The mosaic evolution of Left Dislocation in Xhosa

This paper demonstrates that the status of Clitic LD in Xhosa is a result of the mosaic evolution of Xhosa grammar. It emerges as an accumulation and combination of two more individual, distinct and, at least, initially separated developments and characteristics – LD sensu stricto and Object Agreement. This view enables the authors to propose a possible solution to the problem of whether the Clitic LD structure in Xhosa (and Nguni) is an instantiation of LD (and its prototypical function as posited by Westbury (2014)) or fronting (and its topical and focal functions). The Clitic LD structure is employed to accomplish the two tasks. The former is derived from the original LD construction, while the latter arose due to the Object Agreement cycle. The mosaic character of LD in Xhosa, in turn, demonstrates the situatedness of LD. LD is determined not only by its own evolutionary baggage (the source from which it has developed) but also by the dynamics of the environment in which it has been embedded (the properties of other constructions).


Introduction
The concept of mosaic evolution is currently one of the tenants of the theory of evolution.The mosaic character of evolutionary mechanisms implies that different parts of a system evolve quasi-separately and at different rates (Futyma 2005: 550).First, evolution is modular.A system does not evolve as a whole, but rather fragmentarily.Features and parts of a system do not develop all together, for instance by following a holistic "master-plan".They evolve in separation by being subjected to, at least originally, distinct developmental processes (Futuyma 2005: 54).Second, one feature of a system (for instance of an organism), or a part of it, can evolve more rapidly than other features and parts.That is, different features and parts develop at distinct speeds.It is the former property (modularity) that is especially relevant, since the latter (distinct rate) may be regarded as one of its consequences.
These two properties of mosaic evolution imply that a new macro-feature or a new taxon emerges not as a well-conceived holistic plan but rather as a summation or convergence of individual and quasi-separated micro-evolutions.Larger features do not evolve as homogenous coherent modifications but as an accumulation of a number of microscopic changes characterized by their own evolutionary and systemic properties.A linear change is a result of a network or web of individual evolutionary processes.As such individual evolutionary processes accumulate and combine, a new pattern emerges at a more macroscopic or global level (for examples of mosaic evolution in biology consult Futuyma 2005: 73).
All of this means, in turn, that significant macro-evolutionary trends are gradual. 1Mosaic evolution eliminates clear-cut boundaries between taxa and species (Futuyma 2005: 78). 2  Higher taxa and species do not emerge in a single step or by saltation.On the contrary, they arise through a set of changes that, as have been noted, are relatively independent and regulated by different speeds (Futuyma 2010: 519).
The theoretical similarity between biological and linguistic evolution has been postulated and investigated by many researchers.Above all, like biological species, language is a dynamic system -it evolves.The present paper explores the mosaic or composite character of language evolution (and language functioning).The macroscopic evolution of a macro-feature or the formation of a new systemic property consists of more atomic, individual changes or developmental processes that may be in principle unrelated.As they accumulate and combine, they may organize into clusters that have a greater impact on the macro-structure, ultimately being able to alter it.Accordingly, components of language evolve not by jumping from one discrete stage to another, but rather by transmuting smoothly due to the accumulation of a number of more microscopic, sometimes disconnected, changes.
In this article, we will demonstrate that the current status of L(eft) D(islocation) in Xhosa partially emerges in response to an (at least originally) independent evolutionary process.That is, while the structure of LD (in which the referent is identified as object in the matrix clause) has virtually persisted as such, its macro-property has changed because of another development which the language has experienced, the so-called O(bject) A(greement) cycle.Therefore, the systemic place of LD in a language depends both on its own genetic baggage (i.e. the evolution from its source) and on the embedding environment in which it has been developing and existing (i.e. the properties and dynamics of other constructions).Consequently, the analysis of LD (both language-specific and cross-linguistic) should not be conducted in isolation from other systemic components of grammar, but must be expanded to the study of many other constructions and functions.
We will begin the discussion with the typologically based explanation of LD and OA.

