nakàn ] in (1c.). In this way, the adjectival reading of

(1b.) is preserved and the adverbial interpretation of (1c.) is accounted for. Schematically, this strategy for representing the occurrence of the twin numerals in (1c.) as distinct from the occurrence of the solitary numeral in

(1b.) is given below in (2).

(2) VP

VP XP

NP V´ gútá

2

Musa V NP

ba NP XP

nakàn gútá

1

The remaining possibility, however, offers a less striking account of this distribution. On this version, the treatment of the numeral in (1b.) retains the status proposed in possibility one. The handling of the numeral reduplicant in

(1c.) however, differs saliently on this account. Possibility two suggests that both lexical items (the numeral and its copy) are base-generated as VP adjuncts. Morphologically, this amounts to the claim that RNCs are built up from a process of reduplication, the target of which is the numeral stem 3 . This possibility is represented below in (3).

(3) VP

VP XP

Musa ba nakàn gútá

1

-gútá

2

Stated as such, possibility one claims that despite being lexically identical and synonymous, gútá

1

and gútá

2

are morphologically distinct and admit of separate identities under NP and VP. Possibility two is less adventurous because it proposes that neither gutá

1 nor gutá

2 has an identity without the other, that is, both numerals are bound to each other as a morphosyntactic unit. As it turns out, the later proposal is to be favored. To convince ourselves of this, we turn now to a series of constituency tests involving intervention and dislocation in Nupe.

2.1 Intervention as Test for Constituency

As a preliminary measure and segue into our discussion on intervention, I would like to justify the use of the term

Reiterated Numeral Construction . As we will see, this term is an accurate description of the phenomena under scrutiny. Note that the following data confirm the claim that the construction must be built up out of a) the numeral root and b) a single copy of that root.

2

(4) a. *Musa ba nakàn gúbà-gútá.

Musa cut meat two-three b. *Musa ba nakàn gútá-gútá-gútá.

Musa cut meat three-three-three

Thus, the morphological doubling operation involved in the derivation of RNCs admits of a single application of reiteration and permits no more. Interestingly enough however, while strings of exactly three identical numerals occurring in a given sentence result in ungrammaticality, the insertion of a pause between the first occurrence of the numeral and the remaining two occurrences can save the grammaticality of the sentence. As we will see, this fact becomes important in establishing the constituency of the RNC. Moreover, in any given sentence, all strings of iterated numerals greater than three in length give rise to judgments of ungrammaticality by native speakers. Thus, given the proper morphosyntactic conditions, a numeral can co-occur with a copy of itself. Additional numerals, however, be they identical or not are prohibited from intervening in the resulting complex housing the original numeral and its copy, unless suprasegmental perceptual cues such as pausing are imposed. That is, once an RNC has been established the intervention of subsequent numerals either internal to or external to the complex is disallowed. The question that now remains is whether any lexical item besides the numeral can intervene to break apart the Reiterated Numeral Construction. If this question can be answered in the negative, then the beginnings of a basis for claiming that RNCs are constituents and morphological wholes (along the lines of possibility two) will start to take shape. Toward the goal of answering this question, consider the following:

(5) a. * Musa ba nakàn gútá tó gútá.

Musa cut meat three and three b. * Musa ba nakàn gútá tsúwó gútá.

Musa cut meat three yesterday three c. * Musa ba nakàn gútá sanyin gútá.

Musa cut meat three quietly three d. * Musa ba nakàn gútá dan efo cigban o gútá

Musa cut meat three in hole tree LOC three

These four sentences represent plausible candidates for intervention, one might think. Indeed, the insertion of adverbs and prepositional phrases similar to those italicized above might be so placed by a non-native speaker attempting to approximate the following glosses for (5b.)-(5d.) respectively:

3

(5b') ‘Musa cut the meat in threes yesterday.’ or ‘Musa cut the meat yesterday in threes’.

(5c.') ‘Musa cut the meat in threes quietly.’ or ‘Musa cut the meat quietly in threes.’

(5d.') ‘Musa cut the meat in threes under the tree.’ or ‘Musa cut the meat under the tree in threes.’

All other attempts to split the original numeral from its copy cause the resulting sentence to crash. This finding falsifies the prediction that possibility one gives rise to; namely, that the loose bond holding gútá

1

and gútá

2 together will allow for intervening structure between the two items. Hence, the fact that there is no lexical or phrasal intervention of any sort within the Reiterated Numeral Construction is the first damaging blow to the possibility that both numerals are morphosyntactically independent.

2.2 Dislocation as Test for Constituency

In this section, convincing evidence that RNCs are constituents will be presented by means of three phenomena: i.

RNCs can be clefted ii.

RNCs can be stranded iii.

RNCs are conjoinable and adhere to the Coordinate Structure Constraint

2.2.1 RNC Clefts

Compelling evidence that arrays such as / gútá-gútá / form a constituent comes from their ability to move as a unit. In what follows, observe that whereas / gútá-gútá / as a whole can be fronted, gútá

1

independently of gútá

2 cannot. This is not predicted to occur under possibility one.

(6) a. Musa ba nakàn gútá-gútá.

Musa cut meat three-three

‘Musa cut the meat in threes.’ b. Gútá-gútá Musa ba nakàn o.

three-three Musa cut meat FOC

‘It was in threes that Musa cut the meat.’

c. *Gútá Musa ba nakàn gútá o.

three Musa cut meat three FOC

Perhaps more striking is the example alluded to in section 2.1.

