2.1 On the Nature of Primary Metaphors

Both the source and the target concept of a primary metaphor are equally basic concepts, which differ from each other only with regard to whether they are ultimately grounded in a particular sensorimotor domain. Primary Metaphor Theory (PMT) claims that the connection between the two concepts is strongly unidirectional, such that the source is used to conceptualize and talk about the target, but not vice versa. In the following, we take up each of these aspects in more detail, also responding to recent findings and suggestions.

2.2 On the Function(ality) of Primary Metaphors

Given that primary-metaphor patterns are so ubiquitous, both within English and across languages, it is natural to ask why. At least three broad categories of hypothetical answers to this question suggest themselves:

1. (i) There is no “function”: The patterns are epiphenomenal, not “useful,” an accidental by-product of some other aspect(s) of experience or brain function that happen to lead to particular linguistic and conceptual associations. (The possibility that primary metaphors are epiphenomenal in this sense can be considered the null hypothesis with respect to function.)

2. (ii) Primary metaphors offer communicative advantages: Source concepts are sensory in nature, and by definition associated with particular “images” – in a broad sense that includes kinetic and auditory images as well as visual ones. Thus, as basic elements of a communicative setting (as opposed to the internal, emotional world of the interlocutors, for instance), they might serve as “anchors” for communication, as speakers are able to point to and agree on their referents. If relatively more subjective or “abstract” target concepts such as ‘happiness’ or ‘difficulty’ can be linked with these concepts, there may be communicative advantages to these associations.

3. (iii) Primary metaphors offer cognitive advantages: If cognition related to source concepts is in any way different in character from cognition related to target concepts, then even if we were not communicating with others, primary metaphors might hypothetically play a useful role in conceptualization – namely by allowing us to harness cognitive capacities associated with source concepts when processing (in whatever sense) target concepts.

Any of these hypotheses may be true, and the last two are not mutually exclusive. But the third is arguably the strongest and most interesting. It suggests a reason to explore possible neural substrates for “sourceness” and “targetness,” in case these point to processing differences that would lead to cognitive advantages of some kind when source and target are associated.

3.1 Localized vs. Distributed Neural Substrates

Having established that we will be focusing on distinctions at the level of neural substrates of perception and phenomenological experience, there are several related hypotheses worth exploring regarding the ways in which the substrates for SEs may differ systematically from those of TEs. The main hypothesis we explore here is that the brain activity patterns representing source experiences are organized based on cortical positions (as sets of localized “hotspots”), which we will refer to as “neural maps.” In other words, we propose that the experiences associated with source concepts such as HEAVINESS, HEIGHT, and BRIGHTNESS are more likely to have neural substrates characterized by neural maps, than are the experiences associated with target concepts such as DIFFICULTY, DOMINANCE, or HAPPINESS.

This position is generally consistent with the known functional organization of the mammalian neocortex pertaining to perception of the body and of the external world (Bednar & Wilson Reference Bednar and Wilson2015). For example, the sensory cortex represents somatic sensation (such as touch, pain, and cold or hot detection) from skin receptors approximately preserving the local organization of the periphery. A similar “body map” in the motor cortex (known as the homunculus) codes for the planning and execution of intended movements (Graziano & Aflalo Reference Graziano and Aflalo2007).

One type of cortical localization is the case of so-called “topographic mapping,” in which nearby cortical locations encode for similar stimuli (Chklovskii & Koulakov Reference Chklovskii and Koulakov2004). In other words, parts of the cortex are organized such that adjacent neurons respond to the same type of stimulus (e.g. edges of visual shapes or frequencies of sound tones), and the spatial positions of the neurons themselves encode variations that can be detected, such as relative inclination angles of an edge or the pitch of a voice. All primary sensory and motor maps in the mammalian neocortex are topographic, but higher associative areas such as the frontal neocortex are also known to follow a topographic architecture (Silver & Kastner Reference Silver and Kastner2009).

At the same time, due to the complex and far-reaching, tree-shaped structure of their input and output extensions (dendrites and axons, respectively), cortical nerve cells can also form non-topographic maps. That is, they can form complexes (so-called “cell assemblies”) that act together to perceive, process, and represent particular types of stimuli in the way that topographic maps do, though they are not physically adjacent in the same sense. Among the most evolutionarily conserved examples of these non-topographic maps (i.e. showing significant similarities among species related to humans and to each other) are the neural systems responsible for spatial representation, which have also been shown to play a role in higher cognition and language (Landau & Lakusta Reference Landau and Lakusta2009). Some of the most striking examples of these spatial maps are given by the so-called systems of “place cells” in the hippocampal formation. Place cells are neurons that are selectively active when the subject is in a particular location within an environment. For example, in the context of a residential setting, a specific hippocampal neuron might be systematically active when the subject stands by the stove in the kitchen, another neuron when the subject is in the shower, and yet another when the subject crosses the entrance doorway. The map is non-topographic because nearby locations are not reflected by nearby neurons and vice versa: the ‘stove’ neuron and the ‘shower’ neuron might be adjacent in the hippocampus (even if the kitchen and bathroom are in opposite locations in the house), whereas a neuron representing the medicine cabinet in the bathroom just next to the shower might be physically located in a relatively faraway position of the hippocampus.

Although the activity of all hippocampal neurons is spatially modulated (McNaughton et al. Reference McNaughton, Barnes, Gerrard, Gothard, Jung and Knierim1996), pyramidal cells (the main neuron types in the hippocampus) display the strongest and most reliable place fields. More generally, activity in the neural circuits of the hippocampus appears to be involved in spatial navigation (Hartley et al. Reference Hartley, Lever, Burgess and O’Keefe2014) and, of relevance to our proposal, in the spatial aspects of imagination (Bird et al. Reference 337Bird, Bisby and Burgess2012).

In both topographic and non-topographic maps, the cortical representations of experience remain fairly localized and encoded in the actual positions of the activation foci. This survey represents only a greatly simplified account of the experimental neuroscientific evidence, as the extent of the topography (and even map-like character) of cortical representation may vary substantially depending on the considered level of analysis, from microscopic neuronal assemblies to macroscopic regional activation (Kanold et al. Reference 351Kanold, Nelken and Polley2014).

The hypothesis we suggest here is that the neural representation of an SE (such as the pattern of activity necessary for representation of an experience of heaviness or brightness) corresponds to a significantly smaller set of such loci than the representation of a typical TE (such as difficulty or happiness). In this sense, the cortical representation of SEs would be substantially more localized than the representation of TEs. For instance, the physical brightness of a light stimulus is encoded as visual intensity in the primary cortex in the occipital lobe organized as a retinotopic map (that is, cortical positions actually reflect the locations in which light entered the retina at the back of the eye). Similarly, heat is encoded in temperature-sensitive patches of somatosensory cortex in the parietal lobe organized as a body map (the homunculus). Likewise, heaviness is encoded as force intensity during movement planning in the premotor cortex in the frontal lobe organized as a directional map.

If SEs are, as just discussed, associated with more localized representations, TEs appear to be associated with more distributed representations and diffuse brain states. As an extreme case of diffuse, distributed representation, many target concepts – specifically those associated with affect – appear to be considerably related to a system of neurotransmitters that modulate cortical activity across the entire cerebral cortex rather than locally (Zaborszky Reference Zaborszky2002). These modulatory neurotransmitters include, among others, acetylcholine (associated with states of purposeful arousal, as during a search activity), dopamine (positive and negative valence), serotonin (satisfaction or aversion to risk-taking), and noradrenaline (sudden and unexpected novelty). It seems that many TEs – such as those associated with happiness and difficulty – may consist of or include affective states represented by a combination of such diffuse representations throughout cortical areas. Although the literature on emotional representation in the brain is immense (and there has also been considerable study of the culturally determined dimensions of emotion), our specific emphasis on target concepts in primary metaphors leads us to focus on the neural correlates of declarative experience (Mitchell & Greening Reference Mitchell and Greening2012), as well as the most basic psychological and physiological dimensions that experiences of happiness, for instance, might share across cultures.

It is tempting to speculate that other primary target concepts, including DOMINANCE or others related to social status, may also be less correlated than source concepts with localized areas of activity, and more correlated with diffuse cortical states (see Table 2.2). If true, such a pattern would add substance to previous suggestions (e.g. Grady Reference Grady2005a) that primary target concepts encode aspects of our response to the environment, while source concepts encode our physical perceptions of it. In effect, SEs would make up something like a “physical map” of a given scenario, i.e. reflect the physical aspects of the scenario, while TEs would make up a “goal-related map,” i.e. reflect various aspects of our goal-orientation, correlated in turn with characteristic balances of neurotransmitters, and possibly other non- or less localized aspects of cortical activity.

3.2 Attentional Mechanisms

An interesting consequence of the proposed neural distinctions between SEs and TEs along the lines of cortical localization vs. neuro-modulatory diffuseness is that SEs may be more closely associated than TEs with attentional mechanisms, making it easier in some sense to focus on source concepts. According to this perspective, attention might be bootstrapped to target concepts via source concepts, potentially yielding a cognitive advantage of primary metaphors.

A brain region relevant to this consideration is the thalamus, which plays a crucial role in selective gating (Shipp Reference Shipp2004): the process of suppressing or allowing activity in a given set of neurons. The thalamus is robustly and bidirectionally connected with all areas of the neocortex and is believed to modulate relative levels of activity in various neocortical locations, giving rise to the ability to selectively experience/attend to some stimuli vs. others.⁷ Given the previous discussion of cortical localization, i.e. the positional encoding of the neural correlates of SEs, it may be the case that the attentional mechanisms associated with the thalamus can act more directly on these localized representations than on the more diffuse states associated with TEs.

An additional notion worth considering related to the aforementioned attentional mechanisms concerns a micro ‘aha’ or ‘eureka’ effect, which may again involve bootstrapping of target concepts by source concepts. An intriguing possibility is that when we “successfully” attend to or focus on a concept – i.e. when the brain successfully stabilizes into a state representing a particular experience, which might happen on the order of many times per second – the resulting neural and cognitive events correspond in some sense to effective thought, as opposed to random noisy activity (Kounios & Beeman Reference Kounios and Beeman2014). If it is easier to achieve this type of stabilization, and the resulting cognitive “reward,” with source than with target concepts – e.g. due to attentional mechanisms discussed earlier – then linking the two may enable this effect with target concepts as well. In short, SEs may be associated with more stable or identifiable representations than target concepts, perhaps mediated in part by the neural map organization discussed above, and this stability could in turn lead to a cognitive distinction and advantage.

4.1 Differences in Nonlinguistic Pitch Representations Not Due to Verbal Labeling during the Task

The pattern of spatial interference on people’s pitch reproduction performance reflected the SPACE–PITCH metaphors in their native languages: Dutch speakers could not help incorporating irrelevant height information into their mental representations of pitch (but could ignore thickness), whereas Farsi speakers could not help incorporating irrelevant thickness information into their mental representations of pitch (but could ignore height). This pattern cannot be explained by differences in overall accuracy of pitch reproduction, or in differences in musical training between groups.

