LINGUIST List 29.3959

Fri Oct 12 2018

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

Editor for this issue: Jeremy Coburn <jecoburnlinguistlist.org>



Date: 01-Dec-2017
From: Sanford Steever <sbsteeveryahoo.com>
Subject: Dravidian Syntax and Universal Grammar (Part I)
E-mail this message to a friend

Discuss this message


Book announced at https://linguistlist.org/issues/28/28-3394.html

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

REVIEWER: Sanford B Steever,

SUMMARY

Dravidian syntax and universal grammar (DSUG) collects 28 papers written individually or jointly over the past 30 years by the husband and wife team of K.A. Jayaseelan (J) and R. Amritavalli (A). Two additional coauthors are involved: Deepti Ramadoss (chap. 14) and M. Hariprasad (chap. 28). The authors limit the Dravidian languages to the four literary languages Kannnada (Kan), Malayalam (Mal), Tamil (Tam) and Telugu (Tel). The first two receive the lion’s share of attention: Japanese is cited more than Tamil, Hungarian more than Telugu; none of the 20+ nonliterary Dravidian languages figures in the discussion. Universal grammar (UG) refers to the creation of models using government and binding, minimalism and principles and parameters. K and A introduce data from Kan and Mal to modify proposals in these models to shed light on UG and the individual languages. Several solutions involve proposing new functional projections in the left periphery to accommodate the data, e.g. MoodP, PerspectiveP, or recharacterizing existing projections to facilitate or block the interaction of certain grammatical elements.

The 28 chapters are divided into five topic-oriented sections: I, Scrambling and word order (chaps. 1-4); II, Syntax of questions and quantifiers (chaps. 5-10); III, Finiteness and negation (chaps. 11-16); IV, Case and argument structure (chaps. 17-24); and V, Anaphors and pronouns (chaps. 25-28). There are three indexes: Languages, Names and Subjects. The content and examples of one chapter significantly overlap those of others, sometimes with subtle, sometimes not so subtle differences. Negation and polarity for example are discussed in chapters 6, 11, 12, 13, 14, 16; finiteness in chapters 9, 11, 13, 15, 16; questions in chapters 2, 5, 6, 8, 9. There is heavy repetition of example sentences in chapters 5, 7 and 8. Due to the overlap, this review focuses on sections rather than individual chapters; for this purpose, the notation II.5.134 refers to Section II, Chapter 5, page 134 of DSUG.

Many chapters originated as seminar presentations, appearing in conference proceedings or journals. Unaccountably, the authors decided not to rewrite the chapters for DSUG (pp. x-xi), a decision that covers copyediting, fact-checking and resolving inconsistencies. This has allowed substantial problems to accumulate, often over decades, gain speed and snowball. As a result, the volume contains proposals that are mutually inconsistent or even contradictory. Since many of the errors function as material premises in the authors’ arguments, they undermine their conclusions. To help readers navigate these issues and better evaluate the arguments, I provide alternative perspectives and data that general, non-specialist readers might not have access to.

The decision not to revise has an immediate impact of the volume’s utility. For example, one finds on page 63 a reference to ‘Jayaseelan (2001b)’, which is now Chapter 5 of DSUG. Citations such as ‘XYZ (this volume)’, referring to where they first appeared, may be traced only with difficulty; citations such as ‘Cinque (to appear)’ not at all. This decision includes the systems for transcribing Kan and Mal: several appear in the volume, sometimes two different ones in one chapter; none is adequately explained. In language examples, proper names are capitalized even though Dravidian writing systems do not distinguish upper from lower case; this is problematic when capitalization is used to indicate retroflexion (e.g., the reflexive form Taan ‘self’ on p. 647 should be taan). From page 6 onward examples use ‘John’ and ‘Mary’ as is, not as Dravidian phonology requires, viz. jaan, meeri. Even Indic names such as moohan appear in their anglicized forms, i.e. Mohan. These examples give the impression of carelessness. The morphological segmentation and interlinear glosses are insufficient and follow no recognizable (e.g. Leipzig) conventions.