4

(7) Musa ba nakàn gúbà (…) gútá-gútá.

Musa cut meat two (pause) three-three

‘Musa cut two pieces of meat in threes.’

Again, in production and perception among native speakers, a slight pause exists between /g úbà / and / gútá-gútá /, serving to demarcate the boundary between the adjectival /gúbà / and the RNC. While this certainly adds to the claim that RNCs are produced and perceived as units, the pattern of movement resulting from the transformation of

(7) (enumerated below in (8)) reaffirms the hypothesis that RNCs move as a complex. Consider the following:

(8) a. Gútá-gútá Musa ba nakàn gúbà o.

three-three Musa cut meat two FOC

‘It was in threes that Musa cut two pieces of meat.’

b. * Gúbà-gútá-gútá Musa ba nakàn o.

two-three-three Musa cut meat FOC

Under the assumption that transformational rules can only dislocate constituents, the grammaticality of (8a.), the ungrammaticality of (8b.), and the existence of the intervening pause between / gúbà/ and / gútá-gútá / all serve to drive home the fact that / gúbà-gútá-gútá / does not form a constituent, unlike [ gútá-gútá ], which does.

2.2.2 RNC Stranding

The derivation involved in the question below and the subsequent stranding of the RNC are sufficient conditions for RNC constituency-hood. In what follows, the underlying representation I assume is given by (9a.) and two outputs of wh-movement by 9b. & 9c.

(9) a. Musa ba ké gútá-gútá.

Musa cut what three-three b. Ké Musa ba gútá-gútá o.

what Musa cut three-three FOC

‘What (did) Musa cut in threes?’ c. * ké gútá-gútá Musa ba o.

what three-three Musa cut FOC

5

The sentences in (9) take on a two-fold significance. On the one hand, the fact that / gútá-gútá / can be stranded, as

(9b.) plainly shows, further substantiates the claim that the RNC is a tight compact unit. On the other hand, (9c.) crucially establishes that the RNC does not form a constituent with ké –‘what’. Thus, given this and previous facts, we conclude that RNCs are constituents comprised of a numeral root and its copy, no more no less.

2.2.3 RNCs and Coordinate Structures

Further evidence that RNCs form constituents stems from the fact that, one, the original numeral together with its copy are conjoinable with other pairs of numerals and copies and two, that in such a conjoined state, they adhere to the Coordinate Structure Constraint (Ross 1967). Let us first examine the claim that RNCs are conjoinable. Note the following examples:

(10) a. Musa ba nakàn gútá-gútá tò gúni-gúni.

Musa cut meat three-three and four-four

‘Musa cut the meat in batches of three and four.’

b. Musa ba nakàn gúni-gúni tò gútá-gútá.

Musa cut meat four-four and three-three

‘Musa cut the meat in batches of four and three.’

Sentences (10a.) and (10b.) make it perfectly clear that RNCs can both be conjoined and reversed. Furthermore, from this fact we can conclude that both conjuncts are themselves constituents and are characteristically of the same categorical type as each other. For evidence that both conjuncts together comprise a larger constituent consisting of the conjunction of both RNC constituents, we turn to the following examples which exploit the results of section

2.2.1 concerning RNC clefts:

(11) a. * Gútá-gútá Musa ba nakàn o tò gúni-gúni.

three-three Musa cut meat FOC and four-four

b. Gútá-gútá tò gúni-gúni Musa ba nakàn o.

three-three and four-four Musa cut meat FOC

‘It was in batches of three and four that Musa cut the meat.’

As is evident from (11a.), the clefting transformation cannot operate on one conjunct alone. This violation of the

Coordinate Structure Constraint confirms that the conjunction of the RNCs in (11b.) is itself a constituent of the same type as the RNC sub-constituents which compose it. That is, the constituency of [[ gútá-gútá ] tò [ gúni-gúni ]] is

6

a function of the constituency of [ gútá-gútá ] and [ gúni-gúni ], both Reiterated Numeral Constructions. Given these facts, we are once again justified in asserting that RNCs are morphologically and syntactically unitary.

3.0 Adverbs and Adverbial Modification in Nupe

Having established the constituency of the RNC, we now turn our attention to an investigation of the construction’s modificational and distributional properties. As was alluded to in the introduction, RNCs are fundamentally adverbial modifiers in Nupe. Demonstrating that this is the case is the goal of this section’s argumentation. Toward this aim, a variety of adverbials will be considered, resulting in the proposal that adverbs in

Nupe admit of a partition into two distinct adverb classes; inflectional type adverbs (I-types) and phrasal adverbs

(P-types), each of which possess unique distributional properties. Reflecting on the distribution of the RNC, it will be argued that Reiterated Numeral Constructions are instances of phrasal adverbs , the more productive of Nupe’s two adverb classes. We begin our investigation with a survey of adverbial modifiers in Nupe, which in addition to being split into two distributional types (I & P-types) take on one of two different morphological forms; single word modifiers , or multi-word (phrasal) modifiers .

