Anaphora in Dravidian
Summary and Keywords
The Dravidian languages have a long-distance reflexive anaphor taan. (It is taan in Tamil and Malayalam, taanu in Kannada and tanu in Telugu.) As is the case with other long-distance anaphors, it is subject-oriented; it is also [+human] and third person. Interestingly, it is infelicitous if bound within the minimal clause when it is an argument of the verb. (That is, it seems to obey Principle B of the binding theory.) Although it is subject-oriented in the normal case, it can be bound by a non-subject if the verb is a “psych predicate,” that is, a predicate that denotes a feeling; in this case, it can be bound by the experiencer of the feeling. Again, in a discourse that depicts the thoughts, feelings, or point of view of a protagonist—the so-called “logophoric contexts”—it can be coreferential with the protagonist even if the latter is mentioned only in the preceding discourse (not within the sentence). These latter facts suggest that the anaphor is in fact coindexed with the perspective of the clause (rather than with the subject per se). In cases where this anaphor needs to be coindexed with the minimal subject (to express a meaning like ‘John loves himself’), the Dravidian languages exhibit two strategies to circumvent the Principle B effect. Malayalam adds an emphasis marker tanne to the anaphor; taan tanne can corefer with the minimal subject. This strategy parallels the strategy of European languages and East Asian languages (cf. Scandinavian seg selv). The three other major Dravidian languages—Tamil, Telugu, and Kannada—use a verbal reflexive: they add a light verb koL- (lit. ‘take’) to the verbal complex, which has the effect of reflexivizing the transitive predicate. (It either makes the verb intransitive or gives it a self-benefactive meaning.)
The Dravidian languages also have reciprocal and distributive anaphors. These have bipartite structures. An example of a Malayalam reciprocal anaphor is oral … matte aaL (‘one person … other person’). The distributive anaphor in Malayalam has the form awar-awar (‘they-they’); it is a reduplicated pronoun. The reciprocals and distributives are strict anaphors in the sense that they apparently obey Principle A; they must be bound in the domain of the minimal subject. They are not subject-oriented.
A noteworthy fact about the pronominal system of Dravidian is that the third person pronouns come in proximal-distal pairs, the proximal pronoun being used to refer to something nearby and the distal pronoun being used elsewhere.
1. Anaphors, Pronouns and Long-distance Anaphors: An Overview
1.1. The Binding Theory of Early GB
Traditional grammars did not—or did not always—distinguish between anaphors and pronouns; pronouns were (more often than not) included among the anaphoric elements of a language, the term ‘anaphoric element’ being used loosely to mean any element that can pick up the reference of another referential expression, either in the sentence or in the discourse. But in the Chomskyan theory, at least since the advent of the Government-and-Binding (GB) theory (Chomsky, 1981), there has been a clear distinction drawn between anaphors and pronouns: anaphors are elements that obligatorily require an antecedent in the sentence. Pronouns, on the other hand, do not require an antecedent in the sentence; they can pick up their reference from the discourse or even from the context. Anaphors in fact are required to obey a stricter condition in GB: they need to be bound in a local domain; for an anaphor which is an argument of a verb, this local domain is the minimal clause containing it. (To put it simply, an anaphor must have its antecedent very close to it, within the same clause.) This ‘local binding’ requirement of anaphors is formulated in GB as Principle A of the binding theory. In contrast, pronouns obey a negative requirement: they should not be bound in the minimal clause. In other words, a pronoun—if it has an antecedent at all within the sentence—should not have it too close to it. This ‘anti-local binding’ requirement of pronouns is formulated as Principle B of the binding theory.
1.2. Long-distance Anaphors
The neat opposition of anaphors and pronouns proposed by early GB—the former requiring that they be locally bound and the latter that they not be locally bound—was soon disturbed by the discovery of long-distance anaphors in several languages. A long-distance anaphor is an anaphor that does not obey Principle A, in that it need not be bound in the minimal clause; its antecedent can be in a superordinate clause. But it is not like a pronoun either, because it requires an antecedent in the sentence; its reference cannot be determined simply from the discourse or the context. Long-distance anaphors were discovered almost simultaneously in several language families, e.g., the Scandinavian languages (Anderson, 1983), Chinese (Huang, 1984), Japanese (Ueda, 1984; Fukui, 1984), the Indo-Aryan languages (Gurtu, 1985), and—importantly for the topic of this essay—the Dravidian languages (Mohanan, 1982; Amritavalli, 1984).
2. The Long-distance Reflexive Anaphor taan
The Dravidian languages have a long-distance reflexive anaphor, which is taan in Malayalam and Tamil, taanu in Kannada, and tanu in Telugu. It is third person, singular, and [+human], but not marked for gender; it has a plural form taŋŋaL in Malayalam (with cognates in the other sister languages). Its antecedent can be at any distance from it; in particular, it can be outside the minimal clause containing the anaphor, cf. (1) (Malayalam data; all the data in this article are from Malayalam unless indicated otherwise):1
In (1), taan can take ‘John,’ ‘Peter,’ or ‘Bill’ as its antecedent.
