Publishing Partner: Cambridge University Press CUP Extra Publisher Login

New from Cambridge University Press!


Revitalizing Endangered Languages

Edited by Justyna Olko & Julia Sallabank

Revitalizing Endangered Languages "This guidebook provides ideas and strategies, as well as some background, to help with the effective revitalization of endangered languages. It covers a broad scope of themes including effective planning, benefits, wellbeing, economic aspects, attitudes and ideologies."

New from Wiley!


We Have a New Site!

With the help of your donations we have been making good progress on designing and launching our new website! Check it out at!
***We are still in our beta stages for the new site--if you have any feedback, be sure to let us know at***

Review of  Endoclitics and the Origins of Udi Morphosyntax

Reviewer: Wolfgang Schulze
Book Title: Endoclitics and the Origins of Udi Morphosyntax
Book Author: Alice C. Harris
Publisher: Oxford University Press
Linguistic Field(s): Syntax
Issue Number: 13.2994

Discuss this Review
Help on Posting
1. General Information
The aim of Harris' monograph is to give a detailed account of a morphosyntactic (or: morphopragmatic) phenomenon that is said to be unique among languages, namely 'endoclitization'. By this term, Harris refers to the strategy to place so-called personal agreement clitics into the stem (or even root) of a verb. This phenomenon goes against standard assumption of the so-called Lexical Integrity Hypothesis according to which ''words are composed according to morphological principles that differ in kind from the syntactic principles responsible for the composition of sentences'' (Harris 2002:3). More concrete: ''[T]he morphological composition of a word is not accessible to the rules of syntax'' (ibid.). In case agreement clitics have syntactic and pragmatic properties, endoclitization of these clitics into a verb stem/root would violate this basic hypothesis. Therefore, the existence of endoclitics has often been denied (e.g. Klavans 1979).

Harris' monograph importantly challenges such views. In order to tell the General and Theoretical Linguist more about endoclitization together with its motivation by and its impact on syntactic structures, the author has chosen Udi as her sample language. Udi is a South East Caucasian (or: Lezgian) language that is currently spoken by roughly 3.000 people in now two villages (Nizh in Azerbaijan and Okt'omberi in Georgia). Until 1989, there has been another important Udi population in the village of Vartashen (Azerbaijan). Udi has two dialectal variant (Nizh and Vartashen). Until 2001, most linguistic descriptions and analyses of Udi have relied upon data from Vartashen (together with its variant spoken in Okt'omberi). Data from Nizh were scant. This picture has changed since the appearance of a collection of Nizh Udi poems and tales published by Kechaari 2001. In addition, Udi is documented by narrative texts(both native and translations from Russian), poems, and samples of the conversational style that had been recorded over the years 1850-2002. A translation of the Gospels has been prepared at the end of the 19th century (Bezhanov & Bezhanov 1902, Schulze 2001). Finally, fieldwork data have been collected by Adolph Dirr (1904), by Vladimir Panc^vidze (1960ies), by Alice Harris and by the author of the present review.

Typologically speaking, Udi is marked for a number of features that are alien to its sister (better: cousin) languages (such as Lezgi proper, Tabasaran, Aghul etc.). These features include the 'personalization' of the agreement system (instead of noun classification), massive presence of verb forms marked for incorporation, and the partial splitting of the relational primitives S (Subjective), A (Agentive), and O (Objective)(see Schulze 2000b
for details): Basically, Udi shows an ergative case paradigm. Nevertheless, S and A can be demoted to the 'Indirect Objective' domain (S/A > IO) with verba sentiendi and to encode a potential mood. On the other hand, S can be promoted to the Agentive function (S > A) to mark a strongly controlling referent in subjective function. The Objective is marked for one of the two Dative cases in case the referent is thought to be (textually) definite. In addition, Udi is characterized by clausal subordination that (in parts) replaces the East Caucasian tandard of participle and converbial subordination.