3.1 Single Word Modifiers

Quite generally, adverbs in Nupe conform to the following generalization; when limited to one word, they are primarily manner or temporal adverbs. Multi-word modifiers in the language, however, in addition to their role as manner and temporal adverbs, function to encode the remaining varieties of adverbial modification that single word modifiers fail to represent, namely instrumental and locative modification (among others). Again comparatively speaking, in terms of frequency of occurrence and sheer breadth of expressability, single word modifiers are slightly

(though not considerably) less productive than multi-word modifiers. Syntactically, single word modifiers differ among themselves with respect to the site at which they attach to VP. That this variance is indeed regular and predictable will be elaborated on in the sections dealing with I & P-type adverbs. Consider the following representative examples of single word modifiers, noting the variation in syntactic position of the adverbs, as well as the absence of any form of modification other than manner and temporal:

(12) a. Gana tsò tangi karayín.

Gana set trap carefully

‘Gana set the trap carefully .’

b. Gana dada du cènkafa.

Gana quickly cooked rice

‘Gana quickly cooked the rice.’ c. Gana à du cènkafa káńdóńdò.

Gana FUT cook rice always

‘Gana will always cook rice.’

7

d. Esìgi è tséká gbó.

Dog PRS sometimes bark

‘The dog barks sometimes .’

3.2 Multi-Word Modifiers

Not to be confused with phrasal-type adverbs, which are defined in terms of their distributional properties (to be discussed in the upcoming section), multi-word modifiers are named solely in virtue of the fact that as units larger than a word they modify VPs. Perhaps if RNCs did not exist in Nupe a more suitable name for these adverbials would be PP-modifiers, since the preponderance of multi-word modifiers are in fact almost exclusively prepositional phrases. However, sensitivity to the likelihood that additional phrasal adverbial modifiers might be discovered (as with the discovery of RNCs) has led me to opt in favor of the more conservative terminology. Given this PP-like tendency, such modifiers are able to take on a number of modificational roles including those of manner, instrumental, and locative adverbials. Furthermore, the pervasiveness of the PP-modifier as the dominant multiword modifier is cashed out in the fact that as a class, they agree with respect to the site of VP adjunction. That is, all multi-word adverbials are distributed identically. Bearing these properties in mind, note the following sentences which highlight multi-word modification.

(13) a. Gana nì eyàn bè kágbó yin.

Gana played drum with strength

‘Gana played the drum forcefully .’ b. Gana ba nakàn bè èbi yin.

Gana cut meat with knife

‘Gana cut the meat with the knife .’ c. Gana tsò tangi dan efo cigban o.

Gana set trap in hole tree LOC

‘Gana set the trap under the tree .’

3.3 Nupe Adverb Classes

O.T. Stewart (1996, 1998) claims that two kinds of adverbs exist in Edo (a geographically related language spoken in mid-western Nigeria); I-type adverbs and N-type adverbs . I-type (inflectional type) adverbs exhibit signs of being tense-related (inflection for tense is borne by the tone-bearing units of such adverbs, as in verbs), while Ntype (noun type) adverbs parallel the behavior of Edo nominals in that they do not tonally inflect for tense. As an additional point of comparison, he notes that whereas I-type adverb adjunction is preverbal , N-type adjunction is postverb phrasal . I propose that a similar distinction holds for adverbs in Nupe. Building on the lower level division into adverb classes developed in 3.1 and 3.2, we are now in a position to add to our Nupe adverb taxonomy.

8

Through an analysis of the licensing and distribution of the various flavors of adverbials introduced in the previous two sections, a more restrictive partition among adverb types can be discerned. This will ultimately give rise to a means for defending the claim that RNCs are fundamentally adverbial modifiers.

3.3.1 I-Type Adverbs

More so than the I-type adverbs of Edo, Nupe I-types are deeply INFL-related, although they are not inflected for tense. Distributionally, however, in certain respects the two adverb classes converge in both languages 4 . The fundamental properties of I-type adverbs in Nupe are given in (14).

(14) A. I-type adverbs follow INFL and occur pre-verbally.

B.

I-type adverbs cannot be fronted.

C.

I-type adverbs are licensed local to INFL.

(14) A. and B. further the claim that the distribution of I-type adverbs in Nupe is similar to the distribution in Edo.

To substantiate this proposal, consider the following set of sentences.

(15) Edo

a. (*Giégié) Òzó giégié rrí èvbàré (*giégié).

(*quickly) Ozo quickly eat food (*quickly)

‘(*Quickly) Ozo quickly ate the food (*quickly).’ (Stewart 1996)

b. * Òzó kó!kó giégié ògó

Ozo gather quickly bottle (Stewart 1998)

Nupe

c. (*Dàdà) Musa dàdà gí cènkafa (*dàdà).

(*quickly) Musa quickly ate rice (*quickly)

‘(*Quickly) Musa quickly ate the rice (*quickly).’

d. * Musa gí dàdà cènkafa.

Musa ate quickly rice

e. (* Tséká) Musa è tséká gí cènkafa (*tséká).

(*sometimes) Musa PRS sometimes eat rice (*sometimes)

‘(*Sometimes) Musa sometimes eats rice (*sometimes).’

9

f. * Musa gí tséká cènkafa.

Musa ate sometimes rice

Given the data in (15), it is clear that the distribution of I-type adverbs in Nupe parallels Edo’s I-type distribution.

Specifically, I-types (which in Edo and Nupe are strictly single-word modifiers) cannot follow the verb or its object in either language. Rather, they seem to be post-INFL heads which take VP complements, similar to auxiliary verbs in English. Furthermore, it is evident that I-type adverbs in neither language can be preposed (cf. 15a,c).

