Who’s Running the Show?

            There’s a reason that ant colonies are a favorite example of complex systems theorists. Complex systems can be roughly defined as a network of heterogenous parts that interact in relatively simple ways with each other on the local level, but as a result of this mass activity more variable and self-regulating global behaviors emerge spontaneously. The functional units respond only to local information, but each individual action transmits further information throughout the system so that the whole can optimize resource and task allocation. Ants embody this notion, as simple individuals occupy distinct roles in an overall division of labor that allows the whole colony to behave cohesively and intelligently. The key to this in ant colonies is dynamic task allocation. Most models and empirical research indicate that the overall behavioral output of an individual ant depends on an interaction between fixed physiological and morphological attributes of the ant (which depends on caste-based ontogenetic differences) and transient environmental cues (often in the form of explicit chemical signals from fellow ants).

Factors Intrinsic to the Individual

            ant-Size: In many species there is pronounced size polymorphism between castes. Different tasks may be better suited by a certain overall size. This also supports the Spatial Fidelity Zone model (SFZ) of social insect organization. The SFZ model states that there is a correlation between ant size (or, for the politically correct among you, “ant corpulence”) and an ant’s pace. The differentially paced ants will have different positions in space relative to the colony center, and these different positions in space will dictate which task the ant will perform. The idea is that intrinsic morphological differences, such as size, cause a natural sort of sifting that allows for differentiation.
            -Worker age polyethism: This is basically the notion that an ant’s age will dictate which task it performs. As individuals mature, they take up new tasks. The mechanism that correlates a specific age to a given task is unknown.
            -Genetics: Oh genes, where don’t you come into play? It has been shown in colonies with multiple germ lines that genetically distinct individuals tend to take up differing tasks, and even vary in how their tasks change with age.


Structural or Stigmergic Factors

            -Chemical gradients: Another important component of the SFZ model is the proposed layout of chemical gradients throughout the colony. These gradients supposedly radiate out from the center, where the queen is, and so at any given moment that ant can tell where it is in the colony via pheromone detection. These gradients give important information about which task an individual should be performing in a certain region of space. It is likely that the gradients do not simply radiate concentrically from the center, but from multiple sites of high activity, such as the colony entrance. Multiple pheromones, communicating different information, may exist in spatially distributed gradients, often overlapping to create a combinatorial vernacular for potential task allocation.

            But wait! These elegant explanations may be necessary (and they most likely are to differing degrees) but they are by no means sufficient for the intricate social organization that ants exhibit. A distinction must be made at this point. There are an estimated 22,000 ant species and these diverse “superorganisms” represent a massive spectrum in terms of colony size, colony function and internal dynamics. Colony sizes range from tens of highly specialized ants to teeming metropolises with millions of less distinct individuals. Roughly speaking, it can be said that in small colonies individual roles are more immutable and individual ants have a higher degree of learning and larger cognitive capacity and are generally more efficient. Massive colonies have such extreme redundancy that individuals can afford to be quite inefficient and boorish and more emphasis is placed on flexibility in social roles, so that the masses can be directed in a regulated fashion. It is the net activity and interactions between individuals that matters in the latter. In smaller colonies with more permanent castes such intrinsic and structural factors probably have a higher influence on social organization. In larger colonies, these play a role, but there are many other mechanisms at work as well.

For the colony to operate efficiently, the number of individuals carrying out a task has to adjust to the colony’s overall needs. There is evidence that exactly how an individual decides which task to take up is largely based off of environmental cues.

The Changing Environment of the Colony

            As the colony interacts with other components of the ecosystem, the overall needs can vary dramatically. Ants have been shown to efficiently deal with these incoming demands so that a relatively invariant setpoint is maintained, much like homeostasis. If a particularly abundant new source of food is found, then there is an increase in foraging ants. Conversely, if a flood washes a bunch of debris into the nest, than more maintenance ants are required. The ants become “aware” of these global needs through local cues from other ants, mainly through chemosensory/olfactory or tactile communication.


Cuticular Hydrocarbons

When two ants encounter one another, they run their respective antennas over each other. They are actively picking up different signals that may be emitted by the other ant. One of the most informative signals is to be found in a dollop of mixed hydrocarbons located on the ants’ cuticle. These cuticular hydrocarbons (CHC) provide unique chemical label that contains information of the ant’s identity, such as its caste and where it’s been recently. Ants can recognize whether another is a foreigner or a nestmate based off of differences in the CHC composition. This exhibits a high degree of sensitivity and discrimination on the part of the ant olfaction system. But it is even more subtle yet! Ants can recognize what task a nestmate has been performing. So, an individual can glean important social information that contributes to the ongoing neural computations that constantly must decide which behavior to engage in. This last bit is absolutely integral to optimization within the colony.

