Dative alternation in Tamil: a non-derivational approach
     An important debate concerning the well known dative alternation between double object (dative-accusative;
D-A) and double complement (accusative-dative; A-D) type structures, reported in languages like English,
German, and Japanese, is whether the two alternants stem from a single argument structure (Larson (1988))
or from different ones (Harley (2002), Miyagawa and Tsujioka (2004), McFadden (2004)). This paper
provides additional support for the latter, nonderivational approach with data from the following:
     I. Passivization tests: Ditransitive verbs in Tamil seem to fall into two classes based on their passiviza-
tion strategies. In Class I, only accusative objects may become subjects in the passive (Example 1). In Class
II, both dative and accusative objects may do so (Example 2). Tamil has dative subjects and extensive scram-
bling, so morphological case and surface word-order cannot be taken as direct evidence about grammatical
function; however, independent subjecthood tests such as ability to bind a reflexive and control of P RO (see
Sarma (1999) for more) confirm that the nominative DPs in these examples are subjects and that a verb
classification based on such passivization strategies is valid. Class I constructions often take an unmarked
A-D surface order and the dative arguments express a non-possessional semantics; the accusatives typically
have experiencer -roles. Such characteristics are typically associated with an underlying A-D structure
(McFadden (2004), Harley (2002), and Pylkkänen (2002)). In Class II, the dative arguments can express
a possessional semantics, and take D-A surface arguments as unmarked. Thus, it appears that there is a
systematic correlation between these verb-classes due to passivization and types of argument structure.
     II. Benefactive applicative morpheme - kuău: The benefactive applicative morpheme kuău attaches
to the verb yielding ditransitive constructions with a standard (see Oehrle (1976)) double-object semantics
(see Lidz (2002) for a very similar pattern in Kannada). That is, kuău constructions imply the existence of
the G OA L NP's referent, and imply that the G OA L receives the T H E M E (Example (3a)). Crucially, in Class
I verbs with kuău, a benefactive dative may be introduced above the A-D arguments yielding constructions
with dative-accusative-dative structure (Example (3b)). This falls out naturally from the assumption that
Class I verbs have an underlying A-D structure and that kuău introduces an additional dative argument
above it; crucially, also, the dative argument from kuău must be thematically different from that in the A-D
verbs so that the same thematic position is not occupied twice by the two datives.
     III. Accusative case drop and frozen idiom chunks: Surface order in Tamil does not tell us much
about underlying structure because Tamil shows extensive scrambling. However, accusative case-drop and
idiom constructions where word-order is partially frozen, might throw some light on underlying structure. In
sentences with accusative case-drop (optional when the accusative object is non-rational and accompanied
by a generic reading), the case-dropped direct object is frozen directly left-adjacent of the verb resulting in a
dative-accusative surface order in ditransitives. This supports a D-A underlying structure in ditransitives (but
says nothing about whether an A-D order is also available). Idiom chunks in Tamil are frozen (scrambling
parts of the idiom elsewhere results in a loss of idiomatic meaning). Both accusative-verb (Example (4a))
and dative-verb (Example (4b)) idiom chunks are available, supporting the idea that both dative-accusative
(D-A) and accusative-dative (A-D) underlying structures, respectively, are available.
     Based on this evidence, I argue that there are two distinct argument structures available for Tamil ditran-
sitives: specifically, accusative-dative and dative-accusative underlying orders. These different structures
encode systematically different semantic/thematic relationships. Specifically, the dative-accusative order
takes a G OA L dative complement which encodes a possessional semantics while the accusative-dative or-
der takes an experiencer accusative object which encodes a directional/locational semantics; the choice
between one or other structure is due to the lexical semantics of the ditransitive verb. With respect to the
kuău applicative constructions, I will argue for an approach like Pylkkänen (2002) that relates them to other
dative-accusative constructions. A transformational account of the dative alternation cannot adequately ex-
plain the systematic semantic/thematic differences between the two argument structures. This paper thus
provides additional support for a non-derivational treatment of the dative alternation.

                                                      1

  (1) Class I verbs: only accusative is passivized:
       a. tiruăan     (ennaal) pooliiskaaran-ukku kaaúúikuăukka-paúú-aan (accusative passive)
          thief-N O M (by.me) policeman-DAT         expose-pass-3ms
          "The thief was exposed to the policeman (by me)"
       b. *pooliiskaaran (ennaal) tiruăan-ai kaaúúikuăukka-paúú-aan (dative passive)
          policeman-N O M (by.me) thief-AC C expose-pass-3ms
          Int: "The policeman was shown the thief (by me)"
  (2) Class II verbs: both dative and accusative are passivized:
                    paris-ai varanga-ppaúú-aan (dative passive)
       a. avan
          he-N O M gift-AC C award-pass-3ms
          "He was awarded a prize."
                      avan-ukku varanga-ppaúú-udu (accusative-passive)
       b. parisu
          prize-N O M he-DAT      award-pass-3ns
          "A prize was awarded to him."
  (3) kuău-constructions
                      koóandagaí-ukku bommai-yai vaangi-kuău-ttaa (implies G OA L NP's exxistence)
       a. seeta
          seeta-N O M children-DAT      doll-AC C buy-B E N-pst
          "Seeta bought a doll for the children." (strongly implies that she has children)
       b. naan raman-ukku tiruăan-aai pooliiskaaran-ukku kaaúúi-kuău-tteen (D-A-D order)
          I-N O M raman-DAT thief-AC C policeman-DAT           expose-B E N-pst.1s
          "I exposed the thief to the police for Raman."
  (4) Frozen idiom chunks (italicized)
                                                      kuău-tteen (accusative-verb idiom)
       a. naan avan-ukku kaariyattu-le kai
          I-N O M him-DAT work-L O C hand-AC C give-pst.1s
          "I gave him a hand with his work." (Int: helped with work.)
       b. naan ad-ai        selavu-kku vaangi-neen (dative-verb idiom)
          I-N O M that-AC C money-DAT buy-pst.1s
          "I bought that for good money" (Int: not for free)


References
Harley, Heidi. 2002. Possession and the double object construction. Linguistic Variation Yearbook 2:29­68.
Larson, Richard. 1988. On the double object construction. Linguistic Inquiry 19:335­391.
Lidz, Jeffrey. 2002. Two structures for Kannada ditransitives. Talk presented at SALA 2002.
McFadden, Thomas. 2004. The position of morphological case in the derivation: a study on the syntax-
  morphology interface. Doctoral Dissertation, University of Pennsylvania.
Miyagawa, Shigeru, and Takae Tsujioka. 2004. Argument structure and ditransitive verbs in Japanese.
  Journal of East Asian Linguistics .
Oehrle, Richard T. 1976. The grammatical status of the English dative alternation. Doctoral Dissertation,
  Massachusetts Institute of Technology, Cambridge.
Pylkkänen, Liina. 2002. Introducing Arguments. Doctoral Dissertation, Massachusetts Institute of Technol-
  ogy.
Sarma, Vaijayanthi M. 1999. Case, agreement, and word order: Issues in the syntax and acquisition of
  Tamil. Doctoral Dissertation, Massachusetts Institute of Technology, Cambridge.



                                                    2