The role of behavioural flexibility in promoting resilience to human environmental impacts

Dear Prof. Chambers, Dr. Šlipogor, and Dr. Vernouillet, 

Thank you very much for reviewing our revised manuscript and responses. We are so happy you think the registered report is improved! Your additional comments are very useful and we appreciate being able to address them to further improve the manuscript.

The revised manuscript is available at http://corinalogan.com/ManyIndividuals/mi1.html (note: the previously submitted version is at http://corinalogan.com/ManyIndividuals/mi1v2.html).

The tracked changes version of the manuscript is available as an editable google doc (https://docs.google.com/document/d/1av1FkTIvav8P1hFrWse40VeBJxp_v_pPGp-PwabFXrY/edit?usp=sharing - to show the outline, click View > Show outline) (Note that the google doc version shows up accurately, while the PDF that was uploaded to PCI RR doesn't show table deletions)

and as an rmarkdown file at GitHub (https://github.com/ManyIndividuals/ManyIndividuals/blob/main/Files/rrs/mi1.Rmd). In case you want to see the history of track changes for the rmd file at GitHub, click the link and then click the “History” button on the right near the top. From there, you can scroll through our comments on what was changed for each save event and, if you want to see exactly what was changed, click on the text that describes the change and it will show you the text that was replaced (in red) next to the new text (in green).

In case you find it useful, we also provide the author response as an editable google doc (https://docs.google.com/document/d/19_bg5_QjdwEcWbzizVNganUEiL9bGZrS3u6oPKVy7aU/edit?usp=sharing).

All our best,

Corina, Rachael, Dieter, and Kelsey

Round #2 

by Chris Chambers, 25 Aug 2022 13:46

Manuscript: http://corinalogan.com/ManyIndividuals/mi1.html version 2022-07-18

Minor Revision
**COMMENT 1:** Two of the reviewers kindly returned to evaluate your manuscript, and I'm happy to report that both are broadly satisifed with your revision and response. There are some final minor points to address, including clarification of various corners of the rationale and methodological details, and some final methodological checks. Provided you are able to respond comprehensively to these points in a minor revision, Stage 1 acceptance should be forthcoming without requiring further in-depth review.

**RESPONSE 1:** This is wonderful news, thank you so much! Please see how we addressed the comments and revised our manuscript below.

Reviews
Reviewed by Vedrana Šlipogor, 12 Aug 2022 21:58
**COMMENT 2:** Dear Prof. Chambers, Dr. Logan, Dr. Shaw, Dr. Lukas and Dr. McCune, 

Thank you for the opportunity to review a revision of “How to succeed in human modified environments”. 

The Authors have carefully and thoroughly revised their registered report, addressing all my comments. The Authors have clarified some further raised points and added text to the RR where appropriate (Responses 5, 7 and 9), added interpretations of the theory that could be shown wrong by the outcomes to the Tables (Response 6), and corrected smaller typos (Responses 10 and 11). In terms of the Response 8, I now understand the Authors’ choice and agree with the Authors’ reasoning that differences in number, duration, and methods of focal follows between grackles and jays can differ a bit as the Authors will not compare these two species directly. I would suggest to just once again please check the calculated duration of grackles’ focal follows in the main text (8 focal follows of 10 minutes per individual for pre- and post-phase equals 80, and not 800 minutes). From my side there are no further comments.

I wish the Authors all the best with the project and am looking forward to seeing the results of these interesting studies. 

With my best wishes, 

Vedrana Šlipogor

**RESPONSE 2:** Thank you very much for your wonderful feedback! We look forward to conducting the study and sharing our results with you! Thank you also for catching this mathematical error regarding the 800 minutes. The error came about because we were thinking in terms of across all birds even though this line specifically indicates per bird. We revised the line to say:

Methods > Assessment of the likelihood… > Focal follows: “8 total 10-min follows for 80 minutes of follow per bird”

Reviewed by Alizée Vernouillet, 31 Jul 2022 16:40
**COMMENT 3:** How to succeed in human modified environments

I have now gone through the revised report and the letter to the reviewers. I applaud the authors for their revisions work on the report, I do think this new version has improved in comparison to the first one! The details that were added strengthen the rationale behind the study and help clarify certain methodological details, especially about the follows.

