ANIMAL BEHAVIOUR, 2001, 61, 205–216 doi:10.1006/anbe.2000.1558, available online at http://www.idealibrary.com on
Can receiver psychology explain the evolution of aposematism? MICHAEL P. SPEED
Liverpool Hope University College (Received 5 August 1999; initial acceptance 22 September 1999; final acceptance 12 August 2000; MS. number: 6316R)
The evolution of aposematism is difficult to explain because: (1) new aposematic morphs will be relatively rare and thus risk extinction during predator education; and (2) aposematic morphs lack the protection of crypsis, and thus appear to invite attacks. I describe a simple method for evaluating whether rare aposematic morphs may be selectively advantaged by their effects on predator psychologies. Using a simulated virtual predator, I consider the advantages that might accrue to dispersed and aggregated morphs if aposematic prey can cause neophobic avoidance, accelerate avoidance learning and decelerate predator forgetting. Simulations show that aposematism is very hard to explain unless there are particular combinations of ecological and psychological factors. If prey are dispersed throughout a locality then aposematism will be favoured only if (1) there is neophobia, learning effects and forgetting or if (2) there are learning effects and warning signals reduce forgetting rates. However, the best scenario for aposematic advantage involves learning rates, forgetting and neophobia when prey are aggregated. Prey aggregation has two important effects. First, it is a highly effective way to maximize the per capita benefits of the neophobia. Second, after an attack on a single prey the benefits of learnt aversions will be immediately conferred on the surviving members of an aggregation without the diluting effects of forgetting. Aggregation therefore provides good protection against forgetting. The simulations thus provide new insights into the complexities of aposematic protection and suggest some important directions for empirical work.
signals may enhance aposematic survival. There are four major empirical findings. First, warning signals can enhance neophobia or otherwise exploit dietary conservatism in predators (Sille´n-Tullberg 1985; Leimar et al. 1986; Gamberale & Tullberg 1996a, 1998; Marples & Roper 1996; Marples et al. 1998; Lindstro ¨ m et al. 1999; Rowe & Guilford 1999). Second, warning signals can accelerate the rate of predator aversion learning, thereby reducing the costs of predator education (Gittleman & Harvey 1980; Roper & Wistow 1986; Roper & Redston 1987; Roper 1993; Marples et al. 1994; Roper & Marples 1997). Third, the distinctiveness of warning signals may enhance detection by experienced predators and subsequently reduce recognition errors (Guilford 1986, 1990). Finally, warning signals may help to reduce or prevent forgetting of aversive experiences by predators (see e.g. Roper & Redston 1987; Roper & Marples 1997; Yachi & Higashi 1998; Speed 2000). A complementary approach has been to use the theoretical framework of population genetics to evaluate how effectively alleles for warning signals may persist and spread through populations. Mallet & Singer (1987) proposed that an aposematic morph that becomes locally common may spread behind a moving cline and thus through the rest of a prey population.
The initial origin of aposematism within a population of defended cryptic prey presents evolutionary biologists with a famously complex challenge. On the one hand, since novel aposematic forms are by definition conspicuous, they have little or no protection from crypsis. On the other hand, novel aposematic morphs are likely to be very rare and may not survive the high mortality costs incurred during the education of naïve predators (Fisher 1958; Edmunds 1974; Turner 1984; Guilford 1985, 1988, 1990; Mallet & Singer 1987; Yachi & Higashi 1998). Prey bearing new warning signals thus face two major barriers to survival, yet aposematism is, paradoxically, a common state in many defended species. One way to solve this problem is to investigate whether warning signals have ‘special effects’ on the psychologies of predator-receivers (Guilford 1990; Guilford & Dawkins 1991). These special effects might give aposematic prey sufficient short-term gains to offset the costs of rarity and conspicuousness. Over the last two decades increasingly fruitful research from a receiver psychology perspective has demonstrated a set of mechanisms whereby warning Correspondence: M. P. Speed, Biology, School of Science and Social Sciences, Liverpool Hope University College, Hope Park, Childwall, Liverpool L16 9JD, U.K. (email:
[email protected]). 0003–3472/01/010205+12 $35.00/0
2001 The Association for the Study of Animal Behaviour
205
2001 The Association for the Study of Animal Behaviour
206
ANIMAL BEHAVIOUR, 61, 1
There is, however, little work that bridges laboratory studies of the psychology of individual predators with wider population studies (but see Sille´ n-Tullberg & Bryant 1983; Leimar et al. 1986; Yachi & Higashi 1998; Servedio 2000 for notable exceptions). Having demonstrated that warning signals can generate phobias, accelerate learning, aid recognition and enhance memorability it remains to be shown whether these benefits are sufficient to give a rare, conspicuous aposeme superiority over ancestral cryptic forms. Mallet & Singer (1987) used simple mathematical models and verbal argument to evaluate the case for a receiver psychology perspective. They concluded, notably, that neophobia and learning effects are unlikely to have been ‘very powerful aids to the evolution of warning signals’ (Mallet & Singer 1987, page 342: see also Mallet & Joron 1999). In this paper I aim to extend the discussion begun by Mallet & Singer (1987) and introduce a simple method for evaluating the role that receiver psychology effects may have in the evolution of aposematism. Using simple computer simulations, I investigated the effects that warning signals may have on prey survival by affecting learning rates, enhancing neophobia and reducing forgetting by predators. Although computer models of this sort can suffer from excessive simplicity, they have the virtue of making clear and explicit the sort of assumptions that go unexamined in verbal arguments. The simulations do predict that accelerated learning rates (e.g. Gittleman & Harvey 1980; Roper & Redston 1987) and neophobia are not in themselves likely to be sufficient to explain the evolution of aposematism. However, consideration of specific combinations of psychological and ecological factors yields some important insights into aposematic evolution. THE PROBLEM OF THE LONE MUTANT The evolution of aposematism within a defended, cryptic prey population can be thought of in at least two phases. In the first stage, the genesis of aposematism, a lone (presumably dominant) aposematic mutant must survive long enough to reproduce. In the second stage, its more numerous descendants must be able to generate higher levels of fitness than that of the more common cryptic competitor. I shall deal mainly with the second phase of aposematism. However, it is worth briefly considering how big a problem the genesis of aposematism presents. An obvious answer is that a lone aposematic prey is so rare that random genetic drift is likely to be the greatest determinant of whether new alleles for aposematism survive to the second stage of aposematic evolution. However, it is also the case that rarity and predator neophobia (or similar apostatic effects; see Thompson 1984) are likely to interact in favour of an aposematic mutant. First, the aposematic morph is by definition extremely rare and may thus come within the detection distance of a predator only infrequently. Second, the lone aposematic mutant is the only individual of its kind, and it will be the sole recipient of any neophobic protection available. Consider a simplified scenario in which the probability that a lone aposematic morph is detected
before its reproductive phase is 0.5. The aposematic morph is novel and it provokes a neophobic reaction in which the probability of an attack is 0.1. The probability that the individual survives to reproduce is then high: 1(0.50.1)=0.95. The possibility of multiple detection events before reproduction of course decreases survival chances. However, since the mutant is the only beneficiary of neophobic protection, its survival chances may still be high enough to provide a credible explanation for the first stage of aposematic evolution. THE SECOND PHASE OF APOSEMATIC EVOLUTION To consider the more complex second phase, I ran a Turbo Pascal program simulating a simple virtual ecology which contains a predator and two prey types in a given locality. The predator’s behaviours are modelled as variable attack probabilities. I assume that hunger levels (and other ecological variables such as alternative prey) are constant. If detected, novel prey are attacked with a naïve probability of 1, unless they provoke a neophobic response. If a neophobic response is shown then attack probability at the time of encounter is zero. Learning about prey defences reduces attack probability from the naïve value of 1 and over time forgetting gradually returns disturbed attack probabilities back towards 1. The two prey types are morphs of the same defended species and are equally endowed with protective toxins (or equivalent defences). Multiple-trial learning is necessary to generate avoidance (e.g. Gittleman & Harvey 1980; Gittleman et al. 1980). One prey, ‘Apomorph’, has an aposematic signal; the other, ‘Cryptomorph’, is nonaposematic, but is not always cryptic. Apomorph and Cryptomorph are treated as completely dissimilar prey by the predator and hence attacks on them are controlled by two separate attack probabilities (contrast this with Yachi & Higashi 1998).