2.
Left Dislocation and Object Agreement

Left Dislocation
LD is a specific grammatical construction that encodes a specific function.This means that LD is defined both in functional and formal terms.First, as far as the function is concerned, a prototypical LD construction activates or reactivates a low-accessible referent and identifies its role in the proposition.This role can be pragmatic (e.g.topic or focus), syntactic (e.g.subject, direct object or indirect object) or any other (Westbury 2014: 201-202, 214, 340).This complex task will be referred as the [R+I] function.Second, as far as the form is concerned, given empirical typological studies, the exemplar of LD exhibits the following properties: 1) the [R+I] function is encoded in a mono-propositional context; 2) the dislocated referent is external to or linearly detached from the matrix clause, being placed initially, i.e. to the left; 3) the dislocate is conveyed by a nominal phrase in an unmarked case form (casus pendens, nominative, absolute, etc.); 4) the dislocate's role is specified in the matrix clause by an overt element that functions as an argument or an adjunct; 5) this element is co-indexed with the dislocate; 6) the co-indexed element is a pronominal and exhibits a total identity link (Westbury 2014: 138-139); 3 7) the dislocate is marked by special intonation or a pause; The fact that scholars have proposed a functional-formal prototype of LD does not mean that all LD constructions actually found in languages are identical and exhibit all the properties postulated for that prototype (and only such properties).Quite the opposite is true -the concept of prototype considerably relaxes the definition of LD.That is, although constructions that, in various languages have been defined as LD, can fully comply with the posited prototype, they may also approximate it to a certain extent.LD is a radial category whose members are connected via family resemblance.
The categorial fuzziness of LD stems from the fact that, as everything in language, LD is not a static concept.It is dynamic, constituting a part of a larger evolutionary process -LD derives from more original structures and evolves into more grammaticalized ones. 4As a result, LD formations that are found crosslinguistically can diverge from the functional and formal prototypes posited above, being able to convey other functions (e.g. the disruption or discontinuity, contrast, foregrounding in cases where the referent is discourse active; Westbury 2014) and adopt less canonical structures (e.g. the co-indexed argument in the matrix clause is a full nominal, the pausal is minimal or absent, or the identity link is partial and/or involves a frame; for examples of this, see section 4).
For the purpose of this research, only one particular class of LD will be relevant.This class consists of LD structures in which the syntactic role of the dislocate is identified in the matrix 3 Although the repetition of the referent is possible, more commonly the co-indexation is achieved by means of pronouns, which are also exemplary tools to refer to discourse active referents. 4LD with its [R+I] function evolves from spontaneous discourse routines or "recognition-search-sequences" composed of a referent and a proposition.In these sequences, the referent is first established and subsequently followed by a proposition by which it is specified (Couper-Kuhlen 2011: 426; see also Ochs and Schieffelin 1976: 240).At this initial stage, the two functions are carried out by two different propositions usually in distinct or subsequent turn-constructional units.By definition, they are separated by a pause (Couper-Kuhlen 2011: 428) and, due to this recognition-search nature, the second part almost naturally lends itself for pronominal uses (See for instance the following example: (Speaker A) Look.Peter! (Speaker B) Yeah, I saw him yesterday).LD as an exemplar grammatical category emerges from "the coalescence and entrenchment of [such] a frequent and flexible conversational routine" (Couper-Kuhlen 2011: 429).That is, LD emerges when the two original interactional sequences are collapsed into an integrated whole or in a single sentence (Couper-Kuhlen 2011: 428).On the other edge of the continuum, as the coalescence increases, pausal effects decrease and the construction can be used to convey functions different from the 'recognition and search' or [R+I] function, namely listing and/or contrast (Couper-Kuhlen 2011: 429).
clause by means of clitics and specified as this clause's object.This means that in the following discussion, the term LD will be understood more narrowly, i.e. as an object and clitic LD (henceforth CLD).

Object Agreement
In its prototypical form, OA is a morphosyntactic harmony between the object of the clause and the verb.That is, the clause's object, expressed by an independent nominal, appears on the verb as an inflectional marker.This parallels S(ubject) A(greement).While SA requires an appropriate verbal form (inflection) depending on the properties of the subject of the clause, OA requires a particular form of the verb given the characteristics of the object (e.g.gender, class, number and/or case).In its canonical (or ideal) form, OA is compulsory (the verb is always marked for object) and morphological (the marker is incorporated in the verb itself as a bound morpheme; the same holds true for SA).
However, similar to LD, OA is a dynamic and fuzzy phenomenon.It both derives from less grammaticalized constructions and develops into other more grammaticalized structures.Therefore, OA exhibits a great crosslinguistic variety of actual manifestations.All such instances of OA in specific languages may be understood as either approaching (diachronically) or approximating (synchronically) the prototype postulated above.
From a diachronic perspective, OA (as SA) derives from pronominalization.Givón (1978: 151) argues that "agreement and pronominalization […] are fundamentally one and the same phenomenon" (see also Givón 1976 and1979: 243-244).This observation is correct, as OA tend to arise from independent (demonstrative, emphatic) pronouns (Van Gelderen 2011a: 491, 494).The conversion of pronouns into OA is referred to as the object cycle -a process that consists of the following stages (cf.Figure 1 below; Van Gelderen 2011a: 493-494).At the most original stage, the construction constitutes a phrasal structure in which resumptive elements are first oblique (emphatic or demonstrative) and, later, personal object pronouns.Subsequently, the pronouns evolve into clitics.They gradually become dependent heads attached to verbs, being successively reanalyzed as agreement markers.At the end, the agreement is lost and the entire cycle possibly renewed (Van Gelderen 2011b: 88, 2011a: 497).5 oblique (emphatic or demonstrative) pronouns > personal pronoun > clitics > agreement affixes > zero

Figure 1: Object Agreement cycle
The OA cycle gives rise to a synchronic variation of possible OA constructions.That is, synchronic variations, which are classified into three main types, reflect the major stages of the OA cycle presented above.Instances where the object pronoun (typically unmarked or oblique) is fully phrasal and appears in a specifier position correspond to the original stage (Van  Gelderen 2011b: 88-89, 93-94). 6This type reflects a common crosslinguistic phenomenon where an element can be moved to a topic position, usually being extrapolated (Van Gelderen 2011b: 90, 92). 7Constructions where an additional co-referential full nominal is unavailable, such that a nominal or an independent (full or phrasal) pronoun on the one hand and a dependent pronoun on the other hand coexist in complementary distribution, are classified as instantiations of a more advanced stage (Van Gelderden 2011b: 89, 95-102). 8Lastly, structures in which the pronoun (functioning previously as a pronominal head) evolves into a higher functional head understood as a set of uninterpretable features on a verb, are regarded as the most advanced stage or the peak of the development of OA.In this class of constructions, the Verbal Phrase contains a pro as its inherent object.A set of pronouns can be renewed since the older pronouns have been reanalyzed as agreement markers (Van Gelderen 2011b: 90, 102-105). 9 The OA cycle relates syntax with morphology (OA is incorporated in the verb itself as its morphology or inflection) and optionality with compulsion (OA becomes compulsory).The two sub-processes are gradual.They do not yield neat dichotomies, but rather form scalar continua.Therefore, the stages and constructions belonging to them may themselves not be uniform.On the contrary, they may exhibit considerable variation. 10 Having set theoretical foundation for this research, we will now discuss how the phenomena of LD and OA have been analyzed in studies dedicated to Nguni languages, including Xhosa -the language that constitutes the object of this article.

