Commentary

Mimicry: An interface between psychology and evolution

James Mallet*

Department of Biology, Galton Laboratory, University College London, 4 Stephenson Way, London NW1 2HE, United Kingdom

very now and then, a new technique Bates also noticed that some Heliconius species. Even though Wallace was proba- Eallows major breakthroughs to be (in the ‘‘Acraeoid Heliconidae,’’ now bly correct about the ultimate benefits of made on a hitherto intractable problem. Nymphalidae: Heliconiinae) also mim- warning color to the population, warning In this issue of PNAS, Leena Lindstro¨m icked Ithomiinae, even though they clearly signals could not simply evolve from cryp- and colleagues (1) have shed light on had the aromatic smells of distasteful spe- sis, unless rare aposematic morphs within long-standing controversies in warning cies, and even though other Heliconius species could overcome the disadvantage color and mimicry. They have achieved were themselves the objects of mimicry of rarity that causes Mu¨llerian mimicry this through experiments on the behavior (3). Essentially, Bates discovered an early (6). A variety of hypotheses were pro- of wild-caught birds (great tits, Parus ma- version of Mu¨llerian mimicry, or mimicry posed in which grouping, particularly of jor), using their ingenious ‘‘novel world’’ between pairs of distasteful species, al- close kin (6, 7), would allow the local setup, a modification though he attrib- frequency to rise above a frequency- of that originally intro- uted this mainly to dependent selection threshold. Others re- duced by Rauno Ala- parallel adaptation garded individual selection as more likely, talo and Johanna Leena Lindstro¨m and colleagues by both species to and proposed that the threshold fre- Mappes in 1996 (2). abiotic conditions. have shed light on long-standing quency could be overcome if attacks were The understanding In 1879, Fritz Mu¨l- of mimicry seems to controversies in warning color ler (4) used a math- not fatal and if conspicuous prey induced predator neophobia, as well as being more me to have developed and mimicry. ematical argument, in a particularly jumpy perhaps the first in memorable (8). manner. Mimicry is ar- evolutionary ecol- This debate about warning color is well guably the oldest Dar- ogy (excepting known to most evolutionists, but a recent winian theory not attributable to Darwin. Malthus’, of course), to show that pairs of surge of interest in warning color shared After reading the recently published On unpalatable species gained mutualistic between unpalatable species (i.e., Mu¨lle- the Origin of Species, Henry Walter Bates benefits from resemblance. By sporting rian mimicry) is perhaps less well known. began to realize that unrelated Amazo- the same color pattern, individuals lost Mu¨llerian mimicry has often been under nian butterflies not only resembled each during predator education would be attack over the years, but belief in the other in any locality, but also changed shared between the two species so that the theory on the whole has prevailed. How- color patterns from place to place, while mortality per capita was lower in each. ever, the new debate seems to cast the preserving the resemblance. This ‘‘mimic- Meanwhile, and rather illogically, Wal- most serious doubt yet on the Mu¨llerian ry’’ could not be due to phylogenetic lace’s adaptive hypothesis of warning hypothesis, and centers around the work inertia; it must be adaptive. (As an aside, color (later called aposematism) was de- of Mike Speed. Speed argues that mimicry the term ‘‘mimicry’’ was coopted from veloped only after mimicry was used to theory has ignored Pavlovian models of pre-Darwinian days. Analogies were pre- explained resemblance between warningly predator psychology. In Speed’s theory, sumed to be the ‘‘osculations’’ that con- colored species. Darwin in 1867 was de- mildly unpalatable species may, by mim- nected Quinarian rings of created taxa veloping his theory of sexual selection to icking highly unpalatable species, increase into a coherent whole. In his Wisdom, explain bright colors in animals, and had the overall attack rate on the model so that God was thought to have arranged organ- come across red-and-black-striped sphinx Mu¨llerian mimicry may be parasitic, or isms in rings of five species connected by homologies, and linked to other rings by caterpillars that, as larvae, would not have ‘‘quasi-Batesian’’ (9), in species of unequal these analogies. Perhaps it is no accident been influenced by sexual selection. Wal- unpalatability. Speed is a former student that we still use the peculiar term ‘‘mim- lace argued that bright colors were adap- of John R. G. Turner, who is well known icry rings,’’ to mean groups of species tive because they warned predators not to for his work on mimicry and Heliconius that are connected by analogous resem- attack distasteful prey (5). A few details butterflies. Turner himself has frequently blances.) Bates then proposed his now were left to be tidied up, but these basic rebutted earlier critiques of classical Mu¨l- famous evolutionary explanation (3), that ideas about mimicry and warning color lerian mimicry (10), but is now convinced nonpoisonous ‘‘Leptalidae’’ (now Pieri- were thereafter accepted, at least by Dar- by Speed’s arguments that parasitic mim- dae: Dismorphiinae) adaptively resemble winists, until the late 20th Century. icry may occur between pairs of unpalat- unpalatable species, models, in the ‘‘Dan- In the 1950’s, the first experimental able species. To Speed and Turner, the aoid Heliconidae’’ (now Nymphalidae: tests of predator psychology (as opposed new form of mimicry can explain anoma- Ithomiinae). If sampled by a predator, the to mere observations of natural behavior) lous phenomena in Mu¨llerian mimicry, palatable mimics will discourage future relevant to mimicry were done; at the attacks, and may enhance attacks on the same time, a reaction against group selec- model; thus, Batesian mimicry can be tion led to a questioning of these and other See companion article on page 9181. parasitic. arguments involving overall benefit to *E-mail: [email protected].

