Ideophones as Linguistic “Rebels”: The Extra-systematicity of Ideophones in Xhosa 1 IDEOPHONES AS LINGUISTIC “REBELS”: THE EXTRA-SYSTEMATICITY OF IDEOPHONES IN XHOSA – Part II

This paper contributes to the study of the structural distinctiveness of the category of ideophony. The author analyzes the extent to which Xhosa ideophones exhibit the so-called extra-systematic properties, which cross-linguistically tend to distinguish ideophones from other lexical classes. The analysis demonstrates that ideophones are relatively extra-systematic in Xhosa, although their extra-systematicity is not unitary. It is the largest in morphology, slightly less visible in phonology, and only residual in syntax. It is proposed that the distinct degrees of extrasystematicity are related to differences in grammaticalization and a gradual integration of ideophones into the Xhosa grammar – with the adjustment in syntax occurring faster than the morphological adaptation. In this paper – the second in a series of two articles – the author introduces evidence related to syntax, answers the research question, and explains the contributions of this research to the general theory of ideophony.


Introduction
The present article is concerned with the structural distinctiveness of ideophonesor their extra-systematicity. To be exact, I examine whether Xhosa ideophones exhibit extra-systematic properties that are associated with the prototype of an ideophone and determine the precise extent thereof. This that related to syntax (Section 2), and next holistically discuss all my dataphonological, morphological, and syntacticwithin the adopted framework (Section 3). I answer my research question (To what extent do Xhosa ideophones comply with the structural prototype of an ideophone and its extrasystematicity?) and show how my results contribute to the general theory of ideophony. I end the article with formulating my main conclusions (Section 4). 5

Extra-clausality (S-1)
Ideophones in Xhosa may occupy an extra-clausal position, appearing at the sentence boundaries. In such cases, they are usually separated from the core clause by a perceivable pause: (1) a.
Qhwi , z-aphuk-ile iivazi crush [pause] 10.SA-be.broken-PERF 10.vases [sound of glass breaking], the vases are broken However, in most cases, ideophones appear inside the clause, occupying a clause-internal position. In the most canonical situation, an ideophone appears as a complement of the verbs thi or tsho, with which it forms a verbal predicate (see Sections 2.2 and 2.3 below; consult also the first part of this study 5 As in the first part of this study, whenever possible, I will compare the properties of ideophones in Xhosa with those found in a closely related language -Zulufor which research on ideophones is significantly more advanced. See FIVAZ, D. Some Aspects of the Ideophone in Zulu; VOELTZ, E. Toward the Syntax of the Ideophone in Zulu. In CHIN-WU, K., STAHLKE, H. (eds.). Papers in African Linguistics, pp. 141-152; VON STADEN, P. Die ideofoon in Zulu; VON STADEN, P. Some Remarks on Ideophones in Zulu. In African Studies, 1977, Vol. 36, No. 2, pp. 195-224; CHILDS, T. Where Have All the Ideophones Gone? The Death of a Word Category in Zulu. In Toronto Working Papers in Linguistics, 1996, Vol. 15, p. 81-103; MSIMANG, C. T., POULOS, G. The Ideophone in Zulu: A Re-examination of Conceptual and Descriptive Notions. In VOELTZ, E., KILIAN-HATZ, C. (eds.). Ideophones,; DE SCHRYVER, G-M. The lexicographic treatment of ideophones in Zulu. In Lexikos, 2009, Vol. 19, p. 34-54. published previously). 6 In such instances, the ideophone is not separated from thi or tsho by a pause. 7 Typically, the ideophone follows the verb thi/tsho and, if applicable, precedes internal arguments (i.e. a primary (iemele 'bucket' in 2.a) or a secondary object (utata 'father' in 2.b)) and adjuncts (e.g. a locative phrase (e-Kapa 'to Cape Town' in 2.c) or an adverbial (kaninzi 'many times' in 2.d)). Often, both an argument and an adjunct follow the ideophone (iemele 'bucket' and etafileni 'on the table' in 2.e; review also various examples introduced in the first part of this study).
