K. A. Jayaseelan
I shall refer to the sentence type in which an NP with the Possessor/Experiencer theta-role has the dative case, by the neutral term ‘dative construction’. I shall prefer this term to ‘dative subject construction’, because I shall be arguing in this paper that this is an intransitive sentence type in which a nominative NP is the subject. (The nominative NP can be sometimes nonovert. The only other apparent exception is when, as in certain modal constructions, there is a clausal subject substituting for the nominative NP.) Therefore an imaginable principle that in a nominative-accusative language, there is always a nominative NP which is the subject, is not seriously disconfirmed by the existence of the dative construction.
There has been a good deal of theoretical activity regarding this construction recently. Shibatani (1999) has argued forcefully for the intransitive nature of the dative construction, a position I am in full sympathy with. He also tries to assimilate this construction to the “mutiple subject” construction of Japanese; the latter analysis however is not very helpful for a language like Malayalam, in which there is no “multiple subject” construction. Ura (2000) proposes an analysis in the minimalist framework. In his theory, which he calls “grammatical function splitting”, the grammatical functions typically associated with the subject can be realized by different NPs. For example, the triggering of agreement may be done by the nominative NP, and the control of PRO may be done by the dative NP. (It seems to me that we have in effect a “squishy” definition of subject, as a result.) I mention all this recent theoretical activity because it at least shows that this construction is still an actively debated subject of research, and remains a problem for the theory.
It may be useful to note (parenthetically) that the notion of analyzing the dative NP of the dative construction as the subject of the sentence, and (as a result) of seeing this as a “quirky subject construction”, is of comparatively recent origin. In the Indian grammatical tradition, the dative NP was never treated as the subject; see Amritavalli (this volume) for this point. Neither did European historical linguistics treat it this way: for example, the “impersonal construction” of Old English (which we look at later in this paper) was treated as a “subjectless” sentence, rather than as a sentence with a dative subject. My analysis of the dative construction therefore can be seen as a vindication of the older Indian and European traditions.
I shall first display the contexts in which the dative construction occurs. These contexts, as is well-known, can be grouped under some rough semantic rubrics (or descriptions).
In a sentence expressing the fact of possession, the possessor is in the dative case:1
Malayalam does not allow the option of any other case, e.g. a genitive case, here:
Observe that I have translated the verb uNDə in (1) as ‘be’, and not as ‘have’; for it is the same verb that occurs in a sentence like:
The facts are not different for instances of inalienable possession, cf.
Again, a genitive possessor is not possible:
When the possessed thing is something which can be carried on one’s person (like money), Malayalam also has a construction like (4), in which the possessor is inside a PP:
In the following sentences, the “experiencer” NP is in the dative case. In (5), the experience is a physical experience; in (6), it is a mental one.
However Malayalam seems to make a distinction between physical and mental experience in its diathetic patterns. In the mental experience case, the dative construction alternates with a nominative construction:
The physical experience case shows no such alternation; it can be expressed only by the dative construction:
The sentence (14b) is an interesting case (note the optionality brackets around the asterisk). A sentence like (14b) is actually possible, but only in contexts where the pain is interpreted as a mental pain. Thus (15) is a good sentence:
(Example (15) confirms our claimed distinction between physical and mental experience.)
In (13), I have glossed the verb taNukk as ‘feel cold’; but in fact the verb is neutral between the meanings ‘become cold’ and ‘feel cold’. Mohanan & Mohanan (1990) contrast the following sentences (see their examples (1); I have changed their glossing format):
The same verb, taNukk- / taNutt- , has the meaning ‘become cold’ when the subject is nominative, and the meaning ‘feel cold’ when the subject is dative.
The alternation of the nominative and dative constructions in the mental experience examples is sometimes claimed to have no effect on the meaning. (Thus, see Verma & Mohanan (1990), pp. 7-8.) But it seems to be the case that the nominative construction can be given an agentive interpretation, which is not possible with the dative construction. We can show this by the test of the Imperative Mood. The Imperative Mood is possible with the nominative construction, but not with its dative alternative:
Since the subject is normally deleted in imperatives, one can try “dropping” the dative NP in (17b) and (18b) by pro-drop; but it makes no difference to their grammaticality, cf.