Like other long-distance anaphors (e.g., zibun in Japanese, ziji in Chinese, sig in Icelandic), taan is subject-oriented:
But there are well-known exceptions; for example, in sentences with ‘psych verbs’ or psychological predicates—these are predicates that denote a feeling—the experiencer argument can be the antecedent of taan:
Here, ‘John’—which is not the syntactic subject of the sentence—can be the antecedent of taan. Again, in logophoric contexts—that is, in a discourse describing the thoughts, feelings, or point of view of a protagonist—the protagonist can be the antecedent of the anaphor; in fact, it is only in such contexts that taan can be discourse-bound:
The logophoric antecedent can also violate the c-command condition on the binding relation, cf.
In fact, the binding of taan by the experiencer argument in a sentence like (3) could also be a case of logophoric binding.
It has been noted that taŋŋaL, the plural of taan, can take split antecedents (Jayaseelan, 1997, p. 191):
It has been claimed that the ability to take split antecedents is a property of pronouns and that anaphors cannot do this, but this is obviously an incorrect claim (see Zribi-Hertz, 1989 for a discussion). The fact of the matter seems to be that there are different classes of anaphors, and it is shown in section 3 that there are indeed anaphors that cannot take split antecedents.
In Malayalam, taan is blocked from being co-indexed with its third person antecedent in a superordinate clause if an intervening subject is non-third person. (Such blocking effects were first described in Chinese by Huang & Tang, 1989.) Cf.:
Data like these—blocking, logophoric binding, being bound by the experiencer argument of a psychological predicate—suggest that taan could actually be co-indexed with the perspective of the clause (rather than with the subject per se) (see Jayaseelan (1998), where a Perspective head that projects a Perspective Phrase is postulated in the C-domain; the parallel claim has been made for Chinese ziji in Huang and Liu (2001); see also Sundaresan (2012)).
There have been attempts to reconcile the nonlocal binding of long-distance anaphors with the claim of GB that anaphors should be locally bound. The suggestions have been that these anaphors move up close to the antecedent by successive-cyclic adjunction in the covert component of the grammar, the debate being about whether the adjunction is to IP, that is, the clause (as maintained by Huang and Tang (1989)), or to I, that is, the inflectional head of the clause (as maintained by Pica (1987), Battistella (1989), Cole and Sung (1994), and others).
But why is taan an anaphor in the first place? Or to put it more generally, what makes a nominal expression dependent on another nominal expression for its interpretation? An early proposal in the GB literature was that long-distance anaphors are nominal elements that lack inherent reference and also one or more ‘phi-features,’ that is, person-number-gender features, and that it is this “deficit” that makes them anaphors (Burzio, 1991; Huang & Tang, 1989; Reinhart & Reuland, 1991, 1993; Jayaseelan, 1997). Thus, Chinese ziji is unmarked for any phi-feature; it is marked only for the feature [+human]. But in the case of taan, phi-feature deficit is not a plausible reason for its anaphoricity, because it is marked third person and singular; the only phi-feature it lacks is gender. In Jayaseelan and Hariprasad (2001), it is argued that all referential nominal expressions, both definite descriptions and pronouns, have as their highest projection a DeixP (Deixis Phrase) headed by a deictic element, and that it is the interpretation of this element that constitutes their referential index. It is also suggested that anaphors differ from pronouns only by the absence of DeixP. Let us say that it is this “deficit” that makes taan an anaphor.
2.1. The Anti-Locality Condition on the Binding of taan
Interestingly, taan is infelicitous if bound within the minimal clause when it is an argument of the verb; that is, it seems to obey Principle B of the binding theory. (This is indeed a paradox, because anaphors are supposed to obey Principle A—and not Principle B—of the binding theory of GB; but of course long-distance anaphors don’t obey Principle A, see section 1.2.) But the infelicity is weaker than with regular pronouns:
Similar weak Principle B effects have been noticed with other long-distance anaphors, like Japanese zibun, Korean caki and Scandinavian seg/sig (see Jayaseelan, 1997 for references).
But in a typical reflexive context, the direct object of a transitive verb refers back to the sentential subject (e.g., John loves himself). To function in this context, and to get around the Principle B effect, taan—and also the regular pronouns—employ two strategies in Dravidian. Or, to put it more particularly, Malayalam employs one strategy, and the other Dravidian languages employ another strategy.
2.2. The Emphasis Marker Strategy of Reflexivization
In Malayalam taan, or a regular pronoun (e.g., awan), takes an emphasis marker tanne; the complex form taan tanne /awan tanne can be coreferential with the minimal subject:
The emphasis marker strategy is parallel to the strategy of European languages and East Asian languages; thus the Norwegian reflexive form is seg selv (‘self-emphasis marker’), and the Japanese reflexive form is zibun zisin (‘self-emphasis marker’).