Ndi-yi-th-e xhuzu kaninzi namhlanje 1 st SG.SA-9.OA-THI-PERF pull many.times today I pulled it many times today 6 ANDRASON. A. Ideophones as Rebels: The Extra-systematicity of Ideophones in Xhosa. Part 1. In Asian and African Studies, 2020, Vol. 29, No. 2;DU PLESSIS, J. A. IsiXhosa 4, p. 290;GXOWA, N. C. Ideophones in Xhosa;NOKELE, A. The Syntax of the Ideophone in Xhosa. 7 If one compares examples (1.a-b) with (2.a-e), Xhosa ideophones suggest an inverse relation between expressiveness and morphosyntactic integration. That is, in an extraclausal position, ideophones are separated from the clause by a pause. In contrast, in a clause-internal position, they tend to be pronounced as a single unit with the verb thi. This corroborates an inverse relation between expressiveness and morphosyntactic integration as proposed by DINGEMANSE (2017)  The predicative ideophone cannot be moved to an extra-clausal position in left dislocation constructions or to a clause-initial position in fronted focus-topic constructions (see 3 below). This behavior contrasts with other complements of verbs, whether arguments or adjuncts, which can be left-dislocated or fronted.
(3) *Jwaxa u-th-e utata give 1a.SA-THI-PERF 1a.father inkwenkwe imali 9.boy 9.money Indented meaning: Father gave the boy money If the verbs thi and tsho are absent, ideophones may occupy a clause-initial (4.a) or a clause-medial position (4.b). 8 In intransitive uses, the clause-initial position is the most common, contrary to the prevalent type of word order found with intransitive verbs, namely SV. Nevertheless, even here, the ideophone is not separated from the adjacent elementtypically its compulsory subjectby the contouring or pause, which tend to be found at clause boundaries. The ideophone and the following element are rather pronounced as a single clause. In transitive constructions, the ideophone is most frequently found between the subject and the object, thus attesting to a clause-medial position (4.c). However, it may also be found in a clause-initial position, with the subject appearing after the object (4.d).
(4) a. To conclude, Xhosa ideophonesboth those that are accompanied by the verbs thi/tsho and those that are used without ittend to appear within a clause frame, occupying an initial or medial position.

Complete utterance (S-2)
In Xhosa, ideophones may constitute complete autonomous utterances. This is relatively common in cases of onomatopoeic ideophones that reproduce or imitate sounds from the real world (5.a-b). However, less onomatopoeic ideophones may also be used holophrastically (see 5.c and especially 5.d).
- Although holophrasticity is attested, much more commonly, Xhosa ideophones appear as parts of clauses or sentences, thus failing to constitute separate utterances themselves. In such instances, they are used as predicates (qolokotho '(they) enter' in 6.a), parts of the predicate (athe khwashu 'they jumped quickly' in 6.b)the entire predicate being formed by thi/tsho and the ideophoneor adverbs (qho 'continuously' in 6.c) (see Sections 2.1 above and 2.3 below). 12 (6) a.
Imvula i-na qho e-busika 9.rain 9.SA-rain continuously LOC-14.winter It rains continuously in the winter Crucially, a large set of ideophones that allow for both intransitive and transitive (often causative) interpretations necessitates a clausal structure to disambiguate their meaning. Such ideophones are, inherently, neither transitive nor intransitive (see guqu 'turn something aside / be turned aside' in 7.a-b). Rather, their valency relies on the type of a clause into which they are integrated. If the object is overtly expressed, the ideophone is interpreted as active, transitive and/or causative (7.a). In contrast, in object-less clausal structures, the interpretation of the ideophone is passive, intransitive/detransitive, often accompanied by the nuance of resultativity (7.b

Asyntagmaticity (S-3)
As is evident from the above discussion, ideophones are not inherently asyntagmatic in Xhosa. On the contrary, ideophones commonly enter in various types of syntactic relationships with the other elements of the clause and sentence.
