Defences of hosts against brood parasitic cuckoos include detection and ejection of cuckoo eggs from the nest. Ejection behaviour often involves puncturing the cuckoo egg, which is predicted to drive the evolution of thicker eggshells in cuckoos that parasitise such hosts. Here we test this prediction in four Australian cuckoo species and their hosts, using Hall-effect magnetic-inference to directly estimate eggshell thickness in parasitised clutches. In Australia, hosts that build cup-shaped nests are generally adept at ejecting cuckoo eggs, whereas hosts that build dome-shaped nests mostly accept foreign eggs. We analysed two datasets: a small sample of hosts with known egg ejection rates and a broader sample of hosts where egg ejection behaviour was inferred based on nest type (dome or cup). Contrary to predictions, cuckoos that exploit dome-nesting hosts (acceptor hosts) had significantly thicker eggshells relative to their hosts than cuckoos that exploit cup-nesting hosts (ejector hosts). No difference in eggshell thicknesses was observed in the smaller sample of hosts with known egg ejection rates, probably due to lack of power. Overall cuckoo eggshell thickness did not deviate from the expected avian relationship between eggshell thickness and egg length estimated from 74 bird species. Our results do not support the hypothesis that thicker eggshells have evolved in response to host ejection behaviour in Australian cuckoos, but are consistent with the hypothesis that thicker eggshells have evolved to reduce the risk of breakage when eggs are dropped into dome nests.

Cuckoo and host species. The parasitic cuckoo species selected for use in this study were chosen based on previous knowledge of their host selection and the egg ejection behaviour of those hosts. We selected four Australian cuckoo species based on the known egg ejection rates of their primary hosts from our earlier studies. Two congeneric species, Horsfield’s bronze-cuckoos (Chalcites basalis) and shining bronze-cuckoos (C. lucidus), exploit hosts that build dome-shaped nests in which detection of foreign eggs is constrained by poor visibility in the dark interior. Hosts of these two cuckoo species rarely eject either naturally-laid cuckoo eggs or experimental, non-mimetic plastic model eggs, of similar size to their own (Table 1). The two other cuckoo species in the study, the pallid cuckoo (Cuculus pallidus) and the Pacific koel (Eudynamis orientalis), exploit hosts that build cup-shaped nests with good visibility, and these hosts routinely eject both naturally-laid cuckoo eggs and experimental, non-mimetic model eggs (Table 1). In addition to their primary hosts included in this analysis, these cuckoos also exploit several secondary hosts whose egg ejection behaviour is unknown. However, previous analyses indicate that there is a strong association between visibility inside the nest and egg ejection behaviour; hosts that build dome-shaped nests tend to accept foreign eggs (100% of Australian hosts [N = 6] ejected ≤ 25% of foreign eggs), whereas hosts that build cup-shaped nests tend to eject foreign eggs (75% of hosts ejected > 25% of foreign eggs, [N = 8]). Therefore, we conducted a second set of analyses that included both primary and secondary hosts of these cuckoos (Table 1), where egg ejection behaviour was inferred based on nest type for the secondary hosts.

Eggshell measurements. All eggshells used in this study were sourced from the Commonwealth Scientific and Industrial Research Organisation (CSIRO) Australian National Wildlife Collection (ANWC) oology research collection (Supplementary data file 1). All eggs had been prepared at the time of collection by drilling a small hole in the shell, through which the egg contents were blown and removed. The eggshells were then washed and stored dry. Collector’s notes and consistently small blow-hole diameters indicate that eggs were sampled early in development and were unlikely to be subject to significant eggshell thinning during development (Supplementary data file 1). The availability of parasitised clutches in the ANWC collection dictated which host species were included and the sample size for this study (Table 1; Supplementary data file 1). Suitable cuckoo-host clutches contained at least one intact host and one cuckoo egg, both identified to the species level. We used a precision industrial wall-thickness measuring tool to directly measure eggshell thickness via Hall-effect magnetic-inference, in a similar approach to Peterson et al. However, unlike this previous study we did not cut or damage the eggshell to take measurements. Specifically, we used the ElectroPhysik MiniTest FH7200 gauge and FH4 magnetic probe, with a 1.5 mm diameter steel ball which was inserted inside the empty eggshell, through the existing blow-hole in the specimen (SI 2.0). Thus, all eggs included in the study necessarily had a blowhole diameter > 1.5 mm. This probe and steel ball combination measures thicknesses up to 2 mm, with an accuracy of Å} 3 μm + 1% of the reading (Check Line., Germany). Thickness data were collected at a rate of 10 measurements per second. We did not place the steel probe in direct contact with the egg. Instead, we inserted a 0.73 mm sheet of plastic (cellulose acetate) in between the probe and the egg to minimise risk of damage (hereafter referred to as the ‘protector’). Eggshell thickness data were collected at two regions on each egg—the apex (the most conical end opposite the air sac) and the meridian (the circumference around the widest part of the egg). Manual handling of the egg specimens during thickness estimation is described in detail in Supplementary Information Sect. 2.0 and Fig. S1. Briefly, we inserted the steel ball through the blow hole and rolled the ball to the apex of the egg. We always approached the magnetic probe (and protector) apex-first because this is the strongest part of the egg. Apex thickness was recorded for five seconds by leaving the egg stationary and untouched on the probe (Fig. S1;Video Supplement 1). We then rolled the ball until it was positioned adjacent to the side blow-hole and rotated the egg slowly, to record the meridian thickness (Fig. S1; Video Supplement 2). The steel ball was removed by rolling it back through the blow-hole, whilst still in contact with the probe (Video Supplement 3). Preliminary method optimisation using 60 unregistered eggs indicated that the risk of breaking an egg during this manual handling was very low if specimens had no pre-existing physical damage (cracks, chips, hairline fractures determined via illuminating the egg with a cold-light source) and weighed > 0.05 g (Fig. S2). No registered collection items sustained damage in this study.