8928–8930 ͉ PNAS ͉ July 31, 2001 ͉ vol. 98 ͉ no. 16 www.pnas.org͞cgi͞doi͞10.1073͞pnas.171326298 Downloaded by guest on September 27, 2021 such as polymorphisms in apparently un- able prey attacked at any prey density. birds were used here). (iii) Black crosses and palatable species (11). Two species that differ in unpalatability squares are completely novel to the birds, This new controversy centers on a be- would then differ in their asymptote. If which distinguish symbols easily but show no guilingly simple assumption made by Mu¨l- they became indistinguishable mimics, the inherent bias (19). (iv) Finding and unwrap- ler. He supposed that predators require a combined asymptote would become an ping the paper packets is sufficiently tricky certain number (n) of unpalatable indi- average of the attack rates on each species. to provide a handling cost analogous to viduals during a season to learn their color The more palatable species would benefit making a mistake and attacking an unpal- pattern. Mu¨ller (4) showed that this re- from mimicry because the asymptote is atable insect prey. (v) The foraging task is quirement always led to a mutualistic ad- lower than if it were a nonmimic, whereas complex: birds must search a 57-m2 arena vantage, even though the less palatable containing around 27,000 cryptic prey sym- the advantages were species would suf- bols, on which only 200 palatable and un- lopsided, with the rela- fer greater per If a species is highly protected palatable almond packets are placed. tive fitness gains (g1, capita mortality. Around 130 symbols must be scanned by g2) due to mimicry to either by unpalatability, because Hence, the mimicry each bird for every prey actually taken. (vi) each species being is parasitic, or quasi- the numbers lost during predator The unpalatable prey are presented at three given by the inverse Batesian (9, 11). separate densities, both as singletons, and square of the relative learning are low, or because of It seems as (vii) in ‘‘gregarious’’ clumps of four. abundances (a , a )of though it should be 1 2 great abundance, then it will As expected from Mu¨ller’s theory, the the two species (i.e., simple to design ex- ͞ ϭ 2͞ 2 benefit very little from mimicry. experiments show that neophobia does not g1 g2 a2 a1 ). Mu¨l- periments to test ler’s relative benefits these ideas about give rare unpalatable morphs an advantage. were soon shown to be predator learning. [An earlier ‘‘novel world’’ experiment that an approximation (12); the correct formu- However, most experiments on the psy- might be taken as support for neophobia lation shows that the benefits are even chology of learning use highly standard- (19) is explicable if birds merely ignored more lopsided than in Mu¨ller’s prediction: ized tests, and rarely assay varying densi- novel packets while foraging for cryptic ͞ ϭ 2 Ϫ ͞ ͞ 2 Ϫ ͞ g1 g2 a2 (1 n a2) a1 (1 n a1). ties of items to be memorized. Standard packets they had already learned were pal- Finally, the same number eaten (n) results psychological theories of learning do not atable.] But the results also appear to dis- only if the two species are equally unpal- provide the understanding we need in the prove the simplistic Mu¨llerian hypothesis of atable. With unequal palatabilities, there evolution of mimicry. Impressive field ex- learning induced by a fixed number of prey will be different numbers of each species periments, in contrast, can show strong in at least two ways. (i) Clumped aposematic (n1, n2) lost in the absence of mimicry, and selection for Mu¨llerian mimicry (15), but prey are avoided much more strongly than the relative gains for mimicry (12, 13) in do not distinguish between different mod- solitary prey presented at the same density, ͞ ϭ 2 2 Ϫ the two species will be: g1 g2 a2 n1 (1 els of learning. Potentially more relevant by a factor of about two. It is possible that ͞ ͞ 2 2 Ϫ ͞ n2 a2) a1 n2 (1 n1 a1). A way of under- experiments on the learning components proximity of the same pattern enhances standing this is that fitness benefits de- of mimicry have been carried out on