You are on page 1of 13

K.

Tuite: Vamling review 13-03-02 1

Review of Complementation in Georgian by Karina Vamling [Travaux de linstitut de linguistique de Lund #23; Lund University Press, 1989] 165 pp.

Long-time readers of this journal will have already made the acquaintance of the Georgian language through the works of the late Hans Vogt, who was one of the first non-Georgians to undertake its detailed investigation. The book under review Complementation in Georgian [henceforth CIG] by Karina Vamling [KV] represents an attempt to devise a formal model for generating a certain class of complex constructions in Georgian. The word complement, as used here, denotes clauses including nominalized clauses which function as the subject or object of a matrix (main) predicate. The data presented in CIG show that the distribution of verb forms in complement clauses correlates with a semantic distinction, described below, between Truth and Action information modalities. The author also compiles and analyzes data on control in Georgian, and, on the basis of these and other facts, proposes a model for one segment of the Georgian lexicon. 1.1. Verb forms and complement types. The Georgian language has a dozen or so verb forms (also known as screeves in the Kartvelological literature). They can be grouped, on the basis of morphological and syntactic characteristics, into three series. Within each series verb forms are distinguished by, among other things, the suffixes they allow. These suffixes are portmanteau forms, indicating tense and mood as well as the person and number of the NP controlling Set V (or subject) agreement. The distribution of verb forms according to series and tense/mood/person/number suffixes is shown in the following chart:

[INSERT TABLE 1 HERE]

The symmetric pattern of Middle Georgian verb forms has altered somewhat. The permansive and (except in highly marked contexts) the perfect conjunctive have disappeared from the modern literary language. On the other hand, the pluperfect, formally and historically a past-

K. Tuite: Vamling review 13-03-02 2

indicative verb form, has come to function as a past modal as well. Almost all of the above verb forms can occur in Georgian complement clauses. Their distribution is dependent on various factors, among them as KV demonstrates information modality. This latter term has been taken from a study by E. Ransom (1986). The distinction it captures is that between complements of the epistemic type (which express knowledge and information) and those of the deontic type (which express an act or state of affairs to be realized). The Truth modality is one of three epistemically-oriented modalities recognized by Ransom (the other two Future Truth and Occurrence are categorized as Truth modality with particular timereference restrictions in CIG); and Action is the single deontically-oriented modality. Some Georgian verbs specify complements of one modality or the other: vici I know requires a Truth-modality complement, and upleba makvs I have the right (a phrase functioning as a lexical unit) must have an Action-modality complement (pp 60, 65). A third group of predicates allows both types, usually with a difference in meaning; compare [Truth modality] davajere vano, rom nino cavida (I convinced Vano [that it is true that] Nino left) and [Action modality] davajere vanoj, rom j cavides (I convinced [i.e. persuaded] Vano to leave) (pp. 92-93). Briefly put, only verb forms from the first two columns in Table 1 can occur in Truth-type complements in Georgian, and conversely, only those from the third column (modal) are acceptable in Action-type complements.1 The temporal relations among the event of utterance, and the events expressed in the matrix and complement clauses specify, in most instances, the particular verb form to be employed in the complement. This information, expressed in time-line diagrams in CIG (pp. 67-72), can be condensed into the following table:

[INSERT TABLE 2 HERE]

Compare the use of verb forms in these two Georgian sentences: [Truth modality] damenana, rom xidi dacves (I regretted [aorist] that they burnt [aorist] the bridge) and [Action modality]

K. Tuite: Vamling review 13-03-02 3

vubrdzane mat, rom xidi daecvat (I commanded them [aorist] that they burn [pluperfect] the bridge). 1.2. Control. The second major body of data presented in CIG concerns constraints upon the reference of complement subjects. Certain verbs which take Action-modality complements require that the subject of the complement be coreferent with either the subject or object of the matrix clause, as in the following sentences (p. 87):

vapireb

rom es gavaketo

I-intend-it that it I-do-it[optative] Ia intend to a do it

*vapireb I-intend-it

rom es gaaketos that it s/he-do-it[optative]

Ia intend that s/heb do it

vtxov

mas

rom

es gaaketos it s/he-do-it[optative]

