3.1 Proto-mimesis

While there is growing evidence for minimal selfhood from birth (Rochat Reference Rochat, Slaughter and Brownell2011, Zahavi Reference Zahavi2014), the child does not in the first months of life always differentiate between self and caregiver (Werner & Kaplan Reference Werner and Kaplan1963, Stern Reference Stern2000, McCune Reference McCune2008). For example, Reddy (Reference Reddy2003:401) characterizes young infants as having ‘a more immersed, less detached focus on the other’. That the body schema, albeit in an underdeveloped form, is already active at birth is evidenced by the phenomenon of neonatal mirroring (Meltzoff & Moore Reference Meltzoff and Moore1983, Suddendorf et al. Reference Suddendorf, Oostenbroek, Nielsen and Slaughter2013). However, this is distinct from true imitation, which requires volitional control of the body (Piaget Reference Piaget1962). At this early stage, the child's embodied intersubjectivity is foremost realized through processes of bodily resonance, reflected in ‘emotional contagion’ (spontaneously picking up the feelings of close others) and mutual attention (prolonged bouts of looking into the eyes of caregivers). The infant's body memory is of the intercorporeal kind, realized in more or less automatic turn-taking exchanges known as ‘proto-conversations’ (Trevarthen Reference Trevarthen and Bullowa1979, Gratier et al. Reference Gratier, Devouche, Guellai, Infanti, Yilmaz and Parlato-Oliveira2015) and other interactional formats like ‘peekaboo’ (Bruner Reference Bruner1983).

3.2 Dyadic mimesis

By nine months of age infants have considerably expanded body schemas, and emerging body images, with increased capacity for self-movement, volitional control of the body and a more distinct sense of self (Stern Reference Stern2000). In addition, by this age many of the earlier, apparently automatic, interactions like turn-taking show evidence of volitional control (Hilbrink, Gattis & Levinson Reference Hilbrink, Gattis and Levinson2015). It is characteristic that around this age, the first pointing gestures and (true) imitations emerge (Tomasello Reference Tomasello1999), possibly driven by the need to maintain intersubjective closeness with caregivers that are now more clearly perceived as ‘others’ (McCune Reference McCune2008). Imitation itself can be seen as developing through three steps. First there is sensory-motor imitation, in which what is imitated is in the here and now. Somewhat later, the child becomes capable of deferred imitation (Mandler Reference Mandler2004), where there is temporal delay between observation and performance, but without any internal model or representation to guide this. Finally, there is what Piaget calls ‘representative imitation’ described as follows:

The first two steps correspond to incorporative memory in which, to repeat, the child ‘does not represent the past, but re-enacts it through the body's present performance’ (Fuchs Reference Fuchs2012:11). In the third step, however, around the middle of the second year of life, the child is able to actively imagine a non-present action or event, thus moving from re-enactment to imagination. Such action-imitations, developing from intercorporeality, through incorporation to imagination have been called mimetic schemas, defined (in their most elaborated forms) as ‘dynamic, concrete and preverbal representations, involving the body image, accessible to consciousness and pre-reflectively shared in a community’ (Zlatev Reference Zlatev and Hampe2005:334). These can naturally be considered as structures of embodied intersubjectivity, as acknowledged by commentators:

At the same time, mimetic schemas must be distinguished from lexical concepts, which in the case of action verbs like hit, run and throw involve not just body-based schematizations, but structured inventories of normative meanings.

3.3 Triadic mimesis

The earlier stage was named ‘dyadic’ despite the presence of objects in the communicative interaction, e.g. desired ones in so-called imperative pointing, or imitated actions with objects (e.g. in feeding), since only from about 14 months, do child, other and object become fully integrated in a ‘referential triangle’. This is reflected in declarative pointing gestures (Look at that!) and gaze oscillations between object and addressee that mark full joint attention, implying an intersubjective form of ‘third-order mentality’ (I see that you see that I see) (Zlatev Reference Zlatev2008). This is a clear manifestation of intersubjective object perception, in which (to remind) external objects gain their fully transcendent status only when they can be the potential focus of joint attention. Furthermore, it is only with triadic mimesis, possibly a uniquely human capacity (Zlatev Reference Zlatev2008) that Gricean intentional communication can take place: intending both to communicate something to an audience, and for the audience to recognize this intention. This implies an embodied second-order intention, enacted through bodily actions like gaze, smile, raised eye-brows, or holding of the gestures reply (Zlatev et al. Reference Zlatev, Madsen, Lenninger, Persson, Sayehli, Sonesson and van de Weijer2013)