EVALUATION

Section I, Scrambling and word order, addresses Clefting, Scrambling, the placement of question words and word-order typologies. Several proposals involve expanding the left periphery by introducing new projections or repurposing existing ones. Traditionally, explicit case marking of nouns facilitates the permutation of major constituents within a clause via Scrambling. To explain IP-internal variations J appeals to the differential movement of constituents into (multiple) topic (ToP) and focus (FP) phrases. Differences in word order are thus referred to movement to different nodes and to scope-like differences in the relative order of ToPs and FPs on the clausal spine. This leads to all nonverbal elements occupying a ToP or FP. Overuse of multiple phrases, suggesting very subtle scope distinctions among the word-order variants, dilutes the utility of ToP and FP in identifying actual topicalized or focused constituents. J develops Mahadevan’s (1988) observation that in Mal non-Clefted questions, question words must appear immediately before the verb: he claims question words occupy FP and FP occurs immediately before the verb. This is likely related to the fact that in Mal Clefts the element focused by Clefting appears immediately before the copular verb aaNe. By contrast, Tam, Kan and Tel, whose Clefts lack a copular verb, permit the in situ placement of question words; the solution for the Mal data does not apply to them.

Going big, in Chapter 4 Jayaseelan extends the idea of movement into different phrases to explain differences between (S)VO and (S)OV word order. Earlier UG proposals privileged VO word order as “basic” while OV word order was “derived” through movement of constituents to before the verb, an implicit criticism of antisymmetry. J attempts to mitigate this by arguing that movement applies in both orders in two steps. First, in both orders movement creates Stacking, bringing a V together with its inflectional suffixes. Second, in VO languages the stacked elements are Stranded while in OV languages they are Pied-piped. However, this difference amounts to one between Pied-piping (OV order) and not Pied-piping (VO order): calling the non-application of Pied-piping Stranding doesn’t make it an independent process. Thus OV word order still undergoes an additional derivational step that VO order does not. This further suggests that all OV word orders are topicalized; that there is no “basic” (or untopicalized) order in Dravidian; and, hence, that Topicalization means something different in the two word-order patterns. Saying VO languages have Stranded constituents and OV languages Pied-piped ones merely re-labels the problem.

The preferential treatment of SVO word order in the authors’ model affects their treatment of relative clauses in Sections I and III. The SVO treatment requires relative clauses be finite and marked by a relative pronoun, which moves in between the head noun and relative clause, leaving a trace behind. That the relative marker is a (pro)nominal form is natural because in canonical order, the head noun directly abuts a position in the relative clause occupied by a nominal, viz. the subject, i.e. N’ (NP V NP). J’s adherence to antisymmetry (I.3.55) requires SOV relative clauses receive the same treatment. However, in SOV relative clauses the head noun abuts the verb of the relative clause, viz. (NP NP V) N’. In Dravidian, subordination of the relative clause to the following head noun is signaled by verbal suffixes (relative participle or conditional) or by such clitics as =oo ‘or, any’ (Steever 2017), neither of which is nominal. The suffixes prevent the verb in the relative clause from being finite; the clitic =o allows for the full expression of finiteness, but is a quantifier not a pro-form. The relevant NP in the subordinate clause is Gapped, not moved. Such differences with SVO languages seem not to fall out from Pied-piping.

In Section I and throughout DSUG the authors use the traditional term ‘relative participle’ (RP) to label the form that subordinate clause verbs assume when they combine with a head noun to form a relative clause, e.g., Mal ñaan kaNT-a kuTTi ‘(the) child whom I saw’ [I.NOM see-PST-RP child.NOM] contains the RP suffix –a. However, this form also occurs in factive and adverbial adjunct clauses. In Kan avanu edd-a saddu ‘the sound of him getting up’ [he.NOM arise-PST-RP sound.NOM], the relative participle edd-a ‘getting up’ combines with the head noun saddu ‘sound’ to form a complex nominal complement; in Tam enakku.t teri-nt-a varai ‘as far as I know’ [I-DAT know-PST-ADN limit] the verb teri-nt-a is a relative participle combining with head noun varai ‘limit’ to form an adverbial phrase. However, in neither sentence does the head noun relate to a theta-role in the subordinate clause: neither is a relative clause. J and A apply RP to examples that are not relative clauses, e.g. exx. (30b), (31a, b) on p. 375. Sometimes (III.16.399) they call the relative marker –a an “augment”; sometimes they do not gloss it at all (on IV.19.460 Mal oLiccu irikkum pooL ‘when X was hiding’ is glossed as ‘hide-sit when’ rather than [hide-CF be-FUT-RP time]). This illustrates an uncritical use of labels. For these reasons, Steever (1988, 2005) prefers the term ‘adnominal form’ (ADN) over RP. Curiously, J claims at several places that the relative participle/adnominal form in –a, e.g. Mal vann-a kuTTi ‘(the) child who came’ [come-PST-ADN child] is the same as the distal deictic marker aa- ‘that’. This fails in the future tense, e.g. Mal var-um kuTTi ‘(the) child who will come’ [come-FUT-ADN child], where the adnominal marker is –um. The appeal to homophony dissolves when the data set is slightly expanded; nor do the full range of Tam and Tel adnominal forms support J’s claim.

