You are on page 1of 8

Finiteness and Reduplication in Dravidian

Rahul Balusu EFLU Hyderabad The problem. This paper analyzes verb reduplication in Telugu. In Telugu, as in the other Dravidian languages like Kannada, Tamil and Malayalam, reduplication of the verb indicates repetition or continuity of action as shown in (1). (1) kootulu egiri-egiri velli.ni.yyi monkeys jump-jump went.PST.3PL the monkeys went jumping repeatedly

The reduplicated verb is non-finite. The finite verb cannot be reduplicated to indicate repetition as shown in (2). This paper attempts to solve the finiteness restriction on verbal reduplication. (2) *kootulu egiri-egiri.ni.yyi monkeys jump-jump.PST.3PL the monkeys jumped repeatedly

The analysis. The reduplication in the verbal domain is parallel to the reduplication in the nominal domain where reduplication of numerals gives rise to distributive readings. In earlier work I have shown that the reduplicated numeral is associated with a distributive operator whose sorting key (Choe 1987) is always an event or an event aspect and that it has a plurality requirement on the distributive share (Choe 1987) such that the distributive readings of a predicate can be viewed as the pluralization of the predicate. This is shown in (3) with the interpretation of the sentence at logical form as given in (4). (3) renDu renDu kootu-lu egir-i-niyyi 2 2 monkey-Pl jump-Past-3PPl lit. 2 2 monkeys jumped 2 monkeys jumped each time monkeys jumped in groups of 2 in each location

Temporal key reading Spatial key reading

(4)

a. e (e) [ e(e) x[two monkeys(x) jumped(x, e)]] b. |x: two monkeys(x) jumped(x, e)|>1

There is a natural connection here between verbal and nominal reduplication semantics. Champollion (2010) finds a large class of constructions in the nominal and verbal domain that impose a parametrized but otherwise identical distributivity presupposition on one of their arguments. This distribution is parametrized on the dimension scale it happens in time, in space, or for a thematic role. The verbal reduplication construction in Telugu tests for the dimension of time. I propose that the reduplicated non-finite verb has a D(istributivity)-operator associated with it which takes the temporal aspect of the finite verb as its sorting key and the non-finite verb as its distributive share. So in (5) the temporal aspect of the going event is the Key over which the eating Share is distributed. (5) kootulu tinTuu-tinTuu velli.ni.yyi monkeys eating-eating went.PST.3PL the monkeys went eating continuously

The Share of the distributive verbal reduplication must be an atelic predicate. A telic predicate cannot be the Share. We can use for-adverbials, which are the most reliable indicators of atelicity (Verkuyl 1989) as a diagnostic. This is shown in (6) (8).

(6)

kootulu aidu nimashaalu egiriniyyi monkeys 5 minutes jumped the monkeys jumped for 5 minutes *kootulu chettu.miidaki aidu nimashaalu egiriniyyi monkeys tree.onto 5 minutes jumped the monkeys jumped onto the tree for 5 minutes *kootulu chettumiidaki egiri chettumiidaki egiri monkeys tree.onto jump tree.onto jump the monkeys went jumping repeatedly on to the tree velliniyyi went

(7)

(8)

Atelic predicates have the property that they hold true of any subinterval of the process (Bennett and Partee, 1972): Whenever they hold at some interval i, they also hold at every subinterval of i. This is not true of telic predicates. This is shown in (9) (adapted from Champollion, 2010). (9) DIVduration([[jump]]) jump has the subinterval property e [[jump]] i < duration(e) e [[jump]].[e < e duration(e) = i]

So every proper part of the temporal trace of a jumping event e is the temporal trace of another jumping event which is a proper part of e. The restriction on an acceptable distributive share in the distributive relation is that the property that holds true of the event also holds true of i-parts of that event, as shown in (10). (10) Distributivity Constraint: Champollion (2010) Every distributive construction presupposes that DIV[Map]([[Share]]). For every part of any Key there is a Share which the Map maps to it. This explains the non-finite atelic restriction on the predicate that is reduplicated. The Share must be compatible with a distributive relation. A distributive construction presupposes that for every part of any Key there is a Share which maps to it. A telic predicate is not acceptable as Share because it applies only to the whole event and fails to apply to parts of the event in its denotation i.e. not every part of a jumping onto the tree is itself a jumping onto the tree, DIVduration([[jump onto the tree]]) fails. An atelic and non-finite predicate is compatible because it applies to parts of the whole event. In a distributive relation the members of the distributive share need not be exhaustively used up when being distributed, whereas the members of the sorting key need to be exhaustively used up in being distributed over. This is a key difference between the distributive share and the sorting key which explains the finiteness restriction on verbal reduplication. So in distributive verbal reduplication the Key has to be finite if it has to be exhaustively used up. This accounts for the finiteness restriction on verbal reduplication seen in Telugu. In sum, the selectional restrictions on the two overt constituents of the distributivity relation established by reduplication are responsible for the finiteness restriction in Telugu verbal reduplication. The restriction on what is an acceptable distributive share requires that the property which holds true of an event should also hold true of parts of that event. This explains the atelic nonfinite constraint on the reduplicated verb. The restriction on what is an acceptable sorting key requires that the members of the Key be exhaustively used up. This explains the finite temporal restriction on the non-reduplicated verb in the verbal reduplication construction. Selected References: Choe, Jae-Woong (1987) Anti-Quantifiers and a Theory of Distributivity, Ph.D. Dissertation, UMass., Amherst. Champollion, Lucas (2010) Parts of a whole Distributivity as a bridge between aspect and measurement, Ph.D. Dissertation, UPenn., Philadelphia.

