Pronominal Vs anaphoric pro in Kannada

Kannada licenses a pronominal pro and an anaphor pro in root and subordinate clauses. As for the pronominal pro in subordinate clauses, it occurs only in argument clauses. Here, pro’s person feature largely determines its pronominal/anaphoric interpretations. Accordingly, there are three types of pronominal pro each with a distinct referential property. The 1 st and 2 nd person pros allow pronominal and anaphoric interpretations. In their pronominal reference, they refer to the speaker and the listener, respectively. In their anaphoric reference, they refer to the matrix subject and the object, respectively, irrespective of the latter’s person feature which results in a feature mismatch between pro and its antecedent. The 3 rd person pro allows only pronominal interpretation. These properties of pro are a direct consequence of the fact that Kannada quotative verbs report direct speech and the null subjects in the embedded clauses of reported speech are basically pronominal because they are “copies” of the pronominal subjects in direct speech. Accordingly, the embedded verbal inflection corresponds to inflection in direct speech which results in a feature clash between pro and its antecedent. The null anaphor, however, occurs only in non-argument (adjunct) endu-clauses. It is bound by an NP in the matrix with which it shares its ɸ-features. It is the semantic relation between the non-argument and main clauses that binds anaphoric pro and its antecedent.


Introduction
In most studies on the pro-drop parameter, pro was characterized as a pronominal empty category based on evidence from European null subject languages.To my knowledge, pro as an anaphor was first proposed for Kannada in Sudharsan (1998Sudharsan ( , 2001) ) and subsequently for Marathi in Holmberg (2005).The pro phenomenon in Kannada is a very complicated one, with hardly any work done on it.This paper attempts to fill this gap by offering an account of pro in Kannada and show that it functions as a pronominal and as an anaphor as well, on par with PRO, the pronominal anaphor.I will show that pro, the null pronoun, and pro, the null anaphor, occur in mutually exclusive contexts.
The null subject (NS) phenomenon was first observed in Romance languages (with the exception of French) such as Italian and Spanish in which the pro subject in a finite clause, be it root or embedded, has a pronominal interpretation.Hence the characterization of pro as a null analogue of an overt pronoun.Here are examples from Italian, taken from D'Alessandro ( 2015 ∅ mató killed al the perro dog 'He said that he/she killed the dog.' In the above examples, the referential features of pro can be recovered from the verbal inflection.In example (1), pro is 2nd person plural, in (2) it is 2nd person singular, in (3) it is 3rd person plural, in (4) it refers to a 1st person plural pronoun, and in (5) the embedded subject pro is 3rd person plural.An overt pronoun in place of pro would also be perfectly grammatical in the above examples.English, on the other hand, does not allow a pro subject in the corresponding positions.Consider the following Italian example from Holmberg & Roberts (2010, p. 7): arrived.'The professor spoke after he arrived.' Holmberg and Roberts observe that the overt pronoun of the adjunct clause in (7) does not exhibit the ambiguity that its English counterpart does in its reference.In other words, the overt pronoun is disjoint from 'professor' whereas the pronoun in the English sentence ambiguously refers either to the professor or to someone else.They further observe that the difference between Italian and English in the interpretation of the overt pronoun is related to the fact that subject pronouns may be left unexpressed in languages such as Italian, that is, to the positive value of the Null Subject Parameter (NSP).So, they observe that the null subject in Italian is an analogue of the overt pronoun.Hence, the general view that pro is a pure pronominal empty category.I will show that in Kannada pro can be a pronominal and an anaphor as well both of which exhibit distinct properties and occur in different environments.
We noted that the referential content of pro in Italian examples discussed above can be determined by the verbal inflection in the sentence.But in null subject languages (NSLs) like Japanese, Chinese, and Malayalam, the referential features of null subjects cannot be determined from the sentences in which the null subjects occur since these languages lack agreement inflection.In these languages the omitted pronoun can be recovered from the discourse only.
Besides, not all NSLs drop pronouns in all sentences.And some languages like German lack referential null subjects but they have only empty expletive subjects, corresponding to the overt 'it' or 'there' in English.So linguists have identified four types of NSLs on the basis of what kind of NSs they have and on the basis of how the referential features of null subjects can be determined.They are Consistent/Canonical null subject languages (CNSLs), Radical pro-drop languages, Partial NSLs, and Expletive NSLs.
In CNSLs, referential pronominal subjects of all persons can remain unexpressed in all tenses and they can be recovered from agreement inflection on the verb.So these are sentential null subjects.So consistent NSLs are consistently 'rich' in agreement inflection.For instance, languages such as Italian, Greek, Turkish, and Basque are 'rich agreement' languages.
In Radical or Discourse-oriented pro-drop languages, both subject and object NP can be null.These Radical pro-drop languages lack agreement inflection and the null pronouns are recovered from the discourse, Kansas Working Papers in Linguistics | Vol.38 (2017) | pp.23-42 not from the verbal inflection.For example, Chinese, Japanese, Korean, Thai, and Malayalam are devoid of agreement inflection and are all Radical pro-drop languages.
In Partial NSLs, some referential pronouns can be dropped and some cannot be dropped.In Finnish, which is a partial NSL, only 1st and 2nd person pronouns can be dropped, but not the 3rd person pronoun.A 3rd person pronoun can be null in embedded clauses only if it is bound by a NP in the higher clause (Holmberg, 2005, p. 10).Russian is another Partial NSL and it has a null indefinite or generic subject.
Finally, in Expletive NSLs, only non-referential subjects can be null.German is an expletive NSL.In languages like English and French, null subjects are not allowed in finite clauses.
Earlier it was believed that rich agreement was mandatory for dropping pronouns.We have seen that languages like Japanese and Chinese lack agreement inflection but nonetheless they allow null subjects in finite clauses.Then what licenses null subjects?A brief discussion of some of the NS accounts would be in order here.
One of the earliest accounts of null subjects was by Borer (1986), in which she proposed what she called the I-subject hypothesis, which is as follows.The I-subject hypothesis states that there is no need to assume a null subject in the canonical subject position, i.e., Spec,IP, which means that the Spec,IP position does not get realized in NSLs.This is because Infl in NSLs has pronominal features and is itself assigned the subject theta-role.We are aware that there are NSLs such as Chinese and Japanese which lack agreement morphology and there are also NSLs which allow only non-referential (expletive) NSs.So Borer's I-subject hypothesis fails to accommodate these languages.Jaeggli & Safir (1989) took an integrative approach to the NS phenomenon and proposed the Morphological Uniformity Hypothesis, which accommodates not only consistent NSLs like Italian and Greek but also Radical pro-drop languages like Chinese and Japanese.According to this hypothesis, NSs are allowed only in languages which possess a morphologically uniform inflectional paradigm.This means that languages must have rich agreement or no agreement at all in order to license NSs.In CNSLs like Italian, the verbal paradigm is uniformly rich and in Radical pro-drop languages, agreement inflection is uniformly absent.What this implies is that languages like English which lack a uniform inflectional paradigm cannot license NSs.This hypothesis, however, does not accommodate Partial NSLs like Finnish since these languages do have agreement inflection, though such agreement is not so rich.
So taking into account Partial NSLs like Finnish, Roberts (2010) makes a distinction between rich agreement and impoverished agreement which is crucial to the NS parameter.He observes that if agreement inflection is impoverished, then T lacks a D-feature, hence it does not allow deletion of the subject pronoun.Roberts' "impoverished" agreement hypothesis accommodates null subjects in partial NSLs.Roberts (2010) proposes a deletion analysis of pro in consistent null subject languages within the framework of the Minimalist theory.Both Roberts (2010) and Holmberg (2005) depart from Borer's (1986) Isubject hypothesis and assume that there is an empty pronominal subject at Spec,IP position in null subject languages.Roberts (2010) assumes that in consistent NSLs like Italian, T is pronominal and that there is a D-feature in T which correlates with 'rich' agreement.Roberts (2010, p. 75, fn. 18) treats the D-feature on T as unvalued and that of pro as valued.He derives pro's silent nature in the following way.The postulation of D-feature on T means pro counts as a defective Goal and its features are "properly included in T." Roberts defines a defective Goal this way (2010, p. 70): (8) A goal G is defective iff G's formal features are a proper subset of those of G's Probe P.
Besides, T has an EPP-feature which pro satisfies by raising to Spec, IP.Roberts observes that defective goals never have PF realization.Central to Roberts' account is the view that pro is just like an overt pronoun, and its non-overtness is "purely a PF matter" (Roberts, 2010, p. 80).
As for 'radical' pros, following Saito (2007), Roberts assumes that empty arguments are copied from the discourse at LF.
In partial null subject languages such as Finnish, finite T does not have an uD-feature.T probes for Dfeature.The subject's F-features are copied by T and the subject has its case feature valued in return.Let's now turn our attention to Kannada.