2. The book's purpose and contents
As has been said above, the main purpose of Harris' book is to provide evidence that endoclitization in a synchronically valid technique of syntactic organization. However, the book goes far beyond this synchronic issue: It aims at the explanation of why and how the endoclitization technique arose in Udi. This diachronic perspective is embedded into the general framework of Diachronic Syntax as presented in Harris & Campbell 1995. Doing so, Harris also exploits comparative evidence stemming from related languages in the Eastern Caucasus. In order to formulate the synchronic mechanisms of endoclitization, Harris makes special reference to Optimality Theory.

The book is organized as follows: A first introductory section (pp.3-19) states the basic problems dealt with in the monograph. Section Two (pp. 23-165) discusses the synchrony of the Udi agreement system both from a paradigmatic and a syntagmatic (functional) point of view. This section also includes an Optimality Approach to the phenomenon at issue (chapter 7). In section Three (which, in fact, represents the core of the book), Harris outlines a complex scenario of the emergence of endoclitization in Udi (pp. 169-284). The book concludes with a brief 'Afterword' (pp.283-4), a comprehensive bibliography and an index.

In sections One and Two, Harris introduces the grammar of Udi which makes the book more than just a comprehensive presentation of endoclitization in Udi: As the subtitle of the book suggests ('Origins of Udi Morphosyntax'), Harris is well aware of the fact that the Udi agreement system is at the core of the grammatical organization of the language: Agreement controls a wide range of syntactic and pragmatic properties (focus, verbal valence, referential tracking, 'subject' alignment etc.), just as it is controlled by such properties (functional cases, communicative and deictic reference, Tense/Mood system etc.). Consequently, Harris has to familiarize the reader with the major features of Udi grammar (including certain aspects of morphophonology), which makes the book also an introduction into the (functional) grammar of Udi as such.

In this sense, the first two chapters are compiled from a didactic perspective. This aspect comes also clear from the fact that here, interlinear glosses are given only for those forms that are under consideration. The further Harris progresses in her argumentation, the more
explicit the intelinear glosses become. The depiction of the Udi grammatical system relies on both standard grammars (there are at five such grammars ranging from Schiefner 1863 to Schulze 1982) and textual data (note Harris does not make use of the corpus presented by the Gospels. The new Nizh materials (Kechaari 2001) were not yet published by the time the author had finished her manuscript). In general, Harris confirms what has been said in the grammatical sources. There is, however, one major exception: None of the grammatical treatments of the Udi agreement system did account for the functional distribution of personal clitics: These can occur both with verbs and with extra-verbal constituents as in (1). [See bottom of file for special orthographic symbols.]

(1) (a) xina''r-en gölö s^um u-ne-k-sa
girl-ERG much bread:ABS eat-3sg-$-PRES
'The girl EATs much bread.'
(b) xina''r-en gölö s^um-ne uk-sa
girl-ERG much bread:ABS-3sg eat-PRES
'The girl eats MUCH BREAD.'
(c) go''lo'' s^um xina''r-en-ne uk-sa
much bread:ABS girl-ERG-3sg eat-PRES
'The GIRL eats much bread.'

(The symbol '$' indicates the second part of a verbal stem preceded by an endoclitic lement. Capital letters indicate focused constituents). Harris is the first who relates the formal distribution of these clitics to a functional scenario. Accordingly, the placement of agreement clitics is governed by both properties of the clausal information flow (constituent or sentence (prepositional) focus) and special features of the constituents. For instance, certain particles and pronouns that are in 'natural' focus (negation, adhortative, question) always call for a personal clitic. In case these particles again have clitic properties, 'piggybacking' can take place: I use this term to describe the fact that the resulting clitic cluster behaves as a single clitic (see Schulze (forthcoming) for a detailed account of the piggybacking process):

(2) (a) g^ar-en s^um-q'a-n uk-sa
boy-ERG bread:ABS-ADH-3sg eat-PRES
'The boy should eat BREAD.'
(b) g^ar-en s^um u-q'a-n-k-(e)sa
boy-ERG bread:ABS eat-ADH-3sg-$-PRES
'The boy SHOULD EAT bread.'

Three Tense/Mood categories always call for a clitic in enclitic position: The Factitive Future (labeled Future2 by Harris), the Modal (called Subjunctive by Harris), and the Imperative (usually derived from the Modal). Harris carefully analyses the distributional patterns in question and convincingly relates them to a set of (preference) rules that allow her to give an account of these 'rules' in the framework of Optimality Theory: ''[I]t is shown that this approach can account elegantly for the complex set of requirements and option for placement of the Udi PM [Personal markers, W.S.]'' (Harris 2002:7).

The fact that Harris succeeded in describing a set of functional conditions for the placement of Udi agreement clitics can be safely termed a 'linguistic discovery'. Harris has opened the door to a truly 'new' perspective for the description not only of the Udi system but also of other systems that, too, show floating agreement clitics (such as some Northwest Iranian languages, e.g. Northern Talysh, see Schulze 2000b). This perspective is characterized by the linkage of syntactic and pragmatic arguments that serve as a descriptive scenario for morphological facts. By 'focusing on focus', Harris shows that the clausal organization in Udi is heavily dominated by non-categorial, but pragmatic features that bounce back on nearly every grammatical 'category'. In addition, the pragmatic domain is also present in the ontology of 'words' in Udi: In Chapter 4, Harris gives an illuminating discussion of the degree of 'wordiness' of Udi verbs. She shows that clitization and prosodic features interact to produce incorporated verb forms. As typologically expected, this process is coupled
with the gradual dereferentialization of the host, compare:

(3) (a) xina''r-en as^-ne b-esa
girl-ERG work:ABS-3sg make-PRES
'The girl does a/the WORK.'
(b) xina''r-en as^-ne-b-sa
girl-ERG work-3sg-make>LV-PRES
'The girl WORKs.'

Harris shows that verb forms marked for incorporation structurally behave like simplex verbs: The same constraints on agreement clitics apply that are characteristic for simplex verbs. Nevertheless, Harris correctly observes that stem-internal endoclitization is blocked with incorporating verbs:

(4) (a) xina''r-en nana-xo xabar-re-aq'-sa
girl-ERG mother-ABL news-3sg-take-PRES
'The girl asks (lit.: takes news from)
(b) *xina''r-en nana-xo xabar-a-ne-q'-sa
girl-ERG mother-ABL news-take-3sg-$-PRES

The fact that incorporated elements represent the preferred host of agreement clitics with sentential focus illustrates that agreement is not lexically determined but conditioned by pragmatic factors: Incorporated elements represent the semantic (or: lexical) 'highlight' in complex verbal structures that then end in rather desemantisized 'light verbs' (LV) no longer accessible to endoclitics.

Both pragmatic and syntactic conditions have given rise to the fact that Udi agreement clitics in parts copy the relational properties of their 'personal' trigger: Accordingly, these clitics are 'bipolar': They identify their host as being in focus and relate it to the referential 'center' of a
clause which then is subcategorized according to the feature 'person' (three persons for both singular and plural). This 'identifying' property of the agreement clitics usually is organized in an accusative way (echoed referents are in subjective/agentive function). In case these referents are demoted to the 'indirect objective' function (with verba sentiendi), the clitic echoes this process in Vartashen: Here, the 'Dative' clitics are then used instead of the S/A-clitics, compare:

(5) (a) xina''r-a s^um a-t'u-k-sa
girl-DAT bread:ABS see-3sg:IO-$-PRES
'The girl sees a bread.'
(b) xina''r-en sa s^um be%-ne-g^-sa
girl-ERG one bread:ABS look=at-3sg-$-PRES
'The girl sees (looks at) the bread.'

Harris (pp.29) calls this constructional pattern 'Inversion'. In Sections 8.2 and 11.4, she illustrates the gradual adjustment of this pattern to the standard transitive pattern claiming that in Nizh, this process has today come to its end.

The assumed 'reformulation' of this constructional pattern relates to the diachrony of Udi morphosyntax (and: morphosemantics). In fact, Harris devotes nearly the totality of Section Three to the diachrony of Udi morphosyntax in order gain an explanatory basis for her analysis. Whereas Chapter Two takes a rule-based perspective, Chapter Three interprets the data in terms of dynamic or processual features.

The explanatory section is divided into five chapters: In Chapter 8, the author gives an outline of the morphological history of those forms that are involved in the make-up of Udi clauses: Case morphemes and agreement clitics. But note that the title of Chapter 8.2 ('Inherited Case Marking') is somewhat misleading: Harris does not talk about historical morphology but illustrates the degree to which the basic case marking patterns in Udi match those of the cognate languages. Chapters 8.3 and 8.4 take a more 'morphological' perspective: Harris argues that the whole set of Udi personal clitics ''developed from independent pronouns, and this is clearly correct, even though some problems remain'' (p.182). In fact, this claim that reflects standard assumptions on the origin of Udi agreement markers comes true for at least the 'oblique' (Dative and Genitive) of the clitics echoing speech act participants. In addition, the same provenience must be described for the 'first person' in general. However, the claim is not easy to support for the remaining clitics. Both phonetic and morphosyntactic problems heavily weigh upon this hypothesis which is based on a perhaps too 'universal' perspective (see below and Schulze (forthcoming)).

In addition to the standard clitics, Udi knows a distinct clitic to echo a questioned third person singular referent (Q-clitic). So far, this clitic has remained unexplained. Harris is the first to propose a diachronic model that is said to have produced this morpheme (pp.183-6): Accordingly, she interprets the clitic -a as a reflex of the Persian conjunction 'ya:' meaning 'or' used in yes/no-questions. As an analogon, Harris takes into consideration the German pattern: 'Sie bleiben hier, oder?' ('Are they staying here?' < 'They stay here, or?'). To this we can add the Turkish pattern 'Ahmet var ya' 'Ahmed is there, or (not)', occasionally used in the sense of yes/no-questions. However, it should be noted that neither the German nor the Turkish pattern represent morphologically marked 'questions'. Rather, we have to deal with shortened 'or' constructions that are marked for an additional prosodic pattern that produces the yes/no-question. The same is true for the rare instances, in which Udi 'ya' (~ 'ye')in
either/or-questions. Harris' analysis is based on the assumption that here, 'ya' lost its initial element 'y-' when following a constituent that ended in '-i'. In a second step, the resulting element '-a' would have been extended to first yes/no-questions, and later to WH-questions. Note that in contemporary Udi, '-a' is (longer) used with yes/no-questions. Although Harris' proposal is rather attractive, it is difficult to support both from a functional point of view and from the diachronics of Udi. For instance, the reduction of 'ya' to '-a' presupposes that -i-final constituents were frequent enough to initiate this process. Although it has been often observed that less frequent paradigmatic types can induce reanalysis and extension, we have nevertheless to bear in mind that out of a lexical corpus of 3.856 Udi words liable to host the 'clitic' '-ya', only 104 are marked by final '-i' (= 2.7 %). In addition, Harris' proposal does not explain why the Q-clitic 'replaces' the standard third person singular clitic, as in:

(6) (a) xina''r-en s^um-ne uk-sa
girl-ERG bread:ABS-3sg eat-PRES
'The girl eats BREAD.'
(b) xina''r-en ek'a-a uk-sa?
girl_ERG what:ABS-3sg:Q eat-PRES
'WHAT does the girl eat?'

Obviously, both clitics stand in complementary distribution (except for the fact that '-a' cannot occur as an endoclitic). This distribution suggests that both element ('-ne' and '-a') have a common categorial background. This hypothesis allows relating the two clitics to two different focal strategies in Proto-Lezgian (as they have, for instance, survived in Tsakhur, see Kibrik 1999). Accordingly, '-ne' < '*-ni' would have been used in so-called 'knowledge-based' (or: cognitive) focal contexts, where '-a' indicated a 'verificational' (or: indexal) focus (see Schulze (forthcoming) for details).

Chapters 9-12 concern the origin of the agreement pattern in Udi. In chapter 9, Harris relates the endoclitic technique to the history of Udi verbal stem formation. Harris carefully discusses possible stem types in Early Udi and in Proto-Lezgian. Here, she refers to the standard hypothesis that many of the Udi simplex verbs are marked for so-called petrified class markers: Accordingly, Udi once knew an agreement system that was based on the semantic subcategorization of a referential noun in subjective/objective function. In Udi, this technique is completely lost. Nevertheless, certain verbs such as 'bak-' 'to be(come)' probably show traces of this paradigm: Here, the first element '-b-' is seen as a reflex of the class marking strategy ('b-' = Class III (basically (grow-up) non-human animates and socially/culturally relevant objects). In addition, a verb stem could be marked by one or two local preverbs (see Harris 2002:197,218). In order to account for the 'endoclitic' slot in Udi verb stems, Harris develops four hypotheses: a) The paradigm of agreement clitics developed in situ: She dismisses this hypothesis, because she assumes that the Udi clitics ''have developed from independent personal pronouns'' (p.211). b) 'Trapping': According to this hypothesis, endoclisis would have resulted from the univerbation of formally distinct lexical structures (incorporated element + verb). In order to account for root endoclisis, Harris refers to a third hypothesis: c) Simple movement of the 'Person Markers'. Here, it is claimed: ''Intramorphemic positions developed as a result of the intermorphemic positions which has come about through univerbation'' (p.212-3). d) The fourth hypothesis describes ''person markers as the 'slot holder' of Proto-Lezgian C[lass] M[arker]s'' (p.213). According to this hypothesis, Udi agreement clitics would have taken over the position of the former class
markers that already occurred in endoclisis. Pp. 215-222, Harris compares the last three hypotheses by referring to the individual history of a number of Udi simplex verbs stems. She concludes that all three hypotheses ''play a role in the explanation of the origins of endoclisis in Udi''.

The chapters 10 to 12 interpret the emergence of the Udi agreement technique in terms of both a formal and a functional diachrony. Harris argues that constituent focus stems from older clefting strategies residues of which are said to be found in Udi sources of the 19th century. To explain this point, let me quote an example from Harris (p.237-240):

(7) xunc^i-mug^-on xorag-ax-q'un ha''zir-b-esa
sister-PL-ERG food-DAT2-3pl prepare-DO-PRES
'The sisters are preparing the FOOD.'

According to the Cleft Hypothesis, such a construction would have resulted from the following pattern (note that here, the past tense is used by Harris in order not to complicate the matter):

(8) *xorag BE [no xunc^i-mug^-on ha''zir-b-i]
food:ABS COP it:ABS sister-pl-ERG ready-do-PAST
'It is FOOD that the sister are (read: were) preparing.'

The structure in (8) differs from that in (7) in that the 'agreement marker' (originally an anaphoric pronoun) copied the clefted constituent in just the case form that is expected by the verb in the dependent clause (Objective > Absolutive). Harris argues that the ''process of reanalysis (here of (8), W.S.) (...) must have consisted of the reinterpretation of the biclausal
cleft as a monoclausal structure'' (p.240). In consequence, ''the case of the FocC (= Focused Constituent, W.S.) changed from absolutive to that determined by its grammatical relation in the monoclausal structure.... [T]he pronoun/PM changed from agreeing with the FocC to agreeing with the subject'' (p.240-1). Harris extensively dwells upon this rather problematic
hypothesis that perhaps is too strongly oriented towards more general assumptions on the developments of Clefts in the languages of the world. Contrary to the preceding section of the origins of endoclitization, Harris does not consider alternative proposals to explain the focal nature of agreement clitics in Udi. This fact renders Chapter 10 somewhat suggestive.
Readers familiar with Cleft typologies will probably happily refer to this chapter in order to draw more general conclusions. However, they are deprived from possible alternative perspectives which would direct their generalizations to another road (see below).

Nevertheless, it must be admitted that once Harris has taken her position, which is well formulated, theoretically well-grounded, and empirically supported by well-chosen examples, the analysis ends in a (by itself) coherent and (by itself) convincing scenario. It ends in the explanation of the positional constraints on Udi agreement clitics (Chapter 12). Here,
Harris pays special attention to those Tense/Mood forms that necessarily call for a clitic and hence disallow constituent focus. The author does not relate these constraints to a single cause but argues that different functional and morphological processes have led to the 'same' result. Most importantly, Harris is the first to suggest an explanation for the fact that
the modal verb forms ('subjunctive' in her terms) are always followed by the agreement clitics. Accordingly, the modal forms stem from the reanalysis of sequences marked by a postponed clitic cluster (adhortative particle 'q'a-' + clitic). In sum, Harris arrives to describe the diachrony of all positional constraints and preferences.

All claims and arguments put forward by Harris are easy to read and to follow. In fact, the book is well organized and full of summarizing paragraphs that allow the reader to check whether (s)he has fully understood the by itself rather complicated matter. The main advantage of the book is that it (also) addresses an audience that is not familiar with East Caucasian linguistics. The careful (nearly pedagogical) way of familiarizing the reader with Udi linguistics makes the book a pleasure to read. It appeals to the analytic interest of reader and to his/her readiness to re-enact proposals to solve the puzzle of Udi morphosyntax and
morphopragmatics. Harris not only tells the thrilling story of how Udi morphosyntax may have emerged, but also constantly helps the reader to locate the analyses in more general theories of language function and language change.

3. Critique
'Endoclitics' is said to ''appeal to theoretical linguists, especially those interested in the interface between syntax and morphology. It will also be of considerable interest to historical linguists and students of Caucasian languages'' (from the cover of the book). This quote illustrates the three basic perspectives the author has taken. It is quite natural that specialists in either of these perspectives will look differently at what Harris' analysis is built upon. In my remarks, I will take the perspective of both a Caucasianist and a Typologist and will leave the debate on whether the Optimality Theory perspective taken by Harris in Chapter 7 is appropriate or not to people more qualified than I am. Nevertheless, it should be noted that arris' theoretical argumentation heavily relies on the correctness of the Udi data that represent the bulk of the empirical background of the book. As has been said above, the corpus exploited by Harris does not represent the totality of what is currently available for
Udi. Crucially, Harris does not take into consideration the Udi Gospels (Bezhanov & Bezhanov 1902). In fact, the Gospels represent more than the half of all Udi data. Although the Gospels are translated from Russian and thus have to be taken with great care, a closer look reveals that the morphosyntax of the Gospels (not necessarily its syntax) comes close to what Udi has been at the end of the 19th century. On the other hand, Harris heavily relies on the texts edited by Schiefner 1863. Most of these texts, however, have a rather obscure history. Again, the bulk of these texts is translated from Russian (and, as for the dialogs even from German (!), it seems). Contrary to what can be said for the Gospels, we cannot safely
describe the degre of authenticity of these texts. As Dirr (1904:v) says, Schiefner's work has to be referred to with great caution. More concrete: ''The texts are neither Udi nor Russian from which they are translated. They resemble so few to the Udi language that I could not continue working with them with my Udi teacher ..., a native from Vartashen. Frequently, he did not understand (the texts) and asked me no longer to bother him with these text'' (Dirr 1904:viii; translation W.S.). This fact is crucial because Harris assumes that Schiefner's texts ''represent a slightly earlier form of the language, with diachronic change accounting for the difference'' (p.134). Here, it would have been good if Harris had taken a more critical position. Only if we have additional material that stems from other authors of the same peiod we can judge whether Schiefner's Udi actually reflects 'true' Udi.

The fact that Harris did not consult the Gospels (which can still be processed by contemporary Udi speakers from Vartashen) has conditioned that the author sometimes arrives at problematic generalizations. For instance, she postulates a set of monoconsonantal verbs that do not allow endoclitization. Among others, Harris refers to the verb 'b-esun' 'to do, make' (stem 'b-'). On p.219, she claims that endoclitization does not occur with this verb. However, the Gospels nicely show examples like

(9) be-z-sa
'I do'
'(S)he did^Å'

Such forms are rejected by Harris (p.219). However, Matthew 26:10, Mark 6:14, John 8:41; 7:3, 10:38 clear evince the possibility to use endoclitics with the verb 'besun'. Also note 'be-q'un-sa' (do-3pl-$:PRES) 'they do' in the native tale Rust'am (1888). Accordingly, 'besun' is not a monoconsonantal verb, but reflects an older stem '*be-_-'-' (-_- is used to indicate the endoclitic slot) that again is derived from a root '*-e^'a-' (preceded by the petrified class marker *b-).