The way the emphasis marker insulates taan or a pronoun, so that the negative condition that a pronoun must not take an antecedent in the minimal clause is set aside, has been sought to be explained as follows (Jayaseelan, 1997): in the complex structure ‘[taan/pronoun tanne],’ taan/pronoun is in the specifier position of a phrase which has its own internal head (possibly tanne) that governs it, and therefore taan/pronoun need be free (i.e., not bound) only in this smaller structure. (The same explanation has been offered for Norwegian seg selv (Hestvik, 1990).) (See Reuland, 2016 for a discussion of the different strategies languages use to insulate a pronoun from the effects of Principle B. The Malayalam strategy falls under what Reuland calls “protection.”)
2.2.1. The Pronominal Use of Proper Names
Malayalam—but not the other Dravidian languages—instantiates the pronominal use of proper names. (Malayalam is not alone among the world’s languages in doing this; it has been reported also in some other oriental languages, e.g., Vietnamese and Thai (Lasnik & Uriagereka, 1988; Lasnik, 1989).) In the sentence:
the second occurrence of ‘John’ seems to be functioning as a pronoun. This conjecture is confirmed by an interesting fact: the repeated proper name, when it is an argument of the verb, exhibits Principle B effects. Thus (11) is infelicitous:
But (11) can be made fine if the emphasis marker tanne is added to the repeated proper name:
In Malayalam, a proper name can be used not only as a third person pronoun; it can be (and very commonly is) used for second person reference. (This is possibly a way to avoid having to choose between honorific and nonhonorific forms of address.) Thus the following sentence can be addressed to ‘John’:
In the second person use also, the repeated proper name exhibits Principle B effects, which can be set aside by the use of the emphasis marker:
Note that a sentence like (12) or (14b) contains a reflexive formed from a proper name. Since any proper name can be turned into a reflexive in this way, the reflexive forms of the language cannot be a closed set, and a syntactic process of reflexivization in the language must be assumed (Amritavalli, 1984).
2.3. The Verbal Reflexive Strategy
Tamil, Telugu, and Kannada, the other major Dravidian languages, do not depend primarily on the emphasis marker strategy to reflexivize a transitive predicate. Instead, they use a ‘verbal reflexive’ in the typical reflexive context. The verbal reflexive is a light verb koL- (‘take’), which is added to the perfect participle form of the main verb:
In Telugu, the direct object can even be dropped (in certain contexts) when the verbal reflexive is present (Subbarao & Lalitha-Murthy, 2000, p. 233):
In colloquial Tamil, the light verb kiT- (‘get’)—instead of koL- (‘take’)—seems to be the verbal reflexive, cf. (17) (Annamalai, 2000, p. 175):
The way the addition of koL makes the predicate reflexive has been sought to be explained in two ways. One explanation is that it brings about a valency reduction of the verb, making it monadic; but while the verb is now like an unaccusative, there is an implied meaning of agency like in the case of passives (Lidz, 2001, 2004). An alternative explanation is that koL affects the perspective of the predicate, giving it a self-benefactive meaning: the interpretation that koL induces is that the highest argument of the predicate ‘comes to hold’ the result state of the event described by the predicate (Sundaresan, 2012).
It should be noted that there are contexts where Tamil, Telugu, and Kannada also use the emphasis marker strategy that we saw in the case of Malayalam reflexive predicates. These contexts are the so-called dative subject constructions (Amritavalli, 2000, p. 65, Kannada example):
The dative subject predicates have been shown to be unaccusatives (a variety of intransitives) (Shibatani, 1999). Being intransitives, they are not amenable to a valency reduction operation by the addition of koL. This could be the reason why, in these contexts, the language has to resort to the emphasis marker strategy.
3. Local Anaphors: Reciprocal and Distributive
The reciprocal and distributive anaphors of Dravidian need to be bound by an antecedent in the local domain, and are in that sense well-behaved anaphors from the point of view of the binding theory of GB.
3.1. The Reciprocal Anaphor
The reciprocal anaphor has a bipartite structure; in this it is like the Italian reciprocal l’uno … l’altro (Belletti, 1982) and the English reciprocals each other and one another. In Kannada, the reciprocal is obbaru … obbaru (‘one person … one person’) (Amritavalli, 1984); in Malayalam it is oraaL … oraaL (‘one person … one person’) or oraaL … matte aaL (‘one person … the other person’). Each part of this structure is a noun phrase in its own right and takes its own case, cf.
In the bipartite structure oraaL1… matte aaL2, one member shows the case that the anaphor receives by virtue of its position in the clausal structure; the case of the other member is a copy of the case of the antecedent. This may not be quite clear in the above example, because the copied case here is nominative, which has a null phonetic realization. But consider (20):
This sentence has a (so-called) dative subject. (An experiencer predicate generally has a dative subject in Dravidian.) Note how this case is copied onto oraaL. The sequence of the two parts—copied case NP–autonomous case NP—can be reversed, but with the proviso that matte ‘other’ should come second; this is only because ‘other’ requires the element it is contrasted with to precede it:
In (22), the copied-case NP follows the NP with its own case.
