Where to go at the end: polylexicalization and polygrammaticalization of KAZ "edge" in Korean

Размещено на http://www.allbest.ru/

Where to go at the end: Polylexicalization and polygrammaticalization of Kaz 'edge' in Korean

Seongha Rhee

Abstract

Nominal lexemes undergo in history extensive development in the lexical as well as grammatical domains. Their semantic change involves diverse threads of conceptualization which shows significant aspects of language development. Despite the series of intriguing changes in form and meaning that the Late Middle Korean lexeme kAz `edge' has undergone, encompassing polylexicalization and polygrammaticalization that merit an in-depth analysis, it has not yet received earnest attention to date. Thus, this paper aims to fill the research gap by analyzing the lexeme's development from Late Middle Korean to Modern Korean from the grammaticalization perspective, drawing upon data from a historical corpus and dictionaries. An extensive diachronic data analysis based on concordance and lexical searches shows that new meanings or functions are so diverse that some of them even form an antonymic relation. Such a wide-ranging semantic and functional diversity is attributable in part to the role of the participating forms in word formation processes such as derivation and compounding but, more importantly, to differential conceptualizations of the source meaning `edge' e.g., entity-internal and entity- external conceptualizations; mapping the `edge' onto different continua, e.g., degree, path, likelihood; and involvement of subjectification, e.g., evaluative-epistemic judgment such as counter-expectation. The innovated meanings resulting from such cognitive operations form a conceptual network. The developmental processes of the lexeme kAz `edge', in general, can be explained with reference to grammaticalization mechanisms, such as desemanticization, extension, decategorialization, and erosion, and some mechanisms are also operative in lexicalization as well. However, erosion, or the reduction of phonetic volume, is not prominent in these changes, suggesting that the principles should be interpreted as tendencies rather than deterministic diagnostics. These findings contribute to a better understanding of the complex relationship between lexicalization and grammaticalization.

Keywords: grammaticalization, polylexicalization, polygrammaticalization, Korean, conceptual motivation, subjectification © Seongha Rhee, 2022

Аннотация

Полилексикализация и полиграмматикализация корейской лексемы Kaz 'край'

Санха РИ

Лексемы номинального типа переживают период интенсивного развития как в сфере лексики, так и грамматики. Их семантические изменения связаны с различными тенденциями концептуализации, указывающими на развитие существенных аспектов языка. Несмотря на ряд любопытных изменений в форме и значении лексемы kAz `край' в корейском языке позднесреднего периода, охватывающих полилексикализацию и полиграмматикализацию, которые заслуживают углубленного анализа, они до сих пор не получили достаточного внимания. Цель статьи - заполнить пробел в анализе развития лексемы kAz в корейском языке от позднесреднего до современного периода с позиций грамматикализации на основе данных из исторического корпуса и словарей. Расширенный диахронический анализ, основанный на сочетаемости и лексическом поиске, показывает, что новые значения и их функции настолько разнообразны, что некоторые из них иногда даже вступают в антонимические отношения. Столь широкомасштабное семантическое и функциональное разнообразие обусловлено отчасти ролью используемых форм в словообразовательных процессах, таких как деривация и словосложение, но, что еще более важно, дифференциальной концептуализацией исходного значения «край», то есть концептуализацией, направленной на внутреннюю или внешнюю сущность понятия «край» в разных континуумах, таких как степень, способ, сходство; и субъек- тивизацией, например, оценочно-эпистемическими суждениями, такими как обманутые ожидания. Новые значения, возникающие в результате таких когнитивных операций, формируют концептуальную сеть. В целом процессы развития лексемы kAz `край' могут быть объяснены с учетом механизмов грамматикализации, таких как десемантизация, расширение, декатегоризация и эрозия; некоторые механизмы также задействованы в процессе лексикализации. Однако эрозия, или редукция фонетического объема, не играет существенной роли в этих изменениях, в связи с чем данные принципы следует интерпретировать как тенденции, а не предопределяющие факторы. Результаты исследования позволяют лучше понять сложные взаимоотношения между лексикализацией и грамматикализацией.

