Cognitive reference points in adjective-noun integration networks
Formal semanticists have often used adjective-noun (AN) combinations as evidence of full composi-tionality. On this view, the meaning of red house, for example, is a mere intersection of the set of houses and the set of red things. Linguists working in the cognitive framework have shown that AN constructions may have non-compositional, emergent properties. Focus of these studies has been on noun modification via “intractable” adjectives, like privatives (fake guns) and non-predicating adjec-tives (hot lids) (Coulson 2001). Less attention has been given to modification via predicating adjec-tives, like red house. In this paper I will show that cognitive operations involved in processing “intrac-table” AN modifications are also relevant to the interpretation of “simpler” cases of noun modification via predicating adjectives.
Following Fauconnier & Turner (1998), I suggest that adjectival modification is carried out through conceptual integration of mental spaces. In case of red house the two input spaces are COLOUR and HOUSE. Note, however, that not everything about the red house is red: prerequisite to conceptual blending is the identification of the active zone (Langacker 1987). A default active zone for HOUSE in red house could be exterior walls of red colour. If the active zone is shifted to salient balcony-rails, red house may mean a grey house with red balcony-rails.
Identifying the active zone of the adjectival space looks more problematic. The landmark of red, for example, is the part of the spectrum stretching from purple to orange. However, in actual use only one instantiation of the landmark is active. Sweetser (1999) suggests that by default the prototypical value (e.g. focal red) is projected to the blend. However, prototypes can not be applied to all adjectivals. For adjectives like short, blunt, or warm prototypes fall short of adequate semantic descriptions. I suggest that prototypes constitute a specific (but not the only) type of cognitive reference points (CRPs), i.e. mentally prominent items partaking in identification of the active zone. Another type of CRPs appli-cable to adjectival semantics is a cognitive zero, an average value varying per comparison class.
In zero-contexts and in contexts where the comparison class is not specified a default CRP (prototype or cognitive zero) is employed. Default CRPs are also used if language users lack foreknowledge of the compound. For example, if the interpretation of red house is not constrained by foreknowledge of red houses having a dark-red colour, the recipient is likely to activate prototypical red as a default CRP. If language users do possess relevant foreknowledge of the combination, a compound CRP is used, which is a blend in itself (e.g. prototypical red wine is not bright-red, the CRP is a dark shade of red).
The results of this study run counter to approaches splitting interpretation of AN combinations into composition by virtue of linguistic meaning, and enrichment though encyclopaedic knowledge. This paper shows that not only blends, but also input spaces in AN integration networks may have emer-gent properties (e.g. emergent CRPs) that cannot be accounted for, unless context and encyclopaedic knowledge are immediately involved in meaning construction.
Coulson, S. (2001). Semantic Leaps: Frame-Shifting and Conceptual Blending in Meaning Construction. New York and Cambridge: Cambridge University Press.
Fauconnier, G. and Turner, M. (1998). Conceptual Integration Networks. Cognitive Science 22(1): 133-187.
Langacker, R. (1987). Foundations of cognitive grammar, vol.1, Theoretical prerequisites. Stanford, CA: Stan-ford University Press.
Sweetser, E. (1999). Compositionality and blending: semantic composition in a cognitively realistic framework. In: Janssen, Th. and Redeker, G. (Eds.). Cognitive Linguistics: Foundations, Scope, and Methodology. Pp. 129-162. Berlin and New York: Mouton de Gruyter.