There is no contiguity requirement on the two parts of the reciprocal; the parts can be embedded in larger phrases, in which case there can be considerable intervening material between the two NPs, cf.
(The separability of the two members of the reciprocal construction can be noted even in English, cf. ‘They should love each other’/‘They should each love the other.’)
Although the NPs oraaL and matte aaL are third person, there is no requirement that the antecedent should be third person:
As has been said, one member of the reciprocal structure must get its case by copying the case of the antecedent. It has been suggested that it is this case dependence that makes the reciprocal an anaphor; the claim is that the case agreement between the two NPs is an anaphoric relation (Amritavalli, 1984). An alternative explanation is that it is only the NP with its own case that occupies a θ-position; the NP with the dependent case is in a non-θ position. In LF, in order to be interpreted, the latter NP must adjoin to the antecedent so that it can share its θ-role (Jayaseelan, 1997; see also Heim et al., 1991 for the same explanation for English each other).
The proposal that the NP in a non-θ position must move up to its antecedent by successive-cyclic movement into all the intermediate Spec positions explains why the reciprocal anaphor is a strict anaphor that obeys Principle A of the binding theory. For if there is a filled Spec position in the path of the movement, it will block the progress of the NP towards the antecedent. Consider (25):
Here, the reciprocal can take only the embedded subject as antecedent; the matrix subject is inaccessible to the anaphor because of the intervening ‘specified subject.’ If the embedded subject is in the singular and therefore not a possible antecedent, the sentence is ungrammatical:
What has been illustrated in (25) and (26) is the opacity induced by an intervening subject. But the opacity created by an intervening tensed S boundary also affects the reciprocal anaphor. (This is comparable to English, cf. * ‘They thought that each other would win.’) Consider (27):
It is clearly the need of the NP in the non-θ position (i.e., the member of the reciprocal structure which has the copied case) to access the antecedent that induces the two opacities, namely the specified subject opacity and the tensed S opacity. The position of the other member of the reciprocal structure, namely the NP in the θ-position which has its own case, is irrelevant to these opacities. Consider (28) and (29):
These sentences are fine, because there is nothing that blocks oraaL from accessing its antecedent.
It should be noted that the reciprocal anaphor is not subject oriented:
Here ‘each other’ is anteceded by the indirect object.
Unlike the reflexive anaphor taan, the reciprocal cannot take split antecedents:
This is obviously because the member of the reciprocal structure which is in the non-θ position (i.e. the NP which has the copied case) needs to adjoin to its antecedent in the interpretive component of the grammar, and it cannot adjoin to two targets at the same time.
3.2. Some Other Reciprocals
There are some other forms that are used to convey the reciprocal meaning. In Malayalam, one can mention three such forms:
This form is a tat-sama borrowing—that is, a borrowing without any change from the original—from Sanskrit, and it retains the Sanskrit case-endings. The underlying form is:
The sandhi rules of Sanskrit give anyoonyam. The underlying form has a bipartite structure consisting of two pronouns, the first one showing the nominative case and the second one the accusative case. Consider (33):
The distribution of this form is very restricted. It occurs typically in the direct object position of a transitive verb, as in (33), but it can also occur as the indirect object:
But this form is impossible if one tries to embed it within an argument, to express a meaning like ‘They love each other’s children.’ This is possibly because it is ‘frozen’ and cannot take any case endings.
3.2.2. tammil (tammil)
This form is underlyingly:
The first part of it is an obsolete plural third person pronominal stem, related to taan and its plural form taŋŋaL; -il, as indicated, is the locative case-marker meaning ‘in.’ It is optionally reduplicated:
Since the stem is obsolete, it is doubtful if tammil is analyzed synchronically as a case-marked pronominal. Even if it were, it cannot be the argument taken by the verb in any of the sentences in which it occurs; thus, in (36), the verb sneehik’k’ (‘love’) is subcategorized for a direct object case-marked accusative and not for a locative PP. It is better to treat tammil (tammil) as an unanalyzed form in the contemporary language which carries the meaning of reciprocity.
3.2.3. aŋŋooTTum iŋŋooTTum
This third form is completely transparent: it literally means ‘that-way this-way.’ Cf.
This form is obviously adverbial.
These last-mentioned two forms, tammil (tammil) and aŋŋooTTum iŋŋooTTum, are clearly adjuncts and do not occupy argument positions in the sentences in which they occur. Their function in the syntax can be compared to that of the English adverb mutually in a sentence like ‘They mutually exchanged gifts.’
Like anyoonyam, these forms have a very restricted distribution. They cannot be used to convey the meaning of a reciprocal that is embedded in a larger phrase, as in ‘We like each other’s company’ or ‘They loved each other’s children.’
3.3. Distributive Anaphors
The Dravidian languages have distributive anaphors, which are all formed by the reduplication of pronouns. In Malayalam one can mention four such anaphors: awar-awar, awan-awan, taan-taaŋŋaL, and at-aatǝ.
