Towards a Scientific Concept of Free Will (PDF)

Proc. R. Soc. B (2011) 278, 930–939 doi:10.1098/rspb.2010.2325...

Proc. R. Soc. B (2011) 278, 930–939 doi:10.1098/rspb.2010.2325 Published online 15 December 2010 Review Towards a scientific concept of free will as a biological trait: spontaneous actions and decision-making in invertebrates Björn Brembs* Freie Universität Berlin, Institute for Biology – Neurobiology, Königin-Luise-Strasse 28/30, 14195 Berlin, Germany Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 Until the advent of modern neuroscience, free will used to be a theological and a metaphysical concept, debated with little reference to brain function. Today, with ever increasing understanding of neurons, cir- cuits and cognition, this concept has become outdated and any metaphysical account of free will is rightfully rejected. The consequence is not, however, that we become mindless automata responding predictably to external stimuli. On the contrary, accumulating evidence also from brains much smaller than ours points towards a general organization of brain function that incorporates flexible decision- making on the basis of complex computations negotiating internal and external processing. The adaptive value of such an organization consists of being unpredictable for competitors, prey or predators, as well as being able to explore the hidden resource deterministic automats would never find. At the same time, this organization allows all animals to respond efficiently with tried-and-tested behaviours to predictable and reliable stimuli. As has been the case so many times in the history of neuroscience, invertebrate model systems are spearheading these research efforts. This comparatively recent evidence indicates that one common ability of most if not all brains is to choose among different behavioural options even in the absence of differences in the environment and perform genuinely novel acts. Therefore, it seems a reason- able effort for any neurobiologist to join and support a rather illustrious list of scholars who are trying to wrestle the term ‘free will’ from its metaphysical ancestry. The goal is to arrive at a scientific concept of free will, starting from these recently discovered processes with a strong emphasis on the neurobiological mechanisms underlying them. Keywords: invertebrate; spontaneous; volition; action 1. INTRODUCTION: THE REJECTION OF THE free will as a metaphysical entity indeed most probably METAPHYSICAL CONCEPT OF FREE WILL is an illusion. Colloquial and historical use of the term What could possibly get a neurobiologist with no formal ‘free will’ has been inextricably linked with one variant training in philosophy beyond a few introductory lectures, or another of dualism. There have been so many and to publicly voice his opinion on free will? Even worse, why thorough recounts of the free will debate, that I will use empirical, neurobiological evidence mainly from only reference some, which can serve to introduce the invertebrates to make the case? Surely, the lowly worm, concepts used here [6– 9]. Psychologists and neurobio- snail or fly cannot even be close to something as logists have rightfully pointed out for decades now that philosophical as free will? The main reason is this neuro- there is no empirical support for any form of dualism. biologist’s opinion that free will is a biological trait and The interactionism proposed by Popper & Eccles was not a metaphysical entity. ‘Free will is a biological pro- probably one of the last prominent accounts of dualism perty, not a gift or a mystery’. Today, neurobiology. Since then, these and related positions have largely has accumulated sufficient evidence that we can move fallen into irrelevance. Today, the metaphysical concept on from speculating about the existence of free will of free will is largely devoid of any support, empirical or towards plausible models of how brains have intellectual. implemented it. On the surface, this statement seems to contradict public statements from many other neurobiol- ogists who fervently deny free will. In fact, it appears that 2. THE REJECTION OF DETERMINISM if neurobiologists feel compelled to write about free will, That said, it is an all too common misconception that the they do so only to declare that it is an illusion [2– 5]. Of failure of dualism as a valid hypothesis automatically course, all of these neurobiologists are correct in that entails that brains are deterministic and all our actions are direct consequences of gene – environment inter- actions, maybe with some random stochasticity added in *[email protected] here and there for good measure. It is tempting to One contribution to a Special Feature ‘Information processing in speculate that most, if not all, scholars declaring free miniature brains’. will an illusion share this concept. However, our world Received 26 October 2010 Accepted 25 November 2010 930 This journal is # 2010 The Royal Society  Review. Free will as a biological trait B. Brembs 931 is not deterministic, not even the macroscopic world. C-start because fishes that perceive sudden pressure Quantum mechanics provides objective chance as a changes on one side of their body bend in a C-shape trace element of reality. In a very clear description of away from the perceived stimulus to escape in the oppo- how keenly aware physicists are that Heisenberg’s uncer- site direction. One of the largest neurons in vertebrate tainty principle indeed describes a property of our world nervous systems is mediating this response, the Mauthner rather than a failure of scientists to accurately measure cell (e.g. ). Recently, Kenneth Catania and colleagues it, Stephen Hawking has postulated that black holes described the hunting technique of tentacled snakes emit the radiation named after him , a phenomenon (Erpeton tentaculatus) [24,25]. The snakes hunt for based on the well-known formation of virtual particle – fishes by cunningly eliciting a C-start response in the antiparticle pairs in the vacuum of space. The process potential prey animal with a more caudal part of their thought to underlie Hawking radiation has recently been body, prompting the fish to C-start exactly into the observed in a laboratory analogue of the event horizon mouth of the snake. [12,13]. On the ‘mesoscopic’ scale, fullerenes have Some of the most important predators of earthworms famously shown interference in a double-slit experiment are moles. When moles dig through the ground, they pro-. Quantum effects have repeatedly been observed duce a very distinctive sound. Earthworms have evolved directly