Category: Word Sensibility Theory

A Theoretical Look at the Role of Words for AI: Meta-Lexicography & Commonsense Ontologies

Mental Content

Mental Content


Liz, Manuel
 Incorporated contributions
Liz (8/11/09)
 Usage domain
mind, semantics, psicology, cognitive science
contenú de l’esprit
 German mentaler Inhalt
It is commonly assumed that mental states can be characterized by a certain psychological attitude and a certain content. The content of a mental state is a mental content. A precedent of that analysis can be found in Russell. Believing, desiring, remembering, feeling, perceiving, etc., are examples of psychological attitudes. What is believed, what is desired, what is remembered, what is felt, what is perceived, etc., are the mental content that in each case is associated with those attitudes.

Very often, it is also assumed that there are two big classes of mental contents: conceptual and non-conceptual ones. Conceptual content is the semantic content that we can find in words, expressions and sentences of a language. The content that beliefs, desires, rememberings, etc., typically have is the same as the content of certain sentences. Mental states with conceptual content are also called “propositional attitudes”, their content being a particular proposition that may be expressed by a certain sentence.

Non-conceptual content is a experiential, qualitative or phenomenological content. It is the alleged content that feelings, perceptions and sensations typically have. Whereas conceptual content is semantically evaluable in a quite direct way, non-conceptual content is not so. However, non-conceptual content can be evaluated as more or less correct or incorrect, or as more or less adequate or inadequate, etc. Mental states with non-conceptual content are usually called “qualitative states”, “experiential states” or “phenomenal states”. Their content eventually is a qualitative, experiential, or phenomenal character not identifiable with any proposition.

A very important thesis with respect to the distinction between conceptual and non-conceptual content is that perhaps there are mental states with both conceptual content and non-conceptual content. Another not least important thesis is that perhaps every mental state with conceptual content has also some kind of non-conceptual content.

The contrast between internism and externism has given place to one of the more dramatic discussions about mental content in recent years. Internism claims that mental contents –and mental states– only depend on factors internal to the mind of the subjects. Externism claims that mental contents –and mental states– essentially depend on factors external to their minds. Those external factors may include linguistic norms of the community, how experts would use certain terms, and the relationships with the external world. Descartes and Frege are two paradigmatic classical examples of internism. Externism was introduced by authors like Putnam, Burge and Kripke.


  • BURGE, T. (1992) “Philosophy of Language and Mind 1950-1990”, Philosophical Review, 95. 
  • CRANE, T. (1992) The Contents of Experience, Cambridge, Cambridge Univ. Press.
  • DRETSKE, F. (1980) Knowledge and the Flow of Information, Cambridge, MIT Press.
  • ——– (1988) Explaining Behavior, Cambridge, MIT Press.
  • FODOR, J. (1987) Psychosemantics, The Problem of Meaning in the Philosophy of Mind, Cambridge, MIT Press.
  • KRIPKE, S. (1980) Naming and Necessity, Cambridge, Harvard Univ. Press.
  • LYCAN, W, (ed.) Mind and Cognition, Cambridge, Basil Blackwell.
  • MCGINN, C. (1989) Mental Content, New York, Blackwell.
  • MILLIKAN, R. (1984) Language, Thought and Other Biological Cathegories, Cambridge, MIT Press.
  • PETTIT, Ph. & J. McDowell (eds.) (1986) Subject, Thought and Content, Oxford, Clarendon Press.
  • PUTNAM, H. (1975) “The meaning of ‘meaning’ ”, inMind, Language and Reality, Cambridge, Cambridge Univ. Press.
  • ——– (1988) Representation and Reality, Cambridge, MIT Press.
  • RICHARD, M. (1990) Propositional Attitudes, Cambridge, Cambridge Univ. Press.
  • STALNAKER, R. (19999 Context and Content, Oxford, Oxford Univ. Press.
  • WOODFIELD, A. (ed.) (1982) Thought and Object. Essays on Intentionality. New York, Oxford Univ. Press.

The bibliographic resources offered by David Chalmers in his website are extremely useful: <>

Intentionality & Representation

Does Separating Intentionality From Mental Representation Imply Radical Enactivism?

  • Institute of Philosophy II, Ruhr-Universität Bochum, Bochum, Germany

Traditionally, intentionality is regarded as that feature of all and only mental states – paradigmatically beliefs and desires – in virtue of which they are directed at or are about something. The problem of intentionality is to explain how it fits into the natural order given the intuition that no physical entity can be intentionally directed in this sense. The basic assumption of this paper, proposed by enactivists, is that failure to naturalize intentionality and mental representation is partly due to the fact that most participants in the debate take intentionality and mental representation to be equivalent. In contrast, it is proposed to treat intentionality as a feature of whole embodied agents (paradigmatically organisms) who can be directed at objects and states of affairs in various ways, while representation should be regarded as a feature of mental states (and their respective vehicles or underlying mechanisms). The present paper develops and motivates the distinction, applies it to Metzinger’s project of naturalizing phenomenal representation, and demonstrates the range of theoretical options with respect to a delineation of cognition given the enactive proposal. It is taken as problematic that enactivism takes the realm of cognition to be identical to the realm of biology. Instead, a constraint on a theory of intentionality and representation is that it should delineate the subject matter of cognitive science and distinguish it from other sciences, also to leave room for the possibility of artificial intelligence. One important implication of the present proposal is that there can be creatures which can be intentionally directed without having the capacity to represent. That is, their intentionality is restricted to being able to be directed at existent things. Only creatures in possession of the right kind of neurocognitive architecture can produce and sustain representations in order to be directed at non-existent things. It is sketched how this approach conceives of intentionality as a developmental and layered concept, allowing for a hierarchical model of varieties of intentionality, ranging from the basic pursuit of local environmental goals to thoughts about fictional objects.

Intentionality and Representation in Cognitive Science

What is the most promising approach to cognitive phenomena within a naturalistic framework? Is there a mark of the cognitive, i.e., a feature that all and only cognitive phenomena share? Can all cognitive phenomena be explained by one and the same approach? This paper considers various possibilities of how intentionality, one of the most important features of the mind, can be naturalized, and thus provide a foundation for our most sophisticated form of intentionality, i.e., phenomenal intentionality involving mental representation.

Until recently, cognitive science was based on the undisputed foundation that cognition must be explained in terms of mental representation and computation. For a long time, the dominating theoretical framework of Cognitivism conceived of cognition as information processing along the lines of digital computers, in particular as constituted by syntactically driven manipulations of representational structures in the brain that are “sandwiched” (Hurley, 1998) between sensory inputs and motor outputs (e.g., Fodor, 1975Pylyshyn, 1984). Thus, when I look at the coffee mug in front of me, sensory information hitting my retina is processed in a piecemeal fashion in specialized modules that eventually produce a detailed three-dimensional image of the mug that can in turn guide possible actions like grasping it. The parallel distributed processing movement presented a departure from the simple computer metaphor by modeling cognition using connectionist networks inspired by the architecture of the brain (Smolensky, 1988Churchland and Sejnowski, 1992). However, even though such artificial neural networks process input subsymbolically, most successful models of this kind still rely on intermediate states which allow for storing, manipulating, and transforming information (about the mug, say) before producing the output (O’Brien and Opie, 2004).

More recently, however, proponents of Enactive and Embodied approaches to cognition challenged both this representationalist framework and its explicit separation of perception and action in favor of a robustly dynamic approach to cognition. Cognition is now conceived as primarily a bodily activity of a whole organism (or more generally, embodied agent) that can be explained without appeal to mental representations (Varela et al., 1991Noë, 2004Chemero, 2009Froese and Di Paolo, 2011Hutto and Myin, 2013Gallagher, 2017). Perceiving the coffee mug not only requires multiple actions like eye-, head-, and body-movements (gaze turning, etc.), perceiving is in the service of detecting action possibilities (like grasping, say) from the start. Enactive accounts come in different varieties, but they all share many assumptions regarding the nature of perception and cognition, apart from rejecting a representational approach, and can be seen as differing mainly with respect to different aspects that are highlighted by them. Proponents of Sensorimotor Enactivism emphasize the action-involving character of perception (e.g., Noë, 2004), while the focus of Autopoietic Enactivism is on the self-organizing and autopoietic character of basic intentionality; this amounts to claiming a continuity of life and mind – phenomena allegedly sharing the same basic organizational features (e.g., Thompson, 2007). Radical enactivists, like Hutto and Myin (20132017), propose to separate intentionality from mental representation and consider basic cognitive capacities as merely intentionally directed without them being representational or content-involving. This is a bold move that introduces a distinction between contentless and content-involving cognition. Although there is much more to be said about all these variants, this brief sketch shall suffice for now, since we will return to some of these claims below.

Finally, proponents of so-called Predictive Processing models of the brain claim to have found “the first truly unifying account of perception, cognition and action” (Clark, 2016, p. 2) by conceiving of the brain as a prediction machine, which is constantly testing hypotheses about the incoming sensory stimulation based on a hierarchical generative model that is constantly updated based on prediction errors signaled by forward neural processing (Friston, 2010Hohwy, 2013Metzinger and Wiese, 2016). On this view, when I perceive the mug, brain processes have already formed a set of top-down expectations or predictions about the incoming sensory information based on the most likely hypothesis given prior knowledge about certain parameters of the situation. These expectations are then matched against the actual sensory input resulting in bottom-up neural signals. Deviations from the prediction constitute prediction errors and result in the relevant update of the brain’s generative model of the situation. The overarching goal (and unifying principle) of the brain’s perceptual and cognitive activities is to minimize prediction error. Given the promise that the Bayesian prediction machine hypothesis could yield a unifying framework for cognition, it is not surprising that both representationalists and enactivists engage in a hot debate regarding whether this framework should be interpreted in representationalist (Hohwy, 2013Clark, 2015) or enactivist terms (Hutto and Myin, 2017).

These roughly three frameworks portray cognitive phenomena quite differently since they suggest explanations using different explanatory tools. One of the central questions that this debate gives rise to is whether cognition is in general representational or content-involving and whether any adequate theory of cognitive phenomena must invoke mental representations. It has been an almost undisputed assumption in cognitive science that intentionality and representation can be used synonymously (or are at least equivalent). Most publications on naturalizing intentionality have thus used these terms interchangeably. To give two random examples, Crane (2003, p. 30) says explicitly that “philosophers have a word for the representational nature of states of mind: the call it ‘intentionality’.” Similarly, Searle (1983, p. 4) holds that “intentional states represent objects and states of affairs.” Since representations have content, they also have accuracy conditions. By contrast, echoing points made by Dreyfus (2002)Hutto and Myin (2017, p. 95) propose a “radical enactivism,” according to which we should “think of the most primitive form of intentionality […] in non-contentful, non-representational ways” but still as an “attitude directed toward an object.” Intentionality is then no longer a feature of contentful mental states that represent but “an attitude of the whole organism expressed in their behavior” (Hutto, 2008, p. 57). The idea is to disambiguate Brentano’s initial characterization of intentionality as directedness and aboutness, reserving the latter for those forms of cognition that depend on socio-cultural capacities and cultural symbol-systems, such that contentful representations with satisfaction conditions only appear on the scene in the wake of linguistic capacities.1 This move leads Hutto and Myin to the even more radical claim, however, that all basic cognitive capacities, in particular perception and action, but also forms of imagination and memory, can be exhaustively explained without the assumption of mental representations and content.

Thus, one could express this radical departure from the tradition by formulating a set of theses that express the traditional view in cognitive science, i.e., the Equivalence Thesis and the Separation Thesis:

Equivalence Thesis (ET)

Intentionality and mental representation are equivalent and co-extensive notions referring to the directedness or aboutness of mental states.

Separation Thesis (ST)

Intentionality and representation can be investigated and accounted for independently of consciousness.

The separation thesis is not a target of this paper, although it is part and parcel of classic cognitive science until today. In line with proponents of Radical Enactivism, the present paper approaches the naturalization of intentionality by rejecting the Equivalence Thesis which can be partly blamed for lack of progress and agreement in this area. As will become clear in a moment, this crucial conceptual move opens up new avenues to the naturalization of intentionality independently of the naturalization of mental representation. But this paper considers a systematic position that is only related, yet importantly different from Radical Enactivism. Intentionality and representation should be treated separately, in agreement with Radical Enactivism. But given that it is questionable whether Radical Enactivism can yield a promising and useful framework for cognitive science, it will be explored whether their rejection of mental representation and content may be the only way of dealing with the separation of intentionality and mental representation. In his review of Hutto and Myin’s recent book, Thompson (2018) observes that “they insist that content entails correctness conditions. They also apparently think that only representations can have correctness conditions. So, any form of intentionality (cognitive directedness) that lacks correctness conditions or is not representational is said to be contentless.” By considering a phenomenological notion of content that is not propositional and thus does not imply correctness conditions, Thompson points at alternatives neither explored nor discussed by Hutto and Myin and other radical enactivists. By pointing at some shortcomings of the radical enactive proposal, this paper discusses different systematic alternatives that consider other routes after separating intentionality from mental representation.

The guiding questions of this paper have been formulated at the outset: They concern the task of identifying a mark of the cognitive. The core intuition supporting the rejection of the Equivalence Thesis is that while intentionality (as directedness) should be conceived as a feature of whole embodied agents, mental representation is a feature of the mechanisms sustaining cognitive states and processes. Consequently, the reason why intentionality and representation should be distinguished is not that (basic) cognition is non-representational, as enactivists would have it. The reason is that these different notions denote features of different entities. It is organisms or agents, not mental states that are directed. And it is mental states (and their underlying mechanisms), not organisms or agents, which represent. Consequently, an organism or agent can be intentionally directed at something without representing it. And if an organism is in possession of mechanisms (and thus mental states and processes) that can represent, then this can enhance its range of possibilities to be directed at objects and states of affairs. For example, representational mechanisms can enable an organism to be directed at fictional objects via the imagination. But importantly, there can be simple creatures that are intentionally directed at something without thereby representing it. This proposal must be spelled out in more detail but it gives rise to a set of further interesting questions that should be explored: For example, (1) Is non-representational intentionality cognitive or not? It may well be that cognition merely constitutes a special case of intentionality that is enabled by mental representations. Then the range of intentionality would be bigger than the range of cognitive phenomena. This would be in stark contrast to the thesis of Radical Enactivism. Choosing this option would leave mental representation as a candidate for delineating the cognitive realm and avoid the consequence that all forms of intentional directedness in nature would have to be considered as cognitive. (2) If non-representational intentionality constitutes a form or forms of cognition after all, then what is the mark of the cognitive and how do these forms of cognition relate to the more sophisticated forms that involve mental representation? (3) How does the most sophisticated form of intentionality of conscious thought, i.e., the one that is phenomenally experienced and produces intensional contexts, related to basic intentionality that is neither phenomenal nor representational? Does it delineate a special domain within the realm of the cognitive? Or is it rather the source of all intentionality, as various philosophers, though rarely any cognitive scientists, maintain (e.g., Searle, 1992Pitt, 2004Strawson, 2004Mendelovici, 2018)?

Given the thematic focus of this Special Issue, this paper situates the present discussion in relation to Thomas Metzinger’s work in order to demonstrate options for providing a naturalistic foundation for his theory of a subclass of mental representation, namely phenomenal representation. The rest of the paper proceeds as follows: By way of setting the stage, the following Section “Phenomenal Representation” outlines Metzinger’s approach to representation and mental representation and highlights those points where the present conceptual proposal can usefully complement his account. Section “The Homeostatic Basis of Phenomenal Subjectivity” then turns to a discussion of the role of the organism’s body and central nervous system for sustaining phenomenal representation, while Section “Self-Organization and Basic Biological Intentionality” sketches how this naturalistic grounding of phenomenal representation in the body can be extended to a more general naturalistic account of intentionality (that is independent of mental representation) by relying on enactivist ideas, while discussing conceptual routes that are quite different from enactivism.

Phenomenal Representation

Many philosophers of mind take it for granted, indeed to be an “obvious fact that biological nervous systems are able to generate representations of the world and its causal matrix by forming internal states which then function as internal representations of this causal matrix” (Metzinger, 2003, p. 18). This is the attitude, expressed paradigmatically by Searle (1992), that biological phenomena cause mental phenomena. Functionalists do not question this claim, but typically take the more liberal attitude of allowing for artificial systems to be capable of generating the same types of mental phenomena using other means, downplaying the biological details needed for causing mental phenomena. In his monumental work Being No OneMetzinger’s (2003) project is not to demonstrate how intentionality tout court can be naturalized. He explicitly states that “intentionality as such is not an epistemic target” of his representationalist approach to the first-person perspective. He develops his theory of subjectivity in terms of a phenomenal self-model and a phenomenal model of the intentionality relation, and this theory takes for granted that organisms are capable of representing. It is explicitly not his “goal to offer a general theory of mental representation” (Metzinger, 2003, p. 595). Therefore, for his theory to be thoroughly naturalistic, it presupposes some naturalistic explanation of representation and intentionality in general, while his goal is merely to demonstrate how his theory can account for the phenomenal experience of intentionality, i.e., phenomenal content. His crucial philosophical step “consists in phenomenalizing intentionality” which may be a “necessary detour […] in the project of naturalizing intentionality tout court.” (Metzinger, 2003, p. 414) Thus, his restricted claim is that

“the phenomenal experience of being an intentional agent, of being a perceiving, attending, and cognizing subject, can be naturalized. Of course, this in no way precludes the possibility that intentional content as such can never, and maybe even for principled reasons, be naturalized. But getting the first obstacle out of the way may greatly help in gaining fresh access to intentionality as such […]. We can separate the issue of consciously experienced intentionality from the more general problem of how something like representational content could evolve in the minds of human beings and other animals at all.” (Metzinger, 2003, p. 414)

Note that the last part of this quote indicates an acceptance of both the Equivalence Thesis and the Separation Thesis, since he argues that conscious or phenomenal intentionality is to be separated from intentionality as such which is then identified with representational content. Thus, the aim of his theory is to address the question how the consciously experienced arrow of intentionality can be explained in terms of a transparent model of the intentionality relation, i.e., a model that cannot itself be recognized by the subject as a model. And although he does not provide a naturalistic theory of intentionality as such, he nevertheless speculates that “the ‘real’ intentionality relation” may be “constituted by an active cognitive agent interacting with its environment” (Metzinger, 2003, p. 113). Indeed, he considers the possibility that “intentionality” may perhaps be anchored “[…] on a prerational level, probably starting with the motor system and early levels of attentional processing […]” (Metzinger, 2003, p. 414). These brief comments suggest that Metzinger could agree in principle with some enactivist claims, e.g., the importance of an embodied agent interacting in various ways with their environment in order to constitute an intentionality relation. Indeed, his acceptance of the importance of the aspect of embodiment is emphasized in Metzinger (2014).

Since Metzinger (2003, p. 415) is “mute about the question whether anything like ‘real’ intentionality exists” or what a naturalistic explanation of it looks like, it is the aim of this present paper to use the proposed broader framework in order to take some steps in the direction of a naturalistic grounding of his theory of phenomenal intentionality. Metzinger (2003, p. 15) is a friend of useful distinctions: As he uses the term, “mental representation is a process by which some biosystems generate an internal depiction of parts of reality.” Such internal representations can be experienced and used by the respective system in order to guide its flexible behavior. In contrast to genuinely mental representations, lots of information processing that has to do with the regulation of heart rate or immune system parameters, can “certainly carry information, but this information is not mental information” (Metzinger, 2003, p. 17). That is, although it tracks states of affairs pertaining to the internal milieu of the body and thus concerns the organism, it may only affect mental phenomena but it is not already identical to the mental phenomena in question. According to Metzinger, such representations do not count as mental representations because they cannot become conscious. They are internal representations in a “purely physical sense” (ibid.). In contrast, mental representations “can, at least in principle, possess a phenomenal kind of ‘inwardness”’ (Metzinger, 2003, p. 18). Thus, Metzinger distinguishes, importantly, between representations and mental representations, while he stipulates that the latter are separated from the former by their phenomenal features. Arguably, if the fact that heart rate and the immune system carry information is sufficient for them to count as representations, then representations abound in nature. Even the infamous tree rings, which carry (or embody) information about the age of the tree, then count as representations. But mental representations do not permeate nature so that we need to draw distinctions anyway. The question is where to draw the line and which criteria we should use to draw explanatorily useful boundaries. The relationship between intentionality and mental representation is thus complicated by the fact that mental representations only form a subclass of representations in general, whether or not we take phenomenality to be the correct criterion or not. Given that representations, merely understood as entities indicating information, (Dretske, 1981Rupert, 2018), seem to abound in nature and are thus insufficient to characterize cognition, the question arises whether intentionality at least coincides with representation if not with mental representation. The conceptual move proposed in this paper implies the following tasks: (1) We must characterize intentionality independently of representation and demonstrate how it can be conceived as a natural phenomenon. (2) We must provide criteria for distinguishing (mere) representation from genuine mental representation in order to capture the most distinctive human mental capacities like conscious thought, imagination and so on. Of course, this twofold task cannot be achieved in this single paper. But it is an important task since it forces us to delineate the cognitive from the non-cognitive realm. Identifying a mark of the cognitive is no trivial task but it has important ramifications for cognitive science and its subject matter, as we will see below.

In order to supplement Metzinger’s approach to phenomenal representation by providing a naturalistic theory of intentionality and representation, the first thing to do is to develop a naturalistic theory of intentionality as such. In this vein, the present paper applies the enactivist proposal of separating intentionality from representation and mental representation to this effect, yielding a much more differentiated framework. In addition to Metzinger’s distinction between representation and mental representation, the central claim that mental representation is a feature of mental states, while intentionality is a feature of whole embodied agents – be they natural (organisms) or artificial (robots) – yields the further important distinction between intentionality and representation. As mentioned at the outset, separating intentionality from representation in this way borrows from the recent development of Enactivism, yet, by retaining the notion of mental representation for cognitive phenomena, it differs from this radical position.

The Homeostatic Basis of Phenomenal Subjectivity

I have quoted quite extensively from Metzinger’s major book Being No One above, partly because these quote illustrate how he takes intentionality and mental representation to be equivalent, while considering kinds of representation that are not mental. Interestingly, Metzinger seems to reserve the category of mental representations for those that can in principle be phenomenal. This is in line with Searle’s (1992) Connection Principle, which considers only those representations as mental that can potentially become conscious. Note that any attempt to explain conscious intentionality in terms of a representational theory presupposes accepting the separation thesis since if representation wasn’t taken to be independent of consciousness, then this project of formulating a representationalist theory of consciousness would be doomed to fail. Thus, Metzinger must subscribe to the Separation Thesis. Likewise, the quote from the foregoing section demonstrate that Metzinger also subscribes to the Equivalence Thesis. He often uses intentionality and mental representation interchangeably, for example, when he speaks of ‘intentional content’ and suggests that his approach could yield progress regarding the task of explaining ‘intentionality as such’ which he seems to identify – in the same sentence – with the task of showing how ‘representational content’ could evolve in animals.