Importantly, this pattern also cannot be explained by participants using language covertly during the task: labeling the pitches they needed to reproduce as “high/low” or “thick/thin.” This explanation was ruled out by the experimental design, in which nine different pitches were paired with each of nine different spatial heights or thicknesses. This “crossing” of all of the levels of pitch and of space meant that variation in each domain was orthogonal to variation in the other: There was no correlation between space and pitch in the stimuli. As such, covertly labeling high pitches as “high” (or “thin”) and labeling low pitches as “low” (or “thick”) could not produce the observed effects of space on pitch reproduction; on the contrary, labeling pitches using the spatial metaphors in one’s native language during the task could only work against the effects we predicted and found.

Rather than an effect of using language “online” during the task, these experiments show an effect of people’s previous experience using either one linguistic metaphor or the other, and thereby strengthening either one mental metaphor or the other (i.e. strengthening a non-linguistic HEIGHT–PITCH or THICKNESS–PITCH mapping in memory). To confirm that the observed effects did not depend on participants covertly labeling pitches during the task, Dolscheid et al. (Reference Dolscheid, Shayan, Majid and Casasanto2013) repeated the height interference task in Dutch speakers with the addition of a concurrent verbal suppression task. On each trial of the task, participants had to rehearse a novel string of digits while perceiving and reproducing the pitches. Secondary tasks like this have been used across many experiments to prevent participants from labeling the stimuli (e.g. Winawer et al. Reference Winawer, Witthoft, Frank, Wu, Wade and Boroditsky2007). As predicted, verbal suppression had no effect on the results of the pitch reproduction task. Dutch speakers still showed strong height–pitch interference, consistent with an “offline” effect of participants’ previous experience using language on their subsequent non-linguistic pitch representations (see also Casasanto Reference Casasanto2008b).

4.2 Does Using Different Linguistic Metaphors Cause People to Use Different Mental Metaphors?

The results reviewed so far show a correlation between people’s linguistic metaphors and their non-linguistic mental metaphors, but they do not provide any evidence that language causes Dutch and Farsi speakers to mentally represent pitch differently. Dolscheid et al. (Reference Dolscheid, Shayan, Majid and Casasanto2013) reasoned that if using THICKNESS–PITCH metaphors in language is what causes Farsi speakers to activate THICKNESS–PITCH mappings implicitly when reproducing pitches, then exposing Dutch speakers to similar THICKNESS–PITCH metaphors in language should cause them to reproduce pitches like Farsi speakers. A new sample of Dutch speakers were recruited and assigned to one of two training conditions: Participants in the “thickness training” group learned to describe pitches using Farsi-like metaphors (e.g. “a tuba sounds thicker than a flute”), whereas the other half in the “height training” group (i.e. the control group) described pitches using standard Dutch metaphors (e.g. “a tuba sounds lower than a flute”). After about 20 minutes of this linguistic training, participants in both groups performed the non-linguistic thickness interference task described above. Whereas “height-trained” participants showed no effect of irrelevant thickness information on their pitch estimates, “thickness-trained” participants showed a thickness interference effect that was statistically indistinguishable from the effect found in native Farsi speakers. Even a brief (but concentrated) “dose” of thickness metaphors in language was sufficient to influence Dutch speakers’ mental metaphors, demonstrating that linguistic experience can cause the differences in non-linguistic pitch representations found across natural language groups.

Notably, Dolscheid et al. (Reference Dolscheid, Shayan, Majid and Casasanto2013) also included a training condition in which Dutch-speaking participants were taught to use “thick” and “thin” to describe pitches in a way that contradicts both the linguistic mappings in languages like Farsi and the relationships between thickness and pitch in the natural world (e.g. they learned to use expressions like “a tuba sounds thinner than a flute”). This training intervention was identical to the “Farsi-training” condition described above in every way except for the pairing of high/low with thick/thin. Although participants learned to use the “reverse-Farsi” mapping with high accuracy (95%), this linguistic training had no effect on their non-linguistic mental representations of pitch.

The contrast between the Farsi-like and reverse-Farsi training supports a prediction of HMMT: People should be able to adopt mappings that are included in a superordinate family of mappings (e.g. the SPACE–PITCH mappings that are evident in the natural world) more easily than they can adopt mappings that are not included in the superordinate family (i.e. mappings that run contrary to, or orthogonal to, the source–target mappings found in the natural world).

4.3 When Does Language Shape SPACE–PITCH Mappings?

The results reviewed up to this point leave open the question: Do SPACE–PITCH metaphors in language cause people to develop the corresponding non-linguistic SPACE–PITCH mappings, or does using linguistic metaphors change how likely people are to use a preexisting mental metaphor? To evaluate these possibilities, Dolscheid et al. (Reference Dolscheid, Hunnius, Casasanto and Majid2014) tested 4-month-old infants on a pair of space–pitch congruity tasks. Infants heard pitches alternately rising and falling while they saw a ball rising and falling on a screen (height congruity task) or a cylinder growing thicker and thinner (thickness congruity task). For half of the trials, changes in pitch and space were congruent with the HEIGHT–PITCH and THICKNESS–PITCH mappings encoded in Dutch and Farsi, respectively, and for the other half of the trials they were incongruent with these SPACE–PITCH mappings. The data showed that infants looked longer at congruent SPACE–PITCH displays than at incongruent displays. This was true both in the height congruity condition (consistent with an earlier experiment by Walker et al. Reference Walker, Bremner, Mason, Spring, Mattock and Slater2010) and in the thickness congruity condition. There was no difference in the magnitude of the congruity effect between conditions, suggesting that there was no difference in the strength of the HEIGHT–PITCH and THICKNESS–PITCH mappings in the infants’ minds.

Four-month-olds are completely unable to produce SPACE–PITCH metaphors in language and are also, presumably, unable to understand them. Yet, they are sensitive to two of the SPACE–PITCH metaphors that are found in languages like Dutch and Farsi, and in their speakers’ non-linguistic pitch representations. These results suggest that people who use different linguistic metaphors for pitch come to think about pitch differently, not because language instills in them one spatial mapping instead of the other, but rather because language strengthens one of their preexisting SPACE–PITCH mappings, at the expense of the other.

4.4 Hierarchical Construction of Spatial Metaphors for Pitch

How could infants who are sensitive to both HEIGHT–PITCH and THICKNESS–PITCH mappings turn into adults who appear to activate only one of these mappings when they represent pitch? This process can be understood in terms of HMMT. First a superordinate “family” of mappings is established, which in the case of space and pitch includes both the HEIGHT–PITCH and THICKNESS–PITCH mappings. These mappings may be constructed, over either ontogenetic or phylogenetic time, on the basis of observable correlations between space and pitch in the natural world. The HEIGHT–PITCH mapping reflects the fact that people involuntarily raise their larynxes, chins, and sometimes other body parts (e.g. their eyebrows) when they produce higher pitches, and lower them when they produce lower pitches. It also reflects a statistical tendency for higher pitches to originate from higher locations, and lower pitches from lower locations (Parise et al. Reference Parise, Knorre and Ernst2014). The THICKNESS–PITCH mapping reflects a pervasive correlation between pitches and the size of the objects or creatures that produce them: Consider the different pitches produced by, for example, strumming thin vs. thick strings on a guitar; banging on a large steel oil drum vs. a small steel can; barking by a small dog vs. a big dog. Although Dolscheid et al.’s (Reference Dolscheid, Hunnius, Casasanto and Majid2014) data confirm that both the HEIGHT–PITCH and THICKNESS–PITCH mappings are present in infants’ minds, they leave open the question of exactly how and when these mappings become established initially.

Whatever the ultimate origin of SPACE–PITCH mappings in pre-linguistic children may be, when children learn metaphors in language, a second process begins. The findings in adults reported by Dolscheid and colleagues (Reference Dolscheid, Shayan, Majid and Casasanto2013) suggest that each time people use a linguistic metaphor like “a high pitch” they activate the corresponding mental metaphor, strengthening this mapping at the expense of competing mappings in the same family of SPACE–PITCH associations. As a consequence, speakers of height languages like Dutch and English come to rely on vertical spatial schemas to scaffold their pitch representations more strongly than on multidimensional spatial schemas, whereas speakers of thickness languages like Farsi come to rely on multidimensional spatial schemas, more strongly than vertical spatial schemas.

According to HMMT, the process of strengthening certain mental metaphors via the use of the corresponding linguistic metaphors results in the weakening of other members of the family of mappings – but this does not cause these dispreferred mappings to be extinguished. This aspect of the theory may explain the representational flexibility demonstrated in the training experiment by Dolscheid and colleagues (Reference Dolscheid, Shayan, Majid and Casasanto2013). Dutch speakers could be induced to use a non-linguistic THICKNESS–PITCH mapping (like Farsi speakers) after only a brief training intervention because no spatial mappings had to be created or destroyed; rather, the new pattern of language experience boosted the strength of the thickness–pitch mapping that had presumably been present in the Dutch speakers’ minds since infancy, causing them to think about pitch in a way that was not new, just rarely used.

5.1 Separating Effects of Language and Culture on SPACE–TIME Mappings

These data show that the experience of reading is sufficient to determine the direction of people’s implicit mental timelines (though they do not rule out the possibility that other culture-specific practices, such as gesturing about time or using calendars, could influence people’s lateral representations of time, as well). But why is this an effect of cultural experience as opposed to linguistic experience? Language can be considered an aspect of culture, and in many cases it is difficult to disentangle linguistic and non-linguistic practices. At first glance, reading might appear to be an ambiguous case, since it is a cultural overlay on natural language. Yet it is notable that reading is extremely recent and rare in human history; although reading may seem integral to language use in our culture, only a tiny fraction of all of the humans who have ever used language have been able to read. More to the point, in Casasanto and Bottini’s experiments, language was held constant across the orthography conditions. The words and phrases in natural language were invariant; all that changed was the direction and orientation of the orthography in which they appeared. Thus, changes in orthography determined the flow of time in people’s minds independently of any changes in the structure or content of language.

5.2 Hierarchical Construction of Spatial Metaphors for Temporal Sequence

How could a few minutes of exposure to a new orthography completely reverse people’s usual mental timeline, established over a lifetime of reading experience? As in the case of language, space, and pitch, HMMT may explain the flexibility of this culture-dependent mental metaphor. To elaborate, Casasanto and Bottini (Reference Casasanto and Bottini2014b) proposed that people’s implicit associations between source and target domains can be characterized, not just as families of mappings but alternatively as a set of nested intuitive hypotheses (Goodman Reference Goodman1983). At the top of the hierarchy is the overhypothesis, which comprises a family of specific hypotheses. In this case, the overhypothesis could be: “Progress through time corresponds to change in position along a linear spatial path.” This correspondence could be learned as children observe the relationship between space and time in moving objects, or it could be innate (Casasanto Reference Casasanto, Evans and Chilton2010; de Hevia et al. Reference de Hevia, Izard, Coubart, Spelke and Streri2014; Srinivasan & Carey Reference Srinivasan and Carey2010). Either way, the overhypothesized association between space and time is presumably universal across cultures, and it should be omnidirectional, since more time passes as moving objects travel farther in any direction.

