Elbow Room

Daniel Dennett is one of the most fascinating philosophers currently living. Although he pursued a traditional (but nonetheless exceptional) course of education, over time he seems to have moved farther and farther away from traditional philosophical methods and styles, and more toward a form of engagement that is simultaneously rigorous, original, and accessible. Not only is he prolific, he is far better informed on scientific topics than any other philosopher I can think of, and seems to have quite intentionally pursued a course through the relevant fields of inquiry in order to develop a sophisticated understanding of what we are as human beings.

I had previously read Consciousness Explained and Darwin’s Dangerous Idea, (both of which are excellent), when a friend of mine proposed re-reading Elbow Room for his book group. Although I might not have opted to read this book otherwise, I enjoy Dennett’s writing enough that I was eager to participate.

The subject of the book is “free will”, which seems to be one of those issues on which most people are certain, despite their disagreements with each other. To some, nothing could be more obvious than the fact that they are conscious beings who can do as they like. Suggest to them that they are constrained, and they will be happy to act in defiance of your observation, as if that proved the point.

To those of a more materialist bent, it is equally obvious that (assuming we are correct about the causal nature of the universe) some set of physical laws determine how things unfold from moment to moment, with the present fully determined by the past, leaving no room for any kind of freedom. As Dennett says on the fist page, “[if determinism is true], then our every deed and decision is the inexorable outcome, it seems, of the sum of physical forces acting at the moment, which in turn is the inexorable outcome of the forces acting an instant before, and so on, to the beginning of time.”

The “it seems” in the above quote hints at Dennett’s position, and the subtitle of the book (“The varieties of free will worth wanting”), gives a clue as to how he will proceed. In particular, it suggests that free will is not a simple unitary concept, and that what we want is an important consideration in determining how best to think about this issue.


As we will see, Dennett is largely working in the vein of natural language philosophy. Rather than assuming there is a single commonly-accepted definition of free will (which may or may not even be coherent), and then making arguments for or against its existence, Dennett works by asking what people mean when they speak of free will, why they may fear a lack of it, and to what extent the world as it is may satisfy people’s desires, and allay their fears.

One of the delights of this book is the focus on philosophical method. Before doing anything, Dennett tells you how he’s going to do it, and then later shows you how it was done. As he explains in the first few pages, this book is not intended to be a comprehensive treatment of the topic, but rather an attempt at approaching the topic through means somewhat akin to what is done by sculptors, that is — by starting with a undistinguished block, removing material by large chunks at first, and later returning to each surface and smoothing out the rough patches.

The first chunk Dennett tackles is not the question of free will itself, but what might be at stake. In particular, he lays out a series of “bugbears” — things which people are afraid of or concerned about, which they think may be implied by the idea that we don’t have free will. To quote Dennett, “if the question of free will matters, it must be because not having free will would be awful. … [but] what exactly are we afraid of?” By laying out some of the metaphors used in past literature, such as the “Nefarious Neurosurgeon” who controls us by manipulating our neurons without our knowledge, Dennett is able to demonstrate how particular features of these examples, features which are not actually relevant to the question of free will, are in fact what produces our aversion to their implications.

This tactic is actually a specific application of a technique (borrowed from Douglas Hofstadter) which Dennett uses widely throughout this book, namely that of “tuning the knobs.” Thought experiments, which are extremely popular in certain branches of philosophy, are necessarily embedded in narratives and specificities. Although we like to pretend that such experiments illustrate an abstract idea in a neutral way, the truth is that the details matter. By manipulating these details, not only can we determine the thresholds people have for certain concerns, but also what parts of the thought experiment hold the real power. Surprisingly often, we find that the power lies in some detail which at first seemed irrelevant.

The most central and powerful analogy in the book is one he also borrows from Hofstadter, that of “sphexishness”. Referring to a particular type of wasp, the Sphex ichneumoneus, Dennett quotes a passage describing it’s behaviour, in which it drags a paralyzed cricket to the threshold of its burrow, goes inside “to check that all is well”, and then drags the cricket all the way in. Although this behaviour seems purposive to our anthropomorphizing eyes (and brains), it can actually be revealed as the result of rather elaborate programming. Specifically, if an experimenter drags the cricket away from the borrow while the sphex is inside, the sphex will proceed to drag the cricket to the threshold again, go inside to check a second time, and then emerge to drag it fully in. This operation can be repeated seemingly any number of times, without the sphex altering its repetitious behaviour.

