SEARCH

SEARCH BY CITATION

Keywords:

  • cheating;
  • cue;
  • evolution of communication;
  • handicaps;
  • honesty;
  • manipulation;
  • signal costs;
  • signalling

Abstract

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

Although communication underpins many biological processes, its function and basic definition remain contentious. In particular, researchers have debated whether information should be an integral part of a definition of communication and how it remains reliable. So far the handicap principle, assuming signal costs to stabilize reliable communication, has been the predominant paradigm in the study of animal communication. The role of by-product information produced by mechanisms other than the communicative interaction has been neglected in the debate on signal reliability. We argue that by-product information is common and that it provides the starting point for ritualization as the process of the evolution of communication. Second, by-product information remains unchanged during ritualization and enforces reliable communication by restricting the options for manipulation and cheating. Third, this perspective changes the focus of research on communication from studying signal costs to studying the costs of cheating. It can thus explain the reliability of signalling in many communication systems that do not rely on handicaps. We emphasize that communication can often be informative but that the evolution of communication does not cause the evolution of information because by-product information often predates and stimulates the evolution of communication. Communication is thus a consequence but not a cause of reliability. Communication is the interplay of inadvertent, informative traits and evolved traits that increase the stimulation and perception of perceivers. Viewing communication as a complex of inadvertent and derived traits facilitates understanding of the selective pressures shaping communication and those shaping information and its reliability. This viewpoint further contributes to resolving the current controversy on the role of information in communication.


Introduction on communication

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

Communication is a key feature of life. The dominant challenge of the study of communication over the last three decades has been to understand the mechanisms that enforce reliability, when it often appears that signallers can gain at least short-term benefits by cheating (Zahavi, 1975; Tibbetts & Dale, 2004; Searcy & Nowicki, 2005; Getty, 2006; Számado, 2011). The handicap principle emerged as the prevailing paradigm explaining the evolutionary stability of honest signalling. It predicts that signals can be reliable and informative because they represent handicaps that are costly to produce or to maintain (Zahavi, 1975; Grafen, 1990; Tibbetts & Dale, 2004). Signal costs allow perceivers to assess individual quality, because they demonstrate the ability to signal and survive in spite of the handicap. Epitomized by the peacock's tail, the handicap principle became a core concept of behavioural ecology and sexual selection theory, but this concept has recently been criticized (Cotton et al., 2004; Getty, 2006; Számado, 2011), and it is unclear how important different alternative explanations to reliable communication might be.

The current controversy on handicap signalling is tied into a more fundamental debate on the conceptualization of communication. Whereas many researchers describe communication as the evolved process of transferring information from senders to receivers (Otte, 1974; Bradbury & Vehrencamp, 1998; Searcy & Nowicki, 2005; Carazo & Font, 2010; Seyfarth et al., 2010), others regard the information-based approach as misguided and conceptually untenable and instead view communication as a process of influencing rather than informing others (Dawkins & Krebs, 1978; Rendall et al., 2009; Owren et al., 2010; Scott-Phillips, 2010). According to the information-based approach, natural selection will favour reliable information because perceivers are selected not to respond to unreliable signallers. According to the influence-based approach, the concept of honest signallers seems at odds with the well-established principle of selfish individuals that aim to gain influence on others (Rendall et al., 2009; Owren et al., 2010; Scott-Phillips, 2010). Both concepts thus differ in their relative emphasis on signallers and perceivers. However, because communication is an interaction between signallers and perceivers, it is unlikely that unilateral advantages gained by one party will persist through time (Krebs & Dawkins, 1984; Owren et al., 2010).

Currently, it is difficult to evaluate the relative merits of the alternative concepts of communication. These difficulties arise because the influence-based approach is logically consistent with the possibility that information is associated with communication. Indeed, Owren et al. (2010) wrote (p. 772) that ‘pressure from perceivers can and likely often does select for courtship displays whose features are correlated with signaler fitness’. Thus, information is not a criterion to distinguish whether a given communication system evolved to transfer information or to influence others.

