Turning back to experience in Cognitive Linguistics via phenomenology

1 Introduction

What attracted me to Cognitive Linguistics in the late 1980s was the promise of bringing language back to experience. Rather than just skeletal trees, meaningless symbols, computational algorithms, possible worlds and mathematical functions, etc. the door was opened toward understanding language as what it felt like: rich in imagination, rooted in the body, socially negotiated and driven by communicative needs. Even more, the pathway of the “three milestones” (Johnson 1987; Lakoff 1987; Langacker 1987) seemed to extend toward a better, and more humane understanding of the mind and the world at large.

Such early enthusiasm was in many ways naïve. Many issues and debates have surfaced during the past 30 years. Characteristically, they have circled around the notion of experience. Is language (primarily) grounded in individual (Talmy 2000) or social (Harder 2010) experience? Is such experience basically prelinguistic and continuous with general cognition (Johnson 1987), or is there a level that is distinct to language (Dor 2015)? Finally, and perhaps most controversially, is the relevant experience predominantly conscious (Zlatev 2007), or rather inaccessible to consciousness (Lakoff and Johnson 1999)? The stances on these issues are by far not only philosophical as they have crucial implications for methodology: what role are methods like intuition and introspection to play, or should they even be abandoned in favor of methods like corpus analysis and psycho- and neurolinguistic experiments? Clearly, different answers to such questions imply different interpretations of the modifier “cognitive” in the label of the field.

While I have been engaged in some of these debates (Zlatev 2008, 2010), in the present article I adopt a more reconciliatory tone, and suggest that they could be clarified, if not resolved, with the help of phenomenology: the movement in philosophy founded by Edmund Husserl over a century ago. The reason is that phenomenology has dedicated itself precisely to “the study of human experience and of the ways things present themselves to us in and through such experience” (Sokolowski 2000: 2). Despite many internal debates, inevitable for any movement, phenomenology has always emphasized the fundamental (or “transcendental”) role of conscious experience – of both subjective and intersubjective nature – for the constitution of an objective life-world that can be subsequently studied by science: “Scientific objectivity is something to strive for, but it rests on the observations and experiences of individuals; it is knowledge shared by a community of experiencing subjects and presupposes a triangulation of points of view or perspectives” (Zahavi 2010: 6).

As these citations indicate, phenomenology is far from being either some kind of “introspectionism”, as it concerns our relationship to other people and the world. And it is not an “idealistic” philosophy as it is a deeply empirical enterprise, though one that operates with a richer notion of experience than in natural science, allowing meticulous investigations into phenomena like embodiment, temporality, empathy, perception, normativity and language. It is for this reason that phenomenology has begun a productive dialogue with cognitive science (Thompson 2007; Gallagher and Zahavi 2008), leading, in the best cases, to “mutual enlightenment”. In the same spirit, I argue here that phenomenology can help Cognitive Linguistics clarify the role of different kinds of experience, and thus fully embrace an experiential reading of the term “cognitive”.

2 Meaning in individual or social experience?

For the most part, cognitive linguists have frankly adopted an individual-psychological, meaning-as-conceptualization notion of linguistic meaning, employing notions like mental imagery (Langacker 1987), image schemas (Lakoff and Johnson 1999) and mental simulation (Bergen 2012). One of the most explicit formulations is Talmy’s (2000: 4): “For cognitive semantics, the main object of study itself is qualitative mental phenomena as they exist in awareness. Cognitive semantics is thus a branch of phenomenology.”

Others have objected that such a step would lead to a kind of subjectivism that is both unnecessary for language use and would make linguistic communication impossible (Wittgenstein 1953; Harder 2010). For example, if the meaning of the word dog consisted of our individual images/schemas/simulations of our extremely variant personal experiences with dogs, then we would understand very different things by concrete statements like (1), or even by categorical definitions like (2).

As one of the most outspoken critics of “mentalist” semantics, Itkonen (2003, 2016) has argued extensively that linguistic meaning requires a “social ontology”, corresponding to what Frege (2008 [1892]) meant by “sense” in contrast to “idea” in the following classical quotation:

The reference and sense of a sign are to be distinguished from the associated idea. If the reference of the sign is an object perceivable by the senses, my image of it is an internal image, arising from memories of sense impressions which I have had, and acts, both internal and external, which I have performed. Such an idea is often saturated with feeling; the clarity of its separate parts varies and oscillates. The same sense is not always connected, even in the same man, with the same idea. The idea is subjective: one man’s idea is not that of another. … This constitutes an essential distinction between the idea and the sign’s sense, which may be the common property of many and therefore is not a part or mode of the individual mind.