on the nano-scale [15,16], and superconductivity to respond to this sound by crawling to the surface, Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 (e.g. ) or Bose –Einstein condensates (e.g. ) are where the moles will not follow them. Kenneth Catania well-known phenomena. Quantum events such as radio- recently reported that the technique of ‘worm-grunting’, active decay or uncertainty in the photoelectric effect employed in order to catch earthworms as fish bait, are used to create random-number generators for crypto- exploits this response. The worm grunters use a combi- graphy that cannot be broken into. Thus, quantum effects nation of wooden poles and metal rods to generate the are being observed also on the macroscopic scale. There- sound and then collect the worms from the surface. fore, determinism can be rejected with at least as much In the third example, another very well-studied escape empirical evidence and intellectual rigor as the metaphys- response is exploited by birds. Under most circum- ical account of free will. ‘The universe has an irreducibly stances, the highly sophisticated jump response of random character. If it is a clockwork, its cogs, springs, dipteran flies is perfectly sufficient to catapult the animals and levers are not Swiss-made; they do not follow a pre- out of harm’s way (e.g. ). However, painted redstarts determined path. Physical indeterminism rules in the (Myioborus pictus) are ground-hunting birds that flush out world of the very small as well as in the world of the dipterans by eliciting their jump response with dedicated very large’. display behaviours. Once the otherwise well-camouflaged flies have jumped, they are highly visible against the bright sky and can be caught by the birds [28,29]. 3. BEHAVIOURAL VARIABILITY AS It is not a huge leap to generalize these insights from AN ADAPTIVE TRAIT escape responses to other behaviours. Predictability can If dualism is not an option and determinism is equally never be an evolutionarily stable strategy. Instead, animals untenable, what other options are we left with? Some scho- need to balance the effectiveness and efficiency of their lars have resorted to quantum uncertainty in the brain as behaviours with just enough variability to spare them the solution, providing the necessary discontinuity in the from being predictable. Efficient responses are controlled causal chain of events. This is not unrealistic, as there is by the environment and thus vulnerable. Conversely, evidence that biological organisms can evolve to take endogenously controlled variability reduces efficiency but advantage of quantum effects. For instance, plants use increases vital unpredictability. Thus, in order to survive, quantum coherence when harvesting light in their photo- every animal has to solve this dilemma. It is no coincidence synthetic complexes [19–22]. Until now, however, it has that ecologists are very familiar with a prominent, analo- proved difficult to find direct empirical evidence in support gous situation, the exploration/exploitation dilemma of analogous phenomena in brains. Moreover, and (originally formulated by March ): every animal, more importantly, the pure chance of quantum indeter- every species continuously faces the choice between staying minism alone is not what anyone would call ‘freedom’. and efficiently exploiting a well-known, but finite resource ‘For surely my actions should be caused because I want and leaving to find a new, undiscovered, potentially much them to happen for one or more reasons rather that they richer, but uncertain resource. Efficiency (or optimality) happen by chance’. This is precisely where the biologi- always has to be traded off with flexibility in evolution, cal mechanisms underlying the generation of behavioural on many, if not all, levels of organization. variability can provide a viable concept of free will. A great invertebrate example of the sort of Protean Biologists need not resort to quantum mechanics to behaviour [31,32] selected for by these trade-offs is yet understand that deterministic behaviour can never be another escape behaviour, that of cockroaches. The evolutionarily stable. Evolution is a competitive business cerci of these insects have evolved to detect minute air and predictability is one thing that will make sure that a movements. Once perceived, these air movements trigger competitor will be out of business soon. There are an escape response in the cockroach away from the side many illuminating examples of selection pressures favour- where the movement was detected. However, which ing unpredictability, but three recently published reports angle with respect to the air movement is taken by the dealing with one of the most repeatable and hence best- animal cannot be predicted precisely, because this com- studied class of behaviours are especially telling. These ponent of the response is highly variable. examples concern escape behaviours. Therefore, in contrast to the three examples above, it is One of the most well-studied escape behaviours is the impossible for a predator to predict the trajectory of the so-called C-start response in fishes. The response is called escaping animal. Proc. R. Soc. B (2011)  932 B. Brembs Review. Free will as a biological trait 4. BRAINS ARE IN CONTROL OF VARIABILITY predictable to random-like—are directly influenced by Competitive success and evolutionary fitness of all reinforcement. For instance, consummatory feeding ambulatory organisms depend critically on intact behaviour of the marine snail Aplysia is highly variable behavioural variability as an adaptive function. [44,45]. Recent evidence suggests that the seemingly Behavioural variability is an adaptive trait and not rhythmic cycling of biting, swallowing and rejection ‘noise’. Not only biologists are aware of the fitness movements of the animal’s radula (a tongue-like organ) advantages provided by unpredictable behaviour, but vary in order to be able to adapt to varying food sources philosophers also realized the adaptive advantages of. In fact, much like the reduced variability in flies behavioural variability and their potential to serve as a trained to avoid heat in the self-learning paradigm model for a scientific account of free will, as long as explained above, Aplysia can be trained to reduce the 25 years ago (e.g. ). The ultimate causes of variability in their feeding behaviour and generate rhyth- behavioural variability are thus well understood. The mic, stereotyped movements [47 –52]. It also takes proximate causes, however, are much less studied. One practice for snails to