Here, we have four aims. First, we want to illustrate that information is often a by-product of mechanisms unrelated to communication. Although the concept of by-product information is not entirely novel, we feel that its importance is underappreciated because selection by perceivers is the most commonly invoked force to explain reliable information in behavioural ecology and evolutionary biology. Accepting that by-product information is common leads, second, to the conclusion that constraints operate on signallers to suppress or reveal information. Constraints have recently been invoked for perceivers only (sensory biases; Owren et al., 2010), and here, we emphasize that they operate on signallers as well. Incorporating constraints on communication is important because the most influential models for the conceptualization of the communication process assume no constraints on signallers and perceivers. Third, we propose that reliable communication does not necessarily invoke costs of signalling (as proposed by the handicap principle). Although many researchers agree that the handicap principle is only one mechanism capable of producing reliable information, it is the overarching framework that guides most empirical research on the reliability of communication even though costs of signalling have often not been demonstrated in empirical studies. Fourth, by-product information provides a general proximate mechanism explaining why communication is associated with information even if signallers are only selected to influence (and not inform) others. In this context, we clarify that information does not typically evolve during the evolution of communication but that reliability can evolve if displays involve behavioural components.

We structure the article by first revisiting the debate on the alternative concepts of communication before defining the terms relevant to understand communication. In section three, we turn to how by-product information shapes the evolution of communication.

Alternative concepts of communication

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

Descriptions of the information-based approach rely implicitly or explicitly on the conduit metaphor (Shannon & Weaver, 1949) which considers a sender to have information, to encode a copy of that information into a signal and to transmit that signal to perceivers, which decode the signal and recover the copy of the information (summarized in Rendall et al., 2009). Major criticisms of this approach are that information is assumed but rarely defined and that the concept of equating communication with information transfer precludes investigations into whether animal signals actually transmit any information (Owren et al., 2010). Further, animals (unlike humans) rarely, if ever, use signals in a representational fashion, and as far as is known, most animals do not possess a theory of mind to infer which information a sender most likely encoded into the signal (Rendall et al., 2009). Consistent with these views, we feel that ‘information sharing’ is not a good description of most communication systems. However, this does not imply that we see no value in the concept of ‘information’ in communication.

We completely agree that signallers influence the behaviour of perceivers. Even if communication is essentially about manipulation by signallers, perceivers will always be selected to gather and use any information valuable to them (Carazo & Font, 2010). Hence, the influence-based approach is challenged to explain why perceivers react to signals if these are not consistently and reliably associated with information. One possible solution is that sensory exploitation of perceivers is common in many communication systems (Owren et al., 2010), but this is currently unknown. Moreover, perceivers are not passive in the evolution of communication, and even exploitive communicative systems may become informative (Garcia & Ramirez, 2005). If exploitation of perceivers is either not common or unlikely to persist through time, the most natural way to explain a perceiver's reaction to detection of a signal is that perceiving the signal changes the perceiver's understanding of some aspect of the world around it. That is, perception of a signal changes the informational state of the perceiver (e.g. Seyfarth et al., 2010). Thus, we feel that information is central to understanding the communication process even though most communication systems cannot be described as a process that evolved for information sharing.

Communication is an interaction, where signallers try to influence perceivers, and perceivers try to obtain information from signallers (Dawkins & Krebs, 1978). This interaction can be coevolutionary, that is, an evolutionary change in player one can occur in response to an evolutionary change of player two and stimulate another evolutionary response in player two. As stated previously, the two concepts of communication emphasize different aspects in the interaction among signallers and perceivers. In general, the balance between signallers and perceivers will vary between communication systems and is also likely to vary over time in coevolutionary communication systems.

We contend that signallers are selected to influence the behaviour of perceivers, and that in many communication systems, perceivers are less important in selecting for reliable communication than some descriptions of the information-based approach have assumed. Nevertheless, we feel that informational terms should be retained in definitions of communication because communication can often be associated with information, even if that information has not been selected for in the evolution of communication. To clarify our view, we turn to definitions in the next section.

