How to Represent Abstract Concepts? From the Perspective of Conceptual Metaphor Theory

For concrete concepts, which have their physical referents, can be perceived directly through people’s physical interaction with them. Unlike concrete concepts, abstract concepts, such as morality, power, and time, can not be seen, heard, touched directly through sense modality, thus do not have direct perceptual characteristics. How abstract concepts are acquired, represented, and processed?

Traditional theories held the view that human think in a purely symbolic language ^{15, 16, 17} and concepts are represented as artificial symbols without perceptual characteristics. These amodal symbols are non-perceptual without sensory-motor information in any modality. That is, amodal symbol representations are stored separately from sensory experiences, even though for concrete concepts with sensory information. Concrete concepts are encoded into symbols, like feature lists, semantic networks, and propositions, which do not own original perceptual information anymore, of course for abstract concepts that do not contain the perceptual information. According to amodal theories, both concrete concepts that contain perceptual information and abstract concepts that superficially do not have any perceptual characteristics, are integrated at the symbolic level of representation and then are processed in high-level cognitive processes.

Embodied theories ^{6, 18, 19, 20} proposed different views that representations of concepts are inherently perceptual building on concrete sensorimotor information. A growing number of empirical evidences have investigated sensorimotor activation in cognitive processing. Several accounts of grounded cognition ^{1, 2, 6, 12, 18, 19, 21, 22} were put forward to explain how abstract concepts are grounded in experience and sensory-motor systems. For example, the perceptual symbol system claims the grounded underlying process of abstract concepts depends on perceptual simulations. Conceptual metaphor theory maintained the understanding of abstract concepts depends on the perceptual information of concrete concepts through conceptual metaphors. Both share the basic view that perceptual information is necessary to conceptual understanding and cognitive processing is embodied and grounded in physical experience consisting of (re)activation of multimodal representations through direct physical interactions with physical referents in the world.

From the perspective of conceptual metaphor theory, I discuss three questions in this review: the first is about the processing of underlying metaphoric mapping from concrete information to abstract concepts. The remaining is about the nature of the underlying processing. Is it automatic or not? Is the metaphoric mapping between concrete (source domain) and abstract concepts (target domain) symmetric or asymmetric (source to target vs. target to source)?

Lakoff and Johnson (1999) ² put forward that concrete concepts (e.g., verticality: up/down) are used to represent and describe abstract concepts (e.g., emotion: positive/negative) through metaphoric mappings (can be called metaphor for abbreviation). A metaphor constitutes two domains: the source domain, often referring to concrete concepts, provides the conceptual source with direct physical experience to target domains which is the subject matter of one sentence and often referring to relatively more abstract concepts ².

The metaphoric mappings from source domains to target domains are characterized as image schemas. In other words, image schemas are perceptual-motor gestalts that structure and provide sensory-motor grounding for abstract concepts. For instance, the image schema “SOURCE-PATH-GOAL” provides relational structures to the experience of moving our bodies from one starting location to the ending location (source domains). The abstract concept (our goal of life, success) is structured through this metaphoric mapping from the experience of source domains with moving experience. Such repeated situations of co-experience between concrete and abstract concepts form the metaphoric mappings over time. Gibbs (2006) ²³ claimed that image schemas are not sensory-motor representations themselves but analog representations of mostly spatial relations and movements. Thus, image schemas are considered to be grounded for their sensory-motor experience originality.

To test the underlying process of metaphoric representation of abstract concepts empirically and directly, two views were put forward to provide the empirically testable hypothesis: the Metaphoric Structuring View ^{24, 25} and the Stroop-like Interference Effects ²⁶. Both are extensions of conceptual metaphor theory, which provided the possibility to test the directionality and automaticity of the underlying process of metaphoric association empirically.

The Metaphoric Structuring View ²⁴ is derived from the metaphoric representation view ². The view claims that metaphors play the role of providing relational structure from concrete domains to abstract domains because relational structures of abstract concepts may not be obvious from world experience and thus only can be “imported from concrete domains” just like analogies.

For example, time does not contain concrete information, which lacks the relational structure in the world. Time is experienced as a unidirectional continuum along which objects and events appeared and disappeared in our experiences. Across language, people use spatial terms, such as “ahead/behind” for English speakers and “up/down” for Chinese speakers, to talk about time, such as “we are looking forward to a brighter tomorrow”, “The meeting is ahead of this Friday”, “next Monday” and so on. According to the Metaphoric Structuring View, the relational structures in spatial schemas are used to understand and organize events in time.

Boroditsky’s experiments tested whether time is structured through spatial metaphors ²⁴. Participants firstly judged several priming questions (either ego-moving or object-moving spatial schemas) about spatial relations of objects in pictures. After that, participants were presented an ambiguous temporal sentence (e.g. Next Wednesday’s meeting has been moved forward two days) and judged which day the meeting had been rescheduled. The meeting should be on Friday if participants were primed by the ego-moving spatial perspective which induced that “forward” is the direction of observers’ movement; whereas the meeting should be on Monday when participants in the time-moving condition where “forward” is considered the direction of time’s motion. Results proved that space and time share similar relational structures, consistent with the Metaphoric Structural View.