Learning When a prey is detected, the predator attacks if a randomly generated number (values between and including 0 and 1) is less than or equal to the current attack probability. Attacks are assumed to be successful and to be fatal to the prey. Learning about the defensive properties of the prey is described according to an algorithm familiar to learning theorists (Rescorla & Wagner 1972; see Speed 1993), which generates a negatively accelerated learning curve (as shown in Speed & Turner 1999, page 286) described by this simple operator: P1 =P0 +l(P0)
(1)
Where P0 the attack probability before a learning event and P1 is the attack probability after a learning event. l is a learning rate parameter (values between 0 and 1) and is the asymptotic value towards which P1 moves. In all simulations has a value 0, indicating that the prey are highly defended. The learning rate parameter (l) varies in different simulations.
SPEED: EXPLAINING APOSEMATISM
is always detected if it is within striking distance of a predator.
Forgetting In behavioural terms, forgetting can be defined as the reversal of a learnt behaviour over time. Forgetting is often considered to be a result of changes in perceived context (such as time or surroundings) which make the retrieval of memories less effective (Bouton 1993, 1994). However, by this definition, forgetting will also be seen when animals do retrieve memories but decide to reverse acquired behaviours anyway. Forgetting curves are generally negatively accelerated (see Anderson & Schooler 1991; Wixted & Ebbsen 1991 for discussions), which means that the most rapid losses of memory happen soon after learning. Hence I have used essentially the same equation to model forgetting as learning (equation 1), which produces curves similar in general form to those used by psychologists (see Wixted & Ebbsen 1991) and behavioural ecologists (e.g. Owen & Owen 1984; Turner et al. 1984; Yachi & Higashi 1998). Forgetting does not happen in all simulations, but where it does occur I assume that there is some forgetting about a specific morph after every time interval, except those in which the morph has been attacked. Forgetting is described using the following equation (see Figure 3 in Speed & Turner 1999).
The Foraging Season and Prey Dispersal I consider the emergence of a novel aposematic form during a foraging season in a single locality of a larger ecology. During a foraging season of 300 arbitrary time units the predator travels around this locality and comes within striking distance of 300 prey. Apomorph has either low densities (30 prey or 10% of the population), moderate densities (20%) or high densities (50 and 90%) and Cryptomorph makes up the remaining prey. In the first set of simulations described, prey are distributed in a dispersed manner throughout a locality such that (1) during a single time interval a maximum of one prey comes within striking distance of a predator and (2) prey are evenly distributed in relation to their frequency. Hence, if the dispersed Apomorph has a frequency of 10%, then one individual comes within striking distance of the predator after every 10 time intervals. A total of 10 dispersed prey therefore come within striking distance of the predator during every 100 intervals.
Measures of Survival P3 =P2 +f(1P2)
(2)
where P2 is the attack probability before forgetting, P3 is the attack probability after forgetting and f is the forgetting rate variable. Within a simulation, forgetting rate is constant. I use five values for the forgetting rate variable: f: 0=no forgetting, perfect memory; 0.001; 0.005; 0.01; and 0.05. These values are deliberately conservative, to avoid exaggerating the effects of forgetting in the simulations. For example, the number of time units taken to reverse learning (i.e. to raise attack probability from say 0.01 to 0.9) by forgetting with f =0.001 is 2991 (and similarly 458 time intervals for f =0.005, 231 for f =0.01 and 45 for f =0.05). Even if a time unit represents a third of a foraging day, forgetting rates (especially ^0.01) are conservative and generally similar to or slower than those observed in laboratory animals (e.g. Schreurs 1993).
