Many cuckoos are specialists and parasitize one or few host species. For example, black-eared cuckoos Chrysococcyx osculans mainly parasitize Pyrrholaemus sagittatus and Pyrrholaemus brunneus, and they lay chocolate brown eggs that closely resemble those of their hosts (Brooker & Brooker 1989). However, some cuckoos are generalists and parasitize a range of host species. Egg mimicry may be problematic for these species, if their various hosts lay eggs of differing colours and patterns. The most studied generalist cuckoo, the European cuckoo Cuculus canorus, has resolved this problem by evolving a number of host-specific races (gentes), each with a distinctive egg type to match its host (Brooke & Davies 1988; Gibbs et al. 2000). Each individual female cuckoo always lays its eggs in the nests of its favoured host species (Nakamura & Miyazawa 1997; Honza et al. 2002; Avilés & Møller 2004), probably the host that reared it, and its eggs closely match those of that particular host (Davies 2000; Avilés & Møller 2004).

Polymorphic, host-specific egg types have been reported in many other generalist cuckoos, including several other Cuculus species (C. saturatus, Higuchi & Sato 1984; C. poliocephalus and C. sparverioides, Baker 1923; C. solitarius, Cherry & Bennett 2001), several Chrysococcyx cuckoos (C. Caprius, Reed 1968; Lawes & Kirkman 1996; C. cupreus and C. klaas, Tarboton 2001), the brush cuckoo Cacomantis variolosus (Beruldsen 2003), and the Jacobin cuckoo Clamator jacobonus (Tarboton 2001). It is unknown how other generalist cuckoos avoid egg rejection by their hosts. If hosts have similar egg types, cuckoos might lay a ‘compromise’ egg type that is intermediate in colour and pattern between those of its primary hosts (Edvardsen et al. 2001). Another possibility is that host-specific egg types are universal among generalist cuckoos, but they have not always been detected.

Assessing egg mimicry is not always straightforward. The great spotted cuckoo Clamator glandarius appears to have eggs that mimic those of a host, the magpie Pica pica. Closer inspection using spectrophotometric methods suggested that the eggs were not significantly related to the appearance of host eggs (Soler et al. 2003). Conversely, in the red-chested cuckoo Cuculus solitarius, each gens lays a different egg type, some of which appear to mimic the eggs of their hosts, while others appear non-mimetic. Spectrophotometry has shown that the cuckoo eggs actually match the eggs of their hosts most closely at wavelengths that cannot be perceived by the human eye (Cherry & Bennett 2001). These studies emphasize the inadequacy of human comparisons applied to the coloration of bird eggs, and the importance of techniques such as spectrophotometry to measure colour objectively.

Although the host eggs were superficially similar, our results showed that their reflectance spectra differed significantly from one another in the variation described by PC1 and PC2. Lichenostomus chrysops and M. affinis scored higher for PC1 than the other two hosts, whereas M. affinis eggs scored highest for PC2, followed by L. melanops. The pallid cuckoo eggs from the four different hosts' nests matched their respective hosts closely in these features. These results indicate that selection has led to differentiation between cuckoo eggs, rather than a ‘compromise’ egg type, even when hosts have similar egg types.

PC1 and PC2 accounted for most of the variation in reflectance spectra (over 95%), but a small amount of variation (2.6%) was explained by PC3, which included variation at ultraviolet wavelengths. Pallid cuckoo eggs did not share the small UV peak shown by hosts. There are three possible explanations for this result. First, mimicry of UV wavelengths may be more difficult than mimicry at other wavelengths in some way, constraining the accuracy of mimicry by cuckoos. Second, it is possible that these host species do not use UV reflectance in egg discrimination, rendering mimicry at UV wavelengths unnecessary. Cherry & Bennett (2001) found mimicry of host eggs by the red-chested cuckoo at some wavelengths but not others, and suggested that egg discrimination by hosts may be performed only at certain combinations of wavelengths depending on the spectral sensitivities of the particular host. Third, the pigments in cuckoo and host eggs may differ (Davies 2000) with the result that age has a greater effect on the reflectance in the UV on pallid cuckoo eggs compared to host eggs. If so, it is possible that freshly laid pallid cuckoo eggs will mimic their hosts in the ultraviolet as well as at other wavelengths. It is also possible that the novelty of cuckoo eggs has prompted greater handling. However, this is unlikely because the difference in UV reflectance was consistent between cuckoo eggs and host eggs, and was constant through all years sampled.