Metzinger draws heavily on the work of Damasio (19992011) in his naturalistic account of phenomenal subjectivity. This is noteworthy since Damasio’s reliance on the organism’s body, homeostasis and processes of self-organization is a specific instance of a more general approach to the origins of intentionality to be outlined in the following section. Like Metzinger, Damasio is concerned with the sense of self, i.e., the subjectivity of conscious experience. While Metzinger approaches this issue from a conceptual point of view by developing his theory of self-models, Damasio is interested in supplying the underlying machinery responsible for generating and sustaining subjectivity. Thus, their approaches can mutually complement each other since, as Metzinger puts it, the common target is how an organism can

“feel itself as itself. As a physical system continuously engaged in the process of self-organization and self-regulation, the organism has to maintain a robust functional boundary with its environment and keep a large number of internal parameters stable and invariant. A large section of self-presentational content precisely generates a continuous flow of information about the degree of invariance that is currently being achieved.” (Metzinger, 2003, p. 344)

As far as the stability and invariance that a conscious subject experiences are concerned, the body is the obvious candidate for providing the conditions under which a stable reference point of conscious experiences can be provided and sustained. To this effect, Metzinger praises Damasio’s theory of consciousness and self since it highlights “the way in which this flow is very likely rooted in elementary bioregulatory processes, those processes concerned with keeping the internal chemical milieu of the body stable in a continuously changing environment.” (Metzinger, 2003, p. 345)

All intentionality involves a self-other distinction, since understood as directedness it implies that something reaches beyond itself, transcends itself. As far as Metzinger’s target is concerned, the subjectivity of phenomenal representation, Damasio argues that the experienced, yet elusive sense of self is grounded in a biological process realized by a neural mechanism distributed over a cluster of connected brain structures (Damasio, 2011, p. 8f). These biological processes are responsible for the regulation of the whole organism’s wellbeing. In order to maintain its identity and to ensure survival, the organism’s overall homeostatic state must remain within certain bounds. One of the brain’s chief tasks is to monitor and regulate bodily processes on the basis of information that it receives from the body at any given point in time. Because of their monitoring and regulating function with respect to this goal, Damasio calls these brain structures the unconscious biological ‘proto-self.’ This is the analog of Metzinger’s notion of “mental or merely internal (i.e., necessarily non-phenomenal) self-presentation” (Metzinger, 2003, p. 345). For Damasio, it is one of the most important ideas of his framework “that the body is a foundation of the conscious mind” in the sense that these “proto-self structures are not merely about the body. They are literally and inextricably attached to the body.” That is, although Damasio’s work is mainly concerned with determining the brain structures sustaining subjective consciousness, these structures only possess the function that they have in virtue of the body and receiving feedback from the body. They are only considered as representing the body because they concern the body. In his framework, these structures ultimately give rise to the phenomenally experienced sense of self, the ‘core self.’ In his attempt to provide such a biological grounding for the sense of self, Damasio (2011, p. 48) emphasizes the connection between “organisms,” “purposes” or “biological needs,” and “value.” Since all organisms must maintain their physiological state “within an optimal homeostatic range,” “management operations” such as “procuring energy, incorporating and transforming energy products…aim at maintaining the chemical parameters of a body’s interior (its internal milieu) within the magic range compatible with life.” This homeostatic process of maintaining the right chemical balance within the body is tied to the organism’s need (or goal) of survival. This notion of need is in turn tied to the notion of a biological value:

“I see value as indelibly tied to need, and need as tied to life. The valuations we establish in everyday social and cultural activities have a direct or indirect connection with homeostasis. That connection explains why human brain circuitry has been so extravagantly dedicated to the prediction and detection of gains and losses, not to mention the promotion of gains and the fear of losses.” (Damasio, 2011, p. 47f)

In this passage, Damasio suggests a direct lineage starting from the most basic intentional activities of the simplest living creatures to our social and cultural activities, connected via the relation between value, (biological) needs, and life, over various stages of complexity. Dependent on the complexity of the organism, different values are assigned to its physical environment. In organisms with brains, such as human beings, the homeostatic process is monitored and regulated by neurons, peculiar kinds of cells that possess the ability to influence other cells, based on inputs that inform the neurons about the state of the body at all times. If necessary, neurons initiate the release of chemical molecules to reestablish the homeostatic balance. Damasio interprets this in representational terms:

“They (the neurons) end up representing the state of the body, literally mapping the body for which they work and constituting a sort of virtual surrogate of it, a neural double […] In brief, neurons are about the body, and this ‘aboutness,’ this relentless pointing to the body, is the defining trait of neurons, neuron circuits, and brains.” (Damasio, 2011, p. 38f)

According to Damasio, neurons and the bodily state that they represent also play an important role in our directedness toward – and representation of – the outside world and its objects.

“[…] when the brain maps the world external to the body, it does so thanks to the mediation of the body. When the body interacts with its environment, changes occur in the body’s sensory organs, such as the eyes, ears and skin; the brain maps those changes, and thus the world outside the body indirectly acquires some form of representation within the brain.” (Damasio, 2011, p. 39)

All this suggests that at least organisms with brains and central nervous systems are capable of producing representations and, more specifically, mental representations, where the latter form the subgroup of phenomenally experienced representations, in Metzinger’s terminology.2 Moreover, what we have arrived at is a story that indicates the importance of the organism’s body, indeed of the whole organism, for mental phenomena. But both Metzinger’s and Damasio’s accounts are focused on the question which features of the brain constitute the basis of our phenomenal sense of self. I agree with Metzinger’s assessment that this rough story, indicated by Damasio, makes a lot of sense. And given that it points toward claims that are associated with proponents of embodied cognition, this story also raises further questions: For example, given that Metzinger and Damasio focus on sophisticated mental phenomena, what about a naturalistic account of intentionality and representation as such? Do all organisms exhibit intentionality and why? And does this mean, given the equivalence thesis, that all organisms represent in virtue of exhibiting intentionality? If not, then where and why should we draw boundaries? Can we formulate criteria that inform us about which instances of intentionality should count as cases of cognition? Thus, the question I want to address now is how we can make headway toward providing this general naturalistic account, extending Damasio’s theory by relying on Autopoietic Enactivism.

Self-Organization and Basic Biological Intentionality

Brentano (1874/1995) famously claimed that intentionality is the mark of the mental: All and only mental phenomena are intentional, no physical phenomenon exhibits intentionality. This gave rise to the project of naturalizing intentionality, i.e., of demonstrating that Brentano was wrong in thinking that no (broadly) physical entity can be intentional. Fodor (1987, p. 97) famously thought that this would have to be a reductive project: He suggests that once physicists will have completed the “catalog of the ultimate and irreducible properties of things,” then “the likes of spincharm, and charge will perhaps appear on their list. But […] intentionality […] doesn’t go that deep […] If aboutness is real, it must be really something else.” Again, note that Fodor takes intentionality and mental representation, paraphrased as aboutness, to be equivalent here. But explaining intentionality in naturalistically acceptable terms, i.e., “in non-intentional, non-semantical, non-teleological, and in general, non-question-begging vocabulary” (Fodor, 1987, p. 126) turned out to be a rather difficult project, and we have not reached anything like a consensus as to how to fit the phenomena under the umbrella term “intentionality” in a naturalistic view.

The move of separating intentionality from mental representation in a way parallels the one proposed by Metzinger. But rather than singling out a significant case of representation – conscious phenomenal mental representation – like Metzinger does, the present paper highlights a distinction between a property that applies on the level of whole embodied agents – intentionality – and a property that applies on the level of individual mental states (and their realizing mechanisms) – representation. While there can be intentionality without mental representation, most likely in simple creatures, those agents which are in possession of representational mechanisms will enjoy a much broader range of possibilities to be directed at objects and states of affairs. Most importantly, such agents3 may be directed at fictional objects like centaurs or Santa Claus, because of their capacities of thought and imagination which are (most likely) realized by representational mechanisms in the brain. Simpler creatures lacking these mechanisms and capacities may be restricted in the sense of only being capable to be directed at things that exist. Whether this is the case depends on the criteria we formulate for such sophisticated directedness, i.e., the directedness that many philosophers take Brentano (1874/1995) to have been concerned with in his original formulation of the intentionality thesis. For example, Morgan and Piccinini (2017) discuss the inadequacy of so-called tracking theories to explain the “distinctively mentalistic phenomenon of directedness toward entities that may not exist that poses the central puzzle of intentionality.” Tracking theories typically posit some tracking relation in virtue of which some mechanism picks out – or represents – a feature or object in the world. Against this move, Morgan and Piccinini (2017) argue that to explain directedness toward a centaur it is insufficient to “posit neurons whose function is tracking centaurs.” Such neurons would in fact never fulfill their function since there are no such things and it is difficult to justify why a mechanism with that function should have evolved. Among the entities that exhibit intentionality, we must formulate criteria in order to delineate the set of entities that are capable of this sophisticated kind of directedness. One route may be to require a certain architecture or functional organization which in turn can only be supplied by certain networks like nervous systems (Rupert, 2018). But there are obviously alternative routes since this is ultimately an empirical issue. Sadly, this issue is also beyond the limited scope of this paper which will focus on the other end of the scale, namely the origins of intentionality.

Damasio seems to suggest that abstracting from the paradigm case of a human being, it is only a question of scale to accept a single cell as the bearer of intentionality, since the cell shares the same basic organization as a complex organism such as a human being. Damasio points to these structural analogies between complex organisms such as human beings and simple organisms such as cells:

“In many respects a single cell is a preview of what a single organism such as ours would come to be. One can see it as a sort of cartooned abstraction of what we are. The cytoskeleton is the scaffolding frame of the body proper, just as the bone skeleton is in all of us. The cytoplasm corresponds to the interior of the body proper with all its organs. The nucleus is the equivalent of the brain. The cell membrane is the equivalent of the skin. Some of these cells even have the equivalent of limbs, cilia, whose concerted movements allow them to swim.” (Damasio, 1999, p. 33)

But the problem is that if one accepts the Equivalence Thesis, like Metzinger and many others, then ascribing intentionality to single cells forces us to also grant that they are capable of representing. Whether they also possess the capacity of mental, i.e., phenomenal representation, would be a further issue. By contrast, rejecting the Equivalence Thesis allows for an ascription of intentionality without granting representational capacities. The main point of this section is that if one accepts, following all enactivists, that autopoiesis (self-organization and self-production) is the defining characteristic of organisms and if one is willing to apply intentionality to whole organisms, it would be arbitrary to accept its application with respect to humans and other higher animals but not to cells and plants, given that they are all organisms.4 One can already envisage questions concerning the commonalities and differences between putative cognitive capacities in cells, plants, animals, and humans. We will return to this below.

Autopoiesis and Nano-Intentionality

Some philosophers have suggested that tracing back the natural origins of intentionality may not lead us all the way down to the level of physical particles (pace Fodor), but at least to the level of biological self-organization. Maturana and Varela (1980) introduced the idea that living organisms are defined by the feature of self-organization or “autopoiesis.” Weber and Varela (2002) refer explicitly to Kant’s (1790/1998) discussion of organisms as natural purposes and the continuing discussion by Jonas (1966). The advantage of this view is that it yields, according to Thompson’s (2007, p. 159) Autopoietic Enactivism, “an explicit hypothesis about the natural roots of intentionality: On this conception, intentionality arises from the operational closure and interactive dynamics of autopoiesis,” and is thus grounded in the structural organization and biological autonomy of living organisms, namely self-organization. Intentionality is conceived as a basic feature of an organism’s embodied interactions with the environment, not as a feature of mental states (ibid., 25). This view can be seen as a descendant, indeed a naturalized version, of Kant’s definition of an organism as a “natural purpose,” developed in his Critique of the Power of Judgment (Kant, 1790/1998, pp. 65–66). A natural purpose is a system in which the parts of the system (1) are only possible through their relation to the system as a whole, and in which (2) the parts of the system, moreover, are “combined into a whole by being reciprocally the cause and effect of their form.” That is, “in such a product of nature each part is conceived as if it exists only through all the others, thus as if it exists for the sake of the others and on account of the whole, i.e., as an instrument (an organ) […]” (Kant, 1790/1998, p. 373). Unlike an artifact, such as a watch, an organism is – simply in virtue of being a natural purpose – not caused by any external rational agent, such as a watchmaker, but by its own formative powers. A popular example illustrating the peculiar feature of self-organization characteristic of organisms – alluded to by Kant – is an organism’s ability to repair itself in response to damage to the body. At the time, Kant seems to have been aware of Abraham Trembley’s discovery that after cutting them in two halves, hydra – which are multicellular organisms found in unpolluted fresh waters – regenerate by developing two complete organisms. As Fitch (2008) observes, many organisms like salamanders and zebra fish can regrow entire body parts like lost limbs. This astonishing fact raises the question how it is possible for these animals (or their parts anyway) to “know” what they should grow, i.e., what the overall animal is supposed to be like in order to supplement what’s left with what’s missing. Thompson dubs this process “circular causality,” a combination of local-to-global determination whereby emergent structures and properties on the macro-level are generated and sustained by the behavior of the components on the micro-level, and “global-to-local determination whereby global structures and processes constrain local interactions” (Thompson, 2007, p. 62). He discusses it in the context of neurodynamics:

“Coherent and ordered global behaviors, which are described by collective variables or order parameters, constrain or govern the behavior of the individual components, entraining them so that they no longer have the same behavioral alternatives open to them as they would if they were not interdependently woven into the coherent and ordered global pattern. At the same time, the behavior of the components generates and sustains the global order. This two-sided or double determination is known as circular causality […]” (Thompson, 2007, p. 62)

Understood in this way, teleology or circular causality is not opposed to causality but introduces a differentiation into the notion of causation in terms of a two-sided dependency. Kant’s notion of a natural purpose delineates the group of those entities which are at the same time products of nature and which necessarily have to be understood teleologically, as being intrinsically directed toward some purpose or goal. Varela recognized that his notion of autopoiesis has an important precursor in Kant’s original discussion of self-organization (Weber and Varela, 2002). But while Kant held teleological descriptions as providing merely an indispensable heuristics rather than an objective explanation (Kant, 1790/1998, p. 389), Varela and Thompson both argue that a modern empirical theory of life (based on the theory of autopoiesis and dynamical systems theory) can be seen as a naturalized version of Kant’s notion of a natural purpose, providing us with a “non-reductionist yet ‘hard’ explanation of the living” (Weber and Varela, 2002, p. 102). According to this modern understanding, the organism is conceived of as a “creator of ‘real teleology”’: […] organisms are subjects having purposes according to values encountered in the making of their living” (ibid.). In the present context of the elaboration of the basic biological intentionality of organisms, it is crucial that such biosystems are in an important sense autonomous, i.e.,

“a cell or multicellular organism is not merely self-maintaining, like a candle flame; it is also self-producing and thus produces its own self-maintaining processes, including an active topological boundary that demarcates inside from outside and actively regulates interaction with the environment” (Thompson, 2007, p. 64, cf. 103 for the defining elements of autopoietic systems).

This organization can be illustrated by a look at the simplest organism, the living cell, out of which all complex organisms are ultimately composed. The cell also serves as a model in Varela and Maturana’s initial arguments for the notion of the autopoietic organization that defines organisms: in a single cell, a biochemical network “produces the metabolites that constitute both the network itself and the membrane that permits the network’s bounded dynamics” (Thompson, 2007, p. 65). So, in the first place, a cell qua self-organized and self-producing system generates a simple biological self-world distinction. By itself, this is still insufficient to count as intentionality. But in order to survive and maintain its identity, the cell must continually exchange matter and energy with its environment. Some molecules are imported through the membrane and participate in processes inside the cell, whereas other molecules are excreted as waste. In this way, the cell produces its own components including its boundary, which in turn produce and maintain it as a unified system, in an ongoing process. Autopoiesis is the term to describe this continual self-production (Thompson, 2007, p. 97ff). This brings us back to the discussion of Damasio’s account. In order to sustain itself, the cell must realize biological purposes. With respect to the cell’s biological needs, its physical environment thus obtains a certain value. For the cell, features of the physicochemical environment “turn into” nutrition; but only in relation to the cell’s metabolism do they acquire the status of food. In this way, by being an entity with an identity that depends on its environment for survival, the cell is directed toward and dependent on its environment, i.e., the relation between cell and environment is asymmetric. By being semi-autonomous, the cell as organism bestows significance and value to the relevant features in the environment. Thus, living itself is a way of bringing forth value and significance. “In this way, the environment becomes a place of valence, of attraction and repulsion, approach or escape.” (Thompson, 2007, p. 158) Accordingly, the semipermeable boundary that enables the cell to exchange matter and energy with its environment can be seen as the “natural root of intentionality: Intentionality arises from the operational closure and interactive dynamics of autopoiesis” (Thompson, 2007, p. 159).

Thompson (2007, p. 159) also illustrates his view by saying that “intentionality first emerges in nature in the form of autopoiesis and sense-making.” To call what the cell is doing here, namely its intentional directedness toward its environment based on its biological needs, “sense-making” suggests that it already exhibits a basic kind of cognition. But there is a danger here that slipping from this notion of biological intentionality to the notions of attitudes and sense-making so quickly blurs rather than illuminates this otherwise persuasive naturalization of intentionality. Thompson (2007, p. 127) is cautious not to take a stand in this matter, but Hutto and Myin (2013, pp. 32–36) object to this suggestive way of describing it. The important point to consider at this juncture is that once we have acknowledged the naturalization of the notion of intentionality (as directedness) provided by Autopoietic Enactivism, various theoretical options how to proceed from here are still available. These options concern not only explanations and restrictions on the notion of cognition, but also specifications of the relations between intentionality, cognition, and representation. The crucial question is whether we want to identify cognition in this broad way with autopoiesis and biological autonomy or if we want to introduce a further restriction on cognition. Although I cannot conclusively discuss these questions in this paper, let me sketch a range of options.

What Is the Mark of the Cognitive?

First of all, note that if we take intentionality to be equivalent to mental representation, then this implies that what the cell is doing counts as cognition. Whoever holds this view, must then answer the question what the cell (or certain elements of it) may represent. And indeed, such questions have been debated at length in discussions about naturalizing intentionality. Dretske (1986, p. 26), for example, introduced the case of marine bacteria which can only survive in the absence of oxygen and contain internal magnets (magnetosomes) which enable them to be directed toward oxygen-free waters (coinciding with geomagnetic north in the northern hemisphere and with geomagnetic south in the southern hemisphere). They basically avoid the surface. It is plausible to assume that this magnetic mechanism has evolved for this purpose, in the service of survival. These bacteria clearly exhibit intentionality in the sense of directedness, but it is much less clear that the magnetosome (or the bacterium) represents anything. Speculations about their representational content included “north” or “anaerobic water” (Millikan, 1989Pietroski, 1992). Dretske (1986, p. 27) suggested that its natural meaning was “that there is relatively little oxygen in that direction” and that when a bacterium from the southern hemisphere were to be transplanted into the northern hemisphere, its magnetosome would effectively lead to its destruction, because it would lead it in the wrong direction. Dretske takes this to be “a plausible instance of misrepresentation.” But note that we are only forced to ponder such questions if intentionality is taken to be equivalent with representation. Although the bacterium is intentionally directed toward oxygen-free water (or however, we want to describe it5), and although it contains a mechanism that enables this (quite rigid, inflexible) directedness, this mechanism does not (need to) represent anything, especially because it does not allow for any flexibility in its behavior.6

If we reject the Equivalence Thesis, various other theoretical options remain. We are then in the position to allow that bacteria may exhibit basic intentionality based on their metabolic needs without thereby allowing this to involve representation. Still, although Enactivists of various stripes may agree concerning the separation of intentionality and representation, they may argue that we should consider this process in the single cell as a case of cognition. As mentioned above, Thompson (2007, p. 126) seems to be open to this idea when he holds that “cognition is behavior in relation to meaning and norms that the system itself enacts or brings forth on the basis of its autonomy.” Insofar as bacteria fulfill this criterion, they are capable of cognition. This option raises the question as to how useful it is for us to maintain such a broad notion of cognition since it has obvious consequences for cognitive science. If the mark of the cognitive is to exhibit intentional directedness in the basic biological sense of autopoiesis, then one consequence is that the subject matter of cognitive science is identical to the subject matter of biology. The study of cognitive systems then is the study of biological systems. To the extend to which we would like to learn more about how cognition works and what it is in our case, this makes sense. Yet, cognitive scientists may wish to delineate their field in a quite different way that is at the same time more restricted in one dimension and more flexible in another. Let me explain.

Thompson (2007, p. 128) holds the view that “the organizational properties distinctive of mind are an enriched version of those fundamental to life.” Yet, it is one thing to say that mind and life form a continuum, and quite another to hold that “mind is life,” that “the problem of mind is that of the problem of life” and that “where we discern life, we have everything we need to discern mind” (Noë, 2009, pp. 41–42). This latter view that slips from the “continuity” of life and mind (and thus, cognition) to the co-extensiveness of life and mind is quite useless for researchers in cognitive science who are considering (and developing) artificial systems and their cognitive capacities. After all, a strong reading of this option makes artificial cognitive systems impossible simply because they do not exhibit the right kind of biological setup. Since artificial intelligence has always been a goal of cognitive science, this view may strike many cognitive scientists as being way too restrictive. Di Paolo et al. (2017) prefer to use the more neutral notion of an “agent” and develop a hierarchy of notions of agency, starting with the minimal biological agency exhibited by bacteria, and culminating in the full-blown flexible and “open sensorimotor agency” of human beings and some other animals. With regard to the possibilities of artificial intelligence, they note that “the challenge for robotics is to create agency directly at the sensorimotor level” while bypassing the biological foundation found in organisms. “Insofar as the robot is capable of supporting the emergence, maintenance, and adaptive regulation of a network of precarious sensorimotor schemes it is a sensorimotor agent” (Di Paolo et al., 2017, p. 172). Yet, they hold that so far no robot meets this criterion since none of them can “self-individuate” so that they remain on the same level of behavioral flexibility like bacteria. I think that some kind of scale like the one Di Paolo et al. (2017) propose is needed and will be a welcome differentiation among the phenomena at issue. After all, “cognition” can also be seen as an umbrella term for a range of capacities that can be hierarchically ordered by complexity. Yet, note that the notions of agent and organism are quite distinct and pick out different sets of entities. On the one hand, the notion of an agent is much more flexible since it allows for the possibility of artificial agents that perform cognitive tasks, while the former restriction to organisms keeps its focus on the biological domain. On the other hand, one may worry that the ascription of something as an “agent,” being itself observer-relative and subject to taking the intentional stance toward any system of interest, will possibly include entities that are obviously non-cognitive.

Bourgine and Stewart (2004) and Bitbol and Luisi (2005) argue that autopoiesis may be sufficient for (or even constitutive of) life but that it is by itself insufficient for cognition. Continuity of mind and life (Thompson, 2007Kirchhoff and Froese, 2017) does not imply that mind and life are co-extensive. But other philosophers are more inclusive, since they also consider the behavior of plants as falling within the realm of cognition. For example, in their defense of plant cognition, Garzon and Keijzer (2011) propose the criteria of motility and sensorimotor organization (first suggested by Jonas, 1966) as sufficient for minimal cognition. Incidentally, bacteria meet their criterion such that their proposal does not significantly differ from Autopoietic Enactivism as discussed above. They list striking instances of adaptive behavior in plants to do with movement, signal integration and other capacities, and argue that plants meet Jonas’ criteria once we differentiate between being free-moving on the one hand (like bacteria) and having self-induced motility on the other. Plants exhibit the latter despite not being free-moving, or so they argue. Jonas, by contrast, defended the stricter requirement that cognition demands free motility because only being free-moving allows for self-initiated action in the environment, as Garzon and Keijzer admit. It seems difficult to adjudicate this issue empirically given that we simply do not know enough yet about how many of these organisms, including bacteria, function. Much more work is needed here and evidence may decide the issue in one or the other direction.7

A way out of this dilemma is to abstract from the biological details and refer to the organizational or functional features of autopoietic systems. But note that this is to leave behind a stronger claim regarding the identity of life and mind by allowing quite different realizations of the very organization characteristic of life and mind. Once we focus on functions in living creatures, we have the choice among various candidates of which self-organization and self-maintenance are only two. This gives rise to the question what kind of functional architecture an organism must possess in order to meet the condition of being not only an “intentional system” but a “cognitive system.” A possible candidate may be the presence of a centralized control system that is responsible for (i.e., has the function of) processing information resulting from diverse inputs coming from multiple sources and putting this to use for executive control (Piccinini, unpublished). This is a way of differentiating the cognitive from the non-cognitive within the realm of (intentionally directed) organisms, since plants, fungi and bacteria may lack such an integrating control system. Although plants, bacteria and other organisms exhibit an autopoietic organization, their behavior then does not constitute a case of cognition because they lack the kind of complex processing and integration of information from various channels that is characteristic of cognition (on such a view).