In sum, the selectional restrictions on the two overt constituents of the distributivity relation established by reduplication are responsible for the finiteness restriction in Telugu verbal reduplication. The restriction on what is an acceptable distributive share requires that the property which holds true of an event should also hold true of parts of that event. This explains the atelic non-finite constraint on the reduplicated verb. The restriction on what is an acceptable sorting key requires that the members of the key be exhaustively used up. This explains the finite temporal restriction on the nonreduplicated verb in the verbal reduplication construction. As Dowty (1979) shows, the entailment properties of atelic predicates with respect to a large range of other phenomena, including tense, the progressive, and

aspectualizers, can be
which can have any of the interpretations in (2). The NP that contains the RedNum is the distributive share (Choe 1987). In (3) the RedNum appears with the subject NP and the interpretations change accordingly. That over which the NP with the RedNum is distributed is the sorting key (Choe 1987). The various readings given in (2) and (4) have different sorting keys as shown in (5). Intuitively, the working of RedNum is similar to that of English binominal each, but there are two important differences. In the RedNum construction, (A) The sorting key can be covert as in (2b, c) and (4b, c), and (B) The overt sorting key can apparently be contributed by a distributive universal. Thus (7) contrasts with English (8). At first sight, (B) rules out an analysis where the RedNum contains a silent counterpart of binominal each. But this is precisely what I will propose, tying (A) and (B) together. The analysis. I propose that the RedNum is/has a D(istributivity)-operator which always takes one of the subevents (sub-event-aspects) of the event as its sorting key, and is associated with a plurality requirement. This accounts for the temporal key readings (2b, 4b) and spatial key readings (2c, 4c). Readings (2b, 2c) are analyzed as in (6); the plurality requirement is that there must be more than one pair of monkeys seen in the whole event e: (6) e[ee [M[two_monkeys(M) the_kids(K) see(M)(K)(e) |{X: two_monkeys(X) Z[see(X)(Z)(e)]}|>1] I argue that (2a), the participant-key reading, is in fact also an event-key reading. It arises when (i)+(ii) obtain. (i) The predicate `saw two-two monkeys is distributive with respect to the subject `the kids. That is, in (2a) the sentence has two D-operators: one coming from the predicate and one coming from the RedNum. (ii) There is just one sub-event per kid: one that comprises all his/her two-monkey sightings in e. This is a coarse grained view of the event structure, as opposed to the fine grained view in (6), where more than one sub-event is associated with each of the kids. The analysis now extends to (7) with `every kid. Its participant key reading, (9a), is obtained as in (2a), the only difference being that the extra D-operator does not come from the predicate but from `every kid. This is again a coarse grained view of the event structure. The fact that the temporal and spatial key readings (9b, c) are also available is evidence supporting this analysis with two distributions and D-operators. (9b, c) are analyzed in (10): (10) E[y[kid(y) eE[ee [M[two_monkeys(M) see(M)(y)(e) |{X: two_monkeys(X) z[see(X)(z)(E)]}|>1]]]]] In (11) with a singular, the temporal and spatial key readings are possible (12b, c) but not the participant key reading (12a). This is because the D-operator requires that the cardinality of the set of monkey pairs seen in E be greater than one, and the participant key reading cannot satisfy this. Constructions involving more than one reduplicated numeral are also possible as shown in (13). The sentence in (13) can have many interpretations depending on what goes into the argument position of each of the D-operators associated with each of the RedNums, the spatial or the temporal aspects of the event. Also depending on whether there is one or more than one of each aspect (i.e. a fine grained or coarse grained view of the event structure) associated with each member of the subject argument, the possibilities multiply. Some of the possible interpretations for (13) are given in (14). This analysis has advantages over earlier proposals on nominal distributivity markers like Oh (2001)

and Gil (1990) in that it can properly explain and predict the possible distribution and interpretations of constructions involving RedNum. It accounts for the novel data involving interaction between universals and the RedNum. It can address the cross-linguistic variation in nominal distributivity markers based on the coarse grained or fine grained way of viewing events that a language adopts. Finally, an interesting parallelism can be drawn between nominal distributivity markers and pluractional markers (Lasersohn 1995) that are seen in the verbal domain. (11) aa (12) aaa Finite verb reduplication The finite verb can be reduplicated. It has an emphatic meaning: (13) vastunnaa, vastunnaa I am coming, I am coming! (14) parigettu, parigettu Run, run

tinTuu tinTuu vacchaanu - I came eating. cadavagaa cadavagaa vacchinidi - reading and reading I got the lesson tini tini laavu ekkeenu - eating a lot became fat neenu TV chuustuu chuustuu podukunnaanu I

