Mimesis and Imitation

 by Paul Dumouchel

Mimesis

Girard in his works uses the word “mimesis”, “mimetic desire” and associated terms in two relatively different ways. On the one hand the terms are used in a causal way, as if mimesis and mimetic desire were some form of psychological force that brings people to copy each other, especially in their behaviour of appropriation. Mimesis in this case is understood as a kind of instinct or as a biologically determined propension to copy others. On the other hand, the terms are also used in a descriptive way. On such occasions, mimesis corresponds to or is present whenever we notice similarities in the behaviour of different individuals. This is particularly the case, but not exclusively, when Girard talks about violence. To take a recent, and uncontroversial example, Ismail Haniyeh, the senior political leader of Hamas declared after the atrocious attack on Israeli civilians: “We have only one thing to say to you: get out of our land. Get out of our sight … This land is ours, al-Quds [Jerusalem] is ours, everything [here] is ours … There is no place or safety for you.” [1] Which essentially means no Israeli civilian is innocent, all are legitimate targets. Thus justifying what he had just ordered. A short time after that murderous attack Isaak Herzog, the president of Israel, responding to criticism of the massive bombardments of Gaza and the consequent lost of civilian lives said, “It is an entire nation that is responsible out there…It is not true this rhetoric about civilians not being aware, not being involved.”[2] In other words, they should all be punished, and it is legitimate to do it. It is unlikely that the president of Israel, who would forcefully deny any resemblance with a known terrorist, is copying the leader of Hamas, in the way in which we can say that I am imitating my colleague’s goal of becoming chairperson of the department. In cases of mimesis like this one, we do not need to posit any real relation between the agents, all we do is notice a troubling similitude in their behaviour.

In the competition that opposes me to my colleague, mimetic desire is understood in a causal way. It is what explains our conflict. Its real cause. It is not the direction of the department as a valuable object that cannot be shared that explains our conflict, but the mimetic relation between us. Mimetic desire is also understood causally in this case because it is conceived as real relation between us, and counterfactually it is assumed that if this relation were absent there would be no such conflict between us. As far as the terrorist and the president are concerned the situation seems quite different. Mimesis in this case does not imply or require any real relation between them. By real relation I mean one in which they are personally involved rather than something that can only be described from the outside. Mimesis in this case is revealed by a symmetry or similitude that we observe in their statements and behaviour, a symmetry that is quite independent from any direct relation between the two persons. In this case, we recognize the mimetic dimension of violence from the fact that, under a certain description, the opponents tend to become doubles of each other. In this case, we do not treat mimesis as if it were a force acting causally, but more like a structural characteristic of systems of interactions.

Are these two ways of understanding mimesis compatible? They are different, but I do not think that they are contradictory. To the opposite, an analogy with classical physics suggests that causal explanations and structural explanations are strictly equivalent. This is a rather distant analogy, so it cannot be taken as a proof of anything. However, the analogy is enough for us not to refrain from jumping to the hasty conclusion that the two aspects of mimesis are incompatible or that the notion is incoherent. It remains nonetheless that the concept of mimesis is to some extent ambiguous, and it is not entirely clear how these different aspects of mimesis are related. How mimesis in its descriptive sense, the symmetry that we observe in agents that are not concretely related to each other, relates to mimesis in the causal sense? Are these two types of phenomena sufficiently alike to be both called mimesis?



Imitation

A different, but in some way related problem can be raised concerning imitation. C.L. Nehaniv in his chapter in the book Imitation and Social learning in Robots, Humans and Animals[3] (2009) – a chapter which is entitled “Nine billion correspondence problems” – considers[4] that imitation is a matching behaviour, and defines a matching behaviour in the following way: 1) A behaviour C (for copy) is produced by an organism or machine; 2) C is similar to another behaviour M (for model); 3) registration of M is necessary for the production of C; and 4) C is designed to be similar to M (Nehaniv, 2009:25). This definition corresponds pretty well to how we commonly view imitation and to how Girard conceives what he call the “mimesis of representation”. Imitation requires the “registration” – that is the perception or in some way having a “representation" – of the model behaviour that is reproduced, and this “reproduction” is “designed” (aims at) at being similar to the original behaviour. The requirement that M needs to be registered does not imply that imitation is always or necessarily conscient, its presence in the definition is essentially to rule out cases where C and M are similar by accident. Cases where we would not be dealing with imitation, but with contingent coincidences.

