Dr. Ritchie is Professor of Communication at Portland State University in Portland Oregon. Recent publications on cognitive theories of metaphor have appeared in various issues of Metaphor and Symbol, and he is the author of a Context and Connection in Metaphor Theory, published by Palgrave-MacMillan in fall, 2006.
Author’s note: The author is indebted to Ray Gibbs, Jr., Molly Major, and two anonymous reviewers for many useful insights and suggestions; any remaining errors and omissions are of course my own responsibility.
This essay compares simulation-based accounts of metaphor processing recently proposed by Gibbs (2006a) and Ritchie (2006), using examples of metaphors based on the metaphor vehicle “journey” from four different texts. From analysis of these different examples, it is concluded that simulation may come into play at different levels, depending on the metaphor and the context in which it is used. Further, it is suggested that the imaginative simulation of the object or action named by a metaphor vehicle, proposed by Gibbs, incorporates a partial subset of detail-level perceptual simulators. This leads to the proposal that the two models describe cognitive processes that operate at different levels or stages in the metaphor interpretation process, and that they might usefully be merged into a single more comprehensive model of embodied metaphor interpretation. The more comprehensive model provides a richer theoretical context for understanding how reuse and modification of a particular metaphor (Cameron, 2007) as well as the use of apparently different metaphors that activate similar simulations can influence comprehension, and how skilled orators can use these effects to accomplish complex communicative objectives (e.g., Blair, 2005; Obst, 2003).
X IS A JOURNEY:
Embodied simulation in metaphor interpretation
The idea that cognitive processes, including language and communication, are embodied has come to be widely accepted, but the meaning of “embodiment” and exactly how it is accomplished remains ambiguous. In recent work, Gibbs (2006a) and Ritchie (2006) have each proposed mechanisms for embodiment in metaphor interpretation. Both approaches are based on the idea of simulation, although applied at different levels or stages in the interpretive process.
Gibbs proposes simulation both in terms of listeners imagining the performance of bodily action described by language and as a mechanism to explain how listeners are able to draw inferences about speakers’ intentions by simulating the inner state of the speaker at the moment when an utterance is produced. Ritchie, drawing on Barsalou’s (1999; 2007) perceptual simulation theory of cognition, proposes that language, including metaphorical language, activates the partial simulation of multiple perceptual experiences associated with the metaphor vehicle. The perceptual simulation model emphasizes the simulation process at a relatively detailed level, in which a listener processing a metaphorical phrase might experience only partial simulations of moods or emotions detached from any particular experience, for example a shape, mass, or flash of color, the heft of an object or the feeling of a texture. Gibbs’s model emphasizes the simulation process at the relatively more global level, in which a listener experiences the event or object identified by a metaphor vehicle as a gestalt, as a coherent set of perceptions, or, at the even more global level of a speaker, the listener experiences the event or object as the speaker experiences it, along with the speaker’s experience of the communicative context within which it is relevant. Gibbs’s model implies that perceptions associated with the state or action described by a metaphor vehicle are simulated as a package; the perceptual simulation model implies that a listener encountering a metaphor vehicle may experience simulations of only a small subset of these perceptions, based on relevance in the immediate context.
All three forms of simulation converge at a higher level of conceptualization, but they have different implications for our understanding of metaphorical language and communication processes more generally. This essay considers similarities and differences between these approaches, and how they might be merged into a unified approach. I begin with a brief conceptual summary of the approaches, in the context of underlying theories, including Conceptual Metaphor Theory (Lakoff & Johnson, 1980). I then apply each approach to several different texts that use the common metaphorical vehicle, X IS A JOURNEY. Finally, I consider the prospect for merging these approaches into a single simulation-based model of metaphor use and interpretation.
Simulation as embodied cognition.
Embodied cognition is often contrasted with amodal computational theories, in which mind and mental activities consist purely of logical operations that could be accomplished by a sufficiently powerful digital computer as readily as by a biological brain, with no detectable differences in outcomes (see Barsalou, 1999; 2007; Clark, 1997; Searle, 1993). Implicit in all theories of embodied cognition is the idea that the sensory and motor control processes of the living organism are fundamental to its cognitive activities. Clark (1997) carries the idea even further, arguing that the physical aspects of the environment are also incorporated into cognitive processes, and that a large part of the peculiar genius of human beings is our ability to transform our physical environment in ways that facilitate and amplify our information-processing abilities.