The statement regarding potential social learning of individuals that would not be trained and how this mechanism is ultimately also part of the birds’ response to the manipulation (Response 2) is a clever way to address the comments raised by the reviewers.

Overall, I now have mostly minor comments to make, as can be found below.

1. In the first paragraph under the section “Determining after which reversal an individual has completed the experiment: serial reversal passing criterion” – remove extra word in “however it is important to set the passing criterion for this experiment using the this particular setup”.

**RESPONSE 3:** We are so glad you like the revision - your feedback was crucial to the strengthening of the registered report! Nice catch on the grammatical error - we deleted “the” from “the this”.

**COMMENT 4:** 2. Regarding Response to comment 13, I was wondering if it might be useful for the reader to be explicitly told that individuals that are already highly flexible (as evidenced by the high number of microhabitats used or prey items eaten) would benefit less from the manipulation, or more generally that the flexibility training might influence individuals differently based on their baseline level of flexibility.

**RESPONSE 4:** We think it would be really useful for readers to have these details, thank you for bringing this up. We added the following paragraph:

Methods > Determining after which reversal…: “There will be individual variation in terms of baseline flexibility before the manipulation such that the flexibility training might influence individuals differently. For example, individuals who are already flexible before the manipulation will not benefit much from the manipulation, while the less flexible individuals will benefit more. Individuals who are already flexible and pass the serial reversals in fewer reversals will still meet the experiment’s passing criterion and be considered to have completed the manipulation, even if they did not improve. Baseline flexibility differences could also be reflected in their pre-manipulation success measures (i.e., individuals with high baseline flexibility might already be successful before the manipulation) if these success measures relate to flexibility. Our statistical models account for these baseline individual differences in success as they might relate to performance on the flexibility manipulation because they include an interaction between the intercept (the value at which individuals start) and slope (by how much they change).”

**COMMENT 5:** 3. For the success measures of the studies, the authors mention adding three to the micro-habitats of all the birds in a ‘natural’ habitat as there are three more micro-habitats in the urban habitat in comparison to the ‘natural’ habitat. Yet, in the description of the micro-habitats, the authors only mention “human structure” and “cement” as the additional micro-habitats. Unless the authors also mean that there will be “high human structure”, “low human structure” and “cement” as differences? It is currently not clear why sometimes only two micro-habitats are explicitly mentioned, while they are referred as three in other parts of the report.

**RESPONSE 5:** We see where the confusion was here - we were counting cement only in the low category and not also in the high category, which made Suburban have 11 microhabitats (compared with Natural’s 8 microhabitats). We now removed this confusion (see Response 6), which makes 12 microhabitats for Suburban. This means we will need to add 4 microhabitats to the Natural numbers for analysis. We updated the R chunk “gtgrhabitatuse” to reflect this change. 

**COMMENT 6:** 4. On a similar note, continuing on the micro-habitats, it is referred that for each one, the authors will also categorize it as “low” or “high”. For which micro-habitats is it referred to, exactly? I imagine “grass” will never have a “high grass” category. Does it apply to other categories such as “cement”? And for “cement”, if it also includes roads, maybe have it change into “impervious surface”?

