All reviews of published articles are made public. This includes manuscript files, peer review comments, author rebuttals and revised materials. Note: This was optional for articles submitted before 13 February 2023.
Peer reviewers are encouraged (but not required) to provide their names to the authors when submitting their peer review. If they agree to provide their name, then their personal profile page will reflect a public acknowledgment that they performed a review (even if the article is rejected). If the article is accepted, then reviewers who provided their name will be associated with the article itself.
I appreciate your attention to detail in making your revisions.
Congratulations on a nice paper!
[# PeerJ Staff Note - this decision was reviewed and approved by Dezene Huber, a PeerJ Section Editor covering this Section #]
Please make additional revisions according to the reviewers’ suggestions.
See General comments
See General comments
See General comments
The authors did a very good job in revising their manuscript that is now way easier to read.
The authors addressed my earlier concerns. I now only have few minor comments and editorial suggestions (mostly on the presentation of the Methods) that the authors may want to follow.
1) In the Introduction, I got a bit confused by what the authors refer to as 'signal penalty' (line 134). Intuitively I thought about costs associated with the signal even when it is hidden. I had to read until line 364 before understanding that the authors actually refer to signaling costs. I would suggest the authors (1) to be more explicit at line 134, and (2) to use the term 'signaling penalty' or 'signaling penalty' to avoid any confusion.
2) I find that the terminology about prey behavior is a bit confusing. In particular, the authors could use another expression to refer to the "post-attack behavior"; the sentence "the final attack envent occurred after the post-attack behavior" (line 247) sounds really weird. I understood that the "post-attack behavior" occurs after the initial attack, and before the final attack. Yet, I am afraid that other readers could be confused by this terminology. I would suggest either to change the terminology, or to draw a flowchart showing how the chronology of prey behavior and attack events.
3) In the Methods, I would suggest to explicitly state how the prey behavior is changing (e.g., from the resting state to the approached state). This would considerably help to understand the order of the processes in the main text
4) line 445-446 "it probabilistically decided to attack it or not", lines 453-454 "was the probability of deciding to attack the focal prey in the next time step", and line 459 "each predator conducted an action"; here the process order is not clear at all, and this is rather hard to understand when the attack take place. I guess that it does not really matter during the simulation; once you have update the motivation variable, you can update the survival chance variable at any time. Yet, it would be clearer for the reader to explicitly state that the predator attack (or the retreat) takes place at some point.
Note that the use of the past tense in the Methods does not ease the reading and the understanding of the process order because it make it hard to refer to past/future events. Personally, I would use the present tense but this is a style difference.
Editorial suggestions
line 56: 'but' is misused given because the second part of the sentence does not conflict with the first part of the sentence
line 72: 'to accept'; should the authors use the word 'ingest' instead?
line 80: the authors can remove 'will'
line 100: To ease the reading, this sentence could be rephrased as "Recently, however, researchers experimentally explored the switchable visual displays of some chemically defended prey"
lines 113-114: The expression 'might also accelerate the predator nervous system's learning about the aposematic prey' sounds weird. This could be rephrased to ease the reading.
lines 112: I am not sure whether the authors start a new paragraph here. If not, this could be done to contrast costs associated with signaling and costs associated with switching.
line 162: I would suggest to add 'Notably'; i.e., "Notably, the interactions...."
line 170: 'the general cost of signaling'; do the authors refer to the cost of switchable signaling here, or to the cost of signaling in general?
lines 178-189: Here, the authors could cite the papers that suggested those existing hypotheses.
line 212 and elsewhere: I would avoid referring to 'silent' signals. This can be confusing; it sounds like the authors are describing sound signals.
lines 323-324: "the prey would assume the approached state instead"; This sounds weird. I guess that there is a typo in this sentence.
lines 422-423: "Please note that..."; I did not get the meaning of this sentence. Could the authors rephrase it?
line 475: The authors could write down explicitly the increment equation for *unconditional stimulus*. While reading, I used to look at the following equation to understand what is the use of variable in the subsequent time steps. I guess that writing explicitly the increment equations could really help other readers to understand the Methods.
line 742: To make their point about aposematism based on cues, the authors could refer to the recent opinion paper by de Solan & Aubier (2019 Front.Ecol.Evol. 7:283. doi:10.3389/fevo.2019.00283)
line 905: I would avoid referring to "intelligent predators", the notion of intelligence being rather difficult to define in animal behaviour.