The reduplicated verb can only be non-finite kootulu egurutu egurutu velliniyyi. But you cant say kootulu egiri egiri niyyi. why? Why does finiteness block pluralization of the event argument? Reduplication of Imperative verb possible: emphatic meaning parigettu parigettu - run run! parigettutunnaanu parigettutunnaanu - I am running, I am running! Reduplication in nominal domain - Distributive Reduplication in verbal domain - Atelic neenu 5min parigetteenu neenu 5min parigetti parigetti alisipooyeenu Link between non-finite and atelic?

The reduplicated verb has a distributive share and key. The share is the atelic verb. The key is the finite interval.

Notes: Steever(1988:11114)definesfinitenessasasyntacticproperty. four types of non-finite verb forms i) infinitive paaDagaa paaDagaa vastundi itll come by singing and singing tinagaa tinagaa arugutundi cadavagaa cadavagaa toolagaa toolagaa teelika avutundi naDavagaa naDavagaa deggiriki vastaamu *aDiginaa aDiginaa ivvaTaleedu enta aDiginaa ivvataleedu *parigettinaa parigettina paaripoolevu enta parigettinaa paripooleevu *ceppinaa ceppinaa vinaleedu ii) verbal participle jaagarattagaa cuusi cuusi ceeseenu vaagi vaagi gontu pooyindi eeDisi eeDisi kallu erragaa ayyiniyyi ongi ongi seeva ceeseeDu ceppi ceppi alavaaTayyindi vaana aagi aagi kurisindi tirigi tirigi vaccaamu iii) conditional (non-past tense only) vast-ee vastaanu *vastee vacceenu cadivit-ee caduvutaanu *cadivitee cadiveenu paaDat-ee paaDataanu *paaDatee paaDeenu

iv)

adjective participle.

ceppi ceppi alisipooyaanu ceptuu ceptuu paniceeseenu vinaka vinaka visugoccindi vini vini visugoccindi pani ceesi ceesi alisipooyaanu an infinitive clause with adverbial function can have various interpretations such as i) purpose ii) cause iii) time and iv) result. Of the four types of infinitive adverbial clauses except purpose infinitive clauses can be reduplicated to express intensity, repetition or continuity of an action. In the case of causal and temporal infinitive clauses there is often ambiguity between a causal and temporal interpretation

neenu vastee vastaanu leekapootee leedu vaaDu ceppinaa cepputaaDu manani ikkaDa vadilinaa vadulutaaDu ceppii ceppaka maTlaadii maTlaadaka

The verbal reduplication construction tests for the dimension of time! impose a parametrized distributivity presuppo- sition on one of their constituents. kootulu monkeys egiri egiri jump-jump share Dimension parameter: Time velliniyyi went key

Hence the key always has to be finite. The share has to be non-finite and atelic. So here - until the going (key), is the duration of the jumping (share) The share cannot be finite. Then every subpart of the share wont be the same as the share.

Telic shares are rejected: *kootulu chettumiidaki egurutuu egurutuu velliniyyi.

inner aspect involves the telic-atelic difference, and outer aspect involves the perfective-imperfective distinction.

Explanation: Temporally telic predicates are ruled out because they lack this property. For example, DIVduration([run to the store]) fails, because not every part of the runtime of a running-to-the-store event is the runtime of another running-to-the-store event.

The Share follows the Plus Principle


kootulu aratikaayalu tinTuu araTikaayalu tinTuu velliniyyi *kootulu 3 araTikaayalu tinTuu 3 araTikaayalu tinTuu velliniyyi kootulu 3 araTikaayalu cuustuu 3araTikaayalu cuustuu velliniyyi kootulu araTikaayalu cuustuu araTikaayalu cuustuu velliniyyi

The Plus Principle: A predicate is unbounded (= compatible with for- adverbials) if it has either an unbounded head, or an unbounded argument. (Verkuyl (1972), reformulated by Kiparsky (1998)) (29) a. b. c. d. *Judith[+B] ate[+B] three sandwiches[+B] for an hour. Judith[+B] wanted[-B] sandwiches[-B] for an hour. Judith[+B] wanted[-B] three sandwiches[+B] for an hour. Judith[+B] ate[+B] sandwiches[-B] for an hour. (30) Snow[-B] fell[+B] throughout the area[+B] for two straight days.3 (31) a. *The carafe of wine[+B] flowed[+B] from the jar to the floor[+B] for five minutes. b. Wine[-B] flowed[+B] from the jar to the floor[+B] for five min- utes. (Beavers, 2008) Question: Why is one [-B] enough to make the whole predicate atelic? Answer: Temporalfor-adverbialspresupposeDIV(Share),notDIV(Share). That is, they presuppose the subinterval property, not divisive reference. In many cases, one unbounded predicate is already enough to guarantee the subinterval property.

You might also like