It has already been mentioned that Xhosa ideophones appear in two main functions: predicative and adverbial. 16 When constituting the predicate, or when forming the predicate together with the verbs thi and tsho, ideophones determine the argument structure of the clauseor its potential scope, if more than one valency pattern is possible (see Section 2.2 above). Accordingly, ideophones project the external argument (i.e. subject; bafundi 'students' in 8.a) and internal arguments, e.g. primary (isinambuzane 'insect' in 8.b) and secondary objects (ititshala 'teacher' in 8.c).   , 1977, Vol. 36, No. 2, pp. 195-224; CHILDS, T. Where Have All the Ideophones Gone? The Death of a Word Category in Zulu. In Toronto Working Papers in Linguistics, 1996, Vol. 15, pp. 81-103; MSIMANG, C. T., POULOS, G. The Ideophone in Zulu: A Re-examination of Conceptual and Descriptive Notions. In VOELTZ, E., KILIAN-HATZ, C. (eds.). Ideophones, c.
Abantwana ba-the-el-e gqada 2.child 2.SA-THI-APPL-PERF jump.hurriedly iilekese 19 10.sweets Children jumped hurriedly for sweets Moreover, ideophones enter into relationships with adjuncts, e.g. prepositional phrases expressing location (elwandle 'into the sea' and eKapa 'in Cape Town' in 10.a-b) or goal/theme (kum 'to me' in 10.c), and copulative phrases (ngamasela 'by thieves' in 10.d). 20 This applies both to predicative complexes with thi/tsho and to cases where ideophones are used without an introductory verb. Impahla i-th-e vuthu nga-masela 21 9.clothing 9.SA-THI-PERF be.finished 6.COP-6.thief The clothes are depleted by thieves The firmest syntagmatic relationship links the ideophone to the verbs thi and tsho, with which it forms a complex predicate. As already explained, ideophones need to be introduced by thi or tsho if the subject or object affixes are to be overtly expressed. Furthermore, the verb thi must be used if specific temporal, aspectual, or modal nuances are to be made explicit, or when the construction requires a negative polarity value. 22 Similarly, if the sense of reflexivity, reciprocity, and passivity need to be expressed, thi is regularly found. In all such cases, inflectional and derivational categories are specified morphologically by the verb thi or, much less commonly, tsho, while the ideophone itself determines the lexical meaning of the predicate, the scope of its argument structure (e.g. intransitivity, mono-transitivity, or di-transitivity), and the potential argument-structure alternations in case of derivational extensions. For example, in (11.a), the ideophone shwaka expresses the action of disappearingwhich typically involves one participant, i.e. the experiencerand, consequently, projects a single argument, i.e. the external subject argument. The verb thi locates the action in the near past, additionally profiling its current relevance by means of the ILE tense. In (11.b), the meaning of the ideophone xhakamfu 'hold' involves two participants, the root often being transitive. As a result, it is possible to use the reciprocal affix an 'each other' on the verb thi, if two subjects exert the action of holding one another. Additionally, by means of its inflection, the verb thi overtly places the activity expressed by the ideophone in the future. 21 GXOWA, N. C. Ideophones in Xhosa,p. 94. 22 The verb tsho can only be used in the A past tense with ideophones. See the first part of the study: ANDRASON. A. Ideophones as Rebels: The Extra-systematicity of Ideophones in Xhosa. Part 1. In Asian and African Studies, 2020, Vol. 29, No. 2. (11) a. Inkwenkwe i-th-e shwaka 9.boy 9.SA-THI-PERF disappear The boy has disappeared b.
USipho no-Landile ba-za-th-ana 1a.Sipho and-1a.Landile 2(a).SA-FUT-THI-REC xhakamfu hold Sipho and Landile will hold each other The close syntactic relationship coupling the ideophone and the verb thi/tsho is evident in the contiguity of these two elements. That is, the word order of this predicative complex is rigid, and the chain is unbreakable. For instance, the sequence [thi/tsho + ideophone] cannot be interrupted by arguments or adjuncts. In canonical word order, the subject regularly precedes the verb thi/tsho, while the internal arguments and adjuncts follow the ideophone (see various examples introduced thus far). 23 Specifically, if an object (uSipho 'Sipho' in 12.a) or two objects (inkwenkwe 'boy' and imali 'money' in 12.b) are to be expressed, they must appear after the complex predicate, and hence after the ideophone. Similarly, prepositional adjuncts and adverbs cannot be placed between the verb thi and the ideophone (egumbwini 'in the room' in 12.c). Furthermore, in case of subject-verb inversion, the subject must occupy the position after the entire [thi/tsho + ideophone] chain. 24 This occurs both in intransitive (see amakhwenkwe 'boys' and imvula 'rain' in 12.c-d) and transitive constructions (see uZola 'Zola' in 12.e). 25 When the arguments or adjuncts intervene between thi/tsho and the ideophone, the two verbs regain their literal lexical value 'to say; do'. In other words, the ideophone is no longer the complement of thi/tsho.