This anaphor is formed by the reduplication of the pronoun awar ‘they.’ Cf.
The anaphor has a bipartite structure, awar1awar2 , of which awar2 shows the case of the anaphor’s position in the sentence. Awar1 shows no case or the null case of the nominative. That is, unlike the first element of the bipartite structure of oraaL oraaL, awar1 does not show the copied case of the antecedent. Again unlike oraaL oraaL, the two parts of the bipartite structure cannot be separated. We are probably dealing with a noun-noun compound here (and not two independent NPs), which would account for the fact that awar1 cannot take a case suffix of its own. The antecedent of the anaphor must be third person and plural (since awar ‘they’ is plural, third person).
In Kannada, however, this anaphor is not necessarily a noun-noun compound but may be two independent NPs, each with its own case. Cf. (39) (examples from Amritavalli, 2000, p. 91):
It must be assumed—as in the case of oraaL oraaL—that awar1 is not in a θ-position, and so must adjoin to (or otherwise access) the antecedent in order to share its θ-role. This would account for the fact that this form is an anaphor; it would also explain why this anaphor is sensitive to the opacity created by an intervening subject. Cf.:
The only possible interpretation of this sentence is that the policemen would save the policemen’s children, not the children of the mothers. This is because awar1 is unable to access the matrix subject ammamaar owing to the intervening embedded subject poliisukaar. However, this anaphor—unlike oraaL oraaL—does not show the opacity created by an intervening tensed-S boundary; cf.:
This anaphor is not subject-oriented; cf.:
Here the antecedent is the indirect object.
A point to note is that when awar-awar is an argument of the verb, it is required to be disjoint in reference from every other argument of the verb, behaving in this respect like a pronoun. But since it is also required to be bound in the clause, the sentence is unacceptable:
But the sentence can be made fine if the emphasis marker tanne is added to it:
In the other Dravidian languages (Tamil, Telugu, and Kannada), which don’t use the emphasis marker strategy primarily but instead use the verbal reflexive strategy, the presence of the verbal reflexive koL is obligatory in this context; cf. (Amritavalli, 2000, p. 90; Kannada example):
In all these respects, awar awar behaves like a regular pronoun.
3.3.2. awan-awan, taan-taaŋŋaL, and at-aatǝ
The form awan-awan is obtained by reduplicating the pronoun awan ‘he,’ which is third person, singular, and masculine. Therefore its antecedent must be third person and masculine; but of course the antecedent must be plural, this being a semantic requirement of all distributive anaphors. Cf.:
Regarding the form taan-taaŋŋaL, it is unclear what we are reduplicating here. But a good guess is that it is underlyingly:
where -kaL is a plural suffix. Its antecedent must be third person and [+human], since taan has these features. An example would be:
Since atǝ is [−human ], the antecedent of this anaphor must be [−human ] too. An example would be:
Everything that was said about awar-awar applies to these three distributive forms as well, so their properties will not be detailed here.
The Dravidian pronominal system has the usual three persons, two numbers, and three genders. Dravidian has natural gender, which means that all nonhuman referents are neuter and human referents are masculine or feminine according to sex. This being the case, it might be more appropriate to postulate a feature [+/− human], and to distinguish [masculine] and [feminine] as subclasses of [+human]. Gender is marked only in the third person forms, and even in these, in the plural forms the masculine-feminine distinction is neutralized. In the first person plural, there is a distinction between inclusive and exclusive in Tamil, Malayalam, and Telugu (but not in Kannada): the inclusive form is used when the person addressed is included in the ‘we,’ and the exclusive form is used otherwise. Here are some sample pronominal forms, in Tamil, Malayalam, Telugu, and Kannada:
The third person forms need a bit more elaboration. Besides marking the three-way gender distinctions, the third person forms come in proximal-distal pairs in all the Dravidian languages; that is, there is one form used to refer to something nearby (proximal) and another used to refer to something remote (distal). The paradigm will be illustrated only in one language, Malayalam.
As can be seen, the alternants in (52) are predictable: they differ only with respect to an initial vowel—/a/ or /i/.The reason for the choice is transparent: the Dravidian demonstratives are aa ‘that’ and ii ‘this.’ The agreement markers in Malayalam—with cognate forms in the sister languages—are -an for masculine singular, -aL for feminine singular, -tǝ for neuter singular, -ar for masculine/feminine plural, and -wa for neuter plural. Thus it is plausible to argue that awan ‘he’ (distal) is underlyingly aa ‘that’+-an (masculine singular agreement), awaL ‘she’ (distal) is underlyingly aa ‘that’+-aL (feminine singular agreement), atǝ ‘it’ (distal) is underlyingly aa ‘that’+-tǝ (neuter singular agreement), etc. We can derive the proximal pronouns by starting with the proximal demonstrative ii as the base to which agreement markers are added: thus iwan ‘he’ (proximal) is underlyingly ii ‘this’+-an (masculine singular agreement), etc. (This analysis is not uncontroversial; there are some traditional analyses that treat a form like awan as morphologically simple, and derive iwan from ii+awan; see Hariprasad (1998).)
