R. Amritavalli
This paper examines the intricacies of disjunction in Kannada, a Dravidian language spoken in southern India.* I argue that disjunctive morphemes like Kannada -oo and English or are connectives that are semantically “sensitive” or “deficient” in the sense of Giannakidou (1997); that is, they “cannot be properly interpreted” except in certain environments. The interpretive environment for the disjunctive connective is provided by a Polarity Phrase, which hosts the licensors for the disjunctive morpheme. These licensors are principally Neg and the Question operator; Neg being an averidicality operator, and Q possibly subsumable under the class of nonveridicality operators that includes modality and genericity. In Kannada, the head -oo itself must be licensed; in English, the element that is licensed is either, which has been variously argued to be the head or the specifier of the Disjunctive Phrase.1
Disjunction can be phrasal or clausal; clausal disjunction in natural languages falls into two major subtypes—the disjunction of questions (whether-disjunction in English) and the disjunction of declaratives (either-disjunction in English). I show that the citation Kannada disjunctive morpheme -oo, although paralleling English either in connecting phrases, has (somewhat surprisingly) the distribution and properties of English whether when it connects clauses. Clausal -oo has the default interpretation of whether, and therefore it cannot connect declarative sentences. The disjunction of declarative clauses is accomplished by a Neg morpheme illa.
I argue that the illa of Kannada disjunction is the overt counterpart of a covert Neg element in English declarative disjunction, which licenses either. Both elements—illa and the covert Neg—affect the interpretive possibilities of disjunction in a particular way: they prohibit the conjunctive reading of disjunction. This reading arises from the interpretation of either as a universal quantifier (Higginbotham 1991) and depends on the overt occurrence of modality or genericity. I analyze these elements as licensors for either and argue that, in their absence, there must nevertheless be present a covert licensor for either. This licensor is a Neg that forces an existential interpretation for either and prohibits the conjunctive interpretation. The corresponding fact is that Kannada illa disjunction lacks the conjunctive reading.
Kannada -oo is licensed by nonveridicality operators such as modality and genericity in phrasal disjunction with a conjunctive interpretation. In clausal disjunction, it is licensed by the Q operator that licenses question words. Further, -oo in embedded questions is shown to precisely demarcate their scope, which is an argument that it must adjoin to the Q operator.
Disjunction in Kannada poses some interesting problems. The Kannada disjunctive particle is standardly cited as -oo. It is illustrated in (1).
As (2) shows, -oo can connect Case-marked DPs. It can also connect putative PPs (postpositional phrases), as (3) illustrates.2
What is intriguing is that the particle -oo cannot connect two declarative sentences.
But (4) is not uninterpretable. It has an interpretation as a disjunction of questions: more precisely, as an alternative question. Example (5a) illustrates a disjunction of matrix questions; (5b) illustrates the same disjunction as an indirect-question complement to the matrix predicate gottilla ‘do not know’.3
What about the disjunction of declarative clauses? I illustrate this in (6). Notice that what expresses the disjunction of declarative clauses is a negative element, illa.
But how does a negative element function as a disjunctive particle in declarative clauses in Kannada? And why does the citation disjunction -oo inevitably yield an interrogative interpretation in clausal disjunction, though not in phrasal disjunction? These are the questions of concern in this paper.
Note that illa in (6) is not the illa of sentential negation. The latter occurs clause-finally and is in complementary distribution with the tense/aspect and agreement morphemes (Amritavalli 2000). The illa in (6) appears in a clause-initial adverbial position. The examples in (7), the negative counterparts of (6), illustrate the illa of sentential negation along with the illa of disjunction.
The illa of disjunction indicates the scope of the disjunction (see (8a,b)), and it must occur in parallel positions in the disjuncts (see (8c)).
In its scope-marking function, this illa of disjunction is parallel to English either (Larson 1985). This suggests that illa may be the left element of a larger Disjunctive Phrase illa … oo, much like either … or, with illa as the specifier and -oo the head. The question then arises whether the full phrase is ever attested; that is, if illa and -oo co-occur.