Once children are exposed to cultural practices with consistent directionality, they accumulate a preponderance of evidence for one specific hypothesis. For Dutch children, reading and writing experience provides evidence for the specific hypothesis “Progress through time corresponds to rightward change in position along a linear spatial path,” strengthening this hypothesis at the expense of its competitors and causing Dutch speakers to use a rightward-directed mental timeline by default. Exposure to a different orthography in the experimental setting increased the weight of evidence for one of the participants’ overhypothesized (but culturally dispreferred) SPACE–TIME mappings, strengthening it to the point that it influenced behavior, transiently weakening their culturally preferred mapping as a consequence.

5.3 Spatial Representations of Emotional Valence: Universals and Body-Specificity

Several years ago I proposed a theory of bodily relativity (Casasanto Reference Casasanto2009a), by analogy to the theories of linguistic and cultural relativity. By hypothesis, the contents of our minds are constructed, in part, through our physical interactions with the environment. People with different kinds of bodies interact with their environment in systematically different ways. Therefore, myriad aspects of their thinking should vary relative to the particulars of their bodies. The spatial mapping of emotional valence has provided one fruitful testbed for this proposal (for reviews, see Casasanto Reference Casasanto2011, Reference Casasanto2014a).

Across languages and cultures, good things are often associated with the right side of space and bad things with the left. This association is evident in positive and negative idioms like “my right-hand man” and “two left feet.” Beyond language, people also conceptualize good and bad in terms of left–right space, but not always in the way linguistic and cultural conventions suggest. Rather, people’s implicit associations between space and valence are “body specific.” When asked to decide which of two products to buy, which of two job applicants to hire, or which of two alien creatures looks more trustworthy, right- and left-handers respond differently. Right-handers tend to prefer the product, person, or creature presented on their right side but left-handers tend to prefer the one on their left (Casasanto Reference Casasanto2009a). This pattern persists even when people make judgments orally, without using their hands to respond. Children as young as 5 years old already make evaluations according to handedness and spatial location, judging animals shown on their dominant side to be nicer and smarter than animals on their non-dominant side (Casasanto & Henetz Reference Casasanto and Henetz2012).

Beyond the laboratory, the association of ‘good’ with the dominant side can be seen in left- and right-handers’ spontaneous speech and gestures (Casasanto & Jasmin Reference Casasanto and Jasmin2010): In the final debates of the 2004 and 2008 US presidential elections, positive speech was more strongly associated with right-hand gestures and negative speech with left-hand gestures in the two right-handed candidates (George W. Bush, John Kerry), but the opposite association was found in the two left-handed candidates (John McCain, Barack Obama).

In summary, a body-specific mental metaphor links lateral space and emotional valence, and influences the way people think and communicate about positive and negative ideas. The observed SPACE–VALENCE mappings cannot be explained by influences of language or culture since, in the case of the GOOD-IS-LEFT mapping in left-handers, the implicit mental metaphor goes against the explicit GOOD-IS-RIGHT mapping enshrined in linguistic idioms and other cultural conventions (e.g. raising the right hand to swear to tell the truth).

5.4 Experiential Basis of Lateral SPACE–VALENCE Mappings

Where do body-specific SPACE–VALENCE mappings come from? All of the results reviewed so far demonstrate correlations, but Casasanto (Reference Casasanto2009a) proposed a causal relationship between the way people use their hands and the way they implicitly spatialize ‘good’ and ‘bad.’ In general, greater motor fluency leads to more positive feelings and evaluations: People like things better when they are easier to perceive and interact with (e.g. Ping et al. Reference Ping, Dhillon and Beilock2009). Bodies are lopsided. Most of us have a dominant side and a non-dominant side and therefore interact with the physical environment more fluently on one side of space than on the other. As a consequence right-handers, who interact with their environment more fluently on the right and more clumsily on the left, come to implicitly associate ‘good’ with ‘right’ and ‘bad’ with ‘left,’ whereas left-handers form the opposite association (Casasanto Reference Casasanto2009a).

To test this proposal, Evangelia Chrysikou and I studied how people think about ‘good’ and ‘bad’ after their dominant hand has been handicapped, either due to brain injury or to something much less extreme: wearing a bulky ski glove. In one experiment, right-handed university students performed a motor fluency task, arranging dominoes on a table while wearing a cumbersome glove on either their left hand (which preserved their natural right-handedness) or on their right hand (which turned them temporarily into left-handers, in the relevant regard). After about 12 minutes of lopsided motor experience, participants removed the glove and performed a test of SPACE–VALENCE associations, which they believed to be unrelated. Participants who had worn the left glove still thought RIGHT was GOOD, but participants who had worn the right glove showed the opposite LEFT-IS-GOOD bias, like natural lefties (Casasanto & Chrysikou Reference Casasanto and Chrysikou2011).

5.5 Hierarchical Construction of Spatial Metaphors for Valence

Even a few minutes of altered motor experience can change people’s implicit associations between space and emotional valence, causing a reversal of their usual judgments. HMMT provides a potential explanation of this representational flexibility. In the case of mental metaphors linking lateral space and valence, the overhypothesis may be: “The fluent region of space is good.” For right-handers, who act more fluently on the side of their dominant hand, typical motor experience provides a preponderance of evidence for the specific hypothesis that “the right side of space is good,” whereas typical motor experience for left-handers increases the strength of the evidence for the hypothesis that “the left side of space is good.” In terms of memory networks, this means that either the association between ‘right’ and ‘good’ or the association between ‘left’ and ‘good’ is strengthened at the expense of the competing associations – which are weakened but not lost and can therefore be strengthened again by new patterns of motor experience.

2.1 Experimental Evidence for Conceptual Metaphors: The Directionality Problem

Until recently CMT relied primarily on linguistic evidence for the consistency of metaphorical expressions in language. Given that the occurrence of such clusters of expressions as “warm person,” “chilly reception,” and “cold shoulder” are unlikely to be a mere coincidence, it was suggested that such metaphors do not constitute an arbitrary group of isolated expressions but are rather the linguistic manifestations of a single, implicit mapping between concepts from different domains (e.g. TEMPERATURE and INTERPERSONAL RESPONSIVENESS), that is, a “conceptual” (or, more specifically, a “primary”) metaphor (Grady Reference Grady1997a; Lakoff & Johnson Reference Lakoff and Johnson1980). In other words, the existence of a pre-/non-linguistic metaphorical mapping was deduced from the regularity of mappings in some groups of verbal metaphors, an argument that was criticized for its circularity (cf. e.g. Murphy Reference Murphy1996).

In recent years, however, various psycho-physical studies in the area of embodied cognition have provided a growing body of psycholinguistic/experimental evidence in support of the theory, as manipulating one domain was found to implicitly affect the perception of another, metaphorically related one. For example, participants who held a warm (rather than a cold) beverage tended to judge target individuals as having a “warmer” (i.e. ‘more friendly’) personality, in accordance with the hypothesized conceptual (/primary) metaphor AFFECTION IS WARMTH (Williams & Bargh Reference Williams and Bargh2008a; see also Citron & Goldberg Reference Citron and Goldberg2014). In another study, participants were found to be more likely to judge a currency to be more valuable when holding a heavy (rather than a light) clipboard, in accordance with the conceptual metaphor IMPORTANT IS HEAVY (Jostmann et al. Reference Jostmann, Lakens and Schubert2009). These and similar studies were interpreted by CMT advocates as evidence for the extension of metaphorical relations beyond the realm of language and, consequently, as support for the CMT model (e.g. Gibbs Reference Gibbs2014b: 175–177; Lakoff Reference Lakoff2014).

However, these recent psycho-physical studies have also uncovered a puzzling discrepancy between the conventional verbal instantiations of such (primary) metaphors as AFFECTION IS WARMTH or IMPORTANT IS HEAVY and their conceptual/mental counterparts: In contrast to the clear unidirectionality of verbal metaphors, behavioral research on the kinds of cross-domain associations underlying primary metaphors in CMT consistently reveals bidirectional effects (IJzerman & Koole Reference IJzerman and Koole2011; IJzerman & Semin Reference IJzerman and Semin2009; Lee & Schwarz Reference Lee and Schwarz2012). While evidence accumulated that manipulating a concrete domain affects individual’s judgments in the correlated abstract domain – in accordance with the predictions of CMT – the reverse pattern was also found. For example, thinking about pleasant or unpleasant social situations likewise tends to alter judgments of room temperature as being warmer or colder, respectively (Zhong & Leonardelli Reference Zhong and Leonardelli2008). Following the logic of the original study, this finding corresponds to the non-hypothesized conceptual metaphor *WARMTH IS AFFECTION, which defies the regular concrete-to-abstract pattern and has no conventionalized verbal equivalent.

The WARMTH-AFFECTION case is not an isolated example, as this bidirectional effect extends to many other associations between abstract (non-perceptual) and sensorimotor (perceptual) experiences (of the kind hypothesized to underlie primary metaphors), including moral judgment and physical cleanliness, verticality and positive or negative mood, importance and weight, and many others (Schneider et al. Reference Schneider, Rutjens, Jostmann and Lakens2011; Zhong & Liljenquist Reference Zhong and Liljenquist2006; for an overview, see Landau et al. Reference Landau, Meier and Keefer2010). In contrast, the corresponding verbal metaphors exhibit a robust unidirectionality, e.g. moral judgment is verbally described in terms of physical cleanliness but not vice versa, mood is described in terms of verticality but not the other way around, importance in terms of weight, and so forth (see also Grady & Ascoli, this volume).

We hold that the discrepancy between the unidirectionality of verbal metaphors and the apparent bidirectionality found in psycho-physical studies poses a serious problem for CMT, especially concerning primary metaphors. As CMT maintains that abstract concepts/domains are “embodied,” i.e. conceptualized via bodily experiences, it predicts that manipulating the experienced source concept should impact the representation of the target concept, but not vice versa. To return to our example, if affection is represented in terms of temperature, then changing experiences of physical warmth should affect judgments about affection/empathy. Since physical warmth, in turn, is not conceived in terms of affection, however, perceiving it should not be affected by our social feelings.

In sum, despite CMT’s long-standing claims that verbal metaphors are mere reflections of underlying conceptual mappings, the reported psycho-physical findings (taken as reliable indications of the properties of their mental/conceptual counterparts) suggest that there is a clear difference between the two, with the former being unidirectional and the latter bidirectional.

2.2 Previous Attempts to Cope with the Directionality Problem

Although largely overlooked, a small number of researchers have noticed this discrepancy and the challenge it poses to CMT. While Landau, Meier, and Keefer (Reference Landau, Meier and Keefer2010: 364) “hold out the hope that future research can resolve this issue while preserving the benefits of a metaphor-enriched perspective,” IJzerman and Koole (Reference IJzerman and Koole2011) as well as Schneider and colleagues (Reference Schneider, Rutjens, Jostmann and Lakens2011) take these findings as refuting the CMT notion of conceptual metaphors with a clear target–source distinction altogether; they do not attempt, however, to account for the directionality of verbal metaphors.

In contrast, Lee and Schwarz (Reference Lee and Schwarz2012) claim that the discrepancy can be fully accommodated within Lakoff’s paradigm, relying on (a) the distinctions between the linguistic and the psychological consequences of conceptual metaphors; (b) the distinction between representational structure and online processing; and (c) the distinction between the two mechanisms involved in the formation of metaphors according to Lakoff and Johnson (Reference Lakoff and Johnson1999: 55–57), “co-activation” and “projection.”