There are honorificity distinctions. As in many languages, the singular ‘you’—in Malayalam and Tamil, nii—is non-honorific; the corresponding plural form is honorific. There are some highly honorific forms, many of which have become archaic. (Formerly, a royal ‘we’—naam in Malayalam—was used by kings and religious heads.)
4.1. Are Proximal Pronouns Really Pronouns?
It needs to be noted that the proximal and distal pronouns do not behave alike syntactically. The distal pronouns show all the familiar properties of pronominal forms in languages. Thus consider (53):
(53a) is ungrammatical with the indicated coreference; this shows that awan obeys Principle B of the binding theory, which says that a pronoun should not be coreferential with a sister argument within the clause. (53b) is fine, because the antecedent of awan is outside the minimal clause containing the pronoun. (53c) is also fine because the antecedent, although it is within the minimal clause, doesn’t c-command the pronoun. (Or to put it differently, John and awan are not co-arguments of the verb.) (53d) is an instance of a pronoun picking up its reference from the discourse, that is, extrasententially. (53e) shows that awan can be bound by a quantifier. (This point needs to be stressed, because there are pronouns in languages that are not amenable to quantifier binding, an example being Japanese kare ‘he.’) Finally, (53f) shows that awan can pick up its reference from the nonlinguistic context, when it is used ostensively (i.e., to point out something). In sum, the distal pronouns of Dravidian bear out the general claim made about pronouns, namely that, except for the negative condition imposed on them by Principle B, they are completely “free”: they can pick up their reference from the extralinguistic context, from the discourse, or from within the sentence. The distal pronouns are standard pronouns.
But when we look at the proximal pronouns, we get a very different picture. Consider the sentences of (54):
At a first glance, it would appear that the only licit use of the proximal pronoun is its ostensive use (to point at somebody or something), as instantiated here by (54f). (It cannot even be called an ‘R-expression’ in the sense of GB, because it apparently cannot be coreferential with a non-c-commanding NP, cf. (54c), or with an NP in the discourse, cf. (54d).) If this is the case, then it is not a pronoun at all but a demonstrative, as indeed its morphology would lead us to expect. (This is the position adopted in Jayaseelan, 2000 and Hariprasad, 1998.)
But a closer consideration shows that the real requirement of the proximal pronoun is that its antecedent be explicitly marked as [+proximal]. Once this requirement is taken care of, the proximal pronoun patterns exactly like the distal pronoun, cf.:
(55a) shows that the proximal pronoun obeys Principle B. And (55b) shows that it does not obey Principle C, and therefore it is not an R-expression. In contexts where neither of these binding principles apply, the pronoun is free to take its antecedent anywhere, as shown by (55c–d). Example (55e) shows that the pronoun can be bound by a quantifier. Thus the proximal pronoun is seen to be a regular, well-behaved pronoun, judging by all the familiar tests. (See Jayaseelan & Hariprasad, 2001.)
5. Dravidian Anaphora and the Binding Principles
The binding theory as originally formulated in Chomsky (1981) divided referring nominal expressions into three classes—anaphors, pronouns, and R expressions—each with its own binding principle. Principle A said that anaphors must have an antecedent in a certain local domain; Principle B said that pronouns should not have an antecedent in the same local domain. Together the two principles predicted that anaphors and pronouns would be in complementary distribution. However, this prediction turned out to be true only about anaphors and pronouns in the argument positions of verbs; the specifier positions of noun phrases seemed to accommodate both anaphors and pronouns indifferently (cf. ‘They love their/each other’s children’). Many proposals were made to fix this problem.
But the more serious challenge for the binding theory was posed by the discovery of long-distance anaphors. These elements had to be classified as anaphors because they required a c-commanding antecedent in the sentence; they contrasted with pronouns, which do not require an antecedent (c-commanding or not) in the sentence. But, problematically, these elements could be bound outside the local domain, in clear violation of Principle A. Long-distance anaphors also showed peculiar restrictions on their choice of antecedent: they were subject-oriented, and they could also be logophorically bound. The binding theory had nothing to say about such restrictions.
But even more embarrassingly for the binding theory, long-distance anaphors seemed to show Principle B effects: they were infelicitous if bound by the subject of the minimal clause containing them. So, were these elements anaphors or pronouns? The distinction seemed to break down. The Principle B effects exhibited by long-distance anaphors were apparently first noticed in the discussion of Dravidian data: Some early accounts of Dravidian taan analyzed this form as a pronoun (Mohanan, 1982; Amritavalli, 1984). This in turn led to the claim that anaphors are a subclass of pronouns (Jayaseelan, 1997). The thrust of research now was to see what makes a pronominal form an anaphor; more specifically, what is the “deficit” that makes a pronominal form require an antecedent in the sentence for its interpretation. The devices that languages use to set aside the Principle B effects—emphasis marker, detransitivizing light-verb, etc.—were also investigated in connection with taan; in Jayaseelan (1997), it was shown that the first-mentioned strategy, namely emphasis marker, was very widespread: it was employed by the Scandinavian languages, Japanese, Korean, Chinese, and English. (There are also many other devices that the languages of the world use—e.g., body-part nouns, other intensifiers, and doubling of pronominal elements—to set aside the Principle B effects; see Reuland, 2016 and references cited there.)