Ключевые слова: грамматикализация, полилексикализация, полиграмматикализация, корейский язык, концептуальная мотивация, субъктивизпция

Introduction

It is an aphorism that linguistic form and meaning are in constant change, though unnoticeable unless seen from a historical perspective. The change may create new meanings resulting in polysemy (of a single word) or `heteronymy' (of multiple, related words across word classes; Lichtenberk 1991), or new grammatical functions, a process known as grammaticalization. A single word may display multiple lines of such changes over time. It is well known in grammaticalization studies that a single lexeme may develop into diverse grammatical markers across multiple functional domains (`polygrammaticalization' Craig 1991, Rhee 2019). Similarly, at the lexical level, individual languages have different strategies for word-formation, often closely related to their typological features (Rhee & Khammee 2022), e.g., using a single lexeme for diverse extended meanings (`the polysemy strategy'), coining entirely new words (`the monosemy strategy'), systematically modifying an existing word, such as derivation (`the derivation strategy'), or creating multi-word expressions (`the phraseology strategy' or `the compounding strategy'). Languages may use different combinations of these strategies to different degrees. All these processes, by virtue of their end results being the creation of new lexical items, are subsumed under the lexicalization strategies. When a word is subjected to multiple lexicalization processes, this may, by analogy with polygrammaticalization, be referred to as `polylexicalization' (cf. Kou 2018).

The Korean nominal lexeme kAz `edge' has undergone change across diverse lexical and functional domains. In the earliest extant data written in Korean script Hangeul (= Hankul), kAz and kAs occur in a mixed way (see 4.2 below). A general pattern is that kAz occurs when followed by a particle without an onset consonant, and kAs occurs when standing alone or followed by a particle with an onset consonant (for discussion on various positional issues regarding /z/ and /s/, see Yi 2021). Evidently, kAc- (as in kAcang/kAcay `extremely, most') is a variant developed from kAz, as is often attested in historical change of certain lexemes, e.g., examples in (6) in 4.3.2 (Jung 1995, Yi 2021). Great spelling variability is a norm in historical data, a situation that continues until the implementation of the First Hangeul Orthographic Regulation (Hankul Macchwumpep Thongilan) in 1933. Unless warranted by the context, the lexeme is represented as kAz not specifying variants. The extent of change is so wide-ranging that it showcases a number of theoretically significant aspects in lexicalization and grammaticalization and, more generally, conceptual change. Despite such merits, the change of kAz `edge' has not received much attention to date and this paper intends to fill this research gap.

Thus, the objectives of this paper are twofold; to describe the lexicalization and grammaticalization paths of kAz `edge', focusing on the extension patterns and conceptual networks, and to discuss select issues that carry significance in grammaticalization theory, and more generally, language change. In particular, it intends to show that changes branching out of a source form are the results of divergent conceptual extension motivated by different conceptualizations of an `edge', and that subsequent changes, being contiguous to the previous stage, form a conceptual network.

This paper is organized in the following way. In section 2, we will briefly describe the Korean language with a focus on typological characteristics that are relevant to the change displayed by kAz `edge'. Section 3 deals with the data sources and research methodology. Section 4 describes the development of kAz from Late Middle Korean to contemporary Korean, focusing on the lexical origin (4.1), formal changes (4.2), polylexicalization (4.3) and polygrammaticalization (4.4). Section 5 discusses the noteworthy issues, i.e., multiple paths of language change (5.1), grammaticalization mechanisms (5.2), and conceptual motivation (5.3). Section 6 summarizes the findings and concludes the paper.

Typological characteristics and word-formation in Korean

Korean is spoken by 81.7 million people worldwide, of which 75.6 million are in the Korean peninsula (Eberhard, Simons & Fennig 2022). Genealogically, it has often been regarded as an Altaic language since eminent Finnish Altaic linguist Gustaf Ramstedt proposed it in 1939 (Ramstedt 1997 [1939]), but in the modern linguistic research tradition the Altaic hypothesis is often contested (see Kim 1992, Sohn 1999, among others, for discussion of the issue). Eberhard, Simons and Fennig (2022) list it as a Koreanic language along with Jejueo, a language variety spoken in Jeju Island. Most recently, it is frequently discussed under the geographically motivated classificatory term `Transeurasian languages' (Johanson & Robbeets 2010, Robbeets 2013), often along with Japanese since the two languages share many structural characteristics (Narrog & Rhee 2013, Robbeets 2017).

Typologically, Korean is an agglutinating, head-final, SOV language. In line with the SOV word order characteristics in morphology, it is suffixal and postpositional. Prepositions do not exist, and prefixes are relatively unproductive. Grammatical forms are predominantly suffixes and postpositions, as shall be seen with the development of kAz `edge'. Argument NPs are often omitted, and NPs may occur without their postpositional particles. The particles most susceptible to omission are primary case-markers, e.g., Nominative, Accusative (and, to a lesser degree, Dative and Allative), and those with more specific semantic content, such as those developed from kAz `edge', tend to be retained for preservation of meaning. Syntactic and thematic functions of NPs are marked with case-markers and postpositions. In terms of verbal morphology, finite verbs are marked with tense, aspect, mood, and modality, whereas non-finite verbs are often marked with `converb'-markers (converbalia, Ramstedt 1903), of which the main function is adverbializing a non-finite verb (Haspelmath 1995).