We know that illa and -oo do not co-occur in clausal disjunction, where they signal respectively the disjunction of statements and the disjunction of questions. In phrasal disjunction, illa does occur, at the left edge (cf. -oo at the right edge, in (1)–(3)).
But here again, there is a preference for only one of the elements, illa or -oo, to occur.4
Summing up, the Kannada disjunctive particle -oo has the interpretation of whether in clausal disjunction, but it occurs with a noninterrogative interpretation in phrasal disjunction. Illa, like English either, occurs in declarative-clause disjunction and in phrasal disjunction; and like either, illa indicates the scope of disjunction.
How and why does -oo have different interpretations at the phrasal and clausal levels? Given that -oo and illa have different interpretations in the disjunction of clauses, do they similarly differ in interpretation in phrasal disjunction? And why (to repeat an earlier question) is a Neg element a disjunctive particle in Kannada? To answer these questions, I turn to an analysis of disjunction by Higginbotham (1991).
Higginbotham (1991) gives an account of the conjunctive interpretation of disjunction. This interpretation is illustrated in (10), where the conjunctive reading is parenthetically indicated. The example is from Higginbotham (his (1)).
As Higginbotham points out, (10) can be paraphrased as ‘John plays both chess and checkers,’ or ‘John plays either of chess or checkers,’ or ‘John plays any one of chess or checkers.’5
Higginbotham (1991:143) argues that in English, “every or is an either/or; i.e., part of a larger constituent headed by either or its interrogative counterpart whether.” Thus in (10), there is a covert either, made overt in (11).
Higginbotham now attributes the conjunctive/disjunctive ambiguity of (10)–(11) to the ambiguity of either when it occurs in isolation. He shows that “either without or has the distribution of any.” (Examples (12)–(14) correspond to Higginbotham’s examples (10), (12), and (14).)
Either in (12)–(14) allows two meanings. In (14), either is the existential quantifier; on this interpretation, either is a negative polarity item. A second interpretation of either, in (12)–(13), is as a universal quantifier. In its universal interpretation, either (like free-choice any) needs a modal or generic environment to license it.
In other words, either is what Giannakidou (1997:1) calls a “sensitive” expression. Giannakidou characterizes sensitivity as a kind of semantic dependency. A sensitive expression is an item that is semantically “deficient,” “which cannot be properly interpreted except in the environments which fulfil (its) interpretational demands.” Either as a quantifier is semantically dependent on one of the elements—Neg, modality, or genericity—which functions as its licensor and determines its interpretation as an existential or as a universal quantifier. The ambiguity of either arises out of the availability of two sets of licensors for it.
Returning to disjunctions, Higginbotham argues that the conjunctive interpretation of disjunction arises from the interpretation of either as a universal quantifier. That is, the ambiguity of (10)–(11) between conjunctive and disjunctive readings is attributable to the interpretations of either as a universal or as an existential quantifier, whether either is covert as in (10) or overt as in (11). Thus Higginbotham generalizes the account of either in isolation to either in disjunctions.
But there is an interesting anomaly in this account of English disjunctions with regard to the licensors for either in its two interpretations. The licensors for universal either must always overtly occur in the disjunction, even when either itself is covert; thus Higginbotham tells us (1991:144) that either “remains restricted in its universal interpretation, needing … a modal or generic environment.” The Neg licensor of existential either, however, does not ever surface in the disjunction. Higginbotham therefore concludes (loc. cit.) that either in the disjunction “is substantially free in its existential interpretation.”
But notice that if the either in disjunction is the same “sensitive” expression as the either that occurs in isolation (as Higginbotham argues), it must be uninterpretable in the absence of a licensor. Because existential either in isolation is a negative polarity item, it follows that existential either in the English disjunction must similarly be licensed by a Neg. I thus conclude that there must be a covert Neg in English disjunction, in alternation with the modal and generic licensors for either, and it is this alternation of licensors (respectively) that gives either its existential and universal readings and results in the conjunctive/disjunctive ambiguity of disjunction.
A more general result is that (some element in) the Disjunctive Phrase is always polarity sensitive, occurring under the scope of a licensor. With this, we have the beginnings of an explanation for the role of a Neg in disjunction in Kannada. Conjecturing that the illa in Kannada disjunction is the overt counterpart of the Neg operator that induces an existential reading for either, I draw the following prediction for Kannada: phrasal illa disjunction must exclude the conjunctive reading. This prediction is tested in the next section.