Such a function of integrative control that is characteristic of cognitive processes may require a nervous system and brain. The underlying intuition here would be that nervous systems and brains provide the adequate architecture and functionality to instantiate mental representations, which in turn are considered as crucial for cognitive phenomena. Rejecting the Equivalence Thesis, this would situate the biological source of intentionality in the autopoietic organization of organisms, while restricting cognition to those processes involving mental representations which can – contingently – only be generated and sustained by nervous systems and brains. Rupert (2018) seems to have something like this in mind in his attempt to delineate “mere” representations from genuinely mental representations. According to his proposal, the latter must meet certain criteria one of which is that they occur in certain structures exhibiting a specific architecture.

This seems to be similar to the quite strict restriction of the realm of cognition that some philosophers prefer, e.g., by determining cognition as constituted by “particular kinds of processes involving non-derived representations” (Adams and Aizawa, 2001, p. 53). Adams and Aizawa propose this criterion in the context of the debate about whether cognitive processes can extend beyond the brain into the body and into the environment (Clark and Chalmers, 1998). Although this claim is silent about the “locus” of cognition, they argue that only certain kinds of mechanisms and processes can meet this criterion, namely, brainbound processes. Of course, this proposal is also not without problems. For one thing, it is controversial whether non-derived content exists (Dennett, 1987). Moreover, as Garzon and Keijzer (2011) point out, some researchers even consider “root-brains” (Baluška et al., 2004) in plants, which, despite being very different from “real brains,” are conceived as control centers regulating the plants’ adaptive behavior. This terminology suggests a certain liberty at work here and it does not only point to the need for conceptual clarifications of the kind proposed in this paper, but also to future empirical work that could decide some of these issues.

This (inconclusive) discussion of possible options further demonstrates the difficulty to go beyond certain terminological preferences, and arrive at the best delineation for cognitive science. What’s certain though is that the Representational Theory of Mind and Enactivism do by far not exhaust the range of theoretical options when it comes to determining the mark of the cognitive and the conceptual relations between cognitive, representational, and intentional phenomena. At bottom, the issue is whether cognitive scientists should favor a very broad notion of cognition or a more restricted one in order to investigate cognitive phenomena and it may well be up to the individual researchers to decide for themselves which notion is more useful to them. In the following last section, I would like to sum up and list some advantages of the general approach to a separation of intentionality and mental representation suggested here.


To recap, I have argued that intentionality can be naturalized if we take it primarily to be a feature of whole embodied agents, in our case, organisms, leaving room for the possibility of artificial intentional agents. This basic kind of biological intentionality is manifest even in a single cell’s autonomous self-production and maintenance of its identity via its dynamic interaction with the environment (see also Dennett, 1995, p. 205). The biological needs of the organism guide and control its intentional directedness and interaction with the environment, not the other way around. But this leaves us with the task of naturalizing mental representation which is conceived as a feature of mechanisms not whole organisms, a task not directly addressed in this paper. Of course, we could simply continue to stipulate that the notions of intentionality and mental representation can be used interchangeably, but we would loose the explanatory advantages that their clear separation offers. This separation allows, first and foremost, for a much more differentiated view of intentionality, representation, and mental representation, as indicated above. Focusing on “agents” of various complexities, Di Paolo et al. (2017) proposal has a similarly hierarchical taxonomy in mind. Such differentiations are especially important if we look at debates – in the context of Enactivism and Embodied Cognition – about the putative cognitive capacities of basic organisms, including cells, lower and higher animals and plants. It opens up the possibility, for example, that on the one hand, organisms can be directed at objects in various ways, the most basic of which do not involve mechanisms that carry representational content. But on the other hand, having such representational mechanisms enables an organism (or more general, a system) to be directed not only at ordinary objects or goals, but even at non-existent things via the (representational) capacity of imagination. In this way, the present framework can integrate what many philosophers take to be “Brentano’s problem,” namely the fact that intentionality involves the capacity to be directed at things that do not exist.

Moreover, a further advantage of this proposal is that it enables us to clearly situate predictive processing views and the impact of the free-energy principle (FEP) in discussions of cognition. Friston (2010) developed his predictive processing view of cognition and action as a special case of the tendency of organisms to resist entropy minimizing variational free energy, making the minimization of prediction error an instance of the FEP. The FEP originates in information theory to explain the self-organizing dynamics in systems that can remain in states far from equilibrium, and its application is thus quite independent from (or at least much more general than) mind and cognition (see also Friston, 2013). The proposed distinction underlying the conceptual move defended here considers the FEP, self-organizing dynamics and autopoiesis as pertaining to intentionality, whereas it considers the particular instance of the FEP, namely neuronal predictive processing, as pertaining to cognition and mental representation, providing an architecture sustaining representations. Thus, the proposed distinction avoids overly generous views according to which all systems that minimize free energy would have to be cognitive systems (see also Kirchhoff and Froese, 2017). In this way, the present proposal can integrate insights from the various frameworks distinguished above much better than Radical Enactivism. For example, it can much better make sense of Predictive Processing accounts of perception than Enactivism.8

Finally, the present conceptual move yields a layered model of intentionality with a phylogenetic as well as an ontogenetic dimension (Barresi and Moore, 1996Schlicht, 2008). The central idea which is outside the scope of this programmatic paper is that there is not only a scale from simpler to complex creatures with an increasing range of intentional forms of directedness, but that also ontogenetically, our range of intentional forms of directedness unfolds during cognitive development. Taking this developmental stance enables the separation of non-representational kinds of intentionality from representational ones and the identification of a “distinctively mental kind of representation” (Morgan and Piccinini, 2017, p. 3) with the help of cognitive neuroscience within this broader framework of biological intentionality. As we have seen, Metzinger’s thematic focus is on mental representation that is also phenomenally experienced, a first-person phenomenon. Whether the “mental” realm is identical to the potentially “conscious” realm is an open issue, but the “cognitive” realm most plausibly goes way beyond and is in certain instances much more basic than that. In this sense, the foregoing considerations can be seen as a way of supporting and complementing Metzinger’s naturalization of phenomenal representation by providing a foundation of intentionality as such.

Intentionality enabled by mental representations allows for further differentiations from a developmental point of view: In their first year, young infants first engage only in dyadic relations with either caregivers or objects but already demonstrate an early understanding of the reciprocal nature of social interaction (Tronick et al., 1978). Only at around 9–12 months do they engage in triadic relations in scenes of joint attention (Eilan et al., 2005). According to Tomasello (1999), this indicates an understanding of others as intentional agents to be distinguished from the understanding of others as mental agents, which he associates with children’s understanding of false belief at 4 years (Wimmer and Perner, 1983). In-between though, children already start using their imagination and engage in pretend play, indicating that they do not only respond to a perceptual object or event, but also simultaneously ‘hold in mind’ a representation of an absent object or event. This capacity for counterfactual thinking is crucial for the development of social cognition (Gopnik and Wellman, 2012). At 18 months, children can complete a goal-directed action that an observed adult fails to complete (Meltzoff, 1995), indicating an understanding of the adult’s intentional relation to the task as well as the goal’s affordances. Various tasks show that children progress in their understanding of different mental states (Wellman et al., 2001), starting with understanding that people have divergent desires and diverse beliefs, via an understanding that they have access to knowledge, to the understanding of false belief and hidden emotion. Once worked out in detail, the present approach has the potential to yield a differentiated theory of intentionality, representation, and mental representation that (a) meets the constraint of providing a mark of cognition, (b) provides a naturalistic theory of intentionality, and (c) can incorporate insights from cognitive neuroscience and developmental psychology.


This paper took off from Metzinger’s account of phenomenal representation in terms of his self-model theory. It emphasized the fact that he leaves unaddressed the project of providing a more general naturalistic theory of representation and intentionality. The paper then provided a sketch of how this could be developed using the means of a restricted version of Autopoietic Enactivism. Thus, the proposal is very much in spirit of Metzinger’s approach but attempts to support it by providing first steps toward a naturalistic theory of intentionality as such. The main conceptual move is to separate intentionality from representation – contra orthodox philosophy of mind and cognitive science – and conceive of intentionality as a feature of whole embodied agents and of representation as a feature of mental states (and their vehicles or underlying mechanisms, respectively). This proposal, defended in a similar fashion by Radical Enactivism, yields a division of labor in the sense that we need a naturalistic account of intentionality and another for mental representation while it is likely that they require completely different sets of explanatory tools. Space provided here only allowed attending to the first task. The upshot was that basic intentionality can be traced back to and explained by the autopoietic organization of organisms. Even single cells are intentionally directed to their environment based on their biological needs. That does not imply that they are thereby already cognitive systems or that this intentionality is cognitive or representational. Biological intentionality may be insufficient for cognition. Formulating stricter demands on cognition leaves open various options. The one suggested here is a functional restriction according to which cognition requires centralized control and integration of information coming from different sources. To the extend that this requires a physical mechanism like the brain and nervous system in organisms (and analogously, a physical mechanism of some other kind in an artificial system), it has the consequence that only creatures with such functional mechanisms can constitute an architecture sufficient for mental representation of the kind that Metzinger considers in his own account. Some of the advantages of this way of conceiving of intentionality and representation have been mentioned. The upshot is that intentionality is a developmental phenomenon that allows for various manifestations and varieties, both phylogenetically and ontogenetically. Rejecting the Equivalence Thesis thus provides us with a much more differentiated account of intentionality and representation than can be provided by theories based on the Equivalence Thesis. Yet, drawing this distinction does not imply Radical Enactivism but still leaves room for theoretical alternatives that accept mental representations in explanations of cognitive phenomena.

Author Contributions

The author confirms being the sole contributor of this work and approved it for publication.


This work was funded by the VolkswagenStiftung for the research project Situated Cognition (Az. 87/105) and the DFG-Graduiertenkolleg Situated Cognition (GRK-2185/1).

Conflict of Interest Statement

The author declares that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.


The author wants to thank the Volkswagen Foundation for sponsoring his research project Situated Cognition. Perceiving the world and understanding other minds (Az 87/105), and Krzysztof Dolega, Tobias Starzak, Luke Roelofs, Elmarie Venter, and the reviewers and editors of this special issue for valuable comments.


  1. ^ Note that Crane (2003, p. 31) takes it that “the phrases ‘intentional inexistence,’ ‘relation to a content’ and ‘immanent objectivity,’ despite superficial differences between them, are all different ways of expressing the same idea: that mental phenomena involve representation or presentation of the world.”
  2. ^ Metzinger restricts the term “mental” for processes and events that can become conscious. Although this decision is more or less arbitrary, some terminological distinction is needed to single out “distinctively mental representations” from other kinds of representations. What is still needed is a rigorous terminology that defines how ‘representations,’ ‘mental representations,’ and ‘phenomenal representations’ are related and, most crucially, how they relate to the notion of ‘cognition.’
  3. ^ Up to this point, I have used the notions “agent” or “embodied agent” to denote the bearer of intentionality. As we will see in a bit, enactivists prefer to consider “organisms” as such bearers. Although certain organisms are embodied agents, it may be disputed that all organisms meet the criteria for being an agent. For that reason, it must be spelled out what we should take to be an agent. Also, not all embodied agents need be organisms. For example, from the start it was one of the chief goals of cognitive science to develop artificial intelligence, i.e., an artificial system that can perform cognitive tasks which have been considered as specifically human. We will discuss this issue further below.
  4. ^ Even though Enactivism comes in many flavors, the grounding of intentionality in autopoiesis, an idea that originates in the work of Maturana and Varela (1980), is widely accepted by the main supporters of Enactivism. Hutto and Myin (2013, p. 33) emphasize their “strong affinities with Autopoietic Enactivism,” while Noë (2009, p. 42) explicitly states that “mind is life.” Moreover, see Froese and Di Paolo (2011) and Kirchhoff and Froese (2017) who defend the mind-life continuity thesis as one of the main pillars of Enactivism.
  5. ^ One reviewer pointed out that if the bacterium’s mechanism was directed at anaerobic water, it would therefore also be directed toward North (and so on). Yes, we can provide many names for that toward which it is directed, but the question is whether we are forced to assume that it represents any of these contents. The suggestion is that it need not and that we need not assume that it does so for the purposes of explaining the bacterium’s behavior or the mechanism’s function within the bacterium.
  6. ^ Explaining the flexibility of behavior exhibited by certain agents has always been one important reason to posit mental representations in the first place (see Pylyshyn, 1984, chapter 2). Behavior being rigid and inflexible can thus be taken as an indication that positing mental representations that guide such behavior is unnecessary. I take the bacterium’s behavior to fall in this latter category. See also Di Paolo et al.’s (2017) recent distinctions among concepts of agency, ranging from the “minimal biological agency” exhibited by all organisms up to the “open sensorimotor agency” exhibited by humans and some higher animals whose behavior can “change in unpredictable, historically and culturally influenced manners” (Di Paolo et al., 2017, p. 171).
  7. ^ See Klumpp and Faivre (2016) for a review of the complexity of magnetotactic bacteria.
  8. ^ While Hutto and Myin only hint at a version of the predictive processing framework in radical enactive terms without alluding to “models” and “representations,” they do not develop this putative positive proposal anywhere (Hutto and Myin, 2017Thompson, 2018).


Adams, F., and Aizawa, K. (2001). The bounds of cognition. Philos. Psychol. 14, 43–64. doi: 10.1080/09515080120033571

CrossRef Full Text | Google Scholar

Baluška, F., Mancuso, S., Volkmann, D., and Barlow, P. (2004). Root apices as plant command centres: the unique ‘brain-like’ status of the root apex transition zone. Biologia 59, 7–19.

Google Scholar

Barresi, J., and Moore, C. (1996). Intentional relations and social understanding. Behav. Brain Sci. 19, 107–122. doi: 10.1017/S0140525X00041790

CrossRef Full Text | Google Scholar

Bitbol, M., and Luisi, P. L. (2005). Autopoiesis with or without cognition: defining life at its edge. J. R. Soc. Interface 1, 99–107. doi: 10.1098/rsif.2004.0012

PubMed Abstract | CrossRef Full Text | Google Scholar

Bourgine, P., and Stewart, J. (2004). Autopoiesis and cognition. Artif. Life 20, 327–345. doi: 10.1162/1064546041255557

PubMed Abstract | CrossRef Full Text | Google Scholar

Brentano, F. (1874/1995). Psychology from an Empirical Standpoint. trans. A. C. Rancurello, D. B. Terrell, and L. L. McAlister (London: Routledge).

Google Scholar

Chemero, A. (2009). Radical Embodied Cognitive Science. Cambridge, MA: MIT Press.

Google Scholar

Churchland, P. S., and Sejnowski, T. (1992). The Computational Brain. Cambridge, MA: MIT Press.

Google Scholar

Clark, A. (2015). Radical predictive processing. South. J. Philos. 53, 3–27. doi: 10.1111/sjp.12120

CrossRef Full Text | Google Scholar

Clark, A. (2016). Surfing Uncertainty. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780190217013.001.0001

CrossRef Full Text | Google Scholar

Clark, A., and Chalmers, D. J. (1998). The extended mind. Analysis 58, 7–19. doi: 10.1093/analys/58.1.7

CrossRef Full Text | Google Scholar

Crane, T. (2003). The Mechanical Mind, 2nd Edn. London: Routledge.

Google Scholar

Damasio, A. (2011). Self Comes to Mind. New York, NY: Pantheon.

Google Scholar

Damasio, A. R. (1999). The Feeling of what Happens. Body and Emotion in the Making of Consciousness. San Diego, CA: Harcourt.

Google Scholar

Dennett, D. C. (1987). The Intentional Stance. Cambridge, MA: MIT Press.

Google Scholar

Dennett, D. C. (1995). Kinds of Minds. New York, NY: Basic Books.

Google Scholar

Di Paolo, E., Buhrmann, T., and Barandiaran, X. E. (2017). Sensorimotor Life. An Enactive Proposal. New York, NY: Oxford University Press. doi: 10.1093/acprof:oso/9780198786849.001.0001

CrossRef Full Text | Google Scholar

Dretske, F. (1981). Knowledge and the Flow of Information. Cambridge, MA: MIT Press.

Google Scholar

Dretske, F. (1986). “Misrepresentation,” in Belief: Form, Content, and Function, ed. R. Bogdan (Oxford: Clarendon), 157–173.

Google Scholar

Dreyfus, H. (2002). Refocusing the question: can there be skillful coping without propositional representations or brain representations? Phenomenol. Cogn. Sci. 1, 413–425. doi: 10.1023/A:1021303723047

CrossRef Full Text | Google Scholar

Eilan, N., Hoerl, C., McCormack, T., and Roessler, J. (eds). (2005). Joint Attention: Communication and other Minds. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199245635.001.0001

CrossRef Full Text | Google Scholar

Fitch, W. T. (2008). Nano-intentionality: a defense of intrinsic intentionality. Biol. Philos. 23, 157–177. doi: 10.1007/s10539-007-9079-5

CrossRef Full Text | Google Scholar

Fodor, J. (1975). The Language of Thought. New York, NY: Harvard University Press.

Google Scholar

Fodor, J. (1987). Psychosemantics. Cambridge, MA: The MIT Press.

Google Scholar

Friston, K. J. (2010). The free-energy principle: a unified brain theory? Nat. Rev. Neurosci. 11:127138. doi: 10.1038/nrn2787

PubMed Abstract | CrossRef Full Text | Google Scholar

Friston, K. J. (2013). Life as we know it. J. R. Soc. Interface 10:20130475. doi: 10.1098/rsif.2013.0475

PubMed Abstract | CrossRef Full Text | Google Scholar

Froese, T., and Di Paolo, E. (2011). The enactive approach: theoretical sketches from cell to society. Pragmat. Cogn. 19, 1–36. doi: 10.1075/pc.19.1.01fro

CrossRef Full Text | Google Scholar

Gallagher, S. (2017). Enactivist Interventions. Oxford: Oxford University Press. doi: 10.1093/oso/9780198794325.001.0001

CrossRef Full Text | Google Scholar

Garzon, P. C., and Keijzer, F. (2011). Plants: adaptive behavior, root-brains, and minimal cognition. Adapt. Behav. 19, 155–171. doi: 10.1177/1059712311409446

CrossRef Full Text | Google Scholar

Gopnik, A., and Wellman, H. (2012). Reconstructing constructivism: causal models, Bayesian learning mechanisms and the theory-theory. Psychol. Bull. 138, 1085–1108. doi: 10.1037/a0028044

PubMed Abstract | CrossRef Full Text | Google Scholar

Hohwy, J. (2013). The Predictive Mind. Oxford: Oxford University Press. doi: 10.1093/acprof:oso/9780199682737.001.0001

CrossRef Full Text | Google Scholar

Hurley, S. (1998). Consciousness in Action. Cambridge: Cambridge University Press.

Google Scholar

Hutto, D. D. (2008). Folk-Psychological Narratives. Cambridge, MA: MIT Press.

Google Scholar

Hutto, D. D., and Myin, E. (2013). Radicalizing Enactivism. Cambridge, MA: MIT Press.

Google Scholar

Hutto, D. D., and Myin, E. (2017). Evolving Enactivism. Cambridge, Mass: MIT Press.

Google Scholar

Jonas, H. (1966). The Phenomenon of Life: Toward a Philosophical Biology. New York, NY: Harper & Row.

Google Scholar

Kant, I. (1790/1998). Critique of the Power of Judgment, ed. Guyer (Cambridge: Cambridge University Press).

Google Scholar

Kirchhoff, M., and Froese, T. (2017). Where there is life there is mind: in support of a strong life-mind continuity thesis. Entropy 19:169. doi: 10.3390/e19040169

CrossRef Full Text | Google Scholar

Klumpp, S., and Faivre, D. (2016). Magnetotactic bacteria. Eur. Phys. J. Spec. Top. 225, 2173–2188. doi: 10.1140/epjst/e2016-60055-y

CrossRef Full Text | Google Scholar

Maturana, H. R., and Varela, F. G. (1980). Autopoiesis and Cognition: The Realization of the Living. Boston Studies in the Philosophy of Science, Vol. 42. Dordrecht: D. Reidel. doi: 10.1007/978-94-009-8947-4

CrossRef Full Text | Google Scholar

Meltzoff, A. (1995). Understanding the intentions of others: re-enactment of intended acts by 18-month-old children. Dev. Psychol. Psychol. 31, 838–850. doi: 10.1037/0012-1649.31.5.838

PubMed Abstract | CrossRef Full Text | Google Scholar

Mendelovici, A. (2018). The Phenomenal Basis of Intentionality. New York, NY: Oxford University Press. doi: 10.1093/oso/9780190863807.001.0001

CrossRef Full Text | Google Scholar

Metzinger, T. (2003). Being no One. Cambridge, MA: MIT Press.

Google Scholar

Metzinger, T. (2014). “First-order embodiment, second-order embodiment, third-order embodiment: from spatiotemporal self-location to minimal selfhood,” in The Routledge Handbook of Embodied Cognition, ed. R. Shapiro (Abingdon: Routledge), 272286.

Google Scholar

Metzinger, T., and Wiese, W. (eds). (2016). Philosophy and Predictive Processing. Available at:

Google Scholar

Millikan, R. G. (1989). Biosemantics. J. Philos. 86, 281–297. doi: 10.2307/2027123

CrossRef Full Text | Google Scholar

Morgan, A., and Piccinini, G. (2017). Towards a cognitive neuroscience of intentionality. Minds Mach. 28, 119–139. doi: 10.1007/s11023-017-9437-2

CrossRef Full Text | Google Scholar

Noë, A. (2004). Action in Perception. Cambridge, MA: MIT Press.

Google Scholar

Noë, A. (2009). Out of our Heads. New York, NY: Hill & Wang.

Google Scholar

O’Brien, G., and Opie, J. (2004). “Notes toward a structuralist theory of mental representation,” in Representation in Mind: Approaches to Mental Representation, eds H. Clapin, P. Staines, and P. Slezak (Oxford: Elsevier), 120.

Google Scholar

Pietroski, P. (1992). Intentionality and teleological error. Pac. Philos. Q. 73, 267–282. doi: 10.1111/j.1468-0114.1992.tb00339.x

CrossRef Full Text | Google Scholar

Pitt, D. (2004). The phenomenology of cognition Or What is it like to think that p? Philos. Phenomenol. Res. 69, 1–36. doi: 10.1111/j.1933-1592.2004.tb00382.x

CrossRef Full Text | Google Scholar

Pylyshyn, Z. (1984). Computation and Cognition. Cambridge, MA: MIT Press.

Google Scholar

Rupert, R. (2018). “Representation and mental representation,” in Philosophical Explorations Special Issue on Enactivism, Representationalism, and Predictive Processing, Vol. 21, eds K. Dolega, L. Roelofs, and T. Schlicht204–225.

Google Scholar

Schlicht, T. (2008). “Ein stufenmodell der intentionalität,” in Zur Zukunft der Philosophie des Geistes, ed. P. Spät (Paderborn: Mentis), 59–91.