This conception of imitation – and Nehaniv thinks it is a good one – leads, he argues, to what he calls correspondence problems, which is the problem for an organism or machine of generating a ‘suitable’ matching behaviour. This problem is particularly difficult for a machine which does not have any intuition about the world and what is relevant or not. In robotics the difficulties of creating a robot able to imitate as defined by the correspondence problem constitute a particular version of what is classically known in cognitive science as the “Frame Problem”.[5]

The correspondence problem can be illustrated with the following example. With my right hand I throw a rock and break a window. We can identify three different aspects of this event: a gesture – the bodily movement one makes in throwing something – an action – that of throwing a rock – and an effect – breaking a window. Now you can with your right hand make the gesture of throwing a rock, but not throw anything. At which point you imitate the gesture, but neither the action, nor the effect. You could also with your left hand throw a rock but not break anything, this time you imitate the action, but neither the gesture, nor the effect. You could also with your feet kick a rock in the window breaking it, and now you imitate the effect, but neither the action, nor the gesture. You can also imitate only two, say gesture and effect, action and gesture, effect and action, combining these different aspects of M differently. To this we can add a further requirement concerning granularity. That is to say how much does C, the copy, need to be like M, the model behaviour, in order to qualify as imitation. Clearly there are many different ways of throwing a rock, overhead, underhand, sideways, with force, barely, etc. This of course also applies to the gesture and the effect combining these in a myriad of different ways, we rapidly get to Nehaniv’s nine billion correspondence problems.

Nehaniv raises the correspondence problem as a technical issue in the context of robotics where we want robots to learn by themselves to imitate a model behaviour. Note that this problem is prior to the question of what we want the robot to imitate or of what it should imitate. However, the correspondence problem also constitutes a fundamental epistemological issue for us. What level of similarity between C and M allows us to consider that imitation has taken place. Requiring that C reproduce all three dimensions of M, the gesture, the action and the effect, at a sufficiently high level of granularity seems much too demanding. However, once we open the door to imperfect or incomplete imitation, it becomes extremely difficult to know where we should stop. How matching must the matching behaviour be? Except perhaps by insisting on the presence of some normally non observable elements, for example, that M is registered or that C is designed to be similar to M, this question seems difficult to answer. Unfortunately, once we leave the world of the machines that we make ourselves, both registration and design in the sense in which these terms are used here become unobservable. Viewed as a correspondence problem, what constitutes, or which behaviours correspond to imitation seems far from clear.



Phenomenology

This said, there are important differences between mimesis and imitation, two in particular need to be mentioned for they have far reaching consequences on the respective phenomenologies of these two types of interaction. A third important difference will appear later on. The first is that imitation is discrete while mimesis is continuous. The second is that imitation is a dyadic relation while mimesis, or at least mimetic desire, clearly is a triadic relation. The disciples desires the object that the model desires. It is not without reason that Girard called this “triangular desires”.

That imitation is discrete is evident from its definition above as a matching behaviour. M the behaviour to be imitated is a discrete or independent behavioral element and imitation takes place when another organism produces a behaviour C that matches M sufficiently. Both elements are discrete and imitation as a process has a beginning, an end and goal that can all be determined. It is also clear from the above definition that imitation is a dyadic relation. It exclusively concerns C and M, and this remains the case whether C and M are considered to be two different behaviours or two different persons or organisms. Imitation essentially only involves what or who imitates and what or who is imitated. Nothing else is necessary to define the relation and presumably to understand it.