Consider now the Kannada -oo and illa phrasal disjunctions in (15).
In my judgment, (15a) is an offer to play whichever (or any) game you choose. But (15b) does not commit the speaker to playing any particular one of the games, say checkers; the offer is only to play one of the two games mentioned.6 Thus to the lay ear, (15b) sounds more authoritative than cooperative, and this is a consistent and perceptible difference between the two types of disjunction. The examples (16a,b) and (17a,b) make the same point. Note that in the (a) sentences, but not in the (b) sentences, it is possible to parenthetically add indefinite wh-expressions such as yellig-aadaru (lit. ‘where-to-happen-COND’) ‘to anywhere, any such place’ and yeen-aadaru (lit. ‘what-happen-COND’) ‘anything, any such thing.’
Examples (17a, b) are particularly clear: (17a) can be taken to assert that you can get meals in this hotel; but (17b) does not assert this. Thus on hearing (17a), the interlocutor can confidently say, “Let’s then go and have a meal there,” but not after hearing (17b).
How does illa block the conjunctive reading? I claim that this illa of disjunction in Kannada is the overt counterpart of the covert Neg in the semantic structure of English disjunction, which licenses existential either. The semantic structure of disjunction is, of course, expected to be independent of particular languages. But there are lexical differences between languages—and in this case, Kannada has no lexical counterpart to the quantifier either. What surfaces from the semantic structure is, I suggest, the Neg licensor of an existential quantifier.7
Anticipating this analysis of -oo as a head licensed by the Q and nonveridicality operators, I speculate that illa licenses an empty head for the Kannada DisjP. This null head is interpreted as an existential quantifier—a plausible conjecture, given the parallel fact noted below that the -oo head of disjunction regularly combines with wh-words to form existential quantifiers in Kannada.
The data is reinforced with the following examples. In (18) we have a disjunction in a generic statement pertaining to the past; (18a), without illa, is fine, but (18b) with illa, insofar as it is interpretable, must refer to periods of past time when the speaker played one game or the other.
In (19a), the disjunction is contained under a universal quantifier ella ‘all’. This results in a conjunctive reading; see Higginbotham (1991:146): “when a disjunction is contained within a universal quantifier we will have a conjunctive equivalent.” Predictably, illa cannot occur (19b).
When illa occurs, as in (19c), the set of people in each disjunct is understood as relativized separately, and a single quantifier ella cannot occur that refers to both sets.8
The conjunctive interpretation of -oo in phrasal disjunction shows that it is here licensed by nonveridical operators, giving a universally quantified reading. This is consistent with the occurrence of -oo in alternative questions such as (5a), where, following Higginbotham (1991:152), I take it to represent universal quantification. Its licensor in alternative questions is Q; I shall show that -oo must in fact adjoin to Q. This parallels the adjunction of English either to a wh in Comp (which is taken to be the Q operator, or the reflex of the Q operator) to yield the whether of interrogative disjunction. However, because Q in Kannada is never overt, the counterpart of whether in Kannada clausal disjunction surfaces as a bare -oo.
Although the parallel between -oo and English whether is suggestive, it is incomplete. Kannada -oo is licensed in the Q-contexts of English whether, but also (as I have shown) in phrasal disjunctions where English attests either. Indeed, we find -oo licensed by Q in clauses that are not disjunctions at all, such as exclamations (20) and dubitatives (21). In (20), -oo co-occurs with a question word; in (21), the interpretation suggests a Q operator again.
English whether is considered “the interrogative counterpart” or the “wh-counterpart” of either (Higginbotham 1991, Larson 1985:225); it originally had the meaning ‘which of either’. Whether, however, is not solely an interrogative complementizer but more generally a marker of syntactic dependency, as seen in (22), where its wh-element is comparable to the wh-element of free relatives. Assuming this wh-element to be its licensing operator, we see that whether in (22), unlike whether in interrogative complements, is not required to be overt. As seen in (23), Kannada -oo similarly occurs in sentence adjuncts.9
Kannada (and Dravidian more generally) is usually considered not to have an interrogative complementizer. This is because the complementizer anta ‘that’ introduces interrogative as well as declarative complements:
But there are occurrences of -oo where it seems to function as an interrogative complementizer. Compare the sentences in (25). Example (25a) has an interrogative complement (the wh-word yaaru ‘who’), and -oo occurs instead of the regular complementizer. The sentence in (25b), on the other hand, has a declarative complement (with avaru ‘they’); -oo cannot occur (and anta must occur).