become efficient and predictable. of the few known properties is that the level of variability The default state is to behave variably and unpredictably. also can vary. Faced with novel situations, humans and The mechanisms to control behavioural variability are most animals spontaneously increase their behavioural in place also in humans. For instance, depression and Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 variability [35 – 38]. Animals even vary their behaviour autism are characterized by abnormally stereotypic beha- when a more stereotyped behaviour would be more viours and a concomitant lack of behavioural variability. efficient. Patients suffering from such psychopathologies can These observations suggest that there must be mech- learn to vary their behaviours when reinforced for doing anisms by which brains control the variability they inject so [53,54]. Also, the interactions between world- and into their motor output. Some components of these self-learning seem to be present in vertebrates: extended mechanisms have been studied. For instance, tethered training often leads to so-called habit formation, repeti- flies can be trained to reduce the range of the variability tive responses, controlled by environmental stimuli (e.g. in their turning manoeuvres. For example, one [55,56]). It is intriguing that recent fMRI studies have such stationary flying fly may be trained to cease generat- discovered a so-called default-mode network in humans, ing left-turning manoeuvres by heating the fly (with an the fluctuations in which can explain a large degree of infrared heat beam) whenever it initiates such actions the individual’s behavioural variability , and that and by not heating it during right-turning manoeuvres. abnormalities in this default network are associated with Before such training, it would generate left- and right- most psychiatric disorders [58 – 60]. turning manoeuvres in equal measure. Protein kinase C activity is required for such a reduction. Interestingly, analogous to the exploration–exploitation dilemma men- 5. WHAT ARE THE NEURAL MECHANISMS tioned above, the mechanism by which the animals learn GENERATING BEHAVIOURAL VARIABILITY? to decrease their behavioural variability (‘self-learning’) It thus appears that behavioural variability is a highly interacts with the learning mechanism by which the ani- adaptive trait, under constant control of the brain balan- mals learn about external stimuli (‘world learning’). Part cing the need for variability with the need for efficiency. of this interaction balances self- and world learning such How do brains generate and control behavioural variabil- that self-learning (i.e. the endogenous reduction in ity in this balance? These studies have only just begun. As behavioural variability) is slowed down whenever the was the case in much of neuroscience’s history, be it ion world-learning mechanism is engaged. This part of the channels, genes involved in learning and memory, electri- interaction is mediated by a subpopulation of neurons in cal synapses or neurotransmitters, invertebrate model a part of the insect brain called the mushroom bodies systems are leading the way in the study of the neural [42,43]. This population of neurons ensures that animals mechanisms underlying behavioural variability as well. preferentially learn from their environment and reduce Two recent reports, concerned another highly reprodu- their endogenous behavioural variability only when there cible (and therefore well-studied) behaviour, optomotor are good reasons for doing so. Such an organization may responses. Tethered flies respond to a moving grating in underlie the need for practice in order to reduce our front of them with characteristic head movements in the behavioural variability when learning new skills, e.g. the same direction as the moving grating, aimed at stabilizing basketball free-throw or the golf swing. The parallel to the image on the retina. By recording from motion- the exploration–exploitation dilemma lies in the balance sensitive neurons in fly optic lobes, the authors found between the endogenous and exogenous processing these that the variability in these neurons did not suffice to interactions bestow upon the animal: learning about the explain the variability in the head movements [61,62]. Pre- world first allows the animal to keep its behaviour flexible sumably, downstream neurons in the central brain inject in case the environment changes, while at the same time additional variability, not present in the sensory input, being able to efficiently solve the experimental task. If, which is reflected in the behaviour. however, it turns out that the environment does not A corresponding conclusion can be drawn from two change, then—and only then—is the circuitry controlling earlier studies, which independently found that the tem- the behaviour itself modified, to more permanently alter poral structure of the variability in spontaneous turning the behaviour-generating process itself and thereby maxi- manoeuvres both in tethered and in free-flying fruitflies mize on efficiency by reducing the endogenous variability. could not be explained by random system noise [63,64]. Animals other than insects also learn to control their Instead, a nonlinear signature was found, suggesting variability using feedback from the environment, such that fly brains operate at criticality, meaning that they that levels of behavioural variability—from highly are mathematically unstable, which, in turn, implies an Proc. R. Soc. B (2011)  Review. Free will as a biological trait B. Brembs 933 evolved mechanism rendering brains highly susceptible to 6. DETERMINISM VERSUS INDETERMINISM IS A the smallest differences in initial conditions and amplify- FALSE DICHOTOMY ing them exponentially. Put differently, fly brains Together with Hume, most would probably subscribe to have evolved to generate unpredictable turning the notion that ‘tis impossible to admit of any medium manoeuvres. The default state also of flies is to behave betwixt chance and an absolute necessity’. For variably. Ongoing studies are trying to localize the brain example, Steven Pinker (1997, p. 54) concurs that ‘A circuits giving rise to this nonlinear signature. random event does not fit the concept of free will any Results from studies in walking flies indicate that at least more than a lawful one does, and could not serve as the some component of variability in walking activity is under long-sought locus of moral responsibility’. However, the control of a circuit in the so-called ellipsoid body, deep to consider