All for one... One for all

SCHUMACHER COLLEGE
An International Centre for Ecological Studies

The following article is published with kind permission from New Scientist magazine. It appeared on 13th June, 1998 (pp 32-35). For more about New Scientist,
see: http://www.newscientist.com

All for one… One for all

Ants behave with a collective intelligence that can’t be explained by their genes alone, says Brian Goodwin. And he’s built a virtual colony to show the team in action.

WATCHING termites construct a highrise, air-conditioned apartment block out of the red Kenyan soil, or a colony of ants in the rainforest sewing leaves together to build a miniature metropolis, makes you wonder where the intelligence for such intricate engineering comes from. Is the colony’s behaviour just the sum of the activities of its individual members, or are we dealing here with a property unique to the colony as a whole?

The first view is favoured by reductionists – those biologists who try to reduce complex behaviour to the activities of simpler elements such as individual organisms or, preferably, their genes. The second has given rise to the concept of the colony as a superorganism, a coherent whole with properties that are distinct from those of its constituents and which cannot be predicted from them. This notion, though still not widely accepted, introduces a new level of order to biology which in turn is helping to improve our understanding of evolution.

The idea that there exists unpredictable “emergent” order in natural processes did not originate in biology. It is, for example, well understood in physics. Take the properties of the elements hydrogen and oxygen. These are thoroughly understood in themselves. Yet nobody has ever succeeded in predicting the properties of water from a knowledge of its constituent elements. Trying to explain why water spirals down a plughole simply from a knowledge of atoms is an impossible task. What’s missing is a knowledge of fluids-a whole level of order above that of individual atoms.

Physicists see water for what it is – a fluid. They develop theories about its “high-level” properties that are consistent with the “low-level” properties of atoms by working back and forth between the levels. There is nothing unscientific about this approach. It simply acknowledges that certain phenomena cannot be predicted. Turning back to biology and social insects, we can ask what is the appropriate level to explain the cooperative behaviour of a colony as a whole. Probably the most popular approach to biological phenomena these days is to try to explain them in terms of the properties of genes. According to evolutionary thinking, if a genetic mutation changes an individual’s structure or behaviour in a way that gives it a competitive edge over others, then it will be favoured by natural selection and passed on to future generations.

Selfish bent

This is an explanation in terms of fitness – the capacity of an individual to survive and reproduce. In recent years, biologists such as Bill Hamilton and Richard Dawkins, both now at Oxford University, have extended this argument to the genes themselves: it is genes that reproduce and are selected, so they have become the ultimate units of natural selection. Evolution is now widely understood in terms of selfish genes that are bent on leaving as many copies of themselves behind as possible.

For ant colonies, this view has been further modified. If all the individuals in a colony are closely related to one another – as they are in many species of ant because they all have the same mother, the queen – then a genetic change that gives an advantage to an individual will tend to be advantageous to the whole colony. This is measured in terms of what is called “inclusive fitness”. The cooperative activity seen in ant colonies is assumed to be just such a characteristic that confers benefit on the group as a whole. And while we might interpret cooperative behaviour as altruistic, Hamilton and Dawkins argue that what we’re really seeing is the result of selfish genes looking out for themselves via the colony.

Though fine in principle, this line of reasoning contains a logical gap. Genes may affect the behaviour of individuals, but just how do they affect the behaviour of the colony as a whole? After all, it is not genes that interact in a colony but individual organisms. Just as hydrogen and oxygen atoms are necessary but not sufficient to explain water spiralling down a plughole, so genes – and even individual ants – alone cannot explain cooperative behaviour. There is a level of order missing. We need to proceed more like physicists and ask the question: what particular form of interaction between ants can generate a particular type of collective behaviour? So let’s get specific.

Order from chaos

At the University of Bath, Nigel Franks and his colleagues noticed rhythmic activity patterns among workers tending the queen and the young in the brood chambers of a species of Leptothorax, small ants that form colonies of between 40 and 80 members. The team of workers is active for a time and then becomes inactive, taking about half an hour to go through one cycle. How does this collective rhythm arise? Does every ant behave in a cyclic fashion so that the colony rhythm is simply an expression of synchrony among intrinsically rhythmic individuals?