Interestingly, the caste differences in CHC label composition often arise because of purely environmental reasons. For example, the CHC of foragers is subject to hotter, drier conditions than those castes that stay primarily in the colony, so there are a different ratio of n-alkanes to n-alkenes (extremely simple and common components of the CHC label) due to all the bond breaking that results from frequent exposure. In terms of another ant’s sensory detection, this results in a phenomenally distinct label that the receiving ant processes in such a way as to bias them towards adopting foraging behavior. Greene and Gordon performed a simple experiment that is highly illustrative of the role of this recognition process in task allocation. In red harvester ants (Pogonomyrmex barbatus), patrollers leave the nest in the morning to survey the foraging area. If the patrollers return, then the foragers are signaled to leave and find food. If the patrollers do not return, the foragers remain near the nest entrance without leaving. The researchers captured the patrollers as they were out and dropped plastic beads coated with appropriate quantities of the relevant CHC label into the nest entrance, in a patroller-mimic assay. They found that this did incite foragers to emerge from the nest, while the no-bead control condition did not.
However, an important corollary finding of theirs was that the presence of the CHC label alone was not sufficient to signal foraging. It had to coincide with the time of morning that this ritual usually occurred, occur some time after the patrollers had left the nest, and be localized right in the nest entrance where the foragers were expecting the patrollers to return. This implies that there is an interaction between the incoming CHC olfactory information and the ant’s internal clock mechanism, memory formation involving the whereabouts of other ants informing expectations, spatial cognition, and probably more. This demonstrates that each ant is constantly taking in information and integrating it to make a decision on when to act.


             Cuticular hydrocarbons contain a lot of situational information that may influence the subsequent behavior of a fellow ant. But CHCs are just the tip of the iceberg when it comes to chemical communication in ants. A complex array of messages are conveyed via the ant exocrine system, through the medium ofpheromones. Across species, there are the metapleural glands, the Dufour glands, the mandibular glands, the venom glands, the hindgut, the postpharyngeal glands, the pyrigidal glnd, Pavan’s glands, the rectal glands and the Tibial glands. Each of these glands produces a wealth of substances, depending on the species. This allows for a species- and colony-specific chemical vernacular, meaning that communication lines are somewhat protected and a given chemical cue can have an arbitrary degree of specificity. This has important implications for communication and social organization. A given gland may produce several pheromones at one time and several glands may be used in conjunction, allowing for combinatorial complexity in potential messages. This is a notion that has only recently been acknowledged, so the depth and complexity of social communication between ants (and indeed, any social insect species) is poorly understood.
            There isn’t a strict and fundamental difference between cuticular hydrocarbons and the exocrine system. Much of the exocrine output has a large proportion of hydrocarbons, and indeed the “hydrocarbons” in CHCs are complicated mixtures with much else besides. The cuticle itself is a gland. The key difference lies in the time scales and cause of the secretions. Cuticular hydrocarbons represent static presentations of an ant’s identity and behavioral state relative to the colony as a whole, while most other exocrine secretions are in response to particular events with the effect of inciting other individuals to a more or less specific set of actions. The olfactory processing of CHCs is a statistical summation of encounters with other individuals over time, a way to tally the activity of other ants, and therefore roughly assess the global state and needs of the colony. Pheromone processing communicates explicit instructions. Both systems are necessary for the mass coordination of individuals in a meaningful way, on different time scales, and for the constant adjustment of activity levels.
            Of course, the hypothesis that pheromones are the locus of task allocation begs the question of whether pheromone-processing leading to behavioral output is strictly deterministic. This is overly simplistic and mislead. First of all, different amounts of a given substance detected by the individual denote differing signal strengths, which have important consequences. In multicomponent signals, the proportion of different substances matters as much as the substances themselves. It has been observed that two mixtures with different ratios of the exact same chemicals result in a different overall behavioral output. Finally, many signals are modulatory in nature, meaning that they alter an ant’s response threshold to other signals but don’t elicit any response in and of itself.


             The recruitment hypothesis asserts that there is a degree of plasticity in the social role of any ant, such that most individuals can be recruited to a new task. The mechanism of this recruitment is hypothesized to be the various signal streams available to an individual at any point in time. Recruitment abounds throughout the social structure of colony, though it is not a totally arbitrary network in which all points are exactly equal. That being said, the most transparent subsystem through which to explore recruitment is that of foraging. Since foraging provides the energy input to the entire colony, it is afforded a high degree of salience. There are several different kinds of recruitment in foraging.