Kannada, an ambivalent language
What type of language is Kannada?Kannada allows null subjects of all persons in all tenses.It licenses a referential null subject of all persons in all tenses; it has arbitrary or generic null subjects, and it has an expletive null subject as well.Above all, it licenses a null anaphor in the object position in reflexive sentences and a null anaphoric subject in finite subordinate clauses.Therefore, in Kannada, there are parallels between pro and PRO, the pronominal anaphor.However, the pronominal pro and the anaphor pro occur in different syntactic environments.
Kannada is 'rich' in agreement inflection.However, negative and modal verbs are invariable, hence they are devoid of agreement inflection.Only non-negative and non-modal indicatives, interrogatives, and imperatives show agreement so much so that agreement can be looked upon as a feature of positivity or affirmation.And subordinate clauses such as adverbial, gerund, and relative clauses also lack agreement since they are not declarative/interrogative in force.But these clauses are tensed and license a nominative subject also and can also have a null subject.As we know, null subjects of all persons can occur in almost all sentences and in all tenses.Besides, Kannada allows almost all types of null subjects, although rich agreement is not present uniformly in all sentences.Therefore, Kannada cannot be called either a 'consistent' null subject language or a 'radical' pro-drop language.It is an inconsistent or an ambivalent language as it possesses properties of both 'consistent' null subject languages and 'radical' pro-drop languages.It is in fact more of a radical type than a consistent type.
It is quite possible that Kannada never belonged exclusively to any one type at any point of time in its history.This is because the lack of agreement inflection in some subordinate clauses is a characteristic of the Dravidian languages in general.But the lack of agreement inflection in negative and modal clauses is partly due to the historical changes that took place in Kannada centuries ago.In Old Kannada (OK), even standard negation (SN) and certain kinds of modal verbs did inflect for agreement.SN in OK was formed on the negative suffix -a, which was attached to the verb and the agreement inflection was assimilated into the negative suffix.And the existential negation was formed on the invariable negative existential verb illa, 'to be not'.Both of these negatives were restricted to present tense.During the Jaina period when Kannada and other Dravidian languages came under the influence of Sanskrit, there were significant changes in the syntax of Dravidian languages.I have argued in Sudharsan (2012) that one of the syntactic changes that took place in Kannada was the reanalysis of the invariant negative existential verb illa into a verbal negator, on the lines of the Sanskrit SN in which the negative particle na was used both in SN and existential negation in all tenses.Since in OK -prior to the influence of Sanskrit -the SN could not be used in past tense, the use of the existential negative illa was extended to verbal negation in past tense.Gradually, the SN formed on illa replaced the earlier SN in which the verb inflected for agreement.Kannada thus developed a full paradigm of negation based solely on the invariable negative verb illa.
In contemporary Kannada the modal negative aar, 'not to be able to', which does inflect for agreement, is also fast disappearing and is being replaced by the negative formed on illa and the linking verb aagu, 'become'.
The case of Kannada suggests that not all NSLs fit neatly into one of the four types of NSLs proposed in Biberauer et al. (2010).Languages can be of mixed type, and Kannada gives evidence for this.