Crypsis, Aposematism and Prey Detectability Prey detectability is a function of both signal design and limitations of predator attention (MacDougall & Dawkins 1998). Similar levels of detection between an aposematic and a nonaposematic form may be uncommon, but exist if the species in question is often conspicuous in its ecology anyway (e.g. diurnal lepidoptera on the wing, or aggregated prey). In other cases the nonaposematic morph will have lowered detection rates because of camouflage and cryptic behaviour. Here prey detectability is simply quantified as the percentage of encounters at which the predator notices the prey. Cryptomorph’s detectability varies in the simulations from 100% (complete detectability) to 60% (moderate detectability) and 10% (very low detectability). I assume that Apomorph
The evolution and maintenance of aposematism depend on the relative survival rates of aposematic and cryptic forms. To quantify the difference in survival rates I consider that ‘fitness’ of a particular morph can be calculated as the percentage of the total population of that morph that survive the season. In the Results section, I show the survival difference between Apomorph and Cryptomorph, that is, the percentage of the Apomorph population surviving after a defined foraging period, minus the percentage of the Cryptomorph population surviving after the same period. A positive survival difference thus indicates a higher fitness for Apomorph than Cryptomorph individuals. Survival differences are calculated from the means of 3000 replications. Standard errors of these means are very small (in all cases 1 SEM<0.05 prey units) and are not therefore shown in the figures. Simulations are considered in three sets. In the first, I evaluate the effects that warning signals may have on aposematic survival when there is variation in learning rates and in the presence or absence of neophobia and forgetting. In the second set I consider the role that prey distributions could play in situations in which predators can forget, warning signals enhance learning rates, and there may be neophobia. Finally, I consider the benefits that might accrue if warning signals have the effect of lowering forgetting rates. LEARNING EFFECTS AND NEOPHOBIA The assumptions made in the scenarios described in this section are generous towards a new aposematic morph in the second stage of its evolution. In particular I assume that (1) there is only a single predator in a locality and (2) in the simulations described here Apomorph has already
207
ANIMAL BEHAVIOUR, 61, 1
Quick learning
0 –10
0.9 0.7 0.5 0.3
Apomorph disadvantaged Quick learning
0.1
0.9 0.5 0.7 0.3 h p 0.1 omor Crypt g rate (α 1) Slow in learning n r a le
(b)
20
0.9 0.7 0.5
–20 –30 –40 –50
0.3 0.1
% Survival difference
10 0 –10
Variations in learning rates
0.9 0.5 0.7 0.3 h 0.1 omorp Crypt g rate (α 1) in n r lea
(c)
20 10 0 –10
0.9 0.7 0.5
h o r p te om ra Ap ning ) r lea (α l
Figure 1a shows the effect on survival difference between Apomorph and Cryptomorph of variations in the learning rates they inspire. It is expected that the rare Apomorph will not have superior fitness when its learning rates are less than or equal to those for Cryptomorph. This is verified. In Fig. 1a the survival difference values in these cases are negative, indicating a lower proportionate survival for Apomorph. An important result is that the effect of Apomorph’s warning signal on learning rates is not sufficient to offset the problems of rarity, except in two instances. In one of these (l =0.7; Cryptomorph l =0.1), the benefit is marginal. In the other instance the difference between parameters is extreme (Apomorph l =0.9; virtually singletrial learning; Cryptomorph l =0.1, very slow learning rate) and here also the net benefit to Apomorph is small (survival difference is only 5%). In this simulation I assumed that there are only 270 Cryptomorph individuals in the prey population. However, it is possible to calculate how large the Cryptomorph population would have to be to remove Apomorph’s advantage. When Apomorph l =0.9 and Cryptomorph l =0.1 then on average 2.1 Apomorphs and 32 Cryptomorphs were attacked. Apomorph’s ‘fitness’ (i.e. ratio of mean total that survive to total number at the start) is (302.1)/30 (or 93%) whereas Cryptomorph’s ‘fitness’ is (27032)/270 (or 88%). Increasing Cryptomorph’s population from 270 to 450 would equalize the fitnesses (i.e. (45032)/450100=93%). Thus, even if in a real situation the learning rates are as extreme as 0.9 and 0.1,
–20 –30 –40 –50
Apomorph advantaged
% Survival difference
20 10
h orp te om ra Ap ning ) r lea (α l
Results
(a)
h orp te om ra Ap ning ) r lea (α l
reached 10% of the population (Cryptomorph is therefore at 90%). I also assume (3) that the predator is naïve with respect to both Apomorph and Crytomorph. If learning and neophobia are not effective aids to the rare Apomorph under these conditions then they are unlikely to be effective when in less favourable situations, such as smaller Apomorph populations and/or larger Cryptomorph populations. I describe six simulations here. In the first three, there is no difference in level of detectability between Apomorph and Cryptomorph. In the first simulation, the learning rate parameters (l) for both Apomorph and Cryptomorph vary systematically (l =0.9, 0.7, 0.5, 0.3, 0.1). In the second simulation, learning rates vary in the same manner and Apomorph has special protection from neophobia. In modelling neophobia I assume simply that the first 24 aposematic prey encountered are not attacked, but that neophobia vanishes after the 24th prey has been seen and attack probability is then raised to 1. In the third simulation, I assume that learning rates vary, the same level of neophobia exists for Apomorph and that there is a low rate of forgetting (i.e. f =0.005). The remaining simulations are respectively the same as the first three, with the exception that Cryptomorph has a moderate level of crypsis (i.e. is detected on 60% of occasions that it passes before a predator).
–20 –30 –40 –50
0.3 0.1
% Survival difference
208
0.9 0.5 0.7 0.1 0.3 tomorph Cryp g rate (α 1) learnin
Figure 1. The effects of varied predator learning rates on the survival difference between Apomorph and Cryptomorph prey. The horizontal axes show values for the learning rate variable (αl) for Apomorph and Cryptomorph. The vertical axis shows the percentage survival difference between Apomorph and Cryptomorph (i.e. percentage of Apomorph individuals that survive the season minus the percentage of Cryptomorph individuals that survive the season). A positive value (light-shaded bars) implies a higher fitness for Apomorph and a negative value (dark-shaded bars) a lower fitness. In a, b, c, Apomorph is rare with a frequency of 10% and both prey are completely detectable. (a) There is no neophobia or forgetting. (b) There is neophobia (protecting the first 24 Apomorph individuals encountered) but no forgetting. (c) There is neophobia and a moderate rate of forgetting (αl =0.005).
SPEED: EXPLAINING APOSEMATISM
(a)
20
Aposematic survival with learning effects, neophobia and forgetting
0.3 0.1
% Survival difference
0.9 0.5 0.7 0.1 0.3 tomorph Cryp g rate (α 1) learnin
20 0 –10
0.9 0.7 0.5
–20 –30 –40 –50
0.3 0.1
% Survival difference
10
0.9 0.5 0.7 0.1 0.3 tomorph Cryp g rate (α 1) learnin
(c)
20 10 0 –10
0.9 0.7 0.5
h orp te om ra Ap ning ) r lea (α l
Consideration of the results from this first set of simulations makes two important points about the survival of new aposematic forms within a defined locality. First, on their own, the effects that aposematic signals may have in accelerating aversion learning are almost certainly insufficient to give a rare, conspicuous morph superiority over its nonaposematic, cryptic competitor. Second, neophobia is also likely to be of very limited help. Both findings support the initial conclusions of Mallet & Singer (1987). However, in contrast to Mallet & Singer (1987), the results do show clearly that neophobia and learning effects can be effective in combination if predators forget about aversive experiences over time. That neophobia is generally ineffective without forgetting may seem surprising, especially since in the
–20 –30 –40 –50
(b)
h orp te om ra Ap ning ) r lea (α l
Summary and Discussion
0 –10
0.9 0.7 0.5
h orp te om ra Ap ning ) r lea (α l
Figure 1b, shows the results of a simulation in which Apomorph induces neophobia in the predator. The addition of neophobia increases the number of cases in which Apomorph has a survival advantage, although the benefit is limited largely to circumstances in which Cryptomorph has a low learning rate (i.e. l =0.1; there is only one other case, Apomorph l =0.9, Cryptomorph l =0.3). The superiority of the most advantaged Apomorph (l =0.9) would be turned into a disadvantage if the Cryptomorph population size increased in this case to about 560 individuals (i.e. Apomorph survival value= 94%; if there were 560 Cryptomorphs then Cryptomorph’s fitness/survival value=(56032)/560 or about 94%). Since neophobia provides substantial protection (i.e. 24 exposures) it may seem surprising that its addition to the system does not have a dramatic effect on the evolutionary chances of aposematism. However, a dramatic difference is seen when a moderate forgetting rate is introduced. In Fig. 1c the addition of forgetting means that 60% of cases simulated show a net benefit to Apomorph and Apomorph is advantaged even in cases in which the learning rate parameters are equal (e.g. Apomorph l =Cryptomorph l =0.3). Neophobia is therefore much more effective in situations in which predators forget than those in which they do not forget. Figure 2a, b, c shows the results of simulations identical to those shown in Fig. 1a, b, c except that Cryptomorph has moderate protection from crypsis, that is, it is detected on 60% of the times that it passes the predator (Apomorph is always detected). Comparison of Figs 1 and 2 shows, predictably, that moderate crypsis levels in the nonaposematic prey degrades aposematic protection. However, the presence of learning effects, neophobia and forgetting are still sufficient to account for the superiority of a rare aposeme in 10 of the 25 cases shown (Fig. 2c).