Please see "General comments for the author".
Please see "General comments for the author".
Please see "General comments for the author".
Dear Authors,
I am really pleased with the new version of the manuscript. The writing now reflects well the strength of the methods and the width of the results of the study. In particular, it is now clear that the paper provides not only a novel modeling software, but also, and arguably above all, insights on possible and testable routes for the evolution of switchable aposematism. The new abstract is clear and well written, and provides a clear description of the main question and the main results. It will make readers of the three categories of reader discussed in the first round of review willing to read the paper. In its new form, the introduction is also much easier to follow and well-structured. The methods have gained in both concision and clarity. Finally, the conclusion is very clear and provides a much broader message than the one in the former version of the manuscript, much closer to what the study deserves.
Overall, I think that the main problem of the paper in the first submitted version, namely that the writing didn’t reflect the wealth of theoretical results, which made the paper less appealing, has now been solved and well solved. I am also happy with the answers to my theoretical questions about the interaction between some of the variables (additivity, multiplicativity), and the way the corresponding information has been included in the manuscript. I have found a few instances in which the English language should be improved, in particular the use of “the”. I am not a native speaker, but I think it is worth looking at the points I have listed below to improve further the understandability of the manuscript. Taken together, I consider that the manuscript is sound, novel and interesting for different types of readers, and therefore worth publishing once some minor language corrections will have been carried out.
Best regards,
Olivier Penacchio-----
Minor comments on language:
-L. 49: The switchable aposematism has been relatively -> Switchable aposematism.
-L. 51: Second, it could facilitate the aversive learning -> facilitate aversive learning.
Third, it could minimize the exposure or -> minimize exposure.
-L. 55: 88 128 -> no space, 88128.
-L. 63-64: explore the evolution of the switchable aposematism -> of switchable aposematism.
-In general: “the switchable aposematism”-> “switchable aposematism”.
-L. 83: We will call this form of aposematism as the “switchable aposematism.” -> remove “as the“.
-L. 84: “switchable aposematism”, without “the”.
-L. 84-85: the commas show come after the references in the whole sentence (e.g., with various adjectives such as “facultative” (Sivinski 1981; Grober 1988), “post-attack” (Umbers & Mappes 2015; Kang et al. 2016) “early-acting,” ),…”).
-L. 195: e. g. -> no space.
-L. 268: hypotheses regarding evolution of the switchable aposematism… -> hypotheses regarding the evolution of switchable aposematism…
-L. 270: formulated in -> remove “in”.
-L. 337: It should be noted that the signal intensity is a -> remove “the”
-L. 464: The detailed simulation structure -> Detailed simulation structure
-L. 544: basal detectability and signal intensity has additive relationship, not multiplicative -> basal detectability and signal intensity have an additive, not a multiplicative, relationship
-L. 545-546: “The rationale behind this design is to prevent one side being zero from negating the other; each of them should be able to invoke predator discovery even if the other one is very small.” Should be rephrased for clarity. Maybe something like: “The rationale for this design is that both basal detectability and signal intensity should be able to contribute to predator discovery even if the other variable is close to zero or zero.”
-L. 557: as the Eq. 4 -> “as in Eq. 4” is probably better.
-L. 701: Learning speed…:, without “the”
-L. 736-738: Consider inverting the order of the sentence for clarity.
-L. 839: May better: “Second, we assumed that increasing threat level can only be associated to increasing levels of signal intensity”
-L. 842: 88 128 -> no space.
-L. 857 and other instance below: e. g. -> no space.
-L. 919: Maybe better: “The results of the model (Figure 3) help our understanding of possible evolutionary transitions…”
-L. 955: a higher chance.
-L. 1217: a more detailed discussion: the prey defense. -> no “the”. Same on line 1218.
-L. 1267-1268: may lead to disappearance of post-attack aposematism -> I think “may lead to the disappearance” works better.
