The study of animal behavior, Nikolaas Tinbergen remarked, prompts four questions. Given that an animal is doing something: how does it help the animal survive; how does it change; what is its history; and what is its cause? Questions about the causes of behavior are notoriously difficult. And it has been surprisingly difficult to define the idea of behavior itself. Tinbergen suggested that behavior is, “the total movements made by the intact animal.”1 This might suggest that a turtle, in being shifted by an ocean wave, is undertaking behavior of a sort.
The definition of behavior has not become clearer since Tinbergen wrote. Daniel Levitis attempted to provide a definition by asking one hundred and seventy four researchers to pick out examples from a list that included a spider building a web, a sponge pumping water, and a person sweating in response to heat.2 “Behavior,” he concluded, “is the internally coordinated responses (actions or inactions) of whole living organisms (individuals or groups) to internal and or external stimuli, excluding responses more easily understood as developmental changes.”3
What is good enough for Levitis is good enough for us.
Research into animal behavior cannot be organized as easily as genetic research. In the case of the genome, Tinbergen’s questions admit of easy answers. DNA stays largely the same from generation to generation, and from the beginning until the end of an individual’s life. Changes in DNA are traceable from species to species. The survival value of DNA is obvious, and the central dogma of molecular biology demonstrates how DNA works.4 Before 1953, the definition of a species was as fuzzy as the definition of behavior, but when it comes to behavior, it is still 1952. The best we can do is to compare similar species. To avoid confusing themselves, ethologists have focused on stereotypical behaviors that are robust to environment changes or are part of fixed behavioral patterns.5
Animals closely related to one another phylogenetically sometimes show completely distinct forms of behavior. A good example is the Nudibranch mollusk, commonly referred to as the sea slug. Nudibranchs are a group of aquatic shell-less mollusks commonly found in the ocean. Many species within the Nudibranch group do not swim. Among those that do, some swim by flexing their bodies in the dorsal-ventral direction, others by bending their bodies in the left-right direction. Tritonia diomedea is a notable dorsal-ventral swimmer, while the most extensively studied left-right swimmer is Melibe leonina.6 While Tritonia and Melibe are monophyletic, they swim in starkly different ways. The difference can be traced to the neural circuits controlling swimming. In Tritonia, neurons from both left and right sides fire synchronously during swimming, while in Melibe, neurons fire in alternation.7 What is missing from these studies is some acknowledgement of the flexibility of fixed behavior. Melibe possesses dorsal-ventral neuronal homologues, although these neurons are mostly silent when Melibe is swimming.8 It is unclear whether Melibe is capable of both dorsal-ventral and left-right swimming, but for unknown reasons, simply prefers one to the other.
Several experiments could shed light on this question. One might selectively activate both the left and right dorsal-ventral neurons in Melibe and see whether dorsal-ventral swimming occurs. On the other hand, one might culture individual Melibe with groups of Tritonia when Melibe hatches, and see whether dorsal-ventral swimming occurs as a result of developmental effects or environmental constraints. These experiments are challenging to perform in Nudibranchs, which cannot be cultured and bred in the laboratory. We do not yet have molecular markers to selectively target dorsal-ventral neurons in living animals.
Unlike the sea slug, the fruit fly, Drosophila, offers an amenable test system for elucidating the mechanisms underlying behavior.9 In our laboratory, we have found that female flies from different subspecies display different levels of aggression when feeding. Drosophila erecta gathers around food sources much faster than Drosophila melanogaster, and they show greater levels of aggression.10 Such behavior is food-specific, and can be easily observed during feeding.11 When reared with D. melanogaster, D. erecta becomes less aggressive than when raised alone or with conspecifics. We measured gene expression levels in flies reared in socialized and nonsocialized contexts, and found that genes related to immune response and olfaction were significantly up-regulated in the socialized flies.
In another study from our laboratory, we trained individual flies to regulate their level of activity. When punished, the flies proved able to modify their behavior. Receiving the same punishment, untrained controls did not lower their activity as much. Learning and memory takes place in the flies’ brains, but more so for the trained than the untrained flies. Exactly what is learned may not be easy to pinpoint.