10
–20 –30 –40 –50
0.3 0.1
% Survival difference
Apomorph’s advantage will be removed if the population of cryptic morphs contains 180 extra individuals.
0.9 0.5 0.7 0.3 h p r 0.1 omo Crypt g rate (α 1) in n r a le
Figure 2. The effects of varied predator learning rates on the survival difference between Apomorph and Cryptomorph when Cryptomorph has moderate cryptic protection (i.e. it is detected at 60% of encounters). (a) There is no neophobia or forgetting. (b) There is neophobia (protecting the first 24 Apomorph individuals encountered) but no forgetting. (c) There is neophobia and a moderate rate of forgetting (αl =0.005).
simulations neophobia was extensive, protecting the first 24 prey encountered. This can best be understood by consideration of the simulation that does not include neophobia or forgetting (Fig. 1a). In this scenario avoidance learning takes place at the start of the simulation and protection of the Aposematic prey remains
209
ANIMAL BEHAVIOUR, 61, 1
0.05 0.01 Fo 0.005 rge tti 0.001 (α ng ra f) te
70 50 30 10 10 –30 –50
High forgetting
0 10
No forgetting
LEARNING, MEMORY, NEOPHOBIA AND PREY DISTRIBUTIONS In the first set of simulations, I assumed that the new aposematic prey were localized into an area, roughly equivalent to the foraging range of a predator. In the simulations described below I consider that, within the foraging range, prey can be evenly dispersed (as in the simulations above) or tightly aggregated into a dense clump. An aggregation of prey contains 10 individuals, all of which are encountered within a single time interval. Aggregations are themselves evenly distributed so that for a prey with an abundance of 30 individuals (i.e. frequency of 10%), an aggregation is encountered after every 99 intervals. Dispersed and aggregated prey thus represent simplified extremes of prey distributions.
(b)
0.05 0.01 Fo 0.005 rge tti 0.001 (α ng ra f) te
50
90
y enc qu ph Fre omor % Ap of
20
70 50 30 10 10 –30 –50
0 10
(c)
50
90
y enc qu ph Fre omor % Ap of
20
70 50 30 10 10 –30 –50
Forgetting Rates I consider here how variations in forgetting rates can contribute to the survival of aposematic and nonaposematic prey that vary in abundance (10, 20, 50 and 90% of the population), in manner of distribution (dispersed or aggregated) and level of detectability. I assume that learning rates for Apomorph are high (l =0.7) and are low for Cryptomorph (l =0.1; these are roughly equal to learning rates seen in Gittleman & Harvey 1980). Five different forgetting rates are included f: 0=no forgetting, perfect memory; 0.001; 0.005; 0.01; and 0.05) and Cryptomorph has three levels of detectability, high (100%), moderate (60%) and very low (10%). Unless stated neophobia has been excluded from the simulations.
0.05 0.01 Fo 0.005 rge tti 0.001 (α ng ra f) te
0 10
% Survival difference
(a)
% Survival difference
undiminished throughout the rest of the foraging season. Adding neophobia puts off predator education for a period, after which the full costs of predator learning must be paid. Neophobia may thus delay attacks but makes little difference to the overall mortality costs suffered. However, if there is predator forgetting then neophobia can be very effective because it can provide a substantial period of protection to the aposematic form which is not degraded by forgetting. During the same period any learnt protection enjoyed by the nonaposematic form may be degraded by forgetting.
50
% Survival difference
210
90
y enc qu ph Fre omor % Ap of
20
Results
Figure 3. The effects of morph frequency and forgetting rates on the survival difference between Apomorph and Cryptomorph: dispersed prey. In all simulations the predator learning rate parameters (αl) are Apomorph=0.7 (fast learning); Cryptomorph=0.1 (slow learning). Forgetting rates and morph frequencies are varied (X axes) and the survival differences between Apomorph and Cryptomorph are shown on the Y axis. The species disperses evenly throughout the locality. (a) Apomorph and Cryptomorph are equally detectable. (b) Cryptomorph has a detectability of 60%. (c) Cryptomorph has a detectability of 10%.
With no forgetting (Fig. 3a: f: 0,) and in the absence of cryptic protection for Cryptomorph, a rare Apomorph (i.e. 10% of the population) has a very small advantage. However, the addition of forgetting (Fig. 3a: f:=0.001, 0.005, 0.01 and 0.05) removes any superiority that the rare dispersed Apomorph can generate from its effect on learning rates. The inclusion of moderate or high levels of cryptic protection for Cryptomorph enhances Apomorph’s inferiority even further (Fig. 3b, c). Rare dispersed aposematic prey are therefore highly vulnerable to the diluting effects of forgetting and the presence of crypsis in the nonaposematic morph.
Figure 4 shows results from simulations that differ only in that the prey are tightly aggregated rather than dispersed throughout the locality. If predators do not forget then the results for aggregated and dispersed prey are identical (f =0; Figs 3a and 4a). However, if the predator does forget then a new, rare aposeme has greater chances of survival if it emerges in the aggregated species (Fig. 4a; f _0.001). When Apomorph is aggregated (Fig. 4a) a single attack on an aposematic prey generates substantial protection because of high learning rates. This protection
SPEED: EXPLAINING APOSEMATISM
Incorporation of Neophobia
(b)
50
90
y enc qu ph Fre omor % Ap of
70 50 30 10 10 –30 –50
0.05 0.01 rge 0.005 tti 0.001 (α ng ra f) te
Fo
0 10
(c)
0.05 0.01 Fo rge 0.005 tti (α ng ra 0.001 f) te
50
90
y enc qu ph Fre omor % Ap of
20
70 50 30 10 10 –30 –50
0 10
50
90
y enc qu ph Fre omor % Ap of
20
I have incorporated neophobia into the scenario that includes tight aggregation of prey. I assume here that predators have a neophobic reaction to whole prey aggregations, and hence specify neophobia in terms of the number of prey aggregations that are protected. Unless stated, a low level of neophobia is assumed, in which neophobia lasts only for the first two exposures to aposematic aggregations and then immediately wanes (other parameters are as in Fig. 4).