The person who answered this question is Blaine Cole, a researcher at the University of Houston in Texas who works with another species within the same genus of ant as that used at Bath. He filmed isolated ants and groups of various sizes, and then analysed the activity patterns of individual ants. Cole concluded that isolated individuals and individuals in sparsely populated groups have a pattern of activity-inactivity that is described as deterministic chaos. This is not a random pattern, but one which is so complex that it is impossible for an observer to predict what a particular ant will do next.

Cole also found that when the density of ants reached a certain level, the group displayed a collective activity-inactivity rhythm with about the same cycle time as that found by Franks – around half an hour. How can collective order emerge from chaotic individuals? Clearly, this must result from the way they interact. But what kind of interaction?

Ants perceive their world primarily through their antennae. They have, on the whole, very poor sight, whereas their antennae have exquisitely sensitive touch and smell receptors. When ants meet, they communicate via their antennae. Active ants may encounter either other active ants, or inactive ants. When the latter happens, the inactive ant is stimulated into action.

Is this type of stimulation enough to produce a collective rhythm in a colony of chaotic individuals? It is not obvious that this is the case, so I and my colleagues Ricard Solé from the Polytechnic University of Catalonia, Barcelona, and Octavio Miramontes, then at Britain’s Open University, designed a computer model of an ant colony to see what we could learn. We described ants as cellular automata, simple software agents that move about on a grid like a chess board (see Diagram below). The pattern of activity-inactivity of our “ants” was driven by a neural network that produced a chaotic output, so that in isolation they behaved like the real ants observed by Cole. But the ants could also interact with one another: whenever an active ant reached a position on the grid next to an inactive ant, the lazy ant was stimulated out of its lethargy.

Every ‘ant’ on the computerised grid has an activity-inactivity cycle that is chaotic. And when an active ant (orange) occupies a position next to an inactive ant (grey), the lethargic insect becomes mobile too. With a sparse population of ants, activity across the grid as a whole tends to be chaotic (see graphs below). But as the density increases, a rhythmic pattern emerges. The colony begins to behave like a single, pulsating superorganism.

Now, as Cole had done with real ants, we began to increase the number of virtual ants on the grid. As the density increased, the amount of activity per individual increased, simply because of the higher frequency of stimulation between individuals. Then, at a particular density, a distinct rhythm began to emerge over the “colony” as a whole. With further increases in density, the rhythm became distinct and well-defined (see Graphs below).

Graphs (above) showing ant activity. The X axis is time. The Y axis on the right shows the rate of activity. The figures up the right-hand side show the number of ants (1 in the top row, rising to 80 on the bottom row).

Unpredictable rhythms

So model ants, behaving chaotically and interacting “socially” by stimulation, can generate a collective rhythm throughout a colony. This is a clear case of emergent behaviour: it was impossible to predict the collective, rhythmic pattern just from a knowledge of the chaotic behaviour of the individuals. Just as in physics, we worked backwards from an unexplained behaviour of a system to its possible origins within the pattern of interactions between the system’s components.

Using our model, we could explore the impact on the colony of different patterns of interaction. What would happen if active ants could stimulate other active ants, for example? We found that the most important type of interaction for generating the rhythmic patterns was stimulation of inactive ants by active ants. Cole found this to be true for real ants too. We could now begin to be more precise about the way in which genes may be influencing observed behaviour through the properties of individual ants.

In the model there is a parameter that defines how sensitive ants are to being stimulated by their companions. We found that if this sensitivity is too small, activation never spreads across the colony to produce collective rhythms, no matter how high the density of ants. On the other hand, if the sensitivity is too high, then above a certain density the colony becomes continuously active and again no collective rhythm emerges.