In one experiment, flies were trained to associate a specific odor with aversive electric shocks.12 When presented with two odors, they showed a marked eagerness to avoid being shocked. Only 80–90% of the flies, however, displayed this preference.13 Why did some flies behave differently? In these examples, individual animals solved problems in their own way; their solutions varied. It is the need to come up with a solution that is universal. This is familiar enough in computer science, where a problem can usually be solved with more than one algorithm.
The ability of an organism to generate solutions to its problems is in the nature of behavior; the algorithmic nature of behavior does not impose an identical requirement on the organism’s hardware. The nervous system is not the prerequisite for behavior. Organisms with no nervous system display a surprising repertoire. The nerveless Trichoplax and most bacteria are capable of motion.14 The invention of the nervous system enabled organisms to achieve a more sophisticated variety of behavior, but there remained room in animal life for much simpler solutions to the problems of behavior This appears to be particularly true for locomotion.
All metazoans are multicellular and motile. But not all motile organisms are animals. The choanoflagellates also move around a lot. So do some bacterial and archaeal species.15 Organisms move for a variety of reasons. Movement itself is the basis for more advanced behaviors, such as chemotaxis and phototaxis. Nature presents every organism with the same deep problem: how to move? Consider the bacterial cell, suspended in liquid and attracted to certain stimuli. The flagellum, a structure consisting of a basal body, filament, and hook, and driven by a proton motive force, represents a common bacterial solution to the problem of motion.16 Counterclockwise rotation forces the filaments in the flagellum to coalesce into a bundle along the long axis of the cell body; as a result, the bacterial cell moves forward. Clockwise motion, on the other hand, makes the filaments less organized, causing the cell to tumble. Rotational direction is usually controlled by a system monitoring chemical cues from the environment.
The flagellum also seems to be the archaeal solution to the problem of motion. Archaeal and bacterial flagella are structurally different.17 Many of the proteins that make up flagella in archaea resemble the proteins that make up the bacterial type IV pili structure.18 Type IV pili are used by bacteria for pulling themselves forward when anchored to a surface.19 But in archaea, type IV pili-like flagella function by rotating their filaments. Despite the differences in structure, bacteria and archaea move in functionally similar ways.
Physics places certain limits on the way in which single cells can move through a homogenous environment. A bacterial cell can swim at 20–25μm/s, micrometers per second. This corresponds to 20 bodies per second, or bps. Some archaeal cells swim at anywhere between 3–380μm/s, or 3–380bps.20 Escherichia coli doubles every 40 minutes or less.21 At 20μm/s, E. coli can move up to 48mm during its lifetime. It would be a challenge for E. coli to move to food sources 100 meters away.
The speed limits of flagellum-based movement are more evident in eukaryotic organisms, such as protozoa and Trichoplax. Eukaryotic flagella are structurally unlike bacterial flagella.22 The eukaryotic flagella may have evolved from pre-existing cytoskeletal components such as tubulin and dynein.23 Paramecium propels itself at 50–200μm/s (1 bps) by moving cilia attached to the outside of its cell membrane.24 Trichoplax adhaerens also uses cilia to move at around 15μm/s, or 0.015bps—much slower than some unicellular organisms.25 Trichoplax does not have any internal organs, tissues, nerves, or muscles; and its cilia beat asynchronously. Unicellular species usually propel themselves by synchronous ciliary movement.
Although efficient in moving prokaryotes, the flagellum reaches a bottleneck in moving multicellular organisms. Flagella are controlled by chemical signals in the cell. No intracellular signaling system can be synchronized among cells without an inherent time delay.
Novel locomotor algorithms are needed for efficient locomotion in multicellular organisms.
Muscles and the nerves that control them allowed animals to have better and finer control of their movements. Derived from the mesoderm, muscles generate force when their actin and myosin filaments slide and contract. All bilaterians have striated muscles, suggesting that the muscles emerged before the bilaterians. Jellyfish also use striated muscles to propel themselves. It is likely that the striated muscles found in bilaterians evolved from ancient cells that were able to contract, but were unable to regulate their contractions.26 The protein coding gene MYH1 may have existed before the emergence of such muscle-controlling cells as the bilaterian troponin complex. Movement algorithms are modular and follow a step-by-step process. Properties at the microscopic level appear sequentially and generate emergent properties, such as neutrally-controlled movements.