learning, or perhaps the birds generalize ͞ COMMENTARY pend on the ‘‘protectedness’’ (ai ni)or chicks and wild birds, using pastry baits during learning most readily to nearby ex- ‘‘dose’’ provided by each species. If a made unpalatable with quinine. These ex- amples of the same pattern. In either case, species is highly protected either by un- periments have produced mixed results: novel warning patterns are less strongly at- palatability, because the numbers lost dur- some suggest a selective advantage for tacked in gregarious than in solitary prey, ing predator learning (ni) are low, or rarity (16) or quasi-Batesian mimicry (17), and the simplistic Mu¨llerian dose effect is because of great abundance (ai), then it whereas others demonstrate frequency disproved. (ii) The per capita attack rate on will benefit very little from mimicry. On dependent selection against rare unpalat- aposematic prey decreases with increasing the other hand, a rare or more mildly able forms, as expected under Mu¨ller’s density, as is expected under Mu¨llerian the- unpalatable species will benefit much theory (18). All could be criticized on the ory, but the total number attacked goes up, more strongly from mimicry with a highly grounds that high densities of pastry baits which is not. However, Mu¨ller’s theory is not protected species, by more than the square are too unrealistic as prey, so that abnor- as convincingly disproved here because of the ratio of protectedness. But the mal predator behavior results. Also, these learning is almost certainly not reaching benefits of mimicry to both unpalatable experiments never tested the predators on completion after 2 days of trials (18): the species are still always positive and mutu- more than a pair of prey densities. birds had each eaten about 17 unpalatable alistic. Mu¨ller’s theory is useful for inter- The new work (1) is much more realis- solitary prey after two trials at the highest preting not only the ultimate benefits of tic. In this ‘‘novel world’’ experiment, wild aposematic prey densities (13% of the to- Mu¨llerian mimicry, but also for modeling great tits are brought into captivity and tal), but only about 4 at the lowest densities warning color or mimicry at intermediate trained to feed on novel prey. Each prey stages of evolution. The form of frequen- item is in fact a 1-cm2 white paper enve- (25% of the total). cy-dependent selection should be highly lope containing a tiny, but evidently tasty These experiments were performed nonlinear, hyperbolic in fact (14). 8-mg almond fragment, which can be only with highly unpalatable prey, and are However, this entire body of warning made unpalatable with quinine. Each not direct tests of Speed’s unconventional color and mimicry theory is built up from packet bears a symbol: black crosses, mimicry theory (9, 11). However, a mono- Mu¨ller’s original assumption that preda- which are cryptic against similar symbols tonic decline in attack fraction with in- tors learn by taking a fixed number ni of a on the floor of the arena, or black squares, creasing experience and unpalatable prey given unpalatable prey type, whatever its which stand out visually and are used as density, rather than an asymptotic attack density. This might well be an oversimpli- aposematic signals. The experiments have fraction, seems evident from the data: on fication, and Speed and others have sug- a number of advantages over previous the second trial day each bird attacked less gested that more realistic Pavlovian mod- work. (i) The birds are naturally inquisi- than three of the unpalatable prey at high els will give quite different results. For tive insectivores, and learn difficult tasks density (about 5%, compared with 22% instance, learning and forgetting might in captivity. (ii) The birds are caught in the on the first trial), and less than a single instead lead to an asymptotic balance con- wild, where they are abundant enough for unpalatable prey at the lowest density sisting of a constant fraction of unpalat- the experiments to be well replicated (84 (about 6%, compared with 50% on the