An issue related to the metaphoric structures is whether the source concrete domains are necessary to structure the abstract domains. Murphy (1996) ²⁵ commented on this issue and claimed that conceptual metaphor theory can make strong or weak versions of metaphoric structuring. The strong version assumes that the source concrete domain is automatically activated to structure the abstract domain which does not have its structure. The weak version claims that source concrete domain is not necessarily activated when understanding the abstract domain. Under this view, the relational structure originated from source domain is stored at the abstract domain, and after repeated use, abstract domain can have its structure, but the structure can be influenced by concrete domain because it is formed and stored through repeated relations with concrete domain. In other words, concrete domain is not necessary for the processing of abstract domain, but processing of the concrete domain can influence the concurrent processing of abstract domain, however, the opposite influence (i.e. processing of abstract domain on the processing of concrete domain) would not appear.

Boroditsky (2000) ²⁴ distinguished the two versions of Metaphoric Structuring investigated by investigating whether spatial schemas are necessary to understand time. Participants did a two-page questionnaire with the first page of either spatial schema (i.e., ego-moving schema and object-moving schema) or temporal schema prime question (i.e., ego-moving schema and time-moving schema) and second page of ambiguous spatial target question, or temporal target question (e.g. Next Wednesday's meeting has been moved forward two days. Which day is the meeting now that it’s been moved?). Schema-consistency effect appeared in space-to-time condition, indicating people can use spatial schema (concrete domain) to represent time (abstract domain); however, participants were not influenced by temporal primes when interpreting ambiguous spatial target questions, which is consistent with the weak metaphoric structuring view (spatial schemas can be used to, but not necessary to think about time). Further, a similar pattern of results was found: spatial distance affected estimates of duration distance, but duration did not affect estimates of spatial distance ²⁷.

Actually, such asymmetricity of weak version is consistent with the view of Lakoff and Johnson (1980, 1999) ^{1, 2} that there is direction in metaphoric associations between concrete domain and abstract domain. Specifically, understanding abstract concepts is based on concrete sensory-motor experience, but concrete concepts do not depend on abstract concepts. Piaget and Inhelder’s (1972) ²⁸ development view also supported this asymmetrical view that sensorimotor achievements occur before abstract thoughts do. Afterward, this asymmetrical view is accepted and called the Strict Directionality Hypothesis ²⁹.

Murphy (1996) ²⁵ also agreed the directionality nature of the conceptual metaphors and the tendency for concrete words taking on more abstract meanings over time ³⁰. He claimed that this asymmetry reflected the underlying conceptual structure of metaphors, as well as discourse and conceptual differences. For example, time must be understood in terms of space, but not vice versa. In daily language use, people often say “love is a rose”, but not “a rose is love”. Murphy explained the asymmetricity that abstract domains are more complex and thus harder to describe than concrete domain, so abstract concepts’ meaning often reflect a progression from concrete to abstract domain, so words describing the concrete sensory experience can be understood as to describe abstract concepts.

Although the Strict Directionality Hypothesis and the weaker version of conceptual structuring were proposed from different perspectives (linguistic analysis and conceptual structure), they share the same view that the direction of metaphoric mappings is asymmetrical.

Lee and Schwarz (2012) ³¹ presented the operational definition of the directionality nature of conceptual metaphors. If (a) manipulation of the concrete domain affects measurement in the abstract domain (concrete-to-abstract), and (b) manipulation of the abstract domain affects measurement in the concrete domain (abstract-to-concrete) are both true, then the metaphoric effect would be considered bidirectional. If either (a) or (b) is true, then the metaphorical effect would be considered unidirectional. The unidirectionality implied by the Strict Directionality Hypothesis and the weaker version of conceptual structuring means concrete-to-abstract, but not abstract-to-concrete, i.e., (a) is true, but not (b). Lee and Schwarz explicitly raised the bidirectionality which is completely different from the common assumption about the directionality of conceptual metaphor theory. They argued primary metaphor’ emergence originated from the conflations between concrete and abstract domains during early life experience. Such experiential correlation causes neural coactivation of concrete and abstract domains. And the cross-domain neural connections are supposed to provide the biological foundation for the cross-domain conceptual structure, i.e. conceptual metaphor. Concrete domain provides image-schemas to understand abstract domain. They also proposed that a conceptual metaphor has both linguistic and psychological consequence and claimed that even though it is unidirectional for linguistic expressions; its psychological consequences may be directional. They claimed that sensorimotor experiences and abstract domain are in dynamic interaction, and infers that sensorimotor experiences can have an impact on abstract domain, but also can be shaped by abstract domain ^{14, 32, 33}.

Even though the unidirectional nature of conceptual metaphor is widely accepted by theoretical explanations, as well as supported by empirical evidence, there is also much research supporting the bidirectionality nature of conceptual metaphor. Table 1 summarized the studies on the directionality of metaphoric mapping between concrete domain and abstract domain.