The first two of these distinctions explore the discrepancy rather than fully explain it. No attempt is made, for example, to address questions such as why there should be a difference between the linguistic and psychological consequences of conceptual metaphors, and, in particular, why these two should differ in directionality. The third distinction presents a more convincing argument. According to Lakoff and Johnson (Reference Lakoff and Johnson1999: 48–49), early life experience involves repeated conflation between different domains – for example, between the feeling of affection while being held or hugged as an infant and the physical warmth of the caretaker’s body. These experiential correlations may lead to neural co-activation of the two domains, a bidirectional process by which experience in one domain activates the other and vice versa. In turn, these correlation-based associations form the basis for metaphors such as AFFECTION IS WARMTH, in which properties are projected from one (concrete) domain onto another (abstract) one. According to Lee and Schwarz’s proposal, it is possible that both mechanisms remain active so that aside from the unidirectionality of verbal metaphors that derives from the underlying process of projection, there may also be a bidirectional metaphorical effect that results from an underlying (bidirectional) process of co-activation.

However, Lee and Schwarz’s proposal does not constitute a full account of the issue at stake, since the very distinction between the two processes (which we fully agree with) does not provide an answer to such questions as: What is the relation between the two processes? How does one shift from one process to the other? And, most importantly, why do verbal metaphors reflect the (unidirectional) process of projection rather than the (bidirectional) process of co-activation? Lakoff (Reference Lakoff2014) himself, apart from pointing to a possible solution to Lee and Schwarz’s findings in neural terms, concludes that their experiment points to the need to better understand the difference between metaphorical mapping and experimental effect.

To conclude, past references to the conflict between the CMT predictions and the empirical findings on directionality have led to unresolved attempts to reconcile the two, dismissal of the reliability of psycho-physical findings as “psychological consequences” of the original conceptual metaphors, or even to the rejection of CMT altogether. In contrast, we propose an alternative account that not only provides a simpler solution for the discrepancy but also allows us (i) to accept the experimental findings reported above as reliable representations of metaphorical relationships at the conceptual level, and (ii) to largely preserve CMT’s central claim that verbal metaphors are reflections of a deeper pre-linguistic/conceptual relationship. The solution we propose relies on the commonly overlooked role of a central factor that distinguishes the conceptual level from the verbal one, namely, language itself.

3.1 Two Phases of Metaphor Processing

We agree with the CMT model that verbal metaphors originate from a conceptual pre-linguistic level so that the relation between ‘affection’ and ‘warmth,’ for example, as represented by expressions such as “She is a warm person,” does indeed reflect an underlying implicit association between the two domains. We disagree with CMT’s claim that this implicit pre-linguistic association represents the same type of metaphorical relation as their corresponding metaphorical expressions. Instead, we claim that it represents a simpler, more symmetric type of linkage in which the two domains are simply “associated,” and which involves no attribution of “target” or “source.” We suggest that for this symmetrical “bare association” between the two domains to become a full-blown metaphor in which aspects of the source are unidirectionally “projected” onto the target, another process is required, in which language plays an essential role. In other words, we claim that the unidirectionality of metaphorical expressions in language is not a direct reflection of a corresponding unidirectionality of pre-linguistic metaphorical associations. Instead, we suggest that language is (at least partially) the generator of this unidirectionality, by turning a latent (conceptual) asymmetry between the two concepts/domains into an active one. As said, our suggestion relies on the separation of metaphor processing into two different phases to be elaborated on next: “bare concept/domain association” and “directional projection.”

Bare concept/domain association is the initial phase of metaphorical association, which is derived from experiential correlation between two domains and corresponds to Lakoff and Johnson’s suggested mechanism of co-activation. We use the word “bare” to emphasize that at this phase the two domains (e.g. ‘affection’ and ‘warmth’) are not assigned target and source functions but are merely associated with each other. Crucially, this bare association between the two domains is, in principle, bidirectional, in that each domain is associated with the other and can therefore prompt or interfere with it, as the aforementioned experiments show. We agree with CMT that this basic association between the two domains then motivates a more developed type of relationship which is expressed in full-blown metaphors such as AFFECTION IS WARMTH. However, we do not claim that the shift from this initial phase to the more developed one is triggered automatically, probably in early childhood (see Lakoff & Johnson Reference Lakoff and Johnson1999: 45–59). Rather, we claim that this bare association remains the prominent type of connection between metaphorically related domains – unless directionality is externally triggered, which occurs when the relationship between the two domains is expressed in language. Put simply, we propose that instead of assuming full-blown metaphors at the conceptual level, it would be more accurate to describe the pre-linguistic level as consisting of “conceptual (bidirectional) associations.”

Directional projection refers to the process of mapping one domain onto the other. This type of relation is built on the simpler, bare association between the two domains and includes the additional components of (i) assigning the TARGET and SOURCE functions to the two domains and (ii) projecting some properties from the source to the target domain, which results in the conceptualization of the target in terms of the source. This relation is a unidirectional one, in that properties of the source are used to conceptualize the target, rather than the other way around.

Bearing in mind this separation between the two “types” of metaphor processing, we propose the following:

1. 1. The bidirectional bare association between concepts from two domains is the initial phase of metaphor formation and comprehension, preceding unidirectional projection.

2. 2. Language plays an essential role in changing this bare association into a directional relation of target and source.

Taken together, these two claims provide a fairly straightforward solution to the discrepancy discussed above: The bidirectionality found in the perceptual and sensorimotor experiments reflects the initial bidirectional relation between two domains. In contrast, the robust unidirectionality of verbal metaphors reflects the output of another, “higher” process, namely, the assignment of a target–source relation to the associated domains, whereby language plays an essential role.

This proposal has direct bearing on the debate about (the depth of) the level at which conceptual metaphors exist. The present proposal suggests a compromise between those who maintain the position that there are primary metaphors at a preverbal (conceptual or even embodied) deep level (Lakoff & Johnson Reference Lakoff and Johnson1980, Reference Lakoff and Johnson1999; this volume: Grady & Ascoli; Winter & Matlock; see also Casasanto’s “mental” metaphors) and those who challenge this position by arguing that there is no evidence for this (e.g. Keysar et al. Reference Keysar, Shen, Glucksberg and Horton2000; for conceptual metaphors generally, see also Murphy Reference Murphy1996).

3.2 Evidence for the Distinction between the Two Phases of Metaphor Processing

Since the empirical investigation of the conceptual level poses major methodological difficulties, the existence of the hypothesized bidirectional bare association between the two domains can only be deduced from converging pieces of evidence, e.g. through analytic reasoning (as done by CMT) and from the findings of relevant psycho-physical studies. However, there is another piece of evidence for the existence of two distinct mechanisms in the process of metaphorical comprehension, which seem much like the types we are proposing, that is, an initial symmetrical relation followed by a more developed, directional one.

For example, Wolff and Gentner (Reference Wolff and Gentner2011: 1456–1457) managed to reconcile two of the most dominant (and competing) views regarding the nature of metaphor in the last years, to which they referred as the “emerging commonalities” and the “directional projection” views: While the former claims that metaphor relies on the detection of commonalities in different domains, the latter states that metaphor is a process of projection, in which properties from one domain (the source) are conveyed to the other (the target).¹ Crucially, these two approaches differ in their view of directionality, since the process of projection is unidirectional by definition, but detecting commonalities is at least theoretically a bidirectional process.

Wolff and Gentner (Reference Wolff and Gentner2011) used a mix-deadline procedure (exposing subjects to the stimuli for either a short or a long time) to show that these two processes should not be taken as separate mechanisms, but rather as two stages of metaphor processing. Participants were asked to judge the comprehensibility of verbal (hence directional) A-IS-B metaphors, such as “some arguments are wars,” and their reversed versions “some wars are arguments,” in either short or long exposing time. In accordance with their model, forward and reversed orders did not differ in comprehensibility at the early stage of processing (600 ms), while at later stages forward metaphors were clearly more comprehensible than reversed metaphors (as demonstrated in numerous studies). These results indicate that, despite the common view of metaphor as inherently unidirectional, directionality is in fact attributed at a later stage of processing. As the researchers themselves point out (ibid.: 1480), their findings pose a problem to Lakoff’s view, which predicts directional processing from early stages (both explicit and implicit). However, these findings are consistent with our modified version of CMT, which also claims that there is a shift from symmetrical to unidirectional processing. According to our model, therefore, the first 600 ms in Wolff and Gentner’s study may be regarded as a “window” into a deeper conceptual level.

A few developmental studies provide further support for a “shift” from a symmetrical to an asymmetrical processing. For example, several studies indicated that children are insensitive to asymmetry in metaphors.² Most noticeably, one study showed that although 4-year-olds respond better to grammatically asymmetrical similes – e.g. are more likely to identify the metaphorical “ground” (i.e. the relevant metaphorical property) of “a boat is like a leaf” than of “a boat and a leaf are alike” – they are insensitive to the order of the two concepts: e.g. “a leaf is like a boat” and “a boat is like a leaf” were equally easy to extract a ground from.³ These findings, too, support the notion that metaphorical directionality is a later development. Since young children do notice the asymmetry of grammatical structure, at least one possible explanation for these findings is that children at a young age simply do not use metaphorical projection but, rather, rely on the simpler association between the two concepts.

An even more long-term “developmental” shift was demonstrated in a study that compared cross-domain interference using human and non-human subjects (Merritt et al. Reference Merritt, Casasanto and Brannon2010), which found that while humans demonstrate asymmetrical space–time interference (with spatial information influencing temporal judgments much more than temporal information influenced spatial judgments), rhesus monkeys demonstrate symmetrical interference (with spatial information strongly influencing temporal judgments, and vice versa). As the researchers point out, one obvious (but not necessary) explanation for the difference between the two species may be the availability of language, which could have long-term effects on the creation of asymmetrical mappings (ibid.: 199).

Combined with the discrepancy between the characteristics of verbal metaphors and the findings of psycho-physical studies, these examples provide some empirical support for the first claim of our model, i.e. that under the directional type of metaphor processing there is a more basic and simpler, non-directional linkage between metaphorically related domains. We now introduce some evidence supporting our second and major claim, namely, that language plays an essential role in enhancing, or even establishing, metaphorical unidirectionality.

4.1 The Importance of Linguistic Form

First, linguistic form is by no means “transparent.” Consider, for example, the way people interpret “reversed” metaphors, in which the more natural candidate for the source concept appears in the target position, as in the reversed (unconventional) A-IS(-LIKE)-B metaphor “An anchor is (like) a friend.” The interpretation of such reversed metaphors may cause processing difficulties (e.g. Glucksberg & Keysar Reference Glucksberg and Keysar1990; Ortony et al. Reference Ortony, Vondruska, Foss and Jones1985) or even produce new meanings (Shen Reference Shen2008), as people rely on the grammatical form in determining the metaphorical target (and source) terms, even when this form is incongruent with the conceptual bias to map from more to less accessible concepts (e.g. interpreting the reversed metaphor “an anchor is like a friend” as ‘an anchor is a sailor’s best friend’). Arguably, these difficulties are caused by a clash between grammatical and conceptual asymmetries, that is, between the tendency to assign the target and source on the basis of the grammatical and the conceptual bias to map from more to less accessible concepts.

