Lexical-semantic structures in French and Croatian with an emphasis on prefixation : a contrastive analysis

U temelju je ovoga rada pitanje načela leksičke izgradnje. To se pitanje istražuje na primjeru prefiksacije, a pristupa mu se supostavljanjem francuskoga i hrvatskog, dvaju jezika različitih grana indoeuropske porodice i različite morfološke složenosti. Za analizu u radu odabrani su leksemi tvoreni pomoću četiriju prefikasa, sur- i sous- (sou-) te nad- (nat-) i pod- (pot-). Pored prefigiranih tvorenica, dodatno se promatraju i prijedlozi sur, sous, nad i pod od kojih su navedena četiri prefiksa nastala. Osnovna je pretpostavka rada da razlike između navedenih dvaju jezika utječu na različitu izgradnju promatranog dijela leksika. Dodatne su pretpostavke da su promatrane leksičko-značenjske strukture produktivne u oba jezika, da su i prefiksi i prijedlozi polisemne jedinice te da dio prefiksalnih značenja, ali i prefigiranih leksema stoji u odnosu značenjske suprotnosti. Građa za analizu prikupljena je iz dvaju usporedivih velikih mrežnih korpusa, frWaCa za francuski i hrWaCa za hrvatski. Korpusnom pretragom koja je ručno pročišćena dobivena su 1142 leksema, od čega 752 imenice, 264 glagola i 126 pridjeva. Riječ je o leksemima tvorenima pripajanjem jednoga od četiriju navedenih prefikasa na bazu koja pripada istoj leksičkoj kategoriji kao i tvorenica budući da se promatra samo čista prefiksacija. Kao teorijsko-metodološki okvir za analizu je građe odabran model morfosemantičkih obrazaca koji počiva na temeljnim strukturalističkim postavkama, ali uključuje i neke osnovne elemente kogntitivne lingvistike, u prvome redu onu o kontinuumu leksika i gramatike. U okviru modela MS obrazaca svaka se motivirana riječ promatra kao konstrukcija, čime se omogućuje ujednačen pristup dvama promatranim jezičnim elementima u ovome radu – prefiksima i prijedlozima – budući da se i prijedlozi promatraju kao dio konstrukcija koje tvore. Analizom su potvrđene sve pretpostavke. Točnije, promatrani su prefiksi produktivni i tvore najveći broj imenica u oba jezika. Prefiksi i prijedlozi grade raslojenu polisemnu strukturu koja se ogranizira oko prototipnoga prostornog značenja koja služe kao polazište za brojna značenjska proširenja, ponajprije na temelju metafore, ali i pojmovne bliskosti. Velik dio tvorenica stoji u odnosu hiponimije prema vlastitoj osnovi, a uz to i između dijela prefiksalnih i prijedložnih značenja kao i parova prefigiranih tvorenica postoji odnos značenjske suprotnosti. Najveće su promjene kao rezultat prefiksacije primijećene kod glagola. Kod dijela glagola dolazi do metaforizacije cijele tvorenice, što se puno rjeđe primjećuje kod imenica i pridjeva. Prefiksacija kod brojnih glagola dovodi do promjena u značenjsko-sintaktičkim obilježjima tvoreći velik broj glagola slabe tranzitivnosti koji najčešće nemaju obilježje zahvaćenosti objekta, iako neki nemaju ni obilježje agentivnosti. Ukupno gledajući, analiza je pokazala da se pomoću promatranih prefikasa tvore brojni leksemi kojima se ne obogaćuje samo leksik francuskoga i hrvatskog jezika, već i grade složeni unutarleksički odnosi (polisemija), a posredno i oni međuleksički (hiponimija i antonimija). Prefiksacija ima složenije učinke na hrvatski jezik, što je u skladu s početnom glavnom pretpostavkom. Ovim se radom daje prinos istraživanju prefiksacije općenito, posebno imenske i pridjevske koja je u literaturi slabije zastupljena, ali i omogućuje nov uvid u načela izgradnje leksičko-značenjskih struktura.This thesis focuses on principles of lexical construal. In order to do so, it explores a particular word-formation process, simple prefixation (e.g. Polinsky 2008:267; Körtvelyessy and Genči 2017:26; Babić 2002:375), defined as the attachment of a prefix to a base. The study presented in this thesis is conducted from a contrastive standpoint by looking into French and Croatian, languages from two branches of the Indoeuropean family which also differ with respect to morphological complexity, with Croatian being morphologically richer than French. The analysis includes lexemes formed with the following four prefixes: sur- and sous- (sou-) (French) and nad- (nat-) and pod- (pot-) (Croatian). In addition to prefixed lexemes, prepositions sur, sous, nad i pod, from which the aforementioned prefixes stemmed, are also studied. The main hypothesis is that general differences between the French and Croatian language entail a different construal of the observed part of the lexicon. Additional hypotheses are that the observed lexical-semantic structures are productive in both languages, that both prefixes and prepositions are polysemous language units and that a part of prefixal meanings, as well as of the meanigs of prefixed lexemes, represent antonyms. The methodological approach in this thesis is a corpus-based one (Biber 2010:162-163), where conclusions are drawn on the basis of an analysis of observed language phenomena and in accordance with a set of theoretical settings. Data for the analysis were retrieved from two comparable large web corpora, frWaC (Baroni et al. 2009) for French and hrWaC (Ljubešić and Erjavec 2011) for Croatian. These two corpora slightly differ in size, but were composed following the same principles. Corpus search was conducted via CQL orders of the following type: the [word="sur.*"] containing [tag="N.*"]. In this particular order, all words belonging to the noun class (N.