As shown in Table 1, it is still debatable about the directionality of conceptual metaphors. To attempt to answer such a question, researchers proposed that metaphoric mappings could be modulated, such as primed types and languages. For example, Lakens (2010) ³⁴ claims that metaphoric representations of abstract concepts are highly contextualized, and different representations can be primed (for example, in Boroditsky (2000), time can be thought differently when participants were differently primed, i.e., in the ego-moving or object-moving frame of reference). Boroditsky explains that language has a powerful role in shaping the metaphoric representations ^{35, 36}. Nevertheless, it is worth for future research taking consideration into these questions about whether, when, and how the metaphoric mappings are unidirectional or bidirectional.

Whether metaphor is a necessary part and automatically be activated while people process concepts is another important question, which was firstly investigated by Meier et al. (2004) ²⁶. They tested the association between affect and brightness by asking participants to evaluated emotional words which were presented randomly in either black or white color. Results showed that participants’ responses were faster and more accurately when positive words in white color and negative words in black color (which we call the metaphor-congruent condition) than positive words in black color and positive words in white color (the metaphor-incongruent condition), suggesting that the brightness of stimuli is activated when doing valence-judgment. The judgment of the stimulus’ valence is facilitated when it’s meaning and color is consistent with common metaphors. Conversely, judgments are interfered when they are inconsistent. This paradigm is assumed to investigate the automaticity of metaphoric representation, because the manipulation of concrete dimension is completely uninformative and irrelevant to the task on abstract concepts, such as power considered to be automatically connected to vertical position ³⁷; the association between affect and vertical position considered to be automatic ³⁸; affective evaluations considered to bias subsequent tone categorization automatically ³⁹; closeness considered to be mapped onto similarity automatically ⁴⁰. Meier and Robinson (2004) explained the metaphor congruency effects as the results of Stroop-like inference effect ^{34, 38}.

The typical Stroop effect ⁴¹ refers to longer response times were needed when participants named the ink color of a color word depicting a color incongruent with the ink color than a color congruent with the link color. This was extended to a general interference which occurs when irrelevant stimuli co-occur inconsistently with relevant stimuli (e.g., the word “red” was presented in green font) ⁴². Automatic processes are fast and do not require attention for execution and thus automaticity account was applied to explain the Stroop effect ⁴³.

In the Stroop effect, there are two dimensions of stimuli, one (naming the ink color) requires much more attention and is processed slower than the other (reading the irrelevant word), thus the relatively slower color naming require much more attention resources which is considered to be controlled) than the relatively fast word reading that is considered to be automatic. Based on this, the automatic process cannot be withdrawn and interfere with the controlled process, but not vice versa. The interference was explained as automatic activation of the irrelevant information (e.g., the meaning of the word) when doing target task (naming the font color of the word). Macleod and Dunbar (1988) ⁴⁴ proposed the view of automaticity as a continuum against the all-or-none view of automaticity, Kahneman and Chajczyk (1983) ⁴⁵ indicated that a process is considered as strongly automatic if it is unaffected by attention-allocation strategy; or as partly automatic if it can occur largely without attention, but is affected by attention.

Similarly, the metaphoric congruent effect occurs in the situation that processing the valence of affective words was influenced by the irrelevant brightness of the words. Likewise, there are two dimensions of stimuli too: one refers to the valence of the stimuli (abstract domain) which require more attention than the other dimension, i.e., the brightness of the stimuli (concrete domain). The brightness of words was automatically activated even when participants were asked to evaluate the valence of the words by ignoring the brightness of them. Two dimensions here refer to the two domains of conceptual metaphors. The source domain is processed automatically without explicit attention even though it is uninformative for the processing of target domain. This automaticity view is consistent with Lakoff’s (1993) view ⁴⁶ that metaphoric mapping is used automatic and obligatory and it is an implicit cognitive process with neither effort nor awareness ^{34, 47}.

Meier and Robinson (2005) ⁴⁸ also advocated testing the automaticity or obligatory of metaphoric associations. Many researchers theoretically explained the process of conceptual mapping is automatic ^{38, 40, 49, 50}. However, this automaticity view has rarely been directly tested in the literature with few exceptions ⁵¹. Therefore, I reviewed the empirical studies related to the automaticity of conceptual metaphor in Table 2.

As shown in Table 2, it is clear that some studies tried to test the automatic in some extend which referred to valance-space metaphor, power-space metaphor, similarity-distance metaphor, and so on. Some studies let participants finish two tasks sequentially or doing conceptual tasks while concrete factor was controlled. However, it is inconsistent in these studies and no one tried to provide a standard about it. Usually, if participants do not have the instruction or desire to use metaphors in experiment and these metaphors indeed are activated, showing that metaphors are automatically activated. However, these inconsistent findings prevent us have a clear conclusion about the automaticity of the activation of conceptual metaphor, which could be further investigated in future studies.