**RESPONSE 6:** Cement does include roads so we took your suggestion and changed this category title to “impervious surface”. The low and high distinctions can be applied to all subcategories of suburban and natural habitats because, for example, buildings can have grass on the roof, and an individual could be walking on the cement of an upper floor parking garage. We can see how this isn’t intuitive though so we added a sentence:

Methods > Assessment of the likelihood… > Observational methods: “(for example, grass and impervious surfaces can occur above 3m if grass is on the roof of a building, and if an individual is walking on the impervious surface of an upper floor parking garage)”

**COMMENT 7:** 5. What is the rationale to separate native vs non-native vegetation? Depending on the recency of the introduction of the vegetation species, why would it be expected for the birds to react differently to non-native vegetation? (e.g., maybe for not consuming the fruits or exploit other resources from the vegetation?)

**RESPONSE 7:** We see that we did not explain the reason for this distinction. We would like to keep this distinction in because of its potential usefulness for future research by us (which is not yet planned) and by others. Non-native vegetation often predominates in human-modified systems and the ability to use it for habitat or foraging might exclude some species. Similarly, use of non-native food or nesting substrate might serve as an ecological trap - jays are able to use it, but the quality is not equivalent to native food or substrate. So, use of non-native vegetation may help explain differential success of blue jays and Florida scrub-jays. We added a few words (in bold) to clarify the reason for this distinction in the Jay background section:

“Nesting and foraging substrates can be drastically different in human modified environments compared to natural areas (**e.g. predominance of non-native vegetation**; Tuomainen & Candolin, 2011)”

**COMMENT 8:** 6. As it is currently phrased, it is not clear if the authors will simply categorize the food items as simply “plant” and “animal” or if the subcategories will also be taken into account. 

**RESPONSE 8:** Good point. We added a sentence to clarify as follows:

Methods > Assessment of the likelihood… > Observational methods: “Data will be collected on the four plant subcategories and nine animal subcategories and used in the analyses.”

**COMMENT 9:** 7. For the food items, my first thought when reading plant-based “human provided” food items was that it was about food from feeders. For instance, if a jay eats peanuts (a common food item at feeders), would that be considered “human provided” or “seeds”? Or “human provided” food items are more likely found while scrounging, like chips or fries?

I wish you great success and best of luck with this ambitious research program! I am really looking forward to any updates on it!

Alizée Vernouillet

**RESPONSE 9:** “Human-provided” includes any food type that comes from a human source: feeders, scrounging chips and fries, etc. We can see how the term is unclear so we updated the text to clarify (see below). Thank you for the well wishes! We look forward to sharing the results with you!

Methods > Assessment of the likelihood… > Observational methods: ““Human-provided” indicates any food item that was acquired from a store at some point and is left out by humans. For example, sunflower seeds would be considered human-provided if they are in the form of bird seed or a human snack. Sunflower seeds would only be counted in the “seed” category if the bird is seen eating it from a plant.”

Dear Prof. Chambers, Dr. Chaverri, Dr. Šlipogor, and Dr. Vernouillet, 

We are so grateful for the time you took to evaluate and give us such amazing feedback on our registered report “How to succeed in human modified environments”. We revised the registered report per your comments and we describe how we addressed your comments in detail below. 

The revised manuscript is available at http://corinalogan.com/ManyIndividuals/mi1.html (note: the previously submitted version is at http://corinalogan.com/ManyIndividuals/mi1v1.html).

The tracked changes version of the manuscript is available as an editable google doc (https://docs.google.com/document/d/1PcNmRW6qHtOjZipNYL-0e4D5PZ8UZ659VrE5gfszyVA/edit?usp=sharing - to show the outline, click View > Show outline) or as an rmarkdown file at GitHub (https://github.com/ManyIndividuals/ManyIndividuals/blob/main/Files/rrs/mi1.Rmd). In case you want to see the history of track changes for the rmd file at GitHub, click the link and then click the “History” button on the right near the top. From there, you can scroll through our comments on what was changed for each save event and, if you want to see exactly what was changed, click on the text that describes the change and it will show you the text that was replaced (in red) next to the new text (in green).

In case you find it useful, we also provide the author response as an editable google doc (https://docs.google.com/document/d/1kjVVTk9J8q8VwQa4LP38RBRFEUTjOXL42FnuSOxJB3c/edit?usp=sharing).

