Quantifying the success of prey crypsis, aposematism, and evasiveness in avoiding predator attack

The authors state that "spatially, Euptychiines preferred the ground, reducing their encounter probability," which is a central claim regarding the efficacy of crypsis. However, the study does not fully account for whether this spatial preference is an inherent property of the prey or a response to the aviary structure. It would strengthen the manuscript to clarify if these butterflies exhibit similar vertical stratification in the wild, as the current results could be an artifact of the aviary's dimensions (4m height).

Regarding the influence of predator experience, the authors note that "immature birds were more likely to attack Heliconius, possibly reflecting a lack of learned avoidance." This is a compelling interpretation, but the manuscript lacks a detailed breakdown of the 'immature' category. Given that this category likely spans a wide range of developmental stages, it would be beneficial to discuss whether the observed patterns are consistent across this spectrum or if they are driven by a subset of very young, inexperienced individuals. Providing more detail on the criteria used to classify 'immature' beyond "plumage and commissure colour" would help assess the reliability of this grouping.

The authors state that 'mortality rates among prey types were largely similar,' and they use this to argue for the adaptive value of diverse defenses. However, in Figure 2, the stacked bar charts show that the 'Subjugated' category (the darkest segment) is quite small for all groups, but the composition of the preceding stages (e.g., 'Sight rejected' vs 'Attacked but not caught') varies significantly. I suggest the authors reframe this claim to emphasize that while *total* mortality is similar, the *selective pressures* acting on these prey are distinct. For instance, Spicauda faces high attack rates but survives through evasion, whereas Heliconius survives through identification and rejection. Characterizing these as 'similar' risks obscuring the very evolutionary mechanisms the paper aims to illuminate.

Regarding the influence of predator age, the authors note that 'immature birds were more likely to attack Heliconius, possibly reflecting a lack of learned avoidance.' This is a compelling finding. However, the manuscript would benefit from a more detailed discussion on the potential for 'social learning' or 'cultural transmission' in the wild bird populations. Since the birds were kept in captivity for up to 30 hours, is there any evidence or observation of whether immature birds might have been influenced by the presence or behavior of other birds in the aviary, or is the behavior strictly considered individual? Quoting the text: 'Immature birds were more likely to attack Heliconius, possibly reflecting a lack of learned avoidance for aposematic signals.' To strengthen this, the authors should clarify if they observed any inter-individual interactions that could confound the 'naivety' assumption.

The most urgent technical issue preventing any attempt at independent replication is that the mathematical formulations are entirely missing or corrupted in the text. For example, in the description of the co-occurrence index on line 160, the text simply reads '(1)' followed by a blank space, and then continues on line 161: 'Where, the first term is the proportion of experimental time...'. Similarly, the generalized linear mixed models (GLMMs) on lines 174 and 176 are presented as '(2a) 1 _' and '(2b) 1 _' with no actual equations. Finally, the log-likelihood equation on line 198 is corrupted as '(3) ∑ 1 1'. Without the exact mathematical structures of these models and indices, it is impossible for peer reviewers or future researchers to verify, let alone reproduce, the statistical findings. The authors must carefully re-render these equations using standard LaTeX or high-resolution embedded symbols.

To align with the open-science standards of a solid specialist journal like Ecology, the authors must revise their data availability statement. Currently, the manuscript states: 'The data that support the findings of this study are available from the corresponding author upon request and will be archived on Zenodo after acceptance' (lines 17-18). In modern ecological research, 'available upon request' is highly discouraged as it severely limits transparency and reproducibility. The authors should deposit their fully curated, trial-by-trial raw scoring data, taxonomic metadata, and R analysis scripts (including the spreadsheet-based log-likelihood calculations mentioned on line 239) into Zenodo or Dryad now, providing a private, anonymous reviewer access link for this evaluation phase, which will automatically transition to a public DOI upon publication.

The manuscript lacks essential methodological details regarding the standardization and reliability of the video analysis. The authors note: 'We thank students who helped processing nearly 300 hours of video material: Tereza Látalová, Sara Turková...' (lines 470-471). When multiple observers score subjective behavioral metrics—such as determining whether a bird is '~10 cm close' (line 182) or identifying 'active attacking behaviour' (line 184)—there is a high risk of systematic observer bias. To ensure the reproducibility of these behavioral classifications, the authors must specify: (1) whether the student observers were blind to the experimental hypotheses and prey categories, (2) how the observers were trained to standardize their scoring, and (3) a quantitative measure of inter-rater agreement (such as Cohen's kappa or an intraclass correlation coefficient) calculated from a subset of trials scored by all observers.