Definitions

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

The debate on communication is troubled by the fact that communication and information are sometimes not clearly and not consistently defined. Here, we adhere to the definition that communication is the interaction between a signaller and a perceiver if some trait values of the signaller evolved to stimulate the sensory system of the perceiver in such a way as to cause a change in the behaviour of the perceiver (compared to a situation where the trait values of the signaller were different; Ruxton & Schaefer, 2011; Schaefer & Ruxton, 2011). Communication thus defined is adaptive for signallers; it can be adaptive for perceivers but is not necessarily so. Consequently, our definition of communication is inclusive comprising coevolutionary communication systems and also communication systems like camouflage. Communication is based upon signals, which are usually defined as traits or behaviours that evolved to influence the behaviour of other organisms (either through manipulation or information, Otte, 1974; Maynard-Smith & Harper, 1995; Carazo & Font, 2010). In contrast to signals, cues are informative stimuli that influence others but did not evolve for that effect (Maynard-Smith & Harper, 1995). Our definition of communication excludes situations in which perceivers react to cues.

The current debate on the relationship between communication and information is not helped if the terms we use and measure have little to do with how animals select communication or information. Because communication is an interaction, it is important to note that communicative traits have more phenotypic and genotypic dimensions than ordinary traits consisting of a single phenotype and a single genotype. For example, there can be a correlation between a quality of a signaller and its phenotype, which has the potential to be informative to others. Unless this information will be perceived by others, it is evolutionary irrelevant. Any correlation between the quality of a signaller and its phenotype thus depends on the phenotype of perceivers (i.e. their sensory system) and its corresponding genotype. Information depends on yet another phenotype of perceivers, their cognitive system. Importantly, it is the interaction between the sensory and cognitive systems of perceivers that drives the evolution of the genotypes of signallers (to produce or hide signals) and perceivers (to better perceive and evaluate information).

Because the reaction of perceivers shapes the evolution of communication, we define information as an averaged correlation in the perceived trait values and a quality that the perceiver is interested in. Information thus arises from a sensory interaction between two parties, but it is necessarily a property of the perceiver. The reason that a sensory interaction between signallers and perceivers induces change in the behaviour of the perceiver is that different sets of sensory stimulation are associated with variation in some quality that is interesting to (i.e. with fitness consequences for) the perceiver. Although information is based upon a correlation between the quality of signaller and its phenotype and only arises from an interaction between signallers and perceivers, it is – from an evolutionary point of view – not inherent to the signal. That is, in contrast to the conduit metaphor, we do not see the signal as carrying a specific package of information. Rather the signal has the potential to impart informational change to perceivers and potentially different informational changes to different perceivers. For example, two dove species consuming saguaro fruits from the giant cactus (Carnegiea gigantea) obtain different nutritional rewards from those fruits (Wolf et al., 2002). The information both doves associate with seeing fruits thus differs according to the nutritional rewards they are interested in. Because of this difference between perceivers, the information gained through perception of a signal is best thought of as a property of the specific perceiver. Recasting information as a property of the perceiver allows us to relate it directly to the selection by the perceiver (and hence the qualities of interest to it) and its contribution to shaping communicative trait values.

We define by-product information as arising from a perceived correlation between the qualities a perceiver is interested in and the sensory stimuli it perceives if this correlation has not been selected for a communicative function by the perceiver. By definition, by-product information is associated with cues. By-product information is also associated with indices, a specific form of signals. An index is a signal whose correlation with quality does not depend on past coevolution between signallers and perceivers but on physical necessity (Maynard-Smith & Harper, 1995). Cues and indices thus characterize different types of informative stimuli that both rely on by-product information.

Signals are usually complex entities consisting of distinct components (Grether et al., 2004). These authors show, for example, that a colour patch used for signalling consists of distinct cell layers, pigments and structural features which all contribute to colour perception and that the components may at least partly change independently from each other. Additionally, it has become evident that most perceivers base their decisions on multiple, and often correlated, signalling traits (Candolin, 2003; Bro-Jørgensen, 2010). Here, we thus consider that it is often useful to think of signals as compound traits consisting of multiple components and where at least one component evolved to influence the behaviour of other organisms.