Mallet PNAS ͉ July 31, 2001 ͉ vol. 98 ͉ no. 16 ͉ 8929 Downloaded by guest on September 27, 2021 first trial). They are therefore quick learn- periment with wild birds and pastry baits Will we ever have a complete, unas- ers! Classical Mu¨llerian learning with an (18) also showed higher attack fractions at sailable theory of warning color and approximately zero attack fraction after a lower densities. Therefore, I believe that Mu¨llerian mimicry? I don’t know, but I whole season would seem to be nearer to simplistic number-dependent selection believe the ingenious ‘‘novel world’’ ex- the mark than ‘‘Pavlovian’’ learning with a will ultimately be disproved; however, periments (1, 2, 19) are the best yet nonzero attack fraction asymptote, even near-zero attack asymptotes can only lead designed to investigate the psychological though the numbers required for learning to conventional, mutualistic Mu¨llerian interface of this unsolved evolutionary may increase with density. An earlier ex- mimicry. problem.

1. Lindstro¨m, L., Alatalo, R. V., Lyytinen, A. & 7. Guilford, T. (1990) in Insect Defenses. Adaptive 13. Mallet, J. (2001) Evol. Ecol. 13, 777–806. Mappes, J. (2001) Proc. Natl. Acad. Sci. USA 98, Mechanisms and Strategies of Prey and Predators, 14. Mallet, J. & Joron, M. (1999) Annu. Rev. Ecol. 9181–9184. (First Published July 17, 2001; eds. Evans, D. L. & Schmidt, J. O. (State Univ. of Syst. 30, 201–233. 10.1073͞pnas.161071598) New York Press, New York), pp. 23–61. 15. Kapan, D. (2001) Nature (London) 409, 2. Alatalo, R. V. & Mappes, J. (1996) Nature (Lon- 8. Ja¨rvi, T., Sille´n-Tullberg, B. & Wiklund, C. (1981) 338–340. don) 382, 708–710. Oikos 37, 393–395. 16. Greenwood, J. J. D., Wood, E. M. & Batchelor, S. 3. Bates, H. W. (1862) Trans. Linn. Soc. London 23, 9. Speed, M. P. (1993) Anim. Behav. 45, 571–580. (1981) Heredity 47, 27–34. 495–566. 10. Turner, J. R. G. (1984) in Symposia of the Royal 17. Speed, M. P., Alderson, N. J., Hardman, C. & 4. Mu¨ller, F. (1879) Trans. Entomol. Soc. London Entomological Society of London 11, The Biology of Ruxton, G. D. (2000) Proc. R. Soc. London Ser. B 1879, xx–xxix. Butterflies, eds. Vane-Wright, R. I. & Ackery, P. R. 267, 725–731. 5. Fleischer, R. C., Rothstein, S. I. & Miller, L. S. (Academic, London), pp. 141–161. 18. Greenwood, J. J. D., Cotton, P. A. & Wilson, (1991) Condor 93, 185–189. 11. Speed, M. P. & Turner, J. R. G. (1999) Biol. J. D. M. (1989) Biol. J. Linn. Soc. 36, 213–226. 6. Harvey, P. H. & Greenwood, P. J. (1978) Linn. Soc. 67, 281–312. 19. Lindstro¨m, L., Alatalo, R. V., Lyytinen, A. & in Behavioural Ecology, eds. Krebs, J. R. & Davies, 12. Blakiston, T. & Alexander, T. (1884) Nature (Lon- Mappes, J. (2001) Proc. R. Soc. London Ser. B 268, N. B. (Blackwell Scientific, Oxford), pp. 129–151. don) 29, 405–406. 357–361.

8930 ͉ www.pnas.org͞cgi͞doi͞10.1073͞pnas.171326298 Mallet Downloaded by guest on September 27, 2021