Abstract Glynn, Geeraerts & Speelman
Frames, Fields, and Parasynonymy. Developing usage-based methodology for Cognitive Semantics
This study examines methodological advances in a multifactorial usage-based approach to conceptual structure. Through a case study of three parasynonymous lexemes in English (hassle, bother, annoy), we demonstrate that within the theoretical framework of Cognitive Linguistics (Lakoff 1987, Langacker 1987) corpus-driven, quantitative, and multifactorial techniques may be employed in the description of abstract lexical concepts. Within Cognitive Linguistics, both the study of polysemy (semasiology) and parasynonymy (onomasiology) have recently profited from a dramatic growth in the awareness of the need for a quantitative usage-based methodology (Dimitrieva 2005, Tummers & al. 2005, Frohning 2005, Heylen & al. 2006, Dvijak 2006, Wulff 2006). This is, in part, due to a range of studies that have demonstrated the unreliability of the analytical models used in cognitive semantic analysis thus far (Sandra & Rice 1995, Tyler & Evans 2001 inter alia). However, it is also due to the success of the cognitive quantitative research in lexical and syntactic semantics, developed by Dirven & al. (1982), Schmid (1993, 2000), Geeraerts (2000, & al 1994), Fischer (2000), Gries (2003), and Stefanowitsch (2003, 2006). This study seeks to develop such an approach so that it may serve as a general method of analysis for Cognitive Semantics. A quantitative usage-based method aims for maximum objectivity and the possibility of result verification. The method is based on the feature analysis of corpus examples, the results of which may be examined statistically. Importantly, such a method allows a multifactorial approach to language description that may include, for instance, the effect of dialect, register, and morpho-syntax on polysemy but also the degree of entrenchment and conceptual salience in prototype structures. However, before this approach may be used in the main of lexical semantics, several analytical issues must be resolved. One of the most difficult hurdles this methodology faces is the abstract nature of many concepts. Concepts that possess no Lebenswelt referent or socially delineated function prove particularly difficult to analyse using this method. The basic problem is that feature analysis requires a tertium comparationis or some constant upon which one may base the analysis. We propose a solution to this limitation by drawing on Frame Semantics (Fillmore 1985). Instead of pursuing a feature analysis of the concept per se, we examine the Arguments and Argument Relations instantiated by given speech events. This indirect context-feature approach gives us a picture of how a lexeme is used even though it is not a direct analysis of the concept associated with the form. Despite this, we show that the method supplies enough conceptual information that we may distinguish different "meanings" of a lexeme as well as how those "meanings" overlap with, and differ from, other parasynonymous lexemes. The procedure is straightforward and reliable to the extent that the corpus is representative and the features described are objective and "operationisable". The coding results in a contingency table of feature frequencies of any number of values and variables. These contingency tables may then be investigated using statistical techniques such as Correspondence and / or Hierarchical Cluster analysis. These exploratory methods give us a map of the feature grouping that may be argued to represent meanings of the given lexeme. The significance of Cluster and Correspondence results may be verified by further statistical techniques such as Log-Linear and Logistic Regression analyses. Furthermore, if the same features are coded for different parasynonymous lexemes, then comparative usages of the lexemes may be treated in the same manner offering an onomasiological map of the lexemes under consideration. Our case study pursues a context-feature analysis of three parameters. These include three extra-linguistic variables (register, text-type, and dialect) and two linguistic variables (semantic roles - argument structure and syntactic roles - morphosyntax). This gives us a three way model that represents the formal and semantic usage relative to basic extra-linguistic variables. Large unparsed corpora are used; these are compiled from Internet sources from a range of media such as blog, Usenet, and news press. They also represent both American and British Englishes. A sample of 500 instances of each of the lemmas in both dialects is coded as well as extra cohort studies on certain parts of speech. The frame-based approach to context-features is more productive for verbal profilings of lexical concepts and thus cohort studies examine in more detail the verbal examples. Despite the labour intensive process of coding occurrences in untagged corpora, the varied and natural language found on the Internet compensates for such difficulties, especially in lexical semantics where data sparseness limits the use of commercial corpora. The case study results demonstrate that the use of context-feature analysis for the semantic description of abstract concepts is possible. The statistical study of the feature frequency reveals patterns of usage as well as important distinctions in register and dialect, elements that are particularly difficult to reveal using alternative methods of semantic analysis. Since these "extra-linguistic" elements to semantic structure must necessarily be considered in a usage-based account of language structure, this addition to the study of polysemy and parasynonymy is vital. However, two shortcomings are revealed. Firstly, although possible, the efficacy of Frame Semantic based analysis of context-features proves to be quite limited for non-verbal lexemes. Secondly, the labour required to examine enough occurrences to reveal the more subtle usages of a given lexeme limits onomasiological research. Although theoretically possible, the number of examples that would be needed to capture less common usages and have them appear statistically significant is not practically feasible using manual coding. Despite these limitations, the quantitative nature of the analysis not only gives us a semantic map of the basic usage of the items, it permits a multifactorial approach that takes into consideration factors such as register and dialect, factors that are sorely overlooked in the main of Cognitive Linguistics.