Verba dicendi or quotative markers (S-4)
I have mentioned several times that Xhosa ideophones, specifically those used in a predicative function, tend to be introduced by the verbs thi and, much less commonly, tsho. The verb thi is highly polysemousthe Great Xhosa Dictionary distinguishes thirty-three senses that thi may express. 31 Nevertheless, it is plausible that the entire semantic potential of thi has arisen from one sense through a chain (or network) of metaphorical and metonymical meaning extensions. Currently, one of the most pervasive groups of senses conveyed by thi -i.e. 'say, produce a sound', 'call', 'state, declare'is related to its usage as a verbum dicendi. 32 This usage is illustrated below: (14) a. USipho u-th-e ma-nd-enz-e 1a.Sipho 1a.SA-THI-PERF must-1 st SA-do-SUBJ ikofu 9.coffee Sipho said (that) I must make coffee b.
U-thi-ni? 2 nd SG.SA-THI-what What are you saying?
In some ideophones, especially those of an onomatopoeic character, thi may still express this speech-related nuance, functioning simultaneously as a lexical speech verb (15.a) and the carrier of inflections and derivations (15.b). In its usage as a speech verb, thi is usually marked by contouring and/or pause that separates it from the ideophone, as well as by special expressive phonation  (15) a.
Inkomo i-th-e mhu-u-u 9.cow 9.SA-THI-PERF moo The cow lowed Apart from its speech-or sound-related meaning, the verb thi may also be used as a verbum facendi. In such cases, it communicates a set of imitative meanings, e.g. '(do) like this, imitate', and is often accompanied by gestures (16.a-c). 34 The use of thi with ideophones that exploit less direct soundmeaning associations might have been prompted by such an imitative value of thi (16.d). The exact origin of thi in Bantuand thus the diachronic relationship between its various sensesis still debated. The traditional viewwhich is now rejected 38links the origin of thi to the meaning 'say'. 39 This hypothesis was already revised by Guthrie, who related Proto-Bantu *-tì (or *-TÈ) to 'that, namely', from which the speech value has only posteriorly been derived. 40 Current scholarship generally favors two postulates of the diachronic source of thi: thi derives either from a deictic (focal) element indicating manner 'thus, like this' 41 or from a generic action verb 'do'. 42 According to both proposals, the use of thi as a speech and quotative verbcurrently pervasive in Xhosais secondary.
As has been explained in the first part of this study, the use of tsho with ideophones is much more restricted. The verb tsho is limited to cases involving the subject prefixes of the (remote) A past tense. Even then, my informants preferred the verb thi, using it either instead of or after tsho (17). 43 As was the case of thi, the verb tsho is highly polysemous. The Great Xhosa Dictionary lists sixteen senses expressed by tsho. In further similarity to thi, two pervasive clusters of meanings can be distinguished: speech-related (e.g. 'say, declare, tell, narrate, produce a sound') and performance-related (e.g. 'perform, do'). The former cluster can be encompassed under the term of verba dicendi, while the latter approximates the category of verba facendi. According to a current view in Bantu scholarship, tsho "emerged via the incorporation of the Common-Bantu previous-reference marker -o into the basic form thi". 44 Thus, the use of tsho with ideophones would arguably have the same origin as was the case with thi described above.

(17)
Inkwenkwe ya-tsho ya-thi 9.boy 9.SA.PAST-TSHO 9.SA.PAST-THI Dyumpu e-manzi-ni plunge LOC-6.water-LOC 'The boy plunged into the water' Even though the elements thi and tsho found with ideophones derive, most likely, from a verbum facendi or a deictic of manner, subsequently grammaticalized as a verbum dicendi, they need not exhibit those senses when heading an ideophone. Indeed, in most cases of their use with ideophones, due to grammaticalization, thi and tsho are lexically empty, 45 used merely as a carrier of inflectional, derivational, or polarity information/markers.