chance and lawfulness as the two mutually in the central brain. The authors tested the temporal exclusive sides of our reality is only one way to look at structure in spontaneous bouts of activity in flies walking the issue. The unstable nonlinearity, which makes back and forth individually in small tubes and found that brains exquisitely sensitive to small perturbations, may the power law in their data disappeared if a subset of neur- be the behavioural correlate of amplification mechanisms ons in the ellipsoid body was experimentally silenced. such as those described for the barrel cortex. This Analogous experiments have recently been taken up inde- nonlinear signature eliminates the two alternatives, Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 pendently by another group and the results are currently which both would run counter to free will, namely com- being evaluated. The neurons of the ellipsoid body plete (or quantum) randomness and pure, Laplacian of the fly also exhibit spontaneous activity in live imaging determinism. These represent opposite and extreme end- experiments , suggesting a default-mode network points in discussions of brain functioning, which hamper also might exist in insects. the scientific discussion of free will. Instead, much like Even what is often presented to students as ‘the simplest evolution itself, a scientific concept of free will comes to behaviour’, the spinal stretch reflex in vertebrates, contains lie between chance and necessity, with mechanisms incor- adaptive variability. Via the cortico-spinal tract, the motor porating both randomness and lawfulness. The Humean cortex injects variability into this reflex arc, making it dichotomy of chance and necessity is invalid for complex variable enough for operant self-learning [68–72]. processes such as evolution or brain functioning. Such Jonathan Wolpaw and colleagues can train mice, rats, mon- phenomena incorporate multiple components that are keys and humans to produce reflex magnitudes either both lawful and indeterminate. This breakdown of the larger or smaller than a previously determined baseline determinism/indeterminism dichotomy has long been precisely because much of the deviations from this baseline appreciated in evolution and it is surprising to observe are not noise but variability deliberately injected into the the lack of such an appreciation with regard to brain reflex. Thus, while invertebrates lead the way in the bio- function among some thinkers of today (e.g. ). logical study of behavioural variability, the principles Stochasticity is not a nuisance, or a side effect of our discovered there can be found in vertebrates as well. reality. Evolution has shaped our brains to implement One of the common observations of behavioural varia- ‘stochasticity’ in a controlled way, injecting variability bility in all animals seems to be that it is not entirely ‘at will’. Without such an implementation, we would random, yet unpredictable. The principle thought to not exist. underlie this observation is nonlinearity. Nonlinear sys- A scientific concept of free will cannot be a qualitative tems are characterized by sensitive dependence on initial concept. The question is not any more ‘do we have free conditions. This means such systems can amplify tiny will?’; the questions is now: ‘how much free will do we disturbances such that the states of two initially almost have?’; ‘how much does this or that animal have?’. Free identical nonlinear systems can diverge exponentially will becomes a quantitative trait. from each other. Because of this nonlinearity, it does not matter (and it is currently unknown) whether the ‘tiny disturbances’ are objectively random as in quantum 7. INITIATING ACTIVITY: ACTIONS VERSUS randomness or whether they can be attributed to system, RESPONSES or thermal noise. What can be said is that principled, Another concept that springs automatically from quantum randomness is always some part of the pheno- acknowledging behavioural variability as an adaptive menon, whether it is necessary or not, simply because trait is the concept of actions. In contrast to responses, quantum fluctuations do occur. Other than that it must actions are behaviours where it is either impossible to be a non-zero contribution, there is currently insufficient find an eliciting stimulus or where the latency and/or data to quantify the contribution of such quantum magnitude of the behaviour vary so widely, that the randomness. In effect, such nonlinearity may be imagined term ‘response’ becomes useless. as an amplification system in the brain that can either A long history of experiments on flies provides increase or decrease the variability in behaviour by accumulating evidence that the behaviour of these ani- exploiting small, random fluctuations as a source for mals is much more variable than it would need to be, generating large-scale variability. A general account of given the variability in the neurons mediating the stimu- such amplification effects had already been formulated lus-response chain (reviewed in ). For instance, in as early as in the 1930s. Interestingly, a neuronal the study of the temporal dynamics of turning behaviours amplification process was recently observed directly in in tethered flies referenced above , one situation the barrel cortex of rodents, opening up the intriguing recorded fly behaviour in constant stimulus conditions, perspective of a physiological mechanism dedicated to i.e. nothing in the exquisitely controlled environment of generating neural (and by consequence behavioural) the animals changed while the turning movements were variability. recorded. Yet, the flies kept producing turning Proc. R. Soc. B (2011)  934 B. Brembs Review. Free will as a biological trait movements throughout the experiment as if there had one end of a tube and a light at the other end, the flies been stimuli in their environment. Indeed, the temporal will run to the light. But I noticed that not every fly will structure in these movements was qualitatively the same run every time. If you separate the ones that ran or did compared with when there were stimuli to be perceived. not run and test them again, you find, again, the same per- This observation is only one of many demonstrating the centage will run. But an individual fly will make its own endogenous character of behavioural variability. Even decision’. (cited from Brown & Haglund (1994) J. NIH though there was nothing in the environment prompting Res. 6, 66–73). Not even 10 years later, Quinn et al. separ- the animals to change their behaviour, they kept