Section II, Syntax of questions and quantifiers, discusses certain uses of such forms as =um ‘and, all’ and =oo ‘or, any’ in Kan and Mal. While J and A’s model concentrates as much of the combinatoric properties of linguistic expressions as it can in syntax, this is an area where both morphology and semantics (including logic) will be welcome. First, morphology. J and A treat the forms =um ‘and, all’ and =oo ‘or, any’ as ‘particles’ or suffixes. As the boundary marker = indicates, these are clitics: they attach to right of a fully formed word; never figure in word-formation rules; and exhibit restrictions independent conjunctions might not show, e.g., =um never attaches to finite verbs. That these conjunctions occur “outside case” (II.5.130) follows from the fact that case is marked by a suffix and conjunction by a clitic; similarly, the ungrammatical sentences in exx. iii and iv on (III.15.381, 383) are ill-formed because clitics cannot occur within a word.

In II.5.132-3, J claims questions and quantifiers belong together because in Mal both can be signaled by =oo; however, in Kan, Tam and Tel, the interrogative clitic is =aa, while =oo serves as a quantifier. J’s argument rests on the homophony of two forms in Mal not on evidence elsewhere in Dravidian. He claims (II.5.148) that in Mal a “correlative clause has the same structure as a question.” However, they differ. In the Mal correlative aarǝ manassǝ aTakkunnu.v=oo avannǝ samaadhaanam kiTTunnu ‘whoever controls his mind obtains peace’ [who.NOM mind control-PRES=OR he-DAT peace get-PRES], the question word aarǝ ‘who’ and the clitic =oo co-occur although they cannot do so in a simple question; further, aarǝ is not adjacent to the verb aTakkunnu as required in non-Clefted questions (see above). II.5.134 states that disjunctive =oo is “completely parallel to English “or.” However, =oo must appear on every conjunct in a list while English permits only one in a list, e.g. ‘Peter, Bill or John’.

My glossing indicates that =um ‘all, and’ and =oo ‘any, or’ are general between quantifiers and conjunctions, differing according as they apply to set descriptions or enumerations, respectively. The authors embrace two policies that mask this generalization. First, they assume a version of unrestricted quantification, which does not define the domain over which the quantifier ranges. Second, they require that conjunctions always apply to exactly two conjuncts at a time (pp. 147, 153, 162) rather than indefinitely many. Both policies hamper linguistic and logical analysis. Even so, they ignore their pair-wise constraint when they write formulas with conjunction and disjunction applying to just a single “conjunct,” e.g., CONJ(DISJ(one (child))) on p. 196 is logically incoherent. The pair-wise restriction on conjunctions also obscures an interesting distinction in Kan between two forms that signal disjunction: =oo ‘(inclusive) or’ and illa ‘(exclusive) or’. This distinction is seldom instantiated in natural language: a language with just ‘and’ and ‘inclusive or’ can signal ‘exclusive or’ by conversational implicature or locutions such as ‘A or B, but not both’. So when A likens Kan illa to English ‘or’ (II.6.162), she misses the fact that inclusive ‘or’ in English has a different truth table. Here, greater clarity in logic could have informed a potentially interesting exploration of inclusive and exclusive ‘or’.

The authors attempt to reduce several basic logical operators to a single ‘Disjunction’ operator. II.7.189 claims that ‘or’ in ⋁(A, B) is defined as ⋀(◊A, ◊B), but without proof. II.7.193 claims ‘if’ is a disjunction operator like ‘or’, although both have different semantics (truth tables) and syntax. II.7.203 states that “In … ‘if p, then q’, ‘if’ applies to a propositional variable and outputs a disjunction of propositions.” However, ‘p’ represents a proposition, not a propositional variable; ‘q’ also represents a proposition, not a disjunctive of propositions. There are objections to taking one of ‘and’, ‘or’ and ‘if’ as basic and defining the other two in terms of it and negation (see McCawley 1993); in any event, deductive equivalence is not identity. J and A (II.10.241) try to set up different readings of ‘and’, based on different historical sources and typological trajectories (outside Dravidian), claiming ‘and’ comes from, e.g. ‘A then B’, ‘A along with B’. This digression supports neither a synchronic analysis nor a semantically coherent explanation. J asks (II.10.246), “What distinguishes conjunction and disjunction?” While he finally answers, “Meaning,” he does not propose how to incorporate that into his analysis. A basic command of semantics and logic would place this question and its answer at the beginning of their investigations, not the end.