The multiple components that displays consist of differ in their evolutionary history (Steiger et al., 2011b). The difference between cues and indices is thus that only indices include a component that evolved to influence others, whereas a cue lacks a component that evolved for that function. Cues and indices can thus be separated by the selective pressures that act upon part of the display. Furthermore, differentiating between terms used for an entire display and the informational part associated to it is essential to advance the current controversy on communication: specifically whether communication is a process that evolved to transfer information or not. For example, differentiating between by-product information and an index is important because by-product information can be associated with cues, with indices or can form the starting point for the evolution of signalling, whereas an index characterizes a certain type of display consisting of evolved and inadvertent traits.

Ritualization of by-product information

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

We argue that signals often involve components that are informational, but whose trait values are primarily influenced by noncommunicative selection pressures. In other words, signals contain reliable information, but this information often has not been primarily selected for a communicative purpose. Importantly, the availability of this information can provide a very widely applicable, but we feel neglected, explanation for the maintenance of reliable communication on which we will focus here.

The concept of by-product information is well known. Information is often inadvertently produced by individuals and is often called public information (Valone & Templeton, 2002; Danchien et al., 2004; Dall et al., 2005). By-product information is a cue, and it has played a major role in explaining animals' decisions in foraging, habitat selection, mate choice and predator avoidance (Valone, 1989; Danchien et al., 2004; Valone, 2007; Danchien et al., 2008; Earley, 2010). For example, the resource quality of a food patch can be evaluated by observing the behaviour of others attempting to exploit that food resource, but crucially their behaviour is not a signal, it is a cue. The behaviour of the feeding animals is a cue because it produces by-product information about food for nearby animals but did not evolve to inform these bystanders and possible competitors.

The evolution of communication is usually seen as a process of ritualization where the communicative aspects of precursors (which are cues) are gradually emphasized so that they are more effective in eliciting a response from the perceivers (Tinbergen, 1952; Maynard Smith & Harper, 2003; Steiger et al., 2011b; Scott-Phillips et al., 2012). Two aspects are important about ritualization. First, traits that are already informative owing to genetic, physiological or ecological mechanisms are particularly prone to ritualization (Bradbury & Vehrencamp, 2000) if this information is perceived by others. Second, ritualization renders signalling displays more conspicuous (e.g. by adding components that are effective in long-distance detection or in alerting others), but it does not generally and necessarily alter the information associated with displays (Steiger et al., 2011b). Ritualization is thus a process that turns cues into signals or adds signals as evolved components to a display consisting of cues that did not evolve for the purpose of communication (Scott et al., 2010; Steiger et al., 2011b). Although ritualization is a well-accepted process, its significance for understanding the relationship between information and communication is not widely appreciated.

Ritualization can explain why communication is associated with information even if it evolves solely to influence others. This is because the noncommunicative selection pressures or constraints that led to the existence of reliable cues and thus to by-product information in the first place are not removed if selection through a communication role is added. This aspect is not trivial, yet influential monographs on communication (e.g. Maynard Smith & Harper, 2003) do not consider these noncommunicative selection pressures. The key aspect is thus that when cues become ritualized into signals, the information that was inadvertently available in the cue will often not be lost, but rather will be retained (Steiger et al., 2011b; Scott-Phillips et al., 2012). Because the original information is often retained during the evolution of communication, one could argue that it is no longer inadvertent once communication has evolved. We thus refer to it as by-product information, to contrast it with any information that evolves in response to selection by perceivers. This distinction is important for the conceptualization of communication because by-product information is consistent with the influence-based concept of communication, whereas the concept of information transfer views selection by perceivers as the most important force to explain why information is associated with signalling displays.

Constraints on signallers can limit cheating

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

Recently, constraints on perceivers have been emphasized in the debate on the conceptualization of communication (Owren et al., 2010), but the constraints acting on signallers have been neglected, likely because the evolution of communication is expected to increase their influence on perceivers. This does not mean, however, that constraints on signallers are unimportant or rare.