The example of the sphex seems to get to the heart of the question of free will. We don’t want to be sphexish! We want to do things for our own reasons, for our own purposes, and because we want to do them, not because we are preprogrammed to do so. Through much of the rest of the book, Dennett carefully unpacks all that I have just carelessly said, revealing the complexity inherent in these sorts of concerns, and arguing how, in the ways that matter to us, we do have a will that is free.

As a result of the sculptural method described above, Elbow Room feels less like a single, sustained course of argumentation than a series of essays, circling round the central question while leaving many related questions less than fully developed. Part of the reason for this is that the question involves so many other fields, from biology to psychology to physics, and an author can only do so much in one relatively short book (just over 200 pages). Indeed, Dennett is remarkable precisely because he has taken on nearly all of these connections over the course of his career, writing many (and much longer) books on topics such as evolution and consciousness with similar levels of clarity and engagement with the scientific literature.

As a result of this, it is difficult to summarize the book in any brief way, and here I will just try to touch on the most important points.


The first important issue to be addressed is that of the clockwork nature of the universe. Despite the obsolete nature of that analogy, the question remains, “how could reason ever find a foothold in a material, mechanical universe?” If we are simply complex systems made of atoms, each of which is governed by deterministic laws of physics, how could our behaviour in aggregate be anything more than an emergent property of those laws? In such a universe, how could we have such a thing as a “good reason” for doing something?

As with much of the book, Dennett’s way of addressing this has more do to with questioning what we mean by “reason” than by trying to find a clever loophole in the argument. In particular, beginning from the basics of physics and evolution, Dennett writes:

“In the beginning, there were no reasons; there were only causes. Nothing had a purpose, nothing had so much as a function; there was no teleology in the world at all. The explanation for this is simple: there was nothing that had interests. But after millennia there happened to emerge simple replicators, and while they had no inkling of their interests, and perhaps properly speaking had no interests, we, peering back from our Godlike vantage point at their early days, can nonarbitrarily assign them certain interests — generated by their defining “interest” in self-replication. That is, maybe it really made no difference, was a matter of no concern, didn’t matter to anyone or anything whether or not they succeed in replicating (though it does seem we can be grateful that they did), but at least we can assign them interests conditionally. If these simple replicators are to survive and replicate, thus persisting in the face of increasing entropy, their environment must meet certain conditions: conditions conducive to replication must be present or at least frequent. (p. 24)

In other words, before the existence of anything we would call life, nothing could be said to matter to anything. Once more complex and coherent systems, such as organisms, emerged, however, the condition of the world certainly made a difference to their continued survival and reproduction. Through the process of evolution, these early organisms eventually developed mechanisms of not just making use of the environment in order to replicate, but actually modifying themselves or their environment in order to increases their chance of replication. To quote Dennett again, “The day that the universe contained entities that could take some rudimentary steps towards defending their own interests was the day that interests were born.” (p. 25)

To some, this may seem to be nothing more than wordplay, but in fact it contains the crucial idea. “Interests” (or meanings or objectives) only have meaning relative to the systems that have them. Even more importantly, not knowing about one’s interests does not prevent one from having them. For Dennett, this is the bedrock from which everything else can be constructed. Evolution produced systems that are capable of acting so as to increase the probability of their own survival. This then almost tautologically defines what is an interest to that system and what it would mean to act with reason. This is the “gradualist bridge” that Dennett uses to span the gulf between the purposeless motion of individual atoms, and the seemingly purposeful behaviour of organisms and even people.