All ideophones that are used as parts of complex predicates may be introduced by thi or tsho. Nevertheless, as mentioned in the section dedicated to morphology in the first part of my study, 46 the presence of these two introductory verbs is not compulsory. In several cases, thi and tsho are (or may be) omitted. This applies to both intransitive (18.a) and transitive constructions ( The presence of thi and tsho is ungrammatical in instances where the ideophone functions as an adverb. In such cases, the ideophone appears on its own, in its bare form. 48 (19) U-hamb-e mbo 1.SA-go-PERF straight 'He went straight'

Accumulation (S-5)
Contrary to many other syntactic elements and/or lexical classes, ideophones can be accumulated and deliver long chains of ideophonic entities. 49 The accumulated ideophones may be formally identical (20.a), similar (20.b), or divergent (20.c-d). They may concern a similar semantic domain (20.b-c) or independent domains (20.d). This usage is especially common if ideophones are employed onomatopoeically (20.a-c) and/or if nuances of emphasis, consecution, or iteration need to be profiled. 50 Nevertheless, in less onomatopoeic uses, and in cases where ideophones function as true predicates, their accumulation into a series is also grammatical. 51 The other lexical class that allows for such accumulation is interjections. 52 It is possible that in some cases, such analytical sequences have given rise to synthetic chains, reanalyzed 48 This can be explained by the fact that, in such cases, the ideophone does not form part of a predicate. The absence of thi in those adverbial instances also suggests that the Proto-Bantu ancestor of thi was most likely a verbal base. This, in turn, provides some support for the hypothesis linking the origin of thi to a generic action verb. See DEVOS, M., BOSTOEN, K. Bantu DO/SAY Polysemy and the Origins of a Quotative in Shangaci. In Africana Linguistica, 2012, Vol. 18, pp. 97-132

Discussion
The evidence introduced in the empirical sections of my studyboth in its first and second partdemonstrates that the issue of the potential extrasystematicity of ideophones in Xhosa is complex and intricate. Therefore, to evaluate this evidence within the adopted framework and propose higher-level generalizations, a brief summary of the main findings is, in my opinion, necessary.
As far as phonology is concerned, ideophones generally do not involve sounds that are aberrant in Xhosa (P-1). However, the pervasiveness of clicks is unusual. Very rarely, truly extra-systematic sounds do occur, especially in onomatopoeic ideophones. Similarly, ideophones do not usually contain extrasystematic configurations of soundsthe noticeable exceptions being the use of consonants in a final position and the occurrence of words composed only of 53 WEAKLEY, A. J. An Introduction to Xhosa Ideophone Derivation and Syntax, pp. 64. 54 OOSTHUYSEN, J. C. The Grammar of isiXhosa, p. 354. consonants (P-2). Ideophones extensively exploit vocalic length (P-3). Four degrees of quantity are possible, i.e. short, long, extra-long, and extra-extra-long vowelsthe two last ones being unusual in Xhosa. Contrary to the rule in Xhosa, the long vowel (as well as the stress) may appear in syllables other than the penultimate one. Moreover, in ideophones, vowel length is exploited phonemically to a significantly larger extent than in other parts of the language. Infrequently, ideophones exhibit long consonants, which are generally absent in Xhosa. As is typical of Xhosa, tone is a regular characteristic of ideophones (P-4). Nevertheless, in ideophones, tone seems to play a more relevant role at the phonemic level. Furthermore, ideophones exhibit a strong preference for a low tone, with the tonal pattern [L-L] being predominant. Ideophones make use of consonantal and vocalic harmony to a significantly greater extent than any other lexical classes (P-5). They also utilize tonal harmony, which contradicts the tendency of tone alternation otherwise predominating in Xhosa. These harmony features often result in rhymes. Lastly, ideophonesespecially onomatopoeic onesmay be accompanied by distinctive phonation (P-6). To conclude, if treated in its totality, the category of ideophones exhibits a semi or moderately extra-systematic phonological profile. That is, even though various features exhibited by ideophones are not radically different from those found in other lexical classes, certain properties or the frequency of their occurrence are unusual and transgress the rules operating in the language.
