- Email: [email protected]
“Conventional” Signals in Avian Agonistic Displays: Integrating Theory, Data and Different Levels of Analysis
J. theor. Biol. (1999) 196, 155–162 Article No. jtbi.1998.0825, available online at http://www.idealibrary.com on
‘‘Conventional’’ Signals in Avian Agonistic Displays: Integrating Theory, Data and Different Levels of Analysis J M. D*† G W. S‡ *Institute of Cell, Animal and Population Biology, University of Edinburgh, Ashworth Laboratories, King’s Buildings, West Mains Road, Edinburgh EH9 3JT, U.K. and ‡School of Environmental Sciences, University College Scarborough, Filey Road, Scarborough, North Yorkshire YO11 3AZ, U.K. (Received on 7 July 1997, Accepted in revised form on 7 September 1998)
We present an integration of communication theory and data, drawing on examples from titmice (Aves: Paridae). We suggest how display actions such as lifting the head, raising the nape feathers, crest erecting and spreading the wings, act in agonistic communication when physical contact between opponents is rare. We propose that such displays largely act as strategic choice handicap signals. By giving these displays the signaller is believed to incur costs which underwrite the reliability of the signals; it may strategically increase these costs (for example by display repetition or adding additional elements) to signal its condition, motivation and hence the subjective value of a resource. It is shown that linking these ideas with earlier theories on the causation of display components, leads to an explanation of why birds have a greater repertoire of signals associated with aggression/winning, than with submission/losing. It is suggested that modellers of communication systems and those interested in the theory of signal design should pay more attention to the evolutionary constraints imposed by signal origin. 7 1999 Academic Press
1. Introduction Some elements of animal displays are labelled as conventional signals because the link between signal design and message appears arbitrary, or because they appear to be a strategic (rather than obligate) correlate of signaller quality (Guilford & Dawkins, 1995). Thus in avian agonistic displays, many putative signals appear to be relatively inexpensive to perform and there appears to be little correlation between the structure of the signal and underlying attributes † Author to whom correspondence should be addressed. E-mail: j.deag.ed.ac.uk 0022–5193/99/002155 + 08 $30.00/0
such as size and strength. Examples from the titmice (Aves: Paridae) include spreading the wings or tail, lifting the head to reveal a throat or chest patch, holding the bill at a particular angle, or lifting the feathers on the nape or crest (Stokes, 1962; Blurton-Jones, 1968; Wilson, 1992a; Scott, 1994; Scott & Deag, 1998). Ethologists and theoretical biologists interested in modelling communication have attempted to explain the evolutionary origin of conventional signals and their role in the resolution of agonistic encounters. Much of the discussion has focused on the signalling of intention (Caryl, 1979, 1982a, b; Hinde, 1981; 7 1999 Academic Press
156
. . . .
Enquist, 1985; Popp, 1987; Hauser & Nelson, 1991; Scott & Deag, 1998), the extent to which such signalling systems are open to cheating (Enquist, 1985; Grafen, 1990; Guilford & Dawkins, 1995; Maynard Smith & Harper, 1995; Johnstone, 1997), and the importance of individuality and context (Wilson, 1992a). We present here an integration of theory and data which incorporates ideas from ethology, game theory, and the handicap principle (Zahavi, 1975; Zahavi & Zahavi, 1997) to explore how conventional signals may act in agonistic communication when physical contact between opponents is rare. It suggests an explanation for a question which modern theoretical studies have found difficult to answer: why within a species do signals associated with aggression/winning out number those associated with submission/losing (Scott & Deag, 1998)? Our proposal is compatible with and incorporates current ideas in signal theory: reduced signalling in situations of prior knowledge/individual recognition (Wilson, 1992a); the importance of signaller–receiver interaction in determining the exchange of signals (Hinde, 1981); signal cost, honesty and evolutionary stability (Enquist, 1985; Grafen, 1990; Maynard Smith, 1991; Dawkins, 1993; Maynard Smith & Harper, 1995); ‘‘receiver psychology’’ (Guilford & Dawkins, 1991); the importance of resource value (Enquist & Leimar, 1987); and error-prone signal theory (Grafen & Johnstone, 1993; Johnstone, 1994). It also suggests some testable hypotheses. When exploring these issues we pay particular attention to the agonistic displays of titmice. Studies of these birds have been influential in assessing ethological ideas on display and the impact of theoretical developments such as game theory (Stokes, 1962; Blurton-Jones, 1968; Caryl, 1979, 1982a, b; Hinde, 1981; Wilson, 1992a) but their system of communication is still not understood. 2. The Cost of ‘‘Conventional’’ Signals Understanding signal cost is now seen as central to the definition of ‘‘conventional’’ signalling (Dawkins, 1993; Guilford & Dawkins, 1995). An important reason for separating conventional signals from signals linked with basic attributes such as size or strength, has been
the assumption that the costs born by the signaller are very different (Maynard Smith & Harper, 1988; Dawkins, 1993). Evolutionary theory suggests that costly signals (e.g. antler size, roaring in red deer stags, tail-beating in fish, repeated jumping up and down in bird courtship displays) are an honest advertisement of the signaller’s physical attributes, reflect prior investment, and are hard to cheat. By contrast, low cost conventional signal systems might be prone to cheating and hence, when the signaller’s and receiver’s interests differ, receivers might be expected to be sceptical and to retaliate against dishonest signallers. There may, however, be circumstances (e.g. potentially large cost/benefit ratio) in which it is not worth a receiver constantly testing signal honesty. It has therefore been proposed that conventional signal systems may include a level of dishonesty (Dawkins & Guilford, 1991; Dawkins, 1993). A different (but not necessarily incompatible) view which we build on here, suggests that the costs associated with conventional signals have been underestimated. It proposes that all signal systems (except relatively pure conventional and symbolic systems such as human language), must be costly to be stable and therefore most so-called conventional systems are a fallacy (Zahavi, 1993). The empirical assessment of the cost [defined as equivalent to loss of fitness by Maynard Smith & Harper (1995)] of conventional signals remains elusive but it is clear that costs may be paid at different times and in different ways (Enquist, 1985; Dawkins, 1993; Adams & MestertonGibbons, 1995; Guilford & Dawkins, 1995). The cost may be paid sometime before signalling (e.g. during growth or moulting) and hence independently of a receiver’s response. Thus feather growth for crests and colourful patches may be dependent upon hormone conditions prior to moulting and these hormonal conditions may in turn depend upon costly past social behaviour such as fighting (Wilson, 1992b). The cost may also be paid at the time of signalling either directly (e.g. energy loss, less time for other important activities such as feeding, increased risk of predation) or as a consequence of the receiver’s response [e.g. increased risk of injury should a receiver choose to challenge the