By and large, the dative construction has a “complex predicate” as its verbal, consisting of a noun and what has been called (Mohanan & Mohanan 1990) a “light verb”. Cf. santooSam aak- ‘happiness become’ of (6)/(7a), dukham var- ‘sorrow come’ of (8a), iSTam aak- ‘liking become’ of (10a). But there are exceptions, e.g. vis’akk- ‘become hungry’ of (5), taNukk- ‘feel cold’ of (13a)/(16b), veedanik’k’- ‘feel pain’ of (14a). (Interestingly, all the exceptions signify physical experience.) We shall come back to the question of how the complex predicates are interpreted.
The ‘know’-class verbs should perhaps not be distinguished from the mental experience class of verbs. For one thing, they are like the latter class in showing systematic dative/nominative alternation.
There is often a difference of meaning in the alternants, however. For example, in the dative construction, aRiy- ‘know’ means a state of one’s knowledge; in the nominative construction, it means ‘come to know’. Thus, to say (e.g.) that someone ‘knows English’, we must use the dative construction:
As we remarked when dealing with the mental experience verbs, the nominative construction is consistent with an agentive interpretation, as can be shown by the ‘imperative test’. In fact, (21b) seems (fairly clearly) to be a causativized version of (21a).
From among the verbs of propositional attitude, the verb toonn- ‘seem’8 is the only verb that occurs only in a dative construction:
(Interestingly, from among the Old English “impersonal” verbs, ‘seem’ is perhaps the only verb which has resisted the change to the nominative pattern. We look at the Old English “impersonal construction” in section 4.2.)
The modal -aam ‘may’ occurs in the dative construction when it has the meaning of ‘permission’, and in the nominative construction when it has the meaning of ‘possibility’.
Unlike -aam which is morphologically a suffix, the modal kazhiy ‘can, or be able’ is an independent verb which takes an infinitival complement; and it occurs only in the dative construction:
The verb veeNam occurs in the dative construction when it means ‘want’ or ‘need’, and in the nominative construction when it means ‘must’. It normally shows up as a suffix -aNam on the lower verb:
But underlyingly it takes an infinitival complement, a fact which becomes clear when we interpose an emphatic particle between the lower verb and veeNam:
We can account for the dative/nominative alternation in (24) in the following fashion. Let us say that -aam with the meaning of possibility is a modal. Like all modals, it takes a VP complement:
(caak ‘die’ is an unaccusative verb and takes its argument in the object position.) The lower V caak adjoins to -aam by head-to-head raising; and the DP John raises to SPEC,IP and gets nominative case (or alternatively, raises to an IP-internal Topic position and has its nominative case checked by a ‘probe’ from I).
By contrast, -aam with the meaning of permission is a ‘full’ verb with two arguments, a Theme and a Goal. The Theme argument is obligatorily clausal for this verb; the Goal argument is marked with dative case:
As shown, the clausal argument has PRO as subject, which is controlled by the dative argument. The same analysis can be extended to the alternation in (26). In (26b), veeNam with the meaning of ‘must’ is a modal which takes a VP complement. In (26a), veeNam with the meaning of ‘want’ or ‘need’ is a regular verb with a Theme argument and a Goal argument. The Theme argument can be clausal, or it can be nominal; if the latter, it has the nominative case, cf.:
But in (26a), the Theme argument is clausal, with a PRO subject controlled by the dative-case-marked Goal argument, parallel to (29).
The conclusion of this section (on the ‘contexts of the dative construction’) seems to be a good place to point out that there is always a nominative NP in the dative construction. The nominative NP can be non-overt, as in a sentence like (5) (repeated below):
Here, the nominative NP is a pleonastic pro (see section 4 below for this analysis). Or, the position of the nominative NP can be occupied by a clausal argument, as in (24a)/(26a), which has the structure shown in (29). But this is not different from (say) the English sentence [That pigs fly] is true, where the nominative position is occupied by a clause. Given (then) that there is always a nominative NP, the only debatable point is whether the dative NP is nevertheless the subject in the dative construction.
Many tests of “subjecthood” have traditionally been applied to the dative NP of the dative construction, to determine whether this NP is indeed the subject of the construction. In this section we shall apply these traditional tests to the Malayalam dative NP.
This test cannot be applied in Malayalam because this language (uniquely among the Dravidian languages) has no subject-verb agreement.
The dative NP can control PRO:
However, as is well-known, PRO can be controlled also by a non-subject; therefore, control of PRO is not a good test of “subjecthood”.
A better test would be a demonstration (if possible) that the dative NP can be a PRO, since (universally) PRO occurs only as the subject of an infinitival clause. Unfortunately, such a test cannot be implemented; this is because a matrix verb which forces the embedded subject to be a PRO—e.g. a ‘try’-type verb—never allows the embedded clause to have the dative construction. This could be for a principled reason: a ‘try’-type verb requires the embedded subject to have agentivity, but the dative construction lacks an agentive theta role.