As far as morphology is concerned, ideophones do not host inflectional markers (M-1). When ideophones constitute parts of the predicate, the verbal inflection is carried out by the verbs thi and, to a much more limited extent, tsho. 55 Exceptionally, in the A-tense, the contracted verb thi and the ideophone yield a more grammaticalized form, in which the ideophone is directly preceded by inflectional markers. In cases where the verbs thi and tsho are absent, inflections are not realized at all. Ideophones are not compatible with derivational morphemes that are otherwise typical of verbs and adverbsthe two functions in which ideophones are used (M-2). In cases where ideophones are employed as parts of the predicate, all the verbal derivative extensions appear on the verb thi. 56 In an adverbial function, a bare form of an ideophone is employed. Ideophones are often roots (M-3). The majority of ideophones are not derived from verbs or nouns. Even though it is still debatable and unsettled, the derivative relationship seems to be rather inverse. The only exceptions are 55 Although all inflectional categories can be used with the verb thi, certain differences with normal inflectional patterns occur. Tsho is only compatible with the subject prefixed of the remote A past tense. 56 In contrast to regular verbs, when heading an ideophone, thi cannot host the causative marker. Tsho cannot host any derivational extensions.
ideophones that are derived from verbs by means of the suffix -iyani. To a significantly greater extent than elsewhere in the language, ideophones make use of repetitive patterns, especially reduplication and multiplication (M-4).
Ideophones constitute an open and productive category. Their productivity is often spontaneous and idiolectal (M-5). Overall, the category of ideophones attests to a highly extra-systematic morphological profile, complying with most of the features postulated for the prototype. However, minimal signs of a lesser degree of morphological extra-systematicity (cf. the contracted form in the A tense and the derivation by the suffix -iyani) can also be detected.
As far syntax is concerned, ideophones tend to be found clause-internally, appearing without pause or contouring separating them from the other parts of the clause, especially from the verb thi or tsho. However, their extra-clausal use is also possible (S-1). Furthermore, even though ideophones may be used as autonomous utterances, they appear, much more commonly, as parts of clauses and sentences, i.e. as predicates (or their part) and adverbs (S-2). Inversely, ideophones are not inherently asyntagmatic. Instead, they tend to enter into various types of syntactic relationships with the other elements of the clause or sentence (S-3). As predicates, they may govern the subject, objects, and adjuncts. They also project thematic roles to internal and external arguments. As adverbs, they modify verbs and descriptives, e.g. adjectives and relatives. Overall, ideophones are well integrated into the clause. In the predicative function, ideophones are typically introduced by verba dicendi or verba facendi thi and tsho 'say / do' (S-4). However, in the same function, they may also appear without thi/tsho. In their adverbial uses, ideophones are not accompanied by thi/tsho. This means that even though the presence of thi and, less so, tsho is a characteristic feature of ideophones, it is not compulsory. Lastly, various ideophones may be accumulated (S-5). On the whole, the category of ideophones exhibits a (very) limited degree of syntactic extrasystematicity. Rather, in various instances, ideophones behave as other fully systematic elements of the language, i.e. verbs or adverbs.
The evidence summarized above demonstrates the following: First, the compliance of Xhosa ideophones with the formal extra-systematicity associated with the ideophonic prototype is uneven. Although various extra-systematic properties are present, others are absent or instantiated to a limited degree. Second, the extent of extra-systematicity is dissimilar in the three levels of the language. It is the most evident in morphology, slightly less palpable in phonology, and residual in syntax. Third, the extent of extra-systematicity may be distinct in specific ideophonic lexemes. That is, extra-systematicity is not unitary for all the members of the ideophonic category in Xhosa. Some ideophones are highly extra-systematic, be it phonologically, morphologically, or syntactically. Many may, in contrast, exhibit a limited degree of extra-systematicity. Several others are confined between these two limits. Even a single ideophone may be more or less canonical and thus extra-systematicespecially with regard to syntaxdepending on its use. As a result, when considered in its totalityi.e. if all ideophonic tokens are envisaged jointlythe category of ideophones in Xhosa spans a fragment of the available network of structural extra-systematicity. It covers the section that links canonical extrasystematicity located in the center of the network to more moderate extrasystematic profiles located in peripheral zones. In the former case, Xhosa ideophones comply with the behavior postulated for the crosslinguistic prototype. In the latter case, ideophones are taxonomically more distant from that prototype, inversely approximating the other, more systematic lexical classesverbs and adverbs. 57 Overall, ideophones seem to be relatively well integrated in the Xhosa language. First, their type and token frequency 58 are both significant. That is, the class of ideophones is numerous, and ideophonic lexemes appear commonly in both colloquial and more formal language, be it a spoken or written variety (e.g. in comics). Second, even though ideophones distinguish themselves from the other elements of the language as far as their morphological behavior is concerned, they seem fairly systematic from a syntactic perspective. In other words, while their morphological extra-systematicity is evident, the extrasystematicity of ideophones is only minimal in syntaxor it is optional. To be exact, even though ideophones may be used as holophrastic, extra-clausal, and/or asyntagmatic elements, their non-holophrastic, inter-clausal, and syntagmatic usage is also possible, and is, in fact, prevalent.