All data were inspected and exported following the manufacturer’s protocols in the software package MSoft 7 Basic (Check Line., Germany). The protector thickness was subtracted from the raw gauge readings to obtain a measurement of eggshell thickness (SI 2.0). Mean thickness (μm) was calculated for both apex and meridian measurements of each egg after removing outliers (classified as data points lying outside 1.5X the interquartile range). Mass (g), length (mm) and breadth (mm) were also measured for each egg. Length and breadth measurements were calculated from a 2D photograph of each egg, following Attard et al. Mass was measured using an electronic balance to the nearest 0.001 g (CT-250 On Balance Digital Scale).

Repeatability Analysis. The repeatability of our Hall-effect magnetic-inference methodology with the ElectroPhysik probe was investigated by conducting replicate thickness measurements (N = 10) for an additional 10 unregistered eggs. Repeatability was calculated using the intra-class correlation coefficient (ICC), in the R package irr (SI 4.0). Significance was determined where p < 0.05.

Comparative analysis of avian and cuckoo eggshell thickness. Previous studies indicate that eggshell thickness is positively correlated with egg size; larger eggs have thicker shells77. To investigate whether cuckoo eggshell thickness deviates from this general relationship, we calculated the mean eggshell thickness in a total of 78 species, comprising 12 avian orders and 34 families (total N = 3134 eggs) (Table 2). Our analysis included previously published data for 12 species. We used a phylogenetic generalised least squares regression (PGLS) in the R package caper, and estimated the relationship between eggshell length and two measures of thickness (apex and meridian). To control for phylogenetic relatedness, we used a maximum clade credibility (MCC) tree based on 100 trees downloaded from birdtree.org. The MCC tree, which is the tree with the maximum product of the posterior clade probabilities, was obtained using the R package phangorn. We extracted both phylogenetic residuals (phylogenetically independent) and residuals obtained from the phylogenetic regression line, and visually evaluated whether these residuals from cuckoo species were extreme values (e.g., were greater than expected by their size and phylogenetic position). We also used a linear regression of egg length versus mean eggshell thickness for 74 non-brood parasitic avians using the package lm in R v. 3.6.0.

Statistical analysis. The distribution of raw and normalised eggshell thickness in cuckoos and their hosts was plotted and visually inspected in ggplot2 (Figs. S3 and Fig. 2). Within a species, outliers in the distribution of mean thicknesses (as defined above) were removed. To account for inter-specific differences in egg size (which is correlated with eggshell thickness) ‘normalised thickness’ was calculated for each sample by dividing the eggshell thickness by egg length. This approach is expected to successfully normalise the data because egg length explains a large proportion of the inter-species variation in eggshell thickness (Fig. 3). We tested for successful normalisation by regressing normalised eggshell thickness against eggshell mass (Fig. S4). We tested whether, overall, the thickness of cuckoo eggshells relative to their length differed from that of their hosts at both the apex and the meridian of the egg using a Wilcoxon signed-rank test on matched pairs of cuckoo and host eggs. For this analysis, any unpaired egg samples, or samples with data missing for either the host or cuckoo of the pair were removed from analysis. Final sample sizes for each treatment can be found in Fig. 1. We then tested whether host ejection behaviour predicted the ratio of cuckoo to host normalised eggshell thickness. We used a Restricted Maximum Likelihood Model (REML), with cuckoo:host normalised eggshell thickness ratio as the response variable, host response to foreign eggs (accept or eject) as the fixed effect and host species nested within cuckoo species as the random effect. For all models, we checked standardised residuals for normality. Log₁₀ transformation of variables improved the normality of residuals (Anderson Darling Tests for Goodness-of-fit, all P > 0.4), so we present these results, although the qualitative results remained unchanged regardless of whether or not data were transformed. We ran four models; (1) eggshell thickness at the meridian including only hosts with known egg ejection rates, (2) eggshell thickness at the meridian including hosts with both known and unknown egg ejection rates, (3) eggshell thickness at the apex including only hosts with known egg ejection rates, and (4) eggshell thickness at the apex including hosts with both known and unknown egg ejection rates. The analyses were run in JMP v.15 (SAS Institute Inc, 2019).

Microsoft Excel