Within cognitive linguistics, embodiment is often understood in a more limited sense, as implying that conceptual knowledge, including linguistic knowledge, is acquired, organized, and used on the basis of the body’s interactions with the physical environment, and with the social environment as it is manifested in the physical environment. Because most of our conceptual and linguistic knowledge is based on bodily experience, using and processing language inevitably involves the activation of bodily knowledge stored in the neural systems of the brain; the particular biological organization and characteristics of the neural systems are often, but not always, included within the meaning of embodiment.
Conceptual Metaphor Theory (CMT). Lakoff and Johnson (1980) argue that experienced correlations within perceptual experience provide the basis for conceptual metaphors, and these in turn provide the basis for abstract conceptual thought. Commonplace expressions such as “a warm relationship,” “a close friend,” or “a big problem” all originate in and provide evidence of correlations between physical sensations (physical warmth and proximity, perceived size) and more abstract concepts (love, friendship, problem-solving). Metaphor is primarily conceptual, and the linguistic expressions we usually think of as “metaphors” are expressions or manifestations of underlying conceptual metaphors. Conceptual metaphors are expressed in coherent systems of linguistic metaphors; to use one of Lakoff and Johnson’s primary examples, expressions such as “win” or “lose a debate,” “attack” or “defend a position,” and “undermine an opponent’s argument” all manifest a single underlying metaphor, ARGUMENT IS WAR. According to CMT, when we use or encounter these expressions, we actually experience argument as war. It follows that a close analysis of systems of metaphors will provide insight into individual cognitive processes (Lakoff & Johnson, 1999) as well as social and cultural systems of belief (Lakoff, 1996).
Some of the bolder claims of Conceptual Metaphor Theory have been disputed on various grounds (see, for example, Barsalou, 1999; Haser, 2005; Howe, 2008; Keysar & Bly, 1999; Ritchie, 2003; Vervaeke & Kennedy, 1996). However, considerable evidence has been amassed in support of the basic claims of the theory (see Gibbs, 1994; 2006a; 2006b). At the very least it seems reasonable to conclude that the metaphorical correlations proposed by Lakoff and Johnson provide a basis for grounding some abstract concepts, as Barsalou (2007) recognizes; the accumulated evidence also supports the conclusion that the complex schemas associated with conceptual metaphors play a role in metaphor use and processing at least some of the time.
Simulating bodily actions. Gibbs (2006a) argues that language interpretation involves embodied simulation in at least two senses. First, language automatically activates “construction of a simulation whereby we imagine performing the bodily actions referred to in the language” (p. 434). Second, understanding language requires listeners to draw inferences about the speaker’s communicative intentions, which can be accomplished by simulating the speaker’s experience and thoughts at the moment of the utterance. These two ideas are implicitly related inasmuch as simulating the performance of bodily actions implied by a speaker’s utterance is one way to simulate the speaker’s thoughts and experience.
The claim that listeners must consider and draw inferences about speakers’ communicative intentions is defensible on logical grounds, if we assume that speakers and listeners actively use and maintain something like common ground (Clark, 1996) or a mutual cognitive context (Sperber & Wilson, 1986). However, an experiment by Barr and Keysar (2005) yielded evidence that communicators do not necessarily consider differences in common ground even when these differences are apparent. Subjects who were not visible to each other were required to communicate about the shape and orientation of visual images, and they quickly developed idiosyncratic terminology to describe the shapes. When one member of the team was replaced partway through the experiment by a new member who had not heard the preceding exchanges, the other subject tended to continue using the specialized terminology that had been developed during the first part of the session, in spite of the newcomer’s lack of experience with it. A clear implication is that listeners are likely to consider speakers’ specific knowledge or speakers’ specific intentions only when something about the exchange is problematic, or perhaps when the outcome is of particular importance.