Some anaphors, it has been claimed, take only a subject as an antecedent; and Dravidian taan ‘self’ (a third person reflexive anaphor) has been claimed to be such an anaphor. Using this as a “subjecthood” test, it has been shown—see (among others) Sridhar (1976)—that taan can be anteceded by the dative NP of the dative construction. Here are some Malayalam examples:
However, as I have argued elsewhere (Jayaseelan 1990, 1997), it is incorrect to claim that taan takes only a subject as antecedent. Consider (33):
In (33a), the antecedent of taan, ‘John’, is a possessive NP within the subject noun phrase; and in (33b), it is a direct object.
In Jayaseelan (1998) I have shown that the antecedent of taan is determined by considerations of ‘perspective’. It follows (therefore) that taan’s taking the dative NP as antecedent in sentences like (32) shows nothing about that NP being syntactically a subject. We must conclude therefore that the “subjecthood” tests which have hitherto been applied to the dative NP are inconclusive.
It seems to me that the strongest reason for our thinking that the dative NP is the subject of the dative construction is that it obligatorily comes first in the word order of the clause:
The nominative NP (when there is one overtly) stays close to the verb, and is therefore thought to be “within the VP”; and the dative NP is therefore taken to be in the subject position.
But observe that the nominative NP (in these cases) is indefinite and non-specific; and the dative NP is definite. In Jayaseelan (1999b; 2001) I have argued for IP-internal Topic positions, to which definite/specific NPs may move. That IP-internal topicalization might be happening in a sentence like (34a) is suggested, when we compare (34) with (35):
The position of ii viiTT-il ‘in this house’ in (35a) is parallel to that of avan-ə ‘he-DAT’ in (34a); but we would not perhaps wish to say that the locative PP is the subject of (35a). That definiteness/specificity factors are at play here can be further shown by replacing the nominative NP of (35) with a definite NP:
The point I am trying to make is that the impression created by the position of the dative NP in a sentence like (34a) could be completely misleading.9
In this section, I advance some theoretical proposals to explain the dative construction.
In Jayaseelan (1990, 1999a) I argued that the analysis of the dative NP (of the dative construction) as the subject of the sentence, was a misanalysis which was facilitated by pro drop (or rather, pro drop in combination with scrambling). The argument went as follows. Consider the English sentence at (37):
Here, ‘to NP’ is the English equivalent of what would be realized as a dative case-marked NP in a language with a richer case system. In other words, (37) has a “dative NP” (in a sense), although it is certainly not the subject. Now, imagine that English were a language that allowed scrambling of the verb’s arguments and adjuncts. (37) would now have a perfectly acceptable alternative realization as (37’):
Imagine that English were also a pro-drop language. Pro-drop languages, we know, do not have pleonastic elements such as it and there; instead, they use a phonetically null pronominal element pro. (37’) would therefore actually be realized on the surface as (37’’), although underlyingly there would be a pro in the subject position, as shown in (37’’’):
Looking at (37’’), if ‘to me’ now were a dative case-marked NP, and not a PP, this would be indistinguishable from what we have called a dative construction; and linguists would be inclined to say that ‘seem’ is a “quirky” verb which assigns a dative case to its subject. In fact the equivalent of (37) in Malayalam is considered an example of the dative construction:
The claim that the dative construction is an epiphenomenon of pro-drop is supported by some facts about the historical evolution of English. Old English was an SOV language which allowed scrambling; and it was also a pro-drop language. And significantly, it had a construction which apparently had “no subject”, and which grammarians consequently referred to as an “impersonal construction”. This construction is illustrated in (39)-(40) (all OE examples from Lightfoot 1979):
The parallelism with the dative construction here is too obvious to need stressing.