Results
% Survival difference
0 10
20
% Survival difference
0.05 0.01 Fo rge 0.005 tti (α ng ra 0.001 f) te
70 50 30 10 10 –30 –50
% Survival difference
(a)
Figure 4. The effects of morph frequency and forgetting rates on the survival difference between Apomorph and Cryptomorph: aggregated prey. All parameters are the same as Fig. 3, except that the prey aggregates into clumps of 10 prey items. Clumps are either aposematic or nonaposematic and each type is evenly dispersed throughout the environment. (a) Apomorph and Cryptomorph are equally detectable. (b) Cryptomorph has a detectability of 60%. (c) Cryptomorph has a detectability of 10%.
is then conferred on the other individuals in the clump without the diluting effects of forgetting. When Cryptomorph has some protection from crypsis (detectability of 60%, Fig. 4b), aggregation can help to give the rare aposematic morph an advantage over the cryptic morph but only if moderate or fairly high levels of forgetting can be assumed (Fig. 4b, frequency of Apomorph=10%, f =0.01, 0.05). Thus if predators do forget, the benefit from aggregation helps Apomorph to offset some of the problems of rarity (compare Fig. 3a with 4a) and of conspicuousness relative to the nonaposematic form (compare Figs 3b, c and 4b, c).
Even without forgetting (f =0) the combination of neophobia, learning effects and tight aggregation of prey makes Apomorph superior to Cryptomorph (Fig. 5). The size of this protection is sufficiently high to offset the problems of rarity (Apomorph frequency=10%) and moderate levels of crypsis in Cryptomorph (Fig. 5b). Although the fitness advantage of Apomorph generally increases monotonically with its abundance, a small number of instances show a ‘U-shaped’ fitness curve (i.e. lowest fitness point is at intermediate densities; e.g. Fig. 5b, f =0.05 and Fig. 5c). Aposematic prey at intermediate densities may sometimes be the least fit because the benefits from neophobia may be quickly lost whilst the dilution benefits of high densities are not yet reached. Figure 5d, shows the same simulation as Fig. 5c in which there is a high level of crypsis for Cryptomorph. The level of neophobic protection is increased such that the first 10 aggregations that the predator encounters are completely protected. Increasing the extent of neophobia in this manner is sufficient to offset the problems of rarity and high relative conspicuousness for the rare aposematic morph in all cases except one (Fig. 5d, f =0.05 and frequency=90%). Note that in generating the data for Fig. 5, I assume that the predator is naïve at the start of each season, even when the prey is common and no longer novel in its ecology. If this assumption does not hold after the first season, then the data from Fig. 4 (aggregation without neophobia) will pertain for Apomorph frequencies of 20% and higher. In practice this makes little difference to the general qualitative predictions made. Figure 6 shows a resimulation of Fig. 5a in which Apomorph and Cryptomorph are equally detectable but with the learning rate parameters for Apomorph and Cryptomorph set at 0.5. With equalized learning rates, Apomorph loses much of its advantage and in many cases its superiority is converted into inferiority. Comparison of Fig. 6 and Fig. 5a illustrates well how combinations of specific receiver psychology effects and of assumptions about prey distributions are essential in generating a credible explanation of aposematic evolution.
Discussion A major finding of these simulations is that the protection conferred by neophobic responses is tied closely to
211
ANIMAL BEHAVIOUR, 61, 1
(c)
0.05 0.01 0.005 Fo rge tti 0.001 (α ng ra f) te
90
70 50 30 10 10 –30 –50
0 10
50
0 10
(d)
90
y enc qu ph Fre omor % Ap of
20
0.05 0.01 0.005 Fo rge tti 0.001 (α ng ra f) te
70 50 30 10 10 –30 –50
0.05 0.01 0.005 Fo rge tti 0.001 (α ng ra f) te
50
90
y enc qu ph Fre omor % Ap of
20
70 50 30 10 10 –30 –50
0 10
50
% Survival difference
0 10
50
y enc qu ph Fre omor % Ap of
20
% Survival difference
0.05 0.01 0.005 Fo rge tti 0.001 (α ng ra f) te
70 50 30 10 10 –30 –50
% Survival difference
(b)
% Survival difference
(a)
90
y enc qu ph Fre omor % Ap of
20
Figure 5. The effects of morph frequency and forgetting rates on the survival difference between Apomorph and Cryptomorph: aggregated prey with neophobia. All parameters are as in Fig. 4, except that neophobia has been included. In (a–c) there is a low level of neophobia which extends to the first two aggregations encountered. In (d) there is a moderate level of neophobia, which extends to the first 10 aggregations encountered. (a) Apomorph and Cryptomorph are equally detectable. (b) Cryptomorph has a detectability of 60%. (c) Cryptomorph has a detectability of 10%. (d) Cryptomorph has a detectability of 10%; there is a moderate level of neophobia.
0.05 0.01 Fo rge 0.005 tti (α ng ra 0.001 f) te
70 50 30 10 10 –30 –50
% Survival difference
212
90
0 10
50 y enc qu orph e r F % Apom of
20
Figure 6. The effects of morph frequency and forgetting rates on the survival difference between Apomorph and Cryptomorph: aggregated prey with neophobia but no learning effects. All parameters are the same as Fig. 5a, except that the learning rate parameters (αl) for Apomorph and Cryptomorph are equal at 0.5.
the geographical structure of the prey population. If neophobic avoidance persists for a defined number of exposures then avoidance of aposematic morphs may dissipate quickly if the species is dispersed. Consider a naïve predator that confers neophobic protection during its first two encounters with the aposematic species in question. If the species is dispersed then the number of individuals protected would be two. If, however, it gave
protection to the first two aggregations encountered then, assuming 10 prey per aggregation, the number of protected individuals would be 20. The simulations therefore make the important point that aggregation can be a highly effective way to maximize the per capita benefits of neophobic rejection.