Note that unlike English whether, which introduces only yes/no-question complements, Kannada -oo introduces a wh-interrogative complement in (25a). Yes/no-question complements are treated as disjunctions with or not (Larson 1985). Wh-question complements with -oo are, I suggest, disjunctions with a wh-word yeenu ‘what’. Even yes/no questions in Kannada exhibit -oo as a disjunct with yeenu ‘what’ instead of illa ‘not’, as the examples in (26) show.10
Notice that in (26b), the yeenu-disjunction is itself introduced by the regular complementizer anta ‘that.’ This suggests that -oo is not an interrogative complementizer, after all.
I propose that the disjunctive head -oo occurs in interrogatives by virtue of being licensed by the Q operator of interrogatives. I shall now present evidence that this head must adjoin in the overt syntax to a Q operator in its local domain. The evidence is that the -oo-disjunction precisely delimits the scope of questions. A wh-phrase in an embedded clause with -oo cannot be interpreted outside the domain of this clause; indeed, it must be interpreted in precisely the -oo-clause. If -oo is required to adjoin to a local Q operator, we explain why the clause in which the -oo appears is also the clause in which the question word must find its interpretation.
Consider the contrasts in (27a,b) and (28a,b). In the (a) examples, with the regular complementizer anta, the wh-word in the embedded complement can take matrix scope. In the (b) examples, where -oo occurs, the wh-word cannot scope out of the -oo clause:
This point is clear when we consider more deeply embedded question complements. In the (a) examples of (29)–(31), with anta as the complementizer, a wh-element that is embedded three clauses down can take matrix scope. In the (b) examples, the clause intermediate between the matrix clause and that containing the question word has -oo. The embedded wh-word can (and must) scope out of its own clause, but it cannot scope out of the -oo-disjunction in the intermediate clause:
Note also that -oo does not simply indicate maximal scope for the question word, allowing it the option of being interpreted in the innermost clause in the above examples. Thus (29b), (30b), and (31b) do not have the respective readings “You alleged that the doctor wondered which medicines to give,” “You suspect that the police were uncertain who killed Rajiv,” or “You thought that the BBC wondered how many people died.”
A second interesting property of -oo-disjunctions is that the Q that licenses the disjunctive head -oo may also license a dummy wh-word in the clause -oo introduces. Thus consider the sentences in (32). These can be interpreted as yes/no (whether-) complements; but we find in them (optionally, but preferably) a wh-word yelli ‘where’ that receives no interpretation. (These examples can, of course, also receive the corresponding literal wh interpretation, with yelli as a place adjunct.)
The disjunctive particle -oo overtly adjoins to question words in Kannada to form existential quantifiers or “wh-indefinites”: yaar-oo ‘someone’ is yaaru ‘who’ + -oo; yell-oo ‘somewhere’ is yelli ‘where’ + -oo, and so on.11 Li (1992:130) shows wh-words in Mandarin Chinese to have an indefinite interpretation “in contexts where the truth of a proposition is not asserted/ implied.” She posits a “non-Question Operator,” a –QOP, to license the indefinite wh-word. But her observation about “contexts where the truth of a proposition is not asserted/ implied” suggests that her –QOP may be subsumed under the nonveridicality operators that license -oo. In Chinese, where the indefinite interpretation of wh-element is not signaled in the morphology by a disjunctive particle (unlike in Kannada), there are restrictions on this interpretation, the question interpretation taking precedence (Li 1992:141ff.).
