K. A. Jayaseelan and M. Hariprasad
This paper examines Place Deixis in the Dravidian pronominal system. The theoretical interest of this topic is that the puzzles it throws up make us reexamine Place Deixis in English and lead us to the claim that (universally) all referring nominal expressions contain place deixis, i.e. are marked as ‘proximal’ or ‘distal’. This is of course contrary to the commonly accepted position, which is that some nominal expressions are marked as proximal, e.g. this boy, some are marked as distal, e.g. that book, and others are unspecified for the place deixis feature, e.g. John, he, thebook.
As a consequence of the aforesaid universalist claim, we postulate a DeixP in the extended nominal projection. We further suggest that the referential index of a referring nominal expression is a matter of the interpretation of DeixP, whose presence therefore is crucial in the semantics of reference.
Superficially, the proximal/distal distinction is not of much consequence in many languages, e.g. English. English has only one set of pronouns (he, she, it etc.), which does not mark this distinction. The distinction is marked only in locative adverbs (here/there) and in the demonstratives (this/that and these/those). But in the South Asian language area, the distinction pervades the pronominal system.
It is a well-known fact about Dravidian languages that its pronominals come in proximal/distal pairs. E.g.1
I.e., in Dravidian, it would seem not possible to use a pronoun without specifying whether the entity referred to is proximal or distal.
As can be seen, the alternants in (1) differ with respect to the choice of an initial vowel— /a/ or /i/. The reason for the choice is fairly transparent: the Dravidian demonstratives are aa ‘that’ and ii ‘this’. There seems to be a general principle operating in Dravidian (argued for in Anandan 1985) that in a DP, if (and only if) the head noun is null, its PNG features are marked on the preceding element in the DP. Thus:
If we take -an, -aL, -tə, and -ar to be the agreement markers for (respectively) 3rd person singular masculine, 3rd person singular feminine, 3rd person singular neuter, and 3rd person plural [+ human], we derive the pronouns of the language straightforwardly. The distal pronouns are derived by adding the agreement markers to the demonstrative aa ‘that’; and the proximal pronouns are derived by adding the agreement markers to the demonstrative ii ‘this’. We also account for the fact that the pronouns of the language come in proximal-distal pairs.
Of course, the pairwise disposition of the pronominal system is not the only one attested in languages. In Bangla—an Indo-Aryan language of eastern India which has been greatly influenced by Dravidian—, there are three sets of pronouns: e (‘this-he/she’), o (‘that-he/she’) and Se (‘previously mentioned-he/she’), the last used exclusively for picking up reference from the discourse (never used ostensively).3 In this system, we find, besides a proximal-distal distinction, a pronoun which is neutral with respect to that distinction.
Let us note (before we proceed further) that there is an air of paradox in this pervasive presence of the place deixis feature in the pronominal system.4 A binding theory conditioned by languages like English has all along assumed that meanings like ‘that’ and ‘this’ (typically accompanied by ostension) are to be found only in R-expressions.5 R-expressions and pronouns obey different binding principles. Therefore, a pronominal system which incorporates these meanings is a problem. Do the elements of this system tolerate, or not tolerate, a c-commanding antecedent in the sentence outside their local domain? (I.e., do they behave like R-expressions or like “true” pronouns?) This is the crucial question. Actually, the answer to this question is not simple or straightforward; and it is to the puzzle posed by this that we now turn.
Our problems begin when we notice that distal and proximal pronouns do not behave alike. The distal pronouns are well-behaved pronouns. They obey Principle B, cf. (3a); but are otherwise free to refer to (or be coreferential with) anything. They can have a c-commanding or non-ccommanding antecedent in the sentence, cf. (3b) and (3c); or can refer (extrasententially) to something in the discourse, cf. (3d); or be bound by a quantifier, cf. (3e); or be used for ostension (i.e., to point out something in the context), cf. (3f).