Ritualization has overlooked consequences for cheating. Because the original selective pressures remain that led to by-product information, these selective pressures constrain at the same time cheating because they will often impose a cost on cheats which are deviations from a naturally selected optimum. The very mechanisms that produce by-product information can even prevent cheating entirely. As an example, consider the regulation of reproduction, arguably the most crucial aspect of insect social societies, which is based upon cuticular hydrocarbon profiles. Cuticular profiles are environmentally determined cues that reliably indicate an individual's reproductive state in ants but also in other insects such as burying beetles (Wagner et al., 2001; Steiger et al., 2011a). Workers do not normally reproduce in the ant Aphaenogaster cockerelli. Cheating workers trying to reproduce are reliably identified by changes in their hydrocarbon profile that indicate their reproductive state. Such changes are not signals, because cheaters are policed (i.e. strongly attacked) in colonies that have a functioning queen, and thus pay a price if they are detected as being reproductively active (Smith et al., 2009). The fertility-related changes in hydrocarbon profiles cannot be suppressed because eggs need to have a reproductive profile, otherwise they are policed as well. A crucial point that prevents cheating is that the profiles of eggs and egg layers are necessarily correlated. This correlation identifies cheaters and reinforces reproductive altruism. Importantly, Smith et al. (2009) conclude that the very low levels of cheating by workers are not maintained by costs associated with producing a fertility signal, but rather the unavoidable consequence of basic biochemical mechanisms of hydrocarbon production and distribution (i.e. a cue that provides by-product information about reproductive state). This is thus an iconic example of a constraint that renders traits informative without costs. It also shows why the costs of cheating and not the costs associated with signal production or maintenance are important to maintaining reliability.

Constraints on cheating are imposed by various external and internal factors. Internal factors that influence signal phenotypes are developmental and genetic factors, such as pleiotropy and gene linkage. These inherent mechanisms are arguably best illustrated by polymorphisms, which are common in the natural world. For example, ~350 species of birds are polymorphic (Roulin, 2004). Morphs – be they heritable or phenotypically plastic – are characterized by co-adapted suites of traits that influence fitness interactively and that are maintained by correlational selection (Sinervo & Svensson, 2002; McKinnon & Pierrotti, 2010). A recent study on correlated selection lists 32 species whose colour polymorphisms are genetically correlated with other traits (McKinnon & Pierrotti, 2010). Genetic correlations are often caused by gene linkage and pleiotropy and are orchestrated by differential production of hormones (West-Eberhard, 2003). Melanin-based coloration is thought to be a prime example of pleiotropy (Ducrest et al., 2008) with various behavioural and physiological traits linked to variants in coloration (Cieslak et al., 2011).

External factors that impose constraints on cheating can be biotic or abiotic. For example, the visual signalling of flower coloration may be pleiotropically influenced by adaptation to edaphic conditions or by defence against herbivores (Irwin et al., 2003; Schemske & Bierzychudek, 2007). Predators also limit the expression of sexually selected colours in animals, such as male guppies (Poecilia reticulata, Endler, 1980). Another example of selection by third parties driving communicative traits is the Hamilton–Zuk model that predicts mate choice to rely on condition-dependent traits that reveal parasite burden or general vigour and health (Hamilton & Zuk, 1982). The information arising from host–parasite interactions is a by-product because it is originally not selected for its influence on conspecifics, but it can be used by mates. The information originates proximally from the multiple functions of pigments as colorants and antioxidants as well as from general resource limitations on signallers (see Blount et al., 2003; Faivre et al., 2003; Schaefer et al., 2008).

Although the constraints summarized above are well known, we feel that there has been insufficient realization in the debate on the conceptualization of communication that such noncommunicative selection pressures, and the information available to perceivers as a by-product of them, can be important in maintaining signal reliability. This limitation is illustrated by the fact that most theoretical models of signalling have been based on the simplifying assumption that the signal perceiver imposes the sole selective pressure on signalling traits. As stated earlier, this need not be so and in general will not be. To our knowledge, no modelling study has attempted to formalize the evolution of communication under the explicit assumption that constraints on signallers partly restrict their options to reveal or conceal information. If the values of traits used in signalling are set at least partly by mechanisms other than communication with any particular perceiver, we predict that cheating will often consist of concealing traits that inherently yield information. Cheating is thus more restricted than current game theory models assume it to be.

To illustrate the difference between whether reliability depends on costs and constraints of cheating or on costs of signalling, we now contrast by-product information and the reliability it provides to the much better known handicap principle.

