«Whether signals are reliable or deceptive has been a central question in the study of animal communication in recent years. The crux of the issue is ...»
William A. Searcy & Stephen Nowicki:
The Evolution of Animal Communication
is published by Princeton University Press and copyrighted, © 2005, by Princeton
University Press. All rights reserved. No part of this book may be reproduced in any form
by any electronic or mechanical means (including photocopying, recording, or information
storage and retrieval) without permission in writing from the publisher, except for reading and browsing via the World Wide Web. Users are not permitted to mount this file on any network servers.
Follow links Class Use and other Permissions. For more information, send email to:
email@example.com 1 Introduction Whether signals are reliable or deceptive has been a central question in the study of animal communication in recent years. The crux of the issue is whether animal signals are honest, in the sense of conveying reliable informa tion from signaler to receiver, or deceitful, in the sense of conveying unreliable information, the falsity of which somehow beneﬁts the signaler. This issue arises in a variety of contexts. When a male courts a female, do his signals honestly convey his quality relative to other males? Or does he exaggerate his quality in order to win over females that would otherwise choose some other male? When one animal signals aggressively in a contest over a resource, does the signaler honestly convey its likelihood of attack? Or does the signaler exaggerate that likelihood in order to intimidate competitors that would other wise defeat him? The question of reliability versus deceit arises even in interac tions that, on the face of things, seem to be predominantly cooperative. When an offspring begs for food from its parents, does it honestly convey its level of need? Or does the offspring exaggerate its need in order to get more food than the parents would otherwise provide?
The issue of reliability and deceit in animal communication resonates with human observers for a variety of reasons. One is that the occurrence of deceit is fraught with moral implications. In the view of many, human communication is permeated with deceit. Do humans stand apart in this regard, or are other animals as bad or worse? The answer might have considerable effect on how we view ourselves, as well as on how we view other animals. A second reason for interest in this issue is that the occurrence of deceit, if deceit is deﬁned appropriately, can have considerable implications for our understanding of ani mal cognition. Some deﬁnitions of deceit are framed so as to require cognitive processes of considerable sophistication, such as the ability to form intentions and beliefs and to attribute beliefs to other individuals. If we employ such a deﬁnition, and if we can then determine that nonhuman animals deceive each other according to this deﬁnition (a big “if”), then we have provided support
what is best for the population or species as a whole, leading to the view that animal communication consists primarily of the cooperative exchange of reliable information. The predominant view nowadays, however, is that selec tion acts largely at the level of the individual, so that behavior evolves toward what is best for the individual performing the behavior, and not toward what is best for the group. If behavior is commonly selﬁsh, in this sense, then it is not always obvious why animals should exchange information cooperatively.
Instead, one might expect many instances in which signalers would attempt to proﬁt individually by conveying dishonest information. But because individual selection works on the receiver as well as the signaler, receivers ought to re spond to signals only if doing so is to their advantage, on average. Therefore, if dishonesty is common, it also is not obvious why receivers should respond to signals.
Taking the argument one step further, if receivers fail to respond to signals, it is not obvious how signaling systems can exist at all. Thus if one accepts the view that selection acts predominantly at the level of the individual, as we do, and if one at the same time accepts the idea that animals do communicate with each other, as seems obvious, then one is left with a series of evolutionary puzzles. Are animal signals in reality reliable or unreliable? If animal signals are reliable, what mechanisms maintain reliability despite the tempting advan tages of dishonesty? If animal signals are deceitful, do receivers respond to them anyway, and, if so, why? Our principal purpose in this book is to work through possible answers to evolutionary puzzles such as these.
Deﬁnitions Before we get to these puzzles, we need to deﬁne some terms. First, we need to deﬁne what we mean by “signal,” in order to delimit the set of traits whose honesty and dishonesty we will examine. In one of the ﬁrst rigorous evolution ary analyses of communication, Otte (1974, p. 385) deﬁned “signals” as “be havioral, physiological, or morphological characteristics fashioned or main tained by natural selection because they convey information to other organisms.” Otte explicitly rejected group-selectionist explanations for the evolution of traits, so in his view the transmission of information had to confer some reasonable advantage on the signaler itself in order to satisfy the deﬁni tion. Thus Otte excluded as signals those traits that convey information to predators or parasites without any beneﬁt to their possessors; he cited the chemicals in human sweat that attract disease-carrying mosquitoes as a possi ble example. Otte also rejected as signals those traits, such as body size, that may be used by other individuals of the species to assess their possessors but did not evolve for that function. Clearly included under Otte’s deﬁnition would be vocalizations, color patterns, and body movements that have evolved be
I N T RO D U C T I O N 3â cause they transmit information in a way that beneﬁts the individual that exhib its those traits. More ambiguous are traits, such as the form of a bird’s tail, that originally evolved for some other function but have been modiﬁed by selection for information transmittal. We will regard such traits, or more pre cisely their modiﬁed properties, as signals; thus the bird’s tail itself is not a signal but the tail’s length is, if that length has been exaggerated beyond its aerodynamic optimum in order to inﬂuence receivers.