In Section III, Finiteness and negation, J and A claim to offer an innovative analysis of finiteness in Dravidian, rejecting what they call the traditional view of finiteness as a combination of tense marking and personal endings for subject-verb agreement (SVA) on verbs. They claim Dravidian languages do not mark tense but aspect, and that SVA marking is not a feature of AGR but a reflex of indicative mood in MoodP. Nonetheless, they continue to treat finiteness traditionally: as a combination of a verbal category and SVA marking, just referencing their sources in different functional projections.

Chapter 11 discusses two Kan negative compound verb forms: baral(u) illa ‘did not come’ [come-INF be-NEG] and baruvud(u) illa ‘does/will not come’ [come-NPST-VN be-NEG]. The first is the negative counterpart of the finite past tense, e.g. ban-d-e ‘I came’ [come-PST-1S], ban-d-evu ‘we came’ [come-pst-1P], etc. It consists of the infinitive of the main verb (baru- ‘come’) and the negative auxiliary illa, a defective verb. The second compound is the negative counterpart of both the present (e.g. baru-tee-nee ‘I come’ [come-PRES-1S]) and the (literary) future tense (e.g. baru-v-enu ‘I will come’ [come-FUT-1S]) paradigms. It consists of the nonpast verbal noun of the main verb and the negative auxiliary illa. Neither compound marks person or number; exaggerating the difference between positive and negative forms, A claims that neither marks tense. This cannot go unremarked. Their paradigmatic opposition to tensed positive counterparts in negative contexts suffices to assure marking for tense, even if there were no explicit markers (J and A allow for ‘zero’ and ‘unpronounced’ elements elsewhere). However, the sequence –uvu- in the verbal noun baruvudu ‘coming’ is an allomorph of the nonpast/future tense morpheme (DSUG’s frequent use of colloquial –oodu for –uvudu obscures this). That it signals nonpast tense reflects the fact that Old Kan had an opposition between past and nonpast tense, which continues in the modern negative compounds. When Kan subsequently innovated a present tense paradigm, it did not automatically innovate a present negative compound verb (across languages negative polarity often has fewer distinctions than positive). Note, however, the South Kannnara dialect has innovated a present/future negative, e.g. maaD-al-ikk(u) illa ‘does/will not make’ [do-INF-DAT be.NEG], restricting maaDuvud(u) illa to the present negative.

Attempting to show that negative markers have a patterning distinct from tense markers, A claims that a negative allomorph –a- is infixed into nonfinite negative verbs, e.g., she segments Kan baaraddu ‘not coming’ [come-NEG-VN] as ba-a-raddu with an infix instead of baar-ad-du. Lengthening the verb root baru- to baar- reflects an ancient pattern in which a handful of verb roots had a long-vowel alternant in the negative (see Steever 1993). Such alternation does not occur in the vast majority of Kan verbs. In finite forms, the negative morpheme occurs in the same position where tense morphemes occur.

Amritavalli claims that negative forms mark aspect, not tense, but that through convoluted (and unconvincing) reasoning, they come to signal tense. The observation that “aspectual” negative compounds do not signal subject-verb agreement is taken as evidence that SVA markers must originate in the indicative mood option within MoodP. However, Kan (and other Dravidian languages) have both modal and negative forms that mark SVA. For example, Kan has a simple negative conjugation, consisting of a verb base, (zero) negative marker and personal ending, e.g. tiLiye ‘I did/do/will not know’ [know-NEG-1S]. The authors call this paradigm ‘now absent’ (p. 257, 329), ‘archaic’ (p. 276), ‘erstwhile’ (p. 401). However, a quarter-hour’s perusal of modern fiction yielded tiLiye ‘I do not know’ (two tokens), oppanu ‘he does not agree’ [agree-NEG-3SM], ariye ‘I do not know’, [know-NEG-1S], kaaNaru ‘they do not see’ [see-NEG-3P]. While such forms are infrequent, the conjugation is robustly preserved by the negative capabilitative auxiliary verb aar- ‘not be able’, which has finite and nonfinite negative forms, e.g. heeLidare niivu nambal aariri ‘If I told (you), you couldn’t believe (it)’ [tell-COND you-PL.NOM believe-INF be able.NEG-2P]. This is one of several examples from a 2017 novella by Viveeka Shaanabhaaga, one of Kan’s hippest contemporary writers. Although, this negative conjugation occurs in Old but not Modern Tam, it is maintained in Modern Tel and other Dravidian languages and cannot be dismissed as an archaism. To avoid the embarrassment that the modern Tel negative conjugation poses to their proposal, e.g. weLLanu ‘he will not come’ [come-NEG-3SM], they simply re-label it as “imperfective” without comment (III.16.401). Additionally, despite A’s claim that it is ‘archaic’ (II.11.273), Modern Kan also has a contingent tense paradigm consisting of a verb base, modal suffix and personal ending, e.g. biddiiye ‘you may fall, lest you fall’ [fall-CONT-2S]; it is well established in the language (and in the Rayalseema dialect of Tel).