3.4 Proto-language and language

While the first words typically appear at the beginning of the second year of life, these are restricted to specific contexts and function more like indexical schemas that are associated with – rather than as symbols that stand for – emotional states and external events (Bates et al. Reference Bates, Benigni, Bretherton, Camioni and Volterra1979). In contrast, around their second birthday most children display a marked increase in the number and variety of their words. This is often described in the developmental literature as a ‘vocabulary spurt’. It is possible to relate this to the emergence of reflective consciousness (Zelazo Reference Zelazo2004) and the symbolic insight that ‘things have names’ and that these names are common, i.e. conventional (Zlatev Reference Zlatev2013). Still, this level of intersubjectivity would have been impossible without the layers of embodied intersubjectivity laid down in the preceding states, and in particular those related to joint attention and imitation-derived representations (mimetic schemas). From this point on, language co-develops and interacts with the use of other semiotic resources, such as gestures (McNeill Reference McNeill2005) and pictures (DeLoache Reference DeLoache2004), gradually increasing in structural complexity (Tomasello Reference Tomasello1999). This makes the distinction between ‘proto-language’ and language proper in ontogeny a gradual one, in line with theories of language acquisition that propose that what children learn are constructions, slowly increasing in complexity and abstractness (Goldberg Reference Goldberg2006, Tomasello Reference Tomasello2009).

In sum, the Mimesis Hierarchy outlined in this section, as well as other layer-based models of semiotic development (Stern Reference Stern2000), can serve as empirical explications of the two main theoretical claims of this article. First, that there are cognitive-semiotic capacities that are more basic than language, and hence necessary prerequisites for its emergence and second, that these are essentially the structures of embodied intersubjectivity that have been ‘unearthed’ in phenomenological investigations, as summarized in the previous section. At the same time, the discussion has hopefully helped to move the notion of sedimentation beyond the level of metaphor. Each layer of the mimesis hierarchy has a ‘horizontal’ structure, in which the various capacities discussed in each subsection form a coherent pattern. At the same time, there is a clear ‘vertical’ structure, where earlier layers are sedimented upon by those following them, eventually forming the ‘ground’ for the normative and systematic intersubjectivity of language.

In the next section, we will see how this plays out with respect to one type of semantic phenomenon: the use of conventionalized expressions of motion to describe situations lacking motion, or what we refer to as ‘non-actual motion sentences’.

4.1 Sedimentation vs. mental simulation

Sentences with motion expressions that denote situations lacking motion (i.e. non-actual motion (NAM) sentences) are common in the world's languages. For example, (3) can be translated more or less directly (though see below) into Swedish, in (4), Bulgarian, in (5), and Thai, in (6), using corresponding NAM-sentences.

1. (3) The road goes into the forest.

2. (4)

   

3. (5)

   

4. (6)

   

It is noteworthy that Thomas Hobbes proposed that such expressions should be banned from philosophy, as they are one of the factors behind ‘absurd conclusions’:

But why do we use such expressions spontaneously in everyday speech, most often without attempting to be the least ‘poetical’? Cognitive linguists have offered a multitude of related explanations, involving notions such as conceptual metaphor (Lakoff & Johnson Reference Lakoff and Johnson1980), fictive motion (Talmy Reference Talmy2000), subjective motion (Langacker Reference Langacker2006), and most recently, mental simulation (Matlock Reference Matlock2010). The common denominator of these, in fact rather different, theoretical constructs is the proposition that the meanings of sentences and other linguistic expressions are not situations or any other worldly entities, but rather one or another kind of mental representations. As expressed in one of these accounts, to (use and) understand sentences such as these, (speakers and) hearers need to form dynamic underlying mental images, or representations:

But there are multiple problems with equating such hypothetical mental processes with the meanings of sentences like (1)–(6). First of all, while some people, on some occasions may indeed carry out such imaginings (or if that implies too much conscious awareness: ‘simulations’) these have the wrong kind of ontological status to serve as linguistic meanings. Mental images, static or dynamic, are private while linguistic meanings are public, conforming to super-individual norms or conventions, and the two should not be conflated (Itkonen Reference Itkonen2016).