initiating ated flies, conditioned to avoid one of two odours, into turning manoeuvres in all directions. Clearly, each of those that did avoid the odour and those that did not. In these manoeuvres was a self-initiated, spontaneous a subsequent second test, they found that both the avoiders action and not a response to some triggering, external and the non-avoiders separated along the same percentages stimulus. as in the first test, prompting the authors to conclude: In fact, such self-initiated actions are a necessary ‘This result suggests that the expression of learning is prerequisite for the kind of self-learning described above probabilistic in every fly’. Training shifted the initial [41–43]. At the start of the experiment, the fly cannot 50–50 decision of the flies away from the punished know that it is its own turning manoeuvres that cause the odour, but the flies still made the decisions themselves— Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 switch from cold to hot and vice versa. To find out, the only with a different probability than before training. fly has to activate the behavioural modules it has available Most recently, in the experiments described above, the teth- in this restrained situation and has to register whether one ered flies without any feedback made spontaneous of them might have an influence on the punishing heat decisions to turn one way or another. These are only beam. There is no appropriate sensory stimulus from out- three examples from more than 40 years in which many side to elicit the respective behaviour. The fly must have a behavioural manifestations of decision-making in the fly way to initiate its behaviours itself, in order to correlate brain have been observed. Like heat, flies can control also these actions with the changes in the environment. Clearly, odour intensity with their yaw torque. They can con- the brain is built such that under certain circumstances the trol the angular velocity of a panorama surrounding them items of the behavioural repertoire can get released not only by yaw torque but also by forward thrust, body independent of sensory stimuli. posture or abdomen bending. In ambiguous sensory The fly cannot know the solutions to most real-life pro- situations, they actively switch between different perceptual blems. Beyond behaving unpredictably to evade predators hypotheses, they modify their expectations about the conse- or outcompete a competitor, all animals must explore, quences of their actions by learning and they can actively must try out different solutions to unforeseen problems. shift their focus of attention restricting their behavioural Without behaving variably, without acting rather than responses to parts of the visual field [83,84]. These latest passively responding, there can be no success in evol- studies prompted further research into the process of the ution. Those individuals who have found the best endogenous direction of selective attention in flies balance between flexible actions and efficient responses [85–89]. Martin Heisenberg realized early on that are the ones who have succeeded in evolution. It is this such active processes entail the sort of fundamental free- potential to behave variably, to initiate actions indepen- dom required for a modern concept of free will and keeps dently of the stimulus situation, which provides animals prominently advocating this insight today. with choices. John Searle has described free will as the belief ‘that we could often have done otherwise than we in fact did’. Taylor & Dennett cite the maxim ‘I could have done 8. FREEDOM OF CHOICE otherwise’. Clearly, leeches and flies could and can The neurobiological basis of decision-making can also be behave differently in identical environments. While studied very well in invertebrate models. For instance, some argue that unpredictable (or random) choice does isolated leech nervous systems chose either a swimming not qualify for their definition of free will , it is pre- motor programme or a crawling motor programme to cisely the freedom from the chains of causality that an invariant electrical stimulus [78 – 80]. Every time the most scholars see as a crucial prerequisite for free will. stimulus is applied, a set of neurons in the leech ganglia Importantly, this freedom is a necessary but not a suffi- goes through a so far poorly understood process of cient component of free will. In order for this freedom decision-making to arrive either at a swimming or at a to have any bearing on moral responsibility and culpabil- crawling behaviour. The stimulus situation could not be ity in humans, more than mere randomness is required. more perfectly controlled than in an isolated nervous Surely, no one would hold a person responsible for any system, excluding any possible spurious stimuli reaching harm done by the random convulsions during an epileptic sensory receptors unnoticed by the experimenter. In seizure. Probably because of such considerations, two- fact, even hypothetical ‘internal stimuli’, generated some- stage models of free will have been proposed already how by the animal must in this case be coming from the many decades ago, first by James , later also by nervous system itself, rendering the concept of ‘stimulus’ Henri Poincaré, Arthur Holly Compton, Karl Popper, in this respect rather useless. Yet, under these ‘carefully Henry Margenau, Daniel Dennett, Robert Kane, John controlled experimental circumstances, the animal Martin Fisher, Alfred Mele, Stephen Kosslyn, Bob behaves as it damned well pleases’ (Harvard Law of Doyle and Martin Heisenberg (cited, reviewed and dis- Animal Behaviour). cussed in ), as well as Koch : one stage generates Seymour Benzer, one of the founders of Neurogenetics, behavioural options and the other one decides which of captured this phenomenon in the description of his first those actions will be initiated. Put simply, the first stage phototaxis experiments in 1967: ‘... if you put flies at is ‘free’ and the second stage is ‘willed’. This implies Proc. R. Soc. B (2011)  Review. Free will as a biological trait B. Brembs 935 that not all chance events in the brain must manifest able to control our freedom up to a certain degree. Com- themselves immediately in behaviour. Some may be elimi- pare, for instance, a line that you quickly drew on a piece nated by deterministic ‘selection’ processes before they of paper, with a line that was drawn with the conscious can exert any effects. Analogous to mutation and selec- effort of making it as straight as possible. However, the tion in evolution, the biological process underlying free