signaller (Zahavi, 1977; Waas, 1991)]. A further and generally unacknowledged problem is that there are differences between signals in the extent to which costs may be allocated specifically to the function of communication: a colourful patch may be only used for communication but Wings Out and Bill Up postures use structures primarily designed for flight and feeding. 3. Agonistic Behaviours as Strategic Choice Handicap Signals We shall concentrate here on signals in which the costs specifically attributable to communication are largely paid at the time of signalling; it is acknowledged that in some displays such signals may be combined with other types of signal. As examples consider the agonistic displays of the great tit Parus major, blue tit P. caeruleus and other Parids competing for food (Stokes, 1962; Blurton-Jones, 1968; Wilson, 1992a; Scott & Deag, 1998). Once an interaction is underway a series of displays are given and the interaction may be viewed as a period of negotiation leading to contest resolution. The displays are made up of posturing and gross body movements towards or away from the opponent with behavioural elements such as spreading the wings (WO), spreading the tail (TA), opening the bill (BO), pointing the head and bill down (HD), lifting the head and bill (HU), erecting the nape feathers (NE), and erecting the crest (CE). Interactions end when one bird leaves (escapes, loses the interaction). Many of the elements may be given independently or in combination (Scott & Deag, 1998). A series of studies have shown that these elements are generally not role specific; they may be given by both winners and losers. Nevertheless the performance of most of the elements (e.g. NE, WO, TA, BO, HD, HU) is more associated with a tendency in the signaller to approach/ attack/win than to retreat/escape/lose; the situation is reversed in one element (CE). CE has been considered to provide a much higher level of prediction of escape by the signaller than the other elements do for attack (Caryl, 1979). However recent work suggests that in blue tits this result is an artefact of recording many simple approaches as ‘‘attacks’’; the performance of all
157
these elements provides a high prediction of whether the signaller will lose (CE; 66% prediction) or win (NE, WO, TA, BO, HD; 67–75% prediction) (Scott & Deag, 1998). The numerical supremacy of aggressive elements over submissive elements is the general rule for this type of display but an explanation for this fact has remained elusive. It is part of the broader and unresolved question of why there are so many threat signals in a species’ repertoire (Andersson, 1980). We suggest that such elements of behaviour in the agonistic displays of titmice and other species should be interpreted as strategic choice handicap signals (Grafen, 1990; Guilford & Dawkins, 1995), and that doing so leads to an understanding of several issues. The following explanation is based on some costs of these signals which have been inadequately investigated, probably as a consequence of the assumption that conventional signals (in contrast to assessment signals) are relatively cost-free. The handicap (in the sense of Zahavi, 1975; Grafen, 1990; see also Johnstone, 1997), is the cost which the signaller is willing to bear in order to underwrite the reliability of its signal. In a strategic choice handicap (Grafen, 1990) the level of handicap has a dynamic component, the signaller varying the magnitude of the handicap according to the circumstances (e.g. some combination of factors such as the subjective value of a resource and the perceived challenge from the opponent). A logical, non-arbitrary relationship is expected between the attribute being signalled and the handicap, such that signalling ‘‘uses up’’ or ‘‘wastes’’ some of the attribute (Zahavi & Zahavi, 1997). In the present context the signaller is indicating its desire to defend (or obtain) the food resource and hence indirectly the value of the food to itself. Leaving aside the small energy costs of display (it is hard to imagine that crest or nape raising are costly with respect to other aspects of the daily energy budget), there are two types of handicap which could keep the signal honest. The signaller could place itself at increased risk from its opponent (Zahavi, 1977), or it could waste energy or delay its acquisition. Both satisfy the requirement that poorer individuals who bluffed would suffer a cost which they are going to be less likely to bear.
158 3.1.
. . . . (‘‘ ’’)
Zahavi (1977, 1980) proposed that a system communicating threat could only be stable if threatening carries information on the motivation to fight and a risk of retaliation which would make bluffing by low quality individuals risky. This removes the instability due to revealing intention identified in earlier theories of motivational signalling. Zahavi (1977, 1980) suggested that body movements towards the opponent may increase the likelihood of retaliation, and that display elements (in the present context these might include NE, WO, TA, HU) which involve extreme muscular contraction make the signaller less able to respond promptly and so more vulnerable. This has not been substantiated. However, the idea has received support from Waas (1991) who undertook a risk–benefit analysis of agonistic displays in little blue penguins disputing ownership of nest caves. He found a positive correlation between the risk of performing a display and its effectiveness, and that birds would escalate or de-escalate their signal according to the response of the opponent. Up to 10% of the interactions involving males escalated to attacks, some with serious injury. This level of escalation must reflect the high value of the resource. Wilson (1992a) recorded a similar level of escalation to attack in displays by great tits over food but in a study of blue tits feeding under similar circumstances, Scott & Deag (1998) found much lower levels: fights in Q1% of interactions and peck attacks in a further 2%. There also appears to be no simple pattern of escalation of displays during blue tit interactions (unpublished observations). This suggests that the risk of retaliation may not always be the active handicap, particularly when the resource is of relatively low value (both in absolute terms and because there are alternatives available). It is possible that the low risk costs suggested for blue tit interactions might be deceptive. If the birds frequently use the same elements in other situations where the risk costs were higher (e.g. in disputes over more valuable resources such as territory), the perception of
risk could carry over from one situation to another. (‘‘ ’’) Blue tits spend up to 85% of their time feeding and searching for food during the winter (Gibb, 1954) and overnight survival depends upon accumulating sufficient reserves during the day. Under these circumstances the purest and simplest handicap a resource holder could impose over food would be to waste time by stopping feeding and do nothing else, a type of War of Attrition developing between the birds to see who gives up first. Given the motivational circumstances, value of the resource, and asymmetries between individuals (discussed below) we feel that this strategy is unlikely to be seen without embellishment in nature. 3.2.
