Most mimicry systems involve imperfect mimicry, whereas perfect and high-fidelity mimicry are rare. When the fidelity of mimicry is high, mimics might be expected to have the upper hand against their antagonists. However, in coevolving systems, diversification of model phenotypes may provide an evolutionary escape, because mimics cannot simultaneously match all model individuals in the population. Here we investigate high-fidelity mimicry in a highly specialised, Afrotropical brood parasite-host system: the African cuckoo and fork-tailed drongo. Specifically, we test whether host egg polymorphisms are an effective defence against such mimicry. We show, using a combination of image analysis, field experiments, and simulations, that 1) egg colour and pattern mimicry of fork-tailed drongo eggs by African cuckoos is near-perfect on average; 2) drongos show fine-tuned rejection of foreign eggs, exploiting unpredictable pattern differences between parasitic eggs and their own; and 3) the high degree of interclutch variation (polymorphic egg ‘signatures’) exhibited by drongos gives them the upper hand in the arms race, with 93.7% of cuckoo eggs predicted to be rejected, despite cuckoos mimicking the full range of drongo egg phenotypes. These results demonstrate the effectiveness of model diversification as a defence against mimics, even when mimicry is highly accurate.

Field work took place in September–November 2009, 2010, 2011, and 2019 within a ~3,500 hectare area centred at 16°45' S, 26°54' E in the Choma district, southern Zambia. Fork-tailed drongo and African cuckoo egg phenotypes were measured from nests found at this study site. Eggshell pattern was quantified from photographs of eggs. Eggshell background colour (i.e. avoiding pattern) was measured using spectrophotometry. Egg length and width were measured with digital calipers.

Egg rejection experiments on fork-tailed drongos were performed in 2009, 2010, 2011, and 2019 (n = 14, 19, 19, and 65 respectively). To simulate a parasitism event, we replaced a host egg with a conspecific egg from another female (hereafter ‘experimental egg’; a surrogate for a cuckoo egg). We monitored experiments daily in 2019 and daily when possible in 2009–2011, for four days to determine whether the experimental egg was rejected or accepted. We took the disappearance of the experimental egg to be evidence of host rejection.

Code is implemented in R version 4.2.1.