Google Scholar

Searle, J. R. (1983). Intentionality: an Essay in the Philosophy of Mind. Cambridge, MA: MIT Press. doi: 10.1017/CBO9781139173452

CrossRef Full Text | Google Scholar

Searle, J. R. (1992). The Rediscovery of the Mind. Cambridge, MA: MIT Press.

Google Scholar

Smolensky, P. (1988). On the proper treatment of connectionism. Behav. Brain Sci. 11, 1–74. doi: 10.1016/S0034-7450(14)60076-7

PubMed Abstract | CrossRef Full Text | Google Scholar

Strawson, G. (2004). Real intentionality. Phenomenol. Cogn. Sci. 3, 287–313. doi: 10.1023/B:PHEN.0000049306.63185.0f

CrossRef Full Text | Google Scholar

Thompson, E. (2007). Mind in Life. Cambridge, MA: Harvard University Press.

Google Scholar

Thompson, E. (2018). Review of Daniel D. Hutto and Erik Myin, evolving enactivism: basic minds meet content. Notre Dame Philoso. Rev.11:2018.

Google Scholar

Tomasello, M. (1999). The Cultural Origins of Human Cognition. Cambridge, MA: Harvard University Press.

Google Scholar

Tronick, E., Als, H., Adamson, S., and Brazelton, B. (1978). The infant’s response to entrapment between contradictory messages in face-to-face interaction. J. Am. Acad. Child Psychiatry 17, 1–13. doi: 10.1016/S0002-7138(09)62273-1

CrossRef Full Text | Google Scholar

Varela, F., Thompson, E., and Rosch, E. (1991). The Embodied Mind. Cognitive Science and Human Experience. Cambridge, MA: Oxford University Press.

Google Scholar

Weber, A., and Varela, F. (2002). Life after Kant. Natural purposes and the autopoietic foundations of biological individuality. Phenomenol. Cogn. Sci. 1, 97–125. doi: 10.1023/A:1020368120174

CrossRef Full Text | Google Scholar

Wellman, H. M., Cross, D., and Watson, J. (2001). A meta-analysis of theory-of-mind development: the truth about false belief. Child Dev. 72:655684. doi: 10.1111/1467-8624.00304

CrossRef Full Text | Google Scholar

Wimmer, H., and Perner, J. (1983). Beliefs about beliefs: representation and constraining function of wrong beliefs in young children’s understanding of deception. Cognition 13, 103–128. doi: 10.1016/0010-0277(83)90004-5

CrossRef Full Text | Google Scholar

Bidirectionality and Metaphor: An Introduction

Bidirectionality and Metaphor: An Introduction

Chanita Goodblatt Ben-Gurion University of the Negev, Department of Foreign Literatures and Linguistics Joseph Glicksohn Bar-Ilan University, Department of Criminology

Abstract The authors first present the interaction theory of metaphor, emphasizing its notion of bidirectionality. They then discuss the relationship between bidirectionality and blending, making explicit the different expectations regarding bidirectionality deriving from interaction theory and blending theory. With this as a suitable background for this special issue on bidirectionality in metaphor, the authors then provide a brief introduction to each of the essays that appear in the issue.

Keywords interaction theory, gestalt, blending, tension, cognition

The Interaction Theory of Metaphor It is not inconsequential that in developing his interaction theory of metaphor, Max Black uses the metaphor “man is a wolf,” a metaphor that is incompatible and even grotesque in its juxtaposition of man and animal. As we shall argue in our own essay in this issue, the grotesqueness of this Poetics Today 38:1 (February 2017) DOI 10.1215/03335372-3716189 q 2017 by Porter Institute for Poetics and Semiotics A preliminary version of this paper was presented at two venues: the research seminar held at Bar-Ilan University, in the Gonda Multidisciplinary Brain Research Center, entitled “The Creating Mind: Interdisciplinary Perspectives,” December, 2013; and at the third international conference on Cognitive Futures in the Humanities: Forging Futures from the Past, Worcester College, Oxford, April, 2015. We were supported by a Bar-Ilan University Faculty Research Grant, awarded to Joseph Glicksohn. Downloaded from by guest on 21 September 2021 image is a key to understanding what we shall term a potential for “bidirectionality” in metaphor comprehension. Let us, however, develop this claim in stages. Max Black (1962: 41) writes: The effect … of (metaphorically) calling a man a “wolf ” is to evoke the wolfsystem of related commonplaces. If the man is a wolf, he preys upon other animals, is fierce, hungry, engaged in constant struggle, a scavenger, and so on. … The wolf-metaphor suppresses some details, emphasizes others — in short, organizes our view of man. … We can say that the principal subject is “seen through” the metaphorical expression — or, if we prefer, that the principal subject is “projected upon” the field of the subsidiary subject (emphasis original). If, at this preliminary stage, we look at the nominative “A is a B” metaphor, then following Black, we can attempt to visualize the process of metaphor comprehension (How is A a B?) by means of Figure 1. Much has been written on the analysis of such a nominative metaphor (e.g., Chiappe, Kennedy, and Chiappe 2003; Chiappe, Kennedy, and Smykowski 2003; Gentner et al. 2001; Glucksberg and Keysar 1990), yet little of it has explicitly adopted Black’s way of seeing the principal subject A (the tenor, Figure 1 Visualizing the process of metaphor comprehension 2 Poetics Today 38:1 Downloaded from by guest on 21 September 2021 topic, target, primary subject) through the lens of the subsidiary subject B (the vehicle, source, secondary subject), or of “project[ing]” A “upon” the field of B. We have tried to capture this projection in Figure 1. As A is seen through the lens of B, A becomes much more similar to B than it previously was. This is a process, and not a static comparison. As Roger Tourangeau and Robert Sternberg (1982: 212 – 13) write, “Interaction theorists argue that the vehicle of a metaphor is a template for seeing the tenor in a new way. … In Black’s view … interpretation involves not so much comparing tenor and vehicle for existing similarities, as construing them in a new way so as to create similarity between them.” Black (1962: 44) continues, “If to call a man a wolf is to put him in a special light, we must not forget that the metaphor makes the wolf seem more human than he otherwise would.” In our understanding, and again with reference to Figure 1, it is true that, in the first stage of comprehending “A is a B,” A becomes much more similar to B than it previously was. But now B is “projected upon” the field of A, which has already undergone transformation in the first stage. So we have a new A and a new B, and these two are in juxtaposition in stage 2 of the process. Again, to cite Tourangeau and Sternberg (1982: 214), “metaphors generally involve seeing something (men) in one domain in terms of something (wolves) in a second domain, with a resulting change in our view of both domains.” And as Raymond Gibbs (1994: 238) stresses, “the whole point of interactionism is that both terms affect the meaning of the other. The ‘seeing as’ often associated with metaphor is multidirectional. If man is seen as wolf, so too is wolf seen as man in ‘Man is a wolf’ (Black, 1979, emphasis original).” This, then, is the interaction theory of Max Black, which is actually a theory about the process of metaphor comprehension. There are various stages in this process: in the first, A is seen through the lens of B; in the second, B is seen through the lens of an already transformed A; in the third, a transformed A is now seen through the lens of a transformed B, as well as a transformed B being seen through the lens of a transformed A. Clearly, then, A and B are shifting percepts or concepts within a dynamic, interactive process. Bidirectionality is an integral part of this interactive process because, while A is seen through the lens of B, B is also seen through the lens of an already transformed A. As Bipin Indurkhya has noted, in “the interaction, the target is structured in terms of the source, as much as this can be done because the target has its own attributes, which constrain how it can be structured, so that the resulting organization is influenced both by the source and the target” (2006: 144). Goodblatt and Glicksohn † Bidirectionality and Metaphor 3 Downloaded from by guest on 21 September 2021 Is there a resolution to this iterative process of metaphor comprehension? To address this question, we turn to Heinz Werner and Bernard Kaplan (1963: 21 – 22) who write: Here we come to an important tenet of organismic theory of symbol formation: the act of denotative reference does not merely, or mainly, operate with already formed expressive similarities between entities. Through its productive nature, it brings to the fore latent expressive qualities in both vehicular material and referent that will allow the establishment of semantic correspondence between the two entities. It is precisely this productive nature of the denotative act that renders possible a symbolic relation between any entity and another. Such a possibility could never be realized if one were dealing with static entities, namely, the symbolic vehicle as an end product and the referent as a preformed “thing out there.” It is only realized because it rests on twin form-building processes, one directed towards the establishment of meaningful objects (referents), the other directed towards the articulation of patterns expressive of meaning (vehicles) (emphasis original). As they insist, this is a dynamic, interactive, developmental process. Indeed, we suggest that if this process of metaphor comprehension is developmental in nature, then in line with Werner’s exposition of a general principle of development (1978: 108 – 9), such development “proceeds from a state of relative globality and lack of differentiation to a state of increasing differentiation, articulation, and hierarchic integration.” Specifically, the process of metaphor comprehension entails the differentiation and articulation of two unidirectional readings, namely, A is seen through the lens of B, and/or B is seen through the lens of A or, as we suggest, of an already transformed A. In addition, there is the possibility of their integration (what Werner terms “hierarchic integration”) that results in a higher-order gestalt. Indeed, as we have previously argued, metaphor is a gestalt (Glicksohn and Goodblatt 1993)—to which we now add that the process of metaphor comprehension might eventually lead to a higher-order gestalt, wherein the bidirectional readings are integrated. This, we stress, is a goal that is not necessarily achieved by readers of a text when interpreting poetic metaphor. Bidirectionality and Blending At this critical point in our presentation, it is necessary to distinguish this concept of bidirectionality from that of blending. To do so, we will first elaborate on the history of those theories of metaphor that are relevant for a discussion of both interaction and blending theories. Max Black’s interaction theory was heavily influenced by Ivor A. Richards’s proposal that “when we use a metaphor we have two thoughts of different things active 4 Poetics Today 38:1 Downloaded from by guest on 21 September 2021 together and supported by a single word, or phrase, whose meaning is a resultant of their interaction” (1936: 93). Prior to Richards, there is a whole philosophical enterprise, including such scholars as Gustav Gerber (1820 – 1901), Philipp Wegener (1848 – 1916), Alfred Biese (1856 – 1930), and Karl Bu¨hler (1879 – 1963), paving the way for such a view of the interactive nature of metaphor (Nerlich and Clarke 2000, 2001). But it is Black (1962, 1979) who is credited with this notion of bidirectionality and of this explicit interaction theory. Yet George Lakoff and Mark Turner (1989: 131 – 32), in presenting their conceptual metaphor theory (CMT) as applied to poetic metaphor, were very thorough in their categorical dismissal of bidirectionality. They write: Unfortunately, this very real phenomenon has been analyzed incorrectly as follows: the target domain is described as “suffusing” the source domain, and it is claimed that the metaphor is bidirectional — from target to source as well as from source to target. Indeed, according to this theory, there is no source or target. There is only a connection across domains, with one concept seen through the filter of the other. Here’s what is wrong with such an analysis. When we understand that life is a journey we structure life in terms of a journey, and map onto the domain of life the inferential structure associated with journeys. But we do not map onto the domain of journeys the inferential structure associated with the domain of life. … We map one way only, from the source domain of journey onto the target domain of life. Yet even while interaction theory and its concept of bidirectionality is attacked, Black himself is not really mentioned; as Ray Jackendoff and David Aaron (1991: 322) note in their review of Lakoff and Turner’s book, “only in the appendix is Max Black cited as an adherent of this approach [interaction theory].” One plausible reason for ignoring Black is that he himself (1981) had reviewed the book Metaphors We Live By (1980), in which Lakoff and Mark Johnson first presented their conceptual metaphor theory, and was not forthcoming with praise. He writes, “Their exposition is endemically slipshod [and] the copious literature on metaphor is almost completely ignored” (1981: 210). Lakoff notes in an interview regarding his subsequent dismissal of Black, “I had read Black and I had no interest in what Black was doing. … What influenced me was the discovery that ordinary, everyday thought and language, and specially ordinary everyday thought, is structured metaphorically” (1998: 89). Subsequently, the line of scholarship developing prior to Richards (Nerlich and Clarke 2000, 2001) and then continuing from Richards and Black (Goodblatt and Glicksohn 2003, 2010) has been virtually sidestepped. CMT was Goodblatt and Glicksohn † Bidirectionality and Metaphor 5 Downloaded from by guest on 21 September 2021 presented as a new theory of metaphor, which dismisses the claims (or misunderstandings) of interaction theory (see, for example, Gibbs 1994: 238; Lakoff and Turner 1989: 133). CMT has itself been replaced by conceptual integration theory, or blending theory, wherein the interactive nature of metaphor has resurfaced as a “new” finding, demanding the modification of CMT, and indeed its evolution into blending theory. As Charles Forceville (2004: 86) notes in his review of this theory developed by Gilles Fauconnier and Mark Turner (2002), one of the alleged assets of blending theory that is repeatedly emphasized by Fauconnier and Turner is that it can explain emergent structure. Inasmuch as the blended space generates aspects of meaning that inheres in neither of the input spaces, conceptual integration yields something more than the sum of the component parts and hence clearly has a creative dimension. That is correct but, again, the notion of novel, emergent features has its roots in metaphor theory — not so much in the book on literary metaphor Turner himself coauthored with George Lakoff, More than Cool Reason (1989), but rather in Max Black’s “More about Metaphor” (1977/1979). Even now, this version of events is being revised, with Scarlet Marquette writing in a footnote that blending theory “is deeply indebted to Max Black, as well as to Black’s predecessors in the formulation of metaphor theory, I. A. Richards and Paul Ricoeur” (2007: 698). We propose that blending theory, with its emphasis on a process model for metaphor comprehension, has supplanted CMT, for essentially two reasons: the limitations of CMT’s concept of a unidirectional mapping, and the inability of CMT to convincingly address the emergent structure underlying poetic metaphor. As two of the major proponents of blending theory, Fauconnier and Turner (1998: 133 – 38) write: Conceptual integration — “blending” — is a general cognitive operation on a par with analogy, recursion, mental modeling, conceptual categorization, and framing. It serves a variety of cognitive purposes. It is dynamic, supple, and active in the moment of thinking. … In blending, structure from input mental spaces is projected to a separate, “blended” mental space. … The blend contains emergent structures not in the inputs. Furthermore, as elaborated by Joseph Grady, Todd Oakley, and Seana Coulson, “material is projected from both the source and target spaces to the blend. This arrangement contrasts with the simple, unidirectional projection posited by CMT, in which mappings are from source to target” (1999: 103). Thus, blending theory is quite sympathetic to the concept of bidirectionality. Even so, Lakoff (2014: 8), in discussing the recent work on bidirectionality in social cognition (Lee and Schwarz 2012), has recently cautioned 6 Poetics Today 38:1 Downloaded from by guest on 21 September 2021 that “bidirectionality of experimental effect may or may not mean bidirectionality of the metaphorical mapping.” Be that as it may, we propose that while blending theory—with its emphasis on emergent structure and, further, of bidirectionality—might well contrast with CMT, it still presents a different theoretical position to metaphor comprehension in comparison with interaction theory. For while blending theory predicts the complete fusion of the two domains, we rather predict a continued potentiality for— and tension among—possible readings (Goodblatt and Glicksohn 2016a, 2016b), as we address in the next section. Bidirectionality and Metaphor The first point to note concerns the tension between the primary subject and the secondary subject of the metaphor. As Eva Kittay (1987: 184) has stressed, if we want to preserve the tension, we cannot give an account of interaction which neutralizes all tension between vehicle and topic … Unless this tension is preserved, the “suppress[ion of ] some details” and emphasis of others does not really organize our view of man, for unless the categories of man and wolf remain distinct we cannot use one distinct entity — with its systematic interconnections — to reconceive the other (emphasis original). This is the same point stressed by Richards (1936)—there is an inherent tension between the tenor and the vehicle of the metaphor. In other words, there cannot in principle be a total fusion of the two concepts; hence their blending can only be a potential that is not usually realized. Ray Jackendoff and David Aaron (1991: 334), in their review of Lakoff and Turner (1989), provide their own take on this interactive nature of metaphor: What is the outcome of creating a relationship between the incommensurable source and target domains? [Lakoff and Turner] claim that it is an understanding of the target domain in terms of the source domain. We suspect there is more, something like a “fusion” or “superimposition” of the source and target domains. … This hypothesis sharply differentiates metaphor from simile, where the source and target domains are merely compared, not superimposed. … According to this approach, the mental representation evoked by a metaphor, as well as its affective power, are the result of superimposing the meaning of the source and target domains. Fine details of a source image that do not find precise correlates in the target domain still contribute to the meaning and affect of the composite. Hence the proliferation of image detail in poetic metaphor is motivated: it contributes to the richness of the interpretation. Their comments on the inherent tension between the two subjects of the metaphor should be highlighted. They continue: Goodblatt and Glicksohn † Bidirectionality and Metaphor 7 Downloaded from by guest on 21 September 2021 In addition, the superimposition operation itself has important effects. The most obvious is the affect contributed by using one entity as a symbol for another. … A second effect of the superimposition operation is the sense of tension conveyed by incongruously fusing two disparate domains. The interpreter seeks to resolve this tension by finding points of contact or structural similarity between the two domains, so that they become point-by-point more congruent — this is the mapping process described by [Lakoff and Turner]. Yet this in turn can lead to a third effect, the production of further tension, as the domains themselves are refocused and restructured in order to bring about greater congruence — this is the “interaction” described by Black (1979). … (1991: 335) The second point, one that was raised by Forceville, is that “the ‘oscillation’ between the subjects may go on and lead to further elaborations of the metaphor” (1996: 21). In other words, as we have stressed here, we are referring to a process of metaphor comprehension, wherein one unidirectional reading may move into a second unidirectional reading, without necessarily achieving some form of resolution of the problem of how to reconcile one reading with the other. It is interesting to consider how such “oscillation” presents itself with respect to “man is a wolf.” As part of our own research, presented in this issue (Goodblatt and Glicksohn 2016b), we asked our participants to consider this metaphor prior to the reading of a poetic text. Table 1 presents two protocols as examples of such readings. Reader S5.S1 presents a unidirectional reading of man as wolf, first looking at physical similarity, but then refers to the wolf as “someone” who is “predatory like a person is predatory.” Reader S4.S1 sees them as echoing each other, with the “dehumanizing of the man” echoing “the personification of the wolf.” This is a fine bidirectional reading of the metaphor, and it echoes Black’s insight that Table 1 Protocols for the metaphor “man is a wolf” Reader S5.S1 Reader S4.S1 Man as Wolf Man as Wolf/Wolf as Man If man is a wolf then maybe looking like a wolf? Hair of a wolf, the eyes the ears and maybe the character of a wolf? A wolf is someone, a predator, sneaky, suspicious, but also loving and caring … a wolf is part of something like a man is part of something … predatory like a person is predatory … man is like a wolf in his character, his personality, but also can be in the notion of the man as a wolf, in the social notion. Because I see them [man, wolf ], it’s like they echo each other. The dehumanizing of the man echoes the personification of the wolf. 8 Poetics Today 38:1 Downloaded from by guest on 21 September 2021 if “to call a man a wolf is to put him in a special light, we must not forget that the metaphor makes the wolf seem more human than he otherwise would” (1962: 44). We further claim that bidirectionality in a metaphor is sustained in two primary ways: a clash of sharp visual images and a use of the grotesque. If Gibbs (1994: 133) is correct in his assertion, that “imagery provides a means by which two previously dissimilar domains can be incorporated into one concept, because the task of comprehending metaphor presumably involves fusing two such domains,” then a clash of sharp visual images will prevent such fusion or blending. Furthermore, the emphasis here is on sharp because fuzzy images might be blended. The use of the grotesque, itself potentially involving such a clash of visual images (Harpham 1982: 11; Thomson 1972: 27), is a particularly interesting case, and we discuss this in our article in this special issue (Goodblatt and Glicksohn 2016b). This Special Issue on Bidirectionality The essays that we have commissioned for this special issue of Poetics Today bear on one or another aspect of bidirectionality in metaphor. We have ordered these essays such that a certain degree of balance is achieved both with respect to the disciplines represented and with respect to the claims made about bidirectionality in metaphor. In the first essay, the semiotician Marcel Danesi presents an overview of various theories of metaphor, suggesting that those implicating the interaction theory of metaphor, as explicated above, invoke notions of interaction, projection, and blending. And yet, as he writes, each model can be described as a “unidirectional” one, since it posits essentially that metaphor is the result of enlisting concrete vehicles in order to shed light on (and even construct) abstract topics. By and large, these models have not entertained the possibility that metaphor is actually a “bidirectional” process, whereby not only does it involve enlisting concrete vehicles to guide abstract conceptualization, but also the reverse — namely, that abstract topics allow us to understand the vehicles. In other words, the parts of a metaphor implicate each other in tandem. In their essay, the psycholinguists Albert N. Katz and Hamad Al-Azary suggest that three properties of semantic space “provide boundary conditions that invite uni- or bidirectionality when concepts are juxtaposed as in metaphor.” These properties are: (1) the distance of concepts A and B in this space; (2) the density of semantic space in which A and B reside; and (3) the semantic richness of this space for concrete and abstract concepts. Goodblatt and Glicksohn † Bidirectionality and Metaphor 9 Downloaded from by guest on 21 September 2021 They argue that “bidirectionality is more likely to be found when topic and vehicle come from semantically distant categories, in part because with distant categories there are fewer features that can be found relevant to both topic and vehicle.” Moving to linguistics, Roy Porat and Yeshayahu Shen address verbal metaphors and comparisons, and argue that the relation between the same two concepts/domains can, in principle, be either bidirectional or unidirectional. They suggest that bidirectionality is, in fact, “more basic than the unidirectional, in that it can be triggered by the mere presence of the two stimuli; in contrast, the unidirectional process requires an additional mechanism for it to be fully realized.” This additional mechanism is the linguistic structure of the verbal metaphor. For as they write: “In a sentence, the grammatical subject and the predicate typically tend to encode the target and the source, respectively.” Hence, unidirectionality stems from the structure of the metaphorical phrase. The cognitive psychologists David Anaki and Avishai Henik address the parallel between bidirectionality in metaphor and in synesthesia. They argue that while discussions of metaphors have suggested that metaphors are unidirectional and in particular move from the concrete to the abstract, current research has suggested that metaphors might work in the other direction also, namely, from abstract to concrete. Similarly, while most studies of synesthesia have documented its unidirectional nature, current research has provided evidence that synesthesia might be bidirectional. They address the question whether these similarities between synesthesia and metaphors are just superficial, or whether they tell us something about our cognitive mechanisms. The computer scientists Bipin Indurkhya and Amitash Ojha address advertisements in their essay, noting that while in “verbal metaphor, the target and the source domains can usually be distinguished clearly, and some features of the source domain are mapped to the target domain, and not vice versa,” this is far from clear in visual metaphor. They argue that “visual metaphors can appear to be symmetric more often than the verbal metaphors, because the lack of copula can turn the focus on the comparison between the source and the target, instead of the target itself.” The literary critic Margaret Freeman focuses in her essay on both Black’s Interaction Theory and Fauconnier and Turner’s Blending Theory. She concludes “that not only is metaphorical bidirectionality possible, it explains how the arts enable us to iconically connect with the world through our embodied cognition, not as objective observers in the Western classical sense, but as participatory sharers of that world.” Her examples range from a detailed exposition of “Man is a wolf,” discussed above, to the analysis of both advertisements and poetic texts. 10 Poetics Today 38:1 Downloaded from by guest on 21 September 2021 In the final essay, Chanita Goodblatt (a literary critic) and Joseph Glicksohn (a cognitive psychologist) discuss the grotesque nature of John Donne’s poetic imagery, which constitutes “a clash of incompatibles, generated by the great distance between the two semantic fields.” We argue that it is this clash that sustains bidirectionality in a metaphor, by preserving the tension between its two subjects, while allowing each to alternatively become the focus of one’s attention while reading. Through both a cognitive-literary and an empirical study of the metaphors in Donne’s poems “The Bait” and “The Flea,” we show how these metaphors enable both embodied simulation and bodily feeling in the reader. We argue that Donne is, in fact, an early advocate of embodied cognition. The reader will thus find in this special issue an up-to-date assessment of the study of bidirectionality in metaphor. Our primary objectives are fivefold. First, we set out the intellectual history focused on the concept of bidirectionality, which is particularly evident in this “Introduction” and in the essay by Marcel Danesi. As these essays argue, the two major theories that promote the concept of bidirectionality in metaphor, the Interaction Theory of Metaphor and Blending Theory, both derive from a rich interdisciplinary heritage; we hope our readers will appreciate both the similarity—and especially the dissimilarity—between these two theories, and their predictions regarding bidirectionality. Secondly, as this special issue demonstrates, the interdisciplinary aspect of this intellectual history continues in current scholarship and research in metaphor. The essays in this issue present points of view from semiotics, psycholinguistics, linguistics, cognitive psychology, computer science and literary criticism, even while ultimately downplaying such disciplinary distinctions. Thus, for example, the next two essays are not entrenched in their respective disciplines of psycholinguistics (Katz and AlAzary) and linguistics (Porat and Shen), but rather bear as much relevance for the fields of cognitive psychology and for literary criticism as they do for the study of verbal metaphor on which the authors focus. Our third objective in editing this special issue is to highlight different methods of research: theoretical analyses, empirical studies, and textual analysis appear in various essays, and sometimes all within the same essay. This is clearly the case for the final three essays in this issue (Indurkhya and Ojha; Freeman; Goodblatt and Glicksohn); both theorists and those conducting empirical research in this domain should benefit from seeing how each type of scholarship can inform the other. Fourthly, visual metaphor, verbal metaphor and adjacent phenomena such as simile, synesthesia, and analogy, are all analyzed and demonstrated in the various essays. Thus, bidirectionality is not analyzed solely with respect to conceptual metaphor, and in turn with respect to Blending Theory, but also with respect to a wider frame of reference. Goodblatt and Glicksohn † Bidirectionality and Metaphor 11 Downloaded from by guest on 21 September 2021 What is more, researchers studying bidirectionality in metaphor should be acquainted with bidirectionality in synesthesia, as presented in the sixth essay in this issue (Anaki and Henik). At the same time, researchers studying synesthesia will now be able to view their work within this wider context of interest. Finally, our fifth goal in editing this special issue was to chart out paths for future research based on the scholarship presented here. Various ideas might come to mind on reading these essays: Can synesthetic metaphor inform our understanding of bidirectionality? If unidirectionality is a linguistic constraint of verbal metaphor, is bidirectionality encouraged by visual metaphor? What are the implications for the study of bidirectionality in metaphor for a specific poetic text—and a specific literary tradition? If our readers come away with an appreciation of the importance of bidirectionality in metaphor, as well as with questions they would like to investigate in their own work, then our objectives will have been fulfilled. References Anaki, David, and Avishai Henik 2016 “Bidirectionality in Synesthesia and Metaphor.” Special issue, Poetics Today 38, no. 1: 141 – 62. Black, Max 1962 Models and Metaphors: Studies in Language and Philosophy (Ithaca, NY: Cornell University Press). 1979 “More about Metaphor.” In Metaphor and Thought, edited by Anthony Ortony, 19 – 43 (Cambridge: Cambridge University Press). 1981 “Metaphors We Live By by George Lakoff, Mark Johnson” (Review), Journal of Aesthetics and Art Criticism 40, no. 2: 208 – 10. Chiappe, Dan L., John M. Kennedy, and Penny Chiappe 2003 “Aptness Is More Important than Comprehensibility in Preference for Metaphors and Similes,” Poetics 31, no. 1: 51 – 68. Chiappe, Dan, John M. Kennedy, and Tim Smykowski 2003 “Reversibility, Aptness, and the Conventionality of Metaphors and Similes,” Metaphor and Symbol 18, no. 2: 85 – 105. Danesi, Marcel 2016 “The Bidirectionality of Metaphor.” Special issue, Poetics Today 38, no. 1: 15 – 34. Fauconnier, Gilles, and Mark Turner 1998 “Conceptual Integration Networks,” Cognitive Science 22, no. 2: 133 – 87. 2002 The Way We Think: Conceptual Blending and the Mind’s Hidden Complexities (New York: Basic Books). Forceville, Charles 1996 Pictorial Metaphor in Advertising (London: Routledge). 2004 “Book Review: The Way We Think: Conceptual Blending and the Mind’s Hidden Complexities,” Metaphor and Symbol 19, no. 1: 83 – 89. Freeman, Margaret H. 2016 “Multimodalities of Metaphor: A Perspective from the Poetic Arts.” Special issue, Poetics Today 38, no. 1: 61 – 92. 12 Poetics Today 38:1 Downloaded from by guest on 21 September 2021 Gentner, Dedre, Brian F. Bowdle, Phillip Wolff, and Consuelo Boronat 2001 “Metaphor Is Like Analogy.” In The Analogical Mind: Perspectives from Cognitive Science, edited by Dedre Gentner, Keith J. Holyoak, and Boicho N. Kokinov, 199 – 253 (Cambridge, MA: MIT Press). Gibbs, Raymond W. 1994 The Poetics of Mind: Figurative Thought, Language, and Understanding (New York: Cambridge University Press). Glicksohn, Joseph, and Chanita Goodblatt 1993 “Metaphor and Gestalt: Interaction Theory Revisited,” Poetics Today 14, no. 1: 83 – 97. Glucksberg, Sam, and Boaz Keysar 1990 “Understanding Metaphorical Comparisons: Beyond Similarity,” Psychological Review 97, no. 1: 3 – 18. Goodblatt, Chanita, and Joseph Glicksohn 2003 “From Practical Criticism to the Practice of Literary Criticism,” Poetics Today 24, no. 2: 207 – 36. 2010 “Conversations with I. A. Richards: The Renaissance in Cognitive Literary Studies,” Poetics Today 31, no. 3: 387 – 432. 2016a “Bidirectionality in Poetic Metaphor: William Carlos Williams and Imagist Poetry,” Versus: Quaderni di Studi Semiotici, 122. 2016b “Discordia Concors and Bidirectionality: Embodied Cognition in John Donne’s Songs and Sonnets.” Special issue, Poetics Today 38, no. 1: 163 – 88. Grady, Joseph E., Todd Oakley, and Seana Coulson 1999 “Blending and Metaphor.” In Metaphor in Cognitive Linguistics, edited by Raymond J. Gibbs Jr. and Gerard Steen, 101 – 24 (Amsterdam: John Benjamins). Harpham, Geoffrey Galt 1982 On the Grotesque: Strategies of Contradiction in Art and Literature (Princeton, NJ: Princeton University Press). Indurkhya, Bipin 2006 “Emergent Representations, Interaction Theory, and the Cognitive Force of Metaphor,” New Ideas in Psychology 24: 133 – 62. Indurkhya, Bipin, and Amitash Ojha 2016 “Interpreting Visual Metaphors: Asymmetry and Reversibility.” Special issue, Poetics Today 38, no. 1: 93 – 122. Jackendoff, Ray, and David Aaron 1991 “More than Cool Reason: A Field Guide to Poetic Metaphor by George Lakoff; Mark Turner,” Language 67, no. 2: 320 – 38. Katz, Albert N., and Hamad Al-Azary 2016 “Principles That Promote Bidirectionality in Verbal Metaphor.” Special issue, Poetics Today 38, no. 1: 35 – 60. Kittay, Eva F. 1987 Metaphor: Its Cognitive Force and Linguistic Structure (Oxford: Clarendon). Lakoff, George 1998 “Cognitive Semantics in the Heart of Language: An Interview with George Lakoff,” Fo´rum lingu¨ístico 1, no. 1: 83 – 119. 2014 “Mapping the Brain’s Metaphor Circuitry: Metaphorical Thought in Everyday Reason,” Frontiers in Human Neuroscience 8: article 958. .2014.00958. Lakoff, George, and Mark H. Johnson 1980 Metaphors We Live By (Chicago: University of Chicago Press). Lakoff, George, and Mark Turner 1989 More than Cool Reason: A Field Guide to Poetic Metaphor (Chicago: University of Chicago Press). Goodblatt and Glicksohn † Bidirectionality and Metaphor 13 Downloaded from by guest on 21 September 2021 Lee, Spike W. S., and Norbert Schwarz 2012 “Bidirectionality, Mediation, and Moderation of Metaphorical Effects: The Embodiment of Social Suspicion and Fishy Smells,” Journal of Personality and Social Psychology 103, no. 5: 737 – 49. Marquette, Scarlet 2007 “Metaphors We Lie by: Cognitive Blending in the Poetry of Elena Shvarts,” Slavic and East European Journal 51, no. 4: 693 – 715. Nerlich, Brigitte, and David D. Clarke 2000 “Blending the Past and the Present: Conceptual and Linguistic Integration, 1800 – 2000,” Logos and Language 1, no. 1: 3 – 18. 2001 “Mind, Meaning, and Metaphor: The Philosophy and Psychology of Metaphor in Nineteenth-Century Germany,” History of the Human Sciences 14, no. 2: 39 – 61. Porat, Roy, and Yeshayahu Shen 2016 “The Journey from Bidirectionality to Unidirectionality.” Special issue, Poetics Today 38, no. 1: 123 – 40. Richards, Ivor Armstrong 1936 The Philosophy of Rhetoric (New York: Oxford University Press). Thomson, Philip 1972 The Grotesque. The Critical Idiom 24 (London: Methuen). Tourangeau, Roger, and Robert J. Sternberg 1982 “Understanding and Appreciating Metaphors,” Cognition 1, no. 1: 203 – 44. Werner, Heinz 1978 “The Concept of Development from a Comparative and Organismic Point of View.” In Developmental Processes: Heinz Werner’s Selected Writings, edited by Sybil S. Barten and Margery B. Franklin, 107 – 30 (New York: International Universities Press). Werner, Heinz, and Bernard Kaplan 1963 Symbol Formation: An Organismic-Developmental Approach to Language and the Expression of Thought (New York: Wiley). 14 Poetics Today 38:1 Downloaded from by guest on 21 September 2021