Difference between by-product information and handicaps

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

The handicap principle predicts that reliability is maintained by strategic costs of signals, which are termed handicaps, and that strategic costs are higher for lower-quality signallers (Zahavi, 1975; Grafen, 1990). Recently, some theoreticians have advocated a paradigm shift because the costs of cheating and manipulation, rather than of signalling, are key to understanding reliability (Lachmann et al., 2001; Getty, 2006; Számado, 2011). Accordingly, reliable signals are not necessarily handicaps; they can be cost-free at an honest equilibrium as long as cheating, that is manipulating the perceiver to its disadvantage, is costly. Handicaps only lead to honest signalling if there is some linkage between the costs of signalling at equilibrium and the costs of cheating; such linkage is possible but certainly not universal. We agree with the logic of this argument. But the title of articles such as ‘The cost of honesty and the fallacy of the handicap principle’ (Számado, 2011) might be misinterpreted as suggesting that the handicap concept should be abandoned. This would be premature. Firstly, because the handicap theorem is logically correct, and secondly, because there is good evidence from a number of communication systems that such costs can indeed occur and prevent cheating. For example, nest decorations in the black kite (Milvus migrans) reliably indicated body condition and the quality of the territory. Reliability is enforced by costs because decorations increased the visibility of nests and therefore the rates of attempted egg predation (Sergio et al., 2011). Only high-quality individuals were able to raise their attack rates on predators to counter these costs. The authors concluded that, more generally, extended phenotypes such as nest decorations can be reliable because they represent cost-enforced handicaps.

In many communication systems, empirical demonstration of the costs required for handicap signalling has proved elusive (e.g. Cotton et al., 2004). Moreover, handicap signalling appears to be rare or absent in major communication systems like prey–predator relationships or communication among mutualists (Edwards & Yu, 2007). The important conclusion, clearly described by Számado (2011), is that costs of cheating are required to stabilize reliable communication and that while handicap signalling can provide one route to this, a number of alternatives exist (Számado provides a nonexhaustive list of ten). These findings raise the question how widespread handicap signalling is and how it can be distinguished from other routes to reliability.

The main difference between handicap signalling and the concept of the costs of cheating is that only handicap signalling expects perceivers to be the selective force driving strategic costs of signalling to produce reliability. This expectation matches the basic assumption of the information-based approach to communication. Conversely, if the costs associated with cheating generate reliability, then any inherent (e.g. genetic or physiological) mechanism underlying by-product information can stabilize a communication system, because it will concomitantly impose some costs on cheating. Surprisingly, even though inherent mechanisms are likely common, they have not received sufficient attention. For example, Számado (2011) included in his list of ten alternative routes to reliability only indices as examples of inherent constraints. As made explicit by Maynard-Smith & Harper (1995), an index is necessarily and physically related to a given quality; it is physically impossible (and not simply costly) for a signaller to produce a misleading value of an index. Because indices prevent cheating entirely, there is no controversy about how signals can evolve along this route. Indices thus represent an extreme case of how costs associated with internal mechanisms shape communication systems. We now focus on less extreme cases because they highlight the interplay between by-product mechanisms and evolved components during the process of ritualization.

Social signalling among primates is a good example of (i) inherent mechanisms being the starting point for the evolution of communication and (ii) the importance of evolved components for the maintenance of signalling. Primates often use red skin colours to communicate. Red skin colour is caused by the blood flow in skin tissues and enhanced directly by oestrogen through peripheral vasodilation and indirectly by testosterone through conversion to oestrogen (Rhodes et al., 1997). Hormones did not evolve for their effect on skin coloration; the information they provide arises as a by-product of physiological mechanisms. These mechanisms provide a reliable intersexual signal of fertility, for example, in the sexual swellings of female primates (Stallmann & Froehlich, 2000). Red skin colour can also function as intra-sexual signal. For example, the most dominant male mandrills (Mandrillus sphinx) are the reddest and the ones with highest circulating testosterone levels (Wickings & Dixon, 1992; Renoult et al., 2011). Thus, both fertility and dominance signals rely on the inherent link between hormone status and skin coloration, which restricts the options for cheating in social hierarchy as well as in feigning or hiding fertility.