Clitic Left Dislocation in Nguni
LD (CLD included) is a common feature of Nguni languages as it is also in the greater Bantu family.As for the Nguni branch, this phenomenon has been studied particularly extensively in Zulu.In Zulu, CLD refers to cases where a dislocated phrase appears at the beginning of a sentence and is co-indexed with a resumptive element in the matrix clause, regularly an object pronoun that agrees with the dislocate in class: [Ushukela]i abantwana bayawuithanda 'Sugar, the children like it' (Cope 1984: 41;Bresnan and Mchombo 1987;Zeller 2009: 131).In fact, it is possible to distinguish three types of object LD in Zulu (Zeller 2009: 132).The type that is the most relevant to our study corresponds to cases where the resumptive element is a pronominal object marker agglutinated to the verb as an infixed bound morpheme: [Le ndoda]i ngiyayiithanda 'This man, I like him' (Zeller 2009: 133).This pronominal element is sometimes analyzed as a clitic (Visser 1989;van der Spuy 1993;Baker 2003; see section 3.2 below). 11As far as Xhosa is concerned, CLD seems to be analogous to CLD in Zulu.That is, CLD in Xhosa 7 Compare the following sentences in Polish: Mam go napisanego, ten list 'I have it written, this letter', Napisalem go, ten list 'I wrote it, this letter' and Ten list, juz go napisalem 'This letter, I have written it already'. 8Constructions belonging to this stage can be found in Persian, English and Arabic (Van Gelderen 2011b). 9This stage may be illustrated by constructions found in Spanish, Macedonian and Bulgarian (Van Gelderen 2011b). 10For instance, in Standard Spanish OA (or doubling) is obligatory only with pronominal objects (Lo vimos a él 'We saw him').It is also compulsory with nominals if there is a pronominal direct object in the clause (Se lo dio a Maria 'He gave it to Maria').As the evolution from the second stage to the third stage began but has not been completed, this LD construction attest to a fuzzy (or transitory) developmental phase (Van Gerldern 2010: 102).
One should note that in some Spanish varieties, OA is also found with nominals (Pedro lo vió a Juan (ibid.).
11 Two other types are LD constructions in which the resumptive element is a full (absolute/emphatic/independent) pronoun ([Le ndoda]i ngithanda yonai 'This man, I like (him)') or in which the dislocate is co-indexed both with a pronominal clitic and a noun phrase and its epithets ([UJohn]i, ngiyiithanda ngempela leyo ndodai 'John, I really like that man' (Zeller 2009; see also Doke 1954;Kunene 1975).
is a construction where a NP is placed in the left periphery and its roles is specified in the matrix by an anaphoric pronoun (a clitic) agglutinated to the verb as a bound morpheme (Zerbian 2004: 73; see also Zerbian 2007).
There are two possible functional analyses of CLD in Nguni and/or Bantu.First, according to Zeller (2009), who works within the Principles-and-Parameters framework, CLD in Zulu behaves like fronting (topicalization) in English.The principal function of the dislocate is to present topic and the dislocate seems to be derived by movement (Zeller 2009).One of the argument for this interpretation is that in Zulu, any type of extrapolation (including CLD) automatically necessitates the use of a clitic object pronoun with the verb (Zeller 2009).12Inversely, the dislocate in CLD is not a free-floating topic or a (re)activated element that is further specified in the matrix clause.In this manner, the CLD construction would not be an instantiation of the LD prototype, as it would not encode the [R+I] function.Second, the dislocate can be regarded not as generated by movement but rather as an extra-clausal element, and CLD not as fronting/topicalization but as an example of the [R+I] function.This different analysis has been posited for Chichewa, a member of the greater Bantu family.In Chichewa, one finds a construction that, from a structural perspective, is fully equivalent to CLD in Zulu and complies with the formal prototype of LD.By studying this formation within the framework of Lexical Functional Grammar, Bresnan and Mchombo (1987) reach the following conclusion.
LD is an extra-sentential construction in which the dislocate corresponds to a free-floating discourse topic whose syntactic role is specified by a pronominal clitic anaphorically linked to the dislocate and incorporated into the verb.That is, contrary to the analysis of Zeller ( 2009), the dislocate does not belong to the matrix clause and is not an argument of the predicate.
According to this interpretation, CLD in Bantu would be an instantiation of the LD exemplary conveying the [R+I] function.13