Pro in Kannada.
Pro has a wide distribution in Kannada and it occurs in root and in embedded clauses.It can be definite or arbitrary in its reference.An object pronoun can also be dropped and can be recovered from the discourse, since the verb does not show agreement with the object noun.
Before we consider pro in detail, an overview of verbal morphology of Kannada would be in order.Kannada shows rich agreement inflection in positive indicative, interrogative, and imperative sentences.The verb shows agreement in person, number, and gender with 3rd person singular subjects.And with 1st and 2nd person singular subjects it shows only person and number features.There is a two-way distinction between human and non-human with 3rd person plural nouns.Kannada has two morphological tenses: past and non-past.There are three aspects: simple, progressive, and perfect.In the positive indicative clauses, the auxiliary verb iru, 'to be' is used only in non-simple forms.It is tense which renders a clause finite and it is tense which licenses a null subject in Kannada.And agreement is not present consistently in all sentences.The negatives are periphrastic in simple and non-simple tenses, since negation is indicated by the invariant negative verb illa, which is on a par with the positive iru, 'be' except that the latter inflects for tense and agreement.In verbal negation, the main verb is in non-finite form and the negator illa is invariably attached to the main verb except when there is emphasis on the main verb, in which case illa is detached from the verb.In the following sections we will take an overview of different types of pro in Kannada.We will first look at pro in root clauses.The 1st and 2nd person pronouns are invariably dropped and they refer to the speaker and the listener, respectively.In the above dialogue, a 3rd person pro is licit in speaker B's utterance since it has an "antecedent" nimm magaLu, 'your daughter' in the dialogue.Similarly, the object pronoun can also be left unexpressed if it is recoverable from the discourse.However, in reflexive sentences where the reflexive verb koL is present, the object pronoun can be optionally null and is bound locally by the subject pronoun, null or otherwise, and hence can be recovered from the sentence itself.So it is the reflexive koL which licenses a null anaphor, i.e. pro in the object position.
Reflexivity in Kannada is generally expressed by the reflexive light verb koL, which can be used with 1st, 2nd, and 3rd person subjects.It is koL which inflects for tense and agreement and the main verb will be in the non-finite form.Kannada has a logophoric reflexive pronoun taan, 'self' which is strictly 3rd person subject-oriented; it lacks gender features; it is marked for case.Because of taan's 3rd person subjectorientation, there is a gap in the paradigm of lexical anaphors in Kannada.But the null subject option seems to have filled this gap by allowing for pro to have 1st and 2nd person antecedents.So, the presence of the reflexive koL invariably renders taan superfluous since the subject and the object positions can be null.Examples (11a-c) are reflexive sentences.In (11a), the choice between the reflexive taan and pro in the object position indicates that taan is optional.When there is emphasis, however, the presence of taan is mandatory, as in (11c).The ungrammatical (11b) shows that taan cannot refer to 2nd (or even 1st) person subjects.When the object NP is disjoint from the subject NP, then koL cannot occur as shown in (12).As stated earlier, the reflexive koL can occur with 1st and 2nd pronominal subjects as well, as in (13a-b).It is perfectly grammatical to use an overt 3rd person pronoun in place of taan or pro with 3rd person subject pronouns, as it is shown in ( 14) in which case the overt 3rd person object pronoun is bound by the 3rd person subject.( 14 We see that in root clauses the object pro can be bound by a subject of any person and that the verb agrees with the subject, as shown in the examples above.On the assumptions of the null subject theory proposed in Rizzi (1980), and in Jaeggli & Safir (1989), a null subject has to be identified.Generally speaking, Agr is supposed to identify the referential content of pro.So in root clauses (with non-negative and non-modal verbs), it is koL which renders the object pro reflexive, which means that Agr(eement) on koL identifies both the object and the subject pros.
In the foregoing discussion, we looked at different types of pro licensed by rich agreement on the finite verb.However, pro can appear only when there is a person feature on the verb.In Kannada, non-verbal predicates like nominals are devoid of person features, hence they cannot have a pro subject.As for adjectives, there are pure adjectives and there are derived adjectives.Most of the adjectives are derived adjectives, derived by attaching the adjectival suffix, -ge to the adjective base.And some of the pure adjectives can also be turned into derived ones by attaching the suffix -ge.And the derived adjectives have to occur with the copula iru, 'to be' which inflects for tense and agreement.These derived adjectives can have a pro subject, since they have a person feature.Examples ( 15) and ( 17) have a null subject since the derived adjectives occur with the copula iru which shows a person feature.In ( 16), the pure adjective jaaNa 'intelligent' has number and gender features but lacks a person feature, so it cannot have a null subject.Thus the person feature is mandatory for licensing pro.This suggests that it is the person feature which renders agreement 'rich' enough to allow a null subject.There can be a null subject even in the absence of agreement, but when agreement is present, a person feature is mandatory for the occurrence of a null subject.We will see shortly that in embedded clauses, the verb shows person and number features but lacks a gender feature, but nonetheless, it can have a null subject.

Arbitrary pro.
Kannada allows a pro subject with arbitrary reference in indicative and imperative sentences.Take a look at the sentences below with a generic pro subject: Sentence ( 18) is an indicative sentence and has a 3rd person plural Arb and ( 19) is an imperative with a 2nd person singular Arb.The nominalised gerund -udu clause in Kannada is devoid of agreement.It is marked for tense and aspect and it also licenses a nominative subject.When the nominalised verb does not inflect for a specific tense and aspect, it lacks a specific time reference.It indicates timeless/generic present by default.In that case, it allows an arbitrary pro.
(  ' In (20), pro has an arbitrary reference.The verb is not overtly marked for a specific tense, in which case it receives a timeless present reading by default.Sentence ( 21) is odd and unacceptable because pro is not bound; it cannot have an arbitrary reference either, because the verb is in present progressive and so has a specific time reference.In ( 22) again, pro in the subject position is unacceptable since there is no binder in the matrix; nor can it be used in an arbitrary sense, since the clause refers to a specific situation.Only the overt pronoun niinu 'you' would be acceptable in the subject position.In indicative clauses, on the other hand, an arbitrary pro is allowed even with specific time reference, as seen in sentence (18) above.23) is a passive expletive construction which is no longer used in Contemporary Kannada.This construction type is similar to the English it-construction such as "It has been said that Jesus rose from the dead."The construction in ( 24) is based on the raising verb annisu 'to feel'.Sentence (25) has been analyzed as an impersonal passive in Sridhar (1979).This sentence is formed on a complex predicate composed of V-al(inf)+aagu(become)+iru(to be), but it can also be analyzed as a biclausal structure, containing an infinitival -al clause with an arbitrary PRO as its subject.The finite verb in each of them carries the default 3rd singular neuter inflection and the subject position is occupied by an empty expletive corresponding to the English expletive it.In each of these examples, the main verb has a caseless endu-clause complement which cannot occur in a case-marked position.Since the subject position is non-thematic and caseless, I assume that there is no empty expletive in the canonical subject position; but the endu-clause is coindexed with the subject position.Since the clause cannot occupy a case position, it does not raise to the canonical subject position but remains in its complement position until LF.It raises to the subject position only at LF.So I assume that in these expletive constructions the Spec, IP position comes into existence only at LF.In speaker B's utterance, the referential features of the pro subject can be recovered only from the discourse, that is from Speaker A's utterance, since the negative verb illa lacks agreement.

The radical pro. All along
We have noted that Kannada has almost a full paradigm of null subjects in spite of the fact that rich agreement is not found in all clauses.There are only two possibilities with regard to agreement.It is either rich or totally absent.

The anaphoric pro.
Pro has been generally regarded as a purely pronominal empty category, unlike the big PRO which has been characterized as a pronominal anaphor.An anaphoric pro was first proposed for Kannada in Sudharsan (1998) and subsequently one was proposed for Marathi in Holmberg (2005).We noted that in reflexive sentences, a null anaphor, i.e. an anaphoric pro occurs in the object position and is licensed by the reflexive verb koL.This null object pro is locally bound by the subject.Similarly, Kannada allows an anaphoric pro, i.e. a null anaphor in the subject position of embedded clauses.Whereas the pronominal null subject occurs in argument clauses, the null anaphor occurs only in the adjunct endu-clause, and also in the appositive endu-clause.So, the pronominal pro and the anaphoric pro occur in mutually exclusive environments.We will consider these two types of pro in detail in section 4.