The above factsi.e. the frequency of ideophones and their extent of integration in the languagesuggest a semi-advanced grammaticalization stage of the category of ideophony in Xhosa. On the one hand, type and token frequencies are generally viewed as overt signs of an increase in grammaticalization, being a concomitant phenomenon to it. 59 On the other hand, the syntax of a language is affected by innovations and changes more rapidly than morphology. This stems from a unidirectional progression operating in all languages whereby morphology develops from syntax. That is, synthetic boundedness constitutes a subsequent developmental stage of analytical unboundednessthus morphology (or a tighter structure) evolving from syntax (a looser structure). 60 As accurately captured by Givón "[t]oday's morphology is 57 Compare with similar results for Zulu (VON STADEN, P. Some Remarks on Ideophones in Zulu. In African Studies, 1977, Vol. 36, No. 2, pp. 195-224). 58 Cf. HOPPER, P., TRAUGOTT, E. C. Grammaticalization, p. 127. 59 Ibid., p. 126. 60 Ibid., pp. 7, 26, 32, 100. yesterday's syntax". 61 Therefore, the assimilation of a category to a given language should proceed more rapidly with respect to syntax than morphology. Indeed, in Xhosa, the adaptation of ideophones to syntax, or their syntacticization, is well advanced. In contrast, the adaptation of ideophones to the morphology of the language, or their morphologization, is significantly less advanced. The initial stage of this process is only visible in the coalesced form of the A past tense, where ideophones appear superficially as if inflected.
The study of ideophones in Xhosa has, at least, four implications for the general theory of ideophony. First, the extra-systematicity of the category of ideophones in a language cannot be estimated in binary terms, i.e. as either extra-systematic or systematic, or even as a single point in a network linking genuine extra-systematicity with the absolute lack thereof. Rather, the category of ideophones in a language spans a section of that network, attesting to cases that are more extra-systematic (these cases can be located in the center of the network) and to those that are less extra-systematic (these cases radiate towards the periphery of the network). Overall, languages may differ in the range of the section covering the network of extra-systematicity or that section's precise boundaries; and in the extent to which determined parts of that fragment of the network are populated by ideophonic lexemes (i.e. how typical ideophones of determined (extra-)systematicity degrees are in a language). 62 Second, the extra-systematicity of ideophones is not only a question of compliance, or not, with the prototype of extra-systematicity posited on typological grounds. It also derives from the idiosyncrasy of the language in which ideophones are used. That is, the compliance with a given feature need not be a sign of extra-systematicity, since other components of the language may also make use of that particular device. Put differently, even though a feature is met, it need not be extra-systematic in a particular language. For example, if the language makes extensive use of tone, vocalic length, marked consonants such as clicks, and reduplication, or if it does not use any inflectional and derivational morphemes, the presence of tone, length, clicks, and reduplication in ideophones on the one hand, and their incompatibility with inflections and derivations on the other hand, are all insignificant for the estimation of ideophones' extra-systematicity.