J applies A’s analysis of ‘tense as aspect’ to the past and present negative paradigms in Mal, but for different purposes. Mal lacks SVA marking (apart, possibly, from some imperative forms). Past tense vannu ‘X came’ contrasts with present tense varunnu ‘X comes’; their negative counterparts are the compounds vann(u) illa ‘X did not come’ and varunn(u) illa ‘X does not come’, respectively. Since vannu ‘came’, varunnu ‘comes’ and illa ‘not be’ are all finite verb forms in their own right, the negative compounds appear to be the juxtaposition of two finite forms. J rejects this analysis of vann(u) illa ‘X didn’t come’ because he argues that a uniqueness constraint allows only one marking of finiteness (III.16.396); for him, vannu ‘came’ in the negative compound marks perfective aspect, not past tense, and varunnu marks imperfective aspect while negative illa alone marks finiteness in both compounds. Such forms cannot a priori be “doubly marked for finiteness” (III.13.302), but this fails to explain how vannu or varunnu are finite when they occur without illa. What J does not recognize is that vann(u) illa and varunn(u) illa are both serial verb formations (SVFs) in the sense of Steever (1988), constructions in which two finite forms may occur, either to signal concord between the component parts or to express combinations of verbal categories which cannot co-occur in any simple verb form of the language. Because polarity and tense may not co-occur in any simple verb of Kan or Mal (but may in Konda and Kuvi), SVFs are deployed (see Steever 1993, 2005). Such constructions occur in each branch of the Dravidian family and commonly beyond it, e.g. in Munda (Gorum), Tibeto-Burman, Indo-Aryan and Ainu. The uniqueness restriction cannot be maintained as the authors use it (Steever 1988 suggests how to accommodate such data in terms of a contrast between functional and formal finiteness). Although the authors ascribe perfective and imperfective aspect to the past and present tense forms, respectively, of Kan and Mal, the examples do not show the classical behavior of aspectual distinctions. Part of the confusion stems from their uncritical acceptance of Stowell’s (1982) treatment of infinitives and gerunds as being valid for Dravidian languages. For example, what J and A call gerunds in Mal and Kan are tensed verbal nouns whose subjects appear in the nominative case, not the genitive case, as in English and, apparently, UG. The combinatorics of the Dravidian and English forms differ, and there is no guarantee that one or the other is entirely transparent to their model of UG.

Though not anthologized in DSUG, J’s (1991) review of Steever (1988), an alternative approach to finiteness in Dravidian, is cited in the introduction to Part III (p. 253) and several other places in that section. It encapsulates the basic problems with J and A’s treatment of finiteness. J claims (p. 253) I represent the traditional approach to finiteness as a combination of tense marking and personal endings on verbs. In fact, I argue such an approach is flawed and propose a fundamentally different one involving not morphology, but syntax: “… syntactic rules identify a specially designated position in the constituent structure of a sentence where only finite predicates may occur. A predicate is finite or not depending as it occurs in that specially designated position or not …. The decision to apply the label of finiteness to some predicates squarely rests, not on such morphological criteria as the ability to bear certain verbal inflections, but on such syntactic criteria as the ability to occur in that position in a sentence which is specially designated for finite predicates [Steever 1988:2-3].” The primary motivation for this is that Dravidian predicate nominals occupy the exact same position in constituent structure as finite verb forms do; as nouns, they cannot take—or be defined by—verbal morphology. While J and A note the occurrence of finite predicate nominals (III.11.256, III.13.300), promising subsequent analysis, they divert them into parentheses where they are never heard from again.