WARNING SIGNALS AND FORGETTING RATES In the simulations described so far, I assumed that the presence of a warning signal does not affect forgetting rates. Survival of Apomorph and Cryptomorph were thus compared only with the same forgetting rate (f) values. However, Roper & Redston (1987) and Marples & Roper (1996) have shown that some aspects of a conditioned stimulus, such as conspicuousness, slow down forgetting after single trial learning. I have therefore examined the effect on survival difference under conditions in which the aposematic form generates faster learning and slower forgetting than the cryptic form. To do this I have taken the values for attack rates on Cryptomorph when it is common (i.e. 90% of the population) for varied forgetting rates (f =0.005, 0.01, 0.05). For each I calculated the survival difference from the rare Apomorph (10% of the population) which has a fixed and slower forgetting rate of f =0.001. Calculations were performed and plotted for each of Cryptomorph’s three
SPEED: EXPLAINING APOSEMATISM
(a) 45 % Survival difference
35
(b)
25 15 5 –5 –15 100 % De 60 of tec Ap tab 10 om ili orp ty h
0.05 ph or te om g ra t yp in Cr gett α f) ( o f r
0.01 0.005
45 % Survival difference
35
(c)
25 15 5 –5 –15 100 % De 60 of tec Ap tab 10 om ili orp ty h
0.05 ph or ate m r o t yp ing Cr gett α f) ( for
0.01 0.005
45 % Survival difference
35 25 15 5 –5 –15 100 % De 60 of tec Ap tab 10 om ili orp ty h
0.05 ph or te om g ra t yp in Cr gett α f) ( o f r
0.01 0.005
Figure 7. Survival difference between Apomorph and Cryptomorph when the Apomorph’s warning signal both accelerates avoidance learning and decelerates predator forgetting. In each simulation, Apomorph is rare (10% of the population) and generates a fixed low forgetting rate (αf =0.001). Cryptomorph is very common (at 90% of the population), but generates higher forgetting rates and has varied levels of detectability (see horizontal axes). The survival difference is calculated between Apomorph and Cryptomorph as stated. (a) Dispersed prey; (b) aggregated prey; (c) aggregated prey with low levels of neophobia.
detectability levels (100, 60 and 10%). Survival differences were calculated for situations in which Apomorph generates a higher learning rate than Cryptomorph and (1) both are dispersed, (2) both are aggregated, or (3) both are aggregated and there is neophobia.
Results The lower forgetting rate that the aposematic signal inspires adds considerable protection to Apomorph relative to Cryptomorph (Fig. 7). Importantly, slowed
forgetting can make aposematism profitable where it was formerly unprofitable (comparison of Figs 1 and 2 with Fig. 7a, b). For example when forgetting rates between Apomorph and Cryptomorph are equal, then the rare dispersed Apomorph loses protection (Fig. 1a; f =0.005, 0.01 and 0.05; Apomorph frequency=10%). However, if the aposematic signal slows down forgetting (to f =0.001, Fig. 7a) then aposematism becomes advantageous and profitable. This reversal of survival difference extends (1) to situations in which the dispersed nonaposematic Cryptomorph has some protection from crypsis (i.e. 60% detectability; Fig. 7a) and (2) to some aggregated prey (forgetting rate of 0.05; Fig. 7b). When the nonaposematic morph is very highly cryptic (Fig. 5c; 10% detectability) slowed forgetting is not sufficient to make Apomorph advantaged unless the simulation includes aggregation, at least some neophobia and high forgetting rates for the cryptic individual (Fig. 7c, f =0.05). By different means these results agree with and extend the general predictions of Yachi & Higashi (1998) whose analysis was limited to dispersed prey and excluded neophobia. GENERAL DISCUSSION Accounting for the evolution of aposematism requires that the costs and benefits of a warning signal are identified and that their relative magnitudes can be compared. In this paper I have developed an idea originally outlined by Mallet & Singer (1987), that it is possible to evaluate some of the costs and benefits of aposematism using simple theoretical models. Mallet & Singer (1987 and also Mallet & Joron 1999) concluded that neophobia and learning effects are insufficient to generate a net benefit to rare, conspicuous aposematic morphs. However, my computer simulations indicate that these conclusions may apply to a limited set of situations, specifically to those with dispersed prey species and without any forgetting by predators (Fig. 1a, b). If forgetting is added to the system then neophobia and learning effects may be sufficient to offset the problems of rarity and conspicuousness even in species that disperse throughout a locality (Fig. 1c). Neophobia will be particularly effective in such circumstances because it provides protection that is not degraded by forgetting. Warning signals may also provide a net advantage to rare dispersed aposemes if they decelerate predator forgetting. Thus, forgetting seems to be generally important in many of the scenarios simulated perhaps because learning about the aversive nature of prey creates periods without attacks during which forgetting can begin to control attack rates. In the time since Mallet & Singer’s paper (1987) there has been growing empirical interest in the role that tight aggregations of individual prey may play in the evolution of aposematism. Aggregation has been shown to heighten neophobia and accelerate avoidance learning by predators (Gagliardo & Guilford 1993; Alatalo & Mappes 1996; Gamberale & Tullberg 1996, 1998). However, the simulations show in addition that aggregation may be a particularly effective way that rare aposematic prey maximize per capita benefits from neophobia. Indeed the set of
213
214
ANIMAL BEHAVIOUR, 61, 1
simulations that provides the most compelling case for the evolution of aposematism is the one that includes aggregation and neophobia as well as learning effects (Fig. 5). A final benefit from aggregation is illustrated when predators forget over time. After an attack on a single prey by a naïve or forgetful predator the benefits of learnt aversions will be immediately conferred on the surviving members of an aggregation without the diluting effects of forgetting.
Limitations and Empirical Questions The simulations are not intended to be exhaustive and some important limitations are considered below. Each limitation raises its own important questions about the psychologies of predators and the ecologies of predators and prey in real situations.
Variation in predator numbers It is often assumed in the literature that, since aposematic morphs are rare, increases in the number of predators will be increasingly detrimental to the evolution of aposematism (e.g. Guilford 1985). In fact this is true only if the effect of one predator is in itself detrimental to the relative survival of the aposematic form. If Apomorph has a higher fitness in the presence of a single predator then, within obvious limits of morph abundance, its superiority is enhanced by the presence of multiple predators. Thus, in Fig. 1a, with Apomorph l =0.9, on average 2.1 individuals are attacked and its ‘fitness’ value is (302.1)/30 (or 93%). With Cryptomorph l =0.1, on average 32 individuals are attacked (giving it a fitness of=(27032)/270 (or 88%). The survival difference between Apomorph and Cryptomorph is therefore about +5% in favour of Apomorph. If there are two predators that attack on average 2.1 Apomorphs and 32 Cryptomorphs each, then Apomorph fitness changes to (304.2)/30 (or 86%), and Cryptomorph fitness is (27064)/270=(or 76%). The fitness difference between Apomorph and Cryptomorph is now much larger (a survival difference of 10% in favour of Apomorph). Survival rates for Cryptomorph relative to Apomorph are correspondingly higher with one predator (i.e. 0.88/ 0.93=0.95) than with two (0.76/0.86=0.88). The possible effects of multiple predators may be important in situations in which the superiority of a neophobic individual is otherwise predicted to be small (e.g. Fig. 1c).