Relevant to our discussion is the verb formation strategy, i.e., the productive use of the light verb (ha- `do/be/say') for lexicalization of property words (e.g., anxiety-do > be anxious (adj.)), a feature widely shared among Transeurasian languages (Robbeets 2017: 595-596). The use of the light verb is also a common strategy for forming a verb). In such verb formation, the Accusative case-marker is deleted from the syntactic construction of [N-acc ha-] `do N', resulting in the verb [N-ha-], e.g., food-ACC do > food-do > cook (v.). In addition, the use of the verbs of existence (iss- `exist') and non-existence (eps- `not exist') in lexicalization and grammaticalization is also common. In such word formation, the Nominative case-marker is deleted from the syntactic construction of [N-nom iss-/eps-] `N exists/does not exist' (e.g., `taste exists/doesn't exist'), resulting in the verb/adjective [N-iss-] (e.g., `be tasty/be tasteless') (see 4.3.2 below for more).

Based on these typological and word-formation characteristics, this study aims to describe the lexicalization and grammaticalization patterns of the lexeme kAz `edge' and analyze their interrelation and conceptual motivations. As shall be made clear in the following discussion, the typological characteristics of headedness play a crucial role in the development of grammatical forms and functions.

Data and methodology

The primary data source for investigation of diachronic change patterns is the Sejong Historical Corpus, a 15-million word corpus derived from the 21st Century Sejong Corpus, a 200-million word corpus, developed by the Korean Ministry of Culture and Tourism and the National Institute of the Korean Language (1998-2006). This corpus encompasses data from the fifteenth century to the early twentieth century (1446-1912). The corpus consists of Buddhist scriptures, commentaries, edificatory textbooks, medical advisories, poems, diaries, interpreter training textbooks, bilingual dictionaries, Biblical texts, Christian commentaries, novels, short stories, newspaper articles, and personal correspondences, among others. The search engine used is UNICONC, developed by Jin-Ho Park of Seoul National University. All spelling variants used in historical sources are first identified and their attestations are retrieved from concordance searches using UNICONC. All concordance hits are individually examined to ensure their relevance by eliminating homonyms or spelling variations of irrelevant lexemes.

In addition to the corpus, a number of historical dictionaries have been consulted (see the list in References). These dictionaries contain extensive data compiled from diverse sources from Late Middle Korean (LMK; 15th - 16th c.) and onward. As is the case with concordance data, dictionaries contain headwords that are not related to the target item, largely because of the great level of spelling variation in historical data (see Figure 1 for all relevant forms). A thorough investigation into the various usages of the target lexeme and its derived forms manifested in the historical sources reveals now-defunct usages that are likely to have worked as bridges for the development of contemporary functions.

The development of kAz 'edge'

The lexical origin

The nominal kAz is attested in LMK, carrying the meaning of `edge' and related concepts (e.g., `shore, limit, boundary, end, etc.'), as exemplified below: The Extended Yale Romanization System (Rhee 1996) is used for romanization of the Korean data. Also, following the convention, the Chinese characters in historical sources are rendered with their Modern Korean pronunciation value, indicated in upper-case.

In contemporary Korean, the surviving nominal form of kAz is ka, which resulted from the loss of the coda consonant (see 4.2 below), as exemplified below The most common nominalizer (or dependent noun) in ModK is kes `thing', the origin of which is hypothesized as kes/kas `skin, surface' (Hong 1983). Semantic affinity between `edge, boundary' and `skin, surface' and formal similarity between kAz/kAs and kes/kas suggest a possibility that the two words are cognates. However, the connection has not yet been established.:

The form has undergone a series of changes in form and function in history. Some of the forms and functions are lost over time and are obsolete in contemporary Korean. We will now look at the changes in more detail, first formal changes, followed by semantic-functional extensions through lexicalization and grammaticalization.

Formal change

The LMK noun kAz, with the primary meaning `edge', has a few variant forms, e.g., kAz, kAs, kA, ka, and kas. In the earliest data, i.e., of the 15th century, it appears as kAz and kAs, which often have the same phonetic representation, i.e., [kwt], due to the general rule of final-consonant neutralization into an alveolar stop. When kAz is followed by a particle with no onset consonant, the /z/ is phonetically realized as [z].

In the 16th century /z/ begins to disappear as part of a historical change, adding kA to the inventory of variants, and the final consonant /z/ is no longer found in Early Modern Korean (EModK; 17th-19th c.) and onward. The coda /z/ is changed either to /c/ (usually when occurring word-medially), or /s/ (phonetically realized as [t], when occurring word-finally), or it is completely eroded. The vowel `a' /л/ in LMK is changed to `a' /a/ in Modern Korean (ModK; 20th c. - the present), and consequently the only nominal form surviving in contemporary Korean is ka /ka/ (apart from the derived nouns).

