Although it has always been assumed that chemical mimicry and camou£age play a major role in the penetration of ant societies by social parasites, this paper provides the ¢rst direct evidence for such a mechanism between the larvae of the parasitic butter£y Maculinea rebeli and its ant host Myrmica schencki. In the wild, freshly moulted fourth-instar caterpillars, which have no previous contact with ants, appear to be recognized as ant larvae by foraging Myrmica workers, which return them to their nest brood chambers. Three hypotheses concerning the mechanism controlling this behaviour were tested: (i) the caterpillars produce surface chemicals that allow them to be treated as ant larvae; (ii) mimetic compounds would include hydrocarbons similar to those employed by Myrmica to recognize conspeci¢cs and brood; and (iii) the caterpillars' secretions would more closely mimic the pro¢le of their main host in the wild, M. schencki, than that of other species of Myrmica. Results of behavioural bioassays and chemical analyses con¢rmed all three hypotheses, and explained the high degree of host speci¢city found in this type of highly specialized myrmecophile. Furthermore, although caterpillars biosynthesized many of the recognition pheromones of their host species (chemical mimicry), they later acquired additional hydrocarbons within the ant nest (chemical camou£age), making them near-perfect mimics of their individual host colony's odour.

Ecological studies have been made of all 5 European species of Maculinea. These confirm that M. nausithous and M. rebeli live underground in Myrmica ant nests for 10 months of the year, as has long been known for the other 3 species. The main discovery was that each Maculinea species depends on a single, and different, host species of Myrmica. This specificity contradicts previous papers and scientific reviews of the relationship between Maculinea and ants. Therefore, early records are re-examined and 3 reasons are given to explain why most are misleading when applied to wild populations. Dependence on a single, rather than any, species of Myrmica explains why Maculinea populations exist in only a small minority of biotopes where their foodplants and Myrmica ants abound. It also explains the puzzling disappearance of Maculinea populations from apparently suitable sites. The discovery that M. alcon and M. rebeli depend on separate species of Myrmica that are not even closely related strengthens the argument that these butterflies are good species.

The behavioural interactions between caterpillars of Maculinea rebeli Hir. and their Myrmica ant hosts were studied, both in the wild at the time of adoption, and inside captive nests of six Myrmica species. In the wild, freshly moulted, final instar caterpillars left their food‐plants at a time of day that coincided with the peak foraging activity of Myrmica (18:00‐20:00 h). Once on the ground, caterpillars made no attempt to search for Myrmica but settled and waited for foraging ants to find them, which took up to 1.5 h. There was no adoption ritual: foragers of any Myrmica species picked up the caterpillars within 1–4 sec of discovery, and carried them directly to their nests. Caterpillars grew from < 2 mg to 110 mg in laboratory ant‐nests. About 60 mg was gained in autumn but 40% of this was lost during the winter, while the temperature was < 14 °C. Although caterpillars survived best with their normal host, Myrmica schencki, they could also survive in the nests of other Myrmica species. The presence of queen ants had no effect upon survival. The behaviour of the caterpillars was described and illustrated: this included the production of secretions that were drunk by the ants, begging for food and direct feeding by ants. The preferred solid food was ant eggs. The results are discussed in terms of the social biology of Myrmica ants. It is hypothesized that Maculinea rebeli caterpillars mimic the touch pheromones of ant worker‐larvae. This would explain the inability of ants to recognize caterpillars before touching them, their immediate adoption by any Myrmica species after discovery, host specificity inside wild ant‐nests, the absence of queen‐effect and the intimate attention of host workers.

This paper discusses the different social parasites of ants, evolutionary routes to inquilism and other specialized forms of social parasitism, and host specificity in social parasites. Case histories of host specificity and the degree of integration in 3 social parasite systems, i.e. Maculinea arion and Maculinea teleius (high virulence, low transmission predators of ant brood, which enter host colonies in their final larval instar and mainly inhabit the outer nest chambers), Maculinea rebeli and Maculinea alcon (high virulence, low transmission cuckoo feeders in ant brood chambers, which enter host colonies in their last larval instar), Microdon mutabilis and Microdon myrmicae (high infestation, low transmission predators in ant brood chambers, which enter host colonies as eggs), and the testing and calibration of 2 host specificity hypotheses are presented.

Caterpillars of Maculinea rebeli have two growth strategies for living underground as social parasites of Myrmica ant colonies. Laboratory experiments and ¢eld data show that 25% of caterpillars live ten months with ants before pupating, whereas 75% grow slowly, parasitizing their hosts for 22 months. Both types of caterpillar form apparently identical similar-sized pupae. This may be the ¢rst description in the animal kingdom of polymorphic growth rates spanning di¡erent years within the same population, yet without resulting (as in salmonid ¢sh) in two morphologically distinct adult types with obvious di¡erences in behaviour. We suggest that a balanced polymorphism has evolved in M. rebeli growth rates, representing the most e¤cient way of exploiting the limited, yet steady, daily supply of food available to cuckoo-feeding parasites of long-lived ant societies. Bet-hedging bene¢ts would also accrue to adult butter£ies producing a mixture of annual and biennial o¡spring. Despite ergonomic and other bene¢ts, partial biennialism is unlikely to evolve unless slow-growing individuals have enhanced survival and can remain attached to their mobile hosts. We show that caterpillars become so closely protected by, and integrated with, their host colonies that slow growers experience no greater mortality over two years than fast growers over one, and are transported in preference to the ants' own larvae when the host colony moves nest site.