In addition to the above there are also correlated costs which are a consequence of display; they should not be considered handicaps because they are not directly associated with the attribute being signalled. They include drawing attention to oneself and a reduction of awareness of other stimuli [e.g. reduced vigilance leading to predation as seen in pigeon guillimots (Nelson, 1984)]. 4. Resource Value, Motivation and the Structure of Displays In small temperate birds competing in winter, numerous factors appear to influence subjective resource value and hence the willingness to defend or obtain a food resource. These include hunger, resource nutritional value and size, site-dependent rank, the outcome of recent encounters and resource ownership. The latter is an advantage but not a guarantee of success (Scott & Deag, 1998). Under these circumstances the agonistic display system must have a large dynamic component. Signal repetition is an obvious feature of agonistic displays. We see it as the way by which the signaller (resource holder or challenger) strategically increases its handicap to signal its condition, its motivation to continue displaying, and reluctance to give up
the resource or determination to obtain it. In this way it signals the subjective resource value (i.e. the value of the disputed resource to itself, Enquist et al., 1985; Enquist & Leimar, 1987). In addition, display repetition may guard against a receiver’s errors in perception. We see selection acting on receivers who also value a disputed resource, to demand evidence of the signallers handicap and hence selection acting on signallers to signal this. In this way the system will be kept largely honest. Birds signalling beyond their motivation and ability, pay risk and energy costs; their correlated costs may be higher if they have to pay increased attention to their opponent. As noted above, the purest and simplest strategic choice handicap a signaller defending food could impose upon itself, would be to waste time by stopping feeding but do nothing else. However, while this would satisfy the theoretical requirements for a handicap it is an unrealistic expectation in nature. The challenger will be motivated to take possession of the food and the possessor will be motivated to defend it. Ethological studies have shown that under such circumstances, many of the signals or unritualised intention movements given are associated with underlying but incompatible tendencies such as approach, attack, escape, stay put (Tinbergen, 1959; Blurton-Jones, 1968; Hinde, 1970). While unfashionable (behavioural ecology seems obsessed with function), these ideas on the causation of display components have proved remarkably robust and have not been superseded (Baerends, 1975; Hinde, 1970, 1981). They have been largely ignored by theorists working on signalling systems but studies of signal design should take account of signal origin. The appearance of behavioural elements in situations of motivational conflict has provided a rich source of material for selection to shape into signals. We believe that in the evolution of agonistic displays, these elements have been picked up unequally by selection and that this has lead to the discrepancy in the relative frequency of aggressive and submissive elements. Selection is seen to lead to the increased repertoire of signals associated with aggression, as signallers (resource holders or challengers) are forced by sceptical receivers to increase their handicap. They do so by showing elements
159
associated with different levels of their prime motivation: aggression, resource defence and unwillingness to leave. The multiple components can thus be viewed as ‘‘backup signals’’ in the sense of Johnstone (1996, 1997). Some, such as Head Up in the great tit (Wilson, 1992a) may be brought mainly into play if resource value and motivation to defend it are both high; they may therefore give the impression that they are associated with signalling attributes such as territory ownership and prior residence. By contrast, there is no equivalent selection acting on a bird to increase the number of signals associated with escape/submission; simply repeating the same element (e.g. Crest Erect) will often be enough and will minimize ambiguity. Further more, we suggest that during a period of negotiation, an increased repertoire size and signal repetition (a frequent aspect of signal ritualization) permits a signaller to increase its handicap in smaller increments than would otherwise be the case. This idea provides an explanation for how and why different behavioural elements are used to increase the duration of the interaction during a period of negotiation over a resource. There is no need to refer here to the evolutionary processes involved in shaping signal form, except to point out that recent ideas on the influence of receiverpsychology (Guilford & Dawkins, 1991), limitations on honesty (Dawkins & Guilford, 1991; Johnstone & Grafen, 1993), and the consequences of error-prone signalling (Grafen & Johnstone, 1993; Johnstone, 1994), are compatible with our proposals. 5. The Information Content of Signals The predictability of a signaller’s future behaviour (however imperfect), runs counter to theoretical studies which suggest that signallers should conceal their intention in agonistic encounters (Caryl, 1979). We believe that a combination of factors are responsible. As noted above, ethological studies have shown that much agonistic communication takes place within a situation where there is at least the potential for motivational conflict. It is likely therefore that the elements shown may be a mixture of signals and of unritualised intention movements that
160
. . . .
have not yet been selected as signals. As Caryl (1982b) noted the latter tend to be ignored by game theorists modelling agonistic displays. We also believe that the iconic nature of many agonistic signals (e.g. those derived from intention movements or associated with autonomic arousal), and the consequences of this have been underestimated (see for example Maynard Smith & Harper, 1995). In a truly conventional signalling system, signals would be abstract and non-iconic. A signal would have a form unrelated to internal factors (motivation) and immediate costs. Ethologists use the concepts of emancipation for the process by which the proximate control of a signal is presumed to be separated from its ancestral motivation during evolution. It is clear however that emancipation is often missing or incomplete in the behavioural elements of titmice (BlurtonJones, 1968) and as a consequence the iconic nature of some behavioural elements automatically carry information on intention. Are we studying young communication systems (i.e. systems in which there has been insufficient time for the evolution of emancipation), or is selection for emancipation in agonistic signals not as strong as previously assumed? We believe that in a strategic choice handicap system the latter is most likely. The fact that it may be adaptive to clearly signal a willingness to escape once this course of action has been decided (Caryl, 1979), suggests that there is little selection for emancipation of behaviours which signal escape. The need to incorporate new elements as additional handicaps will also act against the assumption that all elements are at an equal stage of emancipation. The problem of predicting a signaller’s future behaviour or predicting interaction outcome is confounded by the same elements being used by both winners and losers. At the causation level this is also attributed to ambivalence during a period of negotiation. However, there is also the problem that the cost of signal probing by receivers may result in a level of cheating (Guilford & Dawkins, 1991). Interactions will also differ in the ratio of the value of the resource to the resource holder and to the challenger; a bird which under other circumstances would win, might find it pertinent to withdraw if a
period of display reveals that the resource has a higher subjective value to its opponent (Enquist & Leimar, 1987; Popp, 1987; Cristol, 1992). Signalling escape may be both submissive (indicates a desire to stay without being threatening) and appeasing (reduces the likelihood of attack before escape). The repetition of escape signals may be a cost of wishing to delay escape. A further possibility which we have not yet considered is that some agonistic displays act as revealing handicaps (Grafen, 1990; Maynard Smith, 1991). In this case the display would be expected to directly reveal some underlying quality of the signaller (e.g. size, condition) which the recipient cannot readily detect in a non-displaying bird. Thus some postures might reveal body size but we suggest that this is unlikely to be important. Wilson (1992a) concluded that in great tits agonistic behaviour was only weakly correlated with size, and body size frequently does not predict rank at a feeding table in great tits and blue tits (Wilson, 1989; Scott, 1994). More plausibly, postures such as Wings Out and Tail Fan could reveal honest information on condition and hence competitive ability. This is not incompatible with the idea that the same elements might be involved in strategic choice handicap displays. 6. Some Predictions One problem with testing hypotheses on the function of display systems is that it is difficult to devise truly competitive hypotheses. For example, the hypotheses proposed to explain why there are so many threat displays (Andersson, 1980) are not mutually exclusive. This makes analysis difficult but it does not invalidate the hypotheses since many aspects of behaviour are subject to more than one selection pressure. A further problem is that the low frequency of attack in blue tits suggests that the risk–benefit approach of Waas (1991) may not be fruitful. With these limitations in mind we suggest the following as examples of predictions the fulfilment of which would support our theory that many signals in blue tit and great tit agonistic displays over food are strategic choice handicaps. Ideally it is desirable to control for
context by observing dyadic displays (not fights since these involve assessment) over food, preferably between same sexed birds who are on neutral ground and have little prior knowledge of each other. We propose that winners should show a greater variety of elements, a greater frequency of each element, and that more elements will be combined with other elements at an instance in time instead of being given on their own. The timing and choice of elements used by one bird should be related to the perceived challenge, for example to the proximity of the opponent and to the choice and frequency of the opponent’s behaviour. It is predicted that the eventual winner will at some point in the interaction take and keep the lead in incrementing the display elements (type of elements and the rate of performance). When the subjective value of the resource to a pair of birds is reversed such that the expected loser becomes the winner, the behaviour of the winner and loser should change to follow the above pattern. Except when an encounter escalates to a direct supplanting approach, attack or fighting, there generally appears to be no simple act which triggers the end of the agonistic encounters of the type described. This suggests that either a bird gives up after an allotted time (linked to subjective resource value and display costs), or either the current or accumulated opponent’s behaviour indicates that the opponent places such a high value of the resource that it would be too costly to continue. The ‘‘scratch pad’’ memory system proposed by Grafen & Johnstone (1993) suggests one way in which to search for display rules, i.e. the link between behavioural performance and interaction outcome. 7. Conclusion Incorporating the handicap principle into explanations of agonistic conventional signalling systems with immediate (rather than past) costs, permits the linking of Tinbergen’s ethological analysis of display causation with a range of modern ideas, suggests explanations for previously unexplained facts and identifies underestimated constraints on display evolution. Modellers of communication systems should
161
clearly pay more attention to these issues; Maynard Smith & Riechert (1984) have made an important start with their conflicting-tendency model. While we have concentrated on ‘‘conventional’’ signals with immediate costs, we acknowledge that in practice mixed signalling systems exist (i.e. systems made up of signals with both past and immediate costs) (Guilford & Dawkins, 1995). For these situations (for example in fulmars competing for food, Enquist et al., 1985), a communication model based on an interaction between conventional signalling and assessment signalling is required.