Second, in some cases linguistic form can even determine the target–source assignment, as in the case of conceptually symmetrical (metaphorically related) pairs, in which none of the concepts can be regarded as the natural candidate to the roles of target and source. A case in point is “This butcher is a surgeon,” which yields a totally different interpretation (highlighting exaggerated precision and care) than its reversed version, “This surgeon is a butcher.” In this case it is the grammatical form that determines the direction of mapping, given that both directions yield equally plausible interpretations (cf. Glucksberg & Keysar Reference Glucksberg and Keysar1990).

Third and most importantly, grammatical asymmetry may not only trigger or enhance the corresponding conceptual asymmetry but rather necessitate it: Even if, as our model predicts, the fundamental (pre-linguistic) association between metaphorically related concepts is non-directional, then still the very placement of the two concepts in a grammatically asymmetrical form would impose some division of labor between them. Consider, for example, the butcher–surgeon case: The concepts of ‘surgeon’ and ‘butcher’ may be related in several respects (e.g. both cut flesh), and hence, in the absence of grammar (or some other directional cue, see also Forceville, this volume), both directions of mapping are possible, from ‘butcher’ to ‘surgeon’ or vice versa. Once grammatical asymmetry is employed, however – for example, when one is asked to verbalize the relation between two domains in a typical (i.e. grammatically asymmetric) metaphorical form, certain choices regarding the assignment of the target and source functions to the concepts have to be made, determining the direction of the mapping.

In that respect, we assume that expressing metaphors in a grammatically asymmetrical structure may illustrate a special type of “thinking for speaking” (Slobin Reference Slobin, Gumperz and Levinson1996), where the very use of certain linguistic forms requires the speaker to apply a certain mode of thinking, in this case, assign target–source functions to an otherwise bidirectional association; the hearer, of course, is also forced to understand the metaphorically related concepts in the way determined by this linguistic form.

It is important to note that though all of these effects are easily demonstrated by using nominal metaphor or simile (A-IS-B/A-IS-LIKE-B) – as done throughout this section – we intend our arguments to equally apply to other types of grammatically asymmetrical constructions usually discussed in CMT, such as noun–adjective constructions of the type “a smooth person,” “a big event,” “a warm welcome,” or “a heated debate” (which reflect the basic, correlation-based associations underlying primary metaphors), or the above-mentioned instantiation of a complex metaphor like LOVE IS A JOURNEY. The important point is that linguistic forms which are used to express metaphorical association are almost always grammatically asymmetrical and therefore impose some division of labor (between target and source functions) on the associated concepts.

4.2 Language as a Prerequisite for Metaphorical Unidirectionality

We have so far stressed that language may trigger, determine, and necessitate unidirectional mappings. But is it also a prerequisite for such mappings? Unlike the ‘butcher’–‘surgeon’ example, the vast majority of metaphors do maintain some type of conceptual asymmetry between the two concepts/domains which determines a preference for a particular target–source assignment. As discussed above, such a division of labor between target and source is a sufficient condition, according to the CMT view, for the creation of a “full-blown” directional metaphor, especially one of the primary type.

But what if the arrangement of the two concepts within a grammatically asymmetrical structure did not merely reflect a pre-linguistic division of labor but rather led to this division of labor in the first place? In other words, could it be that the mere existence of conceptual asymmetry between ‘fear’ and ‘cold,’ for example, is not sufficient to trigger a directional process of mapping – and that, therefore, an external intervention in the form of grammatical asymmetry (or some other directional cue) is needed?

To test this prediction, we investigated the effect of grammar on directional judgments of novel metaphorical associations (of the A-IS-LIKE-B type) with a clear abstract–concrete asymmetry, such as “childhood memories” (coding for an abstract concept) and “migrating birds” (coding for a concrete concept).⁴

The experiment was conducted in two phases. In the first phase 10 independent judges (mean age 40) were presented with two directional comparisons for each pair of the aforementioned terms (e.g. a pro-hierarchical order in which the source term is more concrete than the target one, as in “childhood memories are like migrating birds” and the corresponding anti-hierarchical order, as in “migrating birds are like childhood memories”). They were asked to judge which of these comparisons seems more natural, and to rate (on a scale of 1–7) the degree of confidence in their choice. Unsurprisingly, the judges demonstrated a very strong preference toward pro-hierarchical ordering (91% of all choices, with average confidence level of 5.3), as would be predicted by CMT. This finding suggests that even without understanding the relation between two given concepts (the judges were not required to interpret those novel similes), the mere conceptual asymmetry between abstract and concrete by itself is sufficient to dictate a division of labor if (and, as soon to be explained, this is a big “if”) grammatical asymmetry is involved.

In the second phase, 43 participants (mean age 33) were presented with the same set of metaphorical pairs. Half of pairs (the “asymmetrical condition”) were embedded in a pro-hierarchical asymmetrical comparison (“A is like B,” as in “childhood memories are like migrating birds”); the other half (the “symmetrical condition”) were embedded in the corresponding symmetrical form (“A and B are alike,” as in “childhood memories and migrating birds are alike”).

Each experimental item was followed by two alternative contexts in which the comparison could have been uttered: the “pro-hierarchical context” (e.g. “a nostalgic writer speaking about his youth,” which determines the abstract concept ‘childhood memories’ as the topic of the comparison), and the “anti-hierarchical context” (e.g. “an enthusiastic ornithologist describing the flight of birds,” which determines the concrete concept ‘migrating birds’ as the topic of the comparison). Participants were asked to decide which was the more likely context for the comparison to be uttered, and to rate (on a 1–7 scale) their level of confidence in their choice. If indeed the preferred ordering of the two concepts reflects a directional mapping existing prior to its representation in language, then the pro-hierarchical context should be preferred regardless of the grammatical form (symmetrical or asymmetrical) representing it in language.

In contrast to this prediction, however, a notable difference between the two structures was found. Comparisons in the asymmetrical condition demonstrated a strong preference for the pro-hierarchical topic-assignment (95.1% of all cases, average confidence level of 5.3); in contrast, the comparisons in the symmetrical condition demonstrated a much weaker tendency for the pro-hierarchical assignment (69%, confidence level of 4.9). The difference between the two conditions was significant for both choice (Z = -4.739 p < 0.001) and confidence level (t = -4.417, p < 0.001).

The results of this experiment suggest that the conceptual asymmetry alone was not sufficient to establish a clear division of labor between the two concepts. Rather, embedding the terms of these comparisons in a grammatically asymmetrical structure is required in order for this conceptual asymmetry (that is, the assignment of the target and source functions to the abstract and concrete terms, respectively) to be fully realized.

One might argue, however, that the novelty of associations in the previous experiment and/or the metaphor type chosen might have tilted the results, as they did not represent the metaphorical relationships discussed as “conceptual” or “primary” in CMT. In order to test this, we replicated the experiment (10 judges, mean age 40; 31 participants, mean age 35) using expressions reflecting conventional metaphorical associations that CMT would regard as instantiations of conceptual metaphors, both complex and primary ones. Example item pairs are given by “a political argument” and “a boxing match” (reflecting ARGUMENT IS WAR, Lakoff & Johnson Reference Lakoff and Johnson1980: 5), or by “fear” and “cold” (reflecting FEAR IS COLD, Kövecses Reference 352Kövecses2005: 289).

This replication yielded a similar pattern of results as the previous experiment. As predicted, a strong preference for pro-hierarchical ordering (92% of all cases, average level of confidence 5.7) and pro-hierarchical topic-assignment was found in the asymmetrical condition (94%, level of confidence 5.5). In contrast, a much weaker tendency toward pro-hierarchical topic assignment was found in the symmetrical condition (72%, level of confidence 4.9). Again, the difference between the different (i.e. symmetrical vs. asymmetrical) verbalizations was significant for both choice (Z = −3.785, p < 0.001) and confidence level (t = −3.604, p < 0.001).

From these findings it follows that even preexisting conceptual hierarchies such as abstract–concrete require a directional task for their actualization and are not in themselves sufficient to prompt a strongly unidirectional mapping. If the very existence of an abstract–concrete asymmetry was sufficient to trigger a directional process, as implied by CMT, we would expect to find the same topic assignment (preferring the abstract concept as the topic of the metaphor) regardless of grammatical structure, i.e. in both the symmetric and the asymmetric condition. The fact that the tendency toward a pro-hierarchical context was much weaker in the symmetrical condition may suggest that the actual division of labor between a given pair of concepts does not derive automatically from their conceptual asymmetry but, rather, are only actualized fully by their insertion into the grammatically asymmetrical form.

To conclude, presenting metaphorically associated words in a grammatically asymmetrical form strongly enhances (and in some cases even triggers) metaphorical directionality in two complementary ways. First, it may actualize a potential asymmetry between the two domains that otherwise might have been remained latent: by inserting relevant verbal expressions into the grammatical slots defining targets and sources, a clear division of labor is imposed. Second, the very placement of the two concepts in these slots may in turn serve as a cue for their metaphorical assignment regardless of their semantic content.

2.1 The Body Schema

In Gallagher’s terms, the concept of body schema refers to the sum of the sensorimotor capabilities that allow us to move and constrain our movements, to navigate and perceive the world. In this sense, the body schema is not the result of our perception of our own body and does not refer to a perceptual or mental image of the body. With this definition, Gallagher refers to the motor abilities that are foundational to action and perception.

As such, the body schema does not usually involve awareness or constant perceptual monitoring. Unless something does not work in the right way or unless we are facing the task of learning a new motor pattern, for example a dance step, we do not generally focus our attentional awareness on the body and on the processes underlying action and perception. Let us think of what we normally do when we walk. We know that we need to move one foot after the other in a certain direction. However, while walking, we do not need to attend to our body and its fine movements, that is, we do not need to focus our attention on our legs and on our feet to calculate the exact distance of space we need to move or to estimate how to balance the weight of our body on one foot when the other is moving forward, and so on. So, even though we are not aware of what we do in order to walk, we are usually able to walk and to effortlessly adjust our movements to the ground. We do not continuously lose our balance if the surface where we are walking is not completely regular, and we do not fall down all the time. To make our movements faster and safer, there is a set of subpersonal mechanisms, proprioception in primis, that allow us to register the inputs we get from the outside and on this basis to regulate what we need to do in order to move in the environment. Of course, we can suddenly have a physical problem, such as a severe pain in a leg that will not allow us to move any more. In such a case we may lose our balance or even fall down, and then need to give conscious attention to our body and its movements to be able to walk again. But this is not the norm. Pain, together with other peculiar circumstances related, for example, to fatigue, sex, sickness, or even to philosophical introspection or medical investigation, are limiting cases in which the body itself becomes the focus of our attention. When the body enters our perceptual field, becoming the object of our conscious perception, we are already dealing with the body image. In sum, the body schema is not an image of the body but the condition of the possibility of action and, thus, also of perception (Noë Reference Noë2004).

According to Gallagher (Reference Gallagher2005: 45), three functional subsystems constitute the body schema. The first is the system responsible for the processing of information about posture and movement. Information about posture and movement comes from different sources. Somatic proprioception is certainly the primary kind of information we use. We get proprioceptive information from kinetic, muscular, articular, and cutaneous sources. Vestibular and equilibrial functions and the visual sense are also sources of proprioceptive information. In the latter case, we automatically register visual information about the movement of our own body in the environment. The visual sense, however, can also be a source of explicit information in cases when we directly focus our perceptual attention on the position of our limbs. In these cases we are again already dealing with a form of body image. Note that proprioceptive information is defined here as a set of physiological, subpersonal, and unconscious processes that lead to proprioceptive awareness. In some pre-reflexive sense, we are always aware of the position of our body, but this awareness is not the result of an explicit act of reflection.