An enactive and dynamical systems theory account of dyadic relationships

Logo of frontpsychol

Link to Publisher's site
 2014; 5: 452.
Published online 2014 May 30. doi: 10.3389/fpsyg.2014.00452
PMCID: PMC4038949
PMID: 24910623

An enactive and dynamical systems theory account of dyadic relationships


Many social relationships are a locus of struggle and suffering, either at the individual or interactional level. In this paper we explore why this is the case and suggest a modeling approach for dyadic interactions and the well-being of the participants. To this end we bring together an enactive approach to self with dynamical systems theory. Our basic assumption is that the quality of any social interaction or relationship fundamentally depends on the nature and constitution of the individuals engaged in these interactions. From an enactive perspective the self is conceived as an embodied and socially enacted autonomous system striving to maintain an identity. This striving involves a basic two-fold goal: the ability to exist as an individual in one’s own right, while also being open to and affected by others. In terms of dynamical systems theory one can thus consider the individual self as a self-other organized system represented by a phase space spanned by the dimensions of distinction and participation, where attractors can be defined. Based on two everyday examples of dyadic relationship we propose a simple model of relationship dynamics, in which struggle or well-being in the dyad is analyzed in terms of movements of dyadic states that are in tension or in harmony with individually developed attractors. Our model predicts that relationships can be sustained when the dyad develops a new joint attractor toward which dyadic states tend to move, and well-being when this attractor is in balance with the individuals’ attractors. We outline how this can inspire research on psychotherapy. The psychotherapy process itself provides a setting that supports clients to become aware how they fare with regards to the two-fold norm of distinction and participation and develop, through active engagement between client (or couple) and therapist, strategies to co-negotiate their self-organization.

Keywords: enactive self, distinction and participation, dynamical systems theory, well-being in relationships, the self as attractor, relationship dynamics, couples counseling


Many social relationships are a locus of struggle and suffering, either at the individual or interactional level. Dyadic exchange and the question of well-being in relationships constitute the core of the psychotherapeutic process, as well as the content of most narratives processed in everyday life. Our goal is to better understand why some couples manage to sustain their interactions whereas others terminate their relationships. We also wish to generate ideas for improving the quality of dyadic interactions and the psychological well-being of the participants. To this end we conjoin a dynamical systems theory perspective with an enactive approach to self and explore the dynamics underlying struggle in couples’ relationships.

Dynamical systems theory (DST) is a branch of mathematics, and as such neither part of the natural sciences nor of the humanities (). Its concepts, heuristics and methods can be used to interrelate theories and findings of the various disciplines and to facilitate the dialogue between them. DST describes the complex behavior of systems over time. It allows us to interrelate experiential findings associated with relationship struggle and to derive implications for improving dyadic interaction and enabling relationships.

However, before assessing problems at the level of the interaction we should clarify our understanding of the individuals involved in it. We need to reconsider their basic nature as individuals and what drives their behavior. We suggest characterizing the individuals in the dyad from an enactive perspective, according to which every individual self is genuinely social and purposeful. The enactive self is social because it exists through engagements with others, and it is purposeful because it thereby strives to survive as a social existence. The self follows a primordial two-fold existential norm: being distinct from, as well as connected to, others. We propose that such a basic normative structure of self exists in all individuals. It guides their behavior and how they evaluate and negotiate their relationships.

Our strategy is as follows. We begin with a brief summary of the enactive self in Section “Distinction and Participation: An Enactive Approach to Self.” Based on this, as an intermediate step, we conceptualize in Section “Socially Enacted Autonomy from a Dynamical Systems Theory Perspective” the enactive self in terms of dynamical systems theory as a non-linear dynamical system. In Section “Dyadic Relationship as Negotiation of Individual and Dyadic Attractor Regions” we introduce two everyday examples of couple relationships using our concepts to describe the dynamics underlying the struggle in these interactions, and to arrive at two simple models of relationship maintenance. In Section “Discussion” we compare the two examples and derive two styles of individual relationship engagement, the passive-closed and active-open style, hypothesizing that the latter is more apt to sustain a relationship and to improve well-being in a dyadic relationship. In the last part we outline how the findings in this paper may inspire research in psychotherapy.


In this section we provide a short summary of the enactive approach to self as a social autonomous system (Kyselo, submitted), a recent development in enactivism. Enactivism is a non-reductionist and integrative epistemological framework for cognitive science that adopts a process-based and biologically grounded perspective on cognition (). It is rooted in the theory of autopoiesis and the idea that living beings can be minimally characterized as self-producing and self-organizing networks of biological processes that create a systemic identity (). Enactivism assumes that biological and mental phenomena are continuous and that therefore the identity of cognitive beings can be conceived as based on similar principles and concepts (). It is thus inspired by the autopoietic idea of self-generated identity, but elaborates on this concept by suggesting the more general notion of autonomy to capture not only biological but also cognitive individuation (). In the enactive view on autonomy there is no clear-cut separation between individual system and environment. Cognitive individuals emerge from active engagement with the environment through which they self-produce an identity. They thereby follow an intrinsic purpose, namely to survive and to maintain their self-generated identity (). This implies a basic tension in the individual: a need to emancipate oneself from the environment as an individual, while at the same time having a structural dependence on it for material resources ().

Through being self-organized in this way, individuals always have their own basic perspective on the world, i.e., they evaluate their interactions with the world according to what these interactions mean with regards to the goal of generating and maintaining an identity. The enactivists call this sense-making, the value-driven active engagement with the environment that in turn creates meaning for the system itself ().

The enactive view on cognitive individuation has been recently elaborated to inspire a new look at the human self (Kyselo, submitted). According to this, the self is essentially a phenomenon of life and a question about the nature of human cognitive individuation. Usually the processes of cognitive individuation have been characterized in terms of embodiment (), but according to the enactive perspective on the human self the body is not the sole source of individuation. The world of humans is a world of others, so our social relations are what matter most to us. Much in line with theories of self that emphasize the social, processual or dialogical nature of self () the enactive approach thus assumes that the social must play a vital role in any description of human cognitive individuation.

The enactive self is operationally defined as a socially enacted autonomous system, whose systemic network identity emerges as a result of an ongoing engagement in social interaction processes that can be qualified as moving in two opposed directions, distinction and participation (Kyselo, submitted). On this account, the self as identity is continuously co-generated through interacting and being related to others and at the time organizes interactions and relations. The individual self is therefore never fully separable from the social environment. It is determined precisely in terms of the types of social interactions and relations of which it is also a part. Yet in order to exist as an identifiable unity, the self also involves an ongoing process of emancipation from others. This basic tension between dependence and emancipation is primordial to the nature of human individuation, and for this reason, it is considered a fundamental drive for human behavior. Whereas living systems strive to survive by avoiding interactions with the environment that threaten their biological survival, the purpose of the human self is additionally to ensure its identity and survive as social existence1. In line with the enactive perspective on autonomy, every individual self thus has its own subjective perspective on the world, a perspective from which social interactions and relations are evaluated according to whether and how they serve the survival, i.e., maintenance, of the self. This maintenance follows a two-fold basic norm that mirrors the tension of the social individual to emancipate itself from the social environment while at the same time structurally relying on it: being able to exist as individual in one’s own right (distinction) while at the same time remaining connected with others (participation). Distinction means that the person experiences herself as both emancipated and yet not fully independent of the social world. Participation means that she feels both connected and open to, but also not fully immersed in, the social world. Both dimensions can overlap. Distinction does not imply that the person does not interact or has to be alone (think of the familiar experience of feeling alone despite being surrounded by others). Participation, on the other hand, does not imply that the person must interact all the time. A person can feel very open or related to another person even when not actually engaged with her. Both distinction and participation are (experienced) types of social interactions and relations, yet they may say little about the amount or actuality of engagement. In every individual the amount and distribution of distinction and participation can come in different degrees: some individuals have a generally strong sense of being an individual in their own right, but feel not so open to others, while others feel equally open to others.

Throughout an individual’s life there are phases when one may feel or strive for more or less distinction or participation; as a child, for instance, there might be a stronger openness to being affected by others and a lower experience of being separated, whereas during adolescence feelings of being or wanting to be separated are more dominant. Distinction or participation furthermore depend on a given cultural context, and on whether a greater value of one or the other is developed because it is socially more accepted (). Furthermore, even though at times one of the dimensions may become extremely dominant and the other appears out of reach, the other dimension can or will, at some point or implicitly, drive the individual’s behavior. Thus, for example, feeling very distinct at some point does not mean that there is no striving for connection and openness anymore.

An excessive degree of distinction would mean that the individual has no sense of openness or connection to others, while excessive participation would mean that the individual is completely immersed in the interaction. Humans thus strive to avoid the double risk of emancipation at the cost of being isolated and of connection at the cost of dissolution of the individual self. We can find examples that approximate such extreme cases in disorders of the self such as schizophrenia () and symptoms like social or self-isolation (no participation) or loss of agency (no distinction). But even though these cases are exceptional, the suffering that accompanies these extreme states could actually be indicative of a persisting striving to balance both dimensions.

Importantly, the maintenance of the self according to the two-fold normative structure of distinction and participation requires constant negotiation with others, that is, engaging with, and disengaging from them. The self is thus co-generated with others and since interactions with others can go wrong and fail to contribute to identity maintenance in the desired way, the self is genuinely vulnerable.

In the next section we conceptualize this view on self in terms of dynamical systems theory and then derive a model of interaction dynamics between two selves.


Dynamical systems theory (DST) allows describing a system at two levels: by variables that denote state changes of systems over time, and by parameters that constrain these changes (gradients). DST has been used in cognitive science to replace the input/output model of cognition, and propose a context- and time-sensitive account of cognition () and neural dynamics (). More recently, researchers in enactive cognitive science have appealed to DST to describe mind, social interaction, and sensorimotor skill-use (). In this section we use DST to conceptualize the self as socially enacted autonomy.

We begin with a brief reminder of some of the main concepts of DST. A core notion especially relevant for our purposes is the concept of attractor, which can be formally defined in terms of the phase space of a system. The phase space is a geometrical space with one or more dimensions, depending on the number of variables needed to fully describe the system (). This can be exemplified by Euclidean space: Euclidean phase space has three spatial dimensions, i.e., the coordinate axes xy, and z. Any state of an object situated in Euclidean phase space can simply be described by the three spatial coordinates. DST captures not only a particular state of the system, but also how the system changes in time.

Imagine a golf ball being hit, flying through space, landing on grass, rolling and ending up in a well or the hole of a golf course. The golf ball represents a simple dynamical system. While flying through the air, rolling on the grass, etc., the ball follows a specific trajectory (its flight or path) through Euclidean space (the three dimensions of space). Until it ends up at the bottom of a hole, the ball trajects the three dimensions of space (x, y, and z), changing its states over time. The state of the ball in the hole, after it has come to a rest, can be represented as a particular single point in phase space, to which the three dimensions of the system’s trajectory through space have converged. Assuming that the ball will always end up in the hole after many different trajectories, we would arrive at a simple illustration of a dynamics in which the system’s three dimensions are always compressed toward zero dimensions, which is indicative of so-called attractor dynamics. The unchanging, stable state, such as in the bottom of the hole in golf, is referred to as a point attractor.

As the example of the golf ball illustrates, DST concepts allow visualizing complex temporal behaviors of particular systems in terms of geometrical representations. We use the notions of phase spacetrajectory and attractor to conceptualize the states and changes of a human self.

From a DST perspective the human self can be seen as a non-linear dynamical system that displays a particular behavior represented as movement through the “landscape” of phase space (). The phase space of the enactive self refers to the states of the self as created in its relations with the social environment. It consists of representations of social interactions and relations, covering idealized engagements with and disengagements from others throughout a life-time. The self’s phase space is therefore a space of the two fundamental variables of distinction and participation as introduced in the previous section.

In order to define the phase space of an enactive self at the most general level we abstract over all possible variations of distinction and participation (individually preferred ranges, different cultures, at different times of life) and use distinction and participation as variables D and P. This is in line with abstract conceptualizations of psychological phenomena, such as Kurt Lewin’s topological psychology (), in which personality and social relations are modeled in terms of regions and barriers in ‘life space’ (). Our model of phase space may also be associated with theoretical psychology (), according to which personality involves an interpersonal space that is similarly constituted by two dimensions, agency and communion (). In terms of psychological development this resonates with the work of Mahler, who described the infant’s self development as a process of individuation and separation through which the infant’s self emerges subsequent to a post-natal period of symbiotic relation with the mother ().

The variables D and P span the self’s phase space, which can thus be illustrated as a plane (Figure Figure11). We denote the states of the self by their locations in this plane (D/P). The higher the value of D, the higher the system’s distinction, and vice versa, the higher the value of P, the higher is the system’s participation. Since the enactive self is always relational, neither D nor P can ever have a value of zero. In addition to its value of D and P, each point (D/P) of the plane has a positive or negative “elevation,” so that the corresponding slope represents repulsion from or attraction toward this point. The self’s behavior is represented by trajectories, i.e., sequences of states in the phase space. Over time a self develops particular tendencies to balance D and P (becoming more or less distinct from others and more or less open to them). In terms of DST we can say that the trajectories tend toward particular regions of the phase space. When these tendencies become manifest, these regions can eventually exhibit attractors we define in terms of a particular value D/P. The attractor regions symbolize the location of the individual’s developed and preferred zones of functioning (balancing D and P). The phase space of each self has regions with negative elevations (“wells” or “troughs” in the landscape of phase space) that indicate attractors toward which the system will move. Some regions are positively elevated and thus repulsive, which indicates so-called repellors (the opposite of attractors). Attractors and repellors emerge due to habitual tendencies and represent goals of the self throughout a particular time span. Attractors of distinction and participation must not mean that the individual evaluates them as positive. Such tendencies and goals may emerge due to the individual’s increased well-being in that region, but they may also emerge due to system-external reasons (not controlled by the individual) or habitual tendencies with a negative connotation.

