Animals are usually considered to behave as complex automata, responding predictably to external stimuli. This study suggests otherwise, showing that even the humble fruit fly can behave spontaneously. The flight paths of flies in a completely featureless environment were neither random nor predictable, but followed a complicated fractal pattern generated within the fly's brain.
Free will and true spontaneity could exist … at least in fruit flies. This is how we provokatively started the press-release for our study in the open-access journal PLoS One.
According to Voltaire, “Liberty then is only and can be only the power to do what one will.” This power to willfully behave differently in identical settings has recently come under fire from neuroscientists (see articles in New York Times, PLoS Biology, Scientific American, Financial Times, The Economist [copy], Süddeutsche, Wired, Mises, Royal Society), so nobody was expecting anything from animals. Animals, especially insects are usually seen as very complex robots which only respond to external stimuli. They are said to be input-output devices. When scientists observe animals responding differently even to the same external stimuli, they attribute this variability to random errors in a complex brain. Using a combination of automated behavior recording and mathematical analyses, we could show that such variability cannot be due to simple random events but is generated spontaneously and non-randomly by the brain. If even flies show the capacity for spontaneity, can we really assume it is missing in humans?
Essentially, the study served to distinguish between the following two alternative models:
Alternative models conceptualizing the fruit fly experiments.
A - According to the robot-hypothesis, there is an unambiguous mapping of sensory input to behavioral output. If the behavioral output is not constant in a constant environment, there are a number of possible sources of noise, which would be responsible for the varying output.
B - In a competing hypothesis, non-constant output is generated intrinsically by an initiator of behavioral activity. Note that the sources of noise have been omitted in B merely because their contribution is judged to be small, compared to that of the initiator, not because they are thought to be non-existent.
Our results caught my collaborator, computer scientist and lead author Alexander Maye from the University of Hamburg by surprise: “I would have never guessed that simple flies who keep bouncing off the same window otherwise have the capacity for nonrandom spontaneity if given the chance.” Previous studies have shown that in nature, flies do not buzz about aimlessly but forage according to a sophisticated search strategy (this is how they find our wine glasses). Our research now suggests that such strategies arise spontaneously rather than being induced by spatial cues.
Tethered fruit fly flying stationarily.
A small drop of glue between head and thorax is used to fix Drosophila flies to small copper hooks. When the flies are tethered by these hooks, they can still beat their wings and perform a variety of other behaviors.
We tethered fruit flies (Drosophila melanogaster) in completely uniform white surroundings and recorded their turning behavior. In this setup, the flies do not receive any visual cues from the environment and since they are fixed in space, their turning attempts have no effect.
The experimental setup And because this is much too cluttered, here a schematic 3D rendering of the most important components: The fly is suspended at the torque meter.
While the fly is restriced in moving its body axis (i.e. it is fyling stationarily), it can still move its wings and legs, extend its proboscis and so forth. The forces the fly generates as it tries to turn around its vertical body axis (i.e. yaw torque) are recorded. The fly's environment is homogeneously white: no other visual stimuli reach the fly other than white, stationary background illumination. In other words, the fly receives a constant, homogenous, featurelss input. In such an environment, the robot-hypothesis predicts constant or monotonous behavioral output.
Thus lacking any input, if the flies were input-output devices, their behavior should resemble random noise, similar to a radio tuned between stations. However, the analysis showed that the temporal structure of fly behavior is very different from random noise. We then tested a whole battery of increasingly complex stochastic computer models all of which failed to adequately model fly behavior.
Raw yaw torque data.
As one could see from the video, flies without visual input rarely stay still. Instead, they constantly change their yaw torque (flight direction). In particular, the trace shows two variable components: a slow, meandering component upon which small, fast fluctuations, so-called torque spikes are superimposed. Yaw-torque is plotted on the Y-Axis and time on the X-Axis. A - 30min total trace. B - 5min enlarged trace from the red section.
Torque spikes correspond to sharp turns in free flight, so-called body-saccades, which can be obsered in freely flying flies: they don't fly smooth curves like airplanes, but zig-zag around, making them hard to catch. Torque spikes have been classified as "fixed-action-patterns" or single units of behavior and are thus well-suited to study the temporal structure of their initiation. ISI: inter-spike-interval.
Only after we analyzed the fly behavior with methods developed by the other co-authors George Sugihara and Chih-hao Hsieh from the Scripps Institution of Oceanography at UC San Diego did we realize the origin of the fly's peculiar spontaneity. Using the so-called "S-Map Procedure" we detected a non-linear signature in the fly behavior. Such a signature can only be found in systems whose indeterminate behavior is not due to noise but originates in their design. This signature indicates that there is a function in the fly brain which evolved to generate spontaneous variations in the behavior. This function appears to be common to many other animals and could form the biological foundation for what we experience as free will.