Spellout analysis of pro licensed by rich agreement
In Kannada the null subject is the basic unmarked option and the overt pronominal subject appears only when there is emphasis or some other special feature on pro.So when pro has emphasis or some special feature, it gets phonetic realization.So I assume that pro is the basic underlying option and the presence of special feature on pro triggers an overt spellout of pro.In this way my analysis departs from Roberts' (2010) deletion analysis, although I do adopt some of Roberts' ideas in developing an account of pro in Kannada. 1ollowing Roberts (2010), I assume that in the case of pro licensed by rich agreement on the verb, finite T is pronominal as in the case of consistent null subject languages.T has an unvalued D-feature (F-features) and also an EPP feature.I treat pro as any argument, except that it is not pronounced.And the F-features on pro are interpretable (intrinsically valued) since F-features are intrinsic properties of arguments (Chomsky, 1995).However, null subjects have to be identified by F-features on T at LF which means that F-features on T should be available at LF.So I assume that F-features on T are interpretable in Kannada.Accordingly, valuation means copying of valued features of pro on to T; it does not mean deletion of features after checking is done.On Minimalist assumptions, T probes for a goal, and an Agree relation between T and pro are established and pro's interpretable F-features are copied on to T and hence T's features are not deleted; they are available for further computation.Besides, pro satisfies the EPP feature by raising to Spec, IP.
I have shown in Sudharsan (2012) that case alternation on arguments in Kannada results in differences in meaning even though the thematic structure remains unaltered.So I assume that case feature is also interpretable.This way F-features on T and case feature on pro are available at all levels of derivation.
As for the 'radical' pro in Kannada which occurs in negatives and clauses with modals and in subordinate clauses which are devoid of agreement, I propose that it is inserted at Spec, IP straight from the discourse at LF.It is at LF that the ambiguous reference of pro is disambiguated.Similarly, null objectswith the exception of null objects in reflexive sentences -are also recovered from the discourse context.This is similar to Saito's (2007) view about null subjects and ellipsis.
Earlier we noted that Kannada has a pro subject with an arbitrary reference.It can be 3rd person plural, or 2nd singular.An overt 1st person plural pronoun naavu, 'we' can also be used with arbitrary reference.These three types of pro with arbitrary reference are discussed in Sudharsan (1998).Pro is specified for F-features as in the case of other types of pro.It is the lack of uD in T that renders pro arbitrary.Another important difference between arbitrary pros and the definite pros is that the former are obligatorily null and the latter are optionally null.The arbitrary pro can be derived in almost the same manner as the definite pro.
We saw in section 2.3 that Kannada licenses an expletive pro as well.The empty expletive construction occurs with a default 3rd person singular neuter inflection on the verb.And generally the unaccusative/passive verb takes an endu-clause as its complement unlike some of the unaccusatives in English which can take an infinitive as their internal argument.I argue that since the expletive pro lacks case and a T-role, there is no need to assume a non-thematic pro subject at Spec,IP.The endu-clause is not a case-marked clause and hence it cannot occupy a case position; so it stays in situ in the complement position until LF.It is co-indexed with the subject position and so it raises to Spec IP at LF.This way, the Spec, IP position in the expletive constructions comes into existence at LF.
In the foregoing sections we have considered briefly different types of pro in Kannada and how they can be derived within the framework of Minimalist theory.Due to space constraints, it is not possible to provide a detailed account of the different types of pro in Kannada.
The main objective of this paper is to examine the (definite) pronominal pro and the anaphoric pro licensed by rich agreement and to account for their distinct properties within the framework of Minimalist theory.The paper will address the following issues: Firstly, it will argue against the general view that pro is a pure pronominal empty category and show that pro in Kannada can be an anaphor also on a par with the big PRO, the pronominal anaphor.The pronominal null subject and the anaphoric null subject occur in root and in subordinate clauses as well.The case of the null subject in subordinate clauses is not as straightforward as in root clauses.These two types of null subject differ in their distribution in that the pronominal category occurs in argument clauses and the anaphoric one occurs in non-argument clauses.Secondly pro's F-features, particularly its person feature, and its referential orientation are closely related.In other words, pro's person feature indicates whether it is pronominal or anaphor or both in its reference.And agreement on the embedded verb plays a crucial role here.It not only identifies pro's referential features but also indicates its pronominal/anaphoric orientation.There has not been any study to date that addresses these issues, hence in the rest of the paper I would like to focus on these aspects of pro.

Pro and the role of rich agreement in embedded clauses
In section 2.1, we considered pro in root clauses and we noted that Kannada allows 1st, 2nd, and 3rd person pros in all tenses.More importantly, we noted that in reflexive sentences the object pronoun can also be null.And this null object is locally bound by the subject (pronoun) which can also be null.The verb in all these cases shows agreement with the subject.Similarly, in embedded clauses pro can also be pronominal or anaphoric depending where it occurs.And agreement plays an important role in relation to these aspects of pro.So we will now consider pro and the role of agreement in embedded clauses.
Pro occurs in a variety of embedded clauses and, in fact, it is the natural choice for the subject position in most of the embedded clauses.It invariably replaces the reflexive pronoun taan also unless there is a need to avoid ambiguity in its reference.
There are three types of CP clauses in Kannada: the quotative endu-clause which cannot be nominalised; the embudu-clause is a nominalised CP clause; and the Complex NP construction (CNPC), i.e. the emba+NP clause.Besides, there are finite IP clauses which also license a pro in the subject position.In Kannada some clauses are case-marked and some are not.The endu-clause is not case-marked.The nominalised embudu-clause, the CNPC, and the gerund udu-clause are case-marked.
Pro can be the subject of any of the clauses mentioned above.Pro can be subject-oriented, or objectoriented, or even speaker or listener-oriented.These different referential orientations are invariably determined by pro's person features.On the basis of these properties of pro, I propose four types of (definite) pro licensed by rich agreement on the embedded verb.Each type of these null subjects is distinguished by its person feature which determines its pronominal or anaphoric status.Secondly and importantly, whether the embedded clause in which the pro occurs is an argument or a non-argument also decides the pronominal or anaphoric status of the pro.The four types of pro are as follows: the 1st person pro, the 2nd person pro, the 3rd person pro, and fourthly, pro, the anaphor.
The 1st and 2nd person pros are null analogues of the corresponding overt pronouns in that they have both pronominal and anaphoric interpretations.When they have pronominal reference, they refer to the speaker and the listener, respectively.When they are anaphoric in their reference they refer to the matrix subject and the object, respectively, irrespective of the person feature of their antecedent.This way, they fail to share their F-features with their antecedent.
The 3rd person pro, however, allows only pronominal interpretation just like its overt counterpart which also allows only pronominal interpretation.
The differences in pro's and its antecedent's referential features which results in a person feature between pro and its antecedent, particularly in the case of 1st and 2nd person pros, are due to the fact that quotative verbs in Kannada generally report direct speech.The person feature of the embedded subject is the same as that in direct speech and accordingly the agreement pattern on the embedded verb corresponds to 1st, 2nd, and 3rd person inflection in the direct speech situation.
The null anaphor, by contrast, occurs in the non-argument (adjunct) endu-clause, and the appositive endu-clause.Secondly, it does not exhibit any feature clash like the other pros.Pro here is bound non-locally by an NP in the main clause.I argue that it is the semantic relation that holds between the subordinate and the matrix clauses that renders pro an anaphor.