This may sound like a purely philosophical argument, but its validity can also be demonstrated on linguistic grounds. If the same kind of mental process makes up the meaning of NAM-sentences across languages, then these should be practically synonymous. But that was not the case even with (3)–(6). As the carful reader may have noted, (3) in English and (4) in Swedish used the verbs go and går, which in terms of motion event typology would be classified as expressing Manner and possibly Deixis (in contrast to come). On the other hand, the motion verb used in the Bulgarian example (5) expresses the category Path and the Thai sentence (6) combines Path and Deixis in a serial verb expression. In other words, the ‘same’ meaning is rendered according to language-specific conventions, resulting in sentences that are not fully synonymous. The point can be made even clearer by looking at the well-known example (7), used by Talmy to show that ‘fictive motion’ sentences are not limited to cases where the grammatical subject denotes objects that may serve as routes for travel.

1. (7) The mountain range goes from Mexico to Canada.      (Talmy Reference Talmy2000:104)

However, this sentence cannot be directly translated to Bulgarian and Thai, where the closest corresponding motion verbs varvi and dqqn ‘walk’ would imply a fairy-tale scenario where the mountain range has gained legs. The Swedish example (8) is a possible (and occurring in actual use) sentence, but it could be a translation loan from English, and a preliminary corpus search shows few cases of motion verbs with figures like bergskedja ‘mountain range’.

1. (8)

   

   (http://www.toppenipyreneerna.se/vandringsleder-i-pyreneerna.html)

Other languages like Yucatec Maya have specific constraints with respect to the use of NAM-sentences, in line with their typological characteristics (Bohnemeyer Reference Bohnemeyer, Barbara and Wolff2010). In short, while it is meaningful to search for experiential motivations that may partially explain such expressions, these motivations need to be clearly distinguished from the corresponding conventional semantics.

Second, it would be at best premature to look for a single experiential motivation as explanatory psychological mechanism. We may wonder with respect to the quotation from Richardson & Matlock (Reference Richardson and Matlock2007) given above: what is it that serves as the ‘protagonist’ and what exactly is being (consciously or unconsciously) simulated? Any one of the following could be the case: one's own imaginary self-motion along a path; the motion of some external imaginary entity (e.g. a car); the extended figure itself ‘as if’ moving; the observer's visual attention along the path, and possibly others. Clearly these correspond to different kinds of experiences. Based on a meta-analysis of how this issue was addressed in some of the most representative studies of non-actual motion in the cognitive linguistic literature, Blomberg & Zlatev (Reference Blomberg and Zlatev2014) distinguished between three experiential factors, serving as possible pre-linguistic motivations for the use of NAM-sentences:

1. (a) the enactive, action-oriented nature of perception. This is mostly plausible for expressions concerning figures that afford motion, as in examples (3)–(6);

2. (b) the correlational nature of intentionality, where every intentional object is linked to one or more intentional acts.Footnote ⁴ This may be reflected in examples that involve ‘visual scanning’, as in (7)–(8);

3. (c) the imagination of counterfactual scenarios. This is closest to a truly metaphorical ‘as if’ reading of a NAM-sentence, as in example (9) below.

1. (9) Insanity runs in my family… It practically gallops!

   (Brandt Reference Brandt2009:573, cited from Arsenic and Old Lace)

As reflected by this classification, some non-actual motion sentences can be tentatively attributed to one or the other of these three motivations. However, there is nothing to restrict the combination of several different motivations such as (a)–(c) above, leading us to the conclusion that the phenomenon of NAM-sentence use cannot be explained by a single factor.