Mimesis and mimetic desire as they are defined by Girard are quite different. First mimetic or triangular desire, as Girard first named it, evidently is a triadic relation involving two subjects and an object. In consequence, it is a more complex relation than imitation because both the model and the disciple are assumed to simultaneously have, so to speak, an eye towards the other and an eye towards the object. Mimetic desire does not simply relate the model and disciple, but it relates them to each other and to the object simultaneously. To put it otherwise, the object mediates their relation to each other and each one mediate the relation of the other to the object. Each pole of the relation constitutes a factor in the relation of the two other poles. The relation is also continuous not having any clearly assignable beginning or end. Mimetic desire does not, unlike imitation, have a well-defined objective whose accomplishment signals the end of the relation. To the opposite, mimesis is an ongoing dynamic relation, that may be broken off by an external event, but there is no particular action or event involving any of the three parties, either the model, the disciple or the object that necessarily constitute the end of the relation.

From these different characteristics of these two types of interactions follow important consequences for their phenomenology. Because of its character as a dyadic relation and its definition as a matching behaviour imitation is essentially repetitive, monotonous, and conservative. It naturally reduces differences and the complexity of the environment. Further because it is repetitive imitation inevitably tends to invades the environment where it takes place unless there is something to stop it. In consequence, in the absence of two conditions which are typically satisfied in evolution by natural selection, imitation reduces the complexity of a system. The first of these two conditions is that there needs to be an independent source of differences or of novelty, which in biology are provided by random mutation and by reproduction mistakes, or imperfect imitation. Note that imperfect imitation or reproduction mistakes are in a way equivalent to Nehaniv’s correspondence problems, but these are seen here from a different point of view. The more correspondence problems arise and less imitation is repetitive, monotonous and conservative. The second necessary condition is accidents of a changing environment, without this moving target populations would rapidly get and remain at fixation. Evolution would stop. However, in the absence of these two external factors, a source of novelty and a moving target imitation by itself is a repetition mechanism that leads to a reduction in complexity. Imitation can only give rise to novelties as part of a complex system of which it constitutes but one element, as is the case in natural selection, or to put it otherwise, imitation can only give rise to novelties if it becomes a triadic relation as in the case of natural selection.

The situation is different when it comes to mimesis. Imitating the desire of another, though it often takes the form of desiring the same object which the other desires, is not itself a definite object. It does not constitute a clear objective or target. To put it otherwise, imitating the desire of another is not a matching problem. There is no behaviour M that being copied satisfies the intended goal of mimesis. Girard often says that the real engine of mimetic desire, what brings us to seek to appropriate the objects that others have or to destroy these others in violence is a desire for being, which he calls metaphysical desire. It is the being of the model that the disciple tries to take over, more precisely the excess of being that he or she attributes to the model. The fact is however, that the being of another is not an object. It is not something that can be represented as a particular goal. Nonetheless, according to Girard, the rivalry does not lead to just anything, to any behaviour whatsoever because each one’s actions are to some extent guided by those of the other.[6] Given the imprecision of this goal, satisfying this desire will take many different forms which will always be constrained by the actions of the other. Hence the phenomena of symmetry that Girard uncovers. Given that here, unlike what is the case for imitation, there is no definite action that determines the end of the relation, the agents are wont to invent different challenges and competitions. They are brought to imagine different means and objects that will become signs of victory or of defeat. Since this search for being is an illusion that nothing can finally or fully satisfy, the relation is incessantly pushed forward. Mimetic desire is continuous and gives rise to a rich phenomenology of behaviour. It inevitably produces novelty and complexity.