A signal contribution that Dravidian data made to the theory of anaphora was the demonstration that syntactic theory needs to incorporate a rule of anaphorization; especially with regard to reflexive forms, it was shown that any pronoun can be turned into a reflexive (Amritavalli, 1984; Jayaseelan, 1997). This showed that the approach of GB, which marked certain lexical items as ‘anaphor’ or ‘pronoun’ in the lexicon, was inadequate.
It must be noted that Principle A of the binding theory was not invalidated by the new developments in the investigation of anaphora. But its purview was restricted: only morphologically complex anaphors which contained a part that needed to copy the case and/or acquire the θ-role of an antecedent (see section 3) obeyed this principle. Again, Principle B has not been invalidated by later research, but the attempt latterly has been to explain it—and also the complementary distribution of pronouns and anaphors—in terms of deeper principles of UG; see for example Safir (2004), Rooryck and Vanden Wyngaerd (2011), Reuland (2011). As Reuland (2016) has aptly put it, the classical binding theory is still “useful as a descriptive tool”; it is an approximation which is “too bad to be true, but too good to be false.”
For general information about the Dravidian languages—their history, geographical spread, population statistics, etc.—the reader may see Chapter 1 (‘Introduction’) of the following title:
Krishnamurti, Bh. (2003). The Dravidian languages. Cambridge, U.K.: Cambridge University Press.Find this resource:
There are typological descriptions of the four major Dravidian languages—Tamil, Telugu, Kannada, and Malayalam—which the reader who wishes to have an overall picture of any one of them may wish to consult:
Asher, R. E., & Kumari, T. C. (1997). Malayalam. New York: Routledge.Find this resource:
Krishnamurti, Bh., & Gwynn, J. P. L. (1985). A grammar of Modern Telugu. Delhi: Oxford University Press.Find this resource:
Lehmann, T. (1989). A grammar of Modern Tamil. Pondicherry: Pondicherry Institute of Language and Culture.Find this resource:
Sridhar, S. N. (1990). Kannada. New York: Routledge.Find this resource:
A “principled typology” of just the anaphoric systems of the above-mentioned four principal languages can be found in:
Lust, B., Wali, K., Gair, J., & Subbarao, K. V. (Eds.). (2000). Lexical anaphors and pronouns in selected South Asian languages: A principled typology. New York: Mouton de Gruyter.Find this resource:
Amritavalli, R. (1984). Anaphorization in Dravidian. CIEFL Working Papers in Linguistics, 1, 1–31.Find this resource:
Amritavalli, R. (2000). Lexical anaphors and pronouns in Kannada. In B. Lust et al. (Eds.), Lexical anaphors and pronouns in selected South Asian languages: A principled typology (pp. 49–112). New York: Mouton de Gruyter.Find this resource:
Anderson, S. (1983). Types of dependency in anaphors: Icelandic (and other) reflexives. Journal of Linguistic Research, 2, 1–23.Find this resource:
Annamalai, E. (2000). Lexical anaphors and pronouns in Tamil. In B. Lust et al. (Eds.), Lexical anaphors and pronouns in selected South Asian languages: A principled typology (pp. 169–216). New York: Mouton de Gruyter.Find this resource:
Battistella, E. (1989). Chinese reflexivization: A movement to INFL approach. Linguistics, 27, 987–1012.Find this resource:
Belletti, A. (1982). On the anaphoric status of the reciprocal construction in Italian. The Linguistic Review, 2, 101–138.Find this resource:
Burzio, L. (1991). The morphological basis of anaphora. Journal of Linguistics, 27, 81–105.Find this resource:
Chomsky, N. (1981). Lectures on government and binding. Dordrecht: Foris.Find this resource:
Cole, P., & Sung, L. M. (1994). Head movement and long-distance reflexives. Linguistic Inquiry, 25, 355–406.Find this resource:
Fukui, N. (1984). Studies in Japanese anaphora. Unpublished manuscript, MIT.Find this resource:
Gurtu, M. (1985). Anaphoric relations in Hindi and English. (Unpublished doctoral diss.). Central Institute of English and Foreign Languages, Hyderabad.Find this resource:
Hariprasad, M. (1998). Anaphorization in Telugu and English. (Unpublished doctoral diss.). Central Institute of English and Foreign Languages, Hyderabad.Find this resource:
Heim, I., Lasnik, H., & May, R. (1991). Reciprocity and plurality. Linguistic Inquiry, 22, 63–101.Find this resource:
