CAOSS und Transzendenz – Zur Repräsentation und Verarbeitung von Nomenkomposita
Allgemeine und Vergleichende Sprachwissenschaft, Experimentelle Linguistik, Typologie, Außereuropäische Sprachen
Zusammenfassung der Projektergebnisse
Compounds are words such as hornbill that consist of two constituent words (here: the modifier horn and the head bill). In the present research project, we investigated the morphological transcendence hypothesis, which postulates that as a result of repeated experience with a word in a specific constituent position leads to the formation of asconstituent meanings which might experience a semantic shift away from the original word meaning (as an example, the head -bill almost exclusively refers to birds, as in hornbill, shoebill, or razorbill). Such as-constituent meanings are very hard to conceptualize using the standard psycholinguistic toolkit; however, they are an inherent part of the meaning-combination process in the CAOSS model, a computational compounding model based on distributional semantics. We first established the validity of measures of semantic shift (the similarity between as-constituent meanings and the corresponding free-word meanings) derived from the model, using qualitative and quantitative analyses. In a series of language-centered studies, we then found that semantic shift is not a direct consequence of more experience with a word in a specific role, that semantic shift is higher for modifiers than for heads, and that the as-modifier and as-head meaning of a word (if it is used in both roles) are more similar to the original meaning than they are to each other. We also found that the higher semantic transparency of heads as compared to modifiers is better described by as-constituent meanings than the free-word meanings of the constituents. When investigating the phenomenon of semantic transparency, we re-conceptualized our initial research program following the realization that any human judgment of semantic transparency is itself a form of behavior to be explained by a psychological model, rather than an objective gold standard. In a series of large-scale behavioral studies, we then found partial evidence for the perspective that these as-constituent representations rather than the free-word representations are the actual psychological units involved in compound processing: On the one hand, the influence of as-constituent meanings on explicit semantic transparency judgments for compound meanings is rather limited. On the other hand, semantic effects in on-line processing are best described as the result of a compositional process combining the constituent meanings in their as-constituent forms. Another series of priming studies to further investigate the psychological status of asconstituent meanings was unsuccessful, since we observed no semantic priming effects. Similarly, we found no evidence for an extended morphological transcendence hypothesis which assumes that as the “end result” of morphological transcendence compound modifiers transcend into affix-like representations. A number of studies resulting from the project, mostly with a focus on response times in compound processing, have already been published.
Projektbezogene Publikationen (Auswahl)
- (2019). Enter sand-man: Compound processing and semantic transparency in a compositional perspective. Journal of Experimental Psychology: Learning, Memory, and Cognition, 45, 1872–1882
Günther, F., & Marelli, M.
(Siehe online unter https://doi.org/10.1037/xlm0000677) - (2019). Vector-space models of semantic representation from a cognitive perspective: A discussion of common misconceptions. Perspectives on Psychological Science, 14, 1006-1033
Günther, F., Rinaldi, L., & Marelli, M.
(Siehe online unter https://doi.org/10.1177/1745691619861372) - (2020). Semantic transparency effects in German compounds: A large dataset and multiple-task investigation. Behavior Research Methods
Günther, F., Marelli, M., & Bölte, J.
(Siehe online unter https://doi.org/10.3758/s13428-018-1143-3) - (2020). Semantic transparency is not invisibility: A computational model of perceptually-grounded conceptual combination in word processing. Journal of Memory and Language, 112, 104104
Günther, F., Petilli, M. A., & Marelli, M.
(Siehe online unter https://doi.org/10.1016/j.jml.2020.104104)