This paper is dedicated to Aubrey Manning on the occasion of his retirement from the Chair of Natural History at the University of Edinburgh. GWS’s work on Parus caeruleus was supported by a S.E.R.C. Postgraduate Studentship at the University of Edinburgh. We thank an anonymous referee for helpful comments.
REFERENCES A, E. S. & M-G, M. (1995). The cost of threat displays and the stability of deceptive communication. J. theor. Biol. 175, 405–421. A, M. (1980). Why are there so many threat displays? J. theor. Biol. 86, 773–781. B, G. P. (1975). An evaluation of the conflict hypothesis as an explanatory principle for the evolution of displays. In: Function and Evolution in Behaviour (Baerends, G.P., Beer, C. & Manning, A., eds) pp. 187–227. Oxford: Clarendon Press. B-J, N. G. (1968). Observations and experiments on causation of threat displays of the Great Tit (Parus major). Anim. Behav. Monogr. 1, 75–158. C, P. G. (1979). Communication by agonistic displays: what can games theory contribute to ethology? Behaviour 68, 136–169. C, P. G. (1982a). Animal signals: a reply to Hinde. Anim. Behav. 30, 240–244. C, P. G. (1982b). Telling the truth about intentions. J. theor. Biol. 97, 679–689. C, D. A. (1992). Food deprivation influences dominance status in dark-eyed juncos, Junco hyemalis. Anim. Behav. 43, 117–124. D, M. S. (1993). Are there general principles of signal design? Phil. Trans. R. Soc. Lond. B 340, 251–255. D, M. S. & G, T. (1991). The corruption of honest signalling. Anim. Behav. 41, 865–873. E, M. (1985). Communication during aggressive interactions with particular reference to variation in choice of behaviour. Anim. Behav. 33, 1152–1161. E, M. & L, O. (1987). Evolution of fighting behaviour: the effect of variation in resource value. J. theor. Biol. 127, 187–205.
162
. . . .
E, M., P, E. & R¨, J. (1985). Aggressive communication in fulmars (Fulmarus glacialis) competing for food. Anim. Behav. 33, 1007–1020. G, J. (1954). Feeding ecology of tits, with notes on treecreeper and goldcrest. Ibis 96, 513–543. G, A. (1990). Biological signals as handicaps. J. theor. Biol. 144, 517–546. G, A. & J, R. A. (1993). Why we need ESS signalling theory. Phil. Trans. R. Soc. Lond. B 340, 245–250. G, T. & D, M. S. (1991). Receiver psychology and the evolution of animal signals. Anim. Behav. 42, 1–14. G, T. & D, M. S. (1995). What are conventional signals? Anim. Behav. 49, 1689–1695. H, M. D. & N, D. A. (1991). ‘‘Intentional’’ signalling in animal communication. Trends Ecol. Evol. 6, 186–189. H, R. A. (1970). Animal Behaviour 2nd Edn. New York: McGraw-Hill. H, R. A. (1981). Animal signals: ethological and games-theory approaches are not incompatible. Anim. Behav. 29, 535–542. J, R. A. (1994). Honest signalling, perceptual error and the evolution of ‘‘all-or-nothing’’ displays. Proc. R. Soc. Lond. B 256, 169–175. J, R. A. (1996). Multiple displays in animal communication: ‘‘backup signals’’ and ‘‘multiple messages’’. Phil. Trans R. Soc. Lond. B 251, 329–338. J, R. A. (1997). The evolution of animal signals. In: Behavioural Ecology: an Evolutionary Approach, 4th Edn (Krebs, J. R. & Davies, N. B., eds) pp. 155–178. Oxford: Blackwell. J, R. A. & G, A. (1993). Dishonesty and the handicap principle. Anim. Behav. 46, 759–764. M S, J. (1991). Theories of sexual selection. Trends Ecol. Evol. 6, 146–151. M S, J. & H, D. G. C. (1988). The evolution of aggression: can selection generate variability? Phil. Trans. R. Soc. B 319, 557–570. M S, J. & H, D. G. C. (1995). Animal signals: models and terminology. J. theor. Biol. 177, 305–311. M S, J. & R, S. E. (1984). A conflictingtendency model of spider agonistic behaviour: hybrid-
pure population line comparisons. Anim. Behav. 32, 564–578. N, D. A. (1984). Communication of intentions in agonistic contexts by the pigeon guillemot (Cepphus columba). Behaviour 88, 145–189. P, J. W. (1987). Risk and effectiveness in the use of agonistic displays by American goldfinches. Behaviour 103, 141–156. S, G. W. (1994). Factors influencing the resolution of agonistic interactions in a winter population of the Blue Tit (Parus caeruleus). Ph.D. Thesis, University of Edinburgh. S, G. W. & D, J. M. (1998). Blue Tit (Parus caeruleus) agonistic displays: a reappraisal. Behaviour (in press). S, A. W. (1962). Agonistic behaviour among blue tits at a winter feeding station. Behaviour 19, 118–138. T, N. (1959). Comparative studies of the behaviour of gulls (Laridae): a progress report. Behaviour 15, 1–70. W, J. R. (1991). The risks and benefits of signalling aggressive motivation: a study of cavedwelling little blue penguins. Behav. Ecol. Sociobiol. 29, 139–146. W, J. D. (1989). The function of agonistic interaction, social dominance and display in a winter population of the Great Tit Parus major L. Ph.D. Thesis, University of Edinburgh. W, J. D. (1992a). Correlates of agonistic display by great tits Parus major. Behaviour 121, 168–214. W, J. D. (1992b). A re-assessment of the significance of status signalling in wild populations of great tits (Parus major). Anim. Behav. 43, 999–1009. Z, A. (1975). Mate selection–selection for a handicap. J. theor. Biol. 53, 205–214. Z, A. (1977). Reliability in communication systems and the evolution of altruism. In: Evolutionary Ecology (Stonehouse, B. & Perrins, C., eds) pp. 253–259. London: Macmillan. Z, A. (1980). Ritualization and the evolution of movement signals. Behaviour 72, 77–81. Z, A. (1993). The fallacy of conventional signalling. Phil. Trans. R. Soc. Lond. B 340, 227–230. Z, A. & Z, A. (1997). The Handicap Principle. New York: Oxford University Press.