While the form is undergoing the reductive change of the loss of the coda, there also occur additive changes over time. Some of the major morpho-syntactic changes are as shown in (2):

a. suffixation of the diminutive -ang

compounding of (a) with cali `place'

syntactic combination of (a) with the light verb hA- `do/be/say'

derivation of (c) with a causative morpheme -i-/-y-

absorption of the genitive -s in the preceding nominal

syntactic combination with the verb ep(s)- `not exist'

We will look at each of the processes listed above in more detail. The suffix - ang, etymologically traceable to ak/aki `child' (Rhee 2001), is a diminutive affix for a noun (Kang 1993: 96). The effect of the affixation of the diminutive suffix to kAz `boundary' is reduction of the size of the `edge' to its extreme and thus creates intensified meaning, i.e., `the very edge, the extreme boundary, the very limit,' from which various `extreme' meanings emerge (see 4.3.2 below). The derived word is further phonetically reinforced by compounding with cali `place', i.e., kacangcali `boundary, edge' from the turn of the 20th century (1915 Sincacen), which is the most commonly used noun denoting `edge' in ModK, and the original stem ka is more frequently used in compounds, e.g., kangka `riverside', pataska `beach, coast', etc.

The diminutivized kacang/kAcang undergoes a syntactic operation by combining with the light verb hA- `do/be/say', whereby the noun is changed into a verb. The verbalized kAcanghA-, literally meaning `do the edge', denotes `reach the point of; act arbitrarily; enjoy to the heart's content; exert self to the limit, etc.' The word undergoes a further morphological change through suffixation of the causative -y (a variant of -i), resulting in kAcanghAy-, which denotes `let someone do at will; let someone act arbitrarily; let someone do something to his/her heart's content; etc.' Still another derivation is through the suffixation of -tAoy- `become, fit', i.e., kAcangtAoy- `become complete'. The derived words kAcanghA-, kAcanghAy- and kAcangtAoy- are all attested in LMK (1459 Welinsekpo, 1475 Nayhwun, and 1482 Kumkamkyengsamkahay, respectively) but are obsolete in ModK (see 4.3.2 below).

The diminutivized kAcang, when it is preceded by a noun, typically occurs with the genitive case-marker -s, in the structural template of X-s kAcang [X-gen edge] `X's edge' (note that Korean is a head-final language with the G-N phrasal order, a common Transeurasian feature, Robbeets 2017). In this configuration, a reanalysis of the morpheme boundary occurs from [X-s kAcang] to [X-skAcang], whereby kAcang becomes skAcang, in which the onset `sk' is the tensified /k'/. Some researchers (Choe 1982, Hur 1975, Kim 1955, Lee 1955, among others) hypothesize that sk was phonetically realized as [sk]. Transeurasian languages, however, tend to avoid word-initial consonant clusters (Deny 1924, Robbeets 2017). The morphosyntactic reanalysis also leads to the functional reinterpretation of the nominal kAcang `edge' into the postposition -skacang `till, up to' (written as - kkacang in ModK). The surviving ModK forms of kAz/kAs contain the initial consonant either /k/, when used as a nominal, or /k'/ (= /kk/), when used as a dependent form, most notably the postposition `to, till'. Thus, the development of the particle can be characterized as [X-s kAcang `X's edge' > X-skAcang `till/to X']. A slight variation is observed here, i.e., in some cases of the dependent forms (i.e., suffix), kAcang is realized as -khAcang, as in sulkhAcang `to the point one hates; to one's heart's content' derived from the verb sulh- `dislike, hate', of which /h/ changes /k/ into an aspirated /kh/. Its reduced form silkhes `maximally, heartily, pamperingly' survives in ModK.

The last extension of kAz is its syntactic combination with the verb ep(s)- `not exist'. As briefly noted in 2 above, an idiosyncrasy of Korean is the productive use of the verbs of existence iss- `exist' (e.g., mas.iss- `be tasty') and non-existence eps- `not exist' (e.g., mas.eps- `be tasteless') in lexicalization (Koo 2005, 2008). The nominal kAz also undergoes this operation, producing kaz-ep(s)- [edge-not.exist] `be limitless, be boundless' (< `the edge does not exist'), which later becomes kayep(s)- `be pitiable' (note, however, that there is no lexicalization of kAz involving iss- `exist') (see 4.3.2 below for more).