The second functional system of a body schema is responsible for the production of motor programs and movement patterns (output). We have a number of motor habits. Developmental studies suggest that some of these are innate, such as swallowing, while others are learned, such as writing (Meltzoff & Moore Reference Meltzoff and Moore1977, Reference Meltzoff and Moore1989; see also Gallagher Reference Gallagher2005). At the neural level, a motor program can be described as the activation of some neural circuits in the motor cortex that enables the execution – at the behavioral level – of the single motor acts that together constitute the chain of a motor action. The processing of proprioceptive information is essential for the production of motor patterns. We constantly need feedback about our current body posture to efficiently move in the environment.

The third functional system of a body schema regards cross-modal communication. We have an innate ability to transform visual inputs into motor competence. This means that when we observe other people’ actions, this visual stimulus is immediately and automatically translated into motor terms. This is achieved by means of mirror neurons, multimodal neurons in the pre-motor cortex which respond both to action observation and to action execution (di Pellegrino et al. Reference di Pellegrino, Fadiga, Fogassi, Gallese and Rizzolatti1992). The motor neurons that enable the movement of the hand when we grasp a cup, for example, will be activated not only when we effectively grasp a cup but also when we observe someone else grasping a cup. The somatotopic activation of motor neurons in the pre-motor cortex in the absence of a corresponding action has been considered as a “motor simulation” of the observed action. It has been suggested that motor simulation is a constitutive process of action understanding; we understand other people’s actions by means of the activation of our own motor competence (Gallese & Sinigaglia Reference Gallese and Sinigaglia2011). Furthermore, it has also been hypothesized that motor simulation is involved in the process of learning by imitation (Rizzolatti et al. Reference Rizzolatti, Sinigaglia and Anderson2008). In this connection, the mirror mechanism determines automatic activation of the motor programs that are necessary to carry out the observed action.

It has lately been discovered that the mechanism of simulation is not limited to motor neurons in the pre-motor cortex. This mechanism is widespread in the brain, and it also characterizes perception- and emotion-related areas. When we see other people’s facial expressions indicating their emotions, for instance, emotion-related areas of our brain will be activated. Furthermore it has also been observed that the mechanism of simulation is activated not only by direct observation of others but also by cognitive tasks such as mental imagery and language comprehension (Gallese & Sinigaglia Reference Gallese and Sinigaglia2011; for a discussion of the latter, see Ritchie, this volume). Given that the process of simulation is neither activated by action observation only nor confined to the motor areas of the brain, it is usually referred to by the more general term of “embodied simulation.”

2.2 The Body Image

As suggested by this definition, different types of intentional relations can be involved in the constitution of a body image. First, the body can be the object of a perceptual state (“body percept”). When we move around perceiving the world, for example when we look at a beautiful landscape from the peak of a mountain, we are usually not aware of our body. It becomes invisible to us while we are absorbed in the breath-taking beauty of the landscape. But if we suddenly have a terrible itch on our left foot, this will become the object of our perceptual attention and of our efforts to relieve the complaint. This intentional relation, clearly, also implies conscious awareness. We are perfectly aware of what our foot feels like when it itches. This kind of epistemic access to our body is from the first-person perspective (Pauen Reference Pauen2012).

Second, the body can be the object of our conceptual knowledge (“body concept”). This happens when it is accessed from the third-person perspective (ibid.). When we open a handbook of anatomy and we start to study the structure and composition of the human body, it becomes the object of our conceptual understanding. In this case, we cannot know how it feels to have neurons firing, but we have a scientific description of this neuro-physiologic process. This is clearly an intentional relation, also involving conscious awareness. Note that our scientific knowledge about the body does not always imply a conscious level of awareness: We do not always have the functions and processes of the nervous system or of other parts of the body consciously present in our mind. According to Gallagher (Reference Gallagher2005: 25), this knowledge is part of a set of beliefs and attitudes that – even unconsciously – entertains an intentional relation with the body considered as an epistemic object. Scientific knowledge is not the only source of conceptual knowledge about the body; commonsense beliefs are an equally important part of our conceptual understanding of the body. Common sense and scientific knowledge constantly interact with one another. Our beliefs are inevitably historically, linguistically, and culturally situated and the progresses and results of scientific knowledge are an integral part of our historical, linguistic, and cultural environment. What scientific research reveals about the body does, in a slow but constant process, influence our commonsense understanding of the body.

Third, the body can also be the object of an emotional attitude (“body affect”). This is the case when our feelings, positive or negative, have the body as their object. We look in the mirror every day, and this routine action is usually associated with a positive or negative feeling, according to our mood on that day and to many other variables. We do or do not like what we see. This feeling toward the body is an intentional relation that can be conscious but is very often unconscious. It is of great importance that our feelings and attitudes toward the body are inevitably mediated by our body concept, that is, by our beliefs about the body. If we live in a society where the standard of beauty is a thin and athletic body and we do not match that standard, this can lead us to have negative feelings toward our body. It is also possible and fairly common that our feelings toward the body influence and affect our perception of our own body and our beliefs and conceptualization of it.

Body affect, body concept, and body percept are three different but deeply interconnected aspects of our image of the body. The body image, then, is a set of interconnected intentional states, whereby each type of intentional relation with the body affects and influences the others and all of them are deeply situated in a historical, linguistic, and cultural environment. The image we have of our own body is inevitably the result of our being situated in a specific context including a particular physical, linguistic, and sociocultural environment.

Finally, it is important to note that the body image is almost always a partial representation of the body: Our perception, emotional attitude, or conceptualization of the body are usually directed toward a part or a single aspect of the body at a time, as in the example of the itch on the left foot discussed at the beginning of this section. We single out one aspect of our bodily experience and this aspect or part becomes the object of a representational state.

2.3 Relations between Body Schema and Body Image

I indicated in the previous sections that, although conceptualized as distinct, the boundary between body schema and body image is not rigid in human cognition, so that the body image can affect the body schema. This is the case when our beliefs and concepts about our own body affect perception and movement in space. Even a basic activity, acquired very early on in life, such as walking, can be affected by our beliefs and thus, in the range of the bio-mechanical possibilities, determined by the structure of our body, it can be to some extent culturally shaped.

Vice versa, it is also possible that in particular circumstances, for example, when we are facing a situation of particular physical effort (e.g. during some sports training), we focus our attention on specific motor patterns that are usually below the threshold of awareness. In such a case, body schema functions become the object of attention and perception.

Finally and most relevant to the present concerns, the body schema and the body image interact in embodied simulation. It has been proposed that embodied simulation is involved in the sensorimotor foundations of the conceptual system because metaphorical thought seems to directly recruit the sensorimotor system (Gallese Reference Gallese and Pineda2009; Gallese & Cuccio Reference Gallese, Cuccio, Metzinger and Windt2016; Gallese & Lakoff Reference Gallese and Lakoff2005). The exploitation of the sensorimotor system, which is part of the body schema, in the structuring of the conceptual system by means of the mechanism of simulation highlights the tight interaction between body schema and body image (see section 4).

3.1 Invisible Metonymies: The Body Schema as a Source Domain

As previously defined, the notion of the body schema refers to the set of (partially innate) sensorimotor functions that enable our active interaction with the environment. In this section I will discuss in depth the hypothesis that the body schema is the basic and ontogenetically primary source of a metonymic mapping from sensorimotor abilities to perception (and, hence, ultimately to cognition, too).

Our perceptual experiences are always functions of our movements. This is the case even for vision, which is often taken to be the most passive of our perceptual modalities and has also served as the paradigm case for thinking about perception in general (cf. Noë Reference Noë2004: 2). There is empirical evidence showing that a paralysis of the eyes causes blindness. Vision requires saccades and microsaccades (eye movements), many times per second, to make us able to see (ibid.: 13). These constant movements of the eyes are constitutive of vision. Furthermore, it has also been shown that perception does not only require bodily movements relative to the surrounding environment. It also requires self-actuated movements, i.e. movements that individuals actively perform, not passive movements (ibid.; Held & Hein Reference Held and Hein1963). From this point of view, perception is grounded in and constrained by our motor abilities (see also Gibbs Reference Gibbs2006a: 42–78). In other words, perception is something that we literally act out (and thus cannot adequately describe as a passive reception of stimuli).

Calling this level of embodied metonymic cognition “transparent” emphasizes that, when we enact perception, our actions and the body that enables them are usually outside the focus of our attention and not visible to us. To have perceptual experiences, we need to map sensorimotor contingencies, here defined as the “interdependence between stimulation and movement” (Gibbs Reference Gibbs2006a: 65), onto the stimuli we receive. Crucially, the stimuli we get from the environment are always partial and incomplete, but our perceptual experiences are not. For example, we usually see objects only from one side but we perceive them as three-dimensional. To do so, we project sensorimotor contingencies onto the partial stimuli we receive. A partial and incomplete stimulus is, thus, always seen in terms of the actions and movements that would allow us to have a more complete perceptual access to it.

The first and primary level of embodiment is at play in our perception of the world, being its condition of possibility, and it is, thus, foundational to any other level of cognition. This is particularly evident when we look at cognitive development. Our body schema is the primary source of knowledge because movement in space is our primary source of knowledge. On the basis of partial stimuli, we map from movement to complete perceptual experiences. This level of embodiment does not reveal itself in the linguistic dimension.

Although taking up vital aspects of Conceptual Metaphor Theory (henceforth CMT) and Primary Metaphor Theory (henceforth PMT), namely, their claim that metaphors are cross-domain mappings from sensory source domains onto more abstract domains (including social and affective ones), the present proposal also goes beyond CMT and PMT insofar as these do not account for the modality of constitution of sensory experiences. Sensory experiences (see Grady Reference Grady1997a; Grady & Johnson Reference Grady and Johnson2002) seem to be a starting point that can be taken for granted. Whether the sensory experiences assumed to provide the source domains of primary metaphors belong to the body schema or to the body image is thus left open.

It is claimed here that we also need to account for the constitution of sensory experiences. A closer scrutiny of the constitution of sensory experiences will allow us to realize that this basic and primary level of cognition implies a metonymic description: The mapping goes from the body schema to perception. The body schema provides the structure to every perceptual experience: Although we are not aware of this, we map patterns of sensorimotor contingencies onto partial perceptual stimuli, and this allows us to have full perceptual experiences. This mapping is neither conscious nor directly visible to us, but it allows perceptual experiences to take place. The level of invisible metonymies is thus of a constitutive nature.

Clearly, the definition of this first level of embodied cognition is profoundly different from the level of primary metaphors introduced by Grady (Reference Grady1997a, Reference Grady2005a). In Grady’s account, primary metaphors are constituted by the association of sensory experiences, referred to as their “image content,” with non-sensory “response” concepts, which include social and emotional assessments of the sensory experience. Both the source and the target of a primary metaphor are equally basic aspects of the holistic experience given by a so-called “primary scene”:

Though PMT does not distinguish between body schema and body image, Grady’s remarks can be taken to place the association between the source and target of a primary metaphor already at the conceptual level. Hence, it can be concluded that, in PMT, the body contributes to cognition being already an object of knowledge, i.e. as body percept.

