LINGUIST List 29.3960

Fri Oct 12 2018

Review: Dravidian; Morphology; Syntax: Amritavalli, Jayaseelan (2017)

Editor for this issue: Jeremy Coburn <>

Date: 12-Oct-2018
From: Sanford Steever <>
Subject: Dravidian Syntax and Universal Grammar (Part 2)
E-mail this message to a friend

Discuss this message

Book announced at

AUTHOR: K. A. Jayaseelan
AUTHOR: Raghavachari Amritavalli
TITLE: Dravidian Syntax and Universal Grammar
PUBLISHER: Oxford University Press
YEAR: 2017

REVIEWER: Sanford B Steever,

Dravidian syntax and universal grammar. (Part Two)

Section IV, Case and argument structure, is primarily concerned with serial verb constructions and “dative-subject” constructions. The latter introduces subsidiary notions of case hierarchy and the incorporation of dative-case markers into nouns, both of which carry over into Section V. “The serial verb construction in Malayalam,” beginning at IV.19.453 is seriously compromised by inaccurate transcription which, with over 100 errors, obscures the distinction between the Mal finite past tense form vannu ‘X came’ [come-PST] and the nonfinite conjunctive form vannǝ ‘coming’ [come-CF] of Class II verbs. While traditional orthography does not distinguish u and ǝ, rendering both as u, they are separate phonemes in the spoken language (Asher and Kumari 1998): the ǝ of the conjunctive form may alternate with u, but the u of the finite form never alternates with ǝ. J’s transcription conceals this difference. (The past tense and conjunctive forms of Class I verbs in Mal are homophonous, e.g. kayari ‘X ascends’ and ‘ascending’, but account for just 30 percent of all verbs.) By distinguishing nonfinite conjunctive forms from finite tense forms, we can solve the puzzle (IV.17.432) that Mal allows “perfect participles” (=nonfinite conjunctive forms) in verb-chaining but does not allow “imperfect participles” in similar contexts: verb-chaining requires nonfinite forms and in *avaL aaDunnu paaDunnu vannu ‘she came singing and dancing’ [she.NOM dance-PRES sing-PRES come-PST] aaDunnu and paaDunnu are both finite present tense forms, not nonfinite imperfect participles.

J describes the Serial Verb Construction as a sequence of adverbial participles (=conjunctive forms), e.g. V1..V2..V3..Vn. He claims (IV.19.472) to be the first to study sequences of conjunctive forms as serial verbs, relating them to other languages in the world, notably Africa. What he is doing, however, is relating things he calls serial verbs in Mal to things called serial verbs elsewhere; there is no guarantee that in virtue of their labels they represent the same grammatical phenomena. What he discusses nowhere matches what Aikhenvald and Dixon (2006) call Serial Verb Constructions: a “sequence of verbs without any overt marking functioning as one predicate.” Even so, such sequences as V1..V2..V3..Vn embrace at least four grammatically distinct constructions: auxiliary compound verbs, descriptive compound verbs, verb-chaining (coordinate) constructions and postpositional structures (Steever 2005, Chapter 4 distinguishes the first three for Tamil).

Taking up Benveniste’s (1960) proposal by that ‘X has Y’ is in some respects equivalent to ‘to Y is X’, the authors pursue the observation that possessive constructions in Dravidian of the type Tam avan-ukku.p paNam iru-kkir-atu ‘to-him money is’ [he-DAT money-NOM be-PRES-3SM] translate English ‘he has money’. Their equation (IV.19. 480) is “to-DP be NP ~ DP(nom) have NP.” They extend this to Dravidian Dative-Subject Constructions (DSC), where certain predicates of feeling (‘know’), physical states (‘be hungry’), etc. subcategorize a subject in dative case. The difference they claim is that in the English construction, the dative case is now incorporated into the verb (‘have’) while the old nominative subject NP may be demoted to an adjective. In effect, they claim that the paucity of adjectives in Dravidian motivates DSCs, and that dative case may be incorporated into Nouns to form Adjectives and Verbs.