In life, as in technology, new machinery requires new controls. In unicellular organisms, movement is mainly controlled by intracellular events; in multicellular organisms, by the nervous system.27 Neurons can roughly be divided into sensory, interneuron, and motor neurons. At neuromuscular junctions (NMJ), motor neurons elicit muscle contraction by transmitting signals to muscles via neurotransmitters—usually acetylcholine in vertebrates and glutamate in invertebrates.28 With NMJs, animals can execute their movements with more precision and accuracy. The impulse for locomotion is of two sorts: something from the outside gets the organism to get going, or it is something from the inside. For the latter, the movement must be generated somewhere inside the nervous system. The result is typically rhythmic, such as breathing and walking. The neurons participating in this type of behavior are referred to as central pattern generators.29 Their identity is different for different forms of behavior. At this level, movement remains largely modular and algorithmic. If any step in the locomotor sequence is blocked, either by chemicals or by physical lesions, movement becomes impossible. This is just what one expects in a sequence of steps involving complicated chemical and physical property changes.
Socialization is another almost universal problem. Although bacteria are unicellular, they often appear in multicellular aggregates known as biofilms.30 Individual cells acquire substantial benefits by being part of a community.31 One cell is unlikely to have any significant effect. The transition from unicellular to community life is regulated by a process called quorum sensing.32 Individual cells produce autoinducing molecules, whose concentration is related to nearby cell density.33 Once their concentration reaches a minimum, cells in this community undergo changes in gene expression. Gram-negative and Gram-positive bacteria behave in comparable ways, but their autoinducers are different. Gram-negative bacteria use acylated homoserine lactones, and Gram-positive bacteria, processed oligopeptides.34
Courtship is a special form of social behavior. To reproduce sexually, individuals have to select their partners. In animals with complex behaviors, sexual selection is usually conducted as courtship. During courtship, male fruit flies display a complex yet stereotyped sequence of actions.35 The male turns towards the female, taps her with his front legs, extends one of his wings and sings a courtship song, licks her genitalia, and then undertakes copulation.36 Female flies are choosy, rejecting males with defective wings or other inadequacies.37
Eusocial species exhibit cooperation in brood care and maintain a clear division of labor. The result is a caste system in which some members are sterile.38 Evolution towards eusociality, some researchers have maintained, is unidirectional and apparently irreversible.39 But the six-banded furrow bee is both eusocial and solitary. In one study, fifteen different species of the genus Halictus were found to have a common social ancestor; at least two were not eusocial.40 If the queen dies in a eusocial system, a worker may become the new queen. We do not know how this individual is chosen. It does not seem likely that her promotion is based on a specific sequence of actions by all individuals in the group. There would seem to be rules in place, but nothing like a deterministic and inflexible algorithm at work.
Bacteria, insects, fish, birds, and mammals exhibit swarming behavior.41 Murmuration in starlings is the best-known example. The phenomenon has puzzled researchers for a long time. How do so many animals manage to move without colliding? And how are they able to quickly change formation? Mathematical models provide one answer; swarming robots another.42 Robots may be made to swarm by being given specific swarming instructions: go there, do this, turn left, turn right. This method is computationally intensive and inflexible. But it is also possible to get a thousand small, identical robots to swarm by relying on three rules: edge-following, gradient formation, and localization.43 Individual robots do not possess global knowledge of the swarm, nor do they have accurate information about their global positions. All they know is the distance to their neighbors. User-defined shapes can be reproduced this way. The solutions reached during this formation are not based on step-by-step instructions. The behavior of large groups of animals supports the idea that swarming is best described by general rules. In physics, even if the initial conditions of a complex system are known, it is often very difficult to calculate the full phase space of the system. The problem is even more challenging in biology. Yet with a minimal set of rules, behavior such as swarming can be reliably reproduced in the laboratory.44
