Greg Detre
Thursday, May 03, 2001
Prof. Steve Simpson, Natural History Museum
When ethologists use terms like �communication�, �information�, �noise� and �deceit�, among others, they mean these in a technical sense. Such terms are useful, because they give an immediate intuitive kick start to our understanding of them, and yet they can later prove obstructive �false friends�, usually through their anthropomorphising connotations. Wilson (1975) broadly defines �communication� well for us: �Communication occurs when one individual�s actions provide a signal that changes the behaviour of another individual�. In this way, we can talk about the �information� transmitted in the signal in terms of the extent and delicacy with which one animal (the �signaller�) can influence the behaviour of another (the �receiver�), or in terms of reduced uncertainty. We can talk about �noise� in terms of attenuation, extraneous environmental distractions and any deficiencies in the signaller�s encoding or receiver�s decoding equipment that hamper or diminish the impact of the signal (Wiley 1983).
Given all this, we can distinguish different types of signals, with different purposes and adapted to different situations and environments. In the case of the rufous-sided towhees (Richards 1981), an initial burst of loud sound of unvarying pitch precedes the more complicated warbling of the song which follows. This �alerting signal� notifies any animals that can hear the initial sound to listen out for the sounds that follow, which become degraded and muffled in the woodland (noise), and might otherwise go unnoticed. Alerting signals are among the four features that ethologists have identified animals using to increase the reliability of detection, including: redundancy, conspicuousness, possession of small signal repertoires and use of alerting components. The analogies between information transfer in animal communication and in digital signal processing are quite explicit.
The terms �honesty� and �deceit� seem ideally suited to describing the behaviour and communication mechanisms employed by many animals to increase their own fitness in some way, often at the expense of the receiver or a nearby partial observer. As we will see, such �honest� or �deceitful� communications are usually explicable in evolutionary terms, without reference to morality or premeditation.
In order to consider how honesty and deceit arise in animal communications, it will help to first look at a few examples of honest animal communications, and to consider how these have evolved.
Dawkins & Krebs (1978) put forward the case for �manipulation�, where every communication balances the selfish attempt of the signaller to increase its fitness, against the selfish attempt of the receiver to increase its fitness. Thus, signallers are trying to influence the behaviour of the receiver, and receivers are trying to gain information that might help them. Hamilton (1964) presented this as a 2x2 matrix, ranging from mutuality (where both signaller and receiver benefit), deceit, eavesdropping and �spite� (where both suffer from the communication).
The means that signallers employ in this �deceit� are varied, and take advantage of the receiver�s decoding mechanism in any way they can. The most human-like deceits might involve simple Batesian mimicry, such as a bigger or more splendid badge, plumage or display. These conventional signals can take many forms, usually based at some point in the evolutionary past on an assessable quality of import to the receiver. However, as soon as �cheats� (deceitful signallers) realise that they can enjoy all the benefits accruing to a great fighter just by wearing the badge, a sort of badge inflation occurs, with bigger and bigger badges being necessary to display prowess. What happens at this point depends on a number of factors:
the costs to the signaller of producing the signal itself (e.g. handicapping signals, which might increase visibility to predators)
the costs to the signaller of being probed and exposed (e.g. losing badly in a fight);
the costs to the receiver of probing the signaller (e.g. time, energy);
the costs to the receiver of being deceived (e.g. copulating with an unsuitable mate).
These factors will determine the frequency of cheats. �Probing� here indicates that the receiver has some means of assessing the signaller in a more rigorous way than just decoding the signal. If the costs to the receiver of probing are high, the genes to cheat will proliferate through the population. However, this will lead to receivers being less trusting, and the conventional system being undermined or devalued, until the receivers start to probe more and more often. If the costs to the receiver of being deceived are high, for example in choosing a mate who is a weak fighter or a poor resource-holder, then the receiver cannot afford to be deceived. Finally, Rohwer�s (1977) experiment with black patches on Harris� sparrows exemplifies how the costs to the signaller of being probed and exposed may be high. Dying a pale bird�s breast black to make it look more like a dominant male increases its success in competition with other subordinate birds, yet young males have not evolved to bluff by evolving larger black patches. The reason for this seems to be that the dyed birds were persecuted by the real dominant males, and presumably bluffers would pay the same price. In a similar way, young male elephant seals (Le Boeuf 1974) who have evolved to look like females in order to sneak past the bull elephant into his harem suffer costs from probing by the female receivers, who emit loud screams during copulation, ensuring that only the most dominant male nearby is able to complete copulation.
