Orð og tunga - 08.07.2019, Blaðsíða 71

Yuki Minamisawa: Metaphor and Collocation 59
a qualitative advantage, since frequently used metaphors are oft en
highly general, whereas metaphors that contribute to more substan-
tially to the structure of abstract concepts are specifi c, such as the flu-
id metaphor. In Turkkila (2014), for example, the four most frequent
anger metaphors were anger is a possession (have anger), a place (in
anger), a moving object (anger toward X), and an object (anger against
X), which seem to be applicable not only to anger but also to other
emotions and abstract concepts.3
To deal with this problem, the present study uses a statistical
method for calculating collocation. In Sinclair (1991:170), collocation
is defi ned as “the occurrence of two or more words within a short
space of each other in a text,” and is oft en measured using statistical
methods (Hunston 2002). Considering Sinclair’s defi nition, a meta-
phorical patt ern is a specifi c type of collocation in which a source-
domain word and a target-domain word co-occur. Thus, it is assumed
that metaphorical patt erns can also be measured using the statistical
methods used for collocations.
Although diff erent measures can be used, the present study uses
the Mutual Information score, which is the observed frequency di-
vided by the expected frequency, converted to a base-2 logarithm
(Hunston 2002). Akano (2009) states that the MI score is appropriate
for extracting semantically associated collocations, so the MI score
should also eff ectively extract metaphorical patt erns. In measuring
the centrality of metaphors, this analysis uses two criteria. The fi rst
is that a metaphorical collocate with a higher score is more strongly
associated with the emotion, that is, more central to the emotion (Cri-
terion 1). This is derived from the fact that the MI score measures
the strength of a collocation (Hunston 2002:71). The second criteri-
on is that the metaphors that contain more signifi cant collocates are
regarded as more central to the emotion (Criterion 2). According to
Hunston, collocations are considered signifi cant when the MI score is
three or higher. By categorizing the signifi cant collocates into various
metaphors, it is possible to create a list of metaphors that contain such
signifi cant collocates. In this way, we can defi ne the metaphors that
have the most signifi cant collocates as being central to the emotion.
As a trial, Table 1 gives the top 30 signifi cant collocates of anger/rage
3 Although Turkkila’s categorization seems to vary slightly from that of many previ-
ous studies, it is nevertheless obvious that the most frequent metaphors are highly
general.
tunga_21.indb 59 19.6.2019 16:55:57