The 1st person pro.
As stated earlier, this pro occurs both in the caseless endu-clause, and the casemarked embuduand emba+NP clauses.
We will first take up pro in the endu-clause.The endu-clause has a wide distribution and has multiple functions.It is generally used in formal written Kannada and its colloquial counterpart is the anta-clause.The complementizers endu and anta are adverbial participials derived from the verbs en and an, respectively, and both mean 'say'.The verb an is the modern form of the earlier en.In this paper, I will use both versions interchangeably.However, wherever there is a significant difference between the written and the colloquial versions, it will be pointed out.This clause occurs with several types of verbs, such as verbs of saying and thinking, performative verbs, stative verbs, etc., and it has different functions.Affirmatives, negatives, questions, and imperatives can all be embedded within the endu-clause.The clause has generally a quotative function when it occurs as an argument of verbs of saying, thinking, etc.Secondly, it can occur as an adjunct of the main verb and function as an adverbial clause of reason.Thirdly, it can also occur in apposition to the main verb, and function as an explanatory equivalent to the verb.Generally, the clause occurs in the preverbal position either before the matrix clause or embedded internally within the main clause.However, the colloquial anta-clause can occur in the post-verbal position in marked word orders in colloquial Kannada.The written version, that is the endu-clause, generally sticks to the preverbal position in formal written Kannada.
In the following discussion, I will examine 1st person pro as the subject of the quotative anta-clause and show that it ambiguously refers to the matrix subject and the speaker.
As we noted earlier, the function of the endu-clause is generally determined by the type of verb it occurs with.Firstly, as an argument of transitive verbs of saying and thinking such as heeLu 'say', keeLu 'ask', nambu 'believe', etc., its main function is to report direct speech.Kannada differs significantly from English in the behavior of reported speech. 2 In English, reported speech exhibits significant changes in the person feature of pronouns and tense feature of the embedded verb.While the tense feature of the embedded verb does not always match with that of the matrix verb, the person feature of the embedded (anaphoric) pronouns almost always changes to match with that of their antecedents in the matrix.It is in this respect that Kannada exhibits a marked difference from English.In Kannada, as we noted earlier, the embedded subjects in reported speech are just "copies" of the pronominal subjects in direct speech situation.They do not differ in their person feature from the subject pronouns in direct speech.And as a result, the agreement inflection on the embedded verb also corresponds to 1st, 2nd, and 3rd person inflection on the verb in direct speech.This results in a mismatch in person feature between pro and its antecedent, particularly in the case of 1st and 2nd person null subjects when they are anaphoric in their reference.Consider  'Jalaja said to me that she herself will write a letter to Dinesha.' In these examples, pro is 1st person as indicated by the 1st person inflection on the embedded verb.Further, pro is coreferential with the matrix subject irrespective of the subject's person feature.Pro can also refer to the speaker.Only in (27, 30) can taan occur in the embedded subject position, although pro is much more natural than taan.When there is emphasis on the subject, taan occurs obligatorily, as in (30).
We also see that an overt 1st person pronoun in the embedded subject position is perfectly grammatical, since the verb has 1st person agreement.Consider Bhatt's (1978)  Bhatt observes that the overt 1st person subject pronoun in the embedded clause is ambiguous in that it refers either to Hari or to the speaker.However, for some speakers, naanu 'I' refers only to the speaker.There were hardly any speakers who said that naanu referred only to Hari.So in either case naanu is not bound in the strict sense of the term since it refers to the subject and to the speaker as well.It allows both pronominal and anaphoric interpretations.Consider again the following sentence from Bhatt (1978) Here again, there are speaker variations regarding the interpretation of pro.Many speakers said that it refers to the speaker and several others said that it refers either to Hari or to the speaker.The consensus that emerged from the informants' judgements is that both pro and the overt 1st person pronoun are ambiguous in their reference.Another important point we need to note is that in this case, the embedded verb has to have 1st person inflection.When the matrix subject is 3rd person, taan can replace pro in the embedded subject position, but the embedded verb has to have 1st person inflection.
In its anaphoric reference, pro in the above sentences does not share its person feature with its antecedent, that is, the subject of the main clause.It exhibits feature clash which goes against Binding Condition A and which disqualifies it to be an anaphor.Secondly, it does not meet the C-command requirement since the subordinate clause generally precedes the main clause.Besides, it has a pronominal interpretation also in that it refers to the speaker.Finally, it can be replaced by an overt 1st person pronoun which suggests that pro here is basically a pronoun.
The 1st person pro can be the subject of an embedded interrogative as well, as in the sentence below: Dinesananna j Dinesh-ACC keeLidlu asked.SF 'Mary asked Dinesh if she (or I) looked good in a red colour saree.' Here pro can refer either to Mary or to the speaker (i.e naanu, 'I').
According to Rizzi's (1980) and Jaeggli and Safir's (1989) theories of null subjects, pro must be identified by the Agr(eement) in the embedded clause, which means that pro in this case is 1st person.We see that pro's antecedent, that is the matrix subject, can be 2nd and 3rd person also, as in (28,30), respectively.In other words, the 1st person pro or even a 1st person overt pronoun is always subject/speaker oriented.So we assume that agreement inflection on the embedded verb in these sentences identifies not only pro's referential content, it also indicates its anaphoric orientation -that is, its subject/speaker orientation.So pro has the features +1st person, +Subj/Sp.According to the theory of null subjects proposed in section 3, a definite pro is specified for F-features and T has uD-feature, and also has an EPP feature which requires pro to raise to Spec, IP.T probes for a goal, and the goal pro forms an Agree relation with T and its F-features are copied on to T. These features are also available at LF, since in Kannada F-features on T are interpretable.
Pro exhibits similar properties in the embudu and emba+NP clauses.In the embudu and emba+NP clauses, the embedded verb shows full agreement when it is affirmative.[35][36] pro is subject-oriented, as the 1st person inflection on the verb indicates.At the same time, pro can refer to the speaker as well.In fact, the speaker orientation is stronger than the subject orientation of pro in (35)(36).This may be because the embudu clause is a DP-clause and the emba+NP clause is a complex NP construction, both of which probably act as islands for coreference.This suggests that pro is probably sensitive to island constraints.However, further investigation is required before we can say anything conclusive about pro in these clauses.
Taking into account these various interpretations of pro, we can conclude that the subject-oriented pro functions like a pronoun and hence it is not bound in the true sense of the term.