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g001.jpg

An illustration of two generic individual phase spaces of the enactive self. The individual self states are symbolized as red and blue balls. Their respective phase spaces are spanned by the variables of Distinction D and Participation P. The attractors are wells in D/P landscape, into which the self systems move. The red and blue arrows indicate the trajectories of these movements. The graphic on the left side illustrates a self (red ball) with an attractor with a greater value of D and a lower value of P. The graphic on the right side illustrates a self (blue ball) with an attractor with a relatively high value of P and a lower value of D.

Attractors may exist in different regions with different values D/P. They represent that an individual self has developed or exhibits a certain degree of distinction in combination with a certain degree of participation. Consider a person who developed a strong preference for high distinction and low participation for particular life times. For example, a novelist who at some point during the writing process escapes her social life, locks herself up in a remote and quiet place in the mountains to finish her new book. The novelist has experienced this kind of solitude as useful, and so whenever she writes a book, she retreats to the cabin. In terms of DST, during the book writing phases the novelist’s phase space shows a particular attractor D/P with a higher value of D and a lower value of P than usually. We may imagine the system starting at some point in D/P landscape. The inclination of the landscape at this point will then determine the direction of the trajectory, which is generally away from repellors and toward the deepest points of attractors. The system will change its distinction and participation values until it has reached the point attractor D/P (solitude and a minimum of engagement with others). The system will remain in the attractor region unless the phase space changes or until perturbations external to the system exert an influence.

To describe how systems’ tendencies change through perturbation, we can refer to another concept in DST, the gradients. Gradients are often referred to as control parameters, but in the context of dynamics of the self’s organization this term, borrowed from physics, is misleading. Gradients here are the environmental fluxes and affordances that drive, but do not control, a system’s self-organization. At certain critical values of a gradient, a system may enter a novel, emergent state of its dynamics, and hence its phase space landscape may become completely modified. Because of these changes new attractors may arise, so that the phase space of the enactive self should be conceived of as a flexible landscape. In case of the enactive self, gradients can refer to the social environment. Imagine for example that during the writer’s exile a friend in need reaches out asking for support in a difficult matter. This perturbates and might also change the novelist’s current states in D/P landscape. In tension to her initially preferred region of low P she reacts to the friend’s perturbation by turning toward a region with higher P values, thus adapting the current range of preference in D/P. Another example is attending some party or other obligatory social gathering, when one would actually prefer spending a quiet evening at home. Because the social event requires higher values of P it can signify a strong perturbation to the current disposition of D/lower P. The prospects of attending the party can therefore cause tension and actual struggle while being there.

Strong emotions and other motivational parameters may also be gradients that affect the self. For example, when the book is finished, the novelist’s attractor may shift back to a different region, with a higher value of participation. Here the gradient is motivational (the author realizes that the book is finished). Persistent sadness may for instance change a person’s D/P landscape and entail avoiding connection with others, shifting her states toward lower P values. Gradients can thus perturbate and change the values of the self’s attractor.

A formerly active attractor may “close,” leaving behind a quasi-attractor in the same location (). An example of this is the perception of bistable stimuli, such as Rubin’s vase-face figure (Figure Figure22). When the first perception is that of a white vase, i.e., the perceiver rests in the attractor “vase,” this will eventually give way to the new perception “black faces.” In terms of DST, the system has altered its display of attractors and the landscape of phase space has changed (). By staying in one attractor (“vase”), the attention gradient that created this attractor becomes depleted (), so that the system will explore other regions of phase space to eventually settle in a different attractor (“faces”).

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g002.jpg

A bistable stimulus according to .

To sum up, we conceptualized the enactive self in terms of DST as an attractor landscape in a social phase space that is organized by the two variables of distinction and participation. Individual changes of distinction and participation are expressed in terms of trajectories through this landscape, and strong behavioral tendencies in terms of attractors D/P. The stability of such patterns of the self are constantly maintained and calibrated by external and system-inherent gradients, such that changes of gradients will generally change the self’s whole landscape. In the following section we explore two examples of dyadic interactions on the basis of these considerations.


In this section we use the conception of the individual self in terms of a D/P attractor dynamics for understanding dyadic relationships. We will introduce two everyday examples of relationship struggle, one in which interaction leads to a breakup the other in which interactions are sustained. We conceptualize the two couples in terms of DST as a dyadic relation between two individual phase spaces. That dyadic relationship can be described as a new kind of dynamical autonomous system (). We conceive of it as a new dynamical system with a phase space that corresponds to sustained interactions between the individuals in the relationship, a joint phase space.

For reasons of simplicity, we assume that the formation of the couple’s joint phase space is a summation of the phase spaces of the individuals: we thus add the elevation values of the individual phase spaces in each point of D and P. This means that when both participants previously had an attractor in the same region of their individual phase spaces, their dyadic joint phase space will have an even deeper attractor in this region.

We then assume that at each point in time the states of the interaction dynamic, represented through particular locations in the dyad’s phase space, affect the partners, in that they act as perturbations on their individual phase spaces D/P. Such perturbations occur at all times during the relationship. It will be a task for the future to elaborate more concrete structures, but we offer a first idea of how a joint state could affect an individual. Firstly, interactions can perturbate one or both dimensions of the individual’s developed or preferred range D/P, distinction and/or participation (they can act as gradients). Secondly, not every perturbation must lead to change in a current state or developed attractor D/P. Thirdly, it will depend on the frequency and the quality of particular interactions or patterns of interactions whether and how each state or attractor is affected. We can assume that for each dimension D or P there will be interaction qualities that currently matter more or less. For example, interactions that are too frequent and aggressive, or not frequent and gentle enough, may perturbate stronger on the dimension of P (openness) in some individuals, while interactions bringing forth a pattern of belittlement and shame on the one hand, or praise and recognition on the other, may be more relevant to the dimension of D (distinction). Whether and how much of the quality of any of such interactions perturbates D or P depends on the individuals. In the following conceptualization of two case examples in terms of DST we chose to refrain from more precise description and restrict the analysis to a fairly general level of interrelating individual and joint action. It will provide a very basic answer to our question: why do couples struggle and what constitutes well-being in a relationship? Each example is approached based on two basic questions: firstly, how the individuals’ particular negotiation tendency, i.e., their respective range of distinction and participation initially match, and secondly, whether and to what extent the actual interaction allows the participants to maintain or to negotiate their individual goals of balancing D and P.

Example 1

She, an artist, has been exploring her inner experiential world continuously in recent years investing considerable time and effort in various practices of mindfulness such as yoga and meditation. Although she is, as a performing artist, used to present herself on stage, she is careful about the exhibition of her private self outside of the roles on stage. She is generally rather inhibited to engage in an intimate relationship. He, a scientist, is used to communicate his personal projects in public and has a strong communion motif privately, being eager to engage in an intimate relationship. After the two met at a workshop and with him taking the initiative, the two soon enter an intense romantic relationship. The initial months of the relationship are full of frequent meetings in a highly participatory mode. Soon, after a few months however, she begins to feel pressured and cornered by him. She experiences fears and crises, mainly as attempts to emancipate herself from the risk of becoming too dependent on him. “Isn’t it true that people should learn to love themselves first?” she asks. To accommodate her struggle the couple decides to try a more detached and individualized style of relationship allowing her to spend more time by herself. After a while however, he begins to feel unhappy about the lack of frequency and intensity of their meetings. In his view they do not see each other enough. The next months of the relationship show a continuous oscillation between attempts to accommodate her need for more time by herself and his need for more time being together. While he experiences her effort for emancipation as too high, she experiences his effort for being together as too much. She insists that over-attachment to the other is not love. He complains that she is pushing him away. They have repeated arguments about the meaning and goals of a relationship, their attempts at improving the relationship do not reach consensus resulting in continuous emotional dissonance for both of them. The couple eventually splits up after about 1 year of being together.

Let us begin with the first question, the individuals’ general tendencies of interrelating D and P with regards to the prospective romantic relationship before they enter the relationship. Based on the above case we derive that she has a stronger tendency toward distinction and toward a sense of self as being a separated individual, whereas his profile shows tendencies in the opposite direction, toward a more participatory mode of identity construction. We can state that the individuals’ attractors dwell in different regions of phase space: as an individual, and with regards to romantic relationship, her attractor resides in a region with a greater value of D and a lower value of P. His attractor is in a region with low distinction and high participation, featuring a lower value of D and a greater value of P. Whenever these two individuals start from mid-range values of distinction and participation, the joint trajectory heads in opposition to her or his previous trajectory. This couple corresponds to the phase spaces of the example that we have given in the previous section “Socially Enacted Autonomy from a Dynamical Systems Theory Perspective” (Figure Figure11).

Let us now consider what happens when the two individuals in the couple of example 1 enter a relationship. To this end we thus create a joint phase space adding the individuals’ preferred attractors of D/P. The new dyadic phase space thus exhibits two attractors that correspond to the former individual attractors (Figure Figure33). Dividing the time of their relationship into temporal windows, we look at three states this coupled system goes through: t1, initial months; t2, adjustment phase I; t3, adjustment phase II.

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g003.jpg

Illustration of the joint phase space of the couple in example 1. The dyadic phase space is a summation of the individual phase spaces (his low D and high P and her high D and low P, see Figure Figure11). The graphic illustrates the couple’s states during the initial months of the relationship (t1). The green ball symbolizes that they move in a region with higher values of participation.

At t1, corresponding to the closeness and intensity experienced in the initial phase of their relationship, we see that the dyad’s states tend to reside in a range close to high values of participation and lower of distinction (Figure Figure33). This is in accordance with his previous individual attractor that showed a greater value of participation. It is in dissonance with her previous attractor that had a higher value of distinction.

At t2 the dyad’s state resides in a new region within D/P plane migrating to an attractor with higher distinction and lower participation levels. At the individual level this means that the dyad’s trajectory thus moves closer to her individual range of preference and farther away from his. However, the system does not remain in this region but moves back again to the previous region of higher distinction and lower participation levels, in accordance with her and in tension with his individual preference.

Subsequently, during the adjustment phase II, the dyad’s states keep oscillating between the two opposite attractors. Except for t1 the trajectories of the dyad never persistently overlap with the individually preferred ranges (Figure Figure44). One might describe this behavior in terms of a bistable quasi-attractor dynamics, as in the example of Figure Figure22. The dyad depletes a current attractor and subsequently revives a former quasi-attractor in the sense of , to then again deplete it and revive the previous one. The couple’s transients between the two attractors eventually result in a collapse of the system at t3.

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g004.jpg

2D illustration of the joint phase space of the couple described by example 1. The attractors are depicted in blue, the repellors in red. Attractors are circled showing that there was no overlap of the basins of the two attractors.

Example 2

She, an artist, has been exploring her inner experiential world continuously in recent years investing considerable time and effort in various practices of mindfulness such as yoga and meditation. Although she is, as a performing artist, used to present herself on stage, she is more careful about the exhibition of her private self outside of the roles on stage and more reluctant to engage in an intimate relationship. He, a scientist, is used to communicate his personal projects in public and also has a strong communion motif privately being eager to engage in an intimate relationship. However, he also likes spending a lot of time by himself. Since the beginning of their relationship the couple experiences short-lasting but intensive crises. In these crises she feels pressure and fears of being overwhelmed and losing control. She would like to be by herself but at the same time she does not want to leave the interaction, afraid to lose the connection or to hurt him. He usually is shocked at the expression of her discontent and feels overwhelmed or afraid of failing to please her. At the same time he also experiences a strong pull to stay in the situation with her, either because he is afraid to hurt her or to lose her. Both are convinced that an intimate relationship requires efforts on both sides and so they try different strategies to deal with their crises. Occasionally the couple decides to briefly interrupt the interaction trying to become aware of individual feelings without worrying what the other does. At other times, overcoming feelings of panic and losing control, they are open and trusting toward the other and remain in the interaction. Both experience these phases as difficult and feel strong emotional dissonance. But they also learn that momentary disconnection does not necessarily threaten their relationship and that what initially seemed frustrating can actually lead to a better mutual understanding. The couple experiences this as nourishing and as deepening their connection.

Let us begin with the first question, the individuals’ general tendencies of interrelating D and P with regards to the prospective romantic relationship before they enter the relationship. Based on the above case we derive that she has a strong tendency toward distinction and toward a sense of self as being a separated individual whereas his profile shows tendencies in the opposite direction, toward a more participatory mode of identity construction. However he also shows relatively high tendencies toward distinction. The individuals thus have different preferences in negotiation of distinction and participation, i.e., the attractors of the individuals are in different, but not opposite regions of phase space: she has a high D/low P attractor, and a repellor at low D/high P. The repellor represents her inhibition for highly participatory states when the range of distinction is low. His attractor is also at greater values of D together with moderate to high P (Figure Figure55, please note that her attractor is identical to the attractor of the “she” protagonist of example 1, cf. Figure Figure11 left), whereas his attractor slightly differs in the two narratives.

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g005.jpg

Illustration of the phase spaces of the individuals in example 2 (left: “She”; right: “He”). The phase spaces are spanned by the dimensions of Distinction D and Participation P. Individual system states are symbolized as locations of red and blue balls. The attractors are wells in the D/P landscape, into which the self systems (balls) tend to move. Arrows indicate trajectories from two arbitrary starting points.

Let us now describe the situation once the individuals of example 2 enter a relationship and the individual phase spaces are merged into one joint phase space (Figure Figure66). Corresponding to the couple’s several instances of crises, the dyad’s states in example 2 oscillate between the two attractor regions. The dyad’s behavior thus shows similarity to that of example 1. However, the transients between the “deepest” points of the attractors here are considerably shorter than in the dyad of example 1. Even though the oscillations occur between different levels of participation, the individuals show an overlap in their previous attractors with a high value of distinction. The couple in this example thus has a region in which the individuals share individual preferences. In terms of DST this is to say that the basins of the two individual attractor regions create an intersection, i.e., a region of overlap (Figure Figure77). Such connections between point attractors are called “saddles” (Figure Figure66). If the couple continues to sustain interactions leading to an overlap of their attractors, a saddle could “deepen” and turn into a new, jointly created attractor indicating the couple’s sustained interaction tendencies.

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g006.jpg

Illustration of the dyadic phase space of example 2. The dyadic phase space corresponds to the individual phase spaces from Figure Figure55. The couple’s state (green ball) is located in the saddle, the region connecting the individual attractor regions.

An external file that holds a picture, illustration, etc.
Object name is fpsyg-05-00452-g007.jpg

2D plane view on the dyadic phase space of example 2. The two attractor regions are depicted in blue, the repellors in red. The attractors are circled. Their inters

Conceptualizing the two relationship examples in terms of dyadic movements away and toward greater distinction or participation that are either in accordance with or deviating from the individually developed attractor, we offer a simple model of co-negotiation of self maintenance in dyadic interaction. In the following section we compare the two couples and discuss what the observed state changes could mean for relationship sustainment and individual well-being.


In example 1 we see that the dyad’s interaction did not lead to a joint region or attractor that was in the same region as the attractors of both individuals. The couple’s states continuously oscillate between two divergent attractor regions. Each attempt to approximate the participants’ respective attractor zone implied a deviation from the developed zone of the other participant. Each experienced deviation was followed by a strong inclination to avoid the jointly enacted quality and to increase it toward the opposite direction and back to an initial preferred range.

Comparing the couple’s states to the individual attractor region we observe that a greater value of D for him is in tension with his preferred zone of well-being that entails lower values of D. Yet when the couple’s states show a greater value of P, then this implies a tension for her. For both individuals the quality of interaction therefore turns out to be in continuous tension with their individual preference for self maintenance (the preferred balance between D/P). The tendencies of the individuals to respond to the tension by fully going back to their own preferred zone of well-being leads to a breakdown of the relationship.

Based on this simple model we hypothesize that individuals whose initial ranges of preferences of distinction and participation are highly opposed are less likely to engage in sustained interactions when for both participants the quality of interaction is in non-negotiable tension to their developed preferences.

In example 2 a joint region (the saddle) was created based on a partial match of the two individual attractor regions and shorter transients back to original individual attractors. As in the previous example the experienced quality of the interaction (more or less D or P) perturbates the individual participants’ preferred range and is in tension with their attractors. However, in contrast to example 1, the individuals do not fully go back to their initial range D/P. Instead they remain within the vicinity of the other’s range of preference. In example 1 the individuals are affected by the interaction and act in accordance with their own individual goals. In example 2 the individuals are affected by the interaction (experienced perturbation to D/P), but they also adapt their own preference in dependence on and through the interaction. As a result, their interaction not only perturbates the individuals’ preferred range but actually alters it. In this example, a higher value of P implied staying within some region of higher P despite a tension with individual preference for low P or, when the interaction showed a higher level of D and this was in tension with a developed low value of D, it implied approaching a higher D than usual. In this way both partners increased the tolerance toward the interaction to act as a gradient on one’s own individually preferred range of D/P. This activity allowed for the development of a new, shared zone of preference.

Based on this simple model we hypothesize that individuals who have attractors that show overlap (here, in the dimension of distinction) are more likely able to negotiate tensions caused by perturbations and to jointly adapt their individually developed attractors so that they allow for sustained interactions.

From the dynamical systems conceptualization of the two relationships we derive two styles of negotiation of D/P. The first style, corresponding to couple 1, shows that both individuals avoid deviations from their original range of preference. We call this style passive-closed as the individuals enter the relationship and react to the tension it creates but do not actively shape the interaction nor adapt their own attractor.

The second style, corresponding to the couple from example 2, shows tolerance for perturbation and potential change for both individuals. We call this style activeopen as the individuals enter the relationship and gradually adapt their movement. An experienced tension is not reacted to independently from the ongoing interaction. The active-open style appears to delegate some of the tension caused by perturbation into the individuals’ joint negotiation of D/P, creating new, shared spaces of balancing their individually developed ranges of D/P.

One can speculate that in this way some interactions can create corrective experiences () to form new, previously unavailable, evaluation and negotiation strategies, thus effectively changing the individuals’ previously developed attractors.

We hypothesize that the active-open style is more likely to ensure well-being in a relationship, i.e., not only that a relationship is sustained but also that the quality of the interaction can meet the needs of both individuals. That said, we do not suggest that the accommodation of participants’ well-being must always imply an ongoing or actual engagement. There are cases, like example 1, where a couple is unable to negotiate the individually experienced tension in a way that still allows for sustaining their relationship. But it is easy to conceive of couples who manage to stay together and are unhappy nevertheless, simply because the negotiation of tension occurs at the expense of the needs of one or both. Negotiation of well-being in a relationship is not only finding continuity in interaction, it is also finding it under specific conditions, namely by considering whether the interaction quality is in tension or in accordance with the individuals’ needs, and their developed preferences for D and P. Compatibility or well-being therefore does not necessarily require individuals to share high values of participation and remain constantly open and ready to be affected by one another. As example 2 shows, it may also involve a greater amount of separateness or even periods of disconnection. In the example both partners might need to strongly feel valuable as a person also independently from the other partner or simply wish to spend more time by themselves. This could make them compatible despite a difference in participation preference. This couple can sustain a relationship with actual interactions and some extent of connectivity but also with spaces of disengagement or disconnection, allowing individuals to experience themselves independently from each other2.

We should emphasize again that our suggestions apply for close relationships and not for every social interaction. There is an abundance of potential and actual social interactions that are not even remotely considered to be relevant for a person’s self maintenance. Certain types of relationships however, such as romantic relationships, friendships or family bonds, but also some relations dictated merely by cultural agreement such as between employer and employee, are usually considered as fundamentally important or closer than others. Based on our model we can speculate that this is the case precisely because they are considered as important sources for self maintenance and spaces for engaging in the existentially needed joint negotiation of both norms of distinction and participation. The more a relationship is deemed to provide such a space the more relevant it will appear. In this sense, being in a relationship is also always an individual choice.

Our account suggests that struggle in a dyadic relationship is in principle unavoidable. This is because any sustained interaction implies that there are two individuals that each have their own goals of social survival and that thus have developed perspectives on how interactions can contribute to them. This leads to constant perturbations that individuals can experience as tension and that can manifest as struggle. Whether or not the couple can maintain the relationship will depend on the individuals’ range of preference and their capacities to tolerate deviations from that range, but also on how the individuals adaptively evaluate and re-evaluate the interaction.


In this paper we conjoined the enactive approach to self with dynamical systems theory to shed light on some basic dynamics underlying struggle and communion in dyadic relationships. We proposed a model of relationship dynamics in terms of a dyadic phase space emerging through the summation of individuals’ phase spaces and assessed struggle or well-being in terms of movements of dyadic states in tension or in harmony with individual attractors. The model predicts that a relationship is sustained when the couple develops a new joint attractor toward which dyadic states tend to move. This is most likely when there is (1) overlap in preferred ranges of distinction and participation in combination with a high estimation of the relationship’s potential to accommodate balancing individual ranges of distinction and participation, as well as (2) an active-open style, in which participants adapt their individual ranges according to their interaction. Because such a relationship has greater potential to meet the needs of both participants to feel more or less connected, and more or less recognized in their own right, it is more likely to lead to well-being.

Presently, we must note some divergences and limitations in conceptualizing an enactive approach to self in terms of DST. In the enactive view, the self is generally co-determined in interaction, and thus already entails perturbations through social interactions. The dimensions of distinction and participation not only mirror the individual’s trajectories but also entail that these trajectories depend on interactions with others. At later stages of the individual’s development, not every interaction matters for self-organization. And yet, at the same time, a self also has developed particular tendencies (dispositions) that constrain to which extent these trajectories are open to perturbations by others, allowing a more flexible evaluation of interactions. Future elaborations on our model have to account for the fact that social interactions and relations matter at different but inextricably linked levels, such as development, dispositional as well as situational enactment of the self. They require clarifications of enactive or dialectical conceptions of identity and the development of corresponding mathematical concepts to arrive at closer approximations for the model and what the model represents. Levins’ work on the relation of dialectical and systemic theory () and Van Geert’s DST approach to cognitive development in children () could serve as inspirations to this end.

Our considerations are exploratory, but we believe they can serve as a starting point to deepen our understanding of the complex interrelation between individual and dyad. They might further help to shed light on interrelations of important phenomena and aspects associated with struggle in dyadic relationships, such as vulnerability and shame, mutual recognition, intimacy, co-dependency, and trauma.

Apart from its potential to assist theoretical integration our proposal may be supported by further quantitative research, for example, through repeated measurements of D and P preferences of people in a relationship. Methods are available for the assessment of communion and agency (the FAMOS: ; the IIP: ), which may be used as an approximation of the enactive concepts proposed here. It also promises applicability to various empirical fields, for example psychotherapy, and a variety of existing methods of measurement could be used or re-evaluated in light of it.