In addition to formal changes, there occurred a number of notable semantic- functional changes. Many changes already occurred in LMK texts, the earliest data written in Hangeul. Some changes led to a dead-end before MoK, leading the forms to disuse, while many others survive in ModK. The kAz-derived words and morphemes are distributed across many word classes. Depending on the nature of the final item created, we discuss the broad two-way categorization of lexicalization and grammaticalization. Put simply, lexicalization refers to the creation of major class words, e.g., verbs, adjectives and adverbs, whereas grammaticalization refers to the emergence of the forms carrying grammatical functions. Creation of adverbs may or may not be considered as lexicalization. For instance, Heine, Claudi and Hunnemeyer (1991: 3) treat adverbs as non-major (closed) class words (contra Anttila 1972: 151) and the lexicon by Heine et al. (1993) includes their development in grammaticalization.

Polylexicalization

The lexeme kAz has undergone a number of lexicalization processes, notably, conversion, derivation, and compounding, and we turn to an elaboration of each of them.

Conversion

The simplest form of lexicalization that kAz has undergone is `conversion' (Haspelmath 2002: 24-26, Valera 2006: 172-175), i.e., the case in which the form of the base remains unaltered in derivation of a word in a different part of speech, e.g., hammer (noun to verb), book (noun to verb), etc. in English. This process is also often referred to as `zero-derivation' for the lack of formal change. In the case of kAz, the most prominent case of conversion is the change from a noun into an adverb, as shown by the examples below, taken from historical texts:

The adverb kAs denotes `just now, recently, etc.', typically modifying an eventive verb by adding the meaning that the event occurred not long ago. The notion of physical vicinity in `edge' is metaphorically transferred to that of temporal vicinity, i.e., recency.

The conversion between major classes is a relatively unproductive strategy in Korean, largely because major word classes, e.g., nouns and verbs, tend to be different in stem forms and derived forms tend to carry morphological trappings that indicate the derivation (see, however, Hong 1947 [1927], Nam & Ko 2011 [1992], Ko & Koo 2018 for discussion of `cross-categorial uses' in Korean). The instance described above, however, occurs early in history, i.e., LMK, at the same time when the nominal form kAz is attested. In ModK, kas as an adverbial is often found as exemplified in (5):

Derivation and compounding

Among a few lexicalization patterns involving morphological derivation is the creation of the adverb kAcay `most, very', i.e., involving the variant form of kAz and the adverbializing suffix -ay `at, to'. The semantic change from `at/to the edge' to `extremely' is intuitively well motivated. This short-lived form is attested in LMK literature (e.g., 1466, Mokwucaswusimkyel 20a; 1527, Hwunmongcahoy II 15b, etc.) but is completely obsolete in ModK. As is evident from its form (kAcay) and its initial meaning (`at the edge'), the word develops into the postposition kkaci `to, toward, etc.' as the cohesion between the preceding noun and the adverbial increases. Thus, the adverbial kAcay `at the edge' is the bridge from the noun to postposition, a state of affairs corroborating the directionality proposed by Heine, Claudi & Hunnemeyer 1991: 167), i.e., lexical > (adverbs >) adpositions > case affixes > zero. The fully developed adverbial meanings are Elative (`extremely, very') and Superlative (`the most'), a noteworthy directionality of change, considering the parallel development of the adverb kAcang (see below).

Another type of morphological derivation is `diminutivization', involving the diminutive suffix -ang, traceable to aki `baby' (see (3a) above). The ModK kacang carries adverbial functions, as exemplified in (6), taken from historical texts:

In LMK, the adverb kAcang is used for a few different, yet related meanings. As illustrated in (6), they all make reference to the notion of `entirety' inclusive of the extreme members. The usage of (6c), i.e., the Elative degree marker function, was very productive in LMK. In such contexts, it modified eventive or stative verbs, as in kAcang tulechi- `(the earth) shakes very violently' (1447 Yekcwu Welinsekpo 20: 52b), kAcyang hyotohA- `be very dutiful to parents' (1514, Soksamkang Yel: 22b), etc. The usages illustrated in (6), however, are no longer current in contemporary Korean. Instead, kacang is used as the Superlative degree marker, as exemplified in (7):

The ModK adverb kacang in (7) is the Superlative marker, modifying a stative verb (= adjective), even though it was an Elative degree marker in LMK. The shift from Elative to Superlative seems to have completed around the 17th c. (e.g., 1670, Nokeltayenhay I: 10). There are rare instances that were amenable to both Elative and Superlative interpretations in LMK, e.g., kAcang motin nimkum `the very/most cruel king' (1459, Welinsekpo 4:6). However, previous to the 17^th century, there are few, if at all, instances carrying the Superlative function. In contrast, the form is invariably the Superlative marker in ModK (see Rhee 2022 for more discussion).