I-request-it him/her that Ia ask him/herb to b do it

*vtxov

mas

rom

es gavaketo it I-do-it[optative]

I-request-it him/her that Ia ask him/herb that Ia do it

Other complement-taking verbs have no such control restrictions: minda rom es gavaketo

I-want-it that it I-do-it[optative] Ia want to a do it

K. Tuite: Vamling review 13-03-02 4

minda

rom es gaaketos

I-want-it that it s/he-do-it[optative] Ia want that s/heb do it

The control properties of most Georgian verbs do not differ significantly from those of their equivalents in other languages, and they can be grouped according to the semantic categories developed by Sag and Pollard (1988:14): commitment type [controller = agent], directive type [controller = goal], attitude type [controller = experiencer](p. 140). Two Georgian verbs, however, are characterized by a double control requirement: the matrix controls the complement subject, and the matrix indirect object controls the complement object (pp. 114-115, 144-145), e.g.

gindomeb I-want-it-for-you

sen yousg

caqvanas taking-away[dative]

Ia want to take youb away (lit. Ia want for youb the a taking away b)

In the case of non-finite complements,2 the reference of the (underlying) subject follows the control pattern for finite complements, i.e. subject-subject coreference for verbs such as vapireb, and object-subject coreference for verbs such as vtxov. Where both coreference and disjoint reference are possible in a finite complement (e.g. minda), a non-finite complement prefers the coreference interpretation, while the disjoint reference interpretation is, at best, marginally acceptable. This is clear evidence that certain control configurations are less marked than others (p. 110). 1.3. Lexical representation. On the basis of facts such as those presented in the preceding two sections, KV formulates a model for the lexical listings of Georgian verbs, and rules for deriving

K. Tuite: Vamling review 13-03-02 5

the surface form of matrix-complement constructions from information contained in the lexicon. She states that her model is in the spirit of the Head-Driven Phrase-Structure Grammar framework of C. Pollard and I. Sag, but with certain modifications. The information contained in the lexical entries is grouped under three headings: phonological, syntactic and semantic. The semantic information for a verb includes its meaning, and the arguments to which it assigns semantic roles. The syntactic component of the lexical entry includes category features (V, N), and, for a Georgian verb, the lexical class to which it belongs. KV recognizes three classes, based on case-assignment and agreement patterns: direct, which corresponds to the 1st and 3rd conjugations of Aronson (1982) and Harris (1981); neutral, corresponding to Aronsons 2nd conjugation; and indirect, or dative-subject verbs (Aronsons 4th conjugation). Direct verbs undergo case shift, that is, the case assigned to their subjects and direct objects shifts according to verb-form series. Once the verb form is selected, that, in conjunction with the verb class specification, determines the mapping of semantic roles onto NPs with specific case and verb-agreement properties (the Georgian verb has two sets of agreement affixes: Set V, or subject, and Set M, or object). Complement-taking verbs have additional information in their lexical entries, which will determine the configuration of the subordinate clause:

[INSERT TABLE 3 HERE]

KV gives an example to illustrate how her model works (p. 142). A complex sentence meaning Nino asked me to paint her, employing the Georgian verbs txov- ask and da-xat- paint, would be generated through the following steps:

Step 1: txov is directive-type verb, which selects an Action modality complement. Step 2: Action modality imposes time reference restriction on complement: [tc > (follows) tm].

K. Tuite: Vamling review 13-03-02 6

Step 3: Verb forms selected according to modality of complement, time reference of matrix, time reference restriction on complement: [tm < (precedes) T (utterance time)] = aorist or imperfect matrix verb; past-tense matrix + Action modality ([tc > tm]) = pluperfect complement. Step 4: Both verbs are of the direct verb class, therefore the aorist form of txov assigns ergative case to its subject, and the pluperfect form of da-xat- assigns its subject dative case. Step 5: Control: Directive-type verb also requires coreference of matrix object and complement subject.