Third, and especially relevant for the present discussion, while most cognitive linguists locate meaning in the individual mind, using terms like ‘mental representation’ with few reservations, the sedimented upon experiential structures listed above are strictly speaking, neither individual nor representational. Why is that so? With respect to (a) affordances of both natural and cultural objects, like the ‘sittability’ of chairs, and the ‘walkability’ of paths, are per definition perceptual features in ecological psychology (Gibson Reference Gibson1979). In the case of (b), figures that do not afford self-motion like mountain ranges, the intentional act of ‘scanning’ their length is still mostly perceptual, or quasi-perceptual (as reflected in the terminology used). The least perceptual is (c), but as suggested in the discussion of incorporative memory and mimetic schemas in Sections 2 and 3, the kind of re-enactment involved in imagination is not to be thought of as a form of pictorial representation, but as dynamic and sketchy re-presentation (Thompson Reference Thompson2007).

Furthermore, we should consider (a)–(c) as fundamentally intersubjective rather than private. With respect to (a) the affordances of paths and highways are the affordances of cultural artefacts, which are not just ‘mine’ but inherently point to other embodied subjects. As nicely expressed by Möttönen (Reference Möttönen2016:160):

With respect to (b), as perceiving any three-dimensional object, and not just cultural artefacts, implies the implicit awareness of other perspectives and possible co-perceivers, we may say that in ‘scanning’ an elongated figure like a fence we are implicitly aware that others could do likewise, or perhaps differently. Scanning is thus something like a perceptual affordance. Finally, as (c) will involve re-enactments of culturally salient actions like walking, running and crawling corresponding to mimetic schemas, their mimetic origins will guarantee a degree or pre-reflective sharing.

An application, and to some degree validation of this reasoning is provided in an experimental study by Blomberg (Reference Blomberg2015). Native language speakers of Swedish, French and Thai were presented pictures such as those shown in Figure 1 and asked to describe what was presented in spontaneous sentence-length utterances. Half of the pictures represented figures that afforded self-motion, and half showed figures that do not. Moreover, the same objects were displayed either from a first-person perspective, or third-person perspective. The reasoning was that if only self-motion affordance (or ‘simulation’) had an impact on the use of NAM-sentences, then perspective would not matter. On the other hand, if ‘scanning’ were the strongest underlying motivation, then figures shown from the side would evoke more NAM-sentences, irrespective of motion-affordance. The results showed that all categories of pictures elicited NAM-sentences on the scale 40% for all three languages, significantly more than control pictures of non-extend figures like houses and trees. Furthermore, again for all three languages, pictures like that shown in Figure 1, elicited more NAM-sentences than the other cases, indicating an interaction between the factors affordance and perspective.

Figure 1. The stimulus picture described in Swedish (10), French (11) and Thai (12), showing a figure that affords self-motion from a first-person perspective, of the type that elicited most NAM-sentences in all three groups.

Moreover, the elicited NAM-sentences consistently followed language-specific conventions for expressing actual motion, but in comparison to these, featured more abstract descriptions, avoiding the use of manner-of-motion expression. In the Swedish group, the two predominantly used verbs were gå ‘go’ and leda ‘lead’, both of which are quite generic, in examples such as (10). The French speakers used predominantly Path verbs such as sortir ‘exit’, as shown in (11). The Thai speakers used serial-verb constructions, but tended to omit the Manner verb in the series, using only a Path verb together with a Deictic verb, as in (12). These descriptions were given in response to the image shown in Figure 1.

1. (10)

   

2. (11)

   

3. (12)

   

4.2 Sedimentation vs. conceptual metaphor

In another study, we compared non-actual motion sentences denoting affective states or processes of consciousness, i.e. emotions, such as the more or less parallel but not synonymous examples from English (13), Swedish (14) and Mandarin Chinese (15). It is possible to analyse such examples in terms of motion–emotion metaphors, i.e. as ‘mappings’ between the domains of MOTION and EMOTION (Zlatev et al. Reference Zlatev, Blomberg, Magnusson, Foolen, Luedke, Racine and Zlatev2012).

1. (13) My spirit soared.