‘‘Conventional’’ Signals in Avian Agonistic Displays: Integrating Theory, Data and Different Levels of Analysis J M. D*† G W. S‡ *Institute of Cell, Animal and Population Biology, University of Edinburgh, Ashworth Laboratories, King’s Buildings, West Mains Road, Edinburgh EH9 3JT, U.K. and ‡School of Environmental Sciences, University College Scarborough, Filey Road, Scarborough, North Yorkshire YO11 3AZ, U.K. (Received on 7 July 1997, Accepted in revised form on 7 September 1998)
We present an integration of communication theory and data, drawing on examples from titmice (Aves: Paridae). We suggest how display actions such as lifting the head, raising the nape feathers, crest erecting and spreading the wings, act in agonistic communication when physical contact between opponents is rare. We propose that such displays largely act as strategic choice handicap signals. By giving these displays the signaller is believed to incur costs which underwrite the reliability of the signals; it may strategically increase these costs (for example by display repetition or adding additional elements) to signal its condition, motivation and hence the subjective value of a resource. It is shown that linking these ideas with earlier theories on the causation of display components, leads to an explanation of why birds have a greater repertoire of signals associated with aggression/winning, than with submission/losing. It is suggested that modellers of communication systems and those interested in the theory of signal design should pay more attention to the evolutionary constraints imposed by signal origin. 7 1999 Academic Press
1. Introduction Some elements of animal displays are labelled as conventional signals because the link between signal design and message appears arbitrary, or because they appear to be a strategic (rather than obligate) correlate of signaller quality (Guilford & Dawkins, 1995). Thus in avian agonistic displays, many putative signals appear to be relatively inexpensive to perform and there appears to be little correlation between the structure of the signal and underlying attributes † Author to whom correspondence should be addressed. E-mail: j.deag.ed.ac.uk 0022–5193/99/002155 + 08 $30.00/0
such as size and strength. Examples from the titmice (Aves: Paridae) include spreading the wings or tail, lifting the head to reveal a throat or chest patch, holding the bill at a particular angle, or lifting the feathers on the nape or crest (Stokes, 1962; Blurton-Jones, 1968; Wilson, 1992a; Scott, 1994; Scott & Deag, 1998). Ethologists and theoretical biologists interested in modelling communication have attempted to explain the evolutionary origin of conventional signals and their role in the resolution of agonistic encounters. Much of the discussion has focused on the signalling of intention (Caryl, 1979, 1982a, b; Hinde, 1981; 7 1999 Academic Press
156
. . . .
Enquist, 1985; Popp, 1987; Hauser & Nelson, 1991; Scott & Deag, 1998), the extent to which such signalling systems are open to cheating (Enquist, 1985; Grafen, 1990; Guilford & Dawkins, 1995; Maynard Smith & Harper, 1995; Johnstone, 1997), and the importance of individuality and context (Wilson, 1992a). We present here an integration of theory and data which incorporates ideas from ethology, game theory, and the handicap principle (Zahavi, 1975; Zahavi & Zahavi, 1997) to explore how conventional signals may act in agonistic communication when physical contact between opponents is rare. It suggests an explanation for a question which modern theoretical studies have found difficult to answer: why within a species do signals associated with aggression/winning out number those associated with submission/losing (Scott & Deag, 1998)? Our proposal is compatible with and incorporates current ideas in signal theory: reduced signalling in situations of prior knowledge/individual recognition (Wilson, 1992a); the importance of signaller–receiver interaction in determining the exchange of signals (Hinde, 1981); signal cost, honesty and evolutionary stability (Enquist, 1985; Grafen, 1990; Maynard Smith, 1991; Dawkins, 1993; Maynard Smith & Harper, 1995); ‘‘receiver psychology’’ (Guilford & Dawkins, 1991); the importance of resource value (Enquist & Leimar, 1987); and error-prone signal theory (Grafen & Johnstone, 1993; Johnstone, 1994). It also suggests some testable hypotheses. When exploring these issues we pay particular attention to the agonistic displays of titmice. Studies of these birds have been influential in assessing ethological ideas on display and the impact of theoretical developments such as game theory (Stokes, 1962; Blurton-Jones, 1968; Caryl, 1979, 1982a, b; Hinde, 1981; Wilson, 1992a) but their system of communication is still not understood. 2. The Cost of ‘‘Conventional’’ Signals Understanding signal cost is now seen as central to the definition of ‘‘conventional’’ signalling (Dawkins, 1993; Guilford & Dawkins, 1995). An important reason for separating conventional signals from signals linked with basic attributes such as size or strength, has been
the assumption that the costs born by the signaller are very different (Maynard Smith & Harper, 1988; Dawkins, 1993). Evolutionary theory suggests that costly signals (e.g. antler size, roaring in red deer stags, tail-beating in fish, repeated jumping up and down in bird courtship displays) are an honest advertisement of the signaller’s physical attributes, reflect prior investment, and are hard to cheat. By contrast, low cost conventional signal systems might be prone to cheating and hence, when the signaller’s and receiver’s interests differ, receivers might be expected to be sceptical and to retaliate against dishonest signallers. There may, however, be circumstances (e.g. potentially large cost/benefit ratio) in which it is not worth a receiver constantly testing signal honesty. It has therefore been proposed that conventional signal systems may include a level of dishonesty (Dawkins & Guilford, 1991; Dawkins, 1993). A different (but not necessarily incompatible) view which we build on here, suggests that the costs associated with conventional signals have been underestimated. It proposes that all signal systems (except relatively pure conventional and symbolic systems such as human language), must be costly to be stable and therefore most so-called conventional systems are a fallacy (Zahavi, 1993). The empirical assessment of the cost [defined as equivalent to loss of fitness by Maynard Smith & Harper (1995)] of conventional signals remains elusive but it is clear that costs may be paid at different times and in different ways (Enquist, 1985; Dawkins, 1993; Adams & MestertonGibbons, 1995; Guilford & Dawkins, 1995). The cost may be paid sometime before signalling (e.g. during growth or moulting) and hence independently of a receiver’s response. Thus feather growth for crests and colourful patches may be dependent upon hormone conditions prior to moulting and these hormonal conditions may in turn depend upon costly past social behaviour such as fighting (Wilson, 1992b). The cost may also be paid at the time of signalling either directly (e.g. energy loss, less time for other important activities such as feeding, increased risk of predation) or as a consequence of the receiver’s response [e.g. increased risk of injury should a receiver choose to challenge the