English lost this construction towards the end of the Middle English period.10 The transition from Middle English to Modern English involved a number of changes. The language lost its case endings and simultaneously changed its word-order from the earlier SOV pattern to the SVO pattern; its word order also became more rigid. In the fifteenth century (as Jespersen points out, see Jespersen 1909-1949, II 10.12), the pleonastic elements it and there made their appearance, signalling (it seems to me) the fact that English had ceased to be a pro-drop language. A result of all these changes was that the impersonal construction disappeared from the language. Lightfoot (1979) notes that one of three things happened to the erstwhile impersonal verbs. A verb like ‘seem’ appeared with a pleonastic it as subject. From our point of view, what happened was (simply) that the underlying pleonastic pro subject of the impersonal construction (see (37’’’)) became realized as a lexical pronoun, English having lost its pro-drop property. (The dative NP of Old English now appeared as a PP, ‘to NP’; so we get a sentence like: ‘It seems to NP that …’) Some of the other impersonal verbs became obsolete (e.g. ‘behove’). The remaining verbs became reanalyzed, with the old dative NP becoming the nominative subject, e.g. ‘To the king, pears liked’ became ‘The king liked pears’. (Simultaneously, the verb underwent a change of meaning from ‘be pleasing to (somebody)’ to ‘find (something) to be pleasing’.11)
The history of English thus supports my claim about the dative construction being dependent on pro-drop. (It is almost like an experiment set up to test my hypothesis: Take a pro-drop language which has the impersonal (dative) construction, remove pro-drop, and see what happens!)12 In fact, these historical facts tell us something else also. They show that the construction in question is determined by purely syntactic factors, and that syntax is the area in which we must look for an explanation of it. It is important to stress this last point, because an idea is often entertained that the dative subject/nominative subject choice is determined by how a language (or a group of languages) chooses to ‘view’ certain types of events. Thus Klaiman (1986), speaking about Bangla, says that the choice between a dative and a nominative subject is determined in that language by the non-volitional/volitional distinction. She goes on to claim that the non-volitional/volitional distinction is an important distinction for South Asian languages but not for certain other language types—and that this constitutes a ‘semantic parameter’.
But if this were right, in view of the history of English, one would have to suppose that the volitional/non-volitional distinction was important to Englishmen in the Old and Middle English period, but it ceased to be important to the modern Englishman! (Perhaps implicit in a suggestion like that of Klaiman is a claim about ‘world views’: certain types of events are viewed as volitional in certain cultures, e.g. European cultures, and as non-volitional in certain other cultures, e.g. South Asian cultures. This neo-Whorfian claim seems clearly untenable when we try to imagine that Englishmen changed their ‘world view’ around the fifteenth century.)
I have suggested that the dative NP (of the dative construction) is not the subject of the sentence. In a sentence like (5) (repeated below), in which there is no NP other than the dative NP:
I am claiming that there is an underlying pro, marked nominative, which is the syntactic subject. As indicated in the translation, it is as if the sentence said, “To him, it hungers”; but since Malayalam has no pleonastic it, the language employs pro in that position. In a sentence like (6),
there is a nominative NP, santooSam ‘happiness’, which is the syntactic subject.
I am not necessarily claiming that the nominative NP (whether pro or lexical) is in Spec,IP. For one thing, it is not at all clear that in SOV languages of the South Asian (or East Asian) type, the subject is ever in Spec,IP. It is more plausible to say that it is in an IP-internal Topic position. (Recall our discussion of IP-internal topicalization earlier.) In fact, if we go along with Kayne’s (1994) suggestion of a “roll-up operation” in the syntax of strictly head-final languages, Spec,IP will be filled by I’s own complement which has moved up, so that it will not be free to accommodate the subject NP.
As we observed earlier, the nominative NP that occurs in the dative construction is invariably non-referential—it is indefinite and non-specific. (Cf. santooSam ‘happiness’ in (6).) It is therefore never topicalized. What does get topicalized is the dative NP. Hence the observed word-order: dative NP—other oblique NPs—nominative NP—verb. Cf. (20a) (repeated below):
The dative case of the dative NP (we must now say) is an inherent case (i.e. a ‘semantic’ case). Let us adopt the claim of Mohanan & Mohanan (1990) that the dative case is assigned to a goal argument in Malayalam. We shall also assume (following Jackendoff (1983), Mohanan & Mohanan (1990)) that each argument can be associated with a “bundle” of theta-roles; so that sometimes the speaker has a choice of case relations for an argument. Thus we have dative case and instrumental case alternating on the same argument in (41):
Observe how, with a different verb (which has a different semantics), the dative/instrumental alternation is not possible:
This strongly argues that the dative NP is an oblique argument, not a subject; since it is well-known that it is the case-marking of a verb’s oblique arguments which are purely semantically determined.
The nominative NP (which I analyzed as the subject of the sentence), when it is not a pleonastic pro but has semantic content, combines with the verb to compositionally determine the number and the case relations of the oblique arguments. The nominative NP (as I said) is always non-referential in this construction; it is therefore “predicative”. The verb is a single-argument verb, which takes the nominative NP as its single argument. The oblique arguments are therefore the arguments of the nominative NP.