But compare the proximal pronouns in the same contexts:
Only the ostensive use, (4f), seems to be licit. (4a), of course, does not distinguish between Principle B and Principle C—the sentence could be out because of the violation of either. Looking at (4b) or (4e) alone, one might conclude that the so-called “pronoun” here is an R-expression. But (4c) shows that it cannot have even a non-c-commanding antecedent; and (4d) shows that it cannot have an antecedent in the discourse. Obviously, the restrictions on the proximal pronoun are not anything which the binding theory can handle. The proximal pronoun (it would seem) cannot pick up its reference from an antecedent at all; it can only refer directly to something in the non-linguistic context.
Now given this picture, the reader might think that the paradox (or problem) that we noted is easily resolved. Our problem (recall) was that the meanings ‘this’ and ‘that’, which are found only in R-expressions, appeared to be incorporated into pronouns. But given the data in (4), one can say that the so-called proximal pronoun is not a pronoun at all; it is simply a demonstrative, as (indeed) its morphology would lead us to expect. It is the distal pronoun which has deviated from the compositional meaning of its morphology: it has lost—at some point in its history—the meaning of aa ‘that’; it has no place deixis feature at all now, but is simply a ‘regular’ personal pronoun (like the English pronouns).6
But a closer consideration shows that this account is not a sufficient solution to the problem. We shall now argue that the right solution is in terms of recognizing that there is a fundamental asymmetry between the distal feature and the proximal feature.
As we know, two R-expressions may corefer. We do not say (in such a case, currently) that one of them picks up its reference from the other, but that they independently pick out the same individual from the domain of discourse. In other words, this is treated as a case of “accidental” coreference.7 This type of accidental coreference can occur across sentences, cf. (5a), or intrasententially, cf. (5b):
Now (as we said), in the binding theory of GB, the demonstratives (in so far as there has been a binding theoretical claim about them at all) have hitherto been classified as R-expressions. And indeed, if we consider only the distal demonstratives, they behave like R-expressions, cf.
But when we look at the proximal demonstratives, the picture is different. Look at the following minimal contrast:
The binding theory has nothing at all to say about this contrast. Recall that in the case of the Malayalam pronouns, we tried to explain away our problem by saying that, whereas the so-called proximal pronouns are actually demonstratives, the distal pronouns are true pronouns, having undergone a shift of meaning whereby they lost the meaning of ‘distal’. But it is impossible to say a similar thing about the English data: that boy and this boy very clearly contain demonstrative elements that have not lost the meaning of ‘distal’ and ‘proximal’.
There is a general principle (uncontroversial, we think) that two nominal expressions that corefer cannot have conflicting features. We need this principle to explain some bad cases of referential dependence (cf. * Mary’simother loves himi), and of “accidental” coreference (cf. * Although Maryiis a brilliant linguist, I don’t like the mani). We would like to appeal to this principle to explain (7). What prevents us from doing this is the widespread assumption that R-expressions—unless they are explicitly marked as ‘proximal’ or ‘distal’ by this or that—are not distal or proximal (i.e. not marked for the place deixis feature). Given this assumption, the name John is unmarked for the feature distal/proximal. Then, there can be no conflict of features between John and either that boy or this boy. And we cannot account for the contrast in (7).
We want to propose that the above-mentioned assumption is wrong, and that all R-expressions are marked as distal by a default rule, unless explicitly marked for the place deixis feature. I.e., we have the following universal principle:
(Given this principle, the boy and that boy are both distal, the only difference being that the boy is marked distal by a default rule, whereas that boy is distal by virtue of a lexical feature of that.8)
An immediate advantage of (8) is that we can account for the contrast in (7), which the binding theory could not handle: that boy can corefer with John because the latter is distal; but this boy cannot corefer with John without conflict of features.