Benzecri, J.-P. 1980. Analyse des correspondances. Bordas: Paris.
Biber, D. 1995. Dimensions of register variation. CUP: Cambridge.
Cibois, P. 1994. Analyse factorielle. PUF: Paris.
Dvijak, D. 2006. Delineating and Structuring Near-Synonyms. S. Gries & A. Stefanowitsch edd. Corpora in Cognitive Linguistics. Mouton: Berlin.
Dmitrieva, E. 2005. .. ......... . ....... ..... .......... . ......... .o. Scando-Slavic 51.
Dirven, R. & al. The scene of linguistic action and its perspectivization. Benjamins: Amsterdam.
Fillmore, C. 1985. Frames and the Semantics of Understanding. Quaderni di Semantica 6.
Fischer, K. 2000. From Cognitive Semantics to Lexical Pragmatics. Mouton: Berlin.
Frohning, D. 2005. Kausalmarker zwischen Pragmatik und Kognition. PhD. University of Feiburg.
Geeraerts, D. 2000. Salience phenomena in the lexicon: A typology. L. Albertazzi ed. Meaning and Cognition. Benjamins: Amsterdam.
- & al. 1994, The Structure of Lexical Variation. Naming, meaning, and context. Mouton: Berlin.
Gries, S. 2003. Multifactorial Analysis in Corpus Linguistics. Continuum: London.
Heylen, K. & al. 2006. Methodology in corpus-based Cognitive Linguistics. G. Kristiansen & R. Dirven edd. Cognitive Sociolinguistics. Mouton: Berlin.
Lakoff, G. 1987. Women, Fire, and Dangerous Things. UCP: London.
Langacker, 1987. Foundations of Cognitive Grammar. SUP: Stanford.
Sandra, D. & Rice, S. 1995. Network analyses of prepositional meaning. Cognitive Linguistics 6.
Schmid, H.-J. 1993. Die Kategorisierung als Grundprinzip einer differenzierten Bedeutungsbeschreibung. Niemeyer: Tübingen.
- 2000. English Abstract Nouns as Conceptual Shells. Mouton: Berlin.
Stefanowitsch, A. 2003. Constructional semantics as a limit to grammatical alternation. G. Rohdenburg & B. Mondorf edd. Determinants of Grammatical Variation in English. Mouton: Berlin.
- 2006. Words and their metaphors: A corpus-based approach. S. Gries & A. Stefanowitsch edd. Corpora in Cognitive Linguistics. Mouton: Berlin.
De Sutter, G. & al. in press. Detecting and balancing determinants of word order variation. Linguistics.
Tryon, J. & Bailey, D. 1970. Cluster Analysis. McGraw-Hill: New York.
Tummers & al. 2005. Usage-based approaches in Cognitive Linguistics. Corpus Linguistics and Linguistic Theory 1.
Tyler, A. & Evans, V. 2001. Reconsidering prepositional polysemy networks. Language 77.
Wulff, S. 2006. Go-V vs. go-and-V in English: A case of constructional synonymy? S. Gries & A. Stefanowitsch