We think the revised version is much improved due to your generous feedback and we look forward to seeing what you think of the revision!

All our best,

Corina, Rachael, Dieter, and Kelsey

Round #1
by Chris Chambers, 11 Jul 2022 12:27

Manuscript: http://corinalogan.com/ManyIndividuals/mi1.html

Revision invited
COMMENT 1: I have now obtained three very detailed and constructive evaluations of your submission, and I have also read it with interest myself. As you will see, the overall tone of the reviews is positive - the reviewers agree (as do I) that this is a high quality Stage 1 submission with strong prospects for achieving in-principle acceptance (IPA). The rationale and design are carefully considered and the programmatic division of the protocol into three separate Stage 2 outputs is well justified given the scope and ambition of the work.

Despite the encouraging assessments, the reviews do identify some areas that would benefit from clarification and revision. Without providing an exhaustive list, some key areas to address include consideration of appropriate controls (and potential confounds or alternative explanations of potential results), additional details concerning the statistical sampling plans, consideration of the volume of data that will be collected (and whether this can be increased), and elaborating on the potential outcomes of the work for theory. One point also highlighted by two of the three reviewers is the extent to which the study is capable of revealing insights at a population level.

Overall, based on these assessments I am happy to invite a revision. The revised manuscript will likely be returned to a subset of the reviewers for re-appraisal.

RESPONSE 1: Thank you very much for your interest in our work and for the opportunity to revise and resubmit! We address all comments in detail below and we revised the RR accordingly. 

Reviews

Reviewed by Gloriana Chaverri, 17 Jun 2022 18:18
The Registered Report, “How to succeed in human modified environments”, by Logan and collaborators, provides a very thorough description of a proposed project that seeks to test whether behavioral flexibility can be modified in various species of birds, and the potential repercussions of these modifications. They intend to increase flexibility through serial reversal learning in ca. 5 species. After focal individuals in each species have been successfully trained, they will measure various responses, including dispersal from natal territories, survival, habitat breadth, and diet breadth, among others. Researchers predict that being able to increase flexibility experimentally might increase an individual’s success in adjusting to human-modified environments, which could eventually be used as a management tool for endangered species.  

COMMENT 2: In terms of the scientific validity of the research question, I believe the researchers have provided sufficient evidence to demonstrate that the question is valid. My only worry on this issue is the basis provided to presume that manipulating flexibility could potentially be used as a conservation tool for entire populations. For the focal species, authors did not provide clues on whether there is the potential for social transmission of information. The latter is a vital condition that must be met for this potential management tool to be successfully deployed. So at the moment I think that you could predict that focal individuals could increase their flexibility and perhaps their success, but how this would translate to population-level effects is debatable. 

RESPONSE 2: Thank you for bringing this up. We agree that non-manipulated individuals might socially learn about the outcome variables from the manipulated individuals, but we do not test this in our study where we are trying to determine which outcome variables are influenced by the flexibility manipulation. Once we know this for each of our species, we could plan a future investigation of social learning by non-manipulated individuals. The implications from the results of our current studies would require conservation managers to manipulate flexibility in the individuals whose success behavior they want to change. In some cases, it is possible to train many individuals in a population or a species because they are so endangered that there are not many left. In other cases, it would be possible to train all of the individuals involved in a conservation management event such as a translocation. Therefore, there can still be significant population consequences even if each individual needs to be trained to achieve the goal. We clarified in the RR as follows:

Introduction: 
“While we do not examine the potential spread of the post-manipulation success behaviors from manipulated individuals to individuals that are not involved in our studies, we acknowledge that this is a possibility worthy of future investigation.” 