In Jayaseelan (1990) I suggested that the nominative NP together with the verb should be treated as a “complex predicate”. (A “complex predicate” is an expression like ‘give a kiss’, in the place of a “simple predicate” like ‘kiss’; as illustrated by the parallel sentences ‘John kissed Mary’ and ‘John gave a kiss to Mary’.) I also demonstrated that the system of theta-marking in complex predicates proposed in Jayaseelan (1988) applied unproblematically to these cases. In this system, the unassigned theta-roles of daughter nodes are “promoted” to the immediately dominating phrasal node, where they are amalgamated according to certain principles of congruence; the phrasal node then assigns them. We can briefly outline how this system works, using a sentence like (43):
Here, uNDə is the existential copula, which takes a single argument, a Theme, to which it assigns the nominative case; in (43), it assigns that case to deeSyam ‘anger’. deeSyam ‘anger’ has two unassigned theta-roles, namely the Experiencer of the anger and the ‘target’ of the anger. These theta-roles are “promoted” up the phrase structure tree, and are realized as two types of Goals, with the oblique cases ‘dative’ and ‘second dative’.
While many theoretical questions remain to be addressed,13 the treatment of these constructions in terms of complex predicates appears to be on the right track. It may be emphasized (in this connection) that a noun-incorporation analysis—e.g. incorporation of deeSyam into uNDə—is not tenable. This is because the noun can have modifiers, cf.
1. The following conventions are adopted in the transcription: t, d, n are dental; t, d, n are alveolar; T, D, N, L, S are retroflex; ñ is palatal; k’, s’ are palatalized; v is a labio-dental approximant; R is an alveolar tap.
2. The apppearance of ‘be’, and not ‘have’, here is as expected. Kayne (1993) claims that ‘have’ is the realization of ‘be’ which has had a preposition – we can instead say ‘case’ (for Malayalam) – incorporated into it. In English-type languages, the dative preposition (or case) is incorporated into the copula and realized as ‘have’; wherefore the possessor NP must move into Spec,IP and get the nominative case. But in Dravidian, the ‘be’ stays as it is; and the possessor NP gets the dative case.
3. The genitive case on the possessor in a sentence like (4) is not a counterexample to what I have said above. ‘John’ is inside a PP here; and the objects of postpositions in Malayalam are normally in the genitive case.
4. The following is also possible:
Note that (i) and (10b) have the same verb form. This is exceptional, considering that in (7)-(9), the (a) and (b) sentences have different verb forms.
5. Note that in this case of alternation, it is not only the subject’s case which changes, but also that of the second argument. (11a) with an accusative in the place of the second dative argument is somewhat marginal (i), while (11b) with a second dative instead of the accusative is totally ungrammatical (ii).
6. Apparently, a parallel sentence is grammatical in Telugu and Kannada (as pointed out by K.V. Subbarao and R. Amritavalli). Cf. the following Telugu sentence:
Therefore, the unavailability of the nominative construction in the physical experience case seems to hold with full generality only in Malayalam (and possibly also Tamil).
7. In the dative construction, aRiy- always occurs with (what looks like) a modal -aam , for reasons which are unclear to me.
8. toonn- also has the meaning ‘feel (pain, happiness etc.)’ and occurs in a sentence like:
9. If the dative NP (of the dative construction) is itself indefinite, it was suggested in the discussion that the dative and nominative NPs could occur in either order. But on closer scrutiny, I seem to detect either a topicalizing effect on the nominative NP, or a focusing effect on the dative NP, if the order is ‘nominative NP -- dative NP’:
This seems to suggest that the ‘dative NP -- nominative NP’ order is a kind of ‘canonical’ order, with implications for the base order that I will not go into here. (But see Jayaseelan 2001.)
10. As pointed out by Howard Lasnik, the construction survived for some time in expressions like ‘Methinks … .’, where the agreement shows that the first person pronoun is not the subject.
11. Possibly some type of agentivity was imported into the meaning of like, so that the erstwhile unaccusative verb became a transitive verb. (Thus it is now possible to have -- although somewhat marginally -- an imperative: ‘Like him!’) As a consequence, the old subject became the direct object. (But why this ‘importing of agentivity’ strategy became the dominant strategy in English for dealing with the impersonal verbs remains to be understood.)