On IV.24.542, J claims that the nominative NP in DSCs designates the experiential state of the subject. However, many DSCs consist solely of a dative-case experiencer and a verb, e.g. Ma. kuTTikkǝ viśakkunnu ‘(the) child is hungry’ [child-DAT be.hungry-PRES]; there is no noun to be demoted or transformed into an adjective. More pointedly, the authors avoid treating DSCs with predicate nominals but no verbs, e.g. Tam avaLukku enniTam peeca payam ‘she is afraid to speak with me’ [she-DAT I-SOC speak-INF fear-NOM] or avanukku oru makan ‘he has a son’ [he-DAT one son.NOM]; these undermine the analogy with English ‘have’ constructions. On IV.18.447, the authors draw an unwarranted caricature of Klaiman’s (1986) analysis of Bangla subjects which claims that the choice between a dative or a nominative subject is based on the choice of the semantic parameter [-volitional] vs. [+volitional], respectively. They misrepresent her claim, insinuating that it requires some ultra-Whorfian model in which changes in languages reflect changes in speakers’ worldviews. First, the Bangla experiencer subject is in the genitive, not the dative case. Second, Klaiman (1986:179) defines parameter as “…something which obtains expression at the overt level of language structure.” Third, she provides five solid grammatical arguments for the differences between genitive (experiencer) and nominative subjects. This ridicule offers no support for their approach; in fact, Klaiman’s analysis explains the ungrammaticality of the Kan sentences (iia, b) in note 4 on page 586.

Chapter 22, “Parts, axial parts and next parts in Kannnada,” is the most successful chapter of this section because it focuses on teasing out distinctions between two kinds of postpositional phrases, e.g. between the “axial” reading ‘There was a kangaroo in the front of the car’ and the “part” reading ‘There was a kangaroo in front of the car’. Not only are different words for ‘before’ used, e.g. Kan munde ‘before’ vs. mumbhaaga ‘before, front’, but many AxPart words incorporate (at least historically) a dative case, e.g. keLage ‘below’ incorporates the dative marker -ge. J uses this as a premise to support the argument that other structures incorporate an unpronounced (and unseen) dative-case marker; he argues in Chapter 22 that Verb and Adjective arise when a nominal element incorporates dative case. How such incorporation differs from the simple marking of dative-case nouns, which stay nouns and don’t become Verbs or Adjectives, remains unclear. J attempts to explain it by proposing a case hierarchy in which DAT>ACC>GEN, claiming that all nouns mark all cases simultaneously and a variety of factors are responsible for the one that finally gets pronounced. He argues the dative form John-inǝ ‘to John’ [John-DAT] in Mal should be segmented as John-in-ǝ [John-GEN-DAT] because elsewhere the sequence in may mark genitive case. Traditional grammarians label such sequences ‘inflectional increments’, i.e., they create a noun stem that can then accept an inflectional suffix. In some dialects, the increment is optional; in others, it is obligatory and fused with a following case suffix so there is no motivation (apart from homophony) to introduce a supernumerary morpheme boundary or case. But J doubles down: on IV.23.519 he claims that the dative ‘to John’ is actually John-in-[e]-ǝ [John-GEN-ACC-DAT] where the accusative case marker –e happens not to be pronounced.

