Metonymy as a usage event
Some properties of metonymy
Domains and subdomains
Standard definitions of metonymy often rely on the notion of ‘single domain’ or ‘domain matrix’ within which the metonymic mapping is supposed to operate (see e.g. Radden and Kövecses 2006).12 More specifically, Francisco Ruiz de Mendoza and his coauthors have forcefully argued in a number of publications (see e.g. Ruiz de Mendoza Ibáñez and Díez Velasco 2002; Ruiz de Mendoza Ibañez and Otal Campo 2002) that metonymy amounts to two kinds of operation: domain expansion (source-in-target metonymy) or domain reduction (target-in-source metonymy). Ruiz de Mendoza and his collaborators reject part-to-part metonymies, claiming that they can be reduced to either domain expansion or domain reduction.
Ruiz de Mendoza’s theory is appealing because it reduces proliferation of analytical categories (Occam’s razor), but it hinges crucially on some operational criterion for distinguishing domains from subdomains. To my knowledge, to date no scholar has devised a satisfactory working definition. A variety of conceptual relations has been lumped together under the rubric ‘subdomain– domain’, and vice versa. The following is an in complete list of such “domain-subdomain” relationships as found in the literature on metonymy. Despite its incompleteness, this list demonstrates that the notions of domain and subdomain are used to refer to distinct conceptual phenomena:
i. Class/set– subclass/subset (hyperonym– hyponym): e.g. flower as a hyperonym of daffodil.
ii. Class– class member: e.g. playwrights with the member G. B. Shaw.
iii.Whole– part: e.g. face– nose, house– room.
iv. Person– attribute (temporary or permanent): John– potbelly, Mozart musical genius.
v. Location– institution– people working in institution: e.g. 10 Downing Street– British government– staff.
These few examples suffice to show that domain-subdomain relations are conceptually very diverse. More seriously, often it seems to be a matter of perspective as what to regard as the superordinate domain and what as the subdomain(s), respectively. In (i) the subdomain DAFFODIL is extensionally properly included in the domain FLOWER. Intensionally, however, the definition of daffodil requires more attributes than the definition of flower. The attribute FLOWER is therefore properly contained in the set of attributes defining the term daffodil. From this perspective, one could claim that FLOWER is a conceptual component (subdomain) of DAFFODIL (domain). As to (ii), extensionally, George Bernard Shaw is a member of, i.e. contained in, the class of playwrights. Intensionally, one could regard Shaw as a human being with a number of permanent and temporary attributes, one of them being that he was a playwright. In the first case, the concept PLAYWRIGHT (extensionally defined as the class of playwrights) would be the superordinate domain and Shaw one of its subdomains; in the second case, Shaw would define a superordinate domain, a component of which is the attribute PLAYWRIGHT. Regarding (iii), the relation between nose and face and room and house, respectively, is one of part-whole: that is, it is completely different from what is described in (i) and (ii). Neither is the set of noses a subset of the set of faces, nor is the set of rooms a subset of the set of buildings. The relation between noses and faces and rooms and houses, respectively, is not one of class inclusion. Again, it seems to be a matter of perspective that determines what is regarded as the superordinate domain and what as the subdomain. One could argue that a proper understanding of house includes some understanding of the concept ROOM, KITCHEN, ATTIC, BASEMENT, etc.; thus house would constitute a superordinate domain vis-à-vis room. Conversely, one could also argue that room requires reference to concepts like HOUSE, BUIDLING, COTTAGE, PALACE and the like, and postulate that ROOM is a superordinate domain from that perspective. The person-attribute relation in (iv) seems to be a clear case, at first sight. Intuitively, one tends to think that an attribute is subordinate to the person that has it. Even if this perspective is the normal one, it does not exclude another perspective where e.g. POTBELLY is a superordinate domain comprising the class of people who display the attribute of having a potbelly. Finally, consider relations such as those exemplified by (v), which also have been claimed to exhibit a domain–subdomain structure. What is the criterion for deciding that LOCATION is the superordinate domain of INSTITUTION, which itself is supposed to be the superordinate domain of PEOPLE (working at the institution)? True, 10 Downing Street, which refers to a location in London and a house at this location, can metonymically evoke the institution and ultimately the human beings working in the institution, but the same arguments as above again apply: One can see an institution as having as one of its features a location; there is no inherent necessity to view INSTITUTION as being a subdomain of LOCATION, although it can certainly be construed that way.13