The symmetry between Edo and Nupe I-type adverbs breaks down, however, when we consider adverb placement non-local to INFL. Given that I-types occupy a pre-verb phrasal position, if we want to examine them distanced from INFL we must have recourse to a sentence with a structure admitting of two VPs; one situated in the typical post-INFL position and the other embedded somewhere below the first. Fortunately, there are such constructions in Nupe, namely, Serial Verb Constructions (SVCs). Thus, we can ascertain the nature of the I-type locality restriction by examining the distribution of I-type adverbs in the Serial Verb Constructions of both languages. Informally, Serial Verb Constructions can be characterized as a complex consisting of two or more finite verb phrases, occurring without marking of coordination or subordination. Although it has been proposed that SVCs themselves admit of a typology (Stewart 1998), I will restrict my discussion of SVCs to the construction that Stewart dubs the Consequential Serial Verb Construction (CSVC). When two verbs are serialized in this construction 5 , V1 and V2 are transitive verbs sharing the same subject and object. As a consequence of this, V2 appears on the surface to be devoid of its object argument, but actually assigns its object theta role to the NP intervening between V1 and

V2. In this way, CSVCs achieve object sharing (Baker 1989). The fundamental feature of the CSVC is that the verbs of the construction are temporally ordered in the following way: the event denoted by VP1 precedes the eventuality denoted by VP2 and so on for the other occurrences of verbs. As such, these verbs and verb phrases denote discrete events, which together compose the macro event denoted by the matrix VP. The data in (16) illustrate the CSVC in Edo and Nupe.

(16) Edo

a. Òzó lé èvbàré ré

Ozo cook food eat

‘Ozo cooked food and (then) ate it.’ (Stewart 1998)

b. Òzó giégié lé èvbàré ré.

Ozo quickly cook food eat

‘Ozo quickly cooked the food and (then) ate it.’ (Stewart 1998)

Nupe c. Musa du nakàn gí.

Musa cooked meat ate.

‘Musa cooked the meat and (then) ate it.’

10

d. Musa dàdà du nakàn gí.

Musa quickly cooked meat ate.

‘Musa quickly cooked the meat and (then) ate it.

Noting that in both languages I-types can occur to the left of VP1 in CSVCs, we might predict that leftward modification of I-type adverbs with respect to VP2 is permissible in both Edo and Nupe, given the independent status of VP2 as denoting a discrete event. This, however, is not the case. Whereas leftward modification of VP2 is allowed in Edo, it is prohibited in Nupe.

(17) Edo

a. Òzó giégié lé èvbàré ré.

Ozo quickly cook food eat

‘Ozo quickly cooked the food and (then) ate it.’ (Stewart 1998) b. Òzó lé èvbàré gié!gié ré.

Ozo cook food quickly eat

‘Ozo cooked the food and (then) quickly

c. Òzó giégié dùnmwún èmà khién.

ate it.’ (Stewart 1998)

Ozo quickly pound yam sell

‘Ozo quickly pounded the yam and (then) sold it.’ (Stewart 1998) d. Òzó dùnmwún èmà gié!gié khién.

Ozo pound yam quickly sell

‘Ozo pounded the yam and (then) quickly sold it.’ (Stewart 1998)

Nupe

e. Musa dàdà du cènkafa gí.

Musa quickly cooked rice ate

‘Musa quickly cooked the rice and (then) ate it.’ f. *Musa du cènkafa dàdà gí.

Musa cooked rice quickly ate

*‘Musa cooked the rice and (then) quickly ate it.’

g. Musa è tséká du cènkafa gí.

Musa PRS sometimes cook rice eat

‘Musa sometimes cooks rice and (then) eats it.’

11

h. * Musa è du cènkafa tséká gí.

Musa PRS cook rice sometimes eat

*‘Musa cooks rice and (then) sometimes eats it.’

Given these facts concerning the possible sites of I-type placement, I propose that there is a locality restriction governing the licensing of I-type adverbs in Nupe, namely, that there can be no intervening structure between INFL and the adverb. That no such restriction is imposed on I-type adverbs in Edo is evident in their ability to occur before VP2 in certain Serial Verb Constructions. Furthermore, this licensing condition explains 1) why I-types in

Nupe must be situated to the left of VP and 2) why they cannot be fronted. If they occurred in a post verb phrasal position, the locality restriction would be violated, as the modified VP projection would intervene between INFL and the adverb. Similarly, in the clefted position, the adverb would c-command INFL, contrary to the licensing constraint previously mentioned.

3.3.2 Phrasal Adverbs

Phrasal adverbs (P-type adverbs) in Nupe are the analogues of N-type adverbs in Edo. As such, they admit of the following distributional properties.

(18) A. P-type adverbs right adjoin to the VP.

B.

P-type adverbs can be fronted.

C.

There is no locality restriction governing the placement of P-type adverbs.

Consider the following examples which illustrate the uniformity of this distribution in both Edo and Nupe.

(19) Edo

a. Òzó (*ègiégié) rrí (*ègiégié) èvbàré ègiégié.

Ozo (*hurriedly) eat (*hurriedly) food hurriedly

‘Ozo (*hurriedly) ate the food hurriedly .’ (Stewart 1996)

b. Ègiégié Òzó rrí èvbàré.

hurriedly Ozo eat food

‘ Hurriedly

, Ozo ate the food.’ (Stewart 1996)

c. Òzó dùnmwún èmà khién ègiégié.