The assumption that listeners always consider speakers’ communicative intentions, however, is not necessary to Gibbs’s overall argument. There is ample evidence that humans, like other primates, are neurally capable of mirroring others’ actions, and that humans do tend to mirror or echo others’ communicative behavior during conversation (Barsalou, 2007; Decety & Grèzes, 2006; Gallese, 2003). Moreover, Gibbs produces extensive experimental evidence in support of the claim that inconsistencies between subjects’ actions and words they are asked to process interfere with (and consistencies facilitate) language processing and comprehension. As only one among many examples, a subject required to signal comprehension by pushing a lever will react more slowly to a verb that implies movement toward the subject than to one that implies movement away, strongly suggesting that motor control neurons consistent with the action of a word or phrase are at least partially activated during processing (Klatzky et al, 1989; see also, for example, Boroditsky & Ranscar, 2002; Glenberg & Kaschak, 2002; Zwaan, 2008). For an extensive review see Barsalou, 2007; Gibbs, 2006b). It seems reasonable to conclude that simulating the action (or state) associated with language plays at least some role in everyday language processing.
Perceptual simulation. Noting that the perceptual neural system aggregates (filters, combines, and summarizes) perceptual experience in convergence zones at ever higher levels of abstraction, up to the conscious experience of objects and action sequences as coherent entities (Damasio,1989; Damasio & Damasio, 1994), Barsalou (1999) suggests that a conceptual neural system parallels the perceptual neural system at every level. Within the conceptual neural system, a huge number of perceptual simulators develop in long-term memory, each linked to particular feature and association areas. Any simulator can produce an infinite number of partial perceptual simulations of the associated concept. For example, a simulator associated with the concept of cave might in one context or situation produce a very limited partial simulation of the visual perception of darkness and the tactile perceptions of coolness and dampness, perhaps accompanied by emotional perceptions of claustrophibia-related panic. In a different context, the cave simulator might produce a more complete simulation of a dark, cool, slightly damp place, with the associated simulations of subdued echoes and emotional responses related to mystery as well as claustrophobia. In yet another context the same simulator might produce a more detailed simulation of a large, open place dimly lighted by electric lamps, with stalactites, the sound of dripping water, muffled voices, and echoes, a musty smell, and so on. A simulator associated with the social category, my significant other, might produce a simulation of quiet conversation or a candle-lit dinner on one occasion, on another occasion might produce a simulation of love-making, and on yet another occasion of holding hands while walking in the park (Niedenthal et al., 2005).
Simulators can be activated by language, and language constitutes a powerful system for activating simulators. Conversely, both direct perception and simulations can activate language, and this process of activation plays a role in language production. Simulators interact with perception at every level; simulations may be compared with perceptual features as part of the process of recognition and identification, and simulators may fill in missing features of a perceived object. For example, the sight of a whole watermelon will, for most of us, activate a simulator that can fill in the color and texture of the flesh inside. If a wild animal is partially seen in the woods and tentatively identified as a deer or elk, the simulator may fill in details including body shape, antlers, and so on. The processes of recognition and filling-in are far from reliable, as is attested by countless hunting accidents every autumn.
In addition to the usual “five senses,” perceptions include proprioceptive and introspective awareness of internal bodily states, emotions and cognitive processes, including the logical comparisons between raw perception and perceptual simulation that constitute recognition, classification, and evaluation. Stipulating a complete set of simulators matching this extended range of perceptions, Barsalou (1999) demonstrated that, in principle, all cognitive processes including abstract reasoning could be accomplished by way of simulations.
In more recent work, Barsalou has explicitly acknowledged the usefulness of a theoretical approach that combines perceptual simulators with amodal computational theories, and has begun to integrate the Perceptual Symbol Systems (PSS) with other approaches to cognition and language, leading to a more comprehensive theory of Language and Situated Simulation (LASS). LASS incorporates more conventional approaches to language processing along with perceptual simulations (Barsalou, 2007; Barsalou et al., 2007). When language is encountered, connections with other words and phrases (e.g. as modeled by Burgess & Lund, 1997 and Landauer & Dumais, 1997) are immediately activated, as are associated perceptual simulators. For simpler tasks, the superficial processing by means of associations between words may be sufficient, and the deeper conceptual processing by means of perceptual simulators may not be required. This helps explain findings, reported by researchers such as Kintsch (1998; 2007), Landauer (2007; Landauer & Dumais, 1997) and Louwerse (2007), that computer programs based on word-word association matrices perform at levels comparable to human agents on multiple-choice vocabulary tests, grading student essays, and many other language tasks. It also helps answer a principle criticism of the Latent Semantic Analysis approach advocated by Kintsch, Landauer and Dumais, and others in their group, by showing how language is grounded in perceptual and motor experience (see Landauer, 2007).