In this vein, a goal of therapy could be to raise individuals’ awareness that they have existential goals (distinction and participation) that are continuously at play and that affect their interactions. At the same time, they should be encouraged to recognize that this equally applies to the partner and that their relationship is thus a jointly negotiated dynamic of their own individual goals (). In therapy a couple’s current relationship status could be assessed in terms of individuals’ current D/P attractors, how the interactions tend to perturbate them, and the strategies that the couple uses to negotiate these perturbations. This could be complemented by an assessment of individuals’ attractors developed before the relationship D/P (e.g., through questionnaires), and by determining the general likelihood of overlap between their attractor regions. For this purpose it might be crucial to evaluate the actual capacities of the individuals for tolerating perturbations and allowing for adaptive change in their developed attractors D/P, taking for instance into account factors such as stress level, emotion regulation capacities, attachment styles, past traumata and how they might constitute hindrances (repellors).

Since psychotherapy is itself a dynamic social interaction, it provides a setting in which participants can develop, through active engagement of client/couple and therapist, novel strategies to co-calibrate their self-organization, i.e., narrow the window of oscillation between the opposing attractors or secure a shared zone of well-being. To this end, especially, systemic or interactional approaches to psychotherapy such as the “open-dialogue” approach (e.g., ) could serve as useful resources. We propose that evaluations and improvements can and should also account for the fact that enactment of relational processes is bodily mediated. Inspiration for reconsidering interventions and assessments in terms of co-negotiation of self maintenance might therefore also come from areas such as mindfulness training, body psychotherapy and dance therapy (e.g., ).

Last but not least, from an ethical point of view our proposal is also meant to encourage a greater tolerance for negativity and struggle as necessary aspects of social life. The self individually is a locus of tension and conflicting tendencies: one needs others, and yet at the same time one also needs to feel capable and recognized independently of them. When two people come together the potential for conflict is increased even more. The recognition that we contribute to one another’s self maintenance, and that this is not an easy endeavor, could be a way of affirming the socially existential basis of life as such. Like life, the self resists rigidity. Like life, it is ever moving and not fully determined as long as it exists. Because of this openness of self, relationship struggle must be a necessary aspect of life.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.


We would like to thank Gabriel Levy and Ezequiel Di Paolo for helpful comments. This work is supported by the Marie-Curie Initial Training Network, “TESIS: Towards an Embodied Science of InterSubjectivity” (FP7-PEOPLE-2010-ITN, 264828).


1This is not to say that the body does not matter but that individuation does not emerge through bodily processes in isolation. The body changes its status becoming a mediator of the self as socially enacted existence. The bodily sense of self can be biologically grounded but its meaning for the self as individual being can only be derived in dependence on how the self is in the world that is, based on and in relation to others. It must be seen as a matrix telling us (e.g., through emotions) how we fare in these interactions with regards to the minimal purpose of human existence: to be someone who is at the same time a person in her own right while also being a person that one can connect to. Bodily experiences thus acquire a social meaning.

2As stated in the famous expression “opposites attract,” individuals can also have no overlap at all and still experience each other as compatible because their different attractors may complement one another. And vice versa, individuals might have a great overlap, sharing high participation and low distinction and still struggle because of a lack of feeling acknowledged as being more than a partner for another person.


  • Abraham R. H., Shaw C. D. (1992). Dynamics – The Geometry of Behavior. Redwood City: Addison-Wesley []
  • Barandiaran X., Rohde M, Di Paolo E. (2009). Defining agency: individuality, normativity, asymmetry and spatio-temporality in action. Adapt. Behav. 17 367–386 10.1177/1059712309343819 [CrossRef[]
  • Buber M. (1947/2002). Between Man and Man. London: Routledge Press []
  • Buhrmann T., Di Paolo E., Barandiaran X. (2013). A dynamical systems account of sensorimotor contingencies. Front. Psychol4:285. 10.3389/fpsyg.2013.00285 [PMC free article] [PubMed] [CrossRef[]
  • Castonguay L. G., Hill C. E. (eds). (2012). Transformation in Psychotherapy: Corrective Experiences Across Cognitive Behavioral, Humanistic, and Psychodynamic Approaches. Washington, DC: American Psychological Association. 10.1037/13747-000 [CrossRef[]
  • Clark A. (2001). Mindware. Oxford: Oxford University Press []
  • De Jaegher H, Di Paolo E. (2007). Participatory sensemaking: an enactive approach to social cognition. Phenomenol. Cogn. Sci. 6 485–507 []
  • Di Paolo E. (2005). Autopoiesis, adaptivity, teleology, agency. Phenomenol. Cogn. Sci. 4 429–452 10.1007/s11097-005-9002-y [CrossRef[]
  • Di Paolo E., Thompson E. (forthcoming) “The enactive approach,” in. In: Shapiro L., editor. The Routledge Handbook of Embodied Cognition. (New York:: Routledge Press); []
  • Di Paolo E., Rohde M, De Jaegher H. (2010). “Horizons for the enactive mind: values, social interaction and play,” in Enaction: Towards a New Paradigm for Cognitive Science eds Stewart J., Gapenne O., Di Paolo E. (Cambridge: MIT Press; ) 33–87 []
  • Froese T, Di Paolo E. A. (2010). Modelling social interaction as perceptual crossing: an investigation into the dynamics of the interaction process. Connect. Sci. 22 43–68 10.1080/09540090903197928 [CrossRef[]
  • Grosse Holtforth M., Grawe K. (2000). Fragebogen zur Analyse motivationaler Schemata (FAMOS). Z. Klin. Psychol. 29 170–179 10.1026//0084-5345.29.3.170 [CrossRef[]
  • Haken H., Tschacher W. (2010). A theoretical model of intentionality with an application to neural dynamics. Mind Matter 8 7–18 []
  • Haken H. (2006). Beyond attractor neural networks for pattern recognition. Nonlinear Phenomena Complex Syst. 9 163–172 []
  • Haken H. (1992). “Synergetics in psychology,” in Self-Organzation and Clinical Psychology eds Tschacher W., Schiepek G., Brunner E. J. (Berlin: Springer; ) 32–54 []
  • Hermans H. J. M., Kempen J. G, Van Loon R. J. P. (1992). The dialogical self. Am. Psychol. 47 23–33 10.1037/0003-066X.47.1.23 [CrossRef[]
  • Horowitz L. M., Wilson K. R., Turan B., Zolotsev P., Constantino M. J., Henderson L. (2006). How interpersonal motives clarify the meaning of interpersonal behavior: a revised circumplex model. Pers. Soc. Psychol. Rev. 10 67–86 10.1207/s15327957pspr1001_4 [PubMed] [CrossRef[]
  • Horowitz L. M., Strauss B., Kordy H. (1994). IIP-D. Inventar zur Erfassung Interpersonaler Probleme – Deutsche Version. Weinheim: Beltz []
  • Jonas H. (1966) The Phenomenon of Life: Toward a Philosophical Biology. Chicago: University of Chicago Press []
  • Kabat-Zinn J. (1994). Wherever You Go, There You Are: Mindfulness Meditation in Everyday Life. New York: Hyperion []
  • Koch S., Morlinghaus K., Fuchs T. (2007). The joy dance. Specific effects of a single dance intervention on psychiatric patients with depression. Arts Psychother. 34 340–349 10.1016/j.aip.2007.07.001 [CrossRef[]
  • Kyselo M, Di Paolo E. (2013). Locked-in syndrome: a challenge for embodied cognitive science. Phenomenol. Cogn. Sci. 10.1007/s11097-013-9344-9 [CrossRef[]
  • Leary T. F. (1957). Interpersonal Diagnosis of Personality. New York: Ronald Press []
  • Levins R. (1998). Dialects and system theory. Sci. Soc. 62 375–399 []
  • Lewin K. (1936). Principles of Topological Psychology. New York: McGraw-Hill. 10.1037/10019-000 [CrossRef[]
  • Luhmann N. (1992). Die Wissenschaft der Gesellschaft. Frankfurt am Main: Suhrkamp []
  • Mahler M., Pine F., Bergman A. (2000). The Psychological Birth Of The Human Infant: Symbiosis and Individuation. New York, NY: Basic Books []
  • Markus H., Kitayama S. (2010). Cultures and selves: a cycle of mutual constitution. Perspect. Psychol. Sci. 5 420–430 10.1177/1745691610375557 [PubMed] [CrossRef[]
  • Maturana H. R., Varela F. J. (1980). Autopoiesis and Cognition: The Realization of the Living. Dordrecht: D. Reidel Pub. Co []
  • Maturana H., Varela F. J. (1987). The Tree of Knowledge: The Biological Roots of Human Understanding. Boston, MA: Shambhala []
  • Mead G. H. (1934). Mind, Self and Society from the Standpoint of a Social Behaviorist ed. Morris C. W. Chicago: University of Chicago Press []
  • Nowak A., Vallacher R. R., Zochowski M. (2002). “The emergence of personality: personal stability through interpersonal synchronization,” in Advances in Personality Science 1 eds Cervone D., Mischel W. (New York: Guilford Publications; ) 292–331 []
  • Parnas J., Sass L. A. (2010). “The spectrum of schizophrenia,” in The Embodied Self eds Fuchs T., Sattel H. C., Henningsen P. (Stuttgart: Schattauer; ) 227–243 []
  • Röhricht F. (2009). Body oriented psychotherapy. The state of the art in empirical research and evidence-based practice: a clinical perspective. Body Mov. Dance Psychother. 4 135–156 10.1080/17432970902857263 [CrossRef[]
  • Rubin E. (1921). Visuelle wahrgenommene Figuren. Copenhagen: Gyldendalske []
  • Salvatore S., Tschacher W. (2012). Time dependency of psychotherapeutic exchanges: the contribution of the theory of dynamic systems in analyzing process. Front. Psychol. 3:253. 10.3389/fpsyg.2012.00253 [PMC free article] [PubMed] [CrossRef[]
  • Seikkula J., Olson M. (2003). The open dialogue approach to acute psychosis: its poetics and micropolitics. Fam. Process 42 403–418 10.1111/j.1545-5300.2003.00403.x [PubMed] [CrossRef[]
  • Seikkula J. (2008). Inner and outer voices in the present moment of family and network therapy. J. Fam. Ther. 30 478–491 10.1111/j.1467-6427.2008.00439.x [CrossRef[]
  • Stern D. (2004). The Present Moment in Psychotherapy and Everyday Life. New York: W. W. Norton []
  • Thelen E., Smith L. B. (1994). A Dynamic Systems Approach to the Development of Cognition and Action. Cambridge, MA: Bradford Books, MIT Press []
  • Thompson E., Stapleton M. (2009). Making sense of sense-making: reflections on enactive and extended theories. Topoi 28 23–30 10.1007/s11245-008-9043-2 [CrossRef[]
  • Thompson E. (2007). Mind in Life: Biology, Phenomenology, and the Sciences of Mind. Cambridge: Harvard University Press. []
  • Thompson E., Varela F. (2001). Radical embodiment: neural dynamics and consciousness. Trends Cogn. Sci. 5 418–425 10.1016/S1364-6613(00)01750-2 [PubMed] [CrossRef[]
  • Tschacher W., Storch M., Munt M. (2014). Tanz: eine psychotherapeutische Technik? Psychother. Psychiatr. Psychother. Med. Klin. Psychol. 19 1–13 []
  • Tschacher W., Munt M. (2013). Das Selbst als Attraktor: das psychologische Selbst aus systemtheoretischer und achtsamkeitsbasierter Sicht. Psychother. Psychiatr. Psychother. Med. Klin. Psychol. 18 18–37 []
  • Tschacher W., Haken H. (2007). Intentionality in non-equilibrium systems? The functional aspects of self-organized pattern formation. New Ideas Psychol. 25 1–15 10.1016/j.newideapsych.2006.09.002 [CrossRef[]
  • Tschacher W., Dauwalder J.-P. (1999). “Situated cognition, ecological perception, and synergetics: a novel perspective for cognitive psychology?” in Dynamics, Synergetics, Autonomous Agents eds Tschacher W., Dauwalder J.-P. (Singapore: World Scientific; ) 83–104 []
  • Tschacher W., Rössler O. (1996). The self: a processual gestalt. Chaos Solitons Fractals 7 1011–1022 10.1016/0960-0779(95)00096-8 [CrossRef[]
  • Van Geert P. (1998). A dynamic systems model of basic developmental mechanisms: Piaget, Vygotsky, and beyond. Psychol. Rev. 105 634–677 10.1037/0033-295X.105.4.634-677 [CrossRef[]
  • Van Gelder T. (1998). The dynamical hypothesis in cognitive science. Behav. Brain Sci. 21 615–665 10.1017/S0140525X98001733 [PubMed] [CrossRef[]
  • Varela F. J., Thompson E., Rosch E. (1993). The Embodied Mind: Cognitive Science and Human Experience. Cambridge, MA: MIT Press []
  • Varela F. J. (1997). Patterns of life: intertwining identity and cognition. Brain Cogn. 34 72–87 10.1006/brcg.1997.0907 [PubMed] [CrossRef[]
  • Vygotsky L. (1986). Thought and Language. Cambridge, MA: MIT Press []
  • Weber A., Varela F. (2002). Life after Kant: natural purposes and the autopoietic foundations of biological individuality. Phenomenol. Cogn. Sci. 1 97–125 10.1023/A:1020368120174 [CrossRef[]

Models of Metaphor in NLP by Ekaterina Shutova

Models of Metaphor in NLP

Ekaterina Shutova

Computer Laboratory University of Cambridge
15 JJ Thomson Avenue Cambridge CB3 0FD, UK


Automatic processing of metaphor can be clearly divided into two subtasks: metaphor recognition (distinguishing be- tween literal and metaphorical language in a text) and metaphor interpretation (identifying the intended literal meaning of a metaphorical expression). Both of them have been repeatedly addressed in NLP. This paper is the first comprehensive and systematic review of the existing computational models of metaphor, the issues of metaphor annotation in corpora and the available resources.

1 Introduction

Our production and comprehension of language is a multi-layered computational process. Humans carry out high-level semantic tasks effortlessly by subconsciously employing a vast inventory of complex linguistic devices, while simultaneously integrating their background knowledge, to reason about reality. An ideal model of language understanding would also be capable of performing such high-level semantic tasks.

However, a great deal of NLP research to date focuses on processing lower-level linguistic information, such as e.g. part-of-speech tagging, discovering syntactic structure of a sentence (parsing), coreference resolution, named entity recognition and many others. Another cohort of researchers set the goal of improving application- based statistical inference (e.g. for recognizing textual entailment or automatic summarization). In contrast, there have been fewer attempts to bring the state-of-the-art NLP technologies together to model the way humans use language to frame high-level reasoning processes, such as for example, creative thought.

The majority of computational approaches to

Proceedings of the 48th Annual Meeting of the Association for Computational Linguistics, pages 688–697, Uppsala, Sweden, 11-16 July 2010. ⃝c 2010 Association for Computational Linguistics

figurative language still exploit the ideas articulated three decades ago (Wilks, 1978; Lakoff and Johnson, 1980; Fass, 1991) and often rely on task-specific hand-coded knowledge. However, recent work on lexical semantics and lexical acquisition techniques opens many new avenues for creation of fully automated models for recognition and interpretation of figurative language. In this paper I will focus on the phenomenon of metaphor and describe the most prominent computational approaches to metaphor, as well the issues of resource creation and metaphor annotation.

Metaphors arise when one concept is viewed in terms of the properties of the other. In other words it is based on similarity between the concepts. Similarity is a kind of association implying the presence of characteristics in common. Here are some examples of metaphor.

  1. (1)  Hillary brushed aside the accusations.
  2. (2)  How can I kill a process? (Martin, 1988)
  3. (3)  I invested myself fully in this relationship.
  4. (4)  And then my heart with pleasure fills, And dances with the daffodils.1

In metaphorical expressions seemingly unrelated features of one concept are associated with another concept. In the example (2) the computational process is viewed as something alive and, therefore, its forced termination is associated with the act of killing.

Metaphorical expressions represent a great variety, ranging from conventional metaphors, which we reproduce and comprehend every day, e.g. those in (2) and (3), to poetic and largely novel ones, such as (4). The use of metaphor is ubiquitous in natural language text and it is a serious bottleneck in automatic text understanding.

1“I wandered lonely as a cloud”, William Wordsworth, 1804.


In order to estimate the frequency of the phenomenon, Shutova (2010) conducted a corpus study on a subset of the British National Corpus (BNC) (Burnard, 2007) representing various genres. They manually annotated metaphorical expressions in this data and found that 241 out of 761 sentences contained a metaphor. Due to such a high frequency of their use, a system capable of recognizing and interpreting metaphorical expressions in unrestricted text would become an invaluable component of any semantics-oriented NLP application.

Automatic processing of metaphor can be clearly divided into two subtasks: metaphor recognition (distinguishing between literal and metaphorical language in text) and metaphor interpretation (identifying the intended literal meaning of a metaphorical expression). Both of them have been repeatedly addressed in NLP.

2 Theoretical Background

Four different views on metaphor have been broadly discussed in linguistics and philosophy: the comparison view (Gentner, 1983), the interaction view (Black, 1962), (Hesse, 1966), the selectional restrictions violation view (Wilks, 1975; Wilks, 1978) and the conceptual metaphor view (Lakoff and Johnson, 1980)2. All of these approaches share the idea of an interconceptual mapping that underlies the production of metaphorical expressions. In other words, metaphor always involves two concepts or conceptual domains: the target (also called topic or tenor in the linguistics literature) and the source (or vehicle). Consider the examples in (5) and (6).

  1. (5)  He shot down all of my arguments. (Lakoff and Johnson, 1980)
  2. (6)  He attacked every weak point in my argument. (Lakoff and Johnson, 1980)

According to Lakoff and Johnson (1980), a mapping of a concept of argument to that of war is employed here. The argument, which is the target concept, is viewed in terms of a battle (or a war), the source concept. The existence of such a link allows us to talk about arguments using the war terminology, thus giving rise to a number of metaphors.

2 A detailed overview and criticism of these four views can be found in (Tourangeau and Sternberg, 1982).

However, Lakoff and Johnson do not discuss how metaphors can be recognized in the linguistic data, which is the primary task in the automatic processing of metaphor. Although humans are highly capable of producing and comprehending metaphorical expressions, the task of distinguishing between literal and nonliteral meanings and, therefore, identifying metaphor in text appears to be challenging. This is due to the variation in its use and external form, as well as a not clear-cut semantic distinction. Gibbs (1984) suggests that literal and figurative meanings are situated at the ends of a single continuum, along which metaphoricity and idiomaticity are spread. This makes demarcation of metaphorical and literal language fuzzy.

So far, the most influential account of metaphor recognition is that of Wilks (1978). According to Wilks, metaphors represent a violation of selectional restrictions in a given context. Selectional restrictions are the semantic constraints that a verb places onto its arguments. Consider the following example.

(7) My car drinks gasoline. (Wilks, 1978)

The verb drink normally takes an animate subject and a liquid object. Therefore, drink taking a car as a subject is an anomaly, which may in turn indicate the metaphorical use of drink.

3 Automatic Metaphor Recognition

One of the first attempts to identify and interpret metaphorical expressions in text automatically is the approach of Fass (1991). It originates in the work of Wilks (1978) and utilizes hand- coded knowledge. Fass (1991) developed a system called met*, capable of discriminating between literalness, metonymy, metaphor and anomaly. It does this in three stages. First, literalness is distinguished from non-literalness using selectional preference violation as an indicator. In the case that non-literalness is detected, the respective phrase is tested for being a metonymic relation us- ing hand-coded patterns (such as CONTAINER- for-CONTENT). If the system fails to recognize metonymy, it proceeds to search the knowledge base for a relevant analogy in order to discriminate metaphorical relations from anomalous ones. E.g., the sentence in (7) would be represented in this framework as (car,drink,gasoline), which does not satisfy the preference (animal, drink, liquid), as car


is not a hyponym of animal. met* then searches its knowledge base for a triple containing a hypernym of both the actual argument and the desired argument and finds (thing, use, energy source), which represents the metaphorical interpretation.

However, Fass himself indicated a problem with the selectional preference violation approach applied to metaphor recognition. The approach detects any kind of nonliteralness or anomaly in language (metaphors, metonymies and others), and not only metaphors, i.e., it overgenerates. The methods met* uses to differentiate between those are mainly based on hand-coded knowledge, which implies a number of limitations.

Another problem with this approach arises from the high conventionality of metaphor in language. This means that some metaphorical senses are very common. As a result the system would extract selectional preference distributions skewed towards such conventional metaphorical senses of the verb or one of its arguments. Therefore, although some expressions may be fully metaphorical in nature, no selectional preference violation can be detected in their use. Another counterargument is bound to the fact that interpretation is always context dependent, e.g. the phrase all men are animals can be used metaphorically, however, without any violation of selectional restrictions.

Goatly (1997) addresses the phenomenon of metaphor by identifying a set of linguistic cues indicating it. He gives examples of lexical patterns indicating the presence of a metaphorical expression, such as metaphorically speaking, utterly, completely, so to speak and, surprisingly, literally. Such cues would probably not be enough for metaphor extraction on their own, but could contribute to a more complex system.

The work of Peters and Peters (2000) concentrates on detecting figurative language in lexical resources. They mine WordNet (Fellbaum, 1998) for the examples of systematic polysemy, which allows to capture metonymic and metaphorical relations. The authors search for nodes that are relatively high up in the WordNet hierarchy and that share a set of common word forms among their descendants. Peters and Peters found that such nodes often happen to be in metonymic (e.g. publication publisher) or metaphorical (e.g. supporting structure – theory) relation

The CorMet system discussed in (Mason, 2004) is the first attempt to discover source-target do-

main mappings automatically. This is done by “finding systematic variations in domain-specific selectional preferences, which are inferred from large, dynamically mined Internet corpora”. For example, Mason collects texts from the LAB domain and the FINANCE domain, in both of which pour would be a characteristic verb. In the LAB domain pour has a strong selectional preference for objects of type liquid, whereas in the FINANCE domain it selects for money. From this Mason’s system infers the domain mapping FINANCE – LAB and the concept mapping money – liquid. He compares the output of his system against the Master Metaphor List (Lakoff et al., 1991) containing hand-crafted metaphorical mappings between concepts. Mason reports an accuracy of 77%, although it should be noted that as any evaluation that is done by hand it contains an element of subjectivity.

Birke and Sarkar (2006) present a sentence clustering approach for non-literal language recognition implemented in the TroFi system (Trope Finder). This idea originates from a similarity- based word sense disambiguation method developed by Karov and Edelman (1998). The method employs a set of seed sentences, where the senses are annotated; computes similarity between the sentence containing the word to be disambiguated and all of the seed sentences and selects the sense corresponding to the annotation in the most similar seed sentences. Birke and Sarkar (2006) adapt this algorithm to perform a two-way classification: literal vs. nonliteral, and they do not clearly de- fine the kinds of tropes they aim to discover. They attain a performance of 53.8% in terms of f-score.

The method of Gedigan et al. (2006) discriminates between literal and metaphorical use. They trained a maximum entropy classifier for this purpose. They obtained their data by extracting the lexical items whose frames are related to MO- TION and CURE from FrameNet (Fillmore et al., 2003). Then they searched the PropBank Wall Street Journal corpus (Kingsbury and Palmer, 2002) for sentences containing such lexical items and annotated them with respect to metaphoric- ity. They used PropBank annotation (arguments and their semantic types) as features to train the classifier and report an accuracy of 95.12%. This result is, however, only a little higher than the performance of the naive baseline assigning majority class to all instances (92.90%). These numbers


can be explained by the fact that 92.00% of the verbs of MOTION and CURE in the Wall Street Journal corpus are used metaphorically, thus making the dataset unbalanced with respect to the target categories and the task notably easier.