(Frege 2008 [1892]:115)

For Itkonen, as for Wittgenstein (1953), the meaning of a word corresponds to the norm for its correct use. Thus, being able to use dog in sentences like (1–2), and countless others, rather than in contexts where cats, chairs or numbers are meant, amounts to knowing its sense. Linguistic norms (or conventions) have a social ontology along with other kinds of rules, like driving on the left/right side of the road, which Itkonen explicates as three-level knowledge: I know that you know that I know that p, with p = the given norm. In such a way the social level (of meaning) is constructed out of instances of individual knowledge, standing in a particular configuration. The metaphor used by Itkonen is the fishing net and the individual threads and knots that make it up. Thus, Itkonen schematically both relates and distinguishes social meaning/experience, which he attributes to Popper’s “world 3”, from psychological meaning, which is a matter of consciousness or “world 2”, both of which are distinct from the “world 1” of physics and biology (Popper 1979; cf. Geeraerts this issue).

From the standpoint of phenomenology, there is something correct in both of these contradictory approaches, but also something crucially lacking: the intentionality of human consciousness, directed toward the world of experience. Meaning is not properly speaking “in” the mind, but consists in the relationship between intentional act (e. g., in perception) and intentional object. Intentional acts can be either “fulfilled” – if the object is present, like my dog sleeping now on the floor – or “empty”, if the object is absent. Perceptual intentions are rich in sensory experience, much of which is indeed individual (though see Section 3). Imagination is like perception, but always directed towards absent objects, which may even be non-existent. The intentional imaginative acts themselves have the properties that Frege lists for “ideas” without the need to postulate an intermediary level of mental representation between act and object (Sokolowski 2000).

What distinguishes language (and other sign systems) from perception and imagination are signitive intentions. These are complex in the sense that there is a sign that is perceptually present, and which “redirects” intentionality toward a referent or intentional object, the latter often absent. Both the signs and their referential functions are crucially based on the social experience of language acquisition and language use, where we engage in “mutual identification” of shared experiences for the sake of communication (Dor 2015). So what is the relation between the individual perceptual experiences we have of dogs and the collective “sense” or “signified” of the sign dog? Dor (2015) links the two levels of meaning explicitly in his semantic theory: on the one hand, individually variable experiential clusters, and on the other, much less variable senses standing in structural relations to one another in a socially shared “symbolic landscape”. The truth (and social meaning) of sentences like (2) can be established entirely on the level of the symbolic landscape, without delving down to level of sensory experience. For sentences like (1) we will have to rely on what we have mutually identified as referents of dog and sleep, constrained by perceptual experiences in a common life-world, leading us to the same kind of dog-sleeping situations. But do we need a separate (ontological) level of “sense” at all?

And yet, although we seem forced to posit meanings and judgments as mental or conceptual things, such things turn out to be philosophically embarrassing and perplexing. We never directly experience them. They are postulated as something we cannot do without but no one has actually seen one of them. … How do they exist? What sort of entities are they? Are they in the mind or in some sort of third realm between the mind and the world?

(Sokolowski 2000: 98)

It is possible to resolve this “embarrassment” with the help of the concept of intentionality again. Actually, the manner of doing so is surprisingly consistent with cognitive semantics: senses are conventionalized, socially shared construals of their intended referential objects. The latter can be absent, but even so, it is they that remain as the targets of intentionality, rather than the “intermediary” concepts or senses. Frege’s evening star and morning star are thus co-referential expressions differing in construal, as well as the much more elaborate dimensions of construal such as profiling and perspective studied in cognitive grammar. The important thing is to understand construal as an intersubjective phenomenon, where individual signitive referential acts over time become entrenched as socially sanctioned perspectives of their corresponding intentional objects and events (see Möttönen 2016).

In sum, phenomenology allows us to draw the distinction between individual and social experience, but without relegating these to different “worlds” and thus making the difference ontological, and more or less unbridgeable. Furthermore, as hinted above, and explained more in the next section, pre-linguistic intentionality is not purely individual or “private” either.

3 Grounding meaning in pre-linguistic or linguistic experience?

Is linguistic meaning grounded in bodily experiences or in language use? Many would probably wish to say “in both”, but how such a synthesis would be worked through is far from obvious, as there is an inherent tension between the “embodiment-based” and “usage-based” ideologies of language (see Zlatev 1997).