signaller (Zahavi, 1977; Waas, 1991)]. A further and generally unacknowledged problem is that there are differences between signals in the extent to which costs may be allocated specifically to the function of communication: a colourful patch may be only used for communication but Wings Out and Bill Up postures use structures primarily designed for flight and feeding. 3. Agonistic Behaviours as Strategic Choice Handicap Signals We shall concentrate here on signals in which the costs specifically attributable to communication are largely paid at the time of signalling; it is acknowledged that in some displays such signals may be combined with other types of signal. As examples consider the agonistic displays of the great tit Parus major, blue tit P. caeruleus and other Parids competing for food (Stokes, 1962; Blurton-Jones, 1968; Wilson, 1992a; Scott & Deag, 1998). Once an interaction is underway a series of displays are given and the interaction may be viewed as a period of negotiation leading to contest resolution. The displays are made up of posturing and gross body movements towards or away from the opponent with behavioural elements such as spreading the wings (WO), spreading the tail (TA), opening the bill (BO), pointing the head and bill down (HD), lifting the head and bill (HU), erecting the nape feathers (NE), and erecting the crest (CE). Interactions end when one bird leaves (escapes, loses the interaction). Many of the elements may be given independently or in combination (Scott & Deag, 1998). A series of studies have shown that these elements are generally not role specific; they may be given by both winners and losers. Nevertheless the performance of most of the elements (e.g. NE, WO, TA, BO, HD, HU) is more associated with a tendency in the signaller to approach/ attack/win than to retreat/escape/lose; the situation is reversed in one element (CE). CE has been considered to provide a much higher level of prediction of escape by the signaller than the other elements do for attack (Caryl, 1979). However recent work suggests that in blue tits this result is an artefact of recording many simple approaches as ‘‘attacks’’; the performance of all
157
these elements provides a high prediction of whether the signaller will lose (CE; 66% prediction) or win (NE, WO, TA, BO, HD; 67–75% prediction) (Scott & Deag, 1998). The numerical supremacy of aggressive elements over submissive elements is the general rule for this type of display but an explanation for this fact has remained elusive. It is part of the broader and unresolved question of why there are so many threat signals in a species’ repertoire (Andersson, 1980). We suggest that such elements of behaviour in the agonistic displays of titmice and other species should be interpreted as strategic choice handicap signals (Grafen, 1990; Guilford & Dawkins, 1995), and that doing so leads to an understanding of several issues. The following explanation is based on some costs of these signals which have been inadequately investigated, probably as a consequence of the assumption that conventional signals (in contrast to assessment signals) are relatively cost-free. The handicap (in the sense of Zahavi, 1975; Grafen, 1990; see also Johnstone, 1997), is the cost which the signaller is willing to bear in order to underwrite the reliability of its signal. In a strategic choice handicap (Grafen, 1990) the level of handicap has a dynamic component, the signaller varying the magnitude of the handicap according to the circumstances (e.g. some combination of factors such as the subjective value of a resource and the perceived challenge from the opponent). A logical, non-arbitrary relationship is expected between the attribute being signalled and the handicap, such that signalling ‘‘uses up’’ or ‘‘wastes’’ some of the attribute (Zahavi & Zahavi, 1997). In the present context the signaller is indicating its desire to defend (or obtain) the food resource and hence indirectly the value of the food to itself. Leaving aside the small energy costs of display (it is hard to imagine that crest or nape raising are costly with respect to other aspects of the daily energy budget), there are two types of handicap which could keep the signal honest. The signaller could place itself at increased risk from its opponent (Zahavi, 1977), or it could waste energy or delay its acquisition. Both satisfy the requirement that poorer individuals who bluffed would suffer a cost which they are going to be less likely to bear.
158 3.1.
. . . . (‘‘ ’’)
Zahavi (1977, 1980) proposed that a system communicating threat could only be stable if threatening carries information on the motivation to fight and a risk of retaliation which would make bluffing by low quality individuals risky. This removes the instability due to revealing intention identified in earlier theories of motivational signalling. Zahavi (1977, 1980) suggested that body movements towards the opponent may increase the likelihood of retaliation, and that display elements (in the present context these might include NE, WO, TA, HU) which involve extreme muscular contraction make the signaller less able to respond promptly and so more vulnerable. This has not been substantiated. However, the idea has received support from Waas (1991) who undertook a risk–benefit analysis of agonistic displays in little blue penguins disputing ownership of nest caves. He found a positive correlation between the risk of performing a display and its effectiveness, and that birds would escalate or de-escalate their signal according to the response of the opponent. Up to 10% of the interactions involving males escalated to attacks, some with serious injury. This level of escalation must reflect the high value of the resource. Wilson (1992a) recorded a similar level of escalation to attack in displays by great tits over food but in a study of blue tits feeding under similar circumstances, Scott & Deag (1998) found much lower levels: fights in Q1% of interactions and peck attacks in a further 2%. There also appears to be no simple pattern of escalation of displays during blue tit interactions (unpublished observations). This suggests that the risk of retaliation may not always be the active handicap, particularly when the resource is of relatively low value (both in absolute terms and because there are alternatives available). It is possible that the low risk costs suggested for blue tit interactions might be deceptive. If the birds frequently use the same elements in other situations where the risk costs were higher (e.g. in disputes over more valuable resources such as territory), the perception of
risk could carry over from one situation to another. (‘‘ ’’) Blue tits spend up to 85% of their time feeding and searching for food during the winter (Gibb, 1954) and overnight survival depends upon accumulating sufficient reserves during the day. Under these circumstances the purest and simplest handicap a resource holder could impose over food would be to waste time by stopping feeding and do nothing else, a type of War of Attrition developing between the birds to see who gives up first. Given the motivational circumstances, value of the resource, and asymmetries between individuals (discussed below) we feel that this strategy is unlikely to be seen without embellishment in nature. 3.2.