Information and the transition from cues to signals

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

Despite the inherent link between red skin colours in primates and their hormonal status, red skin colour is not simply an unfakeable index. Experiments revealed that red skin coloration is not solely a passive by-product of hormonal status because only skin areas used for social signalling redden after administration of hormones, whereas the colour of other skin areas remains unchanged (Rhodes et al., 1997). These experiments document a heightened condition dependency to hormonal status of sexually selected skin areas in primates (compared to other skin areas). Such a heightened condition sensitivity has recently been demonstrated in the large horns of Rhinoceros beetles (Trypoxylus dichotomus). Growing horns were more sensitive (than wings and genitalia) to the insulin/insulin-like growth factor pathway that regulates tissue growth and body size (Emlen et al., 2012). Importantly, heightened sensitivity to an inherent physiological mechanism increases relative variability in trait size between low- and high-condition individuals (compared to other traits), allowing perceivers to assess signallers more accurately. Emlen et al. (2012) thus conclude that exaggerated morphological traits, which are most commonly seen as handicaps, are not handicap signals maintained by the costs of signalling. Rather the authors conclude that exaggerated traits possess ‘intrinsic reliability’, a heightened sensitivity to a basic physiological mechanism, which is informative to others.

Primate social signals and the exaggerated size of the horn of the Rhinoceros beetle document two fundamental issues for the evolution of communication. First, information is often a by-product that is based on an inherent mechanism which does not evolve during ritualization. Thus, information typically does not evolve. Rather, we predict that ritualization will usually preserve those traits of the original cue that provide information in essentially unchanged form. Second, traits can evolve during ritualization that allow signallers to gain influence (e.g. by becoming more conspicuous) and perceivers to better assess information. Such traits can be amplifiers, which do not change or add information but facilitate the assessment of the perceived information provided by pre-existing cues (Hasson, 1989). Amplifiers thus testify to the importance of viewing information as a property of perceivers. The abdominal patterns of the jumping spider Plexippus paykulli are a good example of an amplifier. These patterns facilitate assessment of difference in physical condition (Taylor et al., 2000). In essence, this is what both red skin colours in primates and large horns in Rhinoceros beetles do; they amplify the pre-existing information of differences in hormonal and insulin-related status. The important insight is thus that communication is not the cause but the consequence of reliability.

A good example for our prediction that behavioural and sensory traits undergo more profound changes during ritualization than informational ones is social communication in ants. Their hydrocarbon profiles are a cue strongly influenced by noncommunicative selection pressures, whereas the behaviour of ants to sense the hydrocarbons with their antennae or postures adopted by ants that allow efficient inspection of hydrocarbon profiles by nestmates will have evolved substantially because of the effect that they have on communication. This example elucidates that signals are complexes of traits that differ in their evolutionary histories. The nonevolved component (hydrocarbon profile and its associated information) is integral to the signal as is the evolved component (such as the postural behaviour).

So far, we have stated that information does not typically evolve during the evolution of a signal from a cue because it relies on a basic mechanism, which has been selected for its noncommunicative role. However, the reliability of information can evolve, and it can be selected by perceivers. The early stages of the evolution of an index provide an example of it. At first, this seems counterintuitive because an index is unfakable, and hence, its reliability should be perfect before the start of ritualization. This assumption needs modification if the index involves a behavioural component. Maynard-Smith & Harper (1995) cite the territorial marking of tigers that leave scratch marks as high as they can on tree trunks as an example of an index. The scratch marks and the associated behaviour are signals; their position on the tree is determined by the size of a tiger, but tiger size did not evolve primarily for the purpose of communication (but is a trait of interest to other tigers). Leaving scratch marks might have been a behaviour originally related to stretching of the muscles or cleaning and sharpening of the claws. There is a constraint that smaller tigers simply cannot make a mark high up the tree. At this stage, before signalling develops, there is some useful information in this cue: marks high up a tree would be indicative of a large individual, whereas lower marks would be less informative. When conspecifics began to respond to this cue, this produces selection pressure for the cue to be ritualized, that is, the cue evolved into a signal. Large individuals benefitted from a change in behaviour, whereby they always scratch as high as they are physically able, and began scratching more frequently and/or in more prominent positions. It is those changes in behaviour that produce a signal, but because smaller tigers are simply incapable of scratching as high as larger tigers, scratching is an index.