First, let us note that this issue is not the same as that discussed in the previous section. To begin with, while linguistic experience may be primarily social (Vygotsky 1978) it is not always so, as testified in the use of language for thinking, as well as the (universal) ability to reflect on language, i. e., linguistic intuition (see Section 4). Correspondingly, prelinguistic experience is to a considerable extent intersubjective rather than private. In fact, phenomenology can help us appreciate how closely linked, or even fused together, intersubjectivity and embodiment are in structures of intercorporeality or embodied intersubjectivity. In recent work (Zlatev and Blomberg to appear) we document the relevance of the following intercorporeal structures for language: the dual nature of the human body as subject and object of experience (Merleau-Ponty 1962 [1945]), inter-bodily resonance and emotions (Fuchs 2005), body memory (Koch et al. 2012), multiple perspectives in object perception (Zahavi 2001) and the perception of affordances (Gibson 1979).

Intercorporeal experiential structures such as these are pre-linguistic first in an ontogenetic sense, as they constitute necessary stepping stones in children’s social-cognitive development, without which language would not be able to emerge (Bråten 2006; Zlatev 2013). Secondly, a phenomenological analysis can uncover how such perceptual/bodily intentionality underlies language micro-genetically. Sokolowski (2000) illustrates this with the following four-step process. We may start by distractedly looking at a given car, noticing features like shape, color, details etc. Then our attention begins to zero in on a particular aspect: a dent on the side door, and thus starting a differentiation of the previously rather seamless perception into foreground (the dent) and background (the rest of the car). This is the transition to the next step where we establish a categorical object: “We now register the whole as containing the part. A relation between whole and part is articulated and registered” (Sokolowski 2000: 90). This is an act of reflective consciousness that makes possible the final step: “At this point we can declare, ‘This car is damaged’.”

Cognitive linguists will easily recognize several construal operations in this informal description, and thus support for a degree of continuity between perception, prelinguistic consciousness and language. No less important, however, are two distinctions. First the step when the categorical object [car-damaged] was established is distinct from the more “smooth” perception from the start. Second, making the linguistic statement on the basis of conventionalized construals as discussed above is distinct from all prelinguistic experiences. In phenomenology, we would say that the statement is “sedimented upon” these experiences. Sedimentation is more than a metaphor, as it describes the fundamentally layered nature of experience, with higher layers being more stable, abstract and dependent on social mediation (including the use of notations and external representations), but resting upon more bodily and analogue processes. The latter can be “uncovered”, or “relived” but are usually latent and not strictly necessary for appreciating meaning on the higher level (Blomberg and Zlatev 2014; Zlatev and Blomberg to appear).

It is arguably one of the main contributions of Cognitive Linguistics to have uncovered a plethora of such processes, for example those theorized under the headings of fictive motion (Talmy 2000) subjective motion (Langacker 2006), and most recently, mental simulation (Matlock 2010). The problem, however, is when (variable) pre-linguistic experiences are conflated with either conventional meanings (i. e., sedimented upon construals), hypothetical psycho-neurological processes, or both: “in understanding an FM [fictive, or non-actual motion]-sentence, people … construct a dynamic representation that mirrors the actual motion of the protagonist.” (Richardson and Matlock 2007: 238)

In contrast, based on a phenomenological re-analysis of different cognitive-linguistic models of “non-actual motion” or “dynamic construal” it is possible to distinguish at least three different kinds of pre-linguistic experiences (Blomberg and Zlatev, 2014) which alone or in combination appear to motivate the use of sentences such as (3–6) across different languages: (a) the enactive, action-oriented nature of perception, related to the notion of affordance, (b) the correlational act-object nature of intentionality, where consciousness can be redirected back from the object (“onstage”) toward act (“offstage”) and (c) the imagination of counter-factual scenarios, as in creative metaphor. It appears that Talmy’s account is most related to (a), Langacker’s to (b) and Matlock’s to (c).

While motivations such as these may explain why dynamic construal is common across different languages, the fact that there are strong cross-linguistic differences (Bohnemeyer 2010) is a stark reminder not to confuse the motivations with the sedimented upon conventions. For example, while (3–6) may be obtained as descriptions of the same situation, and serve as “translation equivalents” they are not fully synonymous. Non-actual motion is expressed in English (3) and Swedish (4) through the related verbs go and går, but while the English verb expresses generic motion, in Swedish it also conflates Manner, as it also translates ‘walk’. In line with expected differences in motion event typology (Blomberg 2015) 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.