            -Tandem running: a patroller returns with food, regurgitates a sample for other foragers, then positions itself so that its venom gland is exposed and produces a drop of pheromone that we’ll call “foraging pheromone” (the literal chemical identity doesn’t matter so much, since this changes across species). A nearby ant will extend its antennae to the region, in an exchange of tactile and olfactory communication. The patroller ant will then return to the food source, with the second ant trailing behind.
           swarm -Mass foraging: This is the official forging moniker for the infamous ant trails that form around picnics. The dynamics of mass communication are quite interesting, and hint at the feedback loops regulating it all. When a food source is discovered, the foragers return with food, laying down an attractant pheromone on its way back. Initially, there is an exponential build-up of ants at the food source, since the attractant is indiscriminate. Since there is a functional limit to how many ants are necessary given the surface area of the food source, ants that cannot reach the food return to the nest. This simple feedback mechanism intrinsic to the task allows the number of ants allocated to the task to settle on an appropriate number. This number continually fluctuates as the attractant is laid down every time an ant returns with food. Most ant pheromones are highly volatile and so the signal decays rather rapidly. This allows signal strength to encode food quantity, because if there is more food, then there is more surface area, and therefore more ants laying down a trail, which cause the attractant to build-up faster and decay slower (in terms of net vaporization - it still has the exact same volatility).
            Trunk trails: These are more fixed trails between a reliable and consistent food source and the nest. These presumably consist of less volatile secretions, and are likely to be added to every time an ant travels it.

Ant Chemosensation

            The vaporization mentioned above, while overall dissipating the signal, is integral to the ants’ ability to use the pheromone for navigation. As it vaporizes, it creates a sort of “vapor tunnel” that the ant is constantly picking up on by waving its antennae around in the air, in a directed fashion. Its movements are guided by something called osmotrpotaxis, which is essentially the detection of concentration levels. To begin to explain this, the anatomy of ant olfaction must be introduced.
            Ants have Olfactory Receptor Neurons (ORN) on their antennae that bind to molecules in the environment. These ORNs project to various glomeruli, which are clusters of specialized neurons in the Antennal Lobe (AL). Ants have around 460 distinct glomeruli, arranged in seven superclusters, further divided into two hemispheres. Each of these seven regions has it’s own antenna-cerebral tract (ACT) that projects to a “higher region.” Apart from the ORN projections, the glomeruli contain extensive enervation by excitatory and inhibitory interneurons, suggesting ample room for a combinatorial space of possible representations. For a while, the reigning theory was that the glomeruli were specialized enough to be localized processing centers for different kinds of stimuli. This was supported by the fact that ORNs of similar structure tended to project to the same glomeruli, implying that different molecular classes were cleanly dealt with by different glomeruli. However, the fact that any environmental odor will be a composite of molecules already marks this theory as too simplistic. In addition, calcium imaging studies show that firing patterns in response to simple molecular mixtures are distributed across glomeruli. This makes sense in light of the nature of chemical signaling as well. Many different signals can be composed of the same chemical constituents, so the functional perceptual unit must be at a more distributed level. This is exemplified by studies on nestmate recognition. The CHC, among other things, conveys a generalized “colony odor” so that ants can differentiate nestmates from foreigners. This recognition is important because it biases the ant to respond to other chemical cues if it is a nestmate, and incites aggression if it is a foreigner. Obviously, all of colony functionality rests on this discrimination. Different colonies and species often use the same or very similar components to make up the colony odor, but in different ratios. Therefore, a distributed representation is necessary to discriminate.
            Top-down back projections are likely important for signal interpretation, as evidenced by modulatory signals, as well as the findings in the CHC label study by Gordon and Greene. Other information coming in from previous experiences as well as other senses are important to construct a context by which to interpret a chemical signal. This other “template” information likely comes in the form of activity coming from other regions in the ants brain, modulating the interconnections among the many glomeruli and changing how future olfactory information is processed.

Whole is More than the Sum of its Parts

            The functioning of the colony as a whole depends on the social divisions between subpopulations within the colony. The overall behavioral output of an individual, and therefore what tasks it is engaged in, depends on constant synthesizing of different kinds of cues that orient and dispose it towards certain patterns of behavior. These cues are overlapping systems of overt communication from other ants, whether it is direct and instructional pheromone emission or static cuticular hydrocarbon information stores, as well as incidental information about the colony’s environment. Ants are fantastic information processing organisms, but in the end incapable of much on its own. However, when such complex systems are linked in ways that give rise to so much opportunity for feedback, then fine-tuned regulation is possible. This is how an ant colony can achieve such incredible degrees of organization.