Both Birke and Sarkar (2006) and Gedigan et al. (2006) focus only on metaphors expressed by a verb. As opposed to that the approach of Krishnakumaran and Zhu (2007) deals with verbs, nouns and adjectives as parts of speech. They use hyponymy relation in WordNet and word bigram counts to predict metaphors at a sentence level. Given an IS-A metaphor (e.g. The world is a stage3) they verify if the two nouns involved are in hyponymy relation in WordNet, and if they are not then this sentence is tagged as containing a metaphor. Along with this they con- sider expressions containing a verb or an adjective used metaphorically (e.g. He planted good ideas in their minds or He has a fertile imagination). Hereby they calculate bigram probabilities of verb-noun and adjective-noun pairs (including the hyponyms/hypernyms of the noun in question). If the combination is not observed in the data with sufficient frequency, the system tags the sentence containing it as metaphorical. This idea is a modification of the selectional preference view of Wilks. However, by using bigram counts over verb-noun pairs Krishnakumaran and Zhu (2007) loose a great deal of information com- pared to a system extracting verb-object relations from parsed text. The authors evaluated their system on a set of example sentences compiled from the Master Metaphor List (Lakoff et al., 1991), whereby highly conventionalized metaphors (they call them dead metaphors) are taken to be negative examples. Thus they do not deal with literal examples as such: essentially, the distinction they are making is between the senses included in Word- Net, even if they are conventional metaphors, and those not included in WordNet.

4 Automatic Metaphor Interpretation

Almost simultaneously with the work of Fass (1991), Martin (1990) presents a Metaphor Interpretation, Denotation and Acquisition System (MIDAS). In this work Martin captures hierarchical organization of conventional metaphors. The idea behind this is that the more specific conventional metaphors descend from the general ones.

3William Shakespeare

Given an example of a metaphorical expression, MIDAS searches its database for a corresponding metaphor that would explain the anomaly. If it does not find any, it abstracts from the example to more general concepts and repeats the search. If it finds a suitable general metaphor, it creates a mapping for its descendant, a more specific metaphor, based on this example. This is also how novel metaphors are acquired. MIDAS has been integrated with the Unix Consultant (UC), the system that answers users questions about Unix. The UC first tries to find a literal answer to the question. If it is not able to, it calls MIDAS which detects metaphorical expressions via selectional preference violation and searches its database for a metaphor explaining the anomaly in the question.

Another cohort of approaches relies on per- forming inferences about entities and events in the source and target domains for metaphor interpretation. These include the KARMA system (Narayanan, 1997; Narayanan, 1999; Feldman and Narayanan, 2004) and the ATT-Meta project (Barnden and Lee, 2002; Agerri et al., 2007). Within both systems the authors developed a metaphor-based reasoning framework in accordance with the theory of conceptual metaphor. The reasoning process relies on manually coded knowledge about the world and operates mainly in the source domain. The results are then projected onto the target domain using the conceptual mapping representation. The ATT-Meta project concerns metaphorical and metonymic description of mental states and reasoning about mental states using first order logic. Their system, however, does not take natural language sentences as input, but logical expressions that are representations of small discourse fragments. KARMA in turn deals with a broad range of abstract actions and events and takes parsed text as input.

Veale and Hao (2008) derive a “fluid knowledge representation for metaphor interpretation and generation”, called Talking Points. Talking Points are a set of characteristics of concepts belonging to source and target domains and related facts about the world which the authors acquire automatically from WordNet and from the web. Talking Points are then organized in Slipnet, a framework that allows for a number of insertions, deletions and substitutions in definitions of such characteristics in order to establish a connection between the target and the source


concepts. This work builds on the idea of slippage in knowledge representation for understanding analogies in abstract domains (Hofstadter and Mitchell, 1994; Hofstadter, 1995). Below is an example demonstrating how slippage operates to explain the metaphor Make-up is a Western burqa.

Make-up =
typically worn by women
expected to be worn by women must be worn by women
must be worn by Muslim women

Burqa à=

By doing insertions and substitutions the system arrives from the definition typically worn by women to that of must be worn by Muslim women, and thus establishes a link between the concepts of make-up and burqa. Veale and Hao (2008), however, did not evaluate to which extent their knowledge base of Talking Points and the associated reasoning framework are useful to interpret metaphorical expressions occurring in text.

Shutova (2010) defines metaphor interpretation as a paraphrasing task and presents a method for deriving literal paraphrases for metaphorical expressions from the BNC. For example, for the metaphors in “All of this stirred an unfathomable excitement in her” or “a carelessly leaked report” their system produces interpretations “All of this provoked an unfathomable excitement in her” and “a carelessly disclosed report” respectively. They first apply a probabilistic model to rank all possible paraphrases for the metaphorical expression given the context; and then use automatically induced selectional preferences to discriminate between figurative and literal paraphrases. The selectional preference distribution is defined in terms of selectional association measure introduced by Resnik (1993) over the noun classes automatically produced by Sun and Korhonen (2009). Shutova (2010) tested their system only on metaphors expressed by a verb and report a paraphrasing accuracy of 0.81.

5 Metaphor Resources

Metaphor is a knowledge-hungry phenomenon. Hence there is a need for either an extenZsive manually-created knowledge-base or a robust knowledge acquisition system for interpretation of metaphorical expressions. The latter being a hard task, a great deal of metaphor research resorted to

the first option. Although hand-coded knowledge proved useful for metaphor interpretation (Fass, 1991; Martin, 1990), it should be noted that the systems utilizing it have a very limited coverage.

One of the first attempts to create a multipurpose knowledge base of source–target domain mappings is the Master Metaphor List (Lakoff et al., 1991). It includes a classification of metaphorical mappings (mainly those related to mind, feel- ings and emotions) with the corresponding examples of language use. This resource has been criticized for the lack of clear structuring principles of the mapping ontology (Lo ̈nneker-Rodman, 2008). The taxonomical levels are often confused, and the same classes are referred to by different class labels. This fact and the chosen data representation in the Master Metaphor List make it not suitable for computational use. However, both the idea of the list and its actual mappings ontology inspired the creation of other metaphor resources.

The most prominent of them are MetaBank (Martin, 1994) and the Mental Metaphor Data- bank4 created in the framework of the ATT-meta project (Barnden and Lee, 2002; Agerri et al., 2007). The MetaBank is a knowledge-base of English metaphorical conventions, represented in the form of metaphor maps (Martin, 1988) contain- ing detailed information about source-target concept mappings backed by empirical evidence. The ATT-meta project databank contains a large number of examples of metaphors of mind classified by source–target domain mappings taken from the Master Metaphor List.

Along with this it is worth mentioning metaphor resources in languages other than English. There has been a wealth of research on metaphor in Spanish, Chinese, Russian, German, French and Italian. The Hamburg Metaphor Database (Lo ̈nneker, 2004; Reining and Lo ̈nneker-Rodman, 2007) contains examples of metaphorical expressions in German and French, which are mapped to senses from EuroWordNet5 and annotated with source–target domain mappings taken from the Master Metaphor List.

Alonge and Castelli (2003) discuss how metaphors can be represented in ItalWordNet for


5EuroWordNet is a multilingual database with wordnets for several European languages (Dutch, Italian, Spanish, Ger- man, French, Czech and Estonian). The wordnets are structured in the same way as the Princeton WordNet for English. URL:


Italian and motivate this by linguistic evidence. Encoding metaphorical information in general domain lexical resources for English, e.g. Word-Net (Lo ̈nneker and Eilts, 2004), would undoubtedly provide a new platform for experiments and enable researchers to directly compare their results.

6 Metaphor Annotation in Corpora

To reflect two distinct aspects of the phenomenon, metaphor annotation can be split into two stages: identifying metaphorical senses in text (akin word sense disambiguation) and annotating source – target domain mappings underlying the production of metaphorical expressions. Traditional approaches to metaphor annotation include manual search for lexical items used metaphorically (Pragglejaz Group, 2007), for source and target domain vocabulary (Deignan, 2006; Koivisto-Alanko and Tis- sari, 2006; Martin, 2006) or for linguistic mark- ers of metaphor (Goatly, 1997). Although there is a consensus in the research community that the phenomenon of metaphor is not restricted to similarity-based extensions of meanings of iso- lated words, but rather involves reconceptualization of a whole area of experience in terms of an- other, there still has been surprisingly little inter- est in annotation of cross-domain mappings. How- ever, a corpus annotated for conceptual mappings could provide a new starting point for both linguistic and cognitive experiments.

6.1 Metaphor and Polysemy

The theorists of metaphor distinguish between two kinds of metaphorical language: novel (or poetic) metaphors, that surprise our imagination, and conventionalized metaphors, that become a part of an ordinary discourse. “Metaphors begin their lives as novel poetic creations with marked rhetorical effects, whose comprehension requires a special imaginative leap. As time goes by, they become a part of general usage, their comprehension be- comes more automatic, and their rhetorical effect is dulled” (Nunberg, 1987). Following Orwell (1946) Nunberg calls such metaphors “dead” and claims that they are not psychologically distinct from literally-used terms.

This scheme demonstrates how metaphorical associations capture some generalisations govern- ing polysemy: over time some of the aspects of the target domain are added to the meaning of a

term in a source domain, resulting in a (metaphorical) sense extension of this term. Copestake and Briscoe (1995) discuss sense extension mainly based on metonymic examples and model the phenomenon using lexical rules encoding metonymic patterns. Along with this they suggest that similar mechanisms can be used to account for metaphoric processes, and the conceptual mappings encoded in the sense extension rules would define the lim- its to the possible shifts in meaning.

However, it is often unclear if a metaphorical instance is a case of broadening of the sense in context due to general vagueness in language, or it manifests a formation of a new distinct metaphorical sense. Consider the following examples.

(8) a.

b. (9) a. b.

As soon as I entered the room I noticed the difference.

How can I enter Emacs? My tea is cold.
He is such a cold person.

Enter in (8a) is defined as “to go or come into a place, building, room, etc.; to pass within the boundaries of a country, region, portion of space, medium, etc.”6 In (8b) this sense stretches to describe dealing with software, whereby COM- PUTER PROGRAMS are viewed as PHYSICAL SPACES. However, this extended sense of enter does not appear to be sufficiently distinct or conventional to be included into the dictionary, although this could happen over time.

The sentence (9a) exemplifies the basic sense of cold – “of a temperature sensibly lower than that of the living human body”, whereas cold in (9b) should be interpreted metaphorically as “void of ardour, warmth, or intensity of feeling; lacking enthusiasm, heartiness, or zeal; indifferent, apathetic”. These two senses are clearly linked via the metaphoric mapping between EMOTIONAL STATES and TEMPERATURES.

A number of metaphorical senses are included in WordNet, however without any accompanying semantic annotation.



Metaphor Identification Pragglejaz Procedure

Pragglejaz Group (2007) proposes a metaphor

identification procedure (MIP) within the frame-

6Sense definitions are taken from the Oxford English Dic- tionary.


work of the Metaphor in Discourse project (Steen, 2007). The procedure involves metaphor annotation at the word level as opposed to identifying metaphorical relations (between words) or source–target domain mappings (between concepts or do- mains). In order to discriminate between the verbs used metaphorically and literally the annotators are asked to follow the guidelines:

  1. Foreachverbestablishitsmeaningincontext and try to imagine a more basic meaning of this verb on other contexts. Basic meanings normally are: (1) more concrete; (2) related to bodily action; (3) more precise (as opposed to vague); (4) historically older.
  2. If you can establish the basic meaning that is distinct from the meaning of the verb in this context, the verb is likely to be used metaphorically.

Such annotation can be viewed as a form of word sense disambiguation with an emphasis on metaphoricity.

6.2.2 Source – Target Domain Vocabulary

Another popular method that has been used to ex- tract metaphors is searching for sentences containing lexical items from the source domain, the tar- get domain, or both (Stefanowitsch, 2006). This method requires exhaustive lists of source and target domain vocabulary.

Martin (2006) conducted a corpus study in order to confirm that metaphorical expressions occur in text in contexts containing such lex- ical items. He performed his analysis on the data from the Wall Street Journal (WSJ) cor- pus and focused on four conceptual metaphors that occur with considerable regularity in the corpus. These include NUMERICAL VALUE AS LOCATION, COMMERCIAL ACTIVITY AS CONTAINER, COMMERCIAL ACTIVITY AS PATH FOLLOWING and COMMERCIAL ACTIVITY AS WAR. Martin manually compiled the lists of terms characteristic for each domain by examining sampled metaphors of these types and then augmented them through the use of thesaurus. He then searched the WSJ for sentences containing vocabulary from these lists and checked whether they contain metaphors of the above types. The goal of this study was to evaluate predictive ability of contexts containing vocabulary from (1) source domain and (2) target

domain, as well as (3) estimating the likelihood of a metaphorical expression following another metaphorical expression described by the same mapping. He obtained the most positive results for metaphors of the type NUMERICAL-VALUE- AS-LOCATION (P Metaphor Source = ààà69, P M etaphor T arget = àà6àà, P M etaphor M etaphor = àààà3).

6.3 Annotating Source and Target Domains

Wallington et al. (2003) carried out a metaphor an- notation experiment in the framework of the ATT- Meta project. They employed two teams of an- notators. Team A was asked to annotate “interesting stretches”, whereby a phrase was considered interesting if (1) its significance in the document was non-physical, (2) it could have a physical significance in another context with a similar syntactic frame, (3) this physical significance was related to the abstract one. Team B had to annotate phrases according to their own intuitive definition of metaphor. Besides metaphorical expressions Wallington et al. (2003) attempted to annotate the involved source – target domain mappings. The annotators were given a set of mappings from the Master Metaphor List and were asked to assign the most suitable ones to the examples. However, the authors do not report the level of interannotator agreement nor the coverage of the mappings in the Master Metaphor List on their data.

Shutova and Teufel (2010) adopt a different approach to the annotation of source – target domain mappings. They do not rely on predefined mappings, but instead derive independent sets of most common source and target categories. They propose a two stage procedure, whereby the metaphorical expressions are first identified using MIP, and then the source domain (where the basic sense comes from) and the target domain (the given context) are selected from the lists of cate- gories. Shutova and Teufel (2010) report interannotator agreement of0.61( ).

7 Conclusion and Future Directions

The eighties and nineties provided us with a wealth of ideas on the structure and mechanisms of the phenomenon of metaphor. The approaches formulated back then are still highly influential, although their use of hand-coded knowledge is becoming increasingly less convincing. The last decade witnessed a high technological leap in


natural language computation, whereby manually crafted rules gradually give way to more robust corpus-based statistical methods. This is also the case for metaphor research. The latest develop- ments in the lexical acquisition technology will in the near future enable fully automated corpusbased processing of metaphor.

However, there is still a clear need in a unified metaphor annotation procedure and creation of a large publicly available metaphor corpus. Given such a resource the computational work on metaphor is likely to proceed along the following lines: (1) automatic acquisition of an extensive set of valid metaphorical associations from linguistic data via statistical pattern matching; (2) using the knowledge of these associations for metaphor recognition in the unseen unrestricted text and, finally, (3) interpretation of the identified metaphorical expressions by deriving the closest literal paraphrase (a representation that can be directly embedded in other NLP applications to enhance their performance

Besides making our thoughts more vivid and filling our communication with richer imagery, metaphors also play an important structural role in our cognition. Thus, one of the long term goals of metaphor research in NLP and AI would be to build a computational intelligence model accounting for the way metaphors organize our conceptual system, in terms of which we think and act.


I would like to thank Anna Korhonen and my reviewers for their most helpful feedback on this paper. The support of Cambridge Overseas Trust, who fully funds my studies, is gratefully acknowledged.


R. Agerri, J.A. Barnden, M.G. Lee, and A.M. Walling- ton. 2007. Metaphor, inference and domain- independent mappings. In Proceedings of RANLP- 2007, pages 17–23, Borovets, Bulgaria.

A. Alonge and M. Castelli. 2003. Encoding information on metaphoric expressions in WordNet-like resources. In Proceedings of the ACL 2003 Workshop on Lexicon and Figurative Language, pages 10–17.

J.A. Barnden and M.G. Lee. 2002. An artificial intelligence approach to metaphor understanding. Theoria et Historia Scientiarum, 6(1):399–412.

J. Birke and A. Sarkar. 2006. A clustering approach for the nearly unsupervised recognition of nonlit- eral language. In In Proceedings of EACL-06, pages 329–336.

M. Black. 1962. Models and Metaphors. Cornell Uni- versity Press.

L. Burnard. 2007. Reference Guide for the British Na- tional Corpus (XML Edition).

A. Copestake and T. Briscoe. 1995. Semi-productive polysemy and sense extension. Journal of Seman- tics, 12:15–67.

A. Deignan. 2006. The grammar of linguistic metaphors. In A. Stefanowitsch and S. T. Gries, editors, Corpus-Based Approaches to Metaphor and Metonymy, Berlin. Mouton de Gruyter.

D. Fass. 1991. met*: A method for discriminating metonymy and metaphor by computer. Computa- tional Linguistics, 17(1):49–90.

J. Feldman and S. Narayanan. 2004. Embodied mean- ing in a neural theory of language. Brain and Lan- guage, 89(2):385–392.

C. Fellbaum, editor. 1998. WordNet: An Electronic Lexical Database (ISBN: 0-262-06197-X). MIT Press, first edition.

C. J. Fillmore, C. R. Johnson, and M. R. L. Petruck. 2003. Background to FrameNet. International Journal of Lexicography, 16(3):235–250.

M. Gedigan, J. Bryant, S. Narayanan, and B. Ciric. 2006. Catching metaphors. In In Proceedings of the 3rd Workshop on Scalable Natural Language Un- derstanding, pages 41–48, New York.

D. Gentner. 1983. Structure mapping: A theoretical framework for analogy. Cognitive Science, 7:155– 170.

R. Gibbs. 1984. Literal meaning and psychological theory. Cognitive Science, 8:275–304.

A. Goatly. 1997. The Language of Metaphors. Rout- ledge, London.

M. Hesse. 1966. Models and Analogies in Science. Notre Dame University Press.

D. Hofstadter and M. Mitchell. 1994. The Copycat Project: A model of mental fluidity and analogy- making. In K.J. Holyoak and J. A. Barnden, editors, Advances in Connectionist and Neural Computation Theory, Ablex, New Jersey.

D. Hofstadter. 1995. Fluid Concepts and Creative Analogies: Computer Models of the Fundamental Mechanisms of Thought. HarperCollins Publishers.

Y. Karov and S. Edelman. 1998. Similarity-based word sense disambiguation. Computational Lin- guistics, 24(1):41–59.


P. Kingsbury and M. Palmer. 2002. From TreeBank to PropBank. In Proceedings of LREC-2002, Gran Canaria, Canary Islands, Spain.

S. Narayanan. 1999. Moving right along: A computa- tional model of metaphoric reasoning about events. In Proceedings of AAAI 99), pages 121–128, Or- lando, Florida.

G. Nunberg. 1987. Poetic and prosaic metaphors. In

Proceedings of the 1987 workshop on Theoretical issues in natural language processing, pages 198– 201.

G. Orwell. 1946. Politics and the english language. Horizon.

W. Peters and I. Peters. 2000. Lexicalised system- atic polysemy in wordnet. In Proceedings of LREC 2000, Athens.

Pragglejaz Group. 2007. MIP: A method for iden- tifying metaphorically used words in discourse. Metaphor and Symbol, 22:1–39.

A. Reining and B. Lo ̈nneker-Rodman. 2007. Corpus- driven metaphor harvesting. In Proceedings of the HLT/NAACL-07 Workshop on Computational Approaches to Figurative Language, pages 5–12, Rochester, New York.

P. Resnik. 1993. Selection and Information: A Class- based Approach to Lexical Relationships. Ph.D. the- sis, Philadelphia, PA, USA.

E. Shutova and S. Teufel. 2010. Metaphor corpus an- notated for source – target domain mappings. In Pro- ceedings of LREC 2010, Malta.

E. Shutova. 2010. Automatic metaphor interpretation as a paraphrasing task. In Proceedings of NAACL 2010, Los Angeles, USA.

G. J. Steen. 2007. Finding metaphor in discourse: Pragglejaz and beyond. Cultura, Lenguaje y Rep- resentacion / Culture, Language and Representation (CLR), Revista de Estudios Culturales de la Univer- sitat Jaume I, 5:9–26.

A. Stefanowitsch. 2006. Corpus-based approaches to metaphor and metonymy. In A. Stefanowitsch and S. T. Gries, editors, Corpus-Based Approaches to Metaphor and Metonymy, Berlin. Mouton de Gruyter.

L. Sun and A. Korhonen. 2009. Improving verb clus- tering with automatically acquired selectional pref- erences. In Proceedings of EMNLP 2009, pages 638–647, Singapore, August.

R. Tourangeau and R. Sternberg. 1982. Understand- ing and appreciating metaphors. Cognition, 11:203– 244.

T. Veale and Y. Hao. 2008. A fluid knowledge repre- sentation for understanding and generating creative metaphors. In Proceedings of COLING 2008, pages 945–952, Manchester, UK.


Koivisto-Alanko and H. Tissari. 2006. Sense and sensibility: Rational thought versus emotion in metaphorical language. In A. Stefanowitsch and S. T. Gries, editors, Corpus-Based Approaches to Metaphor and Metonymy, Berlin. Mouton de Gruyter.

S. Krishnakumaran and X. Zhu. 2007. Hunting elusive metaphors using lexical resources. In Proceedings of the Workshop on Computational Approaches to Figurative Language, pages 13–20, Rochester, NY.

G. Lakoff and M. Johnson. 1980. Metaphors We Live By. University of Chicago Press, Chicago.

G. Lakoff, J. Espenson, and A. Schwartz. 1991. The master metaphor list. Technical report, University of California at Berkeley.

B. Lo ̈nneker and C. Eilts. 2004. A Current Re- source and Future Perspectives for Enriching Word- Nets with Metaphor Information. In Proceedings of the Second International WordNet Conference— GWC 2004, pages 157–162, Brno, Czech Republic.

B. Lo ̈nneker-Rodman. 2008. The hamburg metaphor database project: issues in resource creation. Lan- guage Resources and Evaluation, 42(3):293–318.

B. Lo ̈nneker. 2004. Lexical databases as resources for linguistic creativity: Focus on metaphor. In Pro- ceedings of the LREC 2004 Workshop on Language Resources for Linguistic Creativity, pages 9–16, Lis- bon, Portugal.

J. H. Martin. 1988. Representing regularities in the metaphoric lexicon. In Proceedings of the 12th con- ference on Computational linguistics, pages 396– 401.

J. H. Martin.
Metaphor Interpretation. Academic Press Profes- sional, Inc., San Diego, CA, USA.

J. H. Martin. 1994. Metabank: A knowledge-base of metaphoric language conventions. Computational Intelligence, 10:134–149.

J. H. Martin. 2006. A corpus-based analysis of con- text effects on metaphor comprehension. In A. Ste- fanowitsch and S. T. Gries, editors, Corpus-Based Approaches to Metaphor and Metonymy, Berlin. Mouton de Gruyter.

Z. J. Mason. 2004. Cormet: a computational, corpus-based conventional metaphor extraction sys- tem. Computational Linguistics, 30(1):23–44.

S. Narayanan. 1997. Knowledge-based action repre- sentations for metaphor and aspect (karma. Tech- nical report, PhD thesis, University of California at Berkeley.

1990. A Computational Model of


A. M. Wallington, J. A. Barnden, P. Buchlovsky, L. Fel- lows, and S. R. Glasbey. 2003. Metaphor annota- tion: A systematic study. Technical report, School of Computer Science, The University of Birming- ham.

Y. Wilks. 1975. A preferential pattern-seeking seman- tics for natural language inference. Artificial Intelli- gence, 6:53–74.

Y. Wilks. 1978. Making preferences more active. Ar- tificial Intelligence, 11(3):197–223.