Blending Theory (BT)
In their comprehensive exposition of BT, Fauconnier and Turner (1996, 1998, 2002) claim that many cognitive operations, including metaphors, involve the integration of different mental structures into a single representation. They call this process ‘blending’ and the resulting representation ‘blend’. According to these scholars, the human mind organises new information in mental spaces, which are:
[S]mall conceptual packets constructed as we think and talk, for the purpose of local understanding and action. They are very partial assemblies containing elements, structured by frames and cognitive models (...). Mental spaces are interconnected in working memory, can be modified dynamically as thought and discourse unfold, and can be used generally to model dynamic mappings in thought and language (2002:102).
Each space contains some elements taken from ‘frames’, or ‘schemata’, which are stored in our long-term memory. Mental spaces are constructed for specific purposes of understanding and are modified by individual situations. As mentioned in the quotation above, spaces are dynamic and can establish connections with other spaces, thus creating networks. If exposed to new external stimuli, our mind will initially activate a default mental space based on entrenched associations. However, it will also be ready to integrate dynamically this default mental space with the elements from other spaces as new information is processed. The set of correspondences created between the elements of the two spaces will produce a new mental space that fits the new given scenario, or experience (2002:102-103).
This general process can be fruitfully used to explain metaphors whose understanding is mainly based on inferential reasoning. For instance, if we consider the metaphor ‘This surgeon is a butcher’ discussed by Grady et al. (1999: 103-105; cf. Figure 1), we soon realise that it cannot be explained by approaches that treat metaphor as a uni-directional process, that is a mapping from the source into the target domain (as in the CMT model).
The problem here lies in the fact that neither the source domain (butchery) nor the target one (surgery) of this metaphor contains the feature of ‘incompetence’ within their organising frame. Rather, its central inference (the surgeon is incompetent) results from merging elements, or structures, of the two domains together. The BT model can account for the inferential process created by this metaphor.
Fauconnier and Turner’s basic model of blending, or ‘integration network’, consists of four main spaces that are hierarchically connected (cf. Figure 1). There is a ‘generic’ space that contains the elements that the two input spaces share. In the specific example of the metaphor above, Grady et al. list elements such as ‘agent’, ‘undergoer’, ‘sharp instrument’ and so on (cf. Figure 1). Nowadays some scholars tend not to reproduce this space because its constituent elements are already present in the input spaces (i.e. Kovecses 2005: 269). I will do the same during my data analysis.
Figure 1: Grady et al.'s (1999:105) conceptual integration network: surgeon as butcher
The composing elements of the two input spaces derive from the domains activated by the relevant linguistic expressions. In the example above, the two (mental) input spaces (surgery and butchery) draw their structure, or elements, respectively from the source and the target domain that are relevant to the local situation presented in it (i.e. ‘surgeon’, ‘butcher’, ‘scalpel’, etc.). It is also possible to establish a set of correspondences between the counterparts in each space (i.e. ‘operating room’ correspond to ‘abattoir’, ‘surgery’ corresponds to ‘butchery’).
The forth space is called ‘the blended space’, or ‘the blend’ and it is where some elements from each input are projected (i.e. ‘scalpel’, ‘healing’, ‘cleaver’ ‘butchery’, etc.). Projection is selective, which means that not all the elements in the inputs reach the blend. Only relevant or central elements for understanding are projected. These elements then integrate and form a new ‘emergent structure’ (in our case, ‘incompetence’). This process is based on composition (new connections are created between the elements of the inputs once projected in the blend), completion (addition of structures not present in the inputs but deriving directly from our knowledge of the elements that shape them) and elaboration (addition of extra structures that might logically fit in the scenario presented in the blend. This process is also called ‘running the blend’). Fauconnier and Turner (2002: 42-44) also remind us that the blend never detaches itself from the other spaces. On the contrary, the blend is an important tool by which it is possible to map back the relations between the counterparts in the two inputs and the features they share in the generic space (Coulson 2001:178-196 on retrospective projection). In their discussion of the ‘surgeon is a butcher’ metaphor, Grady et al. (2001: 106) show how the notion of incompetence is the result of the composition process that combines the butcher’s ‘means’ and the surgeon’s ‘goals’. Clearly, the blending process is context-dependent and highly selective. Even though it is possible to create multiple compatible or incompatible spaces from the same domain, the context will help us to select the most adequate or relevant.
Throughout their work, Fauconnier and Turner offer some compelling evidence of the validity of BT. However, the high flexibility of this model may cast some doubts on its validity (cf. for example Gibbs 2000 on the scientific validity of BT in terms of falsifiability).
In spite of the methodological issues I have pointed out above, both CMT and BT seem to be valid approaches to the study of metaphors. More specifically, each of them focuses on a different aspect of the same phenomenon (entrenched metaphors for the former and novel metaphors for the latter). Therefore, it is not surprising that some scholars have suggested their combined use, as discussed in the next section.