Output:

ninom Nino[ergative] damexata

mtxova she-asked-it-of-me[aorist]

rom that

I-paint-her[pluperfect]

2. Assessment of the formal model in CIG. KVs representation of the lexical entries for Georgian verbs appears to be descriptively adequate, and represents, as far as I know, the first attempt to formulate a model which generates more than just case and agreement morphology. The model does need some refinement, and I believe the following components need to be corrected: 2.1. Verb classes. The classification of verb types employed by KV sets case-shifting verbs in one group (direct), while dividing the non-case-shifting verbs into two sets, depending on whether their subject receives nominative case (neutral verbs) or dative case (indirect verbs). The assumption here and it is one that many Kartvelologists have made until recently is that there are no indirect case-shifting verbs. Well, there are, and while the number is not high, it does include some of the complement-taking verbs mentioned in CIG, such as m-a-intereseb-s it interests me and m-a-pikreb-s it makes me think (p. 50). Tests for subjecthood (e.g. the

K. Tuite: Vamling review 13-03-02 7

binding of reciprocal pronouns) indicate that the formal direct object of these verbs is the semantic subject (Tuite 1987). 2.2. Form and function. According to the model, the verb form of an Action-type complement predicate is to be determined by its temporal reference, and in some cases that of the matrix clause (p. 71; see Table 2). Note, however, what Aronson & Kiziria (1989: Dialogue #5 note 32) refer to as the automatic use of the pluperfect after past-tense matrix verbs, regardless of their temporal reference. This occurs, for example, when the conditional is used, with present-tense meaning, to soften requests:

mindoda

tkvengan

zogierti

inpormacia information

I-would-want[conditional] youpl-from some mime!o. I-obtain-it[pluperfect] I would like to obtain some information from you

Contrast with the unsoftened request:

minda

tkvengan

zogierti

inpormacia information

I-want[present] youpl-from some mivi!o. I-obtain-it[optative]

I want to obtain some information from you

This indicates, I believe, that one should not always convert sequence-of-tense data directly into time lines. In cases such as the above, form follows form, not function, even when the determining form is employed with a marked time reference.

K. Tuite: Vamling review 13-03-02 8

3. Comparison of CIG with Hewitt (1987). The majority of works concerning complex sentences in Georgian and there are quite a few have been written in the Georgian language. The only other book devoted to this topic written in a West European language is The Typology of Subordination in Georgian and Abkhaz (1987) by the British linguist B. G. Hewitt. Since Hewitt (1987) and CIG are at present the only monographs on Georgian subordination accessible to most of the readers of this review, I will compare and contrast them briefly here. 3.1. Breadth and depth. Hewitts book is considerably broader in terms of the data it covers. In addition to complements in the strict sense (i.e. clauses functioning as subjects or objects), adverbial and adjectival subordinate clauses are discussed. Alongside facts from contemporary Georgian, examples collected from Old Georgian texts are considered, and, for purposes of typological comparison, data from the Northwest Caucasian language Abkhaz (which is unrelated to Georgian). Hewitt also draws parallels, from time to time, with Greek and Latin complex constructions. CIG, on the other hand, is a narrower and more focussed work, which attempts to exhaustively describe the facts concerning Modern Georgian complements, and tends not to stray far from its stated purpose. Only sporadic mention is made of the differences between Modern and Old Georgian complementation; the only languages from which data are drawn for comparison are English and, in one example, Swedish. Nor are adverbial purpose clauses, which are formally and semantically very similar to Action-type complements, discussed in CIG. 3.2. Approach to complementation. A reader interested specifically in complementation will not find the topic discussed in one place in Hewitt (1987). The section on noun-clause complements is in fact almost exclusively given over to Truth-modality clauses. Action-type complements are scarcely presented at all, because of their similarity to purpose clauses, which are analyzed elsewhere in the book (Hewitt 1987:251-252). Since inter-clausal coreference restrictions are almost exclusively confined to Action modality complements in Georgian, there is of course no discussion of control phenomena in Hewitts monograph.