At the first level of embodiment, in contrast, the body is not an object of knowledge. It is, instead, the condition of possibility of knowledge because our motor abilities, together with the sensory apparatuses we are equipped with, are the condition of possibility of every perceptual experience. The relationship between the physical body and motor abilities, on the one hand, and perceptual experiences, on the other, is constitutive and considered to be metonymic. To have perceptual experiences, we map sensorimotor contingencies onto the partial stimuli we get from the environment. Only this allows us to build complete perceptual experiences.

However, the boundary between this level of embodied metonymic cognition and the level of visible metaphors is not so rigid. The lived body and its motor abilities can enter our perceptual field in different situations and can be exploited, by means of the mechanism of embodied simulation, to carry out cognitive tasks that also involve the use of representations of the body. I will discuss this in more detail in section 4.

3.2 Visible Metaphors: Body Image as a Source Domain

The power of metaphors having a body image as a source domain can be considered audible and visible, in comparison with the silent and transparent power of embodied metonymies based on the body schema. To be more precise, in metaphors from the body image the source of the metaphor is our body as an object of representation and knowledge (i.e. a body percept, body concept, or body affect), and the metaphorical mapping takes place independently of the presence of these metaphors at the level of language, although language may reflect them. As such, the metaphorical mapping is now from one representation to another representation, and the contribution of the body is mediated by our cultural, environmentally situated, and linguistically structured representations of the body itself.

According to my proposal, Grady’s (Reference Grady1997a) primary and complex metaphors both take their source domains from the body image. As noted before, a primary metaphor is based on a correlation between a concept in the conceptual domain (matrix) of sensory experiences and another one in an abstract conceptual domain. The metaphorical mapping takes place at the conceptual level, and hence, in this account, the body contributes to metaphorical cognition as an object of knowledge. Metaphors such as INTIMACY IS CLOSENESS or AFFECTION IS WARMTH are classic examples of experientially motivated primary metaphors in which we associate two different conceptual domains.

Complex metaphors are not considered to be directly experientially motivated. They are a combination of primary metaphors recruiting wider frame knowledge (including believes, norms, etc.) (e.g. Grady Reference Grady2005b; Lakoff Reference Lakoff2008a). Standard examples of complex metaphors are long-discussed conceptual metaphors such as LOVE IS A JOURNEY or THEORIES ARE BUILDINGS. In these cases, the metaphorical mapping doubtlessly takes place at the conceptual level. It is thus suggested here that both primary and complex metaphors imply a mapping that takes place at the conceptual level and have (aspects of) the body image as a source domain.

A key element to discuss with respect to the difference between primary and complex metaphors is the notion of context. I propose that both primary and complex metaphors are contextually determined, although in very different ways. In order to highlight the contextually mediated nature of metaphors from the body image, we need to address a classic and hotly debated topic in CMT, which concerns the universal or nearly universal nature of conceptual metaphors, especially primary ones (e.g. Lakoff Reference Lakoff2008a). However, universality claims seem to be undermined both by theoretical considerations about the role of context in metaphorical conceptualization and by the huge amount of data on cross-cultural variation.

Kövecses (Reference 352Kövecses2005, Reference Kövecses2010a) stressed that there is a great amount of data on universal conceptual metaphors, and their universality seems to rely precisely on the universality of our bodily experiences. As Lakoff (Reference Lakoff2008a: 26) suggested, we all have the same bodies and almost the same relevant environments, and during our childhood the same connections between domains are likely to be built. We all have lived the physical experience of being up when we are happy. We all have lived the physical experience of warmth when feeling affection. Hence, this first level of metaphorical conceptualization is universal and deeply grounded in our bodies – although, in Lakoff’s account, it does not follow that those conceptual metaphors are expressed exactly in the same way in all languages. Nor does it follow that cross-cultural variation does not play any role at higher levels of conceptualization. In Lakoff’s view (Reference Lakoff2008a, Reference Lakoff2014), primary metaphors are atomic units that can be the basis for wider systems of contextually determined complex metaphors.

However, empirical evidence abounds on the side of cross-cultural or within-culture variations that can even be applied to primary metaphor (Kövecses Reference Kövecses2010a).⁴ According to Kövecses (ibid.: 204), this tension between universality and cultural variations in conceptual metaphors is the result of two different and concomitant pressures underlining our metaphorical conceptualization: One is the pressure of embodiment that forces metaphorical conceptualization toward its universal characterization; the other is the pressure of context, which is determined by local culture and forces metaphors in the direction of cross-culture and within-culture variations. To reconcile these different pressures, Kövecses (ibid.) proposes the notion of ‘differential experiential focus.’ Based on the idea that our bodily experiences consist of many different components, it assumes that we can single out and emphasize different aspects of that experience when using the same bodily experience as the source domain of a metaphor – depending on the context in which the metaphorical conceptualization takes place. Context is defined broadly by Kövecses (ibid.: 204) as the set of physical, social, cultural, and discourse aspects, the last of these also in relation to factors such as topic, audience and medium, that can affect the conceptualization of metaphors. Differences in experiential focus result from the role of context in the creation of primary metaphors. Although we all share a large number of physical and subjective experiences, which of these experiences that we focus on when establishing a primary metaphorical connection depends on the context in which these experiences are embedded. Hence, at this level, the role of context is that of selecting some aspects of our universally shared experiences. The context works as a lens that allows us to isolate some features of primitive universally shared human experiences that often and very early on occur together in our life (for a related view, see Casasanto, this volume).

Things are different with complex metaphors. Here, the role of context is much more pervasive. The blending of elements (primary metaphors, frame knowledge, beliefs, other complex metaphors, metonymies etc.; see Grady Reference Grady2005b; Lakoff Reference Lakoff2008a; Ruiz de Mendoza, this volume) that is involved in building complex metaphors is largely determined by contextual factors. Complex metaphors completely rely on cultural practices.

Issues related to the universality vs. culture-specificity of metaphorical conceptualization clearly need to be addressed in great detail because the metaphorical mappings take place at the conceptual level, from concept to concept – and even in the case of primary metaphors – from a conceptual representation of the body to another conceptual representation.

When considering the mapping from motor abilities to perception, the issue of the different pressures applied by embodiment and culture has a very different import. On the one hand, it is true that there are motor activities that are deeply culturally embedded and that these activities, for example, the practice of a specific dance or sport, have specific motor programs underlying them. As such, these motor programs are culturally determined.

In this context, it should furthermore be noted that the mechanism of embodied simulation is sensitive to these cultural differences. If I have never seen someone dancing the capoheria, I will likely not have any motor simulation of the observed action or a very low activation of the corresponding motor areas when I finally watch someone dancing it. The motor programs that in our motor system implement and underlie any culture-specific practices are the product of our learning activities in specific contexts.

On the other hand, when considering the contribution that our body schema makes to cognition, we primarily think of the set of processes and mechanisms that are the condition of possibility of any learning activity. These mechanisms and processes are innate and universal. Hence, the contribution the body schema makes to perception is the matrix of species-specific and universal aspects in human cognition.

2.1 MORE IS UP

We first discuss linguistic, gestural, cultural, and cognitive reflections of primary metaphors with MORE IS UP. This metaphor is expressed when English speakers make statements like “high tax rates,” or when German speakers make statements like “Die Preise sind gestiegen” (‘The prices have risen’). Besides these linguistic reflections, MORE IS UP is expressed via gesture. Metaphor is often expressed in manual gestures (Chui Reference 340Chui2011; Cienki Reference Cienki and Koenig1998; Cienki & Müller Reference Cienki and Müller2008b; Sweetser Reference Sweetser1998). Figure 6.1, a still image from the TV News Archive, shows a gesture by Michael Hayden, former CIA director (2006–2009).¹ Hayden is talking about employment in the CIA and, specifically, about a division of the CIA that he classifies as “core support.” According to Hayden, there is a “disturbingly high number of contractors” in this core support division. In Figure 6.1, the palm of his left hand is facing toward the audience and toward the camera. As he says “high number,” he moves the hand upward. This movement is consistent with MORE IS UP. That the movement is time-locked with the verbal phrase suggests that gesture and the metaphorical semantics are tightly coupled in this example (for a close discussion of further examples, see Winter et al. Reference Winter, Matlock, Knauff, Pauen, Sebanz and Wachsmuth2013).

Figure 6.1 MORE IS UP expressed in co-speech gesture on the phrase “high number.” The hand starts low and moves up to a higher position (a), with the end point shown in (b).²

Cultural reflections of MORE IS UP are pervasive. Floors in tall buildings are numbered with smaller numbers at the bottom. Doctors measure humans using scales with smaller numbers at the bottom and increasingly larger numbers going upward. Beakers and measuring cups have small numbers at the bottom and large numbers at the top, and so do thermometers. Cultural reflections of MORE IS UP are prevalent in graphs (Tversky Reference Tversky2011), where it is a convention to put larger numbers higher on the y-axis than lower numbers. This characterizes particularly bar plots and line graphs, which are often used in science and newspapers. Empirical work on the understanding of graphs has furthermore shown that vertical bar plots that embody MORE IS UP are easy to interpret, more so than horizontal bar plots (Fischer et al. Reference Costantini, Galati, Ferretti, Caulo, Tartaro and Romani2005).

Some researchers have pointed out exceptions in the cultural patterns of MORE IS UP (e.g. Holmes & Lourenco Reference Holmes, Lourenco, Carlson, Hölscher and Shipley2011; see also Tversky Reference Tversky2011). Consider numbers on cell phones, where smaller numbers are at the top rather than the bottom, or the rank orders of tournaments, where the first rank is listed at the top and lower ranks at the bottom. However, crucially, these number uses contrast with measurement cups and thermometers in that they do not imply a true sense of quantity, which is commonly called a “cardinal” use of number. Instead, cellphones present a primarily “nominal” use of numbers (e.g. the phone number +1-998-532-9193 is not “more” than the number +1-138-777-6124), while tournament ranks represent an “ordinal” use of number (see Nieder Reference Nieder2005 for excellent discussion). Hence, cultural reflections that do map true quantity onto verticality tend to obey the MORE IS UP principle (see Winter et al. Reference Winter, Matlock, Shaki and Fischer2015).

Experiments have provided abundant behavioral evidence to support the claim that people have a mental association between verticality and quantity (for a review, see Winter et al. Reference Winter, Matlock, Shaki and Fischer2015). For example, when people are asked to generate a sequence of numbers as randomly as possible, they tend to generate larger numbers after having moved their eyes upward (Loetscher et al. Reference Loetscher, Bockisch, Nicholls and Brugger2010), or after having moved their head upward (Winter & Matlock Reference Winter, Matlock, Knauff, Pauen, Sebanz and Wachsmuth2013). Similar associations between verticality and quantity have been found in more classic button-press paradigms. For example, when participants are asked to indicate whether a number is “even” or “odd” through using two buttons, they tend to respond faster to a larger number when the button is located at a relatively higher position (Hartmann et al. Reference Hartmann, Gashaj, Stahnke and Mast2015; Ito & Hatta Reference Ito and Hatta2004; Müller & Schwarz Reference Müller and Schwarz2007).