Egg mimicry by cuckoos has been shown to evolve in response to egg rejection by hosts (Davies 2000). The accuracy of egg mimicry by pallid cuckoos suggests that many of their hosts are likely to be egg rejectors. This has been confirmed for two pallid cuckoo hosts, willie wagtails and rufous whistlers Pachycephala rufiventris (Langmore et al. 2005). It is likely that many honeyeater species are also egg rejectors because they possess the typical characteristics of egg rejector species; they are favoured hosts and they build cup-shaped nests that provide good visibility for egg discrimination (Langmore et al. 2005). This is yet to be tested experimentally.

Museum collections provide a potentially biased source of parasitized clutches (Moksnes & Røskaft 1995; Aviles & Møller 2004). If egg mimicry is very accurate, egg collectors may detect parasitism only when the cuckoo egg is a relatively poor match. This effect is counteracted if hosts reject non-mimetic eggs, because egg collectors are then more likely to find clutches containing cuckoo eggs that are a good match. The former possibility can be largely discounted here because pallid cuckoo eggs are larger than host eggs, so they are detectable regardless of the degree of colour mimicry.

The second possibility, that clutches containing the most mimetic eggs are over-represented in museum collections, is of more concern because it would give the same correlation between host and cuckoo egg colour as is predicted by the gentes hypothesis. However, we consider this an unlikely explanation for our results. If the mimetic cuckoo eggs in this study are not the product of gentes, there are two alternative, but less adaptive, explanations. First, individual female cuckoos could parasitize multiple hosts and produce multiple egg types. However, studies of egg-laying behaviour in the congeneric European cuckoo suggest that females are host-specific (Nakamura & Miyazawa 1997; Honza et al. 2002; Avilés & Møller 2004). Moreover, avian egg coloration appears to be controlled genetically (Collias 1993; Gosler et al. 2000), constraining females to production of a single egg type. Finally, such a system would be less adaptive than gentes, because the probability that the cuckoo would lay the appropriate egg type for a particular host simply by chance would be substantially lower than when a cuckoo gens targets its appropriate host. Second, there could be multiple egg types in the pallid cuckoo population, but individual females produce a constant egg type that is laid randomly in the nest of any host species. Our results could then be explained by host rejection of any cuckoo eggs that do not match their own egg type. However, this strategy would be very wasteful from the cuckoo's point of view (Davies 2000) and would rapidly lead to the evolution of gentes, because a cuckoo's reproductive success would be strongly biased towards the host with the matching egg type. Most eggs laid in other host nests would be rejected, selecting against parasitism of those hosts. The existence of gentes in the pallid cuckoo could be confirmed in future studies using genetic analysis of mtDNA, as demonstrated by Gibbs et al. (2000) for the European cuckoo.

Two results show that egg coloration may be affected by handling, suggesting that it is necessary to control for both the age and origin of museum clutches in studies of parasitism. First, PC3 decreased with the age of the egg. This may indicate a decline in UV reflectance caused by handling (J. Endler 2005, personal communication), because older eggs are more likely to have been handled frequently than younger eggs. Second, we found a significant effect of the random term ‘nest ID’ for PC1, indicating a within nest correlation. This could be explained by each nest being subjected to slightly different treatment in the form of handling and storage, reflecting the need for care in selecting collections for analysis. Alternatively, the effect could indicate regional variation in host eggs. For example, L. penicillatus is reported to lay white, pink or buff eggs (Slater et al. 2003). If pallid cuckoos mimic such regional variation in host eggs, this could explain the within nest correlation. A larger sample size of parasitized clutches would be needed to explore this possibility further.