Suppose we began with a Mal noun from a different declension, say maram ‘tree’. The dative is maratt-innǝ ‘to the tree’ [tree-DAT], but since mara-ttǝ by itself is a locative ‘in the tree’ [tree-LOC], J’s analysis requires us to reinterpret the dative as mara-tt(ǝ)-innǝ [tree-LOC-GEN-DAT]. This implies a different case hierarchy: the choice between the two depends on the phonological shape of the noun, which determines membership of a declension. However, no such homophony exists in Kan or Tel declensions, suggesting the inflectional increment is an accident of Mal’s morphological history, not an argument to incorporate multiple inflectional endings into a single noun stem. Elsewhere J and A (p. 314) recognize the existence of ‘augments’ for verbs, but apparently not for nouns. Homophony also plays a role in J’s claim that the Mal dative case marker –kkǝ is the same sequence –kk- in verbs borrowed from Sanskrit, e.g., samsaarikk-uka ‘to converse’ [converse-INF] putatively incorporates a dative marker. There is in modern Mal, however, no motivation apart from homophony to introduce any morpheme boundary before –kk. Consider the Mal intransitive ~ transitive pair tiir- ‘be completed’ [complete-INTRNS] vs. tiir-kk- ‘complete’ [complete-TRNS]: intransitive tiir- is a verb even without the incorporation of a “dative case” marker. Further, the dative marker and so-called verbalizing suffix have distinct allomorphy. The dative case has the allomorphs –kkǝ ~ -inǝ ~ -ǝ; the verbalizing suffix (or increment) does not: there appear to be no verbs incorporating -inǝ or –ǝ allomorphs. The so-called verbalizing suffix may take additional verbal inflections, e.g. an infinitive marker, but the dative may not. Further, in some Mal causative verbs, the sequence –kk- appears twice, e.g. karakkikk- ‘make X rotate Y’ (< transitive karakk- ‘rotate’ [rotate-TRNS]) so by homophony a causative verb incorporates two dative-case markers. This sets the stage for a reductio ad absurdum. Evidence from Tam, Kan and Tel indicates that dative case and so-called verbalizing suffixes/increments are phonologically, morphologically and syntactically distinct, rebutting the proposal.

Chapter 24, ‘Rich Results’ attempts to decompose verbs into event types à la Vendler by adding a set of extended projections [initial], [process] and [result]. A claims (IV.24.556) that ‘to me fear’ of Kan nanage bhaya aayitu ‘I became afraid’ [I-DAT fear.NOM become-PST-3sn] appears under the [result] projection. This seems to be another argument that dative NPs may be “incorporated” into verbs. However, event types such as activities or accomplishments typically range over entire clauses not simple Vs, so this is not an instance of a verb incorporating a dative-marked NP. Here as elsewhere, the discussion loses focus from skipping from one language to another, e.g., IV.19.483 begins a digression on passive constructions in English and Hungarian (with no examples from Hungarian).

Section V, Anaphors and pronouns, aims to show that within UG anaphors are a subclass of pronoun. In its treatment of reflexives, this section draws welcome attention to partial similarities with reciprocals and distributive forms which other analyses ignore. In some instances, reflexives, reciprocals and distributives are all complex, e.g. the Kan reduplicated reciprocal obban-annu obbanu ‘each other’ [one-M-ACC one-M.NOM], and the Mal reduplicated distributive avar-avar ‘each one’ [they.NOM they.NOM]. Reciprocals and distributives are bound in the minimal S. Reflexives, on the other hand, function sometimes as anaphors, sometimes as pronouns, a problem for binding theory. When reflexive taan ‘self’ is not reduplicated, it requires disjoint reference, i.e., its antecedent stands outside the minimal S and c-commands it. However, use of certain devices, including reduplication, appears to block disjoint reference, requiring taan to take an antecedent in the minimal S. In Kan, insertion of auxiliary koLLu ‘take, buy’ permits local binding and in Mal insertion of tanne after the reflexive form does the same. Perhaps seeking to minimize its differences with Mal tanne, A calls koLLu a clitic, but it is clearly a verb with a full range of verbal inflections. J ultimately settles on treating tanne as a determiner that follows the reflexive form (V.26.598-9), without clarifying how determiners block disjoint reference. (Most forms with determiner-like function, e.g. Tam oru ‘one, a’, Kan ii ‘this’ and Tel aa ‘that’ precede the noun.) Reflexive forms thus modified appear to be hybrid anaphor-pronouns. Chapter 26 begins with a discussion of Mal phenomena but in section 2 segues into a medley of long-distance reflexive forms in other languages including Japanese, Korean and Icelandic. Here more attention might have been paid to relevant Dravidian phenomena. In Mal, reflexivity within the minimal S can also be conveyed by use of the pronouns svayam [self-ACC] and svantam [self-GEN], which are locally bound and noncomplex. As for reciprocals, Mal also has parasparam ‘mutually’ and anyoonyam ‘ibid’ while Kan has paraspara ‘mutually’, all of which are locally bound and noncomplex. (Though they may be analyzed as composite in their source language, Sanskrit, e.g., paraspara ‘mutually’ is a compound of para ‘other’, they are noncomposite in modern Dravidian.) In any event, it is a problem for their analyses that languages as similar as Kan and Mal resort to such different devices to block disjoint reference. And since neither device actually converts the minimal S into an internally complex structure so that reflexive taan would be an anaphor in all its uses, the original quandary remains intact.