Guilford & Dawkins summarise the range of signalling systems which could all be described as �honest� to a greater or lesser degree. These range from the most honest, where a given signal provides a perfect and reliable (i.e. unbluffable) predictor for a quality that the receiver particularly wants to know, i.e. provides important information. However, in reality, such a perfect signal rarely exists, or has a prohibitive cost � the very best way to find out which of two males will definitely win in a fight, for instance, is for them to have a costly, full-scale fight. The case of the red deer stags (Clutton-Brock & Albon 1979) provides one such example, where their roar seems to be an accurate, unbluffable indicator of fighting prowess. However, often there are reliable signals which are correlated (again, to a greater or lesser degree) with the quality that the receiver is trying to assess, and these are used instead. These signals may have costs for both the signaller and the receiver. Guilford & Dawkins stress that most discussion of deceit had not sufficiently considered the effect of the cost to the receiver, usually in time or effort.
This extra pressure increases the propensity to evolve towards conventional systems. These are potentially the least �honest�, given that the signals are lower in cost and arbitrary, allowing Batesian mimics to cheat extremely easily. However, there must be inherent safeguards in conventional systems, otherwise they would become valueless over time as receivers learnt not to trust the signals. Another type of unbluffable signal is one where the costs of bluffing further simply exceed the benefits gained from the bluff. However, the original emphasis placed by ethologists on handicapping signals (Zahavi 1975) has since been undermined (Maynard Smith 1976).
Blumberg & Alberts (1992) strike a cautionary note by distinguishing functions and effects when discussing adaptiveness in ethology. They give the example of humans babies with respiratory distress syndrome, where the baby emits little grunts with every breath. Paediatricians have learnt to use this symptom to identify the condition, and so the grunt is clearly a communication signal (by Wilson�s broad definition), though it is only the acoustic product of a homeostatic mechanism. However, when ethologists have observed similar laryngeal braking in the Norway rats (Blumberg & Alberts 1990), and the mother�s acute auditory sensitivity in this range, they assume that the ultrasonic squeaks produced are some sort of �distress call�. They have confused an exaptation, the pups� involuntary ultrasonic emissions, with an explicit signal whose function is to alert the mother, simply because this time it is the receiver who has driven the evolutionary changes.
Axelrod & Hamilton (1981) discuss the evolution of honesty and deceit in terms of the age-old game theory study, the Prisoners� Dilemma. Briefly, two partners in crime are being interviewed by the police, separately. They each have the choice of cooperating by refusing to name the other, or defecting, i.e. blaming the other. This results in a well-known 2x2 matrix, where the rewards are deliberately weighted such that dual cooperation results in the best situation for them both. If one cooperates but the other defects, the defector goes completely free and the cooperator loses out worst of all. If both defect, then the rewards for both are somewhere in the middle. Various strategies have been revealed, depending on the particular parameters of the game. The results for one game are very different from the results for many games. When playing once, the best option is to defect. However, when the two have to play an indeterminate number of games, a tit-for-tat strategy emerges, ultimately settling in cooperation. This mirrors ethological evidence, where seemingly evolutionarily unsound altruistic strategies frequently emerge amongst social groups, where memories of past wrong-doings are long, and the entire genetic pool benefits from cooperation. In a similar way, honesty proves the best policy in communications, though there are significant difficulties in drawing analogies with the much more complicated situations that emerge in real animal behaviours.
Wiley (1983) outlines these three scenarios of deceit in animal communication as:
escalation by counteracting inflation and devaluation of signals;
mimicry limited by disadvantages of incidental consequences for the signaller;
mimicry limited by disadvantages of increased discrimination by the receiver.
In the first case, signals may inflate, perhaps becoming unbluffable. Receivers may devalue, or become more discriminating (through more subtle observation or more frequent full-scale probing). In the second case, honesty is again enforced by costs directly to the signaller, unless some other form of signalling more open to bluffing emerges. In the last case, the level of dishonesty stabilises according to all the factors listed above, restrained in signallers by the occasional probing by receivers, and restrained in receivers by the costs to them of probing.