“In the event that social learning is not used by a given population to spread the behaviors of manipulated individuals, investing in the training of specific individuals to increase their success in the wild could still have conservation impacts. In some cases, it is possible to train many individuals in a population or a species because there are not many individuals left [@greggor2021pre]. It is also possible to train all individuals involved in a conservation management event such as a translocation [@greggor2021pre]. Therefore, there can still be significant population consequences even if each individual needs to be trained to achieve the goal.”

COMMENT 3: The logic, rationale, and plausibility of the proposed hypotheses seem very sound. Authors provided, in my opinion, a good basis to understand each hypothesis and associated predictions. They also seemed to have considered sufficient outcome-neutral conditions, which I believe ensure that the obtained results will test the stated hypotheses.

The proposed methods seem feasible, and the analysis pipeline adequate. The clarity and degree of methodological detail is certainly sufficient to closely replicate the proposed study procedures and analysis pipeline. In the section Design 1 – Reversal learning experiment in visual isolation (Methods, Planned sample) you mention the following: “Once a preference for the rewarded option is reached (17/20 trials correct)…” If I understood correctly, you would the 10/12 passing criteria to decide when an association has been learned, right? If so, then you should change the text in the section. 

I wish you great success with the project, and look forward to its results!

Gloriana Chaverri

RESPONSE 3: Thank you very much for your positive feedback and for the wishes of success! Good catch on the outdated criterion of 17/20. It was a typo - we meant 10/12 and we now corrected it.

Reviewed by Vedrana Šlipogor, 24 Jun 2022 13:54
“How to succeed in human modified environments” is a registered report seeking to find out whether behavioral flexibility, that is potentially instrumental to coping with rapid change especially in human-modified environments, can be increased by empirically increasing environmental heterogeneity, and whether this improves success of impacted species in such environments. The project will be done on four bird species, two of them successful in human-modified environments (great-tailed grackles and California scrub-jays or blue jays) and two of them vulnerable (toutouwai and Florida scrub-jays), to focus on how flexibility is linked with success and whether flexibility can be trained to improve species’ success (i.e., as defined for relevant behaviour in respective population) in human-modified environments. As authors note, the project will also serve as a kick-start for the ManyIndividuals initiative.

COMMENT 4: This registered report was truly a pleasure to review. All aspects of the report are written in a very clear, thorough, insightful, and informative way. The authors have carefully considered many aspects of their research, given elaborate hypotheses and predictions on the outcomes confirming or disconfirming their predictions, proposed detailed methods, video coding criteria, reliability analyses, as well as power analyses and statistical analyses. The research question is timely and makes sense in the light of both theory and applications, is clearly defined and the proposal includes ample hypotheses that could answer the research question. To my reading, the protocol is sufficiently detailed and enables replication by an expert in the field. There is exact mapping between the theory, hypotheses, sampling plan (with sufficient sample size), allowing for two types of design: a within-subject design with higher power and a between-subject design with lower power. The authors propose Bayesian modelling for their statistical analyses, and also provide possible interpretations given different outcomes of the tests. The authors also have some excellent ideas on how to improve reliability of the automatically collected data; for instance, they will be introducing a verification step for the reliability of the feeder data, that will be compared with data obtained by hypotheses-blind video coders. Launching the ManyIndividuals initiative within this registered report is also very exciting. 

There are only several relatively minor points that I think would be good to expand a bit more in this Stage 1 report, and I list them below.

RESPONSE 4: Thank you so much for your positive feedback on the RR and for your enthusiasm about the launch of ManyIndividuals! We are very glad you enjoyed reviewing the RR and we are excited to incorporate your contributions to improving our work. 

COMMENT 5: One point that should be added is a bit clearer prediction and more theory on how flexibility, as achieved by the authors’ proposed manipulation in a serial reversal learning task can be transferred socially in a group and/or population. The authors touch on this in Introduction, but do not elaborate much, which I think would be helpful to a reader. Perhaps an example of two would also be helpful.

RESPONSE 5: This is a good point and was brought up by another reviewer as well. Please see our Response 2 above for how we addressed this.