Croft (2002: 166) seems to offer a viable solution to the problem of distinguishing domains from subdomains. He proposes the following definition for domain: “We can […] define a domain as a semantic structure that functions as the base for at least one concept profile (typically, many profiles).” The two notions of profile and base (adopted from Langacker 1987) are especially relevant here: According to Croft, a base is “the domain immediately presupposed by the profiled concept” (167). Croft calls the base of a profiled concept its ‘base domain’. This definition seems, at least in part, to coincide with Ruiz de Mendoza’s use of the term matrix domain. Notice that Croft uses the term ‘presupposition’ to characterize the relation between profile and base. It does not become clear, however, whether he has the notion of semantic presupposition in mind; if so, it would render statements about domain structure testable by means of the well-known (internal) negation test for presupposition.14 But the application of this test leads to unsatisfactory results: For example, the notion ARC is, in Croft’s sense, profiled against, or presupposes, the base CIRCLE. However, neither from the proposition This is an arc nor from its negation This is not an arc follows the proposition This is a circle. As another example consider the relations between NOSE, FACE, and HEAD. Does NOSE presuppose FACE, which, in turn, would presuppose HEAD? The (frame-internal) negation test fails here: Neither from x is a nose nor from x is not a nose does it follow that x is a face; analogously, for FACE and HEAD. The presupposition test works somewhat better with taxonomies on condition that negation is restricted to frame- or domain-internal negation: From This is a rose the proposition This is a flower follows; frame-internal negation (This is not a rose) would also imply This is a flower. In conclusion, even Croft’s attempt at providing a more constrained characterization of the relation between domains and subdomains is inconclusive. Presupposition, as defined in standard semantics, does not provide a sufficient criterion for distinguishing superordinate domains from subordinate domains.
An additional complication results from the fact that Croft– different from e.g. Ruiz de Mendoza– does not propose a characterization of metonymy in terms of domain expansion and domain reduction but in terms of highlighting an originally “secondary” subdomain: “[…] metonymy makes primary a domain that is secondary in the literal meaning” (Croft 2002: 179). This conception of metonymy is not compatible with the idea that metonymy is a domain-subdomain or subdomain-domain relationship.
What counts as a domain and a subdomain might be a matter of perspective or construal on the part of the conceptualizer. Take the example of french fries, which is supposedly a subdomain of the domain of customers (see Ruiz de Mendoza Ibáñez and Otal Campo 2002: 132 for an analogous example). I think it is just a matter of perspective to reverse the sub-domain–domain relationship in this case. French fries are a specific kind of cooked food (potato sticks fried in oil), often (but not necessarily) served in fast food restaurants to customers who are in a hurry and do not want to spend much time and money on a meal. There is thus a whole knowledge structure concerning french fries and, one could argue that part of this knowledge structure is the concept of customer. From this perspective, FRENCH FRIES constitutes the superordinate domain of RESTAURANT CUSTOMER. There is thus again no clear objective sense that would allow one to distinguish domains from their subdomains.
The preceding discussion should not be interpreted as implying that I argue against the existence of frames, domains, etc., as such. On the contrary, the present author believes that the development of semantic frame theory constitutes a major breakthrough in the history of natural language semantics. My intention has been to point out that the concepts of super- and subordination, with respect to frames or domains, are in need of more precise explication.
For the time being, a semiotically based definition of metonymy (and metaphor) seems preferable to a definition relying on the notions of domain and subdomain.15 Furthermore, I have proposed that metonymy is an operation of conceptual elaboration. In elaborating a source concept, metonymy relies in general on pre-established inferential patterns. The kind of conceptual realms in which metonymic shifts operate are not necessarily whole cognitive domains (and subdomains) stored in long-term memory, but they might be more like mental spaces, i.e. “small conceptual packets constructed as we think and talk, for purposes of local understanding and action” (Fauconnier and Turner 2002: 40). Mental spaces tap portions of frames stored in long-term memory, but it seems implausible that whenever a metonymic operation takes place a whole conceptual frame is activated.
12. A domain can be equated roughly with what other scholars have called frame (Fillmore 1982) or idealized cognitive model (Lakoff 1987).
13. See e.g. the analysis proposed for Wall Street by Ruiz de Mendoza Ibáñez and Díez Velasco (2002: 513) and Ruiz de Mendoza Ibáñez and Otal Campo (2002: 65), where it is assumed that the location Wall Street is the matrix do main for an institution, which itself constitutes the matrix domain for people (working in the institution).
14. The kind of negation that is relevant for semantic presupposition is frame-in-ternal negation (see Fillmore 1982).
15. This conclusion supersedes the definition of metonymy provided most recently in Panther (2005) and Panther and Thornburg (2003).