Another type of lexicalization involving kAz is compounding. Compounding is presumed to be universal (Greenberg 1966[1963]: 93, Guevara & Scalise 2009: 102), even though the level of productivity may vary across languages. As briefly discussed in 3.1, kAz survives as ka in ModK as a full-fledged nominal, still carrying the original meaning `edge'. There is a more commonly used derived nominal, i.e., kacangcali, which is a fortified form through diminutivization (with -ang) and compounding (with cali `place'), as exemplified in (8):

When the examples of kacangcali in (8) are compared with those of ka in (2) above, it is obvious that the two nominals are nearly synonymous and can often be used interchangeably. The former, however, being the phonetically and semantically fortified form of the latter, tends to refer to a more expansive, often circumferential, region. Thus, nwun ka `the side of eyes' tends to refer to a smaller region, the outer corner of the eyes (e.g., where the crow's-feet lines are visible) whereas nwun kacangcali `region around the eyes' refers to a wider region surrounding the eyes (e.g., where a dark circle is visible). Similarly, ip ka `the side of mouth' tends to refer to the two end points of the lips (e.g., where smiling is most noticeable), whereas ip kacangcali `region around the mouth' tends to refer to a wider region surrounding the mouth (e.g., where some unsightly food particles may smudge while eating).

Another type of compounding involving kAz is one in which the light verb ha- (historically hA-) `do/be/say' and the verbs of non-existence ep(s)- `not exist' (see (3c) and (3f) above) participate. Haspelmath (2002: 85-86) notes that the type of compounds formed by the notional object and the verb (called `incorporation' e.g., drawbridge or babysit in English) is unproductive in English but productive in morphologically rich polysynthetic languages. Compounding of kAz and the light verb hA- constitutes an instance of an `incorporation' process. However, compounding involving the notional subject and its predicate eps- `not exist' is a syntactically peculiar case (note, however, that morphosyntactic operations developing into morpholexical operations over time is not infrequent; Sadler & Spencer 2001). The instances of the first type attested in historical data are exemplified below:

In ModK the lexicalized verbs and adjectives, exemplified above, are obsolete. But compounding involving the verb of non-existence eps- `not exist', an instance of `univerbation' (Lehmann 2015[1982]) of a syntactic construction, survives in ModK. The source construction is a clause ka-i eps- [edge-NOM not.exist] `the boundary does not exist; limitless', which later changes to `be pitiable', in the sense that `the pity is boundless', as is exemplified in (10):

Incidentally, the univerbation process, whereby the internal composition becomes increasingly opaque, is still ongoing, which is observable in the fact that there are two adnominal inflected forms, i.e., the conservative kayeps-un (in which /s/ of eps- `not exist' survives) and the more eroded kayew-un (from which /s/ disappeared; note that [p] > [w] is a lexically-controled phonological rule, applicable to many words in Korean). semantic lexeme derivation language

Polygrammaticalization

Another important group of changes that kAz has undergone is grammaticalization, i.e., the creation of linguistic forms that carry grammatical functions. In a now classical work, Kurylowicz (1976 [1965]: 69) defined grammaticalization as “an increase of the range of morpheme advancing from a lexical to a grammatical or from a less grammatical to a more grammatical status,” according to which the development of derivational morphemes is relatively less grammatical than that of inflectional morphemes. In this spirit, we can distinguish weakly grammaticalized kAz-forms, e.g., prefixes and suffixes, and strongly grammaticalized kAz-forms, e.g., postpositional particles.

Derivational prefix

The first type of grammaticalization is found in kAz used as a derivational prefix, as exemplified in historical and modern sources:

The lexeme kas also develops into an adjectival prefix in ModK, as shown below:

The prefix kas- in (12) is a development from the adverbial use of kas `just now, recently', which is evident not only in meaning but also from the fact that its host is usually a deverbal adjective, i.e., it includes a verbal stem, e.g., ik- `ripen' in (12a), nem- `pass' in (12b), cap- `catch' in (12c), etc. However, kas- in these forms is better analyzed as a prefix because it occurs with these deverbal adjectives with a high degree of interlexical cohesion (cf. Paek 2012: 42).

Derivational suffix

The ModK nominal ka `edge, limit' develops into a nominal suffix, which adds the meaning of `-side, -shore, etc.' as shown in (13):

It is obvious that the suffix -ka in (13) is the development from compounding, as exemplified in (2) above. The increased cohesion between the two compounded nominals through frequent use must have demoted the nominal status of ka, whereby ka is reinterpreted as a suffix, analogous to the development of the English seaside (< sea-side < sea side), seashore (< sea-shore < sea shore), etc. Incidentally, many words involving ka in contemporary Korean occur without interlexical spacing (in which -ka is regarded as a suffix), or with spacing (in which ka is regarded as an independent noun in compounding), or with the Genitive -s inserted between the two nominals (in which ka is the head of an NP of a syntactic construction), in dictionaries and in popular writing.