Jayaseelan (2001) identifies the disjunctive particle -oo with the homophonous yes/no-question morpheme in Malayalam. The Kannada yes/no-question marker is -aa, but -oo can occur instead of -aa in a yes/no question as well. However, two facts (illustrated in (33)) argue that -oo is not simply a variant of aa. First, aa cannot occur in constituent questions, but -oo can (as has been illustrated). Second, matrix constituent questions with -oo have an “embedded” or indirect reading; they are understood as complements to an unexpressed verb like ‘wonder’:
That matrix -oo questions are not requests for information but expressions of puzzlement (“rhetorical questions”) is suggested by a context in which (33) can be uttered—for example, on perceiving a clue that someone had been there in one’s absence (“Who could have come?”). Thus these questions are used in polite contexts such as the following: Suppose at a concert you don’t know who the violinist is. You whisper to your neighbor. Rather than asking, “Who’s the violinist?”, you say: “Any idea who the violinist is?” Analogously, (34a) is more appropriate than (34b).
Yes/no questions with -oo (again) have an air of rhetoricity or challenge absent from yes/no -aa questions.
Also note that -oo does not occur in constituent questions in Malayalam, except in a disjunction with be, when it has a similar indirect reading (Jayaseelan 2001:68).12
I have argued disjunction to be polarity sensitive, building on and generalizing Higginbotham’s (1991) account of the conjunctive interpretation of disjunction. The polarity operators I have identified are principally Neg and Q (in declarative and interrogative disjunction), and modality and genericity (in the conjunctive interpretation of disjunction). English either can be licensed by any of the three operators modality, genericity (nonveridicality operators), or negation (an averidicality operator). English whether is generally considered to be wh+either, which suggests that either is compatible with the Q operator as well.
In Kannada (and perhaps in multiterm disjunction in English), it is the head -oo of disjunction that must be licensed. When -oo occurs in phrasal disjunction with a conjunctive interpretation, its licensors are modality and genericity. We must also presume that -oo can be licensed by Neg, to account for its occurrence in the garden-variety phrasal disjunction in (35), where we take Neg to be the default licensor for (nonconjunctive) disjunction.
Clausal -oo disjunction, however, has the interesting property that it must inevitably receive an interrogative interpretation; hence declarative-clause disjunction is accomplished by a Neg element illa. Thus the clausal counterpart of (35), shown in (36), is an alternative question.
I have shown that -oo must adjoin to a Q generated in Comp. Question disjunction always takes clausal scope; in disjunction with less than clausal scope, therefore, Q is either not generated as a licensor or is uninterpretable if generated. Hence phrasal -oo disjunction in (35) receives a noninterrogative interpretation.13
Q in Kannada is always nonovert; thus the Q+oo of interrogative-clause disjunction surfaces as a bare -oo. In English, on the other hand, wh is a strong feature, and Q+either surfaces as wh (the reflex of the Q operator) + either (i.e., whether). We now have an interesting difference between the two languages with regard to the default interpretation of disjunctions with clausal scope. English or when it coordinates clauses is interpreted as either-or (i.e., as declarative disjunction) unless whether explicitly surfaces.14 Kannada -oo when it coordinates clauses has the interpretation of whether-or, and a Neg illa must explicitly occur for declarative clause disjunction to obtain. Now, I have said that Q in Kannada is never overt; and I have proposed that there is a Neg in English (the licensor for either) that is never overt. Hence in English as in Kannada, the default interpretation of clausal disjunction is the one which is licensed by a covert element; and this interpretation correlates inversely with the strength of the wh-feature in these languages. Intuitively, a weak feature may always be presumed to be present, unless otherwise indicated. Moreover, the principle “Assume a Q licensor unless Neg is overt” for Kannada, and its converse in English, is reminiscent of the Elsewhere principle in phonology (extended by Kiparsky 1998 to semantics), where a more general rule is “blocked” by the application of a specific rule.
Blackwell Publishing Ltd, 2003. Published by Blackwell Publishing, 9600 Garsington Road, Oxford OX4 2DQ, UK and 350 Main Street, Malden, MA 02148, USA
* An earlier version of this paper was presented at GLOW at Nanzan University in 1999. I thank Probal Dasgupta and K. A. Jayaseelan for comments, and Jim Huang for drawing my attention to Audrey Li’s paper; I alone am responsible for my use of their comments. I also thank the two anonymous referees, whose comments and puzzlement helped me to clarify my ideas and the presentation.