COMMENT 6: In terms of Tables 1-3, it would be great if the authors could expand a bit more on aspects in which the theory could be shown wrong by the outcomes, as the theory is only listed, but not explained, thus not all aspects are immediately clear from the current reading. 

RESPONSE 6: We can see where it isn’t obvious how to interpret the theory that could be shown wrong in the study design tables -  thank you for pointing this out. We added details to this column at all of the study design tables to address this.

COMMENT 7: Ethical permissions are already obtained (or in some cases pending), but I am wondering if the authors have considered some possible negative aspects of their manipulations? Could the birds, by increasing flexibility, also increase other behaviors, that would make them more risk-prone, and thus perhaps less successful, as they would be more susceptible to predators or human-induced risks? It would be nice if the authors could briefly touch on this aspect in their report. 

RESPONSE 7: Thank you for bringing this up. Because of the way we set our serial reversal passing criterion, the manipulated individuals are only manipulated to be as flexible as the fastest 20% of tested individuals. This means that we are not making any individuals more flexible than what already exists in their population. Therefore, we do not expect any newly introduced negative effects as a result of the manipulation. We added the following to the RR:

Methods > Determining after which reversal an individual has completed the experiment: serial reversal passing criterion: “We do not expect that the serial reversal manipulation will introduce new negative effects because the passing criterion is set such that the manipulated individuals are only as fast as the fastest 20% of tested individuals. This means that we are not introducing an unnatural amount of flexibility because we are not making any individuals more flexible than what already exists in their population.”

COMMENT 8: Baseline behavior tracking for grackles: if possible/feasible, it would be great to increase either the number of minimum focal follows or their duration per individual, as this would also make observations more comparable to the other species studied (while grackles will be observed for 80 minutes, jays will be observed for 480 minutes in total). Perhaps I have missed it, but please also mention when in the day the baseline focal data will be collected, and/or if it will be always collected at a particular time of the day (e.g., lunch period). For jays, please mention how the feeding data will be collected. 

RESPONSE 8: For the minimum number/duration of focal follows for the grackles, the planned minimum, if sampling 20 individuals, is 4 focals per bird before the manipulation, which is 80 follows multiplied by 10 min per follow = 800 minutes of focal follow time (and then repeat this after the manipulation for a total of 1600 minutes of focal follow data). We will likely have to sample many more than 20 individuals in total because we won’t know until after the experiment how many of the banded and tagged individuals complete the experiment and therefore would be eligible for post-manipulation data collection. In comparison, the jay data is collected as all occurrences every one minute over the course of a 60 minute session, so it is a different type of data. We do not have plans to directly compare the grackle and jay data so this should not be a problem. A longer follow isn’t as useful because all of the points within a follow are autocorrelated, so increasing the number of focal follows would be the better solution for getting more data coverage. Given our experience with conducting focal follows at two other sites on grackles, and that we hope to be able to complete the pre-manipulation data collection in one month, this would mean that we would do one follow per week per bird. The follows need to be spaced apart in time to avoid autocorrelation of data points (e.g., foraging for three days in a row at an open garbage bag laying on the ground), so we would not want to decrease the amount of time between follows, which would result from increasing the number of focals per bird.

Please see Response 21 for details on the counterbalancing of the timing of focal follows and the jay feeding data collection methods.

COMMENT 9: Toutouwai flexibility manipulation. Would it be possible/feasible to put both feeders at the same height? 

RESPONSE 9: We want the toutouwai to attend to height when foraging because one of the main reasons toutouwai fail to persist in human modified environments is that they tend to forage on the ground where they are easy prey for invasive mammals like cats. So the goal of the flexibility manipulation is to get toutouwai to learn to forage at a diversity of heights, which is why the rewarded feeder needs to be either low or high, but not both at the same time. That way, when individuals switch between low and high rewarded options as they progress through their serial reversals, the key cue they are attending to is height. We added the following to the toutouwai background section to make it more clear why the height of the feeders matters:

Research Questions > Toutouwai > Background: “One potential reason for the failure to persist, is that toutouwai tend to forage on the ground and are thus at high risk from invasive mammalian predators, which they fail to recognise. It is possible that the more flexible individuals that disperse outside the sanctuary might forage at a diversity of heights and/or more readily learn to recognise novel predators, therefore having a higher likelihood of survival post-dispersal, but currently this hypothesis is untested.”