The 2nd person pro.
A 2nd person pro can occur in the subject position of an embedded declarative, interrogative or an imperative as well.Just like the 1st person pro, this pro also ambiguously refers either to the object NP or to the listener.The embedded verb has 2nd person inflection.Here is an example in which 2nd person pro occurs in the indicative endu-clause.Here, pro is coreferential with the object NP irrespective of the person feature of the object NP. said.3SM'Dinesh said to Jaanhavi/me/you that (pro: you) look good in a red colour saree.' We notice that the embedded subject can refer to the dative NPs Jaanhavige or nanage, 'me' irrespective of the person feature of the object NP.Besides, it can also refer to the listener represented by ninage 'you', mentioned within the brackets.And the object NP can be dropped and recovered from the context.But pro's referential features will be 2nd person singular.
Pro can be the subject of an embedded interrogative or imperative endu-clause which occurs as an argument of verbs like keeLu 'ask' or heeLu 'say'.heeLidaru said.3PL'My mother told me to come back before it gets dark.' In (38)(39)(40), pro is co-referential with the object of the matrix verb, that is the person addressed by the subject.Hence, pro stands for you (in direct speech situations) and accordingly the embedded verb has 2nd person inflection.In ( 38), an overt 2nd person pronoun niinu 'you', in place of pro would also ambiguously refer either to the object or to the listener.In (39) pro unambiguously refers to the direct object nannannu, 'me'.If the object is not expressed then pro implicitly refers to the listener.In the case of (40) the verb is in the imperative form and hence pro is 2nd person singular.Since the object pronoun is not expressed, pro implicitly refers to the listener, unless it is decided otherwise from the discourse context.This way, the 2nd person pro shares several characteristics with the 1st person pro.Both can have anaphoric and pronominal interpretations.In their anaphoric reference, they exhibit a feature clash in that they do not share their person feature with their antecedent.Besides, corresponding overt pronouns can be used in place of the null subjects without affecting the meaning of the sentences.This shows that they are null analogues of the corresponding overt pronouns.

4.
3. The 3rd person pro, a pure pronominal category.The 3rd person pro, on the other hand, is a pure pronominal empty category, in the sense that it allows only pronominal interpretation unlike the 1st and 2nd person pros.'Rama Rao has carried out an investigation of Kannada language from a broad perspective.It is clearly seen in the present work that he has adopted ideas of Indian linguists and of Western traditions also .' In (42), the endu-clause in the second sentence has a 3rd person pro subject as indicated by the verb inflection and pro refers to Rama Rao in the previous sentence.Here pro's "antecedent" is in the discourse keeLida asked.3SM 'Dinesh asked Ravi if (pro/he) is coming to the party.' Here, the 3rd person pro cannot refer either to Dinesh or to Ravi.It can only refer to someone else as indicated by the index.This is true of an overt 3rd person pronoun also.
(44) dineSa i Dinesh 'Dinesh asked Ravi if he is coming to the party.' As sentence (44) shows an overt 3rd person pronoun is perfectly acceptable and the sentence means the same as (43).So following these observations, we assume that a 3rd person pro and a 3rd person overt pronoun can have only pronominal interpretation.In this respect, the 3rd person overt and null pronouns differ from the 1st and 2nd person overt and null pronouns.The overt 3rd person pronoun differs from the corresponding overt pronoun in English also, since in English an overt 3rd person pronoun subject (in embedded clauses) can have both pronominal and anaphoric interpretations.
In sum, 1st and 2nd person null subjects allow both anaphoric and pronominal interpretations and they also exhibit feature clash, especially in their anaphoric reference.The 3rd person null subject, on the other hand, is exclusively pronominal in its reference.However, they all occur in argument CP and DP clauses.This way, these null subjects share a structural relationship with the matrix clause.Thirdly, an overt pronoun can be used in place of pro in all of these cases.Hence, they are basically pronominal and they obey Binding Condition B.
In section 4.2, I have described how a 1st person pro is derived.The same analysis holds for 2nd and 3rd person pros as well.Just like the 1st person pro, the 2nd and 3rd person pros are also specified not only for F-features but for their anaphoric/pronominal reference as well.The 2nd person will have the feature +object, +listener and the 3rd person pro will have the feature +pronominal.When an Agree relation is established between T and pro, pro's (interpretable) features are copied on to T so that they are available at LF.In both cases, pro raises to Spec, IP of the embedded clause in fulfillment of the EPP requirement.