Object Agreement in Nguni
Similar to CLD, the agreement between the object and the verb is typical of Nguni and, more generally, Bantu languages (Zeller 2012;Bearth 2003;Nurse 2008).Even though in most members of these groups, object markers are not obligatory on verb (if the object is already expressed and occupies a canonical position), one observes a development towards their more compulsory use (Wald 1979(Wald : 506, 1993: 331): 331).That is, in recent time, there is an evolutionary tendency towards an obligatory use of object markers on verbs and, thus, towards a more prototypical variant of OA (Riedel 2009;Marten, Kula and Thwala 2007). 14Accordingly, Zulu and many other Bantu languages (e.g.Swahili) are classified as developing towards Stage 3 of the OA cycle (Van Gelderen 2010: 106; cf.section 2.2 above).
In Zulu, the OA construction exhibits two properties that approximate it to the peak stage of the OA cycle (cf.section 2.2), without however reaching it entirely.These properties are related to the obligatority of agreement and to the status of resumptive elements.
First, although the agreement is compulsory in various instances (e.g. if the nominal referent is dislocated or moved from the canonical or the closest position to verb, as well as in relative clauses; Du Plessis 2010: 24), there are instances where it is not required.Importantly, in canonical positions, animate, inanimate, definite and specific objects can but do not require to be found with object markers (Zeller 2012).If a full nominal appears with a co-indexed object marker, a definite and/or specific interpretation of the object is imposed.Many of such cases convey the idea of emphasis and indicate the focal or topical function of the lexical object (Du Plessis 2010: 26, 141).Accordingly, the referent of the object is established as a discourse topic or focus (Bearth 2003).All of this inversely means that "inherently indefinite objects cannot appear with an object marker" (Zeller 2012: 226; see also Adams 2010).Thus, even though object agreement parallels the cases where the subject is lexically realized and appears on the verb in form of agreement marker (Du Plessis 2010), it also exhibits certain dissimilarities.Most importantly, while subject agreement is (almost absolutely) obligatory, the object agreement is compulsory to a lesser extent (Bearth 2003).
Second, the agreement is encoded by a bound morpheme incorporated into the verb.This morpheme is distinct from independent pronouns, also referred to as emphatic pronouns (Buell 2002;Van Gelderen 2010: 106).However, it is still debated whether this object marker on verb is a genuine agreement marker or a clitic pronoun.The analysis of object markers as cliticswhereby the use of an object marker and a co-indexed nominal is "clitic doubling" (Zeller 2012: 230) -has been proposed by Van der Spuy (1993) and Adams (2010;cf. also Visser 1984:125 and1986 for Xhosa).The analysis in terms of agreement -which regards the use of pronominal elements as inflectional affixes that agree with the nominal object -has been developed by Buell (2005), Henderson (2006) and Zeller (2012).The complexity and the solution to this problem lies in the dynamic nature of language.Namely, since object markers in Zulu are currently developing from pronominal clitics to agreement markers -the language being in a transition phase (Van Gelderen 2010; Zeller 2012) -they synchronically exhibit properties of both types, thus instantiating an intermediate stage of evolution (Zeller 2012: 219).Accordingly, object markers in Zulu are neither clitic pronouns nor agreement markers, but constitute another categorial type with distinct characteristic and distinct grammatical function (Zeller 2012: 232).
The two views regarding the categorial status of resumptive elements on the verb reflect two extremes of possible variations attested to in Bantu.Although, in Bantu, object markers may be used as pronominal clitics or as agreement markers, the majority of languages exhibit mixed properties and may be viewed as intermediate stages on a pronoun-agreement continuum (Zeller 2012: 232).These languages correspond to transitional stages of a grammaticalization process whereby pronominal object (first full and later cliticized) develop into agreement morphemes (cf.Henderson 2006). 15Zulu arguably fits into this group (Zeller 2012).Diachronically (and in harmony with the OA cycle), the object affixes in Bantu (defined as clitics or agreement markers) derive from independent pronouns (Nurse 2008: 65).That is, object markers, 15 To compare, object markers in Swahili are most commonly obligatory.They are obligatory with animate objects as well as with inanimate objects if these are definite or specific (Barrett-Keach 1995).That is in cases where the nominal referent is definite, the agreement is obligatory (ni-li-ki-soma kitabu 'I read the book').Given their inherent definiteness and animacy, the 1 st and 2 nd person objects necessitate object agreement.Nevertheless, indefinite inanimate objects may be unmarked on verb (ni-li-soma kitabu 'I read a book'; Van Gelderen 2011b: 106 quoting Givón 1976: 159).Other Bantu languages, may be in stage 2.This is exemplified by Kinande where the object marker is optional (Baker 2003: 109;Van Gelderen 2011b: 105).Nevertheless (as in other Bantu languages), it is regular if the referent is extrapolated (Van Gelderen 2011b: 106).
nowadays agglutinated to the verb in a preverbal position, descend from a more original, analytic structure postulated for Proto-Niger-Congo and possible in Proto-Bantu.For pronominal objects, this Proto-Bantu syntactic structure was S OP V.This means that pronominal objects appeared preverbally and were originally independent (Nurse 2008: 68-69, 223-224; see also Hyman 2007;Good and Güldemann 2006). 16egarding Xhosa, the standard view is that object agreement uses clitics as agreement markers (Visser 1984(Visser : 125, 1986).17Thus, the marking of object concord on verb constitutes a case of cliticization.An overt nominal object, which is co-indexed with the clitic, commonly appears in postverbal position, albeit it can also be located before the verb. 18Even though the preverbal object markers or clitics are optional in the canonical position (cf.Du Plessis 1997), in cases of any extrapolation, the agreement is obligatory.The extrapolation may locate the lexical object in various places in the sentence, even outside its own clause, for instance at the end of a subordinated or coordinated clause.Furthermore, the agreement regularly appears in cases where the object is definite or specified: Abafundi bayaluibhala uviwoi 'The students are writing the examination' (Visser 1984(Visser : 135, 2008(Visser , 2010)).In such instances, it is also assumed that the nominal object "receives greater emphasis than in a sentence where it appears without the clitic" (Visser 1984(Visser : 125, 2005(Visser , 2008)). 19However, in cases of extrapolation, the pronominal elements on the verb provide "the object feature", thus acting in a manner highly similar to genuine agreement markers (Visser 2005). 20