2. (14)

   

3. (15)

   

The reason behind the caution expressed above, is that it is rather controversial how to define the notion of metaphor. One current debate concerns the fundamental level on which metaphors exist. For many cognitive linguists, this is the level of so-called conceptual metaphors, defined as ‘a cross-domain mapping of structure from a source domain to a target domain, where the two domains are regarded as different in kind’ (Johnson Reference Johnson2010:407). Others have proposed that we should privilege language use, and focus on recurrent discourse metaphors such as our European home, intermediary in their conventionality between novel metaphors relying on analogical reasoning and literalized expressions (Zinken Reference Zinken2007). Or perhaps metaphors should be studied even on the most specific, contextual level, so that

It should be noted that it is what is ‘most basic’ that is contested, as there is a degree of consensus that the levels of (universal) bodily experiences and cognitive processes, (conventionalized) cultural practices, and situated language use are not exclusive but rather complementary, in accordance with the Integral Linguistics of Eugenio Coseriu.Footnote ⁶

The second debate concerns the balance between universality and language/culture specificity of key metaphors, such as that relating the domain of time and space. As may be expected, the stance taken in this debate correlates with that taken in the previous one: if metaphors, and especially so-called primary metaphors (Grady Reference Grady1997) such as AFFECTION IS WARMTH ‘are acquired unconsciously through our bodily engagement with our environment’ (Johnson Reference Johnson2010:410) a considerable degree of universality may be expected. If, on the other hand, metaphors are essentially discursive constructs, cultural specificity follows almost by definition. Thus, while proponents of the universalist stance may accept a degree of cultural variation (Kövecses Reference Kövecses2000), e.g. ‘social constructions are given bodily basis and bodily motivation is given social-cultural substance’ (ibid.:14), the controversy will persist as long as the deeper, definitional disagreement does.

In these debates it is not hard to see conceptual problems that may be at least partially resolved with the help of the notions of embodied intersubjectivity and meaning sedimentation. To begin with, many suggested ‘primary metaphors’ clearly involve intercorporeal experience. This is hardly surprising as they are hypothesized to be ‘acquired by children simply because of the nature of the bodily experience (in perception and bodily movement) for the kinds of the structured environments they inhabit’ (Johnson Reference Johnson2010:410).

In the paradigmatic example of AFFECTION IS WARMTH, these ‘environments’ are crucially interpersonal. Such ‘correlations’ (between affection and warmth, etc.) correspond to felt, reciprocal qualities in the Leibkörper of infant and mother, i.e. to intercorporeal body memory.

However, is it really appropriate to analyse such structures, emerging through bodily interaction and involving (in most cases) emotion, as ‘cross-domain mappings’? As suggested in Section 2, unlike the way they are usually represented in cognitive semantics (Kövecses Reference Kövecses2000), emotions hardly constitute an ‘abstract target domain’, and even less so one that is ‘different in kind’ (see Johnson's definition above) from their bodily expressions. Rather, following Fuchs's analysis, the internal and external sides of emotion are intimately connected through bodily resonance, allowing them to be perceived directly, without the need for inference or simulation. Of course, emotions can be conceptualized in many different ways, in different cultures and languages, e.g. as ‘forces’ or ‘fluids’, but these are secondary, language-mediated, and indisputably metaphorical constructions. If it is the pre-conceptual, and (largely) universal bodily roots of metaphors that concern us, then an analysis in terms of ‘cross-domain mappings’ would seem to be placing the cart before the horse.