Ozo pound yam sell quickly

‘Ozo pounded the yams and (then) sold it quickly .’ (Stewart 1996)

12

Nupe

d. Musa (*karayín) tsò (*karayín) tangi karayín.

Musa (*carefully) set (*carefully) trap carefully

‘Musa set the trap carefully .’

e. Karayín Musa tsò tangi o.

carefully Musa set trap FOC

‘ Carefully , Musa set the trap.’

f. Musa du cènkafa gí karayín.

Musa cooked rice ate carefully

‘Musa cooked the rice and (then) ate it carefully .’

g. Musa (*bè kágbó yin) nì (*bè kágbó yin) eyàn bè kágbó yin.

Musa (*with strength) played (*with strength) drum with strength

‘Musa played the drum forcefully

.’

h. Bè kágbó yin Musa nì eyàn o.

with strength Musa played drum FOC

‘ Forcefully , Musa played the drum.’

i. Musa si eyàn nì bè kágbó yin.

Musa bought drum played with strength

‘Musa bought the drum and (then) played it forcefully .’

Note that in each batch of sentences for each language, the following are respectively illustrated: the right-edge of

VP as the site of P-type adverb adjunction; the ability of P-type adverbs to front; and the lack of a locality restriction governing P-type placement (as manifest in the modification of VP2 in CSVCs). As is evident from the data, both single word modifiers and multi-word modifiers (PPs) in Nupe share the distributional properties of Edo N-type adverbs. That is, in addition to one-word adverbials, phrases in Nupe can act as N-type modifiers. This is clearly opposed to Nupe I-type adverbs, which are exclusively syntactic heads and differ saliently from Edo I-types in the one crucial distributional property (locality to INFL). Thus, from the facts surrounding the syntactic environments and licensing conditions of I-type and P-type adverbs, we conclude that in Nupe both classes of adverbs are in complementary distribution.

3.4 The Reiterated Numeral Construction as a Phrasal Adverb

Although they behave as single morphological units, I claim that RNCs are instances of phrasal-type adverbs. To drive this point home, consider the behavior of RNCs in the following syntactic environments.

13

(20) a. Gana (*gútá-gútá) ba (*gútá-gútá) nakàn gútá-gútá.

Gana (*three-three) cut (*three-three) meat three-three

‘Gana cut the meat in threes .’

b. Gútá-gútá Gana ba nakàn o.

three-three Gana cut meat FOC

‘It was in threes that Gana cut the meat.’

c. Gana ba nakàn gí gútá-gútá.

Gana cut meat ate three-three.

‘Gana cut the meat (any which way) and (then) ate it in threes .’

Sentences (20 a. & b.) confirm that RNCs right adjoin to VP and can be preposed. Furthermore, the occurrence of the RNC in the Consequential Serial Verb Construction in (20)c affirms that its placement is not bound by any locality restriction. Clearly, the distributional properties of the RNC mirror those of P-type adverbs. We are justified in asserting the claim that RNCs are adverbial modifiers.

4.0 RNC Licensing

Thus far, our study has centered on the issues of RNC constituency, distribution, and modification. In the interest of providing a well-rounded syntactic analysis, we now turn our attention to the issue of uncovering the constraints licensing RNCs in Nupe. These constraints, although interesting in their own right, will provide further evidence that RNCs are indeed adverbial modifiers, in addition to rounding out our investigation into the syntax of the construction. Given the results of section three, which revealed that the domain of RNC modification is the verb phrase, we begin our search for licensing elements within the VP.

4.1 RNCs and Object-Taking Verbs

Consider the following sentences of Nupe with RNCs and transitive verbs.

(21) a. Musa pa ecizì gútá-gútá.

Musa pounded yams three-three

‘Musa pounded the yams in threes.’ b. *Musa pa eci ńdoci gútá-gútá.

Musa pounded yam certain three-three

*‘Musa pounded a yam in threes.’ c.

Musa ba nakàn gúbà (…) gútá-gútá.

Musa cut meat two (pause) three-three

‘Musa cut two (pieces of) meat in threes.’

14

d.

*Musa ba nakàn nini (…) gútá-gútá.

Musa cut meat one (pause) three-three

*‘Musa cut one (piece of) meat in threes.’

As the data in (21) suggest, the presence of an RNC seems to be correlated with the presence of a non-singular transitive verb object. Thus, we might propose that in sentences with transitive verbs, RNCs are licensed by some non-singular lexical item internal to the VP. As a test for this hypothesis, consider the following sentences with transitive verbs, singular objects, and plural subjects.

(22) a. *Egizì ya èzè gútá-gútá.

children gave response three-three

*‘The children answered the question in threes.’ b. *Yinzàgizì lo dzukó gútá-gútá.

women went market three-three

*‘The women went to the market in threes.’

Clearly then, our initial hypothesis is on the right track. For if we were to propose that RNCs are licensed by some non-singular antecedent without specifying which syntactic position was crucial, we would predict the sentences in

(22) to be grammatical. Thus, in sentences with transitive verbs, it appears that plural noun phrases condition the grammatical appearance of RNCs. These plural NPs, however, are themselves restricted to object position if they are to license RNCs within sentences with transitive verbs. At this point, we might inquire whether these same facts hold if RNCs were to occur in sentences with ditransitive verbs. Must both objects be plural if RNCs are to occur grammatically? Which object actually does the licensing? The answers to these questions will shed light on the nature of the RNC as well as the facts surrounding its licensing. Toward an answer to these questions, consider the following sentences.