A similar re-analysis could be extended to much in Cognitive Linguistics, including the key notions of image schemas and conceptual metaphors (Zlatev to appear). The (hypothetical) experiential structures of body-based gestalts and analogical mappings across “experiential domains” could possibly motivate linguistic meanings, but should not be conflated with them. For example, the conventional meanings of the prepositions/particles in and out are possibly grounded in a corporeal schema of CONTAINMENT, while those of from and to in the schema PATH, without being reducible to them, in line with the concept of sedimentation. Likewise, discourse metaphors (Zinken 2007) like those in (7) and (8) are the linguistic constructions themselves, not “underlying” cross-domain mappings or (other) motion-related experiences.

The notion of experiential grounding, quite actively discussed in the early days of Cognitive Linguistics (Lakoff and Johnson 1980), can be understood as the converse of sedimentation: X grounds Y if Y is sedimented upon X. What phenomenology helps with is to see both the continuity and discontinuity between X and Y. The major difference is that grounding structures are in general perceptual and analogue while the sedimented upon are signitive and categorical. On the other hand, it is not the case that the first are private while the second are social, as human embodiment is interlinked with intersubjectivity.

This has the important consequence of making the transition from bodily intersubjectivity to linguistic normativity easier to explain. As pointed out at the end of the previous section, the gap seems unbridgeable if it concerns two different “worlds”. But from the standpoint of phenomenology, linguistic conventions are not fully detached from their bodily roots, while the grounding experiences are not private “mental images” but intersubjective motivations.

4 Implications for methodology

While the issues discussed so far could be put aside by some as “too philosophical”, they have very clear implications for what everyone in Cognitive Linguistics seems to care about: proper scientific methodology. The dilemma is often framed as “to be empirical or to be introspective”?

If we understand empirical methods to refer to forms of research (like corpus linguistics, experimentation, and neurological modeling) that do not rely on introspection and intuition but that try to ground linguistic analysis on the firm basis of objective observation, then we can certainly witness a growing appeal to such empirical methods within Cognitive Linguistics.

(Geeraerts and Cuyckens 2007: 16)

Phenomenology can help us see that this is in fact a false dilemma. As we pointed out in the introduction, phenomenology aims to place science in perspective: objectivity is possible, but only on the basis of “the observations and experiences of individuals” and “a community of experiencing subjects”, as stated in the introduction. Everything that we know is given to us through experience, and the best we can do is to make the investigation of this experience and its intentional objects as systematic as possible. What are the intentional objects (i. e., what we aim to study) in the case of Cognitive Linguistics? If they were easily observable like physical objects we could isolate and count them, but are they? The material signs of sound, gesture or marks on paper are to some degree so, but they are literally meaningless without their “senses”. And these, we tentatively concluded, are the socially sedimented patterns of construal of signitive intentions: the meanings of dog and canine differ in how a type of intentional object is conventionally construed, and likewise for the meanings of the sentences (9) and (10): they denote the same perceptual situation, but construe it as two different kinds of categorical objects: one about the table and one about the carpet.

How could we even start on this path of analysis, if we had stayed on “the firm basis of objective observation”? We could not, but does that mean that we have to fully endorse introspection? Continuing the citation from Section 2, Talmy (2000: 4) seems to answer this in the affirmative: “As matters stand, the only instrumentality that can access the phenomenological content and structure of consciousness is that of introspection”. But this leaves many uneasy, since if introspection literally turns the direction of consciousness “inwards” towards what is “in the head” of the linguist, then the potential for intersubjective corroboration of one’s findings seems minimal. Itkonen (2008) therefore understandably distinguishes introspection from intuition, defining it as a process of consciousness that is (a) immediate, i. e., not the product of an argument, (b) categorical, i. e., providing rather discrete data and (c) having norms as its (intentional) object. Arguably, all speakers of all languages have such linguistic intuitions, on the bases of which they can distinguish correct (i. e., in accordance with communal conventions) from incorrect usage (Coseriu 2000; Zlatev 2008), though speakers may differ extensively in how “categorical” their intuitions are, and how easy it is to elicit them (Dąbrowska 2010).

However, this concept of linguistic intuition is somewhat restricted, corresponding to the meaning-as-use theory endorsed by Itkonen, as discussed in Section 2. As a litmus test, we can use it to help separate the grammatical from the ungrammatical, but it does not help us much in deciding what the meanings (as construals) are, why some are more wide-spread than others across languages and how they relate to pre-linguistic experience – the kind of questions that we have discussed so far.