Third, the different extents of the extra-systematicity of ideophones exhibited in a language and/or across languagesand thus different positions 61 GIVÓN, T. Historical Syntax and Synchronic Morphology: An Archaeologist's Field Trip. In Chicago Linguistic Society, 1971, Vol. 7, p. 413. 62 For a similar observation regarding Siwu see DINGEMANSE, M. Expressiveness and System Integration. On the Typology of Ideophones, with Special Reference to Siwu. In STUF -Language Typology and Universals, 2017, Vol. 70, No. 2, p. 377. occupied by them in the network of extra-systematicityarguably attest to different stages of grammaticalization and the gradual integration of ideophones into grammar. During this integrational grammaticalization process, ideophones that are para-grammatical or para-lexemic, gradually evolve into items that are grammatically and/or lexically more canonical. My research suggests that this adaptation is first visible in syntax from which it spreads to morphology. This conclusion would corroborate the evolutionary view on the integration of ideophones in languages. According to that hypothesis, ideophones begin their grammatical life as an independent class characterized by a highly extrasystematic profile. Subsequently, they are grammaticalized, in a gradual manner, into other classes (mostly verbs, adverbs, and adjectives), losing, at the same time, their typical ideophonic profile and thus lowering the degree of their extra-systematicity. 63 The final stage of this process may be a total diffusion of ideophones into another category (or other categories) and their indistinguishability from the other members of the hosting class(es). 64 Xhosa arguably attests to an advanced, albeit non-final stage of this process, where "the ideophonic words augment the class into which they are [being] integrated". 65 That is, ideophones increase the size of the class of verbs and adverbs, while still maintaining their own categorial integrity.
Fourth, the results of my study corroborate a potential relationship between the frequency of ideophones and the increase of their morpho-syntactic integration or deideophonization. 66 That is, the more numerous and the better represented the category of ideophones is, the more likely it is to undergo the process of adaptation to the general grammatical profile of a language, and, thus, to acquire grammatical features typical of other lexical classes. My study shows that at least as far as syntax is concerned, the large number of ideophones in Xhosaideophones being the third largest lexical class in this languageis correlated with their relatively low extra-systematicity and, conversely, advanced integration.
To conclude, the present study demonstrates that ideophones in Xhosa are, in several aspects, structurally distinct from other lexical classes. This corroborates, to an extent, the typologically driven hypothesis of the structural extra-systematicity of the ideophonic prototype. However, the issue of extrasystematicity has also proven to be more nuanced, as the extra-systematic profile of ideophones is not equal in the different modules of the Xhosa language. Overall, given my findings, and the evident integration of ideophones especially as (a subtype of) verbsin the clause syntax of Xhosa, 67 Nguni, 68 and (Southern) Bantu, 69 the postulate of the syntactic extra-systematicity of ideophones across languages and, thus, the ideophonic prototype itself, may be questioned. While the phonology and morphology of ideophones and their prototype may indeed be deeply extra-systematic, the extra-systematicity of their syntax need not be the case. This last statement, resulting from the present research, remains a working hypothesis, which needs to be verified by more indepth typological studies.

Conclusion
The present study analyzed the compliance of Xhosa ideophones with the extrasystematic properties that structurally distinguish the prototype of an ideophone from the other components of the language.
The analysis demonstrates that the extra-systematicity of ideophones is not unitary in Xhosa. It is the greatest in morphology, slightly less visible in phonology, and only residual in syntax. It also varies for specific ideophonic lexemes. When considered in its totality, the category of ideophones spans a fragment of the cline of extra-systematicity: it ranges from the pole of a fully extra-systematic profile to the stage in which extra-systematicity is only moderate. Overall, ideophones constitute a characteristic and common feature of Xhosa. This explains their relative integration into the Xhosa language systemthe adjustment being profound in syntax, albeit only minimal in morphology.
In light of my findings, I propose that the extra-systematicity of the category of ideophones should be represented as a fragment of a network rather than a single unitary value; that different degrees of extra-systematicity available on that network attest to different stages of grammaticalization and, in particular, to a gradual integration of ideophones into the grammar of a hosting languagewith the adjustment to syntax occurring faster than a morphological adaptation; that the extent of extra-systematicity of ideophones heavily relies on the properties of a particular language-system, not only on complianceor non-compliancewith the prototype; and that the syntactic extra-systematicity of ideophones and their prototype may indeed be largely exaggerated.