Re-encounter between predators and prey Another pertinent question is how often a predator re-encounters individual prey or prey aggregation in its ecology. This is important because neophobic protection is, by definition, a limited resource; hence re-encounters with the same individual will deplete this ‘neophobic resource’. If there is a very high rate of re-encounter then neophobia will wane quickly, removing much of the protection that is crucial to a rare aposematic morph. By excluding the possibility of re-encounter the simulations may to a greater or lesser degree overestimate the importance of neophobia in aposematic evolution (Fig. 5). A good response to this problem would be to carry out field
studies to help evaluate (1) what kinds of animals may aggregate sufficiently well to reap the benefits of clumping and (2) what rate of predator–prey re-encounter is typical within an ecology.
Encounter rates and receiver parameters I have assumed within each simulation that the rate at which a predator encounters a prey does not alter the psychological effects of the prey in question. However, this assumption begs some important (and generally unanswered) empirical questions. Thus for instance we have little idea whether there is an effect of the temporal spacing of experience on the strength of neophobia. However, one could easily establish whether the number of prey protected by neophobia is different if for example: (1) 20 aposematic prey are seen each day or (2) if one aposematic prey is seen each day for 20 days. To model neophobia more accurately we also need to know whether it wanes gradually with repeated presentations or, as modelled here, in an abrupt, discontinuous fashion. Aspects of learning are similarly unknown. Thus there are few data that indicate whether learning and long-term forgetting rates are affected by the temporal proximity of learning events (although Goldstein 1976, page 285 does show some evidence that spacing can lower learning rates). Experiences of events may be too closely or too widely spaced to be informative which might then be reflected in (low) learning rates and possibly (high) forgetting rates. If this were generally the case then the implications for aposematism would be important, suggesting another reason why rare, dispersed aposemes are particularly disadvantaged. Forgetting presents studies of aposematism with perhaps the biggest ‘receiver mystery’ because there are few studies that systematically record forgetting rates in adult predators. Although some authors claim that predator forgetting is unlikely on empirical grounds (e.g. Waldbauer & Sheldon 1971; Waldbauer 1988) the data presented are often anecdotal (e.g. Rothschild 1964). Evidence from the psychology literature (with mammals and birds as subjects and which use a variety of aversive reinforcers) is mixed. In some cases, forgetting of learnt aversions is absent or slight over periods of 6–60 days (e.g. Hendersen 1985; Grieg-Smith 1987; Bouton & Peck 1992). In contrast, other experiments do show forgetting with significant lapses in levels of acquired behaviour occurring between 1–2 days and 3 months (e.g. McAllister & McAllister 1968; Alcock 1970; Kraemer 1984; GriegSmith 1987; Roper & Redston 1987; Schreurs 1993; Roper & Marples 1997). It is also known that ‘distractions’ in the form of environmental enrichment can increase forgetting rates in captive animals (Parsons & Spear 1972). Since most laboratory animals are not housed in enriched environments, estimates of memory retention may underestimate forgetting in wild animals.
Conclusions Can receiver psychology explain the evolution of aposematism? The answer is a qualified, cautious yes;
SPEED: EXPLAINING APOSEMATISM
however, the qualifications are not trivial. On their own, the kinds of phenomena reported in laboratory investigations are almost certain to be insufficient to explain aposematic evolution beyond the genesis of aposematism. Learning effects, neophobia and heightened memorability may each contribute to aposematic survival, but in isolation are very unlikely to assure it. The simulations suggest three scenarios in which receiver effects plus ecological properties of prey may be sufficient to explain the evolution of aposematism within a single locality. First, with dispersed prey aposematic superiority may emerge if there are neophobia, learning effects and some level of predator forgetting. Second, aposemes in dispersed and aggregated populations may succeed if they are able to reduce the normal rate of predator forgetting. Finally, aggregated prey may present the best scenario for aposematism, but only if there are learning effects, some degree of forgetting and, crucially, if neophobic protection is concentrated on to all members of an aposematic aggregation. Acknowledgments I thank R. V. Alatalo, J. Mappes, L. Lindstrom and A. Latinian for their hospitality during the Aposematism: past, present and future meeting; G. D. Ruxton for discussions; O. Leimar and an anonymous referee for helpful advice and the Arts & Sciences Deanery of Liverpool Hope University College for a Sheppard Worlock Fellowship which supported the work. References Alatalo, R. V. & Mappes, J. 1996. Tracking the evolution of warning signals. Nature, 382, 708–710. Alcock, J. 1970. Punishment levels and the response of black-capped chickadees (Parus atricapillus) to three kinds of artificial seeds. Animal Behaviour, 18, 592–599. Anderson, J. R. & Schooler, L. J. 1991. Reflections of the environment in memory. Psychological Science, 2, 396–408. Bouton, M. E. 1993. Context, time and memory retrieval in the inteference paradigms of Pavlovian learning. Pychological Bulletin, 114, 80–99. Bouton, M. E. 1994. Conditioning, remembering and forgetting. Journal of Experimental Psychology: Animal Behaviour Processes, 20, 219–231. Bouton, M. E. & Peck, C. A. 1992. Spontaneous recovery in cross-motivational transfer (counterconditioning). Animal Learning and Behavior, 20, 313–321. Edmunds, M. 1974. Defence in Animals. Harlow: Longman. Fisher, R. A. 1958. The Genetical Theory of Natural Selection. 2nd edn. New York: Dover. Gagliardo, A. & Guilford, T. 1993. Why do warning-coloured prey live gregariously? Proceedings of the Royal Society of London, Series B, 251, 69–74. Gamberale, G. & Tullberg, B. S. 1996. Evidence for a more effective signal in aggregated aposematic prey. Animal Behaviour, 52, 597–601. Gamberale, G. & Tullberg, B. S. 1998. Aposematism and gregariousness: the combined effects of group size and coloration on signal repellence. Proceedings of the Royal Society of London, Series B, 265, 889–894. Gittleman, J. L. & Harvey, P. H. 1980. Why are distasteful prey not cryptic? Nature, 286, 149–150.