Although indices are not the only situation where by-product information is important, the example of tiger scratches elucidates (i) that by-product information of body size provided the starting point for a signal (the marking behaviour) to evolve and (ii) that signal evolution did not alter the mechanism producing information (small tigers are physically incapable of scratching as high as larger tigers). Crucially, however, the evolution of communication led to more reliable, that is, more accurate information, because all tigers scratch as high as they physically can after scratching became ritualized. This improvement in the quality of information is achieved because not scratching at all or scratching lower than the physical limit would be disadvantageous for any tiger, but scratching higher than the limit set by the trait of interest (body size) is impossible. Again, constraints on cheating, not costs of signalling, determine reliability in an index.

Summary

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

The debate on the conceptualization of communication concentrates on whether information transfer is an integral part of the communication process. Our contention in this study is that information is commonly associated with communication, but that it evolves as by-product of mechanisms other than communication. Specifically, and unlike previous articles, we envision communication to be the consequence and not the cause of reliability. We contend that informational traits associated with many signals will be affected by noncommunicative selection pressures as well as by the response of any particular signal receiver. These noncommunicative mechanisms will naturally impose constraints on the possible values that signals can take and thereby impose a cost to cheating that will help preserve the reliability of signals. We can make considerable progress in understanding the conceptualization of communication by decomposing communicative traits into their different components and considering the different evolutionary histories and current selection pressures on these traits.

Outlook

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

The relationship between information and communication can be best understood by clearly differentiating between alternative models of communication. The model of sensory exploitation predicts that traits evolve which maximally stimulate the sensory system of the perceiver. These traits may or may not be informative. Handicaps, by-product information and index signals represent different mechanisms that can produce and maintain reliable signalling. Each concept differs in its prediction on the relationship between information and communication and on how information is selected. The handicap principle predicts reliability to be caused by strategic costs to signalling. Specifically, the handicap principle implies that information arises because it is selected by perceivers and is beneficial for high-quality signaller to reveal. According to the concept of by-product information, information preceded the evolution of communication and did not change during that process. Consequently, reliability is maintained by the selective forces that shaped the original cue and that restrict the options for cheating and not by strategic costs of signalling. The concept of index signalling predicts a direct and unfakeable relationship between signal and information. During the evolution of index signalling, perceivers can select for increased accuracy of information if the index involved a behavioural component. Only by distinguishing clearly among these different concepts can we hope to determine the relative role of information for the evolution of communication.

Acknowledgments

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

We thank Øistein Haugsten Holen, Tom Reader, Julien Renoult and Mike Speed for discussions and comments. We are particularly grateful to Drew Rendall and Thom Scott-Phillips for careful and thoughtful critiques of previous versions of this work. HMS is supported by DFG grants (Scha 1008/5-1, Scha 1008/6-1), and GDR is supported by NERC.

References

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

Glossary

  1. Top of page
  2. Abstract
  3. Introduction on communication
  4. Alternative concepts of communication
  5. Definitions
  6. Ritualization of by-product information
  7. Constraints on signallers can limit cheating
  8. Difference between by-product information and handicaps
  9. Information and the transition from cues to signals
  10. Summary
  11. Outlook
  12. Acknowledgments
  13. References
  14. Glossary

By-product information: is the perceived correlation between the qualities a perceiver is interested in and the sensory stimuli it perceives if this correlation has not been selected for a communicative function by the perceiver.

Communication: is the interaction between a signaller and a perceiver if some trait values of the signaller evolved to stimulate the sensory system of the perceiver in such a way as to cause a change in the behaviour of the perceiver.

Cues: are traits that are informative but did not evolve to influence the behaviour of other organisms (either through manipulation or through information).

Information: is the averaged correlation in the trait values perceived by the perceiver and a quality that the perceiver is interested in.

Reliability: The strength of the averaged correlation in the trait values perceived by the perceiver and a quality that the perceiver is interested in.

Signals: are traits or behaviours that evolved to influence the behaviour of other organisms (either through manipulation or through information).