Evidence from Xhosa
Given the problem of the pragmatic function associated with CLD in Xhosa (cf.section 3.1 above), a crucial issue is to determine whether this construction conveys the [R+I] function and thus complies with the functional prototype of LD (cf.section 2.1).
In this section, we will demonstrate that, in Xhosa, the CLD construction (in which the resumptive element is the object clitic or object agreement marker agglutinated to the verb) is commonly used to convey the [R+I] function.Accordingly, it does conform to the functional exemplar of LD.In all such cases, CLD first activates or reactivates a referent, a frame or a referent's profile, all characterized by a low degree of accessibility, and subsequently specifies the role of the referent in the proposition (4.1).Apart from this, CLD in Xhosa can also be used to introduce elements whose accessibility is high.As other LD constructions found crosslinguistically, in such instances, CLD in Xhosa marks discontinuity and causes various effects typically associated with this function (4.2).
Our empirical research builds on Xhosa (1975) translations of the Hebrew Bible.To be exact, we analyzed whether the Biblical Hebrew construction that had been identified by Westbury (2014) as both structurally and functionally LD, was translated into Xhosa by means of CLD.
Since, in all such instances, the LD of the underlying Biblical Hebrew text doubtlessly conveys the [R+I] function typical of LD (or other less exemplary functions that are crosslinguistically associated with LD), it is obvious that CLD (if used in the Xhosa texts) would exhibit the same functional value in respective renditions.This functional equivalence is assumed due to the fact that the Xhosa texts are translations of the Biblical Hebrew original. 21

(Re)activation
Our data show that CLD in Xhosa can be used as a strategy to first activate or reactive an element that entertains a low degree of accessibility for the addressee and, second, to interpret its semantic and pragmatic role in the proposition.Inversely, the complex processing task of recognizing the referent (remembering or inferring) and interpreting its pragmatic and syntactic role can be mitigated in Xhosa by CLD.This stands in agreement with the principle posited by Westbury (2014: 314), whereby it is easier to express the composite and cognitively costly [R+I] task by portioning it into two slots -first, (re)activation and, second, interpretation.In this manner, the (re)activated referent is treated independently from the matrix clause, where the role is subsequently interpreted.In Xhosa, as it is the case in other languages (cf.Westbury 2014), the [R+I] task encoded by LD, more specifically, by CLD if the referent is specified in the matrix clause as the objet.
From a discourse-pragmatic perspective, when CLD is used in the [R+I] function and concerns a referent that entertains a low degree of accessibility, it exhibits three main subtypes depending on the properties of the dislocate and the resumptive element.That is, the dislocate may be exhibit a total identity link with the resumptive element (cf.section 4.1.1.(Re)activation of a referent), a relationship of relevance or restriction (cf.section 4.1.2(Re)activation of a frame), and a partial identity link with the resumptive item, thus relating to a new profile of it (cf.section 4.1.3(Re)activation of a profile).

(Re)activation of a referent
Given the pragmatic role of the referent (which entertains a low degree of accessibility) specified by the co-indexed pronominal in the matrix clause, the CLD construction can be used to announce topic or focus.
First of all, CLD can be viewed as announcing the topic of the matrix clause.Accordingly, the dislocate activates or reactivates a referent which is subsequently identified in the proposition as the topic. 22In our corpus built on Westbury's examples of explicit and prototypical LD constructions in Biblical Hebrew (Westbury 2014), the following may be observed.LD structures that in Biblical Hebrew convey the [R+I] function and are co-indexed with a resumptive element specified as a pragmatic topic and syntactic object, are commonly rendered in Xhosa by means of the CLD construction with clitic pronouns or object agreement markers.
21 For a more nuanced discussion of complexities involved in translations of the Hebrew Bible consult Van der Merwe (2016). 22As convincingly argued by Westbury (2014), the topic of the proposition cannot be the dislocate as "the topic relation exists within the pragmatically structured proposition" (Westbury 2014: 314).It is rather encoded by a resumptive element found in the matrix clause.
Due to the fact that the syntactic role of the referent is specified as the object, that referent's pragmatic function surfaces as secondary topic.Crosslinguistically, the subject role relates to the referent that acts as a primary topic of the proposition.For an object type of LD, this topic is usually secondary (Westbury 2014: 314). 23 "As for the Merarites, you will list them by their clans" (Num 4:29) Although the cases of topic-announcing CLD seem to be especially abundant, the CLD construction in Xhosa can also introduce a referent whose pragmatic role is specified by the resumptive element (with whom the referent possesses a total identity relationship) as the focus of the proposition.In other words, CLD in Xhosa is used to render the focus-announcing LD in Biblical Hebrew.As was the case with the topic announcing CLD, the type that announces the focal function of the referent complies with the primary [R+I] function of LD.CLD construction mitigates the equally costly process of (re)activating and specifying the focal role of the referent by detaching it from the matrix clause and indicating its focal role by a resumptive element (cf.Westbury 2014: 318).An illustration of this subtype of CLD may be found in example 6, where the dislocated referent (yonke impahla edelekileyo, egxugxileyo in Xhosa) is identified as the focus of the sentence. 24This use of CLD seems to be less common among the various subtypes of this construction in Xhosa.This means that other constructions are employed to convey this function.