(23) a. (V1 object = singular, V2 object = plural)

*Musa ya Gana ebìzì gútá-gútá.

Musa gave Gana kola nuts three-three b. (V1 object = plural, V2 object = singular)

*Musa ya egizì eze gútá-gútá.

Musa gave children response three-three c. (V1 object = singular, V2 object = singular)

*Musa ya Gana eci gútá-gútá.

Musa gave Gana yam three-three

15

d. (V1 object = plural, V2 object = plural)

*Musa ya egizì ebìzì gútá-gútá.

Musa gave children kola nuts three-three

Thus, despite the facts that obtain in simple sentences with transitive verbs, RNCs are not licensed in sentences with verbs taking more than one object. That is, the licensing requirement previously proposed is in need of strengthening. RNCs not only require a non-singular NP to occur in object position in sentences with transitive verbs, they require that this non-singular object position be unique and that it be filled by a non-singular noun phrase. That these facts are not the result of a particular quirk of Nupe involving the verb ‘give’ is given by the ungrammaticality of sentences similar to those in (23) involving other ditransitive verbs, such as ‘sold’ and ‘show’.

Armed with these results, we now consider the issue of RNC licensing in sentences with intransitive verbs.

4.2 RNCs and Non Object-Taking Verbs

Given the results of the preceding section, we might predict that RNCs are unable to occur in sentences with intransitive verbs. This, however, is not the case. Rather than discrediting the licensing hypothesis put forth so far, the ability of RNCs to occur in such sentences provides the basis for a principled theory of RNC licensing (to be formulated later). Consider the following.

(24) a. *Ekungùncì á tsu gútá-gútá.

warrior PRF die three-three

*‘The warrior has died in threes.’ b. Ekungùncìzì á tsu gútá-gútá.

warriors PRF die three-three

‘The warriors have died in threes.’ c. *Egi be gútá-gútá.

child came three-three

*‘The child arrived in threes.’ d. Egizì be gútá-gútá.

children came three-three

‘The children arrived in threes.’ e. *Egi nìkin gútá-gútá.

child fell three-three

*‘The child fell in threes.’

16

f. Egizì nìkin gútá-gútá.

children fell three-three

‘The children fell in threes.’

Given this data and the results of the preceding section, we might propose that the RNC licenser depends on the type of verb. That is, we might claim that when transitive, RNCs are licensed if and only if the verb’s object is unique and non-singular and that when intransitive, a non-singular NP in subject position does the licensing. In the interest of evaluating the plausibility of this claim, consider the following sentences.

(25) a. *Egizì lotùn gútá-gútá.

children work three-three

*‘The children worked in threes.’ b. *Egizì lele gútá-gútá.

children sleep three-three

*‘The children slept in threes.’ c. *Egizì yànyàn gútá-gútá.

children danced three-three

*‘The children danced in threes.’ d. *Egizì dazàn gútá-gútá.

children walked three-three

*‘The children walked in threes.’

Contrary to the claims inferred from the data in (24), the presence of a plural subject cannot save the grammaticality of such RNC-containing sentences with intransitive verbs. Thus, in the following manner we might argue that the hypothesis offered at the outset of this section cannot be entertained. Taking the data in (24) and (25) into account, along with the findings of section 4.1, we might formulate a unified theory of VP-internal RNC licensing that does not require the positing of a separate licensing constraint for each type of verb. Such a unified account, we might claim, becomes apparent once we consider the nature of the verbs in (24) and (25).

The sentences in (24) differ significantly from those in (25) despite the fact that both sets contain sentences with intransitive verbs. Whereas the verbs in (24) are unaccusative, those in (25) are unergative. The cash-value of this difference is that unaccusative/unergative verbs differ in the syntactic position in which their arguments are basegenerated (Burzio 1986). Thus, in (24b) for example, it is hypothesized that ekungùncìzì-‘warriors’ is underlyingly generated in object position and raises to subject position for case-related reasons. On the other hand, in sentences with unergative verbs, agents are underlyingly generated in subject position. Thus, in (25a) for example, egizì-

‘children’ occupies the subject position underlyingly, leaving the object position vacant. Taking into account the

17

facts governing RNC licensing in sentences with transitive verbs as well as the data in (24) and (25), a clear pattern regarding VP-internal licensing emerges. Namely, if RNCs are to be licensed in simple non-serialized sentences, the verb must have a unique object position that underlyingly gets filled by a plural noun phrase. This proposal nicely accounts for the data presented so far. As for sentences with transitive verbs, RNCs are licensed just in case the verb takes no more than one non-singular argument. This stems from the fact that transitive verb objects are basegenerated in object position. RNCs can also exist in sentences with unaccusative verbs, in so far as the verb’s argument is non-singular. This is a consequence of the position in which unaccusative verb arguments are basegenerated. Since RNC licensing is a function of a sentence’s underlying structure, as we have proposed, the underlying object-position of an unaccusative verb’s argument gives rise to acceptable RNC sentences just in case the verb’s argument is plural. Finally, given the licensing constraint proposed, we predict that RNCs cannot exist in sentences with unergative verbs in virtue of the fact that both underlyingly and at the surface, the object position of an unergative verb is vacant. Although this accounts for the data in (25), this prediction is not borne out in the language. Consider sentence (26) below.

(26) Elúgizì fù gútá-gútá.

birds flew three-three

‘The birds flew in threes.’