4.4.
Pro as an anaphor: Evidence from the adjunct anta-clause.Pro as an anaphor occurs only in adjunct endu/anta-clauses and also in the appositive endu-clause.The adjunct clause is non-quotative and invariably functions as an adverbial clause of reason.So, there is a cause-and-effect/result relation between the enduclause and the main clause.Secondly, pro in the adjunct clause does not exhibit any feature clash.Nor does it have pronominal interpretation like the other pros in the argument endu-clauses.We have seen that the null (and overt) subjects in the argument endu-clause are mere "copies" of the pronominal subjects in direct speech and hence they are basically pronominal in character.The null anaphor, on the other hand, occurs in a non-quotative (adjunct) anta-clause and hence it is not pronominal in character.Furthermore, pro in the adjunct clause is non-locally bound by an NP in the main clause and I argue that the cause-and-effect relation that holds between the adjunct and the matrix clause is responsible for an anaphor-antecedent relation between pro and an NP in the main clause.In (45), the 1st person pro is bound by the object pronoun nannannu, 'me' and the subordinate verb shows 1st person inflection accordingly.In ( 46), the 2nd person pro refers to the 2nd person object pronoun ninnannu, 'you'; here again the verb has 2nd person inflection.In (47), the 3rd person pro is bound by the oblique NP GitaLa.In (48), the 3rd person pro is bound non-locally by tannaheNDati, 'his wife'.Interestingly enough, if the 3rd person pros in (47-48) are replaced by the corresponding overt pronouns, the overt pronouns will not to refer to GitaLa and tannaheNDati, respectively.In other words, the 3rd person anaphoric pro and the corresponding overt pronoun are not interchangeable in these sentences.However, if the subordinate clauses in (47)(48)  COMP 'Ramu brought a maid for his wife (thinking) since she does all the work alone(by herself) at home.' In the above sentences, the overt pronominal subjects in the subordinate clauses are co-referential with the genitive NP and the dative NP, respectively, in the main clauses.This shows that in order for an overt 3rd person pronoun to be anaphoric, the binder has to precede the pronoun.In other words, it has to meet the Leftness Condition.This constraint/condition can be stated as follows: (51) When there is a potential binder in the main clause, an overt 3rd person pronoun in the subordinate clause can be bound non-locally iff it meets the Leftness condition.
This condition does not hold for the null subjects in (47-48).A pro is always bound irrespective of whether it precedes or follows its binder.
We noted in sections 4.1 and 4.2 that both 1st and 2nd person null subjects and the corresponding overt pronouns can replace each other without affecting the meaning of the sentence.And in section 4.3, we saw that the 3rd person pro as well as the 3rd person overt pronoun cannot have anaphoric reference.They can have only a pronominal interpretation.So the Leftness Conditon does not hold for 3rd person pro and overt 3rd person pronoun, or for the reflexive taan 'self'.Only bound pronouns which occur in adjunct clauses have to meet the Leftness Condition.
Let's sum up our observations about pro in the adjunct anta-clause.Pro here has the same F-features as its antecedent, which means there is no feature mismatch between pro and its antecedent.The matrix Kansas Working Papers in Linguistics | Vol.38 (2017) | pp.23-42 verbs in (45)(46)(47)(48) are not verbs of saying or thinking and the anta-clause is not an argument of these verbs.The anta-clause here functions like an adverbial clause of reason.This means that the relation between pro and its antecedent cannot be defined structurally.There is, of course, a semantic relation between the adjunct and the main clauses.It is probably the cause-and-effect relation between the adjunct and the main clauses which forces pro to be bound by an NP in the matrix clause.This is again confirmed by the following example in which pro is bound in the matrix clause in spite of the fact that the verb in the negative adjunct clause lacks agreement.Here pro unambiguously refers to henDati 'wife'.I conclude that it is the cause-effect relation between the adjunct and matrix clauses that renders pro unambiguously bound.Lack of agreement in the adjunct anta-clause does not seem to affect the anaphoric status of pro.

4.5.
Pro as an anaphor in the appositive anta-clause.We will now consider the appositive anta-clause which also licenses an anaphoric pro.Generally, when the anta-clause occurs with verbs like hogaLu 'praise', Tiikisu 'criticize', bai 'scold' (and other verbs such as salahe koDu 'give suggestion' prakaTisu 'announce'), etc., it is in apposition to these verbs as their complement, not as their argument, and serves as an explanatory equivalent to the verb.These verbs also have a quotative or a reporting function; but the anta-clause does not function as an argument of these verbs.This way the anta-clause has some kind of semantic relation with the matrix verb.
Interestingly enough, this clause has both a direct speech version and a reported speech version.However, in both versions, pro allows only anaphoric interpretation.In the direct speech version, there is a mismatch in person feature between pro and its antecedent.And in the reported speech version, pro and its antecedent share the same person feature.The reported version is similar to reported speech in English in that the anaphoric subject pronoun in the subordinate clause shares its person and number features with its antecedent, unlike in the case of 1st, 2nd, and 3rd person pronominal pros in argument embedded clauses.
In the direct speech version, the appositive clause can license either a 1st person or a 2nd person pro in the subject position referring either to the speaker or to the listener, respectively, in which case the verb will have a 1st or a 2nd person inflection, respectively.Despite the 1st or the 2nd person inflection on the verb, pro here does not seem to show any ambiguity in its reference.Consider the following direct speech version in which the 2nd person pro can be coreferential with an NP of any person in the main clause.
( The 1st person pro and the 2nd pro in (54, 55) unambiguously refer to the speaker nannannu 'me' and the listener ninnannu 'you' respectively.There is no feature mismatch between pro and its antecedent, nor is there any ambiguity in pro's reference.An important point we notice in (54), is that the 1st person pro is not subject-oriented, as it generally happens in argument endu-clauses.As shown in the above examples, an overt pronoun can replace pro in (54, 55) without affecting the meaning of the sentence.However, an overt 3rd person pronoun cannot replace pro in (56), as indicated by an asterisk in front of the pronoun.We noted in section 4.4 that a 3rd person overt pronoun can replace a 3rd person pro in the adjunct enduclause, if it meets the Leftness Condition.This is because the subject position of an adjunct endu-clause is a bound position, hence only a bound pronoun is licit in this position.So in order to meet the Leftness condition, the subordinate clause has to be extraposed.This seems to hold for the 3rd person overt pronoun in the appositive clause also, as shown in the grammatical sentence below in which either a pro or an overt pronoun is bound by the object NP Shantina in the main clause, when the subordinate clause is extraposed.To sum up our discussion in this section, the appositive anta-clause is not an argument of the main verb just like the adjunct anta-clause.It is a complement in apposition to the matrix verb.In a way, it is semantically tied to the matrix verb.We also saw that the appositive clause occurs with verbs like hogaLu 'praise', bai 'scold', Tiikisu 'criticize', all of which have quotative function.So the anta-clause has a direct speech version and also a reported speech version.Pro in the former shows person feature clash with its antecedent; but it has only anaphoric interpretation.And pro in the reported speech version does not exhibit mismatch in person feature with its antecedent.I feel that pro in the reported version of the antaclause alone meets the most important requirement of an anaphor that it match its F-features with those of its antecedent.So it is a real anaphor, but the one in the direct speech version is a marginal one since it exhibits a feature clash.

