Corpus-Based Approaches to Metaphor and Metonymy 227annotation studies. Therefore, their contribution is of particular significance. The authors aim
to establish a framework for metonymy annotation. They first of all propose seven principles
for building a qualified annotated corpus. The principles include ―platform independence‖,
―domain and genre‖, ―annotation extent‖, ―regularities‖, ―underspecification‖, ―coverage‖,
and ―hierarchical structure.‖ Then these principles are elaborated in terms of metonymy
annotation. Firstly, a word corpus is chosen and encoded into a markup language (here
Extensible Markup Language). Secondly, the data should cover domains and genres as many
as possible. Thirdly, the annotation extent is the word level. Fourthly, semantic classes and
metonymic patterns are used as criteria for annotation categorization. Fifthly, both base
classes and metonymic patterns are annotated. Sixthly, both conventional and unconventional
metonymies are annotated. Seventhly, the categories are arranged in hierarchy (156-160). To
illustrate how the annotation framework operates, the contributors focus on two class-specific
annotation systems for location names and organization names. In the system for location,
patterns of PLACE-FOR-PEOPLE, PLACE-FOR-EVENT, and PLACE-FOR-PRODUCT are
introduced and explained. In the system for organization, ORGANIZATION-FOR-MEMBER,
ORGANIZATION-FOR-FACILITY, ORGANIZATION-FOR-PRODUCT, ORGANIZATION-FOR-
INDEX, and ORGANIZATION-FOR-EVENT are distinguished and exemplified. As for class-
independent patterns, the authors set the categories OBJECT-FOR-NAME, OBJECT-FOR-
REPRESENTATION, OTHER, and MIXED. In order to verify the validity of the annotation
systems, the authors conduct an annotation experiment. The findings show that their system is
of high reliability. Their annotation research also enriches the study of metonymy in that
some mappings are discovered, for instance, ORGANIZATION-FOR-INDEX and
ORGANIZATION-FOR-EVENT. Despite the striking achievements, they acknowledge that their
work is facing problems raised by the fuzziness of categories. In the conclusion, they predict
the expansion of their annotation schemes and the cooperation with other annotation systems.
The chapter ―On groutnolls and non-heads: A case study of the interaction between
culture and cognition in intelligence metaphors‖ by Kathryn Allan has provided an example
of empirical diachronic investigations into metaphors. Based on historical data extracted from
the Historical Thesaurus of English at Glasgow, which is characterized by its categorization
of lexical items by concept, Allan concentrates on the target domain INTELLIGENCE and one
of its three most important source domains—DENSITY. The analysis shows that all entries in
the source domain identify stupidity (177) and three subgroups, i.e. WOOD, EARTH, and
FOOD take up three fourths of the total entries. How has the property of these everyday
things evolved to metaphorize intelligence? Allan turns to cognitive theories, culture and
linguistic successes for possible answers. In his view, the conceptual mapping STUPIDITY IS
DENSITY is motivated by an integration of different events (179). If something is dense, it is
hard to go through; the mind is like a thing in that it also has texture; if the mind is dense, it is
hard to let ideas in. Allan thinks culture plays its role in the mapping too. Wood, earth and
food are common and of everyday use and not of high value. However, why have these
substances entered the source domain instead of other things of low value? Allan believes that
there are two possible reasons: ―linguistic failures‖ and convention. In my view, Allan‘s
discussion is thought-provoking but not very convincing. Etymological, typological, and
cross-cultural evidences would sound necessary.
In their joint paper ―Sense and sensibility: Rational thought versus emotion in
metaphorical language‖, Päive Koivisto-Alanko and Heli Tissari also make similar
chronological explorations into metaphors, though their interest lies in the complex emotion-