COMMENT 10: Introduction. When referring to the cited literature, “e.g.” is used a lot, so please check and delete where this abbreviation is redundant.

RESPONSE 10: Thank you for helping us write more clearly! We deleted many of the unnecessary “e.g.’s” in the introduction.

COMMENT 11: Great-tailed grackles, jays & toutouwai. Check whether to use singular or plural (“I will collect” or “we will collect”). 

I am very much looking forward to reading the Stage 2 report and once again congratulate the authors on their thorough Stage 1 report.

RESPONSE 11: Thank you for catching the typos! We intended to change all the I’s to we’s, but we missed some - sorry for not catching this earlier. We made the changes. Thank you so much for your help in making this a better research program!

Reviewed by Alizée Vernouillet, 13 Jun 2022 20:06
COMMENT 12: The registered report proposes an ambitious research project aiming to evaluate whether an individual’s flexibility can be manipulated using serial reversal learning experiments, and whether individuals whose flexibility has been trained will be more successful in a human modified environment (e.g., in a suburban environment). The authors describe a series of experiments that can be implemented in multiple species and that can be used in different research contexts (individual vs in group). Here, they focus on three species (or group of species), each with targeted research questions that derived from the main ones of this research project. As such, this registered report will lead to three full manuscripts, one on each of the three species (or group of species), and can be used as a starting point for the reproducible program ManyIndividuals. 

Overall, this report is well-written and extensively detailed for each of the three projects, without making it repetitive. I enjoyed reading how the different projects were adapted to be relevant for each species by taking into account their ecology. These can be used as examples on how to use the proposed experimental design and to make it relevant for other study species.

On a very small note, due to the different parts being written by different authors, there are occasional changes in personal reference within the text between “I” and “we”.

RESPONSE 12: We are so glad you enjoyed reading about our planned research! We intended to change all the I’s to we’s, but we missed some - sorry for not catching this earler! We made the changes. 

A. Scientific validity of the research question(s) 

The research questions are both valid and scientifically grounded in the literature, yet general enough to allow many avenues to be explored. The introduction is logical and follows recent advances in urban ecology and animal cognition.

B. Logic, rationale, plausibility of proposed research hypotheses 

COMMENT 13: The authors describe how each research question is applied for the three research projects and their study system. There are clearly defined hypotheses for each project, with predictions and descriptions of theories that would be supported/invalidated depending on the results obtained. This section is quite complete and well supported. Yet, I would like more information on some of the measures of successes evaluated. 

1. For both the jays and the grackles, one of the research questions focuses on the number of microhabitats used before and after the manipulation. Can the authors expand on what is meant by ‘used’ in that context? If an individual is resting in a microhabitat, is it considered as ‘used’? Also, given the definition of the microhabitats, would there be any risk to reach a ceiling effect, for example if individuals ‘use’ all the microhabitats before the manipulation?