Jayaseelan (1991) rejects my approach for doctrinaire reasons. First, my analysis permitted nonfinite relative clauses, which his model prohibited: relative clauses had to be finite. After 1994, however, his approach to finiteness in the Dravidian relative clause pivoted 180 degrees, claiming (III.13.320) that “relative clauses … in Dravidian must be non-finite [my emphasis].” His newer account now fails to account for sentential relative clauses (aka correlatives) which have finite predicates in relative clauses (and are exemplified in DSUG). This thwarts the attempt to displace Dravidian tense into aspect with the aim of preserving a spurious relation between tense, finiteness and relative clauses. Adherence to antisymmetry requires J to correlate Relative Clause Formation with finiteness while the Dravidian data show them to be independent variables (Steever 2017). Steever (1988) analyzes the “double marking” of finiteness in SVFs such as Old Tam cel-v-eem all-eem ‘we will not go’ [go-NPST-1P become-NEG-1P] as well as in many similar forms throughout Dravidian. Unable to reconcile these counterexamples with the uniqueness constraint, J ignores them as inconvenient. Despite claims of novelty, J and A’s approach to finiteness remains traditional: it ignores predicate nominals and pegs finiteness to a combination of verbal category and personal endings on verbs. Claims of their model’s superiority founder on it inability to provide even descriptively adequate analyses of finite predicate nominals, finiteness in relative clauses, serial verb formations, the negative conjugation or the contingent tense. The unsuccessful assimilation of tense to aspect and SVA to MoodP has the unwelcome side-effects of emptying tense and AGR of their usefulness in describing finite verb forms and deploying aspect and mood to explain the behavior of forms that may not intrinsically exhibit these categories.

Chapter 14, “The acquisition of negation in Tamil,” indicates Tamil children acquire negative verb forms more readily than their English counterparts acquire ‘not + V’. With a dearth of subjects (two primary, two secondary), the chapter makes for an interesting pilot study. Though Amritavalli and Ramadoss assert without proof that the Tamil form maaTTeen ‘I will not V, I refuse to V’ is not negative (it marks SVA), it routinely functions as the future negative auxiliary in the modern language, e.g. vara maaTTaan ‘he will not come’ [come-INF FUT.NEG-3SM] (Steever 2005, Chapter 3). On III.15. 365, J claims relative clauses cannot be coordinated. While they cannot be coordinated with a clitic such as =um ‘all, any’, they may be coordinated by simple juxtaposition in Kan, Mal and Tam or, in Kan, by use of the independent conjunction mattu ‘and’, e.g. haNa eNisuva mattu byaaŋkige kaTTuva huDugaru elli ‘where are the boys who count the money and deposit it in the bank?’ [money count-NPST-ADN and bank-DAT deposit-NPST-ADN boy-PL where]. Some Dravidian compounds join two expressions which are interpreted with a conjunction even though there is no overt marking, e.g. Kan avana-nanna sneehita ‘his and my friendship, the friendship between him and me’ [he-GEN-I-GEN friendship.NOM]. Juxtaposition as a means of coordination obviates the need for an overt conjunction, whether independent word or clitic. The authors ask (III.13.325) why the Dravidian languages have so many different markers of negation; one reason is a long history of restructuring the Proto-Dravidian verbal system (see Steever 1993:129 ff.).

In the end the authors undermine their own analysis of finiteness and tense. On III.11.266, A describes the three-way contrast of the Kan perfect auxiliary verb iru- ‘be’, viz. idd-anu ‘he was’ ~ iddaane ‘he is’ ~ iruttaane ‘he will be’, as perfective ~ present ~ imperfective, mixing the categories of tense and aspect. On III.15.366, J glosses Mal conjunctive forms (= ‘adverbial participles’) as PAST, but then says they have no past tense meaning. On III.13.313, A claims that tense and aspect morphemes in Kan are “homophonous.” These equivocations set the scene for drawing any conclusion one wishes from contradictory premises, a lapse in argumentation not confined to just this topic. In chapters written later they revert to using the labels ‘past’ and ‘present’ rather than ‘perfective’ and ‘imperfective’ in the glosses or simply utilize English past and present verb forms in the glosses, suggesting low confidence in their earlier analysis.

References appear in Part Two.


ABOUT THE REVIEWER

Sanford Steever has been engaged in the study of the Dravidian languages over the past thirty-five years. He has written several books on the structure and history of Tamil and other Dravidian languages. His latest book is ''Dravidian syntactic typology'', 2017 (Dravidian Linguistics Association).



Page Updated: 12-Oct-2018