This suggests that genes affecting an individual’s response to stimulation need to be regulated. Is this range large or small? If small, then there may be a significant price to be paid for accurately regulating the relevant genes to keep the sensitivity within the required bounds. However, the model showed that there is in fact a wide range over which the sensitivity can vary and still produce collective rhythms. This suggests that rhythmic activity patterns are robust consequences of colonial living. They may, in fact, be hard to avoid. Collective rhythm is an emergent property that has been described as “order for free”.

Unexpected altruism

This result may also help to explain another aspect of colonial life. For many species, including bees, wasps and termites, the members of a colony may not be closely related but still cooperate. This seems to run counter to the spirit of the selfish gene. The theory’s proponents solve this puzzle by arguing that the cost of recognising alien genes in other colony members is too high and would lead to losses in efficiency. Order for free offers a more direct and simple solution.

There is one other important property of colonial rhythms that the model revealed. Rhythmic activity emerges suddenly above a critical density. But what type of discontinuity are we dealing with? The model showed that it has the characteristics of what physicists call a phase transition: a sudden change of state from one type of order to another. One example is the sudden appearance of magnetic properties in iron as it is cools below a critical temperature. Above this temperature the kinetic energy of the atoms, each of which behaves as a magnetic dipole, is too high for them to become aligned, so they move independently of one another. Below the critical temperature, the kinetic energy falls low enough to let the dipoles line up, north pole to south pole along the metal, so that a collective magnetic field emerges.

The ants are doing something similar, but now it is density, and hence the amount of stimulation between them, that is the critical parameter. We see from the model that at the critical density, what was a collection of individuals doing their own local thing begins to change as global order emerges via activation waves that propagate through the entire colony. It’s time to go back to real colonies and ask if they actively regulate their densities. And if so, where on the density spectrum do they lie: close to the lower transition to disorder – at the edge of chaos – or well into the ordered regime? Again, Franks and his colleagues have made important observations here.

They have shown that ants do regulate the mean density of workers in the brood chamber, though it is not clear how they do it. The preferred density of the colony appears to be near the edge of chaos. They have also suggested what the adaptive value of rhythmic activity may be to the colony. Within the brood chamber, the workers look after the queen and the young, feeding and cleaning them and organising the brood so that the youngest are closest to the queen. Franks contends that if workers had chaotic patterns of activity-inactivity in the brood chamber, there would be a haphazard distribution of care over the brood as a whole. But with a rhythmic pattern, the workers are all active within the brood chamber at the same time. They then distribute care more uniformly because workers tend to seek untended brood to look after.

So, the suggestion goes that the rhythmic behaviour improves a colony’s method of nurturing its young and increases its fitness, giving it a competitive edge. This is a plausible suggestion that fits well with mainstream evolutionary thinking. Of course, we could have reached the same conclusion just through discussion of genes and fitness, with no discussion of collective rhythm. But fitness explanations explain only why a particular pattern of behaviour persists, not how it is generated and how difficult or easy it is to produce. This is where the science of complexity and the study of emergent phenomena come into play. They bridge a serious gap in evolutionary thinking. In the process, they are shifting the emphasis from genes and fitness to emergent order as a primary source of evolutionary novelty.

Brian Goodwin is a course tutor on the MSc in Holistic Science at Schumacher College. He is author of How The Leopard Changed Its Spots (Weidenfeld and Nicolson, 1994), and coauthor with Gerry Webster of Form and Transformation: Generative and Relational Principles in Biology (Cambridge, 1997), and co-author with Richard Solé of Signs of Life: How Complexity pervades Biology (Basic Books, 2002).

Copyright New Scientist, RBI Limited 1998

For more information about Schumacher College and its courses, please contact :

The Administrator, Schumacher College,
The Old Postern, Dartington, Totnes, Devon, TQ9 6EA.
Tel : +44 (0)1803 865934 Fax : +44 (0)1803 866899
Email : admin@schumachercollege.org.uk
Website : www.schumachercollege.org.uk

Schumacher College is part of the Dartington Hall Trust, a company limited by guarantee, registered in England and as a charity (company no. 1485560, charity no. 279756). Registered office: The Elmhirst Centre, Dartington Hall, Totnes, Devon TQ9 6EL, UK.