K. Tuite: Vamling review 13-03-02 9

3.3. Assessment. The two books have very different goals, and each makes a useful contribution to Georgian studies in its own way. Hewitt (1987) is certainly more controversial, in that the author argues that many Georgian hypotactic constructions originated as paratactic syntagms, in particular, conjunctions of a verb of saying or cognition with a direct quote (Hewitt 1987:219, 246-251, 260-261). By contrast, the conclusions drawn in CIG concerning the correlations between semantic classes of Georgian verbs and morphosyntactic phenomena such as verb form and coreference do not go beyond what is clearly shown by the data, and the formal model proposed by KV represents these conclusions in a perspicuous manner. It is to be hoped that the extremely focussed nature of this book is an indication that the author intends to build, brick by brick as it were, a firm descriptive base from which in the course of time more wide-ranging conclusions will be drawn. 4. Comments on the text. The book is written in flawless English, and KVs presentation of the Georgian data and the exposition of her ideas is clear and easy to follow. I have noted a few mistakes and/or misprints, none of which, save one, presents a significant obstacle to readers unfamiliar with Georgian. The exception is on p. 125, figure 11, in which some of the components of the diagram are given in the wrong order. The following diagram (fig. 12) appears to be correct, however.

K. Tuite: Vamling review 13-03-02 10

10

TABLES:

Table 1: Georgian verb forms, grouped by 3rd-person tense/mood suffixes [V3sg/V3pl]

1 (PRESENT) [-s/-en], [-a/-an]

2 (PAST) [-a/-es], [-a/-nen]

3 (MODAL) [-s/-n], [-(e)s/-nen]

Series I

present future

imperfect conditional/ iterative

pres. conjunctive fut. conjunctive

Series II

(permansive)

aorist

optative

Series III

pres. perfect

pluperfect

(perf. conjunctive) > pluperfect

K. Tuite: Vamling review 13-03-02 11

11

Table 2: Chart of complement clause verb forms according to information modality and temporal reference.

[T = time of utterance; tm = temporal reference of matrix clause; tc = temporal reference of complement clause]

Constraint: tc > (follows) tm in Action clauses

tc

Action

Truth future

Future unmarked: optative marked: pluperfect*

Present

pluperfect

present

Past

pluperfect

unmarked: aorist marked: conditional** pluperfect*** pres. perfect****

*) optional if T > tm **) required when T > tc > tm ***) optional if T > tm > tc ****) adds resultative/evidential meaning

K. Tuite: Vamling review 13-03-02 12

12

Table 3: Partial lexical entry for complement-taking verbs in Georgian. 1. Modality of complement: Action and/or Truth. 2. Time reference restriction on complement (tc): <a> must follow matrix (tm), <b> must coincide, or <c> no restriction. [Generalization: if Action-type complement, tc > tm] 3. [For verbs taking Action-type complements]: Type of control: commitment, directive, attitude, other. 4. [For some classes of control verbs]: Constraint on reference of complement subject (disjoint or coreferent).

K. Tuite: Vamling review 13-03-02 13

13

REFERENCES:

Aronson, H. I. 1982. Georgian: A Reading Grammar. Columbus, Ohio: Slavica. Aronson, H. I. & Kiziria, D. 1989. Georgian: A Reading Grammar, Part II. Unpublished ms, University of Chicago. Harris, A. 1981. Georgian syntax: a study in relational grammar. New York: Cambridge University Press. Hewitt, B. G. 1987. The Typology of Subordination in Georgian and Abkhaz. Berlin: Mouton de Gruyter. Ransom, E. 1986. Complementation: Its Meaning and Form. Typological Studies in language, Vol. 10. Amsterdam: John Benjamins Publishing Company. Sag, I. & Pollard, C. 1988. A Semantic Theory of Obligatory Control. Unpublished ms. Tuite, K. 1987. Indirect Transitives in Georgian. Proceedings of the Berkeley Linguistics Society 13, 296-309.

NOTES:

It follows that the pluperfect can appear in complements of both modalities: in

Truth clauses in its ancient function of denoting past-anterior taxis, and in Action clauses in its newer role as past modal (pp 75-77). The other two modal verb forms in Table 1 the present and future conjunctive appear in certain types of irrealis clauses, but not in complements.
2

As KV notes (pp 102-4), not all complement-taking verbs allow non-finite

complements, and a few require their complements to be non-finite (p 113).

You might also like