12. Currently, the dative construction in languages like Icelandic is analyzed as involving the dative NP moving into Spec,IP (see Chomsky 1998 and references cited there). (In locative inversion also, it has been claimed that the locative NP moves into Spec,IP. Both constructions involve unaccusative verbs, as pointed out by Anoop Mahajan.) The difficulty about adopting this analysis is that it would lose us the corelation between the dative construction and pro-drop. This corelation was substantiated by the history of English. If an oblique argument can be attracted to Spec,IP, this could happen irrespective of a language being pro-drop or non-pro-drop. One may now ask why present-day English doesn’t have a dative construction.
One may query—by way of probing our analysis—why, if there is a pro in the dative construction, its position is never realized by a lexical pronoun. The answer could be that “true” pro-drop languages do not have an expletive it or there. (In a semi-pro-drop language like German, a lexical pronoun does appear, cf. Es hungert mich ‘It hungers to me’.)
13. The theoretical issues have to do with how the “theta-role promotion” analysis can be conceived of within a Larsonian “VP shell” structure (Larson 1988) or within a system like that of Hale & Keyser (1997). I do not address these larger questions in this paper. (Pertinent also are the many issues raised in Dasgupta (2000).)
References
Chomsky, N 1998. Minimalist inquiries: the framework. Ms., MIT.
Dasgupta, P 2000. Review of K.A. Jayaseelan, Parametric Studies in Malayalam Syntax (Allied, New Delhi, 1999). In Indian Linguistics 61(1-4): 183–191.
Hale, K. and J. Keyser 1997. On the complex nature of simple predicators. In A. Alsina, J. Bresnan and P. Sells, eds., Complex Predicates. Stanford, California: CSLI Publications.
Jackendoff, R. 1983. Semantics and Cognition. Cambridge, Mass.: MIT Press.
Jayaseelan, K.A. 1988. Complex predicates and theta theory. In W. Wilkins, ed., Thematic Relations (Syntax and Semantics, Vol. 21). New York: Academic Press.
Jayaseelan, K.A. 1990. The dative subject construction and the pro-drop parameter. In M.K. Verma and K.P. Mohanan, eds., Experiencer Subjects in South Asian Languages. Stanford, California: CSLI Publications.
Jayaseelan, K.A. 1997. Anaphors as pronouns. Studia Linguistica 51(2):186–234.
Jayaseelan, K.A. 1998. Blocking effects and the syntax of Malayalam taan. In R. Singh, ed. The Yearbook of South Asian Languages and Linguistics. New Delhi: Sage Publications.
Jayaseelan, K.A. 1999a. The dative subject construction in parametric syntax. In K.A. Jayaseelan, Parametric Studies in Malayalam Syntax. New Delhi: Allied Publishers.
Jayaseelan, K.A. 1999b. A focus phrase above vP. Proceedings of the Nanzan GLOW. Nagoya: Nanzan University.
Jayaseelan, K.A. 2001. IP-internal Topic and Focus phrases. Studia Linguistica 55(1):39–75.
Jespersen, O. 1909-1949. A Modern English Grammar on Historical Principles, vols. I-VII. London: Allen & Unwin.
Kayne, R 1993. A modular approach to auxiliary selection. Ms., New York University.
Kayne, R 1994. The Antisymmetry of Syntax. Cambridge, Mass.: MIT Press.
Klaiman, M. H. 1986. Semantic parameters and the South Asian linguistic area. In Bh. Krishnamurthi, Colin P. Masica & Anjani K. Sinha, eds., South Asian Languages: Structure, Convergence and Diglossia (Motilal Banarsidass, Delhi), pp. 179–194.
Larson, R 1988. On the double object construction. Linguistic Inquiry 19:335–391.
Lightfoot, D. 1979. Principles of Diachronic Syntax. London: Cambridge University Press.
Mohanan, K.P. and T. Mohanan 1990. Dative subjects in Malayalam: semantic information in syntax. In M.K. Verma and K.P. Mohanan, eds., Experiencer Subjects in South Asian Languages. Stanford, California: CSLI Publications.
Shibatani, Masayoshi. 1999. Dative-subject constructions twenty-two years later. Studies in the Linguistic Sciences 29:2, 45–76.
Sridhar, S.N 1976. Dative subjects, rule government, and relational grammar. Studies in the Linguistic Sciences, 6:1, University of Illinois, Urbana.
Ura, Hiroyuki. 2000. Checking Theory and Grammatical Functions in Universal Grammar. New York: Oxford University Press.
Verma, M.K. and K.P. Mohanan, eds. 1990. Experiencer Subjects in South Asian Languages. Stanford, California: CSLI Publications.