Please revise your article according to the reviewers' concerns.
see general comments
see general comments
see general comments
In this paper, Song et al. performed an exhaustive simulation analysis to dissect the conditions under which 'switchable aposematism' can be favoured. Switchable aposematism corresponds to an intriguing anti-predatory strategy where defended prey deter predators by suddenly switching warning signals that are normally hidden. While the benefit of switching conspicuous display is well understood in *undefended* prey (a sudden switch can surprise predators, allowing the prey to escape; a process called 'startling'), we lack a clear understanding of the underlying evolutionary forces favouring the evolution of such switching conspicuous display in *defended* prey. The evolution of aposematism and cryptism are rather complex by themselves, and switchable aposematism corresponds to a mixed anti-predatory strategy. Switchable aposematism can bring about multiple benefits that are not mutually exclusive ('startling', facilitation of aversive learning, reduction of the costs associated with signal display), and multiple costs (higher detectability, lower learning speed, intrinsic costs associated with switching). As such, I agree wholeheartedly with the authors that a theoretical investigation of the underlying evolutionary forces favouring 'switchable aposematism' is required.
In this study, the authors choose to constrain their sensitivity analysis to specific parameters linked to the costs associated with signaling or switching (the action of switching and the switching device alike), and highlight the conditions under which cryptism, constant aposematism or different kinds of switchable aposematism (pre- or post-attack switchable aposematism) can be favoured. I believe that the model is making important predictions. For instance, it shows that all sort of switchable aposematism cannot be favoured unless the costs associated with signaling or switching are explicitly considered. Intuitively, one might expect that accounting for one or the other cost is enough to explain the benefit of any kind of switchable aposematism. More importantly, they provide some theoretical evidence that the evolutionary transitions from cryptism to switchable aposematism or from constant aposematism to switchable aposematism are not equally likely, and depend on the intrinsic costs and on the nature of the predator community.
Overall, I think that the theoretical model is sound, and that the authors made a good job to tackle a rather complex question.
Nonetheless, I list below some important comments that the authors should tackle to make their case even stronger, and to make the paper easier to read.
1) I found the paper rather difficult to read, and I was a lot less enthusiastic at first read. The main issue is that the authors do not highlight their interesting results on switchable aposematism. They choose to emphasize the usefulness of their modeling software ('ApoSim'). I would strongly advice the authors to highlight their results instead. First, I am sure that theoretician and empiricists would be delighted to learn about the findings shown in this paper; this would make an overall more appealing paper. Second, I am afraid that very few researchers would use this simulation code because it is tackling a rather specific antipredatory strategy. If all theoretician would name their model and would ask other researchers to use it because it is "user friendly", I fear that it would not facilitate the reading of the theoretical literature. Third, all theoretician may not find such model with a graphic interface so 'user friendly'. Just to be clear, I am not stating that the model is useless. On the contrary, I hope that the authors of the paper (or other researchers) will use this same theoretical framework to tackle other questions related to switchable aposematism (e.g., deimatism and learning facilitation)! I think that the authors would emphasize the potential of the theoretical framework by being clear about the important and unexpected results they got.
To tackle this comment, the authors should consider to re-write almost all the manuscript to emphasize their results.
2) I appreciate that the simulation run time may have been limiting in such individual-based model. Yet, the authors may consider performing few additional analyses (as Supplementary Information) to test the robustness of the results to other parameters. In particular, I would be curious to see if the results hold for different values of intensity associated with signals 'L' and 'H'. Likewise, testing whether the results hold to variations (i.e., same figure as Figure 3, but with two different values around the default combination of parameter) of generation length, prey frequency, predator frequency... etc... would allow the reader to be more confident regarding the robustness of the main conclusions.
3) The authors should really think about removing the 3D representation. By distorting scales, 3D plots are very hard to read. In subfigures 3A, 3B and 3C, each subplot (for specific predator turnover values) could be represented as a separated 2D scatter plot.