Historical texts also show other kAz-derived nominal suffixes, e.g., -kAs and -skAcang, as exemplified in (14):

In examples in (14), the nominal suffix adds the meaning `to the utmost of, to the fullest of, with all of, etc.' to the nominal host. Thus, mAum-kAs in (14a) and mAzAm-skAcang in (15b) mean `to the satisfaction in heart, to one's heart's content, as much as the heart desires, etc.' The second suffix -skAcang has become obsolete, but the first survives in the form of -kkes in contemporary Korean, as shown in (15), and of a rare variant form -khes, derived from a verb-based construction, in (16):

The function of the suffixes -kkes and -khes in (15) and (16) is not to make reference to an entity (as `edge') but a degree, i.e., the extent of the exertion denoted by the host, e.g., `with all one's might', `with all sincerity', `to the point of disliking', etc. When the suffix is used with a temporal noun as in (15c), it adds the durativity meaning in addition to the limit, as in `up until now, throughout until now'. The suffix -khes in (16) is a variant of -kkes. The added /h/ is from the verb stem silh- `dislike, hate' (historically sulh-), as briefly noted in 4.2 above.

Postpositional particle

A notable grammatical class of kAz-derived linguistic forms is postposition. There are a dozen different dialectal variations, but the primary postpositional particle (in `standard' Korean) is -kkaci, with a variety of meanings. This development has been noted, though mostly in passing, by a number of researchers (Ahn 1967, Lee 2001, Narrog & Rhee 2013, Park 2015, Lee 2016, among others), by virtue of the fact that in the history of kAz, the lexeme's most grammaticalized form, is the postposition -kkaci. The development occurs early in history, as exemplified in the following, in its historical forms -kAcang and -skAci:

In contemporary Korean, the postposition survives in the form of -kkaci (though -kkacang also survives in certain dialects) and functions as a signal of diverse functions of forward-direction marking, e.g., goal (Allative), addition (Additive), inclusion (Inclusive), or end point (Terminative). It also marks counterexpectation (Mirative). These uses are exemplified in the following:

The five sub-functions of the postposition -kkaci illustrated above are not mutually exclusive and cannot always be delineated clearly. For instance, some of these, especially, Additive, by virtue of signaling addition of the extreme member on a scalar continuum, tend to carry the overtone of counterexpectation (Mirative).

The development of kAz and its derived forms in history can be diagrammatically presented as in Figure 1. The diagram shows that the lexeme changes in a number of ways either reductively or expansively and that a few forms are lost in history. One caveat is that the distance between the forms and Middle Korean or Present-Day Korean and their relative positions are not proportional to the time of their emergence in history.

Figure 1. Formal change of kAz from LMK to Present-Day Korean

Discussion

We noted in the above that the lexeme kAz has undergone various changes in history. The description warrants discussion of a number of theoretically interesting issues, but we will limit our discussion to three major issues, i.e., multiple paths (5.1), grammaticalization mechanisms (5.2), and conceptual motivation (5.3).

Multiple paths

The changes of kAz described above encompass lexicalization (4.3) and grammaticalization (4.4). Diverse semantic and grammatical functions that the nominal kAz acquired over the course of time can be summarized as follows (note that certain meanings are not exemplified in 3):

Lexicalization processes, inclusive of conversion, derivation, and compounding, listed in (19), are some of the ways whereby a lexicon is enriched, either through spontaneous shifts of functions in different word classes (conversion) or morphosyntactic operations involving a single stem (derivation) or multiple stems (compounding). In these lexical operations, kAz tends to designate the `limit' or `end', often physical but sometimes abstract. When the operation recruits a negative word (e.g., eps- `not exist'), the word denotes the absence of such `limit' or `end'. On the other hand, in the case of adverbs, kAz often denotes `edge' or `periphery', thus not having advanced enough toward the center, even though it also denotes `limit' or `end', thus maximal degree in some instances (see 5.2 for more discussion of conceptual motivation).