On the one hand, it is likely that words vary in the degree to which they activate perceptual simulators. As Landauer (2007; Landauer & Dumais, 1997) point out, many words are encountered first, and often only, in reading and may not be grounded at all in embodied experience; these are unlikely to activate many perceptual simulators if they activate any at all. On the other hand, when a word or phrase activates other words that are closely associated in semantic memory, these may in turn activate perceptual simulators in addition to those activated by the initial word or phrase (Barsalou, 2007; Barsalou et al, 2007). Thus, depending on the requirements of the task and the cognitive resources available, these systems may operate independently or may interact in complex ways.
Context-Limited Simulation Theory (CLST). Ritchie (2003; 2006), while accepting the fundamental claims of Conceptual Metaphor Theory about the embodiment of everyday metaphors, has rejected the implications that encountering a metaphorical expression necessarily activates the full conceptual metaphor and that we necessarily always experience the metaphor topic as the metaphor vehicle (Lakoff & Johnson, 1980). Drawing on Barsalou’s work, Ritchie suggests that, along with associated words and phrases, an array of perceptual simulators may be activated by metaphorical language. Especially in the case of frequently-encountered words or phrases, both the associated words and the activated simulators vary in the degree to which they are associated with and considered to be part of the commonplace or definitional meanings. When any word or phrase is encountered, the words and simulations that are not relevant in the present context, that cannot be readily connected with ideas already activated in working memory (Sperber & Wilson, 1986) may not be activated in the first place and if they are activated, are reduced in activation or suppressed altogether. Those words and simulators that are relevant in the present context will be more highly activated (Gernsbacher et al., 2001; Kintsch, 1998), and will produce context-relevant simulations that will attach to the topic as part of the meaning in this context.
In the case of metaphorical usage, both the words and the perceptual simulators that are most closely associated with the customary “definition” are more or less irrelevant in the current context, and are unlikely to become and remain very highly activated. Conversely, some of the words, and some of the simulations produced by the simulators associated with nuances of experience and activated by the metaphor vehicle are highly relevant in the current context. These definitionally less central but contextually more relevant words and simulations will be more highly activated and will become connected with the topic of the metaphor as part of its meaning in the current context. None of this necessarily involves active processing of the underlying conceptual metaphor. Thus, a phrase such as “attack her argument” is likely to activate perceptual simulations of emotional nuances such as hostility and anger associated with attack, regardless of whether a detailed WAR or PHYSICAL CONFLICT schema1 is activated; other simulators associated with attack may or may not become and remain activated, depending on the overall context.
A primary criticism of Conceptual Metaphor Theory is that commonplace expressions can often be mapped onto any of a number of underlying conceptual metaphors, and that these mappings are consequently indeterminate. For example, Vervaeke and Kennedy (1996) point out that many of the expressions which Lakoff and Johnson (1980) associate with ARGUMENT IS WAR (“win,” “strategy,” “victory”) can as readily be mapped onto an alternative conceptual metaphor such as ARGUMENT IS BRIDGE (see also Haser, 2005; Howe, 2008). Semino (2008) shows that a more general conceptual metaphor, ANTAGONISTIC COMMUNICATION IS PHYSICAL CONFLICT better explains the data. Ritchie (2003; 2004; 2006) proposes that these expressions are often connected in “fields of meaning” on the basis of similar sets of perceptual simulators they potentially activate. Particularly expressive metaphor vehicles often develop as generic metaphors (Tourangeau & Rips, 1991) that can be applied to a wide range of topics.
X IS A JOURNEY.