4) Terminology "Cost"; "cost-reducing". In the introduction, the authors should be more specific and consistent concerning the terminology associated with costs. For instance, in line 122, I was very confused at first read because this is the first time the authors discuss the intrinsic costs associated with switchable aposematism, while the authors only discussed about 'cost-reduction' before. Given that these costs are the most important components of the model, the authors should be more explicit about all those costs and how this relates to the benefit/costs associated with switchable aposematism (for instance, signal cost brings about an advantage to switchable aposematism over constant aposematism but not over cryptism). I believe that introducing the concepts required to understand the results (instead emphasizing the novelty of the method), the introduction would be more easy to read, and less confusing on this aspect.
5) In the Material and Methods, the authors should state that all prey individuals are mimetic (i.e., there is one type of signal, with more or less intensity)
6) In the Material and Methods, I was confused about which condition each signal corresponds to (resting, approached, attacked). In particular, for the last 'prey react', it is not clear whether the prey re-adopt the resting signal. This should be clarified throughout the methods.
7) In the 'Overview of the model structure', it should be stated that the predator is tasting the prey upon attack ant that it may affect subsequent attacks.
8) In the Material and Methods (including in the 'Overview'), the initial prey community should be described (I guess it is a mix of all strategies)
9) In the Material and Methods, I did not understand exactly when the instinctive repulsiveness is reset to 0. At each interaction frame? Note also that this terminology is confusing because it seems to refer to innate avoidance, which is not modeled (as said line 172)
10) I did not understand why a prey should do anything after the last attack within an interaction frame (step 3). Why switching signal (and paying a cost) if the predator is going to leave? I guess I missed something in term of process order.
11) It is very odd to refer that often to Figures in the Discussion. In particular, it makes the reading very unwieldy.
Other minor comments:
The authors should care about the use of "the". I found many sentences where it is misused (eg. lines 51, 52, 68, 69, 71, 75)
lines 55-56: This sentence is odd. How one can verify a theory? This sentence would be clearer if the authors were stating their results (cf my first point)
line 73: "one aspect" should be defined in this sentence
line 94: "startling" should be defined (as in lines 100-101)
line 96: "As introduced above": it has been done in the abstract. It may be worth to be more explicit
line 148: citations should be outside quotation marks
throughout the paper, it would help if the citations were ordered chronologically
Figure 3: the authors could consider adding a legend with the predatory strategies (even if it the same legend than in Figure 2)
line 228-230: this is very speculative. I would remove this sentence
Equation 1: What is the current survival chance initially? One?
Equation 9: Same for the initial current motivation? Zero?
line 325-329: This should be discussed further in the Discussion. It is an important assumption that leads to an overestimation of the parameter space with constant or switchable aposematism.
lines 426-427: This should be moved in the section 'Prey-pay-cost' below. Same lines 481-482
line 434: "after predator approach" is confusing because an attack has already occurred. "approach" should be clarified throughout the manuscript.
lines 580-584 and lines 585-588: these are repetitive paragraphs; this could be shortened
Please see 'General comments for the author' below.
Please see 'General comments for the author' below.
Please see 'General comments for the author' below.
The authors develop a model called ApSim to study the different contexts in which the benefit of switchable aposematism can outweigh its cost and therefore become a winning evolutionary strategy. The model considers a large number of parameters that can constraint the output of prey-predator interaction, including predator turnover and learning speed, prey basal detectability and cost of switching on the anti-predator signal.
The modeling proposed by the authors is highly original, allows the full complexity of the interaction between several aspects of prey-predator interaction to be taken into account, and, importantly, does so by circumventing the hard question of measuring empirically the cost of all the behavioral traits involved in the interactions between prey and predator. The last point allows the authors to make several (qualitative) predictions regarding the situations in which switchable aposematism -and other alternative strategies- may be advantageous.
While I am totally convinced that the manuscript is worth publishing, for the reasons given above and below, I do think it could be greatly improved by being partly rewritten. Three types of readers may be interested by the study: 1. readers who will skim over the abstract, the introduction and jump to the conclusion to get the take-home message (in particular, something in the lines of “switchable aposematism may be advantageous when predator learning speed is high enough and the cost of switching low enough”); 2. readers who, in addition, will try to get an idea of the modeling used to show these theoretical results; 3. readers willing to, in addition, reproduce the results and explore related questions using ApSim. The authors give a very detailed description of the model and of the rationale for most of their choice in the modeling. However, the writing of the manuscript, as it is, would make the reading of categories 1 and 2 very hard, and the improvements I propose below would, I think, also benefit readers in the three categories, as well as boost citations.
