Summary:

The authors study tri-trophic antagonistic and mutualistic interactions within an island system. They find that specialization depends on island size and is driven by species turnover. I commend the multitrophic approach of the study. The paper is very well written and easy to follow. The study system is well suited for studying the questions of the paper. The methods are thorough and appropriate for answering the study questions. Thank you for the opportunity to review such an interesting and coherent paper!

General comments:

My only bigger concern is the bottom-up effect results. As you state in the results, the AICs of bottom-up and top-down models were very similar. As the differences in the models are rather small, I'm not fully convinced by the confidence how the bottom-up effect is discussed. At least the limitations of the methods and the uncertainty in the results should be addressed. I also wonder if there were other methods to use to confirm the bottom-up effect in the study system. You cite Marjakangas et al. 2022 that provides a trait-based method for quantiying relative influence of bottom-up and top-down effects on networks. I understand that you don't have trait data and therefore can't apply the method, but maybe you could at least discuss the potential next steps that could be taken to confirm the bottom-up effect that you find in your paper?

Intuitively, I would think that in a tri-trophic network the antagonistic interactions have a stronger effect than mutualistic interactions, regardless of their position being at the top or at the bottom part of the three trophic levels. What's your take on this comparison between the interaction types beyond the speacialization aspect? Could you elaborate on this ecological aspect of your study system a bit more?

I'm curious about your study system in a slightly larger fragmentation landscape context. Is there knowledge of the three trophic levels on the mainland? That is, do you know if some specific interactions or species have disappeared from the islands that exist in the mainland?

I think that the discussion on the broader context of the results could be expanded. What are the broader implications of your results? Can the results be applied to other antagonistic or mutualistic interaction among other taxa and other interaction types? What can researcher working on other systems learn from your study for their own work? Do you think your results will hold for other fragmented study systems? I'm especially interested in knowing if you think similar patterns could occur in systems where the edge between habitat patch and matrix isn't as "hard" as between forest and lake.

I also think that the discussion on the limitations is lacking and they should be included in the discussion. What are the weaknesses of your approach? Are there any limitations?

Finally, it would be great if you provided the code and the data you used to conduct the analyses. I believe many would benefit from following your thorough analysis steps in future studies. In general, data and code availability would increase transparency, replicability and impact of your work. Currently, raw data or code aren't available (except for intermediate step outputs).

Line-specific comments:

I suggest specifying somewhere that by interaction rewiring you mean rewiring across space and not over time.

Line 58: Consider expanding on and defining the island biogeography concept here.

Line 233: I suggest reminding the reader about the definition/meaning of sample coverage here.

Line 316: "led" to "lead"?

Line 322: "The decreasing of island area was accompanied by a decreasing in species richness" to "The decrease of island area was accompanied with a decrease in species richness"?

Line 322: Abundance-driven mechanisms have been observed in other mutualistic systems as well. Maybe you would want to expand the discussion to cover mutualistic systems beyond aphid-ant one? Example of an additional reference: Chacoff et al. 2018 DOI: 10.1002/ecy.2063 that shows abundance-driven patterns in mutualistic interactions.

Line 365: "may occur caused by" to "may be caused by"?

Line 366: Note the different citation style of a reference that isn't in the reference list.

Line 380: I suggest to rephrase "regardless of antagonistic or mutualistic interactions" to "regardless of interaction type".

Figure 1: Could you use a color to highlight the island circles on the map? Currently it's very difficult to see the circles against the greyscale map.

Figure S5. I like that you have a graphical explanation of a key variable of the paper, nice figure!