1. Kayne (1994:143, n. 2) suggests that English both, either, and neither occur in the specifier of the conjunction or disjunction. Alternatively, we may treat either and whether as heads: either being a variant of or that appears in the first disjunct when there are only two disjuncts and that has the morphological property of combining with wh to yield whether. The idea that initial or may be converted to either is credited by Anandan (1993:22) to J. R. Ross.
2. The coordinated XPs in (3) are also arguably DPs and not PPs, however. The “P-head” in (3b) takes a nominal Case marker, and the “P-objects” in (3) are marked genitive, like the specifiers of DP. Perhaps the category P is unattested in Kannada.
3. Anandan (1993:202, n. 1) notes the parallel fact in Malayalam, a sister Dravidian language, that a tensed VP conjoined with -oo yields a yes/no question. Earlier (p. 162), he appears to consider a “split” between phrasal and clausal conjunction, suggesting that coordinators in Malayalam select a [–V] category. (See also Jayaseelan 2001:65, n. 1.)
4. In (9), the first illa need not be overt. (Sridhar [1990:103] considers the use of illa in both disjuncts to be characteristic of the spoken language.) -oo must occur in both disjuncts, or not at all. I return to a discussion of the co-occurrence of illa and -oo in n. 13.
5. A simple diagnostic is that one can draw the valid inference ‘John plays chess’ from (10), whereas from (i), one cannot draw the inference ‘John played chess yesterday’:
My thanks to Jayaseelan for suggesting this diagnostic to me.
6. It is tempting, but I believe erroneous, to consider (15b) an instance of “exclusive” disjunction (under the reading ‘I shall play one of the two games, but not both.’) Whether or not the exclusive/inclusive distinction of logic obtains in natural language has been debated by philosophers and logicians, but it is perhaps the case that natural language has only the inclusive disjunction. (Pragmatics may force the exclusive reading, as in ‘It is a boy or a girl.’) The difficulty in extending the truth-conditional approach to multiterm disjunction in natural language has been discussed in the relevant literature, as has also the dubiety of putative instances of exclusive disjunction, and the absence of a clinching example (of a disjunction that is false because both its disjuncts are true).
I claim that illa disjunction does not allow us to conclude p from p or q; that is, it disallows the conjunctive reading.
7. The lexical meaning of either restricts its range to precisely the two-disjunct case. Hence we must allow disjunction in general to carry quantification, in multiterm disjunction in English as well.
8. This analysis of illa disjunction entails that declarative-clause disjunction in Kannada does not have a conjunctive reading, which appears to be the correct prediction. In English as well, when either appears in initial position with the entire clause in its scope, it seems to me to lack the conjunctive reading: cf. ‘Either John plays chess, or he plays checkers.’
9. Correlative clauses in Kannada also instantiate -oo:
Jayaseelan (2001) notes the fact that -oo functions as a relative particle for Malayalam, and seeks to offer a unified explanation for the various occurrences of -oo.
10. There is a subtle semantic difference between (26a) and (26b), comparable to what obtains in the corresponding English matrix yes/no-question disjunctions “Is he hungry or not?” and “Is he hungry, or what?” Notice that English permits what-disjunctions in matrix yes/no questions, where whether is not present. This suggests that interrogative whether-complements cannot take or what because of the presence of either in whether.
11. See Jayaseelan 2001 and Madhavan 1997 for an account of parallel Malayalam data. As Jayaseelan notes (pp. 71–72), the disjunctive particle does not have to be immediately suffixed to the question word to yield a quantifier reading; Case markers and lexical material can intervene: “-oo gives a question reading only when it is clause final (i.e., in C); otherwise it yields a quantifier.”
12. Given the parallels between -oo and whether, it is interesting that English does not allow matrix whether (‘*Whether we can have some chocolate?’). A possible explanation is the site of finiteness in these languages. Kannada situates finiteness in a MoodP in the C-system, allowing “matrix infinitive” verbs in negative clauses (Amritavalli 2000). Note also that if finiteness is required to license matrix Q+oo, FinP must be higher than ForceP in Kannada (in Rizzi’s [1997] articulated complementizer system).