These two types of relation – imitation and mimesis – are then radically different and give rise to completely distinct phenomenologies. Imitation is discrete. It is repetitive, monotonous and conservative. It leads to a reduction of complexity. Mimesis is continuous and dynamic. Dynamic in the sense that past results of the relation feedback unto the relation itself and move it forward. It has no recognizable goal or objective. Mimetic theory is a morphogenetic theory because mimetic relations give rise to a complex phenomenology of objects and interrelations. To summarize the argument until now: appearances notwithstanding, mimesis is not a form of imitation. Though imitation may be a way in which some agents try to address their mimetic rivalry. That is to say, mimesis may lead to imitation, but mimesis is not a species of imitation.



An objection and a response

An objection which some may have already silently formulated is the following. If, as argued above, in mimetic relations the actions of each agent are guided by those of the other, then why don’t their behaviour progressively converge towards the same actions and given that mimesis, according to Girard, is contagious, this process would progressively engulf the whole community. In which case rather than giving rise to a complex phenomenology of behaviour, mimesis would turn out to be just like imitation, repetitive, monotonous and conservative and it would lead to a reduction of complexity. The answer to this objection is that this is precisely what sometimes happens, but fortunately not always. The question therefore is that of what are the conditions for this to happen. Or to put it otherwise, under what conditions does mimesis give rise to a rich complex phenomenology and under which does it inevitably lead to a reduction of complexity?

The convergence of the rivals unto the same behaviour corresponds to what Girard describes as mimetic doubles or doubles of violence. Rivals become doubles of violence when in a conflict each one’s action are commanded by those of the other, to such a point that the two opponents become mirror images of each other. This happens, argues Girard, when the rivals lose sight of the object around which their conflict first began. Each focusing on the other as the only cause of the quarrel, aims to destroy or annihilate him or her. Or, as Clausewitz said of war understood as a dual, a fight whose goal is to force the adversary to comply with our will. As more and more opponents become doubles of each other the intensity of the violence becomes greater and this leads to a reduction of complexity.

As Girard says, violence destroys differences, it erases them making both the environment and the relations between the agents more simple. It destroys differences and reduces complexity in two ways. On the one hand directly, physically by destroying objects, bridges, libraries, houses, train stations, etc. Reducing a world that was structured in a way that supported and facilitated the fulfillment of numerous different objectives that is focused on only one: to survive and to win. One the other hand it erases differences as cognitive objects, as in the converging statement of the political leader of Hamas and the President of Israel, differences between children, woman, elderly people, party goers, journalists and combatant have disappeared. All and everyone from now on are only enemies.

This violent lost of complexity corresponds to the transformation of what until then was a triadic relation mediated by an object external to the opponents, into a dyadic relation of pure sociality. Of pure sociality in the sense that the relation now involves as a meaningful component nothing other than the two agents who violently interact with each other. Without any object between them, as can literally be the case in a fists fight. Or, less literally, where all objects lost their reality as independent objects and only exist as means or obstacle to the conflict, as we commonly see in wars. Progressively as the conflict intensifies whatever exists only has meaning through its effect on the reciprocal opposition of the adversaries. So it comes to pass that hospitals, children, party goers, family dwellings become legitimate targets to the advantage of one or the other party. Objects, whose sole meaning now is in relation to pursuing a conflict, have now disappeared as independent objects, and when differences are only meaningful in relation to the conflict they have disappeared as differences, for as Quine once wrote, “a difference that does not make any difference is not a difference”. This metamorphosis of objects and difference into pure means of conflict, corresponds to weaponizing them and constitutes a good definition of what is a weapon.

It is therefore, when mimesis becomes, or is simplified into a dyadic relation that it leads to a reduction of complexity, to an impoverished environment. To the opposite, as long as it remains mediated by an external object. External in the sense that it maintains an independence relative to the rivalry, whether this object is physical, cultural, legal, or intellectual, whether this object is a game, a competition, an enterprise, or a project that the mimetic relations between agents can give rise to a rich and complex world and phenomenology of behaviour.