This brings us to our deﬁnitions of reliability and deceit. In everyday English, “reliable” means that “in which reliance or conﬁdence may be put; trustworthy, safe, sure” (Little et al. 1964). An animal signal, then, would be reliable if one could have conﬁdence in its veracity, or truthfulness—if, that is, one could trust the signal to convey whatever it is supposed to convey. The difﬁculty with this formulation is in ascertaining what the signal is “supposed to” convey. “Sup posed to” in this context must be interpreted from the viewpoint of the receiver rather than the signaler; what matters is whether the signal conveys something that the receiver would beneﬁt from knowing. If we are certain what it is that the receiver beneﬁts from knowing, such as some attribute of the signaler or its environment, then we can ascertain the reliability of the signal by measuring the correlation between the signal and the attribute of interest.
Suppose, for example, that we think that female frogs are interested in the size of conspeciﬁc males, and we ﬁnd that calls communicate information on male size by a negative correlation between call frequency and caller size (males with deeper croaks are larger). We can then determine the reliability of this information by measuring the correlation between call frequency and caller size. The trouble is that we can never really be certain that caller size is what the females “want” or “need” to know. Even if we can show that call frequency is well correlated with caller size, and that the females show a behav ioral preference for calls of lower frequency, we cannot be sure that their true interests are not in some other characteristic—perhaps, in this example, male age. The best we can do is to measure as carefully as we can the beneﬁts that the receivers obtain from different types of information. If we can show that female frogs beneﬁt from mating with larger males but not from mating with older ones, we at least can have some conﬁdence that size is what matters to the receivers, and then evaluate reliability of call frequency in terms of its correlation with signaler size.
To formalize this deﬁnition, we suggest that an animal signal is reliable if:
1. Some characteristic of the signal (including, perhaps, its presence/ absence) is consistently correlated with some attribute of the signaler or its environment; and
2. Receivers beneﬁt from having information about this attribute.
A remaining problem is how to specify what we mean by “consistently corre lated.” We can never expect a perfect correlation between signal characteristic
4 CHAPTER 1â and the attribute being signaled. Even if signalers are striving for perfect hon esty, errors must be expected in the production of the signal and in our mea surements of it, either of which would prevent our observing perfect reliability.
How good, then, does the correlation have to be for us to conclude that the signal is on the whole reliable? One answer is provided by the concept of “honest on average” (Johnstone and Grafen 1993, Kokko 1997). A signal can be considered honest on average if it contains enough information, sufﬁciently often, that the receiver on average is better off assessing the signal than ignor ing it. Consider again the example of male frogs communicating their size to females via the frequency of their call. The correlation between male size and call frequency can never be expected to be perfect, and in reality is often rather low (see chapter 4). The male’s call can be considered honest on average if the correlation between male size and call frequency is good enough that the female beneﬁts on average from using the call to assess male size, instead of ignoring this signal feature. In practice, it will be difﬁcult to determine whether this criterion is being met, but at least it provides a theoretical standard against which reliability can be judged.
A simple way to deﬁne “deceptive” would be as the opposite of reliable, but for many the concept of deception carries more baggage, and consequently requires a more complex deﬁnition. A relatively simple deﬁnition of deception is provided by Mitchell (1986, p. 20), who suggested that deception occurs
1. A receiver registers something Y from a signaler;
2. The receiver responds in a way that is appropriate if Y means X; and
3. It is not true here that X is the case.
Note that the deﬁnition requires specifying what the signal (Y) means to the receiver. The meaning of Y to the receiver is judged by the response of the receiver to Y together with an observed correlation between Y and X, across many such signals. In other words, we infer that Y means X to the receiver because signalers usually produce Y in association with X, and because the receiver responds to Y in a way that is appropriate if X is true. To make this more concrete, let Y be an alarm call given by the signaler. The alarm call is usually produced when a predator (X) is present, and the receiver typically responds to the alarm call by ﬂeeing, an appropriate (i.e., beneﬁcial) response if a predator is indeed nearby. Deception occurs if the signaler produces the alarm and the receiver reacts by ﬂeeing when in fact no predator is present.
A difﬁculty with Mitchell’s (1986) deﬁnition, which he himself points out, is that deception so deﬁned cannot be distinguished from error on the part of the signaler. If the signaler has produced an alarm in error, would we want to call such an action deceptive? This problem can be solved if the deﬁnition of deception further stipulates that the signaler beneﬁts from the receiver ’s re sponse to the signal. Mitchell (1986) himself is uncomfortable with the notion
I N T RO D U C T I O N 5â of beneﬁt, remarking that the “idea of beneﬁt is taken from human affairs” and when applied to nonhuman animals typically refers to what a human observer “believes is good for them.” For an evolutionary biologist, however, “beneﬁt” has a straightforward meaning—an individual beneﬁts from an action if that action increases the individual’s ﬁtness, in the sense of the representation of the individual’s genes in subsequent generations. Beneﬁt in this sense is not an anthropocentric idea, but one that applies equally well to all organisms.
With the added stipulation about a beneﬁt to the signaler, we will deﬁne decep
tion as occurring when:
1. A receiver registers something Y from a signaler;
2. The receiver responds in a way that a. beneﬁts the signaler and b. is appropriate if Y means X; and
3. It is not true here that X is the case.
Deception deﬁned in this way has sometimes been termed “functional decep tion” (Hauser 1996), meaning that the behavior has the effects of deception without necessarily having the cognitive underpinnings that we would require of deception in humans.