One familiar generic metaphor, “X IS A JOURNEY,” is used across many languages, and is applied to a multitude of experiences and processes, ranging from the human life-span to working a crossword puzzle (Feldman, 2006; Lakoff & Johnson, 1980). Gibbs (2006a) cites a highly evocative example from an essay by Obst (2003), “GRIEF IS A JOURNEY.” Other examples appear in Cameron’s (2007; Cameron & Stelma, 2004) analysis of metaphors from a series of reconciliation talks, Tony Blair’s spring 2005 address to the Labour Party at Gateshead (Ritchie, 2008) and the data from a focus group conversation among a group of scientists analyzed by Ritchie and Schell (2008). Since the metaphor is used in different ways, applied to a variety of topics, and developed to different extents in each of these texts, these texts collectively provide a basis for comparing various approaches to embodiment. In this section, I will analyze each in turn, and discuss how alternative concepts of simulation might be applied to each.
“Grief is a journey” Obst (2003) provides an elaborated and fully-developed example of the “journey” metaphor in an essay written for a web-page devoted to providing information and support for professional counselors and laypersons interested in the grieving process. Obst artfully exploits the metaphor both as a device to organize her essay and as a vehicle for expressing several aspects of the grieving process. As such, this example tells us little about the spontaneous use of the “journey” metaphor in everyday conversation, but it is quite revealing about how a widely-known generic metaphor can be deployed as a rhetorical resource.
Obst begins by describing a literal journey, a vacation that is “invigorating and full of adventure.” She then mentions two metaphorical “journeys,” college, “toward the goal of a career” and a new relationship, “filled with discoveries,” before introducing the unifying metaphor, “grief is a journey.” Thus, Obst describes life itself as a series of “journeys,” some literal (vacations), others metaphorical; some goal-directed (college), others exploratory and open-ended. The more general metaphor, “life is a journey,” is never explicitly stated, but provides an implicit frame for the entire essay.
Obst returns to the “journey” metaphor repeatedly, weaving it around several other conventional metaphors. In the double-metaphor, “getting over,” grief is simultaneously a set of “wounds” (“obstacles” to health) from which we eventually “recover,” and an “obstacle” to continued “progress” along the “journey” of our life that we must “get over” before we can “move forward” again. In the fourth paragraph, grief is transformed from an “obstacle” or “wound” to “get over” into a “deep, dark tunnel” (which Obst contrasts with a “cave” that has no “exit”), something to “get through.” Finally, the possibility is raised, that the person we have “lost” may be preventing us from “moving on” because we are unable to “let go.” Obst raises this possibility only to deny it. Instead, she assures us that we can create a new relationship through which we can remain “connected” and, in effect, “take the loved one with us” as we continue along the “journey” of life.
Analysis of this essay in terms of Conceptual Metaphor Theory would be tricky, since Obst continually shifts the metaphoric mappings. Grief is mapped onto a series of conceptual metaphors – it is a “journey,” an “obstacle” on our life “journey,” a “deep, dark tunnel” we must “pass through,”“wounds,” a “tyrant” that “oppresses” us, then a “murderer” that “strangles any joy.” In addition to grief, other concepts including life and learning are also mapped onto “journey.” The result is a complex web of mappings that could be detailed and explained, but it would be a complex explanation, and it would not do much to explain why Obst chose this particular mélange of metaphors, or how grief could be experienced “as” so many different things in so short a span.
If we look at the perceptual simulations associated with Obst’s sequence of images, the picture seems much clearer. By starting with a literal journey, an adventurous vacation, Obst activates expectancy, excitement, and happiness, simulations that are reinforced by comparisons to college and beginning a new relationship. Ending the opening sequence with beginning a new relationship is particularly telling, since the essay is all about the grief of the termination of a relationship by death. Opening with the upbeat, joyous simulations associated with beginnings and discoveries frames the essay in a positive and affirmative mood, primes positive emotions associated with “journey,” and lays the conceptual groundwork for the idea that death and grieving can transform rather than terminate a relationship. The series of metaphors leads the reader through all of the negative emotions, ideas, and bodily states associated with grieving, back to the positive, hopeful entailments of “journey,” of “going somewhere,” of activities “filled with discoveries.”