In Chapter 27, J observes that anaphoric taan ‘self’ in Mal requires a third-person antecedent, which may occur arbitrarily many clauses up. However, the antecedent-anaphor relation is blocked when a first- or second-person pronoun intervenes in a tensed clause, e.g. one with COMP and realized by S[…]S ennǝ ‘that S’ (literally ‘S saying’). J proposes a PerspectiveP projection to capture this. An alternative is that each COMP clause behaves like direct discourse with the potential to reset person assignment, thereby breaking the link between antecedent and anaphor. (Whether the matrix in a complex sentence in Dravidian has a verb of propositional attitude, ordering or speaking, all such constructions are analogized to—or modeled on—direct discourse.) Each tensed clause with S[…]S ennǝ induces a new deictic center potentially different from that of the higher clauses; the presence of a first- or second-person pronoun confirms that change. Discussion of PerspectiveP in Section V and Anchoring in Section III, where anchoring refers to the use of tense/aspect to link the narrated event to the speech event, suggests an alternative of introducing into the sentential spine a projection that in effect serves as a performative clause. Chapter 28, “Deixis in pronouns and noun phrases,” proposes adding a deictic projection, DeixP, to the left periphery to account for the fact that third-person pronouns incorporate a morpheme to mark proximal (ii- ‘this’) or distal (aa- ‘that’) reference, e.g. Tam i-van ‘this man’ vs. a-van ‘that man’. Jayaseelan and Hariprasad do not explain why deictic markers should not be treated as determiners and included as possible expansions of D´ in DP. Perhaps J’s earlier appeal to a postposed determiner to label Mal tanne, noted above, pre-empts this possibility; an alternative analysis for tanne would free up DPs for deictic and other determiners.

Besides the issues with these large topics, the decision not to edit the individual chapters has left many problems; only a small representative sample is noted here. Chapter 2 twice claims that Sinhala has Clefts, but gives neither examples nor references (since the Language Index omits Sinhala, readers can’t cross-reference potential examples). Page 56 claims the Mal complementizer ennǝ [say-PST-CF] ‘saying’ never co-occurs with the relativizer –a, which reflects the fact that both the conjunctive and ADN forms must combine directly with a verb stem. (Nonetheless, the complementizer does have adnominal forms, e.g. Mal avan aviTel uNTǝ enna satyam ‘the truth that he is there’ [he-NOM there be-PRES say-PST-ADN truth].) Page 65 says, “Likewise in German, a sentence like ‘I wonder who left why/how’ is acceptable,” instead of offering an actual German sentence. Examples 19b and 20b on page 80 lack translations, which are, respectively, ‘those who saw me’ and ‘that which fell on me’. The authors claim on page 90 that Dravidian relative clauses may contain no Mood Phrase; however, Tam has the following adnominal forms of modal auxiliary verbs: veeNTiya ‘which is required, which must V’ [be necessary-PST-ADN] and muTinta ‘which may V’ [be possible.PST-ADN]. On page 168 Kan elligaadaru ‘to anywhere’ is glossed as ‘anywhere+DAT’ instead of correct elli-g(e)=aa-daru [where-DAT=become-CND], obscuring morphology and distorting the relative scope of elements. On page 171, the Mal example 24, translated as ‘I asked who came’, is glossed as “… lit. ‘I asked that who came’” when a truly literal translation would be ‘I asked saying who came’. Page 178 has ‘in Mysore’ as Kan mysur-alli instead of maisuur-alli. Page 198, ex. 70 has Tel coosyæDu ‘he saw’ (correct cuusææDu) instead of the cuusindi ‘they saw’ required by SVA. Page 210 translates Sinhala mahattea-Tǝ tee dǝ koopi dǝ oonǝ [sir-DAT tea=OR coffee=OR want] as ‘Does the mister want tea or coffee’ where mahattea ‘sir’ is better rendered as honorific ‘you’. Page 304 claims the infinitive is purposive, a meaning not consistent with its use in, say, the Kan past negative compound verb. Page 376 claims that Tam vantaan illai ‘he didn’t come’ [come-PST-3SM be-NEG] is ungrammatical; it is possible in spoken Tamil and conveys the meaning, “He was expected to come (but didn’t).” Page 376 also has Mal “grrr! enna śabdam” ‘the noise grrr!” [grrr! say-ADN noise.NOM] with an English ideophone, ignoring the hundreds of Mal ideophones available. Page 408 reports a claim that certain nonliterary Dravidian languages have agreement markers that distinguish ‘visible’ from ‘nonvisible’; review of the grammars, e.g. of Kuvi, shows the distinction applies in certain deictic pronouns, not in SVA markers. Page 499 (and elsewhere) has Kan yoochane ‘thought’ for yoocane. Kan example 12b on page 506 adiya [foot-GEN] should be aDiya. Page 654 glosses Japanese okanne o as ‘to money’, implying an indirect object when it is a direct object. Page 661 has “non-ccommanding” instead of non-c-commanding. Page 671 claims that the Munda languages are part of the Dravidian family; they are in fact a branch of the unrelated Austroasiatic family.