Hestvik, A. (1990). LF movement of pronouns and the computation of binding domains. (Unpublished doctoral diss.). Brandeis University, Waltham, MA.Find this resource:
Huang, C.-T. J., & Tang, C.-C. J. (1989). The local nature of the long-distance reflexive in Chinese. In Proceedings of the North Eastern Linguistics Society 19 (pp. 191–206). Amherst, MA: GLSA.Find this resource:
Huang, C.-T. J., & Liu, C.-S. L. (2001). Logophoricity, attitudes, and ziji at the interface. In P. Cole et al. (Eds.), Long distance reflexives (pp. 141–195). New York: Academic Press.Find this resource:
Huang, Y.-H. (1984). Reflexives in Chinese. Studies in English Literature and Linguistics, 10, 163–188.Find this resource:
Jayaseelan, K. A. (1997). Anaphors as pronouns. Studia lingüística, 51(2), 186–234.Find this resource:
Jayaseelan, K. A. (1998). Blocking effects and the syntax of Malayalam taan. In Rajendra Singh (Ed.), The yearbook of South Asian languages and linguistics 1998 (pp.11–27). New Delhi: SAGE.Find this resource:
Jayaseelan, K. A. (2000). Lexical anaphors and pronouns in Malayalam. In B. Lust et al. (Eds.), Lexical anaphors and pronouns in selected South Asian languages: A principled typology (pp. 113–168). New York: Mouton de Gruyter.Find this resource:
Jayaseelan, K. A., & Hariprasad, M. (2001). Deixis in pronouns and noun phrases. Linguistic Analysis, 31(1–2), 132–149.Find this resource:
Lasnik, H. (1989). On the necessity of binding conditions. In H. Lasnik (Ed.), Essays on Anaphora (pp. 149–167). Dordrecht: Kluwer.Find this resource:
Lasnik, H., & Uriagereka, J. (1988). A course in GB syntax: Lectures on binding and empty categories. Cambridge, MA: MIT Press.Find this resource:
Lidz, J. (2001). The argument structure of verbal reflexives. Natural Language and Linguistic Theory, 19(2), 311–353.Find this resource:
Lidz, J. (2004). Causation and refexivity in Kannada. In V. Dayal & A. Mahajan (Eds.), Clause structure in South Asian languages (pp. 93–130). Dordrecht: Kluwer.Find this resource:
Lust, B., Wali, K., Gair, J., & Subbarao, K. V. (Eds.). (2000). Lexical anaphors and pronouns in selected South Asian languages: A principled typology. New York: Mouton de Gruyter.Find this resource:
Mohanan, K. P. (1982). Grammatical relations and anaphora in Malayalam. MIT Working Papers in Linguistics, 4, 163–190.Find this resource:
Pica, P. (1987). On the nature of the reflexivization cycle. Proceedings of NELS, 17, 483–497.Find this resource:
Reinhart, T., & Reuland, E. (1991). Anaphors and logophors: An argument structure perspective. In J. Koster & E. Reuland (Eds.), Long-distance anaphora (pp. 283–321). Cambridge, U.K.: Cambridge University Press.Find this resource:
Reinhart, T., & Reuland, E. (1993). Reflexivity. Linguistic Inquiry, 24, 657–720.Find this resource:
Reuland, E. (2011). Anaphora and language design. Cambridge, MA: MIT Press.Find this resource:
Reuland, E. (2016). Grammar of binding in the languages of the world: Unity versus diversity. Cognition. doi:10.1016/j.cognition.2016.01.020Find this resource:
Rooryck, J., & Vanden Wyngaerd, G. (2011). Dissolving binding theory. Oxford: Oxford University Press.Find this resource:
Safir, K. (2004). The syntax of anaphora. Oxford: Oxford University Press.Find this resource:
Shibatani, M. (1999). Dative subject constructions twenty-two years later. Studies in the Linguistic Sciences, 29(2), 45–76.Find this resource:
Subbarao, K. V., & Lalitha-Murthy, B. (2000). Lexical anaphors and pronouns in Telugu. In B. Lust et al. (Eds.), Lexical anaphors and pronouns in selected South Asian languages: A principled typology (pp. 217–273). New York: Mouton de Gruyter.Find this resource:
Sundaresan, S. (2012). Context and (co)reference in the syntax and its interfaces. (Unpublished doctoral diss.). University of Tromsø.Find this resource:
Ueda, M. (1984). On a Japanese reflexive zibun: A non-parametrization approach. Unpublished manuscript, University of Massachusetts, Amherst.Find this resource:
Zribi-Hertz, A. (1989). Anaphor binding and narrative point of view: English reflexive pronouns in sentence and discourse. Language, 65, 695–727.Find this resource:
(1.) A note on the transcription: /t, d, n/ are dental; /t, n/ are alveolar; /T, D, N, L, S/ are retroflex; /s’/ is palato-alveolar; /k’/ is palatalized; and /R/ is an alveolar tap. /w/ is a labiodental approximant without lip-rounding; it is shown as /v/ in certain transcriptions.