The first improvement I would suggest is to start each of the very precise, long descriptions of the model setting (there are a lot given the complexity of the model) with one, or some, summary sentence(s) providing the meaning of what follows in plain, non-technical language. For doing so, a possible way is to first make the main claim of the paragraph and next back the claim up with detailed descriptions, rationale or references.
Studies based on computer simulation to understand anti-predator coloration are becoming more and more frequent. They are based on online experiments in which a great number of (human) observers play the role of predators and “forage” to determine the relative fitness of prey. This study is totally different. There is no “perceptual interaction” between prey and predators. Prey-predator interactions are determined computationally by the settings of the model. This should be spelled out very early in the text. I found it very hard to understand what the interaction between prey and predator was before the lines around line 270! I think, and that’s perhaps my main concern about the writing, that a clear, plain description of what ApoSim does in terms of prey and predator interaction should be provided very early on in the introduction. An ideal location would be on line 113, after the first sentence of the paragraph (“We developed ApoSim … this subject”). This added explanation would state that prey and predator interact according to a modeled predation that depends on several parameters etc., which is the core of what ApoSim does.
By reading the conclusion and the abstract only, one would get the feeling that the manuscript is solely about the presentation of a model, which is not true. The work realized offers much more: clear predictions about when switching aposematism should be expected according to the cost of different aspects of the problem. The main predictions of the modeling should spelled out in the abstract of the paper and in the conclusions to make sure readers in categories 1 and 2 (and perhaps 3) above will read the paper. It is clear that the modeling is purely theoretical in the sense that the values for the cost considered are abstract. However, the manuscript does a good job in explaining that these costs are very hard to measure empirically, so the theoretical conclusions of the modeling are a new contribution to the field and should be emphasized as much as the fact that a new model has been proposed, in both the abstract and the conclusion.
Other points:
-Lines 224-230: The short paragraph states first that coevolution is not considered, but goes on saying that the modeling can enhance our understanding of coevolution. I don’t understand how both statements can be true. Please clarify.
-Eq. 4, line 341: I find the negative interaction between signal intensity and basal detectability in the product “signal intensity * (1-basal detectability)” hard to understand. Is that arbitrary? I understand this formula guarantees that discovery chance will be between 0 and 1, but other formulae are possible such as, simply, “discovery chance = signal intensity * basal detectability”, and are, to my understanding, more intuitive. Could you clarify this point?
-Eq. 5, line 353: same as for Eq. 4. Why not, simply, “conditional stimulus = basal cognitive cue * signal intensity”? I don’t understand the negative interaction in the formula.
-A general remark saying that the model is fully qualitative is missing. Other choices for the parameters could lead to a totally different picture in Figures 2 and 3.
-Lines 531-533: “We asked … into the system”. This question is very important and really helps the reader to understand the study. Why not highlighting it earlier, in the introduction?
-Lines 616 and following lines of the Discussion: please explain what you mean by evolutionary transitions and how this idea relates to the modeling. It is not clear from the manuscript.
-Figure 3 really does a good job at summarizing the whole complexity of the study in a single visual representation! This figure should be displayed in the published manuscript on a full page. It would therefore be useful to show the color code and the short-three letters notation given in figure 2 in figure 3 as well.
-Line 731: “would be almost random” -> “wouldn’t favor any strategy”?
Minor point:
-Abstract, line 50: “it” is not precise enough and may be replaced by “the model” (“To illustrate how the model can be used”).
-Abstract, line 56: same remark for “its”.
-Line 130: “these mechanisms” instead of “those mechanisms”?
-Line 134: comma missing after Srygley 2004).
-Line 147: “In the literature the switchable aposematism…”
-Line 162: “cost-saving aspect of the switchable aposematism”.
All text and materials provided via this peer-review history page are made available under a Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.