neural principle underlying the process generating the will can be conceptualized as a creative, spontaneous, variability is beyond total conscious control, requiring indeterministic process followed by an adequately deter- us to use rulers for perfectly straight lines. Therefore, mined process, selecting from the options generated by the famous experiments of Benjamin Libet and others the first process. Freedom arises from the creative and since then [2,4,5,98 – 100] only serve to cement the rejec- indeterministic generation of alternative possibilities, tion of the metaphysical concept of free will and are not which present themselves to the will for evaluation and relevant for the concept proposed here. Conscious reflec- selection. The will is adequately determined by our tion, meditation or discussion may help with difficult reasons, desires and motives—by our character—but it is decisions, but this is not even necessarily the case. The not pre-determined. John Locke (1689, p. 148) already degree to which our conscious efforts can affect our separated free from ‘will’, by attributing free to the agent decisions is therefore central to any discussion about the and not the will: ‘I think the question is not proper, degree of responsibility our freedom entails, but not to Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 whether the will be free, but whether a man be free’ the freedom itself.. Despite the long tradition of two-stage models of free will, only now are the first, tangible scientific pieces of evidence being published. For instance, the independent 10. THE SELF AND AGENCY discovery of nonlinear mechanisms in brains from different In contrast to consciousness, an important part of a scien- phyla is compatible with such two-stage models [63,74]. tific concept of free will is the concept of ‘self ’. It is Essentially, the existence of neural circuits implementing important to realize that the organism generates an a two-stage model of free will ‘would mean that you can action itself, spontaneously. In chemistry, spontaneous know everything about an organism’s genes and environ- reactions occur when there is a chemical imbalance. The ment yet still be unable to anticipate its caprices’. system is said to be far from thermodynamic equilibrium. Importantly, this inability is not due to inevitable stochas- Biological organisms are constantly held far from equili- ticity beyond control; it is due to dedicated brain processes brium, they are considered open thermodynamic that have evolved to generate unpredictable, spontaneous systems. However, in contrast to physical or chemical actions in the face of pursuit–evasion contests, open systems, some of the spontaneous actions initiated competition and problem-solving. by biological organisms help keep the organism away from equilibrium. Every action that promotes survival or acquires energy sustains the energy flow through the 9. CONSCIOUSNESS AND FREEDOM open system, prompting Georg Litsche to define biological It thus is no coincidence that we all feel that we possess a organisms as a separate class of open systems (i.e. ‘sub- certain degree of freedom of choice. It makes sense that jects’; ). Because of this constant supply of energy, depriving humans of such freedom is frequently used as it should not be surprising to scientists that actions can punishment and the deprived do invariably perceive this be initiated spontaneously and need not be released by limited freedom as undesirable. This experience of free- external stimuli. In controlled situations where there dom is an important characteristic of what it is like to be cannot be sufficient causes outside the organism to make human. It stems in part from our ability to behave variably. the organism release the particular action, the brain Voltaire expressed this intuition in saying ‘Liberty then is initiates behaviour from within, potentially using a two- only and can be only the power to do what one will’. stage process as described above. The boy ceases to play The concept that we can decide to behave differently and jumps up. This sort of impulsivity is a characteristic even under identical circumstances underlies not only of children every parent can attest to. We do not describe our justice systems. Electoral systems, our educational sys- the boy’s action with ‘some hidden stimuli made him tems, parenting and basically all other social systems also jump’—he jumped of his own accord. The jump has all presuppose behavioural variability and at least a certain the qualities of a beginning. The inference of agency in degree of freedom of choice. Games and sports would be ourselves, others and even inanimate objects is a central predictable and boring without our ability of constantly component of how we think. Assigning agency requires a changing our behaviour in always the same settings. concept of self. How does a brain know what is self? The data reviewed above make clear that the special One striking characteristic of actions is that an animal property of our brain that provides us with this freedom normally does not respond to the sensory stimuli it causes surely is independent of consciousness. Consciousness is by its own actions. The best examples are that it is diffi- not a necessary prerequisite for a scientific concept of cult to tickle oneself and that we do not perceive the free will. Clearly, a prisoner is regarded as un-free, irre- motion stimuli caused by our own eye saccades or the spective of whether he is aware of it or not. John Austin darkness caused by our eye blinks. The basic distinction provides another instructive example ‘Consider the between self-induced (re-afferent) and externally gener- case where I miss a very short putt and kick myself because ated (ex-afferent) sensory stimuli has been formalized I could have holed it’. We sometimes have to work by von Holst & Mittelstaedt. The two physiologists extremely hard to constrain our behavioural variability studied hoverflies walking on a platform surrounded by a in order to behave as predictably as possible. Sports com- cylinder with black and white vertical stripes. As long as mentators often use ‘like a machine’ to describe very the cylinder was not rotated, the animals seemed to efficient athletes. Like practice, conscious efforts are behave as if they were oblivious to the stripes. However, Proc. R. Soc. B (2011)  936 B. Brembs Review. Free will as a biological trait as soon as the cylinder was switched on to rotate around to at least start a thought process that abandoning the the flies, the animals started to turn in register