2.2 Social Distance is Physical Distance

People often talk about social distance in terms of physical distance. For example, in discussing friendships or romantic relationships, English speakers make statements such as “The couple is slowly drifting apart,” or “Bill and Marco have gotten closer lately.” Such cases imply a change in social distance, not physical distance. German speakers also do this, as seen in “Wir waren uns einmal sehr nah” (‘We were very close once’). In these and other linguistic reflections of the primary metaphor SOCIAL DISTANCE IS PHYSICAL DISTANCE (aka INTIMACY IS CLOSENESS), people talk about aspects of social relationships in terms of physical space. More precisely, the amount of space separating people reflects the nature of their relationship, such that larger distances indicate larger degrees of estrangement/alienation, etc.

An example of gesture related to SOCIAL DISTANCE IS PHYSICAL DISTANCE is shown in Figure 6.2, also from the TV News Archive. In this example, Sir Paul McCartney is being interviewed by David Letterman. In describing his relationship with pop star Michael Jackson he says, “So, we kinda drifted apart.” He describes how the relationship deteriorated, and in doing so, he gestures. He raises both hands to his chest, where they are momentarily held close together, and quickly moves them away from each other. Critically, the distance between McCartney’s two hands is smaller at the beginning of the gesture than it is at the end, reflecting increased distance, showing how people spontaneously use the spatial modality of gesture to express state changes in social relationships.

Figure 6.2 Co-speech gesture expressing SOCIAL DISTANCE IS PHYSICAL DISTANCE. First the hands are close together (a), then farther apart (b).³

Above and beyond gesture, we see other cultural reflections of SOCIAL DISTANCE IS PHYSICAL DISTANCE. Social scientists have discussed “segregation effects” and “peer effects” in the context of human relationships.⁴ Segregation effects refer to cases in which people physically move closer to others they perceive as similar to themselves (Miller & Page Reference Miller and Page2007: 143–146; see also Bishop Reference Bishop2008). Peer effects refer to cases when people get physically close to each other and then begin to pick up certain behaviors from each other (Christakis & Fowler Reference Christakis and Fowler2009). So people tend to move toward others they perceive as similar, and to become more like them once they are physically close. These two tendencies can lead to large-scale correlations of social distance and physical distance. This principle is not only characterizing modern societies but has also been shown for old hunter-gather settlement sites (Wiseman Reference Wiseman2014). Thus, culture at large reflects the principle of SOCIAL DISTANCE IS PHYSICAL DISTANCE.

The association between ‘social distance’ (‘intimacy’) and ‘physical distance’ is also reflected in film, another form of cultural representations. Here we discuss a scene from Before Midnight, the third movie in a trilogy about two characters in stormy relationship.⁵ The main characters, Jesse and Celine, husband and wife, are on vacation on a Greek island and, instead of having a romantic evening, have a heated discussion about Jesse’s teenage son, Henry. The argument drags on for 20 minutes, spanning a wide range of heated topics, such as irrational thinking in men versus women, personal sacrifices in marriage, and infidelity. This part of the argument fluctuates in emotional intensity, becoming very loud and caustic at times, but calm and subdued, at others. In the end, Celine leaves the room when she becomes enraged with Jesse.

Over the course of the argument, the physical distance between Jesse and Celine changes in ways that are consistent with SOCIAL DISTANCE IS PHYSICAL DISTANCE. When the argument becomes aggressive and heated, there appears to be greater physical distance between the characters. Figure 6.3 shows two successive moments at the beginning of the argument. Jesse sits on the bed in (a), and Celine, on a couch in the adjacent room, in (b). Jesse pointedly asks, “So this is how you want to be spending this evening? I mean, this is what you wanna do tonight?” to which Celine curtly responds, “Well, you started it!” At this time, Jesse and Celine appear to be as far from each other as possible in the hotel room. Importantly, the distance between the viewer and the characters in the film is also accentuated via the camera. In (a), Jesse appears to be very far away, both from Celine and from the viewer. The physical distance is highlighted by the fact that the door frame is included in the shot, showing physical separation between the two. Moreover, both shots in (a) and (b) show the full body (total shots). In the previous scenes where the couple was intimate, there were more close-ups, focusing on the face and suggesting physical proximity. Thus, both the relative positioning of the characters and the position of (and perspective taken by) the camera adhere to SOCIAL DISTANCE IS PHYSICAL DISTANCE.

Figure 6.3 Spatial positions of characters expressing SOCIAL DISTANCE IS PHYSICAL DISTANCE in Before Midnight. (a) Jesse criticizes Celine for arguing when they should be making love instead; (b) Celine retaliates by blaming Jesse for having started it.

Figure 6.4 shows two shots from before the argument, where physical (and hence also social) distance appears to be smaller, even when Jesse and Celine do not share the same frame. One way that physical proximity is suggested is by the size of the characters in each shot: Rather than the full total shots in Figure 6.3 (showing each body almost entirely), the characters are shown up close, with most of the image space taken up by their bodies. Further, Figure 6.4b shows the same door frame as Figure 6.3a, but separation seems less because Celine is standing at the door.

Figure 6.4 Spatial positions of characters expressing SOCIAL DISTANCE IS PHYSICAL DISTANCE in Before Midnight. Jesse (a) and Celine (b) in a friendly interaction about a gift from their friends.

So far, we have talked about linguistic, gestural, and cultural reflections of SOCIAL DISTANCE IS PHYSICAL DISTANCE. Several experimental studies suggest that the domain of physical distance is automatically accessed when thinking about likeability and intimacy. For example, in Matthews and Matlock (Reference Matthews and Matlock2011), people drew a path through a park depicted by a map. On the map, stick figures represented characters described as “friends” (low social distance) or “strangers” (high social distance). On average, lines intended to represent paths were closer to the “friends” than to the “strangers” on the map. In another experiment, Williams and Bargh (Reference Williams and Bargh2008b), people drew two points on a sheet of paper, either very close to each other or very far away from each other. People in the “far” condition reported that they felt less of an emotional attachment to family members than did people in the “close” condition (although Pashler et al. Reference Pashler, Coburn and Harris2012 report a failure to replicate this result). These and other studies indicate that when performing spatial tasks or when making social judgments, people automatically consider social distance and physical distance together. Such work supports the idea that SOCIAL DISTANCE IS PHYSICAL DISTANCE is not merely expressed in linguistic, gestural, and cultural content but is part of our conceptual system.

2.3 SIMILARITY IS PROXIMITY

English speakers often talk about SIMILARITY in terms of how PROXIMAL or DISTAL things are relative to each other. For example, your friend describes her political views as being “very far from” or “very close to” your political views (cf. examples in Casasanto Reference Casasanto2008a: 1047), or a chef tastes her sauce and says, “It’s getting close now,” referring to how similar the sauce is to the sauce she made a week ago. A comparable German example is seen in statements such as “Diese Ansichten sind weit voneinander entfernt” (‘These views are far away from each other’).

Just as with SOCIAL DISTANCE IS PHYSICAL DISTANCE, the primary metaphor SIMILARITY IS PROXIMITY can be expressed through gesture. The speaker in Figure 6.5, Michael Powell, CEO and president of the National Cable & Telecommunications Association, is asked by an interviewer whether wired and wireless markets should be regulated in the same way. He responds that regulations should be “harmonized” and that wired and wireless markets are “increasingly trending toward being more similar, not more different.” When he is talking about wired and wireless technologies becoming “more similar,” he moves his hands, palms facing toward each other, toward the middle of his body. While saying “not more different,” he moves his hands apart. This sequence is integrated, with his hands continuously approaching each other and retracting again, beginning with “increasingly” in the utterance. The gesture makes it clear that two spatial positions are prominent, one being close (coinciding with the “similar” part of the sequence) and one being far (coinciding with the “far” part of the sequence). In this example, distance is primarily marked through dynamic movement toward or away from the midpoint of the body. The amount of distance between the hands is associated with the degree of similarity or difference.

Figure 6.5 Co-speech gesture expressing SOCIAL DISTANCE IS PHYSICAL DISTANCE. Both hands move toward a location in the center of the body for expressing SIMILARITY (a). For expressing DIFFERENCE, the hands move apart (b).⁶

SIMILARITY IS PROXIMITY also appears in the spatial location and configuration of cultural artifacts. We see this in how rooms in houses are arranged. Things that are similar by virtue of function are grouped together. For instance, substances for cleaning the body are co-located in bathrooms. Edible things are found in kitchens. Clothing items are found together in bedrooms. This pattern is also seen in design, including web design, where things that perform similar functions, such as menu buttons, are positioned close to each other. This pattern is also seen in the design of virtual worlds (e.g. Waterworth et al. Reference Waterworth, Lund, Modjeska, Munro, Höök and Benyon2003). Humans intentionally use space to sort things in their environment according to similarity (cf. Kirsh Reference Kirsh1995). Grouping like things together in space helps people perceive them as similar (Wertheimer Reference Wertheimer and Ellis1938).

Again, these are relatively small-scale reflections of a larger cultural principle that is evident in society-wide scales. For example, people within a city tend to self-organize into districts that are relatively homogenous with respect to factors such as ethnicity or socioeconomic status (Schelling Reference Schelling1971; Bishop Reference Bishop2008). Often, similar people are located near each other because of necessity. Students are located near other students because they attend the same university and live near campus. Nurses are located in or close to hospitals or other medical facilities because they provide care for patients. Commuters are near other commuters because they share the same road or mode of transportation.

What is the experimental evidence for the conceptual nature of the metaphor SIMILARITY IS PROXIMITY? Casasanto (Reference Casasanto2008a) asked participants to rate unrelated words while they were being presented on a computer screen at varying distances. When participants were presented with two words that were close to each other, they rated them as more similar than when they were presented with words that were relatively far from each other. Boot and Pecher (Reference Boot and Pecher2010) found that participants were quicker to judge colors (e.g. two shades of blue) as being similar when the colors were presented close together, and quicker to judge them as different when they were presented relatively far apart (see also Breaux & Feist Reference Breaux, Feist, Love, McRae and Sloutsky2008). Winter and Matlock (Reference Winter, Matlock, Knauff, Pauen, Sebanz and Wachsmuth2013) showed that cities or people that were described as similar were subsequently placed closer to each other in a drawing task. Finally, Guerra and Knoerferle (Reference Guerra, Knoeferle, Carlson, Hölscher and Shipley2012) showed how visual depictions of various distances can affect the comprehension of sentences involving similarity. In their task, participants were shown concepts, i.e. words such as “stupidity” and “wisdom,” on two cards that were near or far from each other, and then read the sentence, “Stupidity and wisdom are certainly different.” Sentences about differences were read more quickly when the words “stupidity” and “wisdom” had been far from each other. Conversely, sentences about similarity were read more quickly when the concepts had been presented close to each other. Similar to our discussion of MORE IS UP and SOCIAL DISTANCE IS PHYSICAL DISTANCE, such research on SIMILARITY IS PROXIMITY suggest that this primary metaphor is deeply entrenched in our cognitive system, and there is much convergent evidence to support the hypothesis that SOCIAL DISTANCE IS PHYSICAL DISTANCE is a mapping not just in language and culture but also in our conceptual systems.