In addition to the above there are also correlated costs which are a consequence of display; they should not be considered handicaps because they are not directly associated with the attribute being signalled. They include drawing attention to oneself and a reduction of awareness of other stimuli [e.g. reduced vigilance leading to predation as seen in pigeon guillimots (Nelson, 1984)]. 4. Resource Value, Motivation and the Structure of Displays In small temperate birds competing in winter, numerous factors appear to influence subjective resource value and hence the willingness to defend or obtain a food resource. These include hunger, resource nutritional value and size, site-dependent rank, the outcome of recent encounters and resource ownership. The latter is an advantage but not a guarantee of success (Scott & Deag, 1998). Under these circumstances the agonistic display system must have a large dynamic component. Signal repetition is an obvious feature of agonistic displays. We see it as the way by which the signaller (resource holder or challenger) strategically increases its handicap to signal its condition, its motivation to continue displaying, and reluctance to give up
the resource or determination to obtain it. In this way it signals the subjective resource value (i.e. the value of the disputed resource to itself, Enquist et al., 1985; Enquist & Leimar, 1987). In addition, display repetition may guard against a receiver’s errors in perception. We see selection acting on receivers who also value a disputed resource, to demand evidence of the signallers handicap and hence selection acting on signallers to signal this. In this way the system will be kept largely honest. Birds signalling beyond their motivation and ability, pay risk and energy costs; their correlated costs may be higher if they have to pay increased attention to their opponent. As noted above, the purest and simplest strategic choice handicap a signaller defending food could impose upon itself, would be to waste time by stopping feeding but do nothing else. However, while this would satisfy the theoretical requirements for a handicap it is an unrealistic expectation in nature. The challenger will be motivated to take possession of the food and the possessor will be motivated to defend it. Ethological studies have shown that under such circumstances, many of the signals or unritualised intention movements given are associated with underlying but incompatible tendencies such as approach, attack, escape, stay put (Tinbergen, 1959; Blurton-Jones, 1968; Hinde, 1970). While unfashionable (behavioural ecology seems obsessed with function), these ideas on the causation of display components have proved remarkably robust and have not been superseded (Baerends, 1975; Hinde, 1970, 1981). They have been largely ignored by theorists working on signalling systems but studies of signal design should take account of signal origin. The appearance of behavioural elements in situations of motivational conflict has provided a rich source of material for selection to shape into signals. We believe that in the evolution of agonistic displays, these elements have been picked up unequally by selection and that this has lead to the discrepancy in the relative frequency of aggressive and submissive elements. Selection is seen to lead to the increased repertoire of signals associated with aggression, as signallers (resource holders or challengers) are forced by sceptical receivers to increase their handicap. They do so by showing elements
159
associated with different levels of their prime motivation: aggression, resource defence and unwillingness to leave. The multiple components can thus be viewed as ‘‘backup signals’’ in the sense of Johnstone (1996, 1997). Some, such as Head Up in the great tit (Wilson, 1992a) may be brought mainly into play if resource value and motivation to defend it are both high; they may therefore give the impression that they are associated with signalling attributes such as territory ownership and prior residence. By contrast, there is no equivalent selection acting on a bird to increase the number of signals associated with escape/submission; simply repeating the same element (e.g. Crest Erect) will often be enough and will minimize ambiguity. Further more, we suggest that during a period of negotiation, an increased repertoire size and signal repetition (a frequent aspect of signal ritualization) permits a signaller to increase its handicap in smaller increments than would otherwise be the case. This idea provides an explanation for how and why different behavioural elements are used to increase the duration of the interaction during a period of negotiation over a resource. There is no need to refer here to the evolutionary processes involved in shaping signal form, except to point out that recent ideas on the influence of receiverpsychology (Guilford & Dawkins, 1991), limitations on honesty (Dawkins & Guilford, 1991; Johnstone & Grafen, 1993), and the consequences of error-prone signalling (Grafen & Johnstone, 1993; Johnstone, 1994), are compatible with our proposals. 5. The Information Content of Signals The predictability of a signaller’s future behaviour (however imperfect), runs counter to theoretical studies which suggest that signallers should conceal their intention in agonistic encounters (Caryl, 1979). We believe that a combination of factors are responsible. As noted above, ethological studies have shown that much agonistic communication takes place within a situation where there is at least the potential for motivational conflict. It is likely therefore that the elements shown may be a mixture of signals and of unritualised intention movements that
160
. . . .
have not yet been selected as signals. As Caryl (1982b) noted the latter tend to be ignored by game theorists modelling agonistic displays. We also believe that the iconic nature of many agonistic signals (e.g. those derived from intention movements or associated with autonomic arousal), and the consequences of this have been underestimated (see for example Maynard Smith & Harper, 1995). In a truly conventional signalling system, signals would be abstract and non-iconic. A signal would have a form unrelated to internal factors (motivation) and immediate costs. Ethologists use the concepts of emancipation for the process by which the proximate control of a signal is presumed to be separated from its ancestral motivation during evolution. It is clear however that emancipation is often missing or incomplete in the behavioural elements of titmice (BlurtonJones, 1968) and as a consequence the iconic nature of some behavioural elements automatically carry information on intention. Are we studying young communication systems (i.e. systems in which there has been insufficient time for the evolution of emancipation), or is selection for emancipation in agonistic signals not as strong as previously assumed? We believe that in a strategic choice handicap system the latter is most likely. The fact that it may be adaptive to clearly signal a willingness to escape once this course of action has been decided (Caryl, 1979), suggests that there is little selection for emancipation of behaviours which signal escape. The need to incorporate new elements as additional handicaps will also act against the assumption that all elements are at an equal stage of emancipation. The problem of predicting a signaller’s future behaviour or predicting interaction outcome is confounded by the same elements being used by both winners and losers. At the causation level this is also attributed to ambivalence during a period of negotiation. However, there is also the problem that the cost of signal probing by receivers may result in a level of cheating (Guilford & Dawkins, 1991). Interactions will also differ in the ratio of the value of the resource to the resource holder and to the challenger; a bird which under other circumstances would win, might find it pertinent to withdraw if a
period of display reveals that the resource has a higher subjective value to its opponent (Enquist & Leimar, 1987; Popp, 1987; Cristol, 1992). Signalling escape may be both submissive (indicates a desire to stay without being threatening) and appeasing (reduces the likelihood of attack before escape). The repetition of escape signals may be a cost of wishing to delay escape. A further possibility which we have not yet considered is that some agonistic displays act as revealing handicaps (Grafen, 1990; Maynard Smith, 1991). In this case the display would be expected to directly reveal some underlying quality of the signaller (e.g. size, condition) which the recipient cannot readily detect in a non-displaying bird. Thus some postures might reveal body size but we suggest that this is unlikely to be important. Wilson (1992a) concluded that in great tits agonistic behaviour was only weakly correlated with size, and body size frequently does not predict rank at a feeding table in great tits and blue tits (Wilson, 1989; Scott, 1994). More plausibly, postures such as Wings Out and Tail Fan could reveal honest information on condition and hence competitive ability. This is not incompatible with the idea that the same elements might be involved in strategic choice handicap displays. 6. Some Predictions One problem with testing hypotheses on the function of display systems is that it is difficult to devise truly competitive hypotheses. For example, the hypotheses proposed to explain why there are so many threat displays (Andersson, 1980) are not mutually exclusive. This makes analysis difficult but it does not invalidate the hypotheses since many aspects of behaviour are subject to more than one selection pressure. A further problem is that the low frequency of attack in blue tits suggests that the risk–benefit approach of Waas (1991) may not be fruitful. With these limitations in mind we suggest the following as examples of predictions the fulfilment of which would support our theory that many signals in blue tit and great tit agonistic displays over food are strategic choice handicaps. Ideally it is desirable to control for
context by observing dyadic displays (not fights since these involve assessment) over food, preferably between same sexed birds who are on neutral ground and have little prior knowledge of each other. We propose that winners should show a greater variety of elements, a greater frequency of each element, and that more elements will be combined with other elements at an instance in time instead of being given on their own. The timing and choice of elements used by one bird should be related to the perceived challenge, for example to the proximity of the opponent and to the choice and frequency of the opponent’s behaviour. It is predicted that the eventual winner will at some point in the interaction take and keep the lead in incrementing the display elements (type of elements and the rate of performance). When the subjective value of the resource to a pair of birds is reversed such that the expected loser becomes the winner, the behaviour of the winner and loser should change to follow the above pattern. Except when an encounter escalates to a direct supplanting approach, attack or fighting, there generally appears to be no simple act which triggers the end of the agonistic encounters of the type described. This suggests that either a bird gives up after an allotted time (linked to subjective resource value and display costs), or either the current or accumulated opponent’s behaviour indicates that the opponent places such a high value of the resource that it would be too costly to continue. The ‘‘scratch pad’’ memory system proposed by Grafen & Johnstone (1993) suggests one way in which to search for display rules, i.e. the link between behavioural performance and interaction outcome. 7. Conclusion Incorporating the handicap principle into explanations of agonistic conventional signalling systems with immediate (rather than past) costs, permits the linking of Tinbergen’s ethological analysis of display causation with a range of modern ideas, suggests explanations for previously unexplained facts and identifies underestimated constraints on display evolution. Modellers of communication systems should
161
clearly pay more attention to these issues; Maynard Smith & Riechert (1984) have made an important start with their conflicting-tendency model. While we have concentrated on ‘‘conventional’’ signals with immediate costs, we acknowledge that in practice mixed signalling systems exist (i.e. systems made up of signals with both past and immediate costs) (Guilford & Dawkins, 1995). For these situations (for example in fulmars competing for food, Enquist et al., 1985), a communication model based on an interaction between conventional signalling and assessment signalling is required.