This may be cold comfort to some, and it may still seem that there is no place in the universe for “reasons”. Indeed, objectively speaking, is hard to see how such reasons could exist, independent of those who have them. But we, as creatures who are the product of evolution, are compelled to behave in ways that are at least partially the product of natural selection, which has thereby implicitly defined interests for us. Granted, we are much more complex than the sphex; one explanation for this is that the mechanisms we possess are much more intricate, and include such things are not just predicting what the world will do, but also how we will in turn respond to the world, and the many higher order interactions than can result from such monitoring. In Hofstadter’s lovely terminology, we are ourselves “strange loops” which are capable of recursively including a model of ourselves in our model of the world.

As a result of this higher order modeling, we can indeed “look inside” our own minds (presumably to a greater degree than the sphex can), and to some extent articulate what we believe is important to us, and what we think is our reason for doing something. There is of course no guarantee that we are correct in this, and indeed, there is almost certainly a mismatch between the interests of our genes, and what we think our interests are as people. Nevertheless, this ability (by the fact that it exists), presumably conveys an evolutionary advantage. To quote Dennett again at length,

“The creation and improvement of intelligence is one of evolution’s most impressive products … The merely behavioural (or “information-processing”) capacity to “wonder” about the evidential pedigrees of one’s own “beliefs” and the soundness and coherence of one’s “desires” is a … major advance in that cognitive arms race. Even an imperfect capacity to “evaluate” some of one’s own cognitive and conative states makes a big difference. Without it, one would be particularly gullible, and hence particularly vulnerable to manipulation by any (secret) agent who could discover ways of causing one to believe this or that.” (p.42)

In other words, we are better able to secure our own interests, because we are better able to monitor (in ways that the sphex is not) for the possibility of an external agent (with its own interests) attempting to control us by exploiting the limitation of our programming. Thus, we are better able to respond to circumstances with actions that will actually favour our interests as organisms. By virtue of our power of self-reflection (and critically, Dennett argues, because we have evolved the use of language), we are increasingly able to behave in powerful, flexible, and advantageous ways that could not be easily pre-programmed by genes. At the same time, however, this naturally introduces the possibility of interests which to some extend conflict across levels. Here Dennett quote Richard Dawkins’ The Selfish Gene:

“We have the power to defy the selfish genes of our birth … We can even discuss ways of deliberately cultivating and nurturing pure, disinterested altruism — something that has no place in nature. … We, alone on earth, can rebel against the tyranny of the selfish replicators.”

As with all organisms, we are the product of both our genes and the environment. The above basically argues that we are unique in the extent to which the complexity of our behaviour is determined by the environment, and even modifiable by our own behaviour. The possible interactions start to get very complex, and as Dennett says, “intellectual vertigo does set in on occasion”. The key, however, is that we do have interests, both from our genetic heritage, and those that have arisen from our own self-reflection, and that by virtue of our genetic heritage, we are actually well-equipped to respond to circumstances in an appropriate way.

At this point, it may seem as if we have ventured rather far afield from free will, without gaining any ground. The idea that we respond appropriately may not seem very comforting, if such behaviour is still entirely deterministic. To address this, Dennett next turns to the ideas of causality and control. Control

An important point is that more “pure” notions of free will are often rather incoherent. We don’t want our behaviour to be dictated by the world, but rather from some isolated, self-directing, rational consciousness, or something like that. Aside from its incompatibility with our present understanding of physical reality, such an idea simply does not make sense. If such an abstract consciousness were outside of causality, how exactly would it unfold? Moreover, if we stop to think about it, we should want to be at least somewhat causally determined. To use Dennett’s analogy, when we consider whether to buy the can of soup in the grocery store, we want that decision to be affected by things in the real world, like the nature of the soup, and perhaps the price. Certainly the idea of all such decisions being made entirely at random should be even more horrifying. What we fear, rather, seems to be the prospect of not being in control, or rather of being controlled by someone else.

Dennett thinks this is a central element of why people resist the idea that we don’t have free will. “What we fear — or at any rate a very important part of what we fear — in determinism is the prospect that determinism would rule out control, and we very definitely do not want to lose control or be out of control or be controlled by something or someone else — like a marionette or a puppet.” (p. 56)

But again, what exactly do we mean by control? Here again, Dennett does some more linguistic philosophy and chooses to make use of a specific meaning of “control” that leads to a productive analysis. In particular, he says that “for something to be a controllee it just needs to have a variety of different states it can be in, but for something to be a controller its states must include desires — or something “like” desires — about the states of something (else).” (p. 57–58) In other words, it is not sufficient for something merely to be part of the chain of causation; we need something like intention to meaningfully speak of control.