I would again propose that phenomenology can help us find a dialectic synthesis: Talmy correctly places the aim on the “phenomenological content and structure of consciousness” but forgets that consciousness is intentionally directed, and primarily so toward the world than closed in within itself. Second, as pointed out repeatedly, the layers of prelinguistic and linguistic meaning (and consciousness) are not clearly distinguished in his analysis. Itkonen gets the directedness, but only towards the norm. What is the common denominator of the two accounts, and is at the same able to avoid their drawbacks: too much subjectivity for Talmy, and too much dependence on sociality for Itkonen? It is in fact the key concept in phenomenology, or the “method of methods”: the phenomenological reduction. This notion has almost been made mystical in some quarters, but informally, it means to attend not to what is immediately given, “the natural targets of our concern” (Sokolowski 2000: 49), but to look back, reflect on and contemplate the different kinds of intentional acts themselves. As we saw earlier, human consciousness is clearly capable of doing so, and what remains is to institutionalize this into a system, a science: “Thus, the ‘phenomenological reduction’ is simply the requirement always to abide by the sense of the proper investigation, and not to confuse epistemology with natural scientific (objectivistic) investigation.” (Husserl 1984: 410 [cited in Zahavi 2010: 7]).

To the extent that they reach valid results – and clearly they do – both Talmy and Itkonen apply special cases of the phenomenological reduction in their (meta)linguistic investigations, while differing on the level of analysis: the social-normative for Itkonen, and the perceptual-experiential for Talmy. The methodology and the conceptual apparatus of Langacker are even closer to phenomenology (Möttönen 2016) as can be seen when Langacker (this issue) justifies his qualitative analysis by appealing to descriptions based on “manifestations of well-established cognitive phenomena” generalizing over “a wide array of data from diverse languages”. However, not heeding the warning by Husserl in the quotation above, neither Talmy nor Langacker distinguish their qualitative explications of language and consciousness (“epistemological”), from natural scientific explanations (“objectivistic”), apparently assuming that with scientific progress their analyses will cash out in hard bio-physical facts: “the structures described in qualitative terms ultimately consist in neural processing” (Langacker this issue).

It is mistaken to make such a prediction, as meaning – on both pre-linguistic and linguistic levels – is a qualitatively different kind of phenomenon from biophysics. Their difference is not so much ontological as epistemological: meaning is, as argued, essentially amenable to study by phenomenological reduction. This is a prototypical first-person method, where the consciousness of the researcher reflects, rather than “introspects”. Brain states, isolated behavioural responses in controlled experiments, instances of signs in corpora are indeed amenable to study by detached observation and quantification. There is no reason to refrain from doing so, but it would be pure hubris to believe that such methods can be self-sufficient: what the “objective data” mean can only be made sense of by matching third-person, observational methods with first-person methods. An explicit research program where this is systematized into a procedure is that of neurophenomenology (Varela 1996). Finally, we need to remind ourselves of the social, intersubjective nature of all scientific enterprises, and especially those where what we investigate are (other) human beings and their cultures, rather than physical objects. Table 1 summarizes the three kinds of methods with respect to the dominant type of perspective that consciousness takes in each: toward itself, toward others and towards objects.

A hierarchy is here implied, so that higher-level methods are possible without lower-level ones, which is clearly testified in the history of linguistics: intuition-based and fieldwork methods have temporal precedence to third-person methods. The latter are necessary for scientific progress, but are never sufficient in themselves as they presuppose the intentionality of both the researcher and the “experimental subjects”. Both of these need to be made thematic, i. e., explicitly analyzed, in a proper investigation that concerns linguistic or non-linguistic meaning.

5 Conclusions

What I have suggested here is a kind of historical spiral in the evolution of Cognitive Linguistics. It started in the 1980s by highlighting the centrality of experience for language, and diversified along trajectories understanding this, and the closely related notion of cognition, in different ways. Having reviewed some of these differences, I have argued that methods and concepts from phenomenology can help re-unite various strands in an epistemologically pluralistic, truly experiential linguistics. My prediction would be that in the coming decades, this would merge with broader fields that likewise adopt a phenomenological perspective in the understanding of mind and meaning, such as enactive cognitive science (Thompson 2007) and cognitive semiotics (Zlatev 2015).