13. In a disjunctive such as (35), nothing prevents illa and -oo from both occurring, although (as mentioned previously) there is a clear preference for only one of these elements to occur. My intuitions are consistent with Sridhar’s (1990:103) observation that “occasionally, both devices [-oo and illa] may be found, somewhat redundantly, in the same sentence.” The example Sridhar cites to illustrate his point is intriguing, however. It is an instance of a clausal -oo disjunction with an interrogative interpretation, where an illa “intervenes between” the disjuncts. (The example here is his example (367).)
Given that -oo and illa signal interrogative and declarative disjunction respectively, (i) should be as bizarre or incoherent as the corresponding English example where an either occurs in a whether-disjunction: *‘One does not know whether either Rupa will come or her husband will come.’ But it is not so. The reliability of the data in (i) is confirmed by (ii), an actual example taken from a letter.
Two observations are in order. Both examples are unambiguously question complements, which license -oo. Omitting the matrix predicate yields an alternative question. A declarative reading (‘They will perform the wedding in Pondicherry or in Mysore’) is not possible, unless -oo is dropped. Thus illa can occur “inside” -oo-disjunction, but the complementary case, with -oo “inside” declarative disjunction, does not occur.
Notice now that the “redundant” illa in (i) and (ii) occurs in the second disjunct. Such a “redundant” illa cannot occur in the first disjunct. The examples in (iii) and (iv), with both illa and -oo in the first disjunct, have the incoherence of the translations with whether and either cooccurring.
If, as Larson (1985:242, n. 15) proposes, the general structure for disjunction allows for “any finite string of α categories, including a string consisting of only a single α,” we might expect the interrogative or noninterrogative feature to be specified in the first disjunct once for all, given that such specification does not usually vary within the terms of the disjunction. (That is, a sequence such as “He is coming. Or are you going?” must be interpreted as two sentences, not as a disjunction.) Then it is illa in the first disjunct that gives the disjunction its declarative interpretation, and this illa cannot co-occur with -oo. (Recall that -oo must occur in both disjuncts, or not at all [see n. 4]; thus -oo cannot be dropped from the first disjunct but allowed to surface in the others.)
14. Or unless the presence of Q is signaled by I-to-C movement of Tense, in which case an initial either cannot occur.
References
Amritavalli, R. 2000. Kannada clause structure. In Yearbook of South Asian Languages and Linguistics, ed. R. Singh, 11–30. New Delhi: Sage India.
Anandan, K. N. 1993. Constraints on extraction from coordinate structures in English and Malayalam: An ECP approach. Ph.D. dissertation, CIEFL, Hyderabad, India.
Giannakidou, A. 1997. The landscape of polarity items. Ph.D. dissertation, University of Groningen, the Netherlands.
Higginbotham, J. 1991. Either/or. In Proceedings of NELS 21, ed. T. Sherer, 143–155. Amherst, Mass.: GLSA Publications.
Jayaseelan, K. A. 2001. Questions and question-word incorporating quantifiers in Malayalam. Syntax 4:63–93.
Kayne, R. S. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press.
Kiparsky, P. 1998. Aspect and event structure in Vedic. In Yearbook of South Asian Languages and Linguistics, ed. R. Singh, 29–61. New Delhi: Sage India.
Larson, R. 1985. The syntax of disjunctive scope. Natural Language & Linguistic Theory 3:217–264.
Li, Y.-H. A. 1992. Indefinite wh in Mandarin Chinese. Journal of East Asian Linguistics 1:125–155.
Madhavan, P. 1997. How do languages create quantified expressions: A cross linguistic investigation. In Phases and Interfaces of Morphology: Vol. 2, eds. M. Hariprasad, H. Nagarajan, P. Madhavan & K. G. Vijayakrishnan, 206–219. Hyderabad, India: CIEFL.
Rizzi, L.1997. The fine structure of the left periphery. In Elements of grammar: Handbook of generative syntax, ed. L. Haegeman, 281–337. Dordrecht: Kluwer.
Sridhar, S. N. 1990. Kannada. New York: Routledge.