Why is this the case? Why does the object do that? Or to give to these questions a different more general formulation: why does the passage from a dyadic to a triadic relation, or vice versa have such dramatic consequences? Why is it that in one case the relation is repetitive, monotonous, reducing complexity while in the other it becomes the occasion of change, of novelty and leads to greater complexity? Why is going from two to three so important?

I do not know the answer to this last, more abstract question, but I can suggest a hypothesis concerning the role of the object based on the differences between imitation and mimesis as analysed above. The complexity reducing aspect of imitation seems to be a direct consequence of its definition as a matching problem and this is inseparable from the fact that there is nothing between C and M, between the copy cat and the model. A behaviour of M that is to be imitated by C is not an external object. On its own it has no reality that is independent of M. Nonetheless, for C it can constitute an object and the target of his or her imitation. However, as a matching behaviour the imitation of M by C does not add anything new to the world. It can only produce a novel behaviour to the extent that it fails as a matching behaviour.

M’s desire for something can constitute a goal for C but that can only be in the form of M’s relation to the object. That relation however is unlike a discrete behaviour, there is no particular gesture, action or effect that defines it. At least that is the case as long as the object retains its reality. If I imitate your action of eating an apple, I can only match your behaviour by eating a different one. If, however, it is your desire of the apple that I imitate, what I want cannot be a different apple. It is this one which your desire has made valuable that I also desire. Our behaviours towards the apple will therefore be different, rather than literally the same, as in imitation. I will try to grab the apple; you will try to prevent me from doing that. Even in such a simple case, my mimetic behaviour will not directly match yours. The symmetry between our two behaviour will only arise in a dynamic sequence through repeated interaction. When I imitate your action of eating an apple, the apple only has meaning and existence as defining your action and not in itself. When I imitate your desire for the apple the apple has an existence and meaning in itself as the object that we both desire and its existence as an object independent of us is what prevents us from both having exactly the same relation to it.

When the object disappears our several behaviours can become strictly the same as in imitation, but not immediately for in reciprocal violence, identical actions first appear alternatively as Girard argues in his analysis of Oedipus Rex. As in a fist fight, when I punch you block and when you punch I block. However, finally at the climax of the mimetic crisis the imitation becomes literal, perfect matching behaviour, when all turn against the same and only victim.




Paul Dumouchel

Université du Québec à Montréal

 

---

[1]  See, S. Tindall, “Angry old men set the middle east ablaze. The young will pay the price.” https://www.theguardian.com/commentisfree/2023/oct/14/angry-old-men-set-the-middle-east-ablaze-the-young-will-pay-the-price-israel-gaza

[2]  See Chis. McGreal, “The language used to describe the Palestinian is Genocidal”   https://www.theguardian.com/commentisfree/2023/oct/16/the-language-being-used-to-describe-palestinians-is-genocidal

[3] C.L. Nehaniv, “Imitation and Social Learning in Robots, Humans and Animals” in Imitation and Social learning in Robots, Humans and Animals, C.L. Nehaniv & K. Dautehahn, Eds, Cambridge University Press, 2009, pp. 35-46.

[4] He tells us that this conception of imitation comes from W.R. Mitchell, “A Comparative Developmental Approach to Understanding Imitation” in Perspectives in Ethology 7: Alternatives, P.P.G. Bateson & P.H. Klopfer, Eds, New York: Plenum Press, 1987, pp. 183-215.

[5] For an introduction to the frame problem, see “The Frame Problem” in the Stanford Encyclopedia of Philosophy, available on line at https://plato.stanford.edu/entries/frame-problem/

[6] I assume here that, as Jean-Pierre Dupuy argued in “Le signe et l’envie”, double mediation, that is where A imitates B desire while B imitates A’s desire is the simplest form of mimetic desire. See P. Dumouchel & J.-P. Dupuy, L’Enfer des choses. René Girard et la logique de l’économie, Paris : Seuil, 1988.