Conclusions
Kannada, as we noted, is an inconsistent language, in that it possesses characteristics of both consistent null subject type and radical pro-drop type.This is because rich agreement is not present consistently in all types of sentences.In Kannada, either there is rich agreement or there is no agreement at all.In either case, a null subject is the basic unmarked option in all subject positions except when there is emphasis or some other feature.So the null subject option is assumed to be the underlying option in my account.Accordingly, I have provided a Spellout analysis of pro according to which pro gets pronounced when there is a special feature like emphasis, etc.
The main objective of this paper was to argue against the general view that pro is a pure pronominal empty category and to show that Kannada licenses a pronominal pro and an anaphoric pro as well.Therefore, pro in Kannada is on a par with the pronominal anaphor, the big PRO.However, as we noted, the pronominal pro and the anaphoric pro occur in different syntactic environments.Besides, we also saw that the agreement inflection plays a crucial role, particularly in subordinate clauses, since agreement not only identifies the referential features of pro but indicates its pronominal/anaphoric status as well.So the main focus of this paper was on the pronominal and the anaphoric pro licensed by rich agreement.Besides, we considered in what respects null subjects are similar to overt pronouns and in what respects they differ from them.
We noted that the case of pro in root clauses was a fairly straightforward one.A null subject of any person is allowed in all clauses and in all tenses provided there is a person feature on the finite verb.This means that predicates such as nominals and adjectives which lack a person feature cannot have a null subject.The object pronoun can also be dropped and recovered from the discourse since in Kannada, (transitive) verbs do not show agreement with the object.In reflexive sentences, the reflexive light verb koL licenses a null pronoun in the object position and this null pronoun is locally bound by the subject pronoun, null or otherwise.
We saw that the case of pro in embedded clauses, by contrast, was a complicated one.The pronominal and the anaphoric pros occur in different environments.The pronominal null subject occurs in argument CP and DP clauses while the null anaphor occurs only in non-argument clauses.Secondly, the person feature of the null subject determines its pronominal/anaphoric status.So four types of pro were proposed, each of them possessing a distinct referential property depending upon its person feature.
The 1st and 2nd person pros can have both anaphoric and pronominal interpretations just like the corresponding overt pronouns.In other words, they refer to the subject and the object, respectively, irrespective of the person feature of the subject and object.So in these cases pro's person feature does not match with that of its antecedent.The 1st and 2nd person pros also refer to the speaker and the listener, respectively.The 3rd pro and the corresponding overt pronoun, on the other hand, are pure pronominal categories in that they allow only a pronominal interpretation.We noted that the feature mismatch in the case of 1st and 2nd person null subjects was due to the following factors: Kannada quotative verbs such as verbs of saying and thinking simply report direct speech.But there is no difference in person feature between the subjects of embedded clauses in reported speech and those in direct speech.The subjects in reported speech are just "copies" of the pronominal subjects in the direct speech situation.Accordingly, agreement on the embedded verb reflects the 1st, 2nd, and 3rd person agreement inflection in direct speech.It is this which leads to a mismatch in person feature between pro and its antecedent, especially in the case of 1st and 2nd person pros when they have anaphoric meaning.It is these parallels between reported speech and direct speech that are responsible for the pronominal character of these pros.An important property that the 1st, 2nd, and 3rd person null subjects share is they all occur in clauses which function as arguments especially of quotative verbs.The 1st and 2nd person null subjects exhibit feature clash which goes against Binding Condition A and disqualifies them to be anaphors.The 3rd person pro allows only pronominal interpretation.This way, these null subjects are basically pronominal categories in keeping with Binding condition B.
Since the null subject is the unmarked option in Kannada, I assumed that the null subject is the basic underlying option and proposed a Spellout analysis of pro according to which a null subject gets pronounced only when it has some special feature like emphasis, etc.Following Roberts (2010), I assumed that T is pronominal in Kannada especially in clauses which show rich agreement, and that it has a uD-feature and also an EPP feature.I also assumed that F-features on T and case feature on arguments are interpretable in Kannada.Pro in Kannada is not only specified for F-features, but also for its pronominal/anaphoric reference, since pro's person feature largely determines its pronominal/anaphoric meaning.Accordingly, 1st and 2nd person pros are marked for the features +subj/speaker, and +obj/listener, respectively, in addition to the F-features.The 3rd person pro will have the feature +Pron.When an Agree relation is established between pro and T, the interpretable F-features are copied on to T so that these features are not deleted, but are visible at LF as well.
Pro, the anaphor, does not share any structural relationship with its antecedent, since it occurs in nonargument adjunct and appositive endu-clauses.The adjunct and the appositive clauses share a semantic relation with the main clause.The adjunct clause in particular is not quotative; it acts as an adverbial clause of reason.Pro in the adjunct clause does not exhibit any feature clash; it shares its F-features with its antecedent although it does not meet the structural C-command requirement; it is bound non-locally by an NP in the main clause.It is the cause-and-effect relation between the adjunct endu-clause and the main clause that licenses an anaphoric pro in the adjunct clause.
As for the appositive endu-clause, it has a direct speech version and a reported speech version as well.In the former, pro shows a feature clash and in the latter, it does not.So I look upon the pro in the reported speech version alone as a real anaphor since it meets the morphological requirement that it has the same referential features as its antecedent.
In either of these clauses, the 3rd person overt pronoun in place of pro can be anaphoric only if the subordinate clause is extraposed to the clause-final position.This shows that the overt 3rd person pronoun in the adjunct and appositive clauses obey the Leftness Condition.This condition does not apply to the overt 3rd person pronoun occurring in the argument clauses.
There has been very little work done on null subjects in Kannada.The present study is a modest attempt to fill the gap and therefore it is by no means exhaustive.Further research is required before we can say anything definite about the two types of pro in Kannada.
constructions.Consider the following empty expletive constructions: .PERF-3SN 'It has been said in the Bhagavata that Krishna is God's incarnation.''It feels to me that our new teacher is married.' said here that Dravidians came first to Bharat (India).' Sentences in(23)(24)(25) are examples of empty expletive constructions.( angry with his wife as she does not do any work at home.' teacher always praises Shanti (saying) that she sings very well.' Kansas Working Papers in Linguistics | Vol.38 (2017) | pp.23-42 The pure/underived ones show only number and gender features and they cannot have a pro subject, as illustrated in the following examples.
Your smoking cigarettes in front of my father is not appropriate.
we have looked at null subjects licensed by 'rich' agreement.As we noted earlier, Kannada also has a radical pro with verbs such as negatives and modals which are devoid of agreement inflection.
'(pro) hasn't joined yet.(pro) has to join within one week.' observations about the interpretations of the overt pronominal subject in the example he cites.3 , in which taanu in the subject position disambiguates this sentence.Here, taanu refers to Hari only and it cannot receive pronominal interpretation.Now let's see what happens if pro is inserted in the subject position in place of taanu.
Dinesh has not told anybody that pro will go to America.
Did Mary ask John if he is coming to the party' (Did Mary ask John if you are coming to the party?) Here is an example:In (41) we notice that the embedded verb inflects for 3rd person only.This kind of sentence with a pronominal pro occurs quite frequently in written Kannada, as shown in the example below: Kansas Working Papers in Linguistics | Vol.38 (2017) | pp.23-42 although its F-features are determined by the inflection on the embedded verb.Consider again this exam- Here are some examples.
are extraposed to the clause-final position, the overt pronouns do refer to GitaLa and tannaheNDati, respectively, as shown in the following examples.
The music teacher always praises you/me/Shanti (saying), pro (you) sing very well' As the above example illustrates, the null subject of the appositive clause is corerential with the object NP in the main clause irrespective of the object pronoun's person feature.Consider again the reported version of the appositive anta-clause: