Harris part

At its core, this is a book about what a word is. Many linguists have assumed thecorrectness of the Lexical Integrity Hypothesis, the hypothesis that words arecomposed according to morphological principles that differ in kind from thesyntactic principles responsible for the composition of sentences. According tothis hypothesis, the morphological composition of a word is not accessible to therules of syntax, and items manipulated by syntactic rules (i.e. words) cannot bemanipulated by the morphology (see, for example Di Sciullo and Williams ,Bresnan and Mchombo ).
The facts of Udi challenge these basic assumptions. In this language, person markers (PMs) may occur between the bound morphemes of a verb, as for ex-ample the third person singular PM -ne- in as#-ne-b-sa [work-SG-DO-PRES] ‘sheworks’, which may not be so surprising, but also inside a monomorphemic verb,as in e-ne-f-sa ‘she keeps (it)’, where -ne- splits the unanalyzable morpheme ef-‘keep’. The problem is that the rules that position -ne- and other PMs must be inpart syntactic rules, given that PMs may occur on words outside the verb, as in ().
(In (a) -ne is attached to the noun yaq’ ‘road’ used adverbially, while in (b) it isattached to the noun ait, the direct object of the sentence.) road-DAT-SG in-LV-PRES
‘On the road he opens it.’
‘Merab keeps his word.’
But if the rules are syntactic, the Lexical Integrity Hypothesis claims that they donot have access to the internal structure of a word and therefore cannot positionthe PM inside the verb, as in the examples as#-ne-b-sa ‘she works’ and e-ne-f-sa‘she keeps (it)’, cited above. Put differently, in Udi we find endoclitics both inter-morphemically (between the morphemes that compose a complex verb word)and intramorphemically (directly inside a single morpheme). Udi PMs havemany properties traditionally associated with affixes, such as marking agree-ment between subject and verb. Yet unlike most familiar affixes, their position in  The notion that these morphemes mark ‘agreement’ is adopted from more traditional grammars uncritically until §., where the notion is discussed. Use of the term is not intended to imply (a) that Endoclitics and the Origins of Udi Morphosyntax the verb is anything but fixed. If linguists are to understand the limits on whatconstitutes a word, we must come to grips with the data from Udi.
1.1. Problems to Be Addressed and Goals
The fundamental problem that motivated this research is the placement of PMs
in Udi. Although it has long been recognized that the PM can occur in various
positions, it has not been possible until now to predict where the PM will occur
in a given clause. The PM may be word-final in the verb, as in bos#t’-al-le [plant-
FUT-SG] ‘she plants it’, or it may occur between morphemes in the verb, as in box-
ne-sa [boil-SG-PRES] ‘it boils’. In some instances the PM may occur inside the verb
root, as in a-ne-q’-sa ‘she takes, receives’, where the morpheme aq’ is split by the
PM -ne- [SG]. In other examples the PM attaches to a word other than the verb,
as in () and ().
äyel-ax-ne
child-DAT-SG bathed
‘The mother bathed the child.’
Note that the fact that linguists have never specified the rules for positioning PMsmeans that it is impossible to teach someone how to form even one grammaticalsentence of the language. The fact that there are at least eight published full-scalegrammars or grammatical sketches of the language (Schiefner , Dirr ,Bouda , Jeiranis#vili , Gukasjan , Panc#viZe , Schulze , andSchulze-Fürhoff b), as well as numerous scholarly articles about it, none ofwhich has stated the position of the PM precisely, indicates how elusive the goalof specifying the position of the PM has proven to be. Thus my first goal is todescribe the position in which PMs occur under specific grammatical conditions;this is partly accomplished in Chapter , where it is shown that PMs are hosted byfocused constituents. The description of the positions in which PMs occur iscompleted in Chapter .
The theoretical context in which the position of Udi clitics becomes important is the claim by Klavans (, ) that endoclitics, clitics that occur inside words,cannot exist. In fact, almost all linguists believe that it is impossible for clitics tooccur inside roots. Although it has been suggested that there are endoclitics inEuropean Portuguese, Crysmann () has shown that the morphemes at issueare not clitics according to the criteria established by Zwicky and Pullum (). Ithas also been claimed that endoclitics exist in Pashto, but Kaisse (, : –) these morphemes ‘belong to’ the verb or to imply (b) any particular analysis of the status of thesemorphemes as clitics vs. affixes. A position on the first issue, what word hosts the agreement markers,follows as a consequence of the discussion in Chapter ; no assumption is made independently of thatanalysis. The second issue is discussed in Chapter . The assumption that these markers are examplesof agreement, rather than constituting arguments, is consistent with the fact that they cooccur withcoreferential independent full noun-phrase or pronominal arguments.
showed that the clitics at issue always follow a morpheme; they do not occurinside a morpheme. (See further discussion in Chapter .) A related goal of this book is to describe the diachronic origins of each of the positions of PMs, explaining how the use of that position came to be part of thegrammar; this is completed in Chapter , and most of the chapters leading up toit contribute to the same goal. In §. it is shown that Proto-Lezgian, thesubgroup proto-language from which Udi descended, had gender-class agree-ment; and later in that chapter we see that the PMs of Udi are a relatively recentinnovation. The newness of these markers makes it especially puzzling that theywould occur in the interior positions described in the preceding paragraph. InChapter  I consider the hypothesis that PMs were ‘trapped’ in an interior pos-ition when two independent words were reanalyzed as a complex verb, as well asthe hypothesis that the position of the innovative PMs represents an ancientposition inside verbs that can be reconstructed to PL or even to Proto-North EastCaucasian (PNEC). It is especially problematic to find morphemes that areclearly innovations inside a word, since there is good reason to believe thatmorphemes of this type, if they move at all diachronically, move to the periph-ery of the word (Haspelmath a).
Part of the explanation of the origins of PM positions lies in the reconstruc- tion of the focus cleft. Some of the other languages of the North East Caucasian(NEC) family have productive focus clefts or monoclausal focus constructionsderived from them. Dargi retains both, as illustrated in ().
SG-ERG brothers.ABSL bring.PAST-SG‘You brought the brothers.’ SG.ABSL FOC[AUX-SG] brothers.ABSL bring-PTCPL.SG
‘You brought the brothers.’
SG-ERG FOC[AUX] brothers bring-PTCPL.SG
‘You brought the brothers.’
(a) is a simple sentence with no focus marked. In (b, c), focus is placed on ‘you’.
In Chapter  I argue that (b) is a cleft, with the first two words, x’o saj-ri ‘youare’ forming the matrix clause, and that (c) is monoclausal. The sentence is thuscomparable to English It was you who brought the brothers. In Chapter  focusclefts are tentatively reconstructed for Pre-Udi.
 If I understand Kazenin () correctly, he argues that sentences comparable to (b) are not clefts because the copula determines the case of the focused constituent (absolutive). He seems toclaim that this is not characteristic of clefts. However, that would be incorrect; in unreanalyzed clefts,the focused constituent is subject of the copula, and this grammatical relation determines its case.
Being uncertain of whether Kazenin views this Dargi construction as a cleft, I have provided my ownanalysis.
Endoclitics and the Origins of Udi Morphosyntax Additional goals of this book include presentation of material relevant to other current theoretical debates in synchronic morphosyntax. First, in Chapter  it isargued that PMs in Udi are clitics, not affixes, according to criteria set out inZwicky and Pullum () and in other works. Klavans ( and ) claims thatendoclitics (clitics that divide a lexical item) cannot exist, but Jeiranis#vili (: )and Panc#viZe (: –) claim that Udi PMs do divide simplex verbs in ex-amples such as those in (), where the stems buq’, c#uk, and aq’ are consideredunanalyzable single morphemes.
() a. bu-za-q’-sa
c. c#u-ne-k-sa
Synchronically the elements glossed as a verb with the subscripts <> and <> are parts of a single morpheme, made discontinuous by the imposition of the PMinside it. However, it is possible that (a) could be explained historically as consist-ing of a fossilized gender-class marker (b(u)-) with a root q’. In a number of verbsin Udi, b- occurs as a fossil of the agreement system inherited from Proto-NEC andProto-Lezgian, in which a prefix marked the grammatical gender-class of the abso-lutive nominal (subject of an intransitive or direct object) (Jeiranis#vili ).
Generally a vowel occurs between the gender-class marker (CM) and the verb rootunder circumstances that are not yet fully understood. Thus b- and u- in (a) werein all likelihood historically a morpheme (or two morphemes) distinct from (theancestor of) the root -q’ ‘love, want’ (see also Nikolayev and Starostin : ,Schulze : ). A related explanation may apply to (b); it could consist of the(different) historical root q’ ‘take’, together with a (different) vowel belonginghistorically to the ancient gender agreement system. However, there appears to beno comparable explanation for (c) and a handful of similar forms. It appears thatneither c#u- nor -k can be related historically to other morphemes. Thus, the follow-ing issues are addressed. Are PMs in Udi clitics or affixes? Do c#uk ‘break’ and simi-lar stems consist of a single morpheme (an unanalyzable base), or does a closerinspection reveal that they are etymologically two (or more) morphemes? If this isa single morpheme, why do PMs split it, making it discontinuous? Apects of thestructure of the verb relevant to this issue are presented in Chapter , and the pos-ition of the PM inside the verbstem is described in Chapter .
 I use ‘verbstem’ here as a technical term in distinction to ‘(verb) stem’. While ‘stem’ is used in different ways by different authors, in relation to a verb it most often means a base (usually more A second synchronic issue dealt with is the theoretical account of the positions of clitics in Udi. In Chapter  an account in the framework of Optimality Theoryis proposed, and it is shown that this approach can account elegantly for thecomplex set of requirements and options for placement of the Udi PM.
Although not taken up as a theoretical issue, the structure of complex verbs also figures in an important way in this work (described primarily in Chapter ,and here and there in Chapter ). Udi provides evidence relative to causatives,potentials, and incorporation of nouns and other elements. It distinguishes truecomplex verbs from lexicalized phrases, which are so similar that they require theuse of specific diagnostics.
The final major goal is presentation of data and discussion of issues relevant to a theory of diachronic morphosyntax. Harris and Campbell (, chapter )make specific predictions relative to the simplification of biclausal structures.
In Chapter  of the present work it is suggested that Udi clefts were simplifiedas monoclausal focus constructions. Comparative data from other NEClanguages were first presented by Kazenin (), when Harris and Campbell() was already in press. Thus the NEC data could not be taken into accountin that work, and they provide data relevant to the hypotheses formulatedthere.
Second, changes in case-marking patterns result in part from changes in the structure of complex verbs. Udi inherited the ergative–absolutive case markingcharacteristic of NEC languages, with subjects of transitive verbs in the ergativecase, and direct objects and subjects of intransitives in the absolutive case. Udiretains that system to an extent, but it has introduced two changes not shared byits sister languages. One of these differences is that in Udi, while direct objects maybe marked with the absolutive, according to the inherited system, they may alter-natively be marked with the dative, according to an innovative system. This isillustrated in () and discussed at greater length in Chapter , where it is shownthat the dative case marking of the direct object has developed the function ofmarking definiteness. (In the remainder of this chapter, Udi verbs are segmentedand glossed only as relevant to the point at hand.) Dative direct objects may precede or follow the verb, and they appear to have the same privileges of occurrence that absolutive direct objects have (but see complex than a root) to which additional affixes are added. I also refer to a ‘present stem’, ‘aorist stem’etc. in this sense (see especially §.). The verbstem contrasts with this and is described more preciselyin Chapter .
Endoclitics and the Origins of Udi Morphosyntax further below). How did they arise diachronically in Udi? In Chapter  I showthat they developed as a ‘second direct object’, illustrated in ().
‘Finally they make a guest of this woman.’ ‘. . . receive this woman as a guest.’ As shown there, q’onag ‘guest’ was historically the direct object of b- ‘make, do’, andme c#ubg-ox ‘this woman-DAT’ was an oblique complement expressed in the dative(roughly ‘make a guest of this woman’). When q’onag-b- was reanalyzed as aunitary verb, it was natural to interpret as its direct object the nominal representedin this example by me c#ubg-ox ‘this woman-DAT’. The use of dative marking fordirect objects was then extended, beside absolutively marked direct objects.
Therefore, variable use of dative or absolutive for direct objects in other construc-tions is in part explained diachronically through the ‘second direct object’construction, though we will see that contact is likely to have played a role as well.
A second change Udi has introduced in the inherited case marking system involves the use of the historical ergative case with verbs that are, loosely speak-ing, active intransitives (unergatives). This change may have been influenced byneighboring Georgian, which also uses the historical ergative case with verbs of asimilar semantic type in certain tenses; however, it is possible to identify an inter-nal mechanism of change, whether or not there was Georgian influence. It isnatural to hypothesize that this phenomenon is related to incorporation, sincemany active intransitive verbs in Udi are of the incorporated type illustrated in(), using the same simplex verbs used to form transitive verbs, while most in-active intransitives use the intransitive formants -bak ‘be, become’ or -c’ (poly-semous), as in ().
The structure of () was historically something like ‘the child says a crying’ and itwas natural for the clause to be treated as transitive, and thus for the subject to bemarked with the ergative case. From this sort of construction developed ergativecase marking with a restricted set of intransitive verbs.
The third issue in the theory of historical syntax to be taken up here is the use of Internal Reconstruction and the Comparative Method in syntax. In Chapters  The verb ‘say’ is easily identified synchronically, and its meaning in complex verbstems should not be taken literally. This is discussed in detail in §....
–, the use of these methods in the reconstruction of the syntax of the NorthEast Caucasian family is illustrated, according to principles discussed in Harrisand Campbell (). In §. cognate sentences, a phenomenon disputed byLightfoot (: ), are used in comparative reconstruction. In §§.–, thepronominal origins of the PMs are identified, as well as the origin of one specialPM, -a. In Chapter  I reconstruct the structure of the PL verb, and in Chapter the Pre-Udi focus cleft.
Lastly, my goals also include relating these diachronic phenomena in NEC languages to the universals of language change; they thus include further an ex-planation of how endoclitics (word-internal clitics) originate and an explanationof how an ergative–absolutive case system could develop intransitive verbs withergative case subjects.
Udi was selected as the basis for a study in diachronic morphology and syntax because (a) the structures of the Daghestanian languages are different in manyways from those of more familiar languages and therefore can contribute signifi-cantly to our understanding of language universals, (b) Udi has a complexmorphology, which supplies ready evidence of syntactic relations, and (c) Udi hasundergone significant morphosyntactic change.
1.2. What is Udi?
Udi is a language of the Lezgian subgroup of the North East Caucasian languagefamily. It is characterized by a number of interesting features, including some thatplay little or no role in this book. For example, its phonetic inventory containsglottalized consonants, [p’, t’, k’, q’, c’, c#’], and pharyngealized vowels, [i., e¢, a¢, o¢, u¢](see Fähnrich ).
Typologically, the language is agglutinative. In the verb, tense–aspect–mood (TAM) is marked by a set of suffixes, while person and number are marked by theclitics that are the main topic of this work. (Person and number are not markedby separate morphemes, as they are in the most highly agglutinative languages.)In aq’-al-zu ‘I will take, receive’, the future is marked by the suffix -al and the firstperson singular subject by the enclitic -zu. In the noun, case and number aremarked by distinct suffixes: äyel-ug-on ‘child-PL-ERG’.
Like other languages in its family, Udi has so-called dual-base noun declension.
This term indicates a declension in which (some) nouns have one base for theabsolutive and ergative cases and a different base, derived from the ergative, forother cases. Table . provides an illustration of the singular. Dual-base declensionis somewhat different for certain other nouns.
Some Udi verbs, including verbs of motion, are characterized by directional preverbs. Only a few verbs have a monomorphemic base, such as aq’- ‘take,receive’. The majority of verbs in the language are composed of an incorporatedelement and a light verb, e.g. as#-b- ‘work’ (literally, ‘work-DO’). The structure ofthe verb is treated in detail in Chapter .
Endoclitics and the Origins of Udi Morphosyntax TABLE .. Dual-base declension in Udi(Jeiranis#vili :  and SixaruliZe : ) The inherited case marking is ergative, with subjects of transitive verbs in theergative case, and with subjects of intransitive verbs and direct objects in the abso-lutive case. However, definite direct objects are marked with the dative case, whichis otherwise used for indirect objects and certain locative functions. Theseconstructions are illustrated and described in detail in §.. Subjects of certainverbs that appear to be intransitive may be in the ergative case; the historicalexplanation of this is discussed in Chapter .
Like other languages of the North East Caucasian family, Udi makes extensive use of non-finite verb forms, including participles, infinitives, masdars, andconverbs. A frequent construction with a participle is illustrated in ().
‘The king’s wife, having gone, enters the village.’ In this construction, the subject of the first verb, the participle, is also subject ofthe second; the participle is also used in other constructions.
Word order in Udi is generally SOV; this is illustrated and described in §..
1.3. A Brief Overview
The reader interested only in synchronic theoretical issues, not their historicalexplanation, needs to read Chapters –. Chapter  provides background on basicpoints of Udi grammar relevant to the issues discussed here, including the use ofthe several sets of PMs. In Chapter  it is shown that the position of the PM isinvolved in marking focus. It is the combined effect of Chapters – that estab-lishes that Udi possesses endoclitics. In Chapter  it is argued that complex verbsare single words; were they not, in some instances the PM would occur betweenwords, rather than between morphemes of a single word. In Chapter  it is shownthat PMs are clitics; were they not, we would have examples of infixes, ratherthan of endoclitics. Chapter  establishes informal rules for the placement ofPMs and shows that they may occur between the morphemes of complex verbsand be internal to the roots of simplex verbs. In Chapter  I argue that a hier-archical set of violable rules provides a simple account of the placement of PMs, while other theoretical approaches cannot provide an elegant account of thesefacts in Udi.
The reader interested only in diachronic theoretical issues must nevertheless acquaint himself with their description, which can be found in Chapters –.
Issues in diachronic morphosyntax are discussed in Chapters –. §. providesa new illustration of the use of the Comparative Method in syntax, while §§.–exemplify use of both the Comparative Method and Internal Reconstruction inmorphosyntax. In Chapter  I discuss the origin of the monoclausal cleft,described in Harris and Campbell (, chapter ). Chapter  discusses theproblem of how the innovative PM could move inside existing verbstems, andChapter  provides a summary discussion of how the various PM orders cameabout historically. Issues in the diachrony of case marking are discussed inChapter .
Naturally I hope that most readers will be interested in both synchronic and The problems with the use of texts are well known. If an example does not occur,we do not know whether its absence is due to its being ungrammatical or whetherit represents an accidental gap. For this reason, and to obtain minimal pairs, I haverelied primarily on elicitation for my analysis. I have elicited data (a) by asking fortranslations from Georgian (using bilingual consultants), (b) by presenting Udisentences for grammaticality judgement, (c) by asking consultants to make up asentence containing a particular form, and through other miscellaneous means. Inthis language, translation cannot be used effectively to determine word order andcertain other phenomena, while Udi sentences I have formulated may presentother problems.
On the other hand, texts provide advantages over elicited data in some respects. First, in texts we find constructions that we would not know existed if werelied on elicitation alone. The best example of this in Udi is compound verbs(§..), which could not have been elicited. Second, under most circumstances wefind more natural examples in texts. For both reasons, I started my research withclose study of texts and have elicited additional texts from consultants.
Selection of examples to cite is a problem distinct from analysis. In this work I have cited elicited examples when no textual example was available and when Ineeded pairs that differed minimally. On the other hand I have cited text ex-amples to show the naturalness of a construction and so that my data can be veri-fied. The latter is important in a work on Udi, while it would not be in one onEnglish, German, or Italian. In order to verify the data in the field, the linguistwould have to travel to the former Soviet Union and have a good knowledge ofUdi, Russian, Georgian, or Azerbaijani. Text examples, on the other hand, can beverified with only a trip to the library and knowledge of German (for Bouda ,Dirr , and Schiefner ) or Russian (for Bez#anov , , Dirr ). I Endoclitics and the Origins of Udi Morphosyntax have prepared some brief texts for publication on the web, with glosses and trans-lation in English to make verifiability that much easier.
1.4. History of Research
Although a considerable amount of linguistic research has been done on Udi (e.g.
Dirr , Gamq’reliZe , , Gukasjan , Jeiranis#vili , Panc#viZe ,Schiefner , Schulze ), PMs have never been fully described. Concerningthe location of the PM, Schiefner (: ), the first to describe the language,writes only The person marking occurs either after the verb or added between the two elements of thecomplex verb; third, however, the personal pronoun that belongs to the verb can occurenclitic to a preceding word.
Dirr (: ), in his description of the location of PMs (‘pronominal infixes’),seems to be the first to identify the secondary function of the position of the PMas indicating what he refers to as ‘logical stress’: These pronominal infixes are usually located between the root and other elements of theverb, . . . but very often they separate from the verb and adjoin to other words of thesentence, mainly to interrogative, negative, or prohibitive particles, and in general to wordson which the logical stress falls.
Elsewhere he notes that they may also be final in the verb form (Dirr : ).
Jeiranis#vili (: –) provides a more extensive discussion, distinguishing the position of PMs in the Vartas#en dialect from that in the Ni # dialect. Regardingthe former he writes in part: The position in which the case-changing pronominal person markers are used is as follows:
) in static verbs it is always final, for example: bu-ne (is) [for glosses, see below] . . . ) in
the root of dynamic (transitive and intransitive) verbs it is used in various places; for ex-
ample . . . ba-ne-ksa ‘is, happens’, c#’e-ne-baksa ‘go across, convert’, . . . baksa-ne ‘is, happens’.
. . . As we see, the person marker is never used in initial position in any verb: When used as
a prefix it must nevertheless follow some other independent (morphemic) element. . . . It
is interesting that grammatical meaning is changed by a change in the position of the
person marker in particular groups of verbs . . .; compare for example box-ne-sa ‘boils
(INTR)’—bo-ne-xsa ‘boils (TRANS)’.
As shown in the adduced examples, in composite verbs [complex verbs] the person marker is used either finally, as a suffix (more precisely, as a postpositional element): as#-
besa-ne ‘works’ . . .; or it turns up between the initial (nominal, adverbial) component and
the root-form of the auxiliary verb that follows it; for example: as#-ne-bsa ‘works’, and thus
in all other verbs of complex structure.
 For a general history of research on Udi, see Schulze (: –).
(Jeiranis#vili does not provide morpheme-by-morpheme glossing; a morecomplete glossing of the structure of the verbs cited is as follows: ba-ne-k-sa
bak ‘become’, -ne- SG, -sa PRES ‘it is, becomes, happens’ c#’e-ne-bak-sa
box-ne-sa
as#-b-esa-ne
as#- ‘work’, -b- ‘do’, -(e)sa PRES Panc#viZe (: –) observes that the PM is ordinarily what he calls an infix; that is, it is both preceded and followed by other parts of the verb complex.
However, with certain verb forms it may be the [final] suffix; his paradigm of verbforms (Panc#viZe : ) helps to make this statement more specific, for theforms listed there show the PM as the final or penultimate suffix only in thesubjunctive I, subjunctive II, imperative, and future II (eg. kar-x-al-zu ‘I will live’,where -zu marks the first person singular subject). He points out that the PMoften precedes the verbal base, as in te-z-cam-p-i ‘I did not write’, literally ‘not-SG-writing-SAID’, where the verbal base is cam-p- ‘write, writing-SAY’. He adds thatsometimes the PM is entirely outside the verb, as in (), where it is attached to thesubject sa q’o #a kaft’ar- ‘an old woman’.
‘An old woman comes to the king.’(Bez#anov : , ; Ni # dialect; cited by Panc#viZe : ) Panc#viZe (: ) suggests that we would have expected instead sa q’o #a kaft’arpascˇ’agun-t’o¢go.l e-ne-sa, with the PM between the two elements of the verb base,as it ordinarily is.
Schulze (: –) formalizes the discussion as a list of environments in which the agreement markers (which he calls PZ, my PM) may occur in relationto the verb (V), tense marker (T), nominal incorporated into the verb (Vn), andauxiliary (HV). Table . provides Schulze’s formulae and his example of each.
TABLE .. Patterns of morpheme order in Udi (from Schulze : –)  The word for ‘king’ varies by speaker; Schiefner () records pads#ax, Dirr () pacs#ax or pac#s#ag, Dirr () p’asc#ax, Jeiranis#vili () pas#c#’ag, and the consultant who wrote ‘Taral’ pac#ag.
 Schulze () does not translate this form at all; in Schulze-Fürhoff (b: ) he glosses it as a perfect, ‘he has said’. In my view, the TAM forms have not been studied thoroughly enough for us todistinguish adequately among the meanings of the several pasts.
Endoclitics and the Origins of Udi Morphosyntax While these sources describe a wide variety of locations available for PMs, theyprovide little in the way of statements to predict where PMs will occur under vari-ous circumstances. That is, these sources do not state when the PM will occur inwhich position. Schulze-Fürhoff (b: ), on the other hand, states There is no evidence of functional criteria for the use of the single slots. Sentence intona-tion may play a role (when the verb is final, the structure VERB–TM–PAM [i.e. VERB-STEM–TAM–PM] is preferred).
In Chapters  and  I establish the criteria that determine the position of the PM,showing that for many speakers these are not preferences, but absolute rules, whilefor some speakers variation is permitted in two out of seven rules. Verb-final orderplays no role in this (see examples above), and in Chapter  I show also that into-nation does not determine the position of the PM.
1.4.2. The Structure of Complex Verbs The most important aspects of the structure of complex verbs were described in thevery first work on Udi grammar, Schiefner (: –). Schiefner notes that there arefew simplex (monomorphemic) verbs in Udi; among them he includes aq’- ‘take,receive’, uk- ‘eat’, tit’ ‘run’, and others. The majority of verbs are what Schiefner callscompounds, which I refer to as complex verbs. He points out that these contain theverb besun ‘make, do’, pesun ‘say, do’, desun, t’esun, or less frequently k’esun (or its vari-ant xesun), where -(e)sun forms a masdar (deverbal noun, often used as the citationform); I refer to these verbs henceforth as light verbs. These light verbs combine withnouns (e.g. q’i.-besun ‘be afraid’, with q’i. ‘fear’), including ones in the ergative-instru-mental case (e.g. q’uful-en-b-esun [castle-INST-DO-MAS] ‘lock up’, with q’uful ‘castle,fortress’). They may also combine with adjectives (e.g. agu-b-esun ‘make bitter’, withagu ‘bitter’), or with adverbial elements (e.g. ala-b-esun ‘raise’, with ala ‘up, above’).
The light verbs combine with verbs in the infinitive form (in -es); e.g. a¢c-es ‘disappear,be lost’ combines with b-esun ‘make, do’ to form a¢ces-besun ‘destroy’. The light verbdesun, t’esun forms causatives in the same way; e.g. ot’bes-t’esun ‘make someoneashamed’, based on ot’-besun ‘shame oneself’, based in turn on ot’ ‘shame NOUN’.
Schiefner observes further that a number of verbs other than light verbs also incorporate nouns (e.g. c#ubux-aq’-sun ‘marry’ from c#ubux ‘woman’ and aq’-sun‘take, receive’) or adverbial particles, many of which are however no longer inde-pendent words (e.g. ba-sak-sun ‘push in, stick in’ from ba- ‘in’ and sak-sun ‘push’).
Dirr (: –) divides Udi verbs into three types: (a) Simple verbs, e.g.
aq’sun ‘take, receive’, (b) verbs composed of a root, a preverbal element, and theformant of the masdar -(e)sun (esun, b(e)sun, p(e)sun, t’esun, desun, k’esun, xesun,kesun); e.g. ba-p-sun ‘put in’, with ba- ‘in’, and (c) verbs composed with a root verband another constituent. In the third type the verb can combine With a noun: as#-besun ‘work’ (with the noun as# ‘work’) (ii) With an adjective: xuru-bsun ‘reduce to fragments’ (xuru ‘small’)(iii) With an adverb: oq’a-sak-sun ‘push down’ (oq’a ‘down’) (iv) With a non-Udi verb: buyurmis#-besun ‘fasten’; or with an Udi verb in -es, the formant of the infinitive: ap’es-besun ‘cook TRANS’ (ap’sun ‘cook INTR’) (v) With an interjection: vay-besun ‘moan, suffer’ (vay ‘woe!’).
Dirr (: ) seems to have been the first to point out that the roots of simpleverbs (his type (i)) are divided in two parts by the pronominal ‘infixes’; e.g. a-ne-q’-sa ‘she takes, receives’ from aq’- ‘take, receive’, -ne- SG, and -sa PRES.
Jeiranis#vili (: –) lists thirty simplex verbs, yet concedes () that many of these are no longer used this way. For example, he lists oc’- as a simplex rootand provides examples of finite forms, such as oc’-ne-sa ‘it is washed’, oc’-ne-c-e ‘itwas washed’, oc’-eg-al-le ‘it will be washed’. The root oc’ actually occurs primarilyin oc’-d-, a complex verb meaning ‘wash’ (see further below).
One subgroup of Jeiranis#vili’s simplex category (numbers –: c#’esun ‘go out’, baysun ‘go in’, esun ‘come’, taysun ‘go (away)’, laysun ‘go up’, cisun ‘go down’) representdirectional movement. He suggests that the adverbial portions of these were onceindependent words and that they are now similar to preverbs of other languages.
Among complex verbs, Jeiranis#vili singles out a group that have an initial consonant which is a fossilized element, possibly with a vowel, followed by a root:ba-k- ‘be, become; be able’, ba-q- ‘have, get’, bi-x- ‘bear, be born’, bi-q’- ‘catch, hold;build’, bo-k’- ‘burn’, bo-s#- ‘satiate, saturate’, bo-t’- ‘cut, break’, bo-s- ‘throw’, bo-x-‘boil’, bo¢-q’- ‘gather’, etc. Elsewhere Jeiranis#vili () has identified this b- as theinherited marker of the neuter gender-class agreement (see §. for discussion).
Jeiranis#vili (: –) notes that either the incorporated element of a complex verb or the verbal component may be limited to use in these complexes. Thecomplex verb as#-b- ‘work’ provides an example of a complex in which bothcomponents are still used independently, as# ‘work’ as a noun, and b- (masdarbesun) as a verb ‘make, do’. An example in which the incorporated element is stillused independently, while the verb is not, is z#ol-d- ‘cork up, stop up’ (z#ol ‘cork,stopper’); -d- is not used today as an independent verb. Jeiranis##vili (: ) iden-tifies s#am-p- ‘kill’ as consisting of productive p- ‘say’ (see note ) and an elements#am which is no longer used outside this complex.
Jeiranis#vili further notes that the nominal constituent can itself be morpho- logically complex. Among his examples are azaru-bak- ‘become ill’ (based on theunderlying form azar-lu ‘ill’ related to azar ‘illness, need’) and is#ex-tay- ‘marry(said of a woman)’ (where is#ex is the dative case form of is#u ‘man’, and tay- is ‘go’).
Schulze (: –) recognizes two types of Udi verb structure: simple and complex. He is careful to define the simple verb structures as those that cannot beanalyzed synchronically, and points out that a number of these are complex from adiachronic point of view. Among the latter he points out two subtypes. The old loca-tive preverbs, he says, combine with the suppletive verb base e(g) ‘go’ to form verbsof directional movement, such as lay- ‘go up’, bay- ‘go in’. Schulze also draws attentionto the fossilized gender-class marker (CM) b-, which shows up in numerous simpleverbs with the structure bVC, such as bak- ‘be’, bix- ‘bear, give birth to’, bok- ‘burn’.
Endoclitics and the Origins of Udi Morphosyntax Complex verbs generally consist of two parts: a nominal base and a verbal element. Schulze notes that the nominal base is generally in unmarked form whenit combines with a recognized auxiliary verb (HV, light verb); on the other hand,when it combines with a simple verb base the nominal base may be inflected.
Formally these are of three types—those formed with one of the intransitiveauxiliaries (see below), those formed with one of the transitive auxiliaries, andthose formed with an independent simple verb (: ):Vn + HVi xabar-aq’sun ‘ask’ (literally ‘take question’) A fourth type consists of combinations with a ‘quasi-nominal’, which was origin-ally verbal, but which no longer occurs independently: gac#’-esun ‘be bound’, gac#’-pesun ‘bind’. In this case the auxiliaries function to make the quasi-nominal basesinto transitive or intransitive verbs.
While the intransitive HVs are only e(g)- ‘go’ and bak- ‘be, become’, Schulze argues that the transitive HVs include practically all of the stops of the phonemeinventory of Udi: b-, d-, p-, k-, q-, p’-, t’-, k’-, q’- (: ). The identical charac-ter of these stems becomes clear, he says, from the fact that transitives formed withany of those named above can be opposed by an intransitive formed with -esun. Schulze (: ) lists the following verbs as ones which also allow incorp- oration of a nominal element: aq’- ‘take’, dug- ‘beat’, tast’- ‘give’, sak- ‘throw’, zap-‘pull’, biq’- ‘catch’.
Lastly, there is a group of composed verbs whose quasi-nominal base is formed with the Turkish morpheme -mis, (-mıs,, -mus,, -müs,), formant of the past parti-ciple. In Udi this element is borrowed with specific verbs; it does not exist as anindependent morpheme. Schulze-Fürhoff (b) presents the same view in anabbreviated form.
Other aspects of the history of research on Udi are discussed below as they 1.5. The Theoretical Basis of This Work
1.5.1. Diachronic Syntax and Morphology The theoretical framework for the historical analysis undertaken here is providedby Harris and Campbell (). Of particular importance is that in historical  Schulze (: ) provides evidence to support this for verbs in k’-, p’-, d-, and b-. However, in the Vartas#en dialect, the word Schulze cites for p’- is not gac#’-p’-esun ‘bind’ but gac#(’)-p-esun; seedictionary entries in Jeiranis#vili (: ) and Fähnrich (: ) and textual examples such as D : , DG : , as well as examples such as D : , D :  that show the suppletion -exa/p-/(u)k’-,which characterizes the light verb ‘say’. Thus, it may be that in the Ni # dialect there are more light verbsthan in the dialect described here.
syntax and morphology we are dealing with just three mechanisms of change:reanalysis, extension, and borrowing.
Reanalysis is defined in Harris and Campbell (: ) as ‘a mechanism which changes the underlying structure of a syntactic pattern and which does notinvolve any modification of its surface manifestation’. Underlying structure isspecified as including ‘(i) constituency, (ii) hierarchical structure, (iii) categorylabels, and (iv) grammatical relations’. Surface manifestation, on the other hand,includes ‘(i) morphological marking, such as morphological case, agreement, andgender-class, and (ii) word order’. Extension is defined as ‘a mechanism whichresults in changes in the surface manifestation of a pattern and which does notinvolve immediate or intrinsic modification of underlying structure’ (Harris andCampbell : ). Borrowing is ‘a mechanism of change in which a replicationof the syntactic pattern is incorporated into the borrowing language through theinfluence of a host pattern found in a contact language’.
The three mechanisms identified apply in both syntax and morphology. It is recognized as especially important that a complete account not limit itself to themorphological source of the phonological content of a morpheme (for example,the independent pronoun zu ‘I’ as the source of the first person singular subjectPM zu), but consider also the syntactic structure of the phrases or clauses involvedin change.
In linguistics of Caucasian languages it has often been a problem that much of the syntax and some morphology is attributed to borrowing, with little evidenceto support specific individual claims. I consider it as essential to present specificevidence of borrowing as to present specific evidence of inheritance or of theapplication of another mechanism of change. Often there is prima facie evidenceof borrowing (that is, a similar construction exists in a contact language), but thisfact alone does not actually prove that borrowing was the source of the construc-tion. In thoroughly studied instances of contact, one can often identify an inter-nal mechanism of change, which likely applied because of the influence of thecontact languages. Because it is often difficult to find specific evidence of borrow-ing, it is necessary to recognize that contact may have been one of several causesof change. It is important to note that even the proven fact that a syntacticconstruction is due to the influence of contact languages does not absolve thelinguist of the responsibility to account for the mechanisms through which itbecame part of the language and through which it developed further. In addition,if a construction was borrowed early and used in a proto-language, then it mustbe reconstructed to that proto-language; that is, the term proto-language does notinclude only what is inherited, but everything found in that stage of the language,whether acquired at some earlier time through borrowing or made up of inher-ited elements. It may be difficult or impossible to determine whether a borrowedconstruction was part of a proto-language or was later borrowed individually bydaughter languages, but this does not mean that it should not be our goal to deter-mine this.
Endoclitics and the Origins of Udi Morphosyntax Reconstruction in syntax is reconstruction of patterns, not reconstruction of specific sentences that have existed in the past (Harris and Campbell , chap-ter ). Syntactic patterns, like morphological patterns, can be reconstructedthrough application of the Comparative Method, with rigorous safeguards forcomparability of data (loc. cit., Harris : –).
Universal properties of the reduction of biclausal structures to monoclausal ones and of changes in case-marking patterns are two issues given particularattention in Harris and Campbell (, in chapters  and , respectively), andthese two are particularly relevant to the Udi changes described in this book. Herethese two questions arise especially in Chapters  and , and more backgroundon theoretical issues is provided there.
To the extent possible, I have kept the descriptive chapters (Chapters –) free ofthe assumptions of any specific syntactic or morphological theory, though I haveassumed syntactic constituency and hierarchy.
In my analysis of focus, I have relied heavily on Lambrecht (), who takes the approach that every sentence has focus; my analysis has also been influencedby recent work on focus in Hungarian and a number of other diverse languages(references in Chapter ).
In analyzing complex verbs, I have been influenced by the framework set out in Ackerman and Webelhuth (), which provides a theory of predicates. Mostpertinent to the present problems is that the approach in their work provides thebasis for treating an auxiliary and a dependent verb form as a predicate. Seenthrough the diachronic lens of Harris and Campbell (), this means that an‘auxiliary’ and a ‘dependent’ verb may originate as independently viable verbs,each governing the syntax of a clause in a biclausal structure. As a consequence ofreanalysis, the once independent verbs may take on their ‘auxiliary’ and ‘depen-dent’ verb status, now forming a single predicate. Through further reanalysis andthrough extension, these may, over time, become a single verb. Though more vis-ible, the many small changes that create a single verb from a two-part predicateare more superficial than the single reanalysis that makes a biclausal structure intoa monoclausal one, and turns two verbs into a single predicate. It is proposed inChapter  that such a series of changes took place in Udi.
To determine whether PMs and other ‘particles’ are clitics, I have turned first to Zwicky and Pullum (), but also have considered important work byother specialists in this area. The analysis of the position of clitics in the wordand in the phrase in Udi is based on Optimality Theory. The theory developedin Harris and Campbell () is assumed in dealing with historical issues,though in many instances I have not discussed here the relevance of Udi datato that framework.
The data presented here represent a challenge to the Lexicalist Hypothesis.
Some theoretical aspects of this are discussed briefly in Chapter  within the framework of Optimality Theory. That analysis has drawn most on McCarthy andPrince (), together with other works cited there.
This work is intended to be primarily descriptive and to provide an explan- ation of why Udi is the way it is, and thus of why it provides a counterexample tothe Lexicalist Hypothesis.

Source: http://udilang.narod.ru/papers/Harris_ch1.pdf