Some of the reconstructions proposed by Harris importantly affect her general analysis. This holds especially for the origin of the clitics themselves. As has been said above, Harris takes the position that the Udi clitics stem from independent (personal or deictic) pronouns. This
hypothesis gives her the clue to establish the Cleft Hypothesis. Without alluding to the problems raised by this hypothesis itself, it must nevertheless be said that the proposal has so many phonetic and functional shortcomings that it is difficult to subscribe to it any longer. Rather, we should think of an interplay between older focal strategies based on constituent focus (Proto-Lezgian '*-ni' ~ '-a') and the gradual development of personal paradigms that started with the first person (a process that is typical for a number of other Lezgian languages). This assumption allows proposing an alternative scenario that does not make use of the Cleft Hypothesis. A simple example is:

(10) xina''r-en s^um-ne uk-sa
girl-ERG bread:ABS-3sg eat-PRES
'The girl eats BREAD.'
< *xina''r-en s^um-ni uk-sa
girl-ERG bread-FOC eat-PRES

According to this hypothesis, the 'local' focus marker '*-ni' once had been used with all persons (or: impersonally). In 'egocentric' contexts (involving a first person), it became replaced by the first person pronoun whereas in the second person, the clitic '*-ni' was (later) accommodated to the phonetic shape of the corresponding pronouns (see Schulze forthcoming for details).

I have elaborated this point in order to show that Harris' Cleft Hypothesis takes a perhaps too narrow perspective. The critique of other hypotheses put forward by Harris is perhaps less relevant for the evaluation of the whole scenario described by the author. Nevertheless, the reader is sometimes left with the impression, that Harris has unnecessarily complicated the matter: For instance, Harris has to describe a rather idiosyncratic sound change (*-i-q'a- > *-i-a > -ai) to account for the constraint on the modal ('subjunctive') (see above). In a second step, she has to describe a process of reanalysis (> -a-_-i) to arrive at the actual paradigm of the Udi modal. Neither the sound change, nor the metathesis can be safely described for
Udi. In addition, the assumed process of reanalysis is without parallels in Udi. In fact, it is much more simple to assume that the Udi 'Past Modal' (marked by -ai and followed by agreement clitics) once represented a modal form for its one (< Conditional), which later (in parts) merged with the past variant of the standard modal in -a (which itself is taken from the
Imperative). This analysis refers to the functional (or: categorial) cluster 'Epistemic < Deontic' which is crucial not only for Udi but also from the point of view of a general theory of Modality.

Finally, it should be added that Harris rarely refers to language contact as a possible clue to understand the morphosyntax of Udi. For instance, it is out of question that the paradigm of personal clitics has been both formally and functionally influenced especially by Northwest Iranian languages, but also by Old Armenian, by Georgian, and, last but not least, by the local
varieties of Azeri. The same probably holds for the emergence of Fluid-O structures (also know as 'Differentiated Object Marking', DOM) the understanding of which is crucial for the discussion of focus, as Harris has convincingly shown herself.

Nevertheless, what Harris tells us is currently one of the best (and most straightforward) proposals we have at our disposal to approach the typology of endoclitization in Udi. But we should be aware of the possibility that progress in Comparative Lezgian (and Udi) linguistics may arrive at a partially or totally different picture of Early Udi morphosyntax. In this respect, we should also bear in mind that the recently discovered palympsest from Mt. Sinai that most likely contains a variant of Early Udi (5th - 7th century) will probably tell us more about the architecture of Early Udi, once the palympsest has been read (see Aleksidze/Mahé 1997, 2001, Aleksidze 1998-2000). It may well be that the language of the palympsest confirms Harris' diachronic assumptions. But it may likewise be the case that the contrary is true. Hence, Harris' should taken as what it is: A remarkable and highly professional study in the morphosyntax of Udi that reflects our knowledge of this language at the turn of the century.