As a specific analogy, Dennett uses the example of a model airplane. We hold the remote control, and can direct its action according to our intentions (within our ability), and so we think of ourselves as the controller of the plane, not the wind or gravity, even though these also affect its behaviour. If there were two remote controls and one plane, we could test which one is actually in control by manipulating each and observing the results. Of course, our ability to control the plane is not complete. If a hurricane came along, our ability to exert control would likely vanish.

This is an important point. There is much subtlety that I am passing over, but basically it comes down to this: for relatively simple creatures (such as the sphex), their programming can (at least in some cases) be reverse engineered, and they can thus be controlled (such as by the experimenter moving the paralyzed cricket). As beings get more complex however, it gets harder and harder to control them, as it becomes more difficult to understand the programming well enough to insure control. When it comes to humanity, our programming is complex enough that even though we may be able to exert control on average (i.e. advertising does work to at least some degree), complete control of the behaviour of an individual may be impossible, even by that individual!

As a friend of mine is fond of saying, we are part of the causal nature of the universe, and are therefore both causal and caused. As such, we are at least partially in control of our own behaviour, and can resist external attempts to control us. This, for Dennett, is what provides us with what we ultimately want: “lots of elbow room. We want a margin of error; we want to keep our options open, so that our chances of maintaining control over our operations, come what may, are enhanced.” (p. 69)

Using a different example,

“The Viking spacecraft is as deterministic a device as any clock, but this does not prevent it from being able to control itself. Fancier deterministic devices can not only control themselves; they can evade the attempts of other self-controllers to control them. If we are also deterministic devices, we need not on that account fear that we cannot be in control of ourselves and our destinies.” (p. 79).

Putting together the parts of Dennett’s argument up to this point, a coherent way to think about free will is not that we are somehow removed from the chain of causality, but rather to ask to what extent we are subject to control. By virtue of our existence as complex, reflective beings, we have interests, — not only those that stem from our genetics, but also conscious interests that have arisen from our self-reflection and interaction with the environment. Thanks to being the products of evolution, we are well attuned to further those interests within our limited (but still profound) capacity as decision makers. And because of our sophisticated brains, and our abilities to model not just the world, but ourselves in the world, we need not fear that we are too sphexish: we can (at least in theory) evade even the most sophisticated attempts of others to control us.


There is much, much more in this book, including greater depth about exactly what we believe is within our control, what we mean by selves (see also Dennett’s Consciousness Explained), the question of quantum randomness in a deterministic world, and more, but I shall leave that for you to discover yourself. The last point I want to touch on, however, is Dennett’s treatment in this book of moral responsibility, which is the only part that I find myself less in agreement with.

As many have observed, determinism seems to present difficulties for any objective sense of moral responsibility. If everything that happens is strictly deterministic, how could it be just to blame people who have done something wrong? First, I would argue, that if we take that argument seriously, it is somewhat self-defeating, in that any blame or punishment is just another part of what is unfolding deterministically, and so fairness or justice is undercut just as effectively. Dennett’s approach, by contrast, is to say that since we do in fact have some control over ourselves, it makes sense to hold people accountable, just as we hold manufacturers accountable for defective products: “I take responsibility for any thing I make and then inflict upon the general public; if my soup causes food poisoning, …, I , the manufacturer am to blame. … [Similarly,] I have created and unleashed an agent who is myself; if its acts produce harm, the manufacturer is held responsible.” (p. 93)

The challenge, as I see it, however, is that although we do have the capacity for self-control and self-improvement, the actual way our life unfolds, the sort of people we become, and our inclinations to try to aim for a certain sort of self are deeply influenced by the circumstances into which we are born.