An additional (and for us, more important) advantage of adopting (8) is that we can now continue to treat the proximal members of the Malayalam pronominal system as pronouns (as traditional grammars have always done). The problematic instances of bad coreference in the sentences of (4) are bad because of a conflict of features. The prediction now is that if we take parallel sentences in which the antecedents are explicitly marked as proximal, we should get different grammaticality judgements. This is indeed the case, cf. (9):
(9a) shows that the proximal pronoun obeys Principle B. And (9b) shows that it does not obey Principle C. In contexts where neither of these binding principles apply, the pronoun is free to take its antecedent anywhere, as shown by (9c-d). (9e) shows that the pronoun can be bound by a quantifier. Thus, the proximal pronoun is seen to be a regular, well-behaved pronoun, going by all the familiar tests.
The English pronouns, we shall say, are lexically marked as [α proximal] in view of data like the following:
Observe that we are not saying that the English pronouns have no feature [+/– proximal]. It is important that all definite pronouns have a place deixis feature, for it is our claim that all nominal expressions that have a referential index—i.e. all R-expressions and definite pronouns—have a place deixis feature.9,10
Malayalam distal pronouns seem (on a superficial examination) to behave like English pronouns, cf.
But the situation is more complicated.
Dasgupta (1992) has pointed out that Bangla e (prox.), which functions both as a demonstrative pronoun and as a demonstrative modifier, shows difference of behavior in the two functions. e, when it is used as a demonstrative pronoun (i.e., when it occurs by itself, unaccompanied by a lexical noun), can corefer only with a [+ proximal] expression; but when it is used as a demonstrative modifier (i.e., when it modifies a lexical noun), the DP it occurs in can have its reference picked up by either a proximal pronoun or a sequent (Se).11 Cf. (12) (adapted from Dasgupta’s examples; er and tar are the genitive forms of e and Se respectively):
Similar facts obtain in Malayalam. iwan (‘he (prox.)’), for example, can corefer only with a [+ proximal] expression, but ii kuTTi (‘this child’) can have its reference picked up by either a proximal or a distal pronoun:
Dasgupta’s solution for Bangla was to say that e (prox.) is [+ deictic], but a lexical noun is [- deictic]. In a phrase like e lokTa (‘this man’), either e or lokTa may percolate its feature upward to the phrasal node, which can therefore end up as either [+ deictic] or [- deictic]. In the former case, only the [+ deictic] e can pick up the DP’s reference; but in the latter case, the sequent Se (which Dasgupta analyzes as [- deictic]) can pick up the DP’s reference.
We cannot (obviously) adopt this solution in our account, because we are claiming that all referring nominal expressions contain a deictic element. (Therefore, the sequent pronoun Se should also be [+ deictic].) Also, the proposed feature percolation mechanism is problematic, cf. (7b) (repeated below) or (15):
In (7b), if this boy can be [- deictic], it should be able to corefer with John. In (15), if both ii kuTTi ‘this child’ and aa maNDan ‘that dolt’ can be [- deictic], there ought to be nothing preventing their coreference.
Looking for another explanation (then), let us say that the Malayalam distal pronoun is actually ambiguous between a “true” distal pronoun and a ‘sequent’ pronoun (i.e. one whose function is to pick up reference from the previous discourse, cf. Bangla Se); see n. 3.12 This assumption takes care of data like (11), (13b) and (14b). In the case of (13a) and (14a), the question now is: why cannot awan, qua sequent pronoun, pick up the reference of iwan?
The solution may lie in recognizing that the proximal pronoun is (in this respect) in a class with first and second person pronouns. Purely at the observational level we can say that a first or second person pronoun can pick up its reference only directly from the context of utterance or from another instance of the same pronoun. Also, its reference can be picked up only by another instance of the same pronoun. The problem presented by this last fact has hitherto been concealed because we have been operating in terms of a theory which postulated three person features in languages. But if it is the case that there are only two person features, namely first and second person (Benveniste (1966), Kayne (1998)), and that the so-called third person pronouns are simply “determiner pronouns” (Kayne’s term), the concealed problem surfaces: why cannot a determiner pronoun pick up the reference of a first or second person pronoun—exactly like it picks up the reference of (say) a definite description or another determiner pronoun? The problem with iwan, we wish to suggest, is the same problem.