The book itself is well done from a formal point of view. The bibliography refers the reader to most of the relevant literature; an index helps him/her to spot points of interest in the text. Unfortunately, the book contains a number of typographical errors that, however, normally do not affect the understanding of Harris' argumentation. In sum, we have to praise the author for having undertaken the enterprise to approach the functional and formal scope of agreement clitics from both a synchronic and a diachronic perspective. This book will surely help to make typologists and theoreticians more familiar with this language, which - as shown by Harris - challenges some of the generalizations current in contemporary linguistics.

4. Bibliography
Aleksidze, Zaza & Jean-Pierre Mahé 1997. Découverte d'un texte albanien: une langue ancienne du Caucase retrouvée. CRAI 1997:517-532.

Aleksidze, Zaza 1998-2000. Remarques sur le déchriffrement de l'écriture albano-caucasienne. RArm 27:423-428.

Aleksidze, Zaza & Jean-Pierre Mahé 2001. Le déchiffrement de l'écriture des Albaniens du Caucase. CRAI 2001 [in press].

Bezhanov, Semon & Bezhanov, Mikhail 1902. Gospoda Nas^ego Iisusa Xrista Svajtoe E.vangelie ot Matfeja, Marka, Luki i Ioanna. Tiflis: Typ. Kancelarii Glav. 1902 (Sbornik Material djla Opisanija Plemen i Mestnostej Kavkaza XXX).

Harris, Alice C. & Lyle Campbell 1995. Historical Syntax in Cross-Linguistics Perspective. Cambridge: Cambridge University Press.

Kechaari, Georgi 2001. Orayin. Baki^: Aze''rbaican Do''vle''t Ne''s^riyyati^.

Kibrik, Alexandr E. (red.) 1999. E.lementy caxurskogo jazyka v tipologic^eskom ocves^c^enii. Moskva: Nasledie.

Klavans, Judith L. 1979. 'On Clitics as Words'. CLS parasession volume, 68-80.

Schiefner, Anton 1863. Versuch über die Sprache der Uden. Mémoires de l'Akadémie impériale de sciences de St.-Pétersbourg, VII série, vol. vi., no. 8. St. Peterburg: Kaiserliche Akademie der Wissenschaften.

Schulze, Wolfgang 1982. Die Sprache der Uden in Nord-Azerbaidzhan. Studien zur Synchronie und Diachronie einer süd-ostkaukasischen Sprache. Wiesbaden: Harrassowitz.

Schulze, Wolfgang 2000a. The Accusative Ergative Continuum. General Linguistics 37:71-155.

Schulze, Wolfgang 2000b. Northern Talysh. Munich: Lincom Europa.

Schulze, Wolfgang 2001. The Udi Gospels. Annotated Text, Etymological Index, Lemmatized Concordance. Munich: Lincom Europa.

Schulze, Wolfgang (forthcoming). A Functional Grammar of Udi (to appear 2003).

Symbols used in this review:
g^ = Voiced Uvular Fricative
s^ = Alveopalatal Voiceless Fricative
a'' = Palatal /a/
o'' = Palatal /o/
e^ = Schwa
e. = Russian 'inverted' /e/
i^ = /i/ without a dot (Turkish)
% = Pharyngealized Vowel
$ = Second part of a discontinuous lexeme
-_- = Endoclitic Slot
Prof. Dr. Wolfgang Schulze is the Head of the Institute for General Linguistics and Language Typology at the University of Munich. His
research interests are Language Typology, Comparative Linguistics, Cognitive Typology, and Language Theory. He has done field work in a number of East Caucasian languages and has published (among others) both articles and books on Udi as well as on other (East) Caucasian languages. In addition, he has developed the framework of a 'Grammar of Scenes and Scenarios' (GSS) that aims at laying the ground for a Theory of Cognitive Typology.