(Re)activation of a frame
Although the type of CLD where the referent is (re)activated and, in particular, its topic function announced, are commonly expressed by CLD in Xhosa, the CLD construction can also be used to announce frames.As the previous types, this variant expresses the primary function of (re)activating and interpreting the role of a referent characterized by a low degree of accessibility.However, in this class, the dislocate does not exhibit a total identity relationship with the co-indexed element in the matrix clause.The dislocate is rather found in the relation of relevance to the resumptive pronominal and restricts the proposition that follows (Westbury 2014: 32).This type of CLD in Xhosa can be illustrated by the following example extracted from the Bible, in which the dislocate (USarayi 'Sarai') creates a frame for the subsequent matrix clause: 24 For an analogous interpretation of ‫ס‬ ‫ָמֵ‬ ‫נ‬ ‫וְ‬ ‫ָה‬ ‫ז‬ ‫בְ‬ ‫מִ‬ ‫נְ‬ ‫ָה‬ ‫ָאכ‬ ‫ל‬ ‫מְּ‬ ‫ָל-הַ‬ ‫כ‬ ‫וְ‬ in Biblical Hebrew, see Westbury (2014).In Biblical Hebrew, the resumptive element is fronted in the matrix clause. 25Compare, for instance, the following structure: Yaba yiHatsore yodwa awayitshisayo uYoshuwa 'It was Hazor only that Joshua burned' (Josh 11.13). 26This example can also be viewed as an instance of topic announcing in a context where topics are contrasted.
The focus of the this construction is indicated in Biblical Hebrew by the infinitive absolute. (

(Re)activation of a profile
The third subtype of LD identified by Westbury is also available in Xhosa.In this use, LD does not (re)activate a referent, but rather (re)activates a new conceptual profile of it.Here, the motivation for the use of LD is not the low accessibility of the referent or frame.Rather, LD is employed to "recast the hearer's/speaker's existing cognitive representation of the referent against a new conceptual profile" (Westbury 2014: 323).Usually, there is a modifying element (for instance, a non-restrictive relative clause) that add new attributes to the referent.This may be illustrated by the following example where the speaker activates not the referent itself (i.e.ilizwe 'the land'), which is in fact already discourse active, but a certain profile of it (i.e.lonke ilizwe olibonayo 'all the land that you see').Such profiles can be further specified as topics (as in example 8 below) or, albeit less commonly, as foci."For all the land that you see, to you I will give it and to your offspring" (Gen 13:15)

Discontinuity
Apart from the types of CLD in Xhosa that fully comply with the functional prototype of LD, CLD in Xhosa (and LD in other languages) can be used in other less exemplary functions, which are also associate with LD constructions crosslinguistically.One of them is a discontinuity function.In this use, contrary to its main [R+I] function, CLD does not mitigate complex cognitive processing, but rather disrupts it (Westbury 2014:324).To be specific, in such cases, the construction is used to produce pragmatic implicatures that arise due to the dislocation of the referent that is already active and, moreover, bestowed with a high degree of accessibility.Since such a context seems to be superfluous for the use of the LD constructionthe processing cost being already mitigated -its presence results in an over-use.This, in turn, creates a new, opposite, effect -disruption or discontinuity (Westbury 2014: 324).As an epiphenomon of this disruption or discontinuity -a type of an implicature -functions such as thematic shift, contrast/comparison or foregrounding emerge.
Our (limited) Xhosa corpus provides examples of such an over-use of CLD for discontinuity purposes.In most cases, this discontinuity is explored to produce the effect of contrast and/or comparison.It also seems that both relations are not encoded by CLD itself, but are rather inferred from the propositional content of two or more clauses (cf.Westbury 2014: 327).This contrastive relationship of CLD may be illustrated by examples 9 and 10 below.In both cases, the use of CLD creates the discontinuity of the discourse.To be exact, CLD enriches the effect of contrast between the Israelites and other nations (cf.example 9), and between the fruit of the tree of knowledge and the other types of fruit (cf.example 10): (9) Ke [wena]i akakuinikanga okunjalo uYehova uThixo wakho "But as for the tree of the knowledge of good and evil, you shall not eat of it" (Gen 2:17) Overall, the evidence provided in section 4 demonstrates that CLD complies with the prototype of LD.The construction not only fulfils formal criteria (cf.section 3.1), but also, as demonstrated in this section, complies with functional conditions enabling it be classified as an instantiation of LD.As for the former, the locution consists of a nominal phrase dislocated from the matrix clause and co-indexed with the resumptive element in the matrix clause that indicates the syntactic and functional role of the dislocated referent.As for the latter, the construction is often employed in order to convey the [R+I] function, thus (re)activating a low accessible referent and specifying its syntactic and pragmatic role.It also exhibits other functions associated with LD crosslinguistically.It is therefore not surprising that CLD in Xhosa is commonly used to render the LD construction in Biblical Hebrew in translation of biblical passages.