In the next subsection we argue that first and second person pronouns have the feature [+ proximal]. The above problem then is one having to do with [+ proximal] pronouns.13
Let us say that a sequent pronoun cannot simply ‘pick up’ the index of a referential phrase in the preceding discourse; it must arrive at the reference by means of the descriptive content of the antecedent. Since pronouns have no descriptive content (they only have grammatical features), it follows that a pronoun cannot be the antecedent of a sequent pronoun.
Now if (as we said) awan is ambiguous between a distal pronoun and a sequent pronoun, we get the following predictions: As a distal pronoun, awan cannot corefer with any nominal expression which is marked [+ proximal], including iwan ‘he (prox.)’ and ii kuTTi ‘this child’. (However it can corefer with another distal pronoun awan.) As a sequent pronoun, awan cannot take any pronoun as its antecedent; and as a subcase, it cannot take iwan as its antecedent. Therefore iwan cannot be the antecedent of awan in either interpretation of the latter. (The same explanation will do for why Malayalam awan or English he cannot be coreferential with the first and second person pronouns.)14
Place Deixis is not always binary, i.e. proximal-distal, in the world’s languages. Some languages have a three-way distinction, proximal-medial-distal; two such modern languages are Spanish and Japanese. Old Dravidian also had this three-way distinction, which was (moreover) incorporated into the pronominal system, so that there were pronominal triplets:
In languages having a ternary system of Place Deixis, one value—possibly ‘distal’ again—may have functioned as the default value for expressions not explicitly marked for this feature.
Given a possible three-way distinction in Place Deixis, it is tempting to assimilate the (customary) three-way Person distinctions of pronominal systems to it. In fact, such claims have been made.16 However, such a reduction—however enticing—appears unlikely to succeed, since the agreement systems of languages (which are sensitive to Person distinctions) are not sensitive to the proximal-distal distinction.17
All that we can say is that personal pronouns, like other referring nominal expressions, contain place deixis. First person pronouns are clearly proximal, since proximality is determined with respect to the Speaker’s position at the time of utterance. With regard to second person pronouns, we wish to suggest (in agreement with Andrewskutty 1989) that they (too) are proximal. Some evidence for this comes from locutions (in Malayalam) in which a demonstrative is used with a first or second person pronoun for emphasis; only ii ‘this’ is permissible, aa ‘that’ is not:
Third person pronouns are [α proximal], or perhaps ambiguous between distal and sequent pronouns, in English (as we suggested earlier). Dravidian third person pronouns (as we know) have different forms when they are proximal and distal; but the distal pronouns (we suggested) are actually ambiguous between distal and sequent pronouns.
We claimed that all referring nominal expressions have Place Deixis. We can express this correlation between reference and Place Deixis if we say that the referential index of a referring nominal expression is not something other than Place Deixis; that the index is the “interpretation” of Place Deixis (of ‘this’ and ‘that’).19 The suggestion (essentially) is that a non-linguistic object like a referential index should have a linguistic encoding. And that Place Deixis is that linguistic encoding.
We postulate a Deixis Phrase (DeixP) in the extended nominal projection, possibly immediately under Case Phrase (KP). It could be the absence of this DeixP that constitutes the “deficiency” of anaphors in comparison to ‘regular’ pronouns (see n. 9).
1. All our Dravidian examples are from Malayalam. The facts of other Dravidian languages are parallel; but see n.2.
2. In English the principle seems to operate only one way: if—but not only if—the head noun is absent, the preceding element is marked for PNG features. Thus, the with a null head noun becomes this/these/that/those (in this case, the preceding element is marked for both number and place deixis). Again, in the rich, the adjective is presumably marked (albeit non-overtly) for plurality, even though English adjectives are (otherwise) not marked for PNG features.