Failure to edit the individual chapters seriously compromises DSUG’s ability to serve as a source book of Dravidian material. There are over 200 misspellings in the language examples; moreover, the morphological segmentation of forms is muddled and the glossing inadequate. Third parties ought to consult specialists when utilizing these examples. The anthology references at least 48 languages; the Language Index (pp. 675-6) includes only 31. Seventeen are missing: Blackfoot (408), Bulgarian (139), Chichewa (431), Classical Armenian (426), Dakhini Urdu (374), Ga (458), Halkomelem (400), Hindi (57, 154, 565, 575), Munda (671), Lezgian (101), Saramaccan (458), Serbo-Croatian (152), Sinhala (45, 159-8, 212, 215), Sranan (465, 460), Thai (597-8), Vietnamese (597-8), Yoruba (428, 473). While I have not made a systematic tally of the two other indexes, the traditional Kan grammarian Keśiraja is mentioned in the text but not the Name Index; neither Topic nor Focus Phrase, which figure so prominently in Section I, appears in the Subject Index.

Structural and rhetorical problems undermine the arguments’ persuasiveness. For example, Chapter 4 has five sentences in Malayalam, five in Hungarian and seven in French, barely more than a squib. Chapter 26 is 51 pages long, with 16 pages taken up by 55 notes and two pages of references. This chapter and several others like it contain numerous digressions that fray the thread of their main arguments. More than once I had the impression that the discussion was moving toward some insight about Kan or Mal only to be turned away short of that goal by a digression. The use of parentheses to enclose a witticism is indulged too freely, e.g. “It may be useful to note (parenthetically)…” (p. 434). The authors’ dismissive use of the terms ‘Dravidianist’, ‘grammarian’, ‘traditionalist’ and ‘typologist’ will be viewed by many as ad hominem distractions, taking up space better occupied by valid arguments.