with the metaphysical concept of free will does not automatically moving stripes, in an attempt to stabilize their orientation entail that we are slaves of our genes and our environ- with respect to the panorama. Clearly, when the animals ment, forced to always choose the same option when turned themselves, their eyes perceived the same motion faced with the same situation. In fact, I am confident I stimuli as when the cylinder was rotated. The two scien- have argued successfully that we would not exist if our tists concluded that the animals detect which of these brains were not able to make a different choice even in otherwise very similar motion signals are generated by the face of identical circumstances and history. In this the flies and which are not and dubbed this the ‘principle article, I suggest re-defining the familiar free will in scien- of reafference’. To test the possibility that the flies just tific terms rather than giving it up, only because of the blocked all visual input during self-initiated locomotion, historical baggage all its connotations carry with them. the experimenters glued the heads of the animals rotated One may argue that ‘volition’ would be a more suitable by 1808 such that the positions of the left and right eye term, less fraught with baggage. However, the current were exchanged and the proboscis pointed upwards. connotations of volition as ‘willpower’ or the forceful, Whenever these ‘inverted’ animals started walking in conscious decision to behave against certain motivations the stationary striped cylinder, they ran in constant, render it less useful and less general a term than free Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 uncontrollable circles, showing that they did perceive will. Finally, there may be a societal value in retaining the relative motion of the surround. From this free will as a valid concept, since encouraging a belief in experiment, von Holst and Mittelstaedt concluded that determinism increases cheating. I agree with the self-generated turning comes with the expectation of a criticism that retention of the term may not be ideal, visual motion signal in the opposite direction that is per- but in the absence of more suitable terms, free will; ceived but normally does not elicit a response. If the remains the best option. visual motion signal is not caused by the animal, on the I no longer agree that ‘ ‘‘free will’’ is (like ‘‘life’’ and other hand, it most probably requires compensatory ‘‘love’’) one of those culturally useful notions that action, as this motion was not intended and hence not become meaningless when we try to make them ‘‘scienti- expected. The principle of reafference is the mechanism fic’’ ’. The scientific understanding of common by which we realize which portion of the incoming sen- concepts enrich our lives, they do not impoverish them, sory stream is under our own control and which portion as some have argued. This is why scientists have is not. This is how we distinguish between those sensory and will continue to try and understand these concepts stimuli that are consequences of our own actions and scientifically or at least see where and how far such those that are not. Distinguishing self from ‘world’ is attempts will lead them. It is not uncommon in science the prerequisite for the evolution of separate learning to use common terms and later realize that the familiar, mechanisms for self- and world learning, respectively intuitive understanding of these terms may not be all , which is the central principle of how brains balance that accurate. Initially, we thought atoms were indivisible. actions and responses. The self/world distinction is thus Today we do not know how far we can divide matter. the second important function of behavioural variability, Initially, we thought species were groups of organisms besides making the organism harder to predict: by using that could be distinguished from each other by anatomical the sensory feedback from our actions, we are constantly traits. Today, biologists use a wide variety of species defi- updating our model of how the environment responds to nitions. Initially, we thought free will was a metaphysical our actions. Animals and humans constantly ask: What entity. Today, I am joining a growing list of colleagues happens if I do this? The experience of willing to do who are suggesting it is a quantitative, biological trait, a something and then successfully doing it is absolutely natural product of physical laws and biological evolution, central to developing a sense of self and that we are in a function of brains, maybe their most important one. control (and not being controlled). Concepts and ideas in several sections of this article have Thus, in order to understand actions, it is necessary to been adapted from a to-be-published presentation of introduce the term self. The concept of self necessarily Martin Heisenberg. I am very grateful for his sharing this follows from the insight that animals and humans initiate presentation with me. I am also indebted to Christopher behaviour by themselves. It would make no sense to Harris, Bob Doyle, Matt Leifer, Sandeep Gautam, assign a behaviour to an organism if any behavioural Andrew Lang, Julien Colomb and two anonymous activity could, in principle, be traced back by a chain of referees for helpful comments on an earlier version of the manuscript. causations to the origin of the universe. An animal or human being is the agent causing a behaviour, as long as no sufficient causes for this activity to occur are coming from outside the organism. Agency is assigned REFERENCES to entities who initiate actions themselves. Agency is cru- 1 Doyle, R. O. 2009 Free will: it’s a normal biological cial for moral responsibility. Behaviour can have good or property, not a gift or a mystery. Nature 459, 1052. bad consequences. It is the agent for whom the con- (doi:10.1038/4591052c) sequences matter the most and who can be held 2 Cashmore, A. R. 2010 The Lucretian swerve: the bio- logical basis of human behavior and the criminal responsible for them. justice system. Proc. Natl Acad. Sci. USA 107, 4499– 4504. (doi:10.1073/pnas.0915161107) 3 Greene, J. & Cohen, J. 2004 For the law, neuroscience 11. WHY STILL USE THE TERM FREE WILL TODAY? changes nothing and everything. Phil. Trans. R. By providing empirical data from invertebrate model sys- Soc. Lond. B 359, 1775– 1785. (doi:10.1098/rstb.2004. tems supporting a materialistic model of free will, I hope 1546) Proc. R. Soc. B (2011)  Review. Free will as a biological trait B. Brembs 937 4 Murphy, N., Ellis, G. F. R. & O’Connor, T. 2009 Down- resolved coherent optical wave-mixing. Phys. Rev. Lett. ward causation and the neurobiology of free will. Berlin, 102, 057402. (doi:10.1103/PhysRevLett.102.057402) Heidelberg: Springer. (doi:10.1007/978-3-642-03205-9). 