The temporal signature in the fly behavior points to a so-called 'unstable nonlinearity' in the fly brain. This in turn means that the brain areas controlling turning behavior must be tuned very precisely to generate unpredictable output and are unlikely to be a by-product of the general complexity of the brain. Unstable nonlinear systems are known from many other natural systems and display a high sensitivity to small perturbations. These sensitive systems provide an evolutionary advantage to animals that possess them not only because they help animals forage, but also for a number of other reasons. For instance, they can lead to unpredictable escape maneuvers when avoiding a predator or to unpredictable moves which confer advantages in almost any competitive social setting (think politics or chess). The biological implementation of this mechanism is currently unknown, but there is evidence from a previous study that a brain area called the ellipsoid body (sometimes called the "fly motor cortex") might be involved.
I particularly like what George had to say about our results: "This nonlinear signature eliminates the two alternative explanations of spontaneous turning behavior in flies that would run counter to free will, namely complete randomness and pure determinism. These represent opposite and extreme endpoints in discussions of brain functioning which mirror the free will debate." To that, I'd only add that our subjective notion of 'Free Will' is essentially an oxymoron: we would not consider it 'will' if it were completely random and we would not consider it 'free' if it were entirely determined. Nobody would attribute any responsibility to our action if it had happened entirely coincidental. On the other hand, if our action was completely determined by external factors such that there was no alternative, again the person would not be held responsible. So if there is anything remotely close to free will, it must exist somewhere between chance and necessity - which is exactly where fly behavior comes to lie. George again finds the right words: "Our results address the middle ground between simple determinism and randomness that is currently not well understood or characterized. We speculate that if free will exists, it is in this middle ground." This leads me to believe that the question of whether or not we have free will appears to be posed the wrong way. Instead, if we ask 'where between chance and necessity are we located?' one finds that this is precisely where humans and animals differ. Humans may not have free will in the philosophical sense, but even flies have a number of behavioral options they need to decide between. Humans are less determined than flies and possess even more options. With this small reformulation, the topic of free will becomes the new biological research area of studying spontaneous behavior and can thus be discerned from the philosophical question.
Of course, I realize that there is no clear and unambiguous definition of what 'Free Will' actually means (see Wikipedia) and that at least historically people have debated of whether it existed as something immaterial and independent of our brains (so-called dualism). I don't think anybody really takes a dualistic position nowadays anymore, so I define free will as the capacity to chose from different options in the same situation with some degree of anticipation of the consequences of each choice. It is perfectly reasonable to speculate about this form of free will in fruit flies, but regardless of our speculations on free will, the most important scientific aspect of our work is the evidence we found for a brain function which appears evolutionarily designed to always spontaneously vary ongoing behavior. There is tentative evidence that such a function may be very widespread in the animal kingdom, including humans. Why would all brains have this function? If this indeed turned out to be the case, we may actually have discovered the first evidence for something truly fundamental to the understanding of brains.
Research on this topic will likely go on for quite a few more years. We have only made the first step. We have shown that even a fly brain possesses a function which makes it easier to imagine a brain that creates the impression of free will. It remains to be shown how this faculty is implemented in the brain and if it plays part in creating the subjective experience in humans of being able to do what one will. The next step in this direction will be to localize and understand the brain circuits responsible for the spontaneous behavior in flies. The results from this step could lead directly to the development of robots with the capacity for spontaneous nonrandom behavior. Eventually this research may help treating disorders leading to compromised spontaneous behavioral variability in humans such as depression, schizophrenia or obsessive compulsive disorder.
Oiginal study:
Maye, A.; Hsieh, C.; Sugihara, G. and Brembs, B. (2007): Order in spontaneous behavior. PLoS One, May 16.
DOI number: 10.1371/journal.pone.0000443
Manuscript abstract:
Brains are usually described as input/output systems: they transform sensory input into motor output. However, the motor output of brains (behavior) is notoriously variable, even under identical sensory conditions. The question of whether this behavioral variability merely reflects residual deviations due to extrinsic random noise in such otherwise deterministic systems or an intrinsic, adaptive indeterminacy trait is central for the basic understanding of brain function. Instead of random noise, we find a fractal order (resembling Lévy flights) in the temporal structure of spontaneous flight maneuvers in tethered Drosophila fruit flies. Lévy-like probabilistic behavior patterns are evolutionarily conserved, suggesting a general neural mechanism underlying spontaneous behavior. Drosophila can produce these patterns endogenously, without any external cues. Such behavior is controlled by brain circuits which operate as a nonlinear system with unstable dynamics far from equilibrium. These findings suggest that both general models of brain function and autonomous agents ought to include biologically relevant nonlinear, endogenous behavior-initiating mechanisms if they strive to realistically simulate biological brains or out-compete other agents.
