Umeå University's logo

The present study investigates the word formation process of lexical blending in the context of written US web news between January 2010–March 2018. The study has two interrelated aims. First, it aims to develop a transparent, rigid, and replicable method of data collection. This is motivated by a lack of systematicity of data collection procedures in previous research. Second, it aims to identify the characteristics of the retrieved blends; both generally and with a special focus on how ambiguity is realized. The data were collected from an offline version of the NOW corpus (News On the Web). A strict algorithm was devised to organize the data and to identify lexical blends among a large body of systematically collected word forms. Both automatic and manual procedures were employed in these tasks. The study is conducted within the framework of Cognitive Linguistics (CL). Semantic analysis is foregrounded in CL and language is considered perspectival, dynamic, non-autonomous, and experience-based. Furthermore, a Langackerian view on meaning is adopted in that symbolic potential is acknowledged in all resources and manifestations of language. Categorization is approached in accordance with the tenets of prototype theory, which acknowledges fuzzy category boundaries and gradual distribution of attributes.

The results of the study show that the data collection methodology is quantitatively robust, which offers the possibility to generalize the observations within the context of the chosen limitations. Consequently, the developed methodology may also be applied in future investigations. Second, quantitative analyses validate some previous assumptions about grammatical functions, semantics, and seriality in blending. Third, a set of qualitative characteristics are identified in the collected set of blends, which offers a comprehensive approach to describing blend formation in the given context. The characteristics structural profiling and domain proximity are suggested as prominent aspects of blend formation. Structural profiling is marked by prominent structural attributes such as similarity of source words and intricate patterns of amalgamation, but figurative strategies are also foregrounded. Domain proximity is described in terms of semantic similarity between the source words and an iconic relation between the fusion of structure and the fusion of concepts. The notion of pseudomorphemic transfer is used to capture blends that fall within the operational definition of the study but also seem to be connected to other morphological processes through the instantiation of morphological schemata. Blends clustered in series based on recycled truncated segments are revisited from a qualitative perspective, and it is claimed that the process of morphemization is likely influenced by the degree of morpheme-like character of a serially distributed segment. Furthermore, four types of ambiguity are identified in the blend data; truncation ambiguity, mode ambiguity, source word ambiguity, and covert source ambiguity. On the basis of the observed impact of ambiguity, it is suggested that the construal of meaning in lexical blending makes use of multistability, which is a perceptual phenomenon observed in, for instance, binocular rivalry. Taken together, the results constitute a background for suggesting a model of categorization divided into two levels of organization. This model of categorization is called the dual model of blend classification.