*) are retireved. Subsequently, the results were manually checked. The corpus search resulted in 1142 lexemes, 752 nouns, 264 verbs and 126 adjectives. The theoretical and methodological framework for the analysis presented in this thesis is the model of morphosemantic patterns (Raffaelli and Kerovec 2008, 2017, Raffaelli 2012, 2013, 2015b, 2017, Filko and Raffaelli 2017, Raffaelli 2018) which is based on the main structuralist tentes. The model also includes some major Cognitive Linguistic features, such as the notion of grammar-lexicon continuum (Langacker 1987), as well as Conceptual Metaphor Theory (e.g. Lakoff and Johnson 1980). Within the model of morphosemantic patterns, every motivated word is considered to be doubly motivated, both formally and semantically, thus emphasizing the semantic approach to word-formation. In addition to that, motivated words are perceived as constructions. This fact enabled us to have a unified approach to both prefixed lexemes and prepositions, the latter also being considered part of constructions they are used in. The analysis has demonstrated that nouns are the most numerous lexical category in which prefixes develop the richest polysemy. Prefixal meanings are mostly more abstract and based on orientational metaphors. A great majority of prefixed nouns are hyponyms of their bases (e.g. sous-vêtement ‘underwear’ is a type of vêtement ‘clothing item’) and they are also endocentric. Verbs are the lexical category that exhibits the most significant changes as a result of prefixation. Some prefixed verbs, especially in Croatian, undergo metaphorization as a whole, which is rarely the case in nouns and adjectives. Such metaphorized verbs are usually not hyponyms of their bases. In a number of verbs, prefixation leads to changes in the semanticsyntactic features. In this thesis, we have observed two such features: argument structure and transitivity, as the most relevant ones. In accordance with Hopper and Thompson (1980), we consider transitivity to be a gradable category in which two elements are especially important for the verbs analysed in this thesis: agentivity and affectedness of object. In both analyzed languages, prefixation usually results in the formation of verbs with weaker transitivity, i.e. lacking one of these two elements (or both), usually affectedness of object (e.g. nadtrčati 'lit. over-run someone; run better than someone', formed from trčati 'run'). A great majority of Croatian prefixed verbs are perfective with respect to verbal aspect. Verbs are the lexical category on which prefixation exercises the most profound influence. They are also the category in which the greatest differences between French and Croatian arise. More precisely, metaphor and semantic-syntactic changes are more present in Croatian verbs, while there is less hyponymy than in French verbs. Adjectives are the least numerous lexical category in the analyzed corpora, and therefore exhibit the least developed polysemy. Their meanings are very similar to those found in prefixed nouns, and are predominantly abstract (e.g. suractif 'very active'). A majority of adjectives formed with the prefixes sur-, sous- and pod- are hyponyms of their bases, while those formed with the prefix nad- are an exception. The latter ones are mostly exocentric and do not represent hyponyms of their bases (e.g. nadljudski 'superhuman' is not human, but 'above human'). When it comes to prepositions, they usually appear in constructions containing a verb (e.g. profesor bdije nad svojim razredom 'lit. the professor stays awake over his class; the professor takes care of his class') or a noun (e.g. sous la protection de l'ONU 'under UN protection'). Some prepositional constructions such as the mentioned bdjeti nad 'watch over' are cases of strong rection between a verb and a preposition and should be studied in more detail in further research. Anaylsis has demonstrated that prepositions too, like prefixes, are highly polysemous language elements. The number of their meanings is similar to those found in prefixes forming nouns, which is higher than both those of prefixes forming adjectives and verbs due to the larger number of nouns in the corpus. Still more important than the number of prepositional meanings is the fact that they differ significantly from the prefixal meanings, thus confirming one of our main initial hypotheses. More precisely, we have demonstrated that the different constructions in which prefixes and prepositions are used have very different effects on their semantic behaviour, leading to new prefixal meanings developing in prefixed lexemes under the influence of the bases. In sum, the analysis has confirmed all the three hypotheses. First, the observed prefixes are productive, with the largest number of noun derivatives and the least number of adjectives in both languages. Second, prefixes and prepositions have a rich polysemous structure organized around a prototype, i.e. the spatial meaning serving as a basis for a number of meaning extensions. These extensions are based on orientational metaphors in a majority of cases, and on conceptual contiguity. Third, antonymy appears between a part of prefixal meanings, as well as between some pairs of prefixed lexemes, mostly nouns. A large part of prefixed lexemes are hyponyms of their bases. These findings demonstrate that prefixation affects not only the intralexical level, leading to polysemy, but also the interlexical level, resulting in antonymy and hyponymy. Our analysis has shown that the observed prefixes form numerous lexemes that not only enrich the lexicon of the French and Croatian language respectively, but they also build complex intralexical (polysemy) and interlexical relations (hyponymy and antonymy). Prefixation has been shown to have more complex effects on Croatian, which is in line with our main hypothesis. This thesis makes a contribution to the exploration of prefixation in general, especially noun and adjective prefixation, which is less present in the literature than verb prefixation. The thesis also offers a new insight into the principles of the construction of lexicalsemantic structures, which remain one of the less explored topics of word-formation