Dennett first appeals to evolution for the idea that we should have high expectations for our fellow humans in general, arguing that evolution has shaped us to have a strong capacity for self-control (with some variation, obviously). Moreover, we expect more from people as they develop. (Certainly we don’t hold infants morally responsible, as they haven’t had the chance to develop that sort of self-control). He then acknowledges the difficulty:

“But still, one may be tempted to say, there are two sorts of differences in an agent’s circumstances that are merely matters of luck: how much initial strength or talent or character one is lucky enough to be born with, and how many lucky breaks one encounters during one’s period of self creation.” (p. 103)

He then, however, argues that moral development is more like a marathon than a sprint, in which, although people may start at different points, the distance is long enough that those initial differences can be ignored once the race has been run for long enough.

“Some people reach the plateau swiftly and easily, while others need compensatory effort to overcome initial disadvantages in one way or another.

But everyone comes out more or less in the same league. When people are deemed “good enough” their moral education is over, and except for those who are singled out as defective — retarded or psychopathic, for instance — the citizenry is held to be composed of individuals of roughly equivalent talents, insofar as the demands of such citizenship are commonly held to average out.” (p. 104)

A few paragraphs later:

“There is a tendency to treat “lucky” and “unlucky” as complementary and exhaustive, leaving no room for skill at all. On this view nothing in principle could count as skill or the result of skill. This is a mistake. Once one recognizes that there is elbow room for skill in between lucky success and unlucky failure, the troubling argument that seems to show that no one could ever be responsible evaporates. Luck averages out and skill will tell in the end.” (p. 105–6)

While I acknowledge that people have the capacity for self-control and self-improvement, and that it is reasonable to expect people to make use of these capacities, I am somewhat uncomfortable with Dennett’s characterization of luck averaging out in the end. I am personally not so convinced that the question of moral responsibility is necessarily the most interesting one, but Dennett’s thinking here seems far too close to those who view their success as entirely self-made, ignoring the numerous ways in which they have benefited from circumstance and good fortune. The differences in opportunities between people born wealthy and those born poor, or between those born in a prosperous country to those born into one torn by war or famine seem so vast that it is remarkable in the rare examples where people have been able to truly transcend these differences. If anything, knowing more about all the ways in which people are limited and lack full self-control (even ignoring more extreme cases like brain tumors), should give us more sympathy for those who behave differently than we would like.

Perhaps a more fruitful way of looking at this is through the lens of conditional probability. Given different starting points, it makes sense to have different expectations, but we should recognize that we will never have the complete picture of any individual; if outcomes are different from what we feel they “should” have been, perhaps that is is an indication that our initial model of them was incomplete or wrong.

Dennett eventually returns to some of these questions in the last chapter, where he seems to take a more pragmatic perspective. Asking why we care about moral responsibility, he essentially argues that for practical reasons, as part of shaping the sort of society in which we want to live, it may be of benefit to continue work with the notion of personal responsibility, whether or not it is entirely appropriate. As Dennett says, “holding people responsible is the best game in town.” (p. 177)

Ultimately, however, Dennett admits that this is a rather academic issue, and that there are more important real world concerns. “There are real threats to human freedom, but they are not metaphysical. There is political bondage, coercion, the manipulation inducible by the dissemination of misinformation, and the “forced move” desperation of hunger and poverty.” (p. 185)


While I don’t agree with everything that Dennett writes, I think this is a fabulous book, and I would encourage everyone to read it. It is tremendously difficult to cast fresh light on a topic that has become so hackneyed, but Dennett succeeds admirably. Moreover, the topic is arguably particularly interesting in light of recent discussion about the quickening pace of the development of artificial intelligence, genetic engineering, etc.

Although there is more to be discovered here, Dennett provides a nice summary of his view in the last few pages:

“My conclusion is optimistic: free will is not an illusion, not even an irrepressible and life-enhancing illusion.

What we want when we want free will is the power to decide our courses of action, and to decide them wisely, in the light of our expectations and desires. We want to be in control of ourselves, and not under the control of others. We want to be agents, capable of initiating, and taking responsibility for, projects and deeds. All this is ours, I have tried to show, as a natural product of our biological endowments, extended and enhanced by our initiation into society.” (p. 184)

Elbow Room: The Varieties of Free Will Worth Wanting, by Daniel C. Dennett, MIT Press, 1984, 2015.