The analysis of forms like nallawan as ‘Adjective + Agreement Marker’ is not uncontroversial; the traditional analysis of these forms is as ‘Adjective + Pronoun’—i.e. nallawan is nalla + awan (see also Mohanan 1982). In fact, Hariprasad (1998) shows that the parallel Telugu forms clearly lend support to the traditional analysis. (Advocates of this view treat a form like awan as morphologically simple; and account for the awan-iwan alternation by deriving iwan from ii + awan; see Hariprasad (1998) for arguments.) Let us merely note here that the issue of this controversy does not really affect the main claims of this paper.
3. Dasgupta (1992) calls Se the “sequent” pronoun. He adds:
Bangla is unusual in its retention of the three-way
distinction of Proximal vs. Distal vs. Sequent; most
other Indo-European languages have merged the Distal
and Sequent series, and there is some evidence that
Bangla is moving in that direction as well.
4. A point made in Dasgupta (1992).
5. Justifiably, in English (it would seem). ‘That’ and ‘this’—with or without a following nominal head—are R-expressions in English. Dasgupta (1992) gives the following examples:
6. This was the position adopted in Jayaseelan (1999a, 2000), Hariprasad (1998).
7. But see Kayne (2001) for a different view.
8. A caveat. We said that an R-expression can be proximal only if it is explicitly marked as proximal. Actually, there is also another way it can be proximal: one can explicitly assert that the entity it refers to is physically close to the Speaker, e.g. ‘John is here’. One can subsequently refer to ‘John’ using a proximal pronoun, cf. (i), or a proximal R-expression, cf. (ii):
One can also “assert” proximality by pointing at the referent of the R-expression while speaking.
Thus, (iii) is possible if the speaker utters it while pointing at ‘John’:
The same principle of perceived nearness may be what underlies a discourse use of ‘this’, which does not mean physical proximity but seems to signify that a fact or entity has been mentioned in the discourse just a short time before:
(The last example is a “found” example.)
9. Anaphors may not have this feature, because anaphors may not have a referential index. (We are going to claim a correlation between the two.) Hariprasad (1998) argues that what distinguishes anaphors from pronouns is not “feature deficit”—as has been suggested by Burzio (1991), Huang and Tang (1989), Reinhart and Reuland (1991, 1993) and Jayaseelan (1997), among others—, but the fact that anaphors have no inherent referential index (‘R-index’, in the sense of Huang and Tang (1989)). (Hariprasad (ibid.) points out that there are many anaphors that are no “poorer” in respect of phi-features than many pronouns.)
10. Two points need to be made about the notion of ‘deixis’ here. Firstly: Although ‘deixis’ is sometimes interpreted narrowly, to mean pointing at an entity in the “real world,” we are obviously using the term in a broader sense. Thus, we have already shown proximal deixis being used to point to the preceding discourse; cf. examples (iv) and (v) of n. 8, the latter example repeated below as (i):
Here, ‘this medical speciality’ refers (as it were) “anaphorically” to ‘reanimatology,’ behaving very much like a pronoun.
A second point about deixis has to do with the philosophers’ distinction between the ‘referential’ and ‘attributive’ use of a definite description (Donnellan 1971). The distinction probably does not come within the purview of linguistic semantics. (Donnellan himself described the two uses as a “duality of function” (Donnellan 1971: 100).) Kaplan (1978) attempts to sharpen the distinction by suggesting that in the referential use a definite description contains a demonstrative element, which he proposes to represent as “Dthat”; so that (ii) (in its referential use) is properly represented as (iii):
(The “Dthat” would be absent if (ii) is used attributively.) Again, it is doubtful if Kaplan’s “Dthat” has anything to do with linguistic semantics. In any case, the natural language demonstrative ‘that’ is uncooperative in functioning as a signal of referential use in the philosophers’ sense. For, it can occur in a definite description which is clearly attributively used, cf.