Gittleman, J. L., Harvey, P. H. & Greenwood, PJ. 1980. The evolution of conspicuous coloration: some experiments in bad taste. Animal Behaviour, 28, 879–899. Goldstein, M. L. 1976. Acquired drive strength as a joint function of UCS intensity and number of CS-UCS pairings: the effect of massed and distributed CS-US pairings. Journal of General Psychology, 95, 257–286. Grieg-Smith, P. W. 1987. Aversions of starlings and sparrows to unfamiliar, unexpected or unusual flavors and colours in food. Ethology, 74, 155–163. Guilford, T. 1985. Is kin selection involved in the evolution of warning coloration? Oikos, 45, 31–36. Guilford, T. 1986. How do ‘warning colours’ work? Conspicuousness may reduce recognition errors in experienced predators. Animal Behaviour, 34, 286–288. Guilford, T. 1988. The evolution of conspicuous coloration. American Naturalist, 131(Supplement), S7–S21. Guilford, T. 1990. The evolution of aposematism. In: Insect Defenses: Adaptive Mechanisms and Strategies of Prey and Predators (Ed. by D. L. Evans & J. O. Schmidt), pp. 23–61. New York: State University of New York Press. Guilford, T. & Dawkins, M. S. 1991. Receiver psychology and the evolution of animal signals. Animal Behaviour, 42, 1–14. Hendersen, R. W. 1985. Fearful memories: the motivational significance of forgetting. In: Affect, Conditioning and Cognition: Essays On the Determinants of Behavior (Ed. by F. R. Brush & J. B. Overmier), pp. 43–54. Hillsdale, New Jersey: L. Erlbaum. Kraemer, P. J. 1984. Forgetting of visual discriminations by pigeons. Journal of Experimental Psychology: Animal Behavior Processes, 10, 530–542. Lindstro ¨ m, L., Alatalo, R. V. & Mappes, J. 1999. Reactions of handreared and wild-caught predators toward warningly colored, gregarious, and conspicuous prey. Behavioral Ecology, 10, 317–322. Leimar, O., Enquist, M. & Sille´n-Tullberg, B. 1986. Evolutionary stability of aposematic coloration and prey unprofitability: a theoretical analysis. American Naturalist, 128, 469–490. McAllister, D. E. & McAllister, W. R. 1968. Forgetting of acquired fear. Journal of Comparative and Physiological Psychology, 65, 352–355. MacDougall, A. & Dawkins, M. S. 1998. Predator discrimination error and the benefits of Mu¨ llerian mimicry. Animal Behaviour, 55, 1281–1288. Mallet, J. & Joron, M. 1999. Evolution and diversity in warning colour and mimicry: polymorphisms, shifting balance and speciation. Annual Review of Ecology and Systematics, 30, 201–233. Mallet, J. & Singer, M. C. 1987. Individual selection, kin selection and the shifting balance in the evolution of warning colours: the evidence from butterflies. Biological Journal of the Linnean Society, 32, 337–350. Marples, N. M. & Roper, T. J. 1996. Effects of novel colour and smell on the response of naive chicks towards food and water. Animal Behaviour, 51, 1417–1424. Marples, N. M., van Veelen, W. & Brakefield, P. M. 1994. The relative importance of colour, taste and smell in the protection of an aposematic insect Coccinella septempunctata. Animal Behaviour, 48, 967–974. Marples, N. M., Roper, T. J. & Harper, D. G. C. 1998. Responses of wild birds to novel prey: evidence of dietary conservatism. Oikos, 83, 161–165. Owen, R. E. & Owen, A. R. G. 1984. Mathematical paradigms for mimicry: recurrent sampling. Journal of Theoretical Biology, 109, 217–247. Parsons, P. J. & Spear, N. E. 1972. Long-term retention of avoidance learning by immature and adult rats as a function of environmental enrichment. Journal of Comparative and Physiological Psychology, 80, 297–303. Rescorla, R. A. & Wagner, A. R. 1972. A theory of Pavlovian conditioning: variations in the effectiveness of reinforcement and
215
216
ANIMAL BEHAVIOUR, 61, 1
nonreinforcement. In: Classical Conditioning II: Current Research and Theory (Ed. by A. H. Black & W. F. Prokasy), pp. 64–99. New York: Appleton-Century-Crofts. Roper, T. J. 1993. Effects of novelty on taste-avoidance learning in chicks. Behaviour, 125, 265–281. Roper, T. J. & Marples, N. M. 1997. Odour and colour as cues for taste-avoidance learning in domestic chicks. Animal Behaviour, 53, 1241–1250. Roper, T. J. & Redston, S. 1987. Conspicuousness of distasteful prey affects the strength and durability of one-trial avoidance learning. Animal Behaviour, 35, 739–747. Roper, T. J. & Wistow, R. 1986. Aposematic coloration and avoidance learning in chicks. Quarterly Journal of Experimental Psychology, 38B, 141–149. Rothschild, M. 1964. An extension of Dr. Lincoln Brower’s theory on bird predation and food specificity, together with some observations on bird memory in relation to aposematic colour patterns. Entomologist, 97, 73–78. Rowe, C. & Guilford, T. 1996. Hidden colour aversions in domestic chicks triggered by pyrazine odours of insect warning displays. Nature, 383, 520–522. Rowe, C. & Guilford, T. 1999. Novelty effects in a multimodal warning signal. Animal Behaviour, 57, 341–346. Schreurs, B. G. 1993. Long-term memory and extinction of the classically conditioned rabbit nictitating membrane response. Learning and Motivation, 24, 293–302. Servedio, M. R. 2000. The effects of predator learning, forgetting, and recognition errors on the evolution of warning coloration. Evolution, 54, 751–763. Speed, M. P. 1993. Muellerian mimicry and the psychology of predation. Animal Behaviour, 45, 571–580.
Speed, M. P. 2000. Warning signals, receiver psychology and predator memory. Animal Behaviour, 60, 269–278. Speed, M. P. & Turner, J. R. G. 1999. Learning and memory in mimicry. II. Do we understand the mimicry spectrum? Biological Journal of the Linnaean Society, 67, 281–312. Sille´ n-Tullberg, B. 1985. Higher survival of an aposematic than of a cryptic form of distasteful bug. Oecologia, 67, 411–415. Sille´ n-Tullberg, B. & Bryant, E. H. 1983. The evolution of aposematic coloration in distasteful prey: an individual selection model. Evolution, 37, 993–1000. Thompson, V. 1984. Polymorphism under apostatic and aposematic selection. Heredity, 53, 677–686. Turner, J. R. G. 1984. Mimicry: the palatability spectrum and its consequences. In: The Biology of Butterflies. (Symposia of the Royal Entomological Society of London 11) (Ed. by R. I. VaneWright & P. R. Ackery), pp. 141–161. London: Academic Press. Turner, J. R. G., Kearney, E. P. & Exton, L. S. 1984. Mimicry and the Monte Carlo predator: the palatability spectrum and the origins of mimicry. Biological Journal of the Linnaean Society, 23, 247–268. Waldbauer, G. P. 1988. Asynchrony between Batesian mimics and their models. American Naturalist, 131(Supplement), S103– S121. Waldbauer, G. P. & Sheldon, J. K. 1971. Phenological relationships of some aculeate Hymenoptera, their dipteran mimics, and insectivorous birds. Evolution, 25, 371–382. Wixted, J. T. & Ebbsen, E. B. 1991. On the form of forgetting. Psychological Science, 2, 409–415. Yachi, S. & Higashi, M. 1998. The evolution of warning signals. Nature, 394, 882–884.