The general argumentation in DSUG falls short in the following areas. One is an uncritical reliance on labels without regard for their applicability to Dravidian, e.g. gerund, participle, relative participle. Two is a tendency toward reductiveness, e.g. reducing adjectives and verbs to nouns, and reducing ‘if’, ‘or’ and ‘not’ to a single DISJUNCTION operator. Three is the tendency to displace one category for another, e.g. shifting tense to aspect, AGR to Mood. Not only does this remove the displaced categories from the tool-kit of devices for linguistic analysis, it often relabels a problem without solving it. Four is a lack of morphological rigor, inserting morpheme boundaries arbitrarily and imputing categories (via glosses) which may not be justified by language-internal evidence. This includes the misuse of such terms as clitic, infix and suffix. Five is a lack of clarity in semantics, particularly as regards quantifiers and conjunctions. Six is a readiness to apply to homophony in syntactic argumentation. Seven is an overzealousness to preserve a model’s theoretical purity against potential counterexamples, e.g., attachment to the “uniqueness” requirement for INFL marking prevents the authors from recognizing verbal constructions with double markings of finiteness. These shortcomings pertain primarily to synchronic analysis but also characterize the authors’ occasional remarks on historical Dravidian grammar, which rarely rise above the level of folk etymology. When taken together, these lapses in argumentation often give rise to pseudo-problems (e.g. whether a relative clause must correlate with finiteness), whose pursuit leads the reader away from satisfactory solutions, not toward them. UG claims in part to enable the ready comparison of syntactic systems across languages. For this to work, it seems to me, care must be taken that the systems or subsystems being compared are effectively analyzed for each individual language. If not, their flaws and deficiencies will ultimately distort the larger claims for the individual languages and for UG.

DSUG successfully memorializes the role that Kan and Mal have played in the development of certain topics in UG over the past 30 years. The authors have introduced these data to a larger audience, embedding them in debates in several interesting areas of syntax. A major contribution of DSUG is to show how rich in possibilities Kan and Mal are for morphological and syntactic research, whatever the model. However, the solutions to many of the problems presented in this anthology fail to convince for the reasons stated above, particularly the decision not to edit inconsistencies or contradictions. These issues might be addressed going forward by condensing and consolidating the relevant chapters of each of the five sections into a state-of-the-art essay. The chance of providing satisfactory solutions will rise as focus sharpens and the number of digressions falls; as critical use of morphology and semantics rises; and as more data from other Dravidian languages and ideas from other linguistic approaches are included.


Aikhenvald, Alexandra and Robert Dixon (eds.). 2006 Serial verb constructions: a cross-linguistic typology. Oxford: Oxford University Press.

Asher, R. and T. Kumari. 1997. Malayalam. London: Routledge.

Benveniste, Emile. 1960. «Être» et «avoir» dans leurs fonctions linguistiques. Bulletin de la Société de Linguistique, LV.

Jayaseelan, K.A. 1991. “Review of Sanford B. Steever, The Serial Verb Formation in the Dravidian Languages, Linguistics 2:247-58.

Klaiman, Miriam. 1986. Some parameters and the South Asian linguistic area. In Bh. Krishnamurti (ed.), South Asian languages: structure, convergence and diglossia, pp. 179-94. New Delhi: Motilal Banarsidass.

Mahadevan, P. 1988. Quantifier structure in Malayalam. CIEFL Working Papers in Linguistics. Hyderabad, India: Central Institute of English and Foreign Languages.

McCawley, James. 1993. Everything that linguists always wanted to know about logic (but were ashamed to ask). Second edition. Chicago: University of Chicago Press.

Shaanabhaaga, Viveeka. 2017. Ghaacar ghoocar. Heggodu: Akshara Prakashana.

Steever, Sanford. 1988. The serial verb formation in the Dravidian languages. New Delhi: Motilal Banarsidass.

Steever, Sanford B. 1993. Analysis to synthesis: The development of complex verb morphology in the Dravidian languages. New York: Oxford University Press.

Steever, Sanford. 2005. The Tamil auxiliary verb construction. London: Routledge.

Steever, Sanford. 2017. Dravidian syntactic typology. Trivandrum: Dravidian Linguistics Association.

Stowell, Tim. 1982. “The tense of infinitive.” Linguistic Inquiry 13:561-70.


Sanford Steever, an independent scholar, has written and edited six books on Dravidian languages and linguistics, including Analysis to Synthesis: The Development of Complex Verb Morphology in the Dravidian Languages (1993), The Tamil Auxiliary Verb System (2005) and Dravidian Syntactic Typology (2017). Research interests include Tamil, Dravidian languages, historical linguistics and syntax.

Page Updated: 12-Oct-2018