Clearly the analysis previously sketched is in trouble, for the non-singular object position constraint previously hypothesized is not across-the-board. Given the acceptability of (26) then, it seems that our initial proposal was on the right track. That is, certain constraints governing RNC licensing vary according to the type of verb. In order to drive this home, we must understand why the sentences in (25) are ungrammatical, while (26) is perfectly acceptable. Such an understanding, however, depends upon grasping the nature of the verbs themselves in the sentences in question.

Note that the verbs in (25) are the first polysyllabic verbs to be encountered in this study thus far. While this might seem trivial, it plays a significant role in the analysis to be sketched. Crucially, the verbs in (25) are not the proper sort to be called “transitive” or “intransitive”. Rather, they are all instances of “complex verbs”, verbs which together with a nominal element form a verbal complex that syntactically functions as if it were a single verb. Nupe admits of two types of verbal complexes; cognate object verbs (as in (25)) and split verbs , in which the verb’s object splits the components of the complex, as italicized in (27) below.

(27) a. Musa dà tákùn gwa .

Musa pushed rock hand

‘Musa pushed the rock.’ b. Musa le tákùnzì yé .

Musa saw rocks eye

‘Musa saw the rocks.’

18

As cognate object verbs , the predicates in (25) can be broken down into their verbal and nominal components. In all cases, cognate object verbs are formed by attaching the verb to its nominal form, after deleting the word-initial vowel of the nominal. This is illustrated in (28) below.

(28) a. lo + etùn → lotùn

‘worked’ ‘work’ → ‘worked’

b. le + ele → lele

‘slept’ ‘sleep’ → ‘slept’ c. yàn + eyàn → yànyàn

‘danced’ ‘dance’ → ‘danced’ d. da + ezàn → dazàn

‘walked’ ‘walk’ → ‘walked’

Given the nature of their formation, the verbs in (25), although not “transitive” verbs in the strict sense of the term, can be analyzed as object-taking verbs. The conclusions drawn in 4.1 and the data analyzed at the outset of this section can now be generalized to account for the unacceptability of the sentences in (25) and the acceptability of those in (24) and (26). If a verb takes an immediate object , that object must be unique and non-singular if RNCs are to be licensed; if a verb does not take an object, then its subject must be non-singular in order to license RNCs.

Because the nominal component ( immediate object ) of a cognate object verb can never be pluralized, we predict

RNCs in Nupe to be disallowed in sentences with complex verbs. This prediction is borne out in the language.

Furthermore, because the nominal component of a split verb is not an immediate object , RNCs are licensed in sentences with split verbs if the immediate object that splits the verbal from the nominal element is non-singular.

Thus, by restricting our attention to the nature of the immediate object of a verb, rather than the verb’s status as

“transitive” or “intransitive”, we can account for the possibility of RNCs in sentences with split verbs as well as the impossibility of RNCs in sentences with cognate object verbs. In this way, our theory of RNC licensing explains the grammaticality judgments of (21)-(27), while making correct predictions about significant classes of sentences in the language. Such findings accord nicely with the adverbial analysis of RNCs previously given. Since RNCs rightadjoin to VP as adverbial adjuncts, it is not surprising that the principal licensing constraint centers on aspects related to the nature of the verb.

4.3 Further Considerations in RNC licensing

Thus far, the examples of sentences with RNCs we’ve seen have been strikingly uniform in several respects. In all cases, the main verb of the sentence has been event denoting and the verb’s subject or object has always assumed the status of a count noun. Furthermore, all sentences considered up to this point have had animate subjects and

19

inanimate objects. The goal of this section is to determine whether or not the sentences we have seen so far have these properties necessarily. That is, do the properties previously mentioned in any way license RNCs? To answer these questions, let us consider the following sentences.

(29) a. *Musa è yebo egizì gútá-gútá.

Musa PRS likes children three-three b. *Musa du cènkafa gútá-gútá.

Musa cooked rice three-three c. Musa bà eyandzòzì gútá-gútá.

Musa counted toys three-three d. *Ebi ba nakàn gútá-gútá.

knife cut meat three-three e. Musa yán nangizì ló èmì gútágútá.

Musa chased sheep be-in house three-three

‘Musa chased the sheep into the house in threes.’

In the interest of being concise, further examples will not be provided. However, it will be mentioned that these samples are quite representative of similar types of sentences in the language. That is, Nupe does not allow RNCs to occur in sentences with stative verbs or inanimate subjects. Furthermore, it seems that in addition to being plural and unique, the verb’s object, which may or may not be animate (as illustrated by (29e) above), can not be a mass noun. These final remarks conclude our discussion of the secondary licensing constraints governing RNCs in Nupe.

(30) below summarizes the various licensing constraints argued for in section four.

(30) Constraints on RNC licensing (for simple non-serialized sentences) i.

If V takes an immediate object , that object must be unique and filled by a non-singular count noun. ii.

If V does not take an object, its subject must be a non-singular count noun. iii.

V must be eventive. iv.

V’s subject must be animate.

5.0 Conclusion

I have argued that adverbs in Nupe admit of a type distinction similar to the one found in Edo. Using this finding as grounds for a taxonomic division of adverbials in Nupe, the question regarding the modificational status of the

Reiterated Numeral Construction gives rise to an intriguing answer. Contrary to the standard role of the numeral as

20

adjectival modifier (semantically speaking), when duplicated and situated adjacent to the original, the resulting numerals together as a morphosyntactic unit take on the modificatory role of adverb as predicate of an event. Thus, in sentence (31):

(31) Musa dà tákùnzì gwa ya gúni-gúni.