This paper is dedicated to Aubrey Manning on the occasion of his retirement from the Chair of Natural History at the University of Edinburgh. GWS’s work on Parus caeruleus was supported by a S.E.R.C. Postgraduate Studentship at the University of Edinburgh. We thank an anonymous referee for helpful comments.
REFERENCES A, E. S. & M-G, M. (1995). The cost of threat displays and the stability of deceptive communication. J. theor. Biol. 175, 405–421. A, M. (1980). Why are there so many threat displays? J. theor. Biol. 86, 773–781. B, G. P. (1975). An evaluation of the conflict hypothesis as an explanatory principle for the evolution of displays. In: Function and Evolution in Behaviour (Baerends, G.P., Beer, C. & Manning, A., eds) pp. 187–227. Oxford: Clarendon Press. B-J, N. G. (1968). Observations and experiments on causation of threat displays of the Great Tit (Parus major). Anim. Behav. Monogr. 1, 75–158. C, P. G. (1979). Communication by agonistic displays: what can games theory contribute to ethology? Behaviour 68, 136–169. C, P. G. (1982a). Animal signals: a reply to Hinde. Anim. Behav. 30, 240–244. C, P. G. (1982b). Telling the truth about intentions. J. theor. Biol. 97, 679–689. C, D. A. (1992). Food deprivation influences dominance status in dark-eyed juncos, Junco hyemalis. Anim. Behav. 43, 117–124. D, M. S. (1993). Are there general principles of signal design? Phil. Trans. R. Soc. Lond. B 340, 251–255. D, M. S. & G, T. (1991). The corruption of honest signalling. Anim. Behav. 41, 865–873. E, M. (1985). Communication during aggressive interactions with particular reference to variation in choice of behaviour. Anim. Behav. 33, 1152–1161. E, M. & L, O. (1987). Evolution of fighting behaviour: the effect of variation in resource value. J. theor. Biol. 127, 187–205.
162
. . . .
E, M., P, E. & R¨, J. (1985). Aggressive communication in fulmars (Fulmarus glacialis) competing for food. Anim. Behav. 33, 1007–1020. G, J. (1954). Feeding ecology of tits, with notes on treecreeper and goldcrest. Ibis 96, 513–543. G, A. (1990). Biological signals as handicaps. J. theor. Biol. 144, 517–546. G, A. & J, R. A. (1993). Why we need ESS signalling theory. Phil. Trans. R. Soc. Lond. B 340, 245–250. G, T. & D, M. S. (1991). Receiver psychology and the evolution of animal signals. Anim. Behav. 42, 1–14. G, T. & D, M. S. (1995). What are conventional signals? Anim. Behav. 49, 1689–1695. H, M. D. & N, D. A. (1991). ‘‘Intentional’’ signalling in animal communication. Trends Ecol. Evol. 6, 186–189. H, R. A. (1970). Animal Behaviour 2nd Edn. New York: McGraw-Hill. H, R. A. (1981). Animal signals: ethological and games-theory approaches are not incompatible. Anim. Behav. 29, 535–542. J, R. A. (1994). Honest signalling, perceptual error and the evolution of ‘‘all-or-nothing’’ displays. Proc. R. Soc. Lond. B 256, 169–175. J, R. A. (1996). Multiple displays in animal communication: ‘‘backup signals’’ and ‘‘multiple messages’’. Phil. Trans R. Soc. Lond. B 251, 329–338. J, R. A. (1997). The evolution of animal signals. In: Behavioural Ecology: an Evolutionary Approach, 4th Edn (Krebs, J. R. & Davies, N. B., eds) pp. 155–178. Oxford: Blackwell. J, R. A. & G, A. (1993). Dishonesty and the handicap principle. Anim. Behav. 46, 759–764. M S, J. (1991). Theories of sexual selection. Trends Ecol. Evol. 6, 146–151. M S, J. & H, D. G. C. (1988). The evolution of aggression: can selection generate variability? Phil. Trans. R. Soc. B 319, 557–570. M S, J. & H, D. G. C. (1995). Animal signals: models and terminology. J. theor. Biol. 177, 305–311. M S, J. & R, S. E. (1984). A conflictingtendency model of spider agonistic behaviour: hybrid-
pure population line comparisons. Anim. Behav. 32, 564–578. N, D. A. (1984). Communication of intentions in agonistic contexts by the pigeon guillemot (Cepphus columba). Behaviour 88, 145–189. P, J. W. (1987). Risk and effectiveness in the use of agonistic displays by American goldfinches. Behaviour 103, 141–156. S, G. W. (1994). Factors influencing the resolution of agonistic interactions in a winter population of the Blue Tit (Parus caeruleus). Ph.D. Thesis, University of Edinburgh. S, G. W. & D, J. M. (1998). Blue Tit (Parus caeruleus) agonistic displays: a reappraisal. Behaviour (in press). S, A. W. (1962). Agonistic behaviour among blue tits at a winter feeding station. Behaviour 19, 118–138. T, N. (1959). Comparative studies of the behaviour of gulls (Laridae): a progress report. Behaviour 15, 1–70. W, J. R. (1991). The risks and benefits of signalling aggressive motivation: a study of cavedwelling little blue penguins. Behav. Ecol. Sociobiol. 29, 139–146. W, J. D. (1989). The function of agonistic interaction, social dominance and display in a winter population of the Great Tit Parus major L. Ph.D. Thesis, University of Edinburgh. W, J. D. (1992a). Correlates of agonistic display by great tits Parus major. Behaviour 121, 168–214. W, J. D. (1992b). A re-assessment of the significance of status signalling in wild populations of great tits (Parus major). Anim. Behav. 43, 999–1009. Z, A. (1975). Mate selection–selection for a handicap. J. theor. Biol. 53, 205–214. Z, A. (1977). Reliability in communication systems and the evolution of altruism. In: Evolutionary Ecology (Stonehouse, B. & Perrins, C., eds) pp. 253–259. London: Macmillan. Z, A. (1980). Ritualization and the evolution of movement signals. Behaviour 72, 77–81. Z, A. (1993). The fallacy of conventional signalling. Phil. Trans. R. Soc. Lond. B 340, 227–230. Z, A. & Z, A. (1997). The Handicap Principle. New York: Oxford University Press.