Discussion
CLD in Xhosa can be used as a LD construction -that is, it can behave as a LD prototype complying both with its formal and functional properties.However, the systemic status of the construction that surfaces as CLD is more complex.This stems from the fact that, from a morpho-syntactic perspective, the same construction can be used for entirely different purposes.
In order to differentiate between these two varieties, we will refer to them as C(onstruction)1 and C(construction)2.The former is CLD, which has been discussed in section 4 and which complies with the LD prototype.The later, albeit formally analogous, does not constitute a case of LD from a functional point of view.
C2 principally functions to communicate the idea of FR(onting) and to convey two functions crosslinguistically associated with it: T(opicalization) and F(ocalization).In contrast to LD, FR consists of moving one constituent of the clause (in our case an object) to the initial position in order to (re)introduce a topic (topicalization) or focus (focalization), as in the English sentence Marry I saw (compare LD Marry, I saw her).
The instances of FR and the usage of C2 in order to convey the T/F functions can be illustrated by analyzing Biblical translations.That is, C2 is employed to render the constituents that in Biblical Hebrew have been fronted and act either as topic (12) or focus (13).In all such cases, the underlying Biblical Hebrew text does not employ an appropriate LD construction (contrary to all the cases discussed in section 4).Rather, fronted construction are used (see Van der Merwe forthcoming).Formally, the most explicit indicator of this in Biblical Hebrew is the absence of resumption and, arguably, movement of a constituent.Functionally, these examples do not (re)activate a referent that entertains a low degree of accessibility.On the contrary, the referent is already discourse active and there is no need for mitigating the complex [R+I] function.
In example 11, the element that occupies the left periphery in Biblical Hebrew ( ‫ים֙‬ ‫שִׁ‬ ‫ַגְ‬ ‫יל‬ ֽ ‫פִּ‬ ‫הַ‬ ‫֤י‬ ‫נֵ‬ ‫בְ‬ ‫לִ‬ ‫וְ‬ 'sons of the concubines') is employed to compare characters or topics that have been introduced previously and that are discourse active (Van der Merwe forthcoming).Moreover, there is no resumptive pronominal specifying the role of that noun -its role is specified by the preposition that governs it.As a result, in the Biblical text, this noun/prepositional phrase is not dislocated but fronted.Pragmatically, it is used in the T function.In order to render this expression, Xhosa employs a construction similar to that found in Biblical Hebrew, namely by resorting to the ante-position of the topic (oonyana bamashweshwe 'sons of the concubines').However, contrary to Biblical Hebrew, the Xhosa text also makes use of a resumptive element because such an element is compulsory given the movement of the constituent out of its canonical position.This means that fronting in Xhosa (and its use for topicalization) is formally indistinguishable from CLD (and its use for the "But to the sons of the concubines, that Abraham had, Abraham gave gifts" (Gen 25:6) As mentioned above, FR in Biblical Hebrew can also be used to communicate the idea of focus.An example of this is Deuteronomy 6:13, where the ante-posed constituent ‫י�‬ ֛ ‫�הֶ‬ ֱ ‫א‬ ‫֧ה‬ ‫הוָ‬ ‫ת-יְ‬ ‫אֶ‬ 'Lord, your God' appears in the focal function.That is, it is God, him only and especially, whom the addressee should fear.This Hebrew construction (which in Hebrew is not a case of LD but of Fronting with no resumption and with movement) is rendered in Xhosa by a locution that is formally analogous to CLD.Once more, this similarity stems from the fact whereby the moved constituent requires object agreement.As a result, any type of fronting, either for the purpose of topicalization or focalization, results in Xhosa in object agreement, rendering the whole morpho-syntactic construction undistinguishable from CLD.All of this means that in Xhosa, a construction that structurally appears as CLD (i.e. the summation of C1 and C2) is more than a LD sensu stricto.It exhibits a systemic status that results not only from its LD origin but also from the advancement of the object agreement cycle experienced by the Xhosa language.To be exact, on the one hand, CLD (C1) grosso modo persists as such, without suffering major modifications since its origin27 -its form complies with the LD prototype and its function likewise fulfils the functional prototype of LD.On the other hand, as OA advanced on its grammaticalization path and became nearly obligatory (being especially compulsory for any extrapolation including fronting), topic or focus fronting requires a resumptive element on the verb.Consequently, fronted construction (C2) became identical to CLD (C1).In other words, a structure with a nominal referent placed in the left periphery and co-indexed with the pronominal object marker agglutinated to the verb can currently convey both [R+I] function (a successor of the original CLD construction -the C1 variant; cf.section 4) and T and F functions (a successor of the original fronted construction, which now requires object agreement -the C2 variant). 28he phenomenon whereby the status of the broader CLD structure has been altered not because of its own properties (as explained the behavior of the original CLD [C1] construction has not changed substantially), but rather because of a distinct evolutionary process, i.e. the progression along the object agreement cycle (and the formation of the C2 variant due to the obligatory use of agreement in fronted constructions) constitutes a case of mosaic evolution.As the two, initially independent, processes combine, they deliver a new global pattern and, thus, a new interpretation of a CLD construction -a summation of C1 and C2 -altering the systemic status of the original CLD form.Nowadays, because of OA, this structure is associated both with the CLD prototype ([R+I]) and with Fronting (Topic/Focus).29