Clearly (then), having a place deixis feature (whether proximal or distal) does not necessarily make a nominal expression ‘directly referential’ (in the sense of being referential to an entity in the “real world”).
But there may be a difference of emphasis between ‘the NP’ and ‘that NP’ (although both, we are claiming, are [+ distal]). This would account for (v) and (vi) (which may be compared with (iv)):
The difference of emphasis would also perhaps account for the following data (pointed out by a reviewer): If you see someone watching the Harvard-Yale football game on TV, you could say (vii) but not (viii):
Again, (ix) but not (x) is okay, as spoken by someone in a smoggy city at sundown:
11. Or even by o (dist.) (Gautam Sengupta, p.c.). (This could indicate that o is acquiring the function of a sequent pronoun, see n.3.)
12. The English third person pronoun (too) is perhaps similarly ambiguous between a ‘distal’ and a ‘sequent’ pronoun; so, in the sentences of (10), the pronoun picks up the reference of ‘this boy’, qua sequent pronoun. (We can, if we adopt this position, abandon the claim that English pronouns are [α proximal].)
13. There is a set of systematic exceptions to the generalization that a [+ proximal] pronoun’s reference can be picked up only by an identical pronoun. The exceptions are the long-distance reflexives which are “unmarked for person”—e.g. Japanese zibun, Chinese ziji, Korean casin. In the case of ziji and casin, the actual reflexive pronoun could be pro –i.e. ziji aand casin could be underlyingly pro-ziji and pro-casin (Jayaseelan 1997). The “empty pronouns” of Malayalam—which can occur in all positions except the object position of a postposition—are also probably pro (Jayaseelan 1999b). This pro can pick up the reference of a first/second person pronoun or iwan, cf.
14. The Bangla pronoun Se must actually be ambiguous between a distal and a sequent pronoun, because it can pick up the reference of the distal pronoun o, cf. (i) (data provided by Mahasweta Sengupta). (tar, to remind the reader, is the genitive form of Se.)
Our claim about sequent pronouns can be restated in terms of Kayne’s (2001) proposal that pronouns and their antecedents are generated in a “doubled” structure (from within which the antecedent is subsequently moved out, see Kayne (2001) for details). A sequent pronoun (we can say) takes only a “full” DP as its double; unlike a distal or proximal pronoun, which can also take a null Topic as its double.
15. Possibly, Old Dravidian had—besides ii ‘this’ and aa ‘that’—a demonstrative uu denoting ‘in-between’ distance, which was the source of uwan.
Ramakrishna Reddy informs me that in the Munda languages (tribal languages which are part of the Dravidian family), there is a five-way distinction: proximal, medial, distal (can be seen), distal (cannot be seen), and one other. I have not been able to investigate these distinctions.
16. Thus Karl Brugmann (Die Demonstrativpronomina der Indogermanischen Sprachen, 1904) identified the First Person with the proximal category of place deixis, the Second Person with the medial category, and the Third Person with the distal category. (Our information about Brugmann is derived from Fillmore (1966).)
Dasgupta (1992) also has pursued this line of thought but suggested different identifications. His schema is:
17. For example, consider Brugmann’s identification of third person with ‘distal’. English agreement treats an R-expression which is explicitly marked as proximal (e.g. this boy) as third person. Again, in Dravidian, a proximal pronoun like iwan induces third person agreement.
Note that in Bangla, Dasgupta’s [- Illocutionary] pronouns (i.e. third person pronouns) also are differentiated w.r.t. place deixis: o (dist.) and e (prox.).
18. Some further suggestive evidence for saying this comes from the semantics of COME and GO. Universally, the place to which one comes is the place where I am or YOU are. (See Fillmore (1966) for a discussion of the semantics of COME.) Also, Andrewskutty (1989) provides dialectal data like the following, which shows an interesting use of the proximal pronoun for referring to the Speaker or the Hearer:
(The point is that the distal pronoun is never used in this function.)