Once we analyze Obst’s images in terms of perceptual simulation, the apparently confused tangle of mixed metaphors makes complete sense. Throughout the essay, her metaphors activate and juxtapose simulations of loss, pain, suffering with change, transformation, and hope. Grief is a “wound,” activating simulations of pain and disablement or disfigurement, but we are “recovering” from it, simultaneously activating simulations of relief, restoration, and “feeling better.” We are “in the depths” and “in the dark” but we are not in a “cave, with no way out”; rather we are in a “tunnel,” that leads “to the other side – to where we can begin again.” Even the analysis of “filled with discoveries” is more straightforward once we abandon the insistence that a “CONTAINER” metaphor must somehow be “blended” (Fauconnier & Turner, 2002) with an “EXPLORATION” metaphor and focus instead on the context-relevant simulations activated by these two metaphor vehicles and how they interact with the stream of simulations activated by Obst’s sequence of metaphors.
At this point it is worth acknowledging a paradoxical element in analyzing perceptual simulations. Metaphors are often used to express nuances of thought and feeling that are difficult or impossible to express in literal language – that is why figurative language is necessary. The perceptual simulations activated by a metaphor like “depths of a dark cave” or “filled with discoveries” are complex and subtle; they will be experienced differently by each reader, and they defy simple labels. Unable to assign simulations to clearly labeled categories, as in a conventional content analysis, the analyst is left with two complementary tactics: Point toward the potential for activation of simulators and producing context-relevant simulations, and, in a text such as this, in which metaphor is piled upon metaphor, point toward the association of very different metaphors with very similar simulations.
Gibbs (2006a) argues that the reader will imagine or simulate the bodily action or state described by the metaphor vehicles in Obst’s essay. Given the vividness of most of these metaphors, it does seem likely that a person who is reading the essay with more than casual attention will imaginatively “experience” (as a more or less detailed simulation) many of these images – a vacation journey, a new relationship, recovering from a wound, being inside a cave or tunnel and looking for a way out. However, these bodily actions and states need not be simulated in much detail. The reader is not likely to smell the dampness of the cave, feel the hardness of the stone floor, or hear water dripping into a shallow pool in the depths of the cave, any of which might be activated by the use of a “cave” metaphor in a different context. Rather, those elements of the experience that are relevant in the overall context of the essay (and, for some readers, relevant in the context of a personal experience of grief) will become more highly activated and will be attached to the topic of grief while the less relevant simulators are suppressed. But the most strongly activated simulations will be the proprioceptive, introspective, and emotional simulations that are activated in conjunction with the visual, tactile, and motor control simulations. These subtle and intangible simulations remain activated even as the prose moves on to other images, other metaphors. The effect of the essay is accomplished by the activation, accumulation, and reinforcement of thought, experience, and emotion, and its interaction with the overt message conveyed in the more literal parts of the text.
In this instance, simulating the writer and recovering the writer’s intention seems more important to the analyst, to the cognitive researcher and theorist, than to the general reader. Perhaps other grief counselors may need to draw reasonably accurate inferences about what Obst intended, but a person who is seeking solace from a recent loss may not find it either useful or necessary. For the grieving person, it seems that the intent of this essay is to “lead” the griever “through” a series of images that will “show the way through” the grieving process.
The meaning of Obst’s essay, for the grieving person, is not to recover Obst’s theory of grief. Rather, for the grieving person the meaning is found in experiencing a sequence of intense perceptual simulations that will connect with the experience of grief, and either alter or replace the perceptual simulations that previously dominated the sufferer’s cognitive environment. Thus, the intention of this string of metaphors is to transform both the sufferer’s concept of grief and the sufferer’s present experience of grief. The reader of this essay will experience the simulations activated by the sequence of words and phrases directly, in the context of her or his own cognitive context, and not necessarily in the context of the author’s cognitive context. Indeed, the author, her voice, and any sense of her communicative intention may disappear entirely as the reader engages with the series of powerful simulations she evokes.
That having been said, a language researcher (or another therapist) might wish to recover Obst’s communicative intention. Experiencing the sequence of perceptual simulations activated by her language (literal as well as metaphorical), in the context of language theory and grief counseling theory and practice, would enable the reader to experience detailed simulations of the grief and counseling experiences on which Obst bases her essay. Ultimately this more detailed simulation would provide the basis for vicariously experiencing the ideas and feelings Obst experienced as she wrote the essay – for simulating, in other words, what it may have been like to have been Obst, writing this essay, as suggested by Gibbs (2006a).