Studying the ‘mapping’ between the ‘domains’ of motion and emotion in more or less related languages confirms this theoretical conclusion. On the basis of native speaker intuitions and corpus analysis, we performed an exhaustive search for conventionalized expression types (i.e. constructions) in English, Swedish, Bulgarian and Thai where a motion verb is used to denote a state of affect/emotion, without any perceived motion, as in my heart dropped. One hundred and fifteen such metaphors (on the intermediary level of conventional metaphor types, see above) were identified and compared. As expected, the findings showed considerable cross-language differences, especially between Thai and the three European languages, but also significant similarities, shown in Table 1. The first two rows reflect a familiar ‘primary metaphor’ in which upward motion is correlated with positive affect/emotion and downward motion with negative emotion. At the same time, this is expressed differently in the different languages, with e.g. Thai requiring a compound in which the word caj ‘heart-mind’ needs to be added to the verb to make this an expression of emotion rather than purely motion expression. This supports the crucial distinction between the level of conventional linguistic expressions and the level of pre-linguistic motivations. What may the latter be in this case? As pointed out early by Lakoff & Johnson (Reference Lakoff and Johnson1980) one likely motivation for this EMOTION IS MOTION ‘primary metaphor’ is that positive/negative emotion corresponds to higher/lower bodily posture. In such cases, this would be yet another reflection of Leibkörper duality, with felt qualities being ‘worn on the sleeve’ of the living body, since Leib and Körper are not to be treated as two distinct domains. To remind, emotions have both an affective, felt Leib component, and an emotive, shown Körper component, which is key for them to be shared.

In other cases, especially when the motion verb expresses motion through a liquid (one typically ‘sinks down’ (sjunker ner) into a depression in Swedish), a cross-domain analogy may be a more appropriate analysis. There is, in other words, nothing to guarantee that the ‘overlap’ of metaphors across unrelated languages is to be explained by the same mechanism.

The same goes for the lower three rows. Here we see that in the European languages, and in the case of STIR also for Thai, the basic form of the motion–emotion construction is transitive: C(ause) moves F(igure). All of these are negative, but there is a kind of hierarchy: STIR < SHAKE < SHATTER, so that as the caused motion on the figure/experiencer becomes more violent, the emotion/feeling becomes more negative. In the case of SHAKE, we may again see this as a Leib–Körper correlation in experience. For STIR and SHATTER, the motivation is more likely in the similarity/analogy of the felt (inner) sensation and the observed transformations of external objects: with brews being stirred, and fragile things shattered. Even in these latter cases, intercorporeal grounding of the ‘source domain’ would be guaranteed by the mimetic schemas corresponding to STIR and SHATTER.

4.3 Summary

In this section we presented the theoretical background for and summarized the main findings from two different studies of non-actual motion expressions across a number of different languages. The first study focused on expressions denoting routes and other extended figures using motion verbs (Blomberg Reference Blomberg2015). The second one considered expressions that ‘mapped between’, or more appropriately conflated or blended motion and emotion (Zlatev et al. Reference Zlatev, Blomberg, Magnusson, Foolen, Luedke, Racine and Zlatev2012). Despite the semantic and methodological differences, both studies confirmed the distinction between experiential motivations and conventional meanings, and the intersubjective-perceptual, rather than individual-representational character of the first.

In both studies, we saw how certain pre-linguistic experiential structures can help understand why different kinds of non-actual motion sentences are widely spread across languages. They are clearly not ‘arbitrary’ social constructions. The main point is that these intercorporeal motivations should not be confused with the conventional expressions that sediment upon them. Concerning the definitional debate on metaphor, it is therefore appropriate to question Lakoff & Johnson's (Reference Lakoff and Johnson1980) redefinition of metaphor as a ‘cognitive rather than linguistic’ phenomenon: metaphors are not primarily in the (individual) mind, as they necessarily involve expression, in language or in some other semiotic resource (Kolter et al. Reference Kolter, Ladewig, Summa, Müller, Koch and Fuchs2012). Expressions such as those in Table 1 above are conventional linguistic constructions that are sedimented upon with only partially recoverable motivational experiences. In certain contexts these experiences may be ‘unearthed’ and extended but whether and how this is done is largely indeterminate, due to the fundamentally creative nature of language use.

In accordance with the concept of sedimentation, which we professed to explicate along the way in this article, we can see how the sedimenting meanings of language are more abstract and objective, and at the same time presuppose the sedimented upon ones, which are more experientially rich, and at least in part recoverable. The fact that the latter are structures of pre-linguistic embodied intersubjectivity, rather than purely private mental processes can help explain both the emergence of linguistic intersubjectivity, and our recurrent claim that the different layers of meaning do not stand in opposition, but in a complementary relationship to one another.