23 Korn, H. & Faber, D. 2005 The Mauthner cell half a 5 Soon, C. S., Brass, M., Heinze, H.-J. & Haynes, D. century later: a neurobiological model for decision- 2008 Unconscious determinants of free decisions in making? Neuron 47, 13–28. (doi:10.1016/j.neuron. the human brain. Nat. Neurosci. 11, 543 –545. (doi:10. 2005.05.019) 1038/nn.2112) 24 Catania, K. C. 2009 Tentacled snakes turn C-starts to 6 Dennett, D. C. 1984 Elbow room: the varieties of free will their advantage and predict future prey behavior. Proc. worth wanting. Cambridge, MA: MIT Press. Natl Acad. Sci. USA 106, 11 183 –11 187. (doi:10. 7 Doyle, R. 2010 Jamesian free will, the two-stage model 1073/pnas.0905183106) of William James. William James Stud. 5, 1 –28. 25 Catania, K. C. 2010 Born knowing: tentacled snakes 8 Kane, R. (ed.) 2002 The Oxford handbook of free will. innately predict future prey behavior. PLoS ONE 5, New York, NY: Oxford University Press. e10953. (doi:10.1371/journal.pone.0010953). 9 Koch, C. 2009 Free will, physics, biology, and the brain. 26 Catania, K. C. 2008 Worm grunting, fiddling, and In Downward causation and the neurobiology of free will charming—humans unknowingly mimic a predator to (eds N. Murphy, G. Ellis & T. O’Connor), pp. 31–52. harvest bait. PLoS ONE 3, e3472. (doi:10.1371/ Berlin and Heidelberg, Germany: Springer. journal.pone.0003472) 10 Popper, K. R. & Eccles, J. C. 1977 The self and its 27 Card, G. & Dickinson, M. H. 2008 Visually mediated Downloaded from https://royalsocietypublishing.org/ on 14 December 2024 brain—an argument for interactionism. Heidelberg, motor planning in the escape response of Drosophila. Germany: Springer. Curr. Biol. 18, 1300–1307. (doi:10.1016/j.cub.2008. 11 Hawking, S. 1996 A brief history of time. New York, NY: 07.094) Bantam. 28 Jablonski, P. G. & Strausfeld, N. J. 2001 Exploitation of 12 Belgiorno, F., Cacciatori, S. L., Clerici, M., Gorini, V., an ancient escape circuit by an avian predator: relation- Ortenzi, G., Rizzi, L., Rubino, E., Sala, V. G. & Faccio, ships between taxon-specific prey escape circuits and the D. 2010 Hawking radiation from ultrashort laser pulse sensitivity to visual cues from the predator. Brain Behav. filaments. Phys. Rev. Lett. 105, 203901. (doi:10.1103/ Evol. 58, 218 –240. PhysRevLett.105.203901) 29 Jabloński, P. G. & Strausfeld, N. J. 2000 Exploitation of 13 Leonhardt, U. 2002 A laboratory analogue of the event an ancient escape circuit by an avian predator: prey sen- horizon using slow light in an atomic medium. Nature sitivity to model predator display in the field. Brain 415, 406–409. (doi:10.1038/415406a) Behav. Evol. 56, 94–106. 14 Arndt, M., Nairz, O., Vos-Andreae, J., Keller, C., Der 30 March, J. G. 1991 Exploration and exploitation in Zouw, G. & van Zeilinger, A. 1999 Wave-particle dua- organizational learning. Organ. Sci. 2, 71–88. (doi:10. lity of C60 molecules. Nature 401, 680 –682. (doi:10. 1287/orsc.2.1.71) 1038/44348) 31 Driver, P. M. & Humphries, N. 1988 Protean behavior: 15 Gaidarzhy, A., Zolfagharkhani, G., Badzey, R. & the biology of unpredictability. Oxford, UK: Oxford Mohanty, P. 2005 Evidence for quantized displacement University Press. in macroscopic nanomechanical oscillators. Phys. Rev. 32 Miller, G. F. 1997 Protean primates: the evolution of Lett. 94, 030402. (doi:10.1103/PhysRevLett.94.030402) adaptive unpredictability in competition and courtship. 16 Ghosh, S., Rosenbaum, T. F., Aeppli, G. & In Machiavellian Intelligence II: extensions and evaluations Coppersmith, S. N. 2003 Entangled quantum state of (eds A. Whiten & R. W. Byrne), pp. 312 –340. magnetic dipoles. Nature 425, 48–51. (doi:10.1038/ Cambridge, UK: Cambridge University Press. nature01888) 33 Domenici, P., Booth, D., Blagburn, J. M. & Bacon, J. P. 17 Sahu, M., Bae, M.-H., Rogachev, A., Pekker, D., Wei, 2008 Cockroaches keep predators guessing by using T.-C., Shah, N., Goldbart, P. M. & Bezryadin, A. preferred escape trajectories. Curr. Biol. 18, 2009 Individual topological tunnelling events of a quan- 1792–1796. (doi:10.1016/j.cub.2008.09.062) tum field probed through their macroscopic 34 Grobstein, P. 1994 Variability in behavior and the consequences. Nat. Phys. 5, 503–508. (doi:10.1038/ nervous system. In The encyclopedia of human behavior nphys1276) (ed. V. S. Ramachandran), pp. 447 –458. New York, 18 van Zoest, T. et al. 2010 Bose –Einstein condensation in NY: Academic Press. microgravity. Science 328, 1540–1543. (doi:10.1126/ 35 Bunzeck, N. & Duzel, E. 2006 Absolute coding of science.1189164) stimulus novelty in the human substantia nigra/ 19 Collini, E., Wong, C. Y., Wilk, K. E., Curmi, P. M. G., VTA. Neuron 51, 369–379. (doi:10.1016/j.neuron. Brumer, P. & Scholes, G. D. 2010 Coherently wired 2006.06.021) light-harvesting in photosynthetic marine algae at ambi- 36 Liu, L., Davis, R. L. & Roman, G. 2007 Exploratory ent temperature. Nature 463, 644– 647. (doi:10.1038/ activity in Drosophila requires the kurtz nonvisual nature08811) arrestin. Genetics 175, 1197–1212. (doi:10.1534/ 20 Engel, G. S., Calhoun, T. R., Read, E. L., Ahn, T.-K., genetics.106.068411) Mancal, T., Cheng, Y.-C., Blankenship, R. E. & 37 Roberts, S. & Gharib, A. 2006 Variation of bar-press dur- Fleming, G. R. 2007 Evidence for wavelike energy transfer ation: where do new responses come from? Behav. Process. through quantum coherence in photosynthetic systems. 72, 215–223. (doi:10.1016/j.beproc.2006.03.003) Nature 446, 782–786. (doi:10.1038/nature05678) 38 Shahan, T. A. & Chase, P. N. 2002 Novelty, stimulus 21 Lee, H., Cheng, Y.-C. & Fleming, G. R. 2007 Coher- control, and operant variability. Behav. Anal. 25, 175 – ence dynamics in photosynthesis: protein protection of 190. excitonic coherence. Science 316, 1462–1465. (doi:10. 39 Krechevsky, I. 1937 Brain mechanisms and variability 1126/science.1142188) II. Variability where no learning is involved. J. Comp. 22 Mercer, I. P., El-Taha, Y. C., Kajumba, N., Marangos, Physiol. Psychol. 23, 139 –160. J. P., Tisch, J. W. G., Gabrielsen, M., Cogdell, R. J., 40 Wolf, R. & Heisenberg, M. 1991 Basic organization of Springate, E. & Turcu, E. 2009 Instantaneous mapping operant behavior as revealed in Drosophila flight orien- of coherently coupled electronic transitions and energy tation. J. Comp. Physiol. A 169, 699 –705. (doi:10. transfers in a photosynthetic complex using angle- 1007/BF00194898) Proc. R. Soc. B (2011)