More intriguing is the grammaticalization listed in (20) in that the change makes the source lexeme depart from its original lexical function and acquire grammatical functions. As we have seen, grammaticalization of kAz is varied and proceeds into different grammatical domains, thus constituting polygrammaticalization. In other words, the single lexeme develops into diverse grammaticalized forms in the domains of derivational morphemes (e.g. prefixes and suffixes) and of postpositions (Allative, Additive, Inclusive, Terminative, Mirative, etc.). In the case of prefixes, as their development is closely related to that of adverbs, kAz tends to denote shallow depth (from `edge, boundary') in time, hence, temporal recency. In the case of suffixes, as their development is closely related to nominal compounding, kAz tends to denote `edge' or `end', i.e., the source meaning of the lexeme. A similar developmental scenario is observed with the grammaticalization of postpositions. As kAz occurred in the nominal compounding configuration, it tends to designate the `end point' with variable foci.

When the meanings of kAz in the lexicalized forms and those in the grammaticalized forms are compared, the semantics of the lexicalized forms are substantially more concrete. This is in line with the observation that grammaticalization involves `abstraction' in meaning (Traugott 1980, Heine, Claudi & Hunnemeyer 1991), which is closely tied to the fact that grammatical meanings in general are abstract (see also Sapir 1921). The abstraction process is so robust in grammaticalization that it is often considered a mechanism or principle, sometimes with a different name, i.e., desemanticization (see 4.2 below).

Grammaticalization mechanisms

In this context, we can briefly review the mechanisms of grammaticalization, as proposed by Heine and Kuteva (2002) and Kuteva et al. (2019), i.e., desemanticization, extension, decategorialization, and erosion. As shall be seen, these grammaticalization mechanisms are often observable in lexicalization processes as well, because grammaticalization is not exclusive of lexicalization. Rather, grammaticalization is often accompanied by lexicalization, and both processes exhibit fluctuation of form and function.

Desemanticization, also commonly called `semantic bleaching' in figurative terms, refers to the loss in meaning content. The lexeme kAz, even at the lexicalization level, has undergone extensive semantic bleaching from its original meaning of `edge', a physical object, to abstract meaning of `limit', `degree', and temporal meaning of `recency', among others. When it develops into grammatical markers, it also acquires more abstract meanings as addition, termination, inclusion, surprise, etc. Also notable is the fact that when desemanticization or abstraction occurs, the previous meanings remain, and thus the process is not meaning shift (e.g., from A to B) but meaning extension (e.g., from A to A&B). In this regard, desemanticization of kAz (and in fact most instances of grammaticalization) is related to `generalizing abstraction', a process whereby the number of distinguishing features of a concept is reduced to its most central characteristics (Heine, Claudi & Hunnemeyer 1991: 43). In this sense, desemanticization (i.e., the reduction of distinguishing features) is in fact semantic enrichment (i.e., the increase in the number of semantic designations).

The mechanism of extension, also known as context generalization, refers to the use of a form in new contexts. In its lexicalization and grammaticalization stages, kAz extends its use context from its original position, i.e., an argument position by virtue of a noun, to other non-argument positions, e.g., modifier of a verb or verb phrase (as an adverbial), which enables its further positional extension, i.e., placement before a deverbal adjective (as an adjectival prefix). When the meaning becomes more abstract, its placement in compounding becomes not restricted to behind a concrete noun with physical edges or boundaries; it can be affixed to an abstract noun such as `heart', `power', `sincerity', etc. (as a nominal suffix).

Decategorialization refers to the loss in morphosyntactic properties characteristic of lexical or other less grammaticalized forms. Since kAz is a noun, decategorialization kAz necessarily involves the loss of the characteristic features of the nominal category. The characteristics of the nominal category typically include pluralizability, modifiability by an adjective, ability to be case-marked, ability to function as an argument, among others (note that the ability to be marked by an article is not applicable in Korean). At the level of lexicalization whereby the nominal kAz has changed its word class into a verb, adjective, or adverb, such categorial features are all lost. Likewise, at the level of grammaticalization, whereby the lexeme has changed to a prefix, suffix, or postposition, such categorial features are lost.

Erosion, also known as phonetic reduction, refers to the loss in phonetic substance. An agglutinating language, Korean has instances of substantial loss of phonetic volume of multiply stacked postpositional particles and verbal suffixes.

However, the change of kAz does not always involve reduction of phonetic volume. The change from the LMK noun kAz to ModK noun kA is an instance of erosion. Lexicalization processes such as compounding and derivation, needless to say, lead to increase of phonetic volume. Even in grammaticalization, the developments of the prefix kas-, the suffixes -khes, -kkes, and -ka, and the postposition -kkaci do not uniformly exhibit erosion. Rather, most of them have clearly increased phonetic volume, as shown by additive processes listed in (3) above. This state of affairs suggests that the mechanisms are by no means the prerequisites or absolute diagnostics of grammaticalization; they indicate a general directionality of grammaticalization processes.

Conceptual motivation

As illustrated in the preceding exposition, the series of semantic changes kAz has undergone in hi...