Musa pushed rocks hand fell two-two

‘Musa pushed the rocks down in twos (pairs).’ the reiterated numeral complex modifies the event of rock pushing, not the substance rocks, following Parsons’ account of adverbs as predicates of events denoted by verbs (Parsons 1990). I take this finding to be of considerable significance for two reasons: 1) it cuts across the generalization that numerals are strictly nominal/adjectival in nature, and 2) it suggests that the fundamental means by which adverbs are classified is not based solely on the adverb’s semantic label (e.g. manner, benefactive, comitative), but rather the conjunction of the adverb’s semantic attributes with its distributional properties.

Notes

*

This work represents a section of my ongoing B.A. honors thesis devoted to the syntax and semantics of numeral constructions in Nupe. I would like to extend my utmost appreciation and gratitude to my warm and patient fieldwork consultant Ahmadu Ndanusa Kawu. While the judgments expressed in this work are his, I do not hold him responsible for any errors or inconsistencies that may appear in this paper. For opening my eyes to relevant data in Malagasy and introducing me to the work of David Gill, I gratefully thank Ed Keenan. Special thanks to D.

Lightswitche for valuable comments and insightful remarks offered along the way. I reserve my deepest thanks for my thesis committee, Maria Bittner and Mark Baker. I am honored to have worked under their watchful scrutinizing eyes and feel that I am a better theoretician for it. The fieldwork for this project was supported by a grant from the

National Science Foundation (NSF grant BCS-9905600) for which I gratefully acknowledge.

1 Nupe is a Nupoid language of the Benue-Congo sub-family of the Niger-Congo language family of Africa. It is spoken by approximately one million people in the Middle Belt region of Nigeria on both sides of the River Niger.

Other Nupoid languages include Ebira, Gade, and Gwari.

2 Numeral constructions such as those in (1c) seem to be common in languages which employ reduplication rather productively. To my knowledge, in addition to Nupe such numeral complexes are attested in Hebrew, Yoruba, Edo,

Kikongo, Lingala, Tagalog, Georgian, Turkish, and Malagasy. See Gill (1982) for others. It is my belief that a significant percentage of the world’s languages employ numerals in this way and that the phenomenon is somehow rooted in facts surrounding reduplication, as opposed to being unique to African and Austronesian languages. In these respects, such numeral constructions are deserving of our attention.

3 Numeral stems in Nupe (e.g. gútá) are composed of a numeral root and an affix. Cardinal numerals are comprised of a prefix plus the root; ordinal numerals consist of the root plus a suffix. Thus, it is argued that in a numeral like gútá, gú is a prefix and tá is the numeral root. Evidence for this proposal comes from the fact that tá occurs in both cardinal and ordinal numerals, while gú does not (Kawu 1997); e.g. gú tá (‘three’) vs. tá tsi (‘third’). Thus, RNCs can be seen as types of stem-reduplicated numerals. In this way, they contrast with another intriguing numeral construction present in the language, in which the root is the target of reduplication. In such constructions, numerals such as gútá-tá act both like universal quantifiers and cardinality predicates. See Kandybowicz (2000) for a full discussion of the syntax and semantics of both stem and root reduplicated numerals in Nupe.

21

4 See Baker, Stewart, et al. (2000) for arguments that adverbs adjoin higher in Nupe than in Edo. Such a claim, if true, does not refute the proposals put forth in section 3, but rather provides a more precise explanation for the phemenon detailed there.

5 Serial Verb Constructions are not limited to chains of exactly two verbs occurring in series. For the sake of clarity and brevity, however, I discuss the properties of SVCs limited to just those complexes housing two verbs. These properties can be generalized to instances when more than two verbs occur serialized.

References

Baker, M.C. (1989) “Object Sharing and Projection in Serial Verb Constructions. Linguistic Inquiry 20, #4: 513-

553.

Burzio, L. (1986) Italian Syntax: A Government-Binding Approach , Dordrecht: Reidel

George, I. (1975) A Grammar of Kwa-Type Verb Serialization: Its Nature and Significance in Current Generative

Theory , Ph.D dissertation, UCLA.

Gill, D. (1982) Distributive Numerals , Ph.D dissertation, UCLA.

Kandybowicz, J. (2000) “Deviant Numeral Constructions in Nupe”, B.A. honors thesis, Rutgers University.

Kawu, A. (1997) “Nupe Reduplication”, ms. Rutgers University.

Parsons, T. (1990) Events in the Semantics of English: A Study in Subatomic Semantics , Cambridge, Mass.: MIT

Press.

Ross, J. (1967) Constraints on Variables in Syntax , Ph.D dissertation, MIT. Cambridge, Mass. Published in 1986 as Infinite Syntax . Norwood, N.J: Ablex.

Stewart, O.T. (1996) “Adverb Placement and the Structure of the Serial Verb Construction,” Proceedings of NELS

26 , 409-423, MIT/Harvard GLSA.

Stewart, O.T. (1998) The Serial Verb Construction Parameter , Ph.D dissertation, McGill University.

Stewart, O.T., Baker, M.C., Kawu, A. “A Comparative Analysis of Serial Verb Constructions in Kwa Languages.”

Talk presented at the 31 st Annual Conference on African Linguistics, March 3 2000, Boston University.

22

23

Download