Conclusion
The present study demonstrated that the current systemic status of the CLD structure can be explained as a case of mosaic evolution.Accordingly, the CLD form emerges as an accumulation and combination of two more individual, distinct and, at least, initially separated developments and characteristics -LD sensu stricto and object agreement.It is not only the evolution of LD as such but also the evolution along the object agreement cycle (which constitutes a different dynamic phenomenon) that shaped the present nature of CLD in Xhosa.
This enables us to propose a possible solution to the problem whether the CLD structure in Xhosa (and Nguni) is an instantiation of LD (the [R+I] function) or FR (the T/F functions).The CLD structure is employed to accomplish the two tasks.The former is derived from the original LD construction, while the latter arose due to the OA cycle.
This mosaic character of LD in Xhosa, in turn, demonstrates that LD operates not only due to its own evolutionary baggage (imposed by the source from which it has developed) but also due to the dynamics of the environment in which it has been embedded.As anything in language (and nature in general), LD is a situated "individual".It emerges and runs in response both to its intrinsic traits and to properties of the external environment. 30In some analyses (as the one presented in this paper), a portion of this exogenous nature of the individual can be interiorized by it.That is, certain relations, by which the individual is coupled to the external environment can be regarded as endogenous, i.e. as inherent to that individual and originating from within it.
Although the evidence and discussion built around it answered our research question, this article did not respond to all possible issues.Among such possible limitations, two seem to be the most important.First, in this study, prosodic information has been ignored and, second, the behavior of other types of LD and agreements has not been analyzed.
As for the former issue, although morpho-syntactically C1 and C2 are identical and, thus, both the category of LD (R+I) and FR (F/T) can be expressed by the same structure, it is sometimes possible to differentiate between the two variants.Namely, it has been argued that in Xhosa, CLD is related to the phonological phrase boundary (Zerbian 2004: 72).This boundary, which shows the right edge of the phonological phrase, is indicated by the lengthening of the penultimate vowel and a tonological change, as well as arguably by pause (Jokweni 1995;Zerbian 2004: 73-74).Accordingly, CLD has the following effect on the dislocate: the penultimate syllable of the dislocate surfaces as long, the word-final high-tone is preserved and the pause may be observed: [incwaːdí]i bayayiivula 'As for the book, they open it' (Zerbian 2004: 74, 77). 31A possible fourth phonological feature could declination reset (Couper-Kuhlen 1996, 2001). 32Consequently, the four diagnostics for the boundaries of a phonological phrase can be used to determine if an element has been dislocated or not (Zerbian 2004;consult also Jokweni 1995: 55;Zerbian 2007: 257, Downing 2011;Feldhausen and Patin 2008). 33 Supposedly, when the same structure is used to communicate Fronting -and where the constituent is moved but not dislocated -the three features would be absent or, at least, less evident. 34  As for the latter issue, the systemic status of CLD in Xhosa may be related to other types of LD -especially, to one in which the dislocated referent is specified as the subject in the matrix 31 An explained in section 2.1.1,pause is one of the most typical features of LD crosslinguistically. 32 Declination units are larger than intonation units.Intonation units are embedded in a declination unit.The former exhibit different shapes depending on their relative position in the latter (Couper-Kuhlen 2001:17).This complex nature of the phonological marking of LD implies its gradience.That is, an intonation pattern can range from being more prototypical of LD (where the extra-clausal position is marked more efficiently) to being less prototypical.
The length of the pause in LD can also range from more extensive to less extensive, additionally contributing to the gradience of LD phonetics. 33This observation on the boundaries of a phonological phrase seems to hold true for Chichewa.However, Bantu languages, in general, differ with respect to the requirement of a phrase break after the dislocated element in CLD.
For the sake of a crosslinguistic comparison one should note that Romance languages generally use a boundary in CLD constructions (Downing et al. 2004, Delais-Roussarie, Feldhausen andPatin 2013). 34A support for this may be found in Zulu.According to Cheng andDowning (2009) andFeldhausen andPatin (2008), although CLD in in Zulu exhibits pausal intonation and the penultimate lengthening (gama-bhayisekiːlii siwainike abantwana 'We gave bicycles to the children'), this lengthening is generally optional.To be exact, when such constructions indicate emphasis of the lexical object and its focal function (typical of Fronting, i.e. cases where the nominal object is moved out of the canonical position), pausal intonation may be absent.Inversely, if the lexical object is accompanied by pausal intonation, the emphasis effect is missing (Du Plessis 2010:26, 141).Thus, both the pause and the lack of emphasis may suggest a dislocation effect in contrast to Fronting.A provisional phonetic analysis of the examples discussed in section 4 seems to be consistent with this proposal.A native speaker interviewed on this account used pause to mark dislocation of a noun phrase exhibiting both the penultimate lengthening and high-tone preservation, corroborating the view expressed by Zerbian (2004).However, the phonetic properties of the fronted variant (C2) require further investigation.
clause.This class of LD can in turn be connected to the phenomenon of subject agreement in an analogous manner to the relation of CLD to object agreement.In fact, since Xhosa (and Nguni and Bantu in general) are more advanced on the subject agreement cycle than on the object agreement cycle, the confusion of subject-LD with subject agreement may be greater than the similar confusion which exists in the case of CLD and object agreement.
Since these two issues -that is, the prosodic properties of CLD in comparison to object agreement or Fronting and the characteristics of subject-LD and its relation to subject agreement -may have important bearings on the status of CLD and LD in Xhosa, they will be researched by the authors of this article in the near future.