However, YOU can apparently be marked as ‘distal’ if one explicitly asserts that it is distal; cf. ‘You are there’, which is not a contradiction.
19. Cardinaletti and Starke (1994) have a suggestion that the referential index is the “interpretation” of Case. But Case—it seems to us—is the NP’s interface with the predicate. If “referential index” is the association of a linguistic expression with a non-linguistic entity, it seems unlikely that it should lodge in Case, since the NP’s relation to “the world” has nothing to do with the NP’s relation to the predicate.
Anandan, K.N. 1985. Predicate Nominals in English and Malayalam. CIEFL, Hyderabad, M.Litt. dissertation.
Andrewskutty, A.P. 1989. Deictic Properties of “Person” in Malayalam Dialects. International Journal of Dravidian Linguistics 18(2):13–26.
Benveniste, E. 1966. Problemes de linguistique generale, Gallimard.
Burzio, L. 1991. The Morphological Basis of Anaphora. Journal of Linguistics, 27:81–105.
Cardinaletti, A. & M. Starke. 1994. The Typology of Structural Deficiency. On the Three Grammatical Classes. Working Papers in Linguistics 4, University of Venice, 41–109.
Dasgupta, P. 1996. Pronominality and Deixis in Bangla. Linguistic Analysis 22:61–77.
Donnellan, K. 1966. Reference and Definite Descriptions. Philosophical Review 75:281–304. [Reprinted in Danny D. Steinberg & Leon A. Jacobovits (eds.) Semantics, Cambridge University Press, Cambridge, 1971, pp. 100–114; page reference to this volume.]
Fillmore, C.J. 1966. Deictic Categories in the Semantics of ‘Come’. Foundations of Language 2:219–227.
Hariprasad, M. 1998. Anaphorization in Telugu and English. Doctoral dissertation, Central Institute of English and Foreign Languages, Hyderabad.
Huang, C.-T.J. & Tang, C.-C.J. 1989. The Local Nature of the Long-distance Reflexive in Chinese. Proceedings of NELS 19, 191–206. GLSA, University of Massachusetts, Amherst.
Jayaseelan, K.A. 1997. Anaphors as Pronouns. Studia Linguistica 51:186–234.
Jayaseelan, K.A. 1999a. The Pronominal System of Malayalam. In K.A. Jayaseelan Parametric Studies in Malayalam Syntax, pp. 39–64. New Delhi: Allied Publishers.
Jayaseelan, K.A. 1999b. Empty Pronouns in Dravidian. In K.A. Jayaseelan Parametric Studies in Malayalam Syntax, pp. 14–25. New Delhi: Allied Publishers.
Jayaseelan, K.A. 2000. Lexical Anaphors and Pronouns in Malayalam. In Barbara C. Lust et al. (eds.), Lexical Anaphors and Pronouns in Selected South Asian Languages: A Principled Typology, pp. 113–168. New York: Mouton de Gruyter.
Kaplan, David. 1978. Dthat. In Peter Cole (ed.) Syntax and Semantics vol. 9, Academic Press, New York. [Reprinted in French, Uehling, Wettstein (eds.) Contemporary Perspectives in the Philosophy of Language, University of Minnesota Press, 1979.]
Kayne, R. 2000. Person Morphemes and Reflexives in Italian, French and Related Languages. In R. Kayne Parameters and Universals, pp. 131–162. Oxford: Oxford University Press.
Kayne, R. 2001. Pronouns and their Antecedents. Ms. New York University.
Mohanan, K.P. 1982. Grammatical Relations and Anaphora in Malayalam. MIT Working Papers in Linguistics 4:163–190.
Reinhart, T. & Reuland, E. 1991. Anaphors and Logophors: An Argument Structure Perspective. In Long-Distance Anaphora, ed. by Koster, J. & Reuland, E. Cambridge: Cambridge University Press. 283–321.
Reinhart, T. & Reuland, E. 1993. Reflexivity. Linguistic Inquiry 24:657–720.