RESPONSE 13: We understand the confusion here and thank you for the opportunity to clarify. We consider that an individual is using a microhabitat if it is present there for more than 5% of the time during the follows. So this excludes instances where individuals are just passing through a particular microhabitat. Even if an individual is just resting, that indicates it is a suitable microhabitat for that behavior. Regarding a potential ceiling effect, yes, it is possible that some individuals will use all microhabitat types before the experiment begins because we have a pre-set number of microhabitat categories. We needed a pre-set number so we could build our models and run power analyses, which is why we implemented this. However, if an individual is at ceiling before the manipulation, this indicates that they could already be very flexible and we would not expect a difference between their before and after microhabitat measures if flexibility is related to this success measure. Our statistical models take this into account because they have an interaction between the intercept and the slope, so they expect individuals who have large values before to change less in the after condition. We added the following text to the microhabitat research question for jays and grackles:

Research Questions > Grackles > Research questions
and
Research Questions > Jays > Research questions

“We only count that a microhabitat was used if the individual had at least 5% of their data points there. This prevents a microhabitat from being counted even if an individual was simply moving through it, and therefore not necessarily using it.”

COMMENT 14: 2. For the grackles (mostly), what is considered a natural habitat in comparison to the suburban habitat and at which scale will that be evaluated? Will there be a threshold in the amount of time an individual need to be found in a human-modified microhabitats to be considered to be a suburban individual? For instance, an individual living in a suburban environment might spend most of its time in a park.

RESPONSE 14: I think the confusion came from us not stating in the RR main text how we classify suburban vs natural habitats. We quantify suburban habitat as that within 100m of a human structure and natural habitat as that farther than 100m from a human structure. We realize we forgot to include this in the text so we now added it (see below). The suburban or natural categorization is recorded at the focal follow level and not at the bird level. While the grackles move between suburban and natural habitats on a daily basis, the territories of the jays will be located in either a suburban or a natural habitat. We realize that we hadn’t yet made these data sheets, so we added a tab for this data to the grackle (https://docs.google.com/spreadsheets/d/1xELxnCJPa0hIXKagbUPgqq8Tap8mh7PFfLFUGCc525s/edit?usp=sharing) data sheets as an example. We want to keep the same suburban and natural categories across species to make the research comparable.

Methods > Assessment of the likelihood of success in human modified environments with regard to the flexibility manipulation > Observational methods:
“Microhabitat types in the suburban habitat (<100m from human structure) include: vertical human structure (e.g. building, bench), native vegetation, non-native vegetation, grass, cement, and dirt. In the natural habitat (>100m from human structure), microhabitat types include all previous categories, but not human structure or cement. All categories can be further defined by whether the subject was high (>3m) or low (<3m).”

COMMENT 15: 3. With both previous points combined, if individuals in a “natural” habitat and individuals in a suburban habitat both use all the microhabitats available, there will still be a difference in the number of microhabitats used due to the difference of availabilities, rather than an effect of the manipulation. Hence, my question is how the authors will take into account this difference in availabilities in their analyses? The same question is raised for food items. It might be that some environments provide a higher abundance of a certain food item and individuals may rely more on that abundant food item rather than switching into different types.

RESPONSE 15: Differences in availabilities between before and after shouldn’t be an issue because of the within-subject design. For these measures, we will target individuals who have already dispersed before we start collecting data on them (e.g., adults) and the research will be conducted within a particular season (e.g., the non-breeding season), therefore the individuals should be performing similar kinds of activities in both the before and after conditions because they will have the same availabilities. We had discussed whether to use a measure that could take into account that individuals might switch to a more abundant food source, such as the Shannon index of diversity. However, these indexes compress some information and can be more noisy, which is why we decided to use the number of microhabitats/number of food types as the measure. In the analyses, we made it so that experimenters will add three microhabitats to all natural habitat data points (because the suburban habitat has three more microhabitats than natural), which means that the habitats will be equalized in the model.

COMMENT 16: 4. In the same line of thought, individuals may specialize in certain micro-habitats and/or food items in human modified environments. They might have been flexible enough to exploit a new niche but they might not have been required to rely on other micro-habitats or food items within the same habitat due to lack of competition. How would such a situation be handled?

RESPONSE 16: Good point - we mentioned this in Research Questions > Prediction 1 alternative 1 and alternative 2, but we had forgotten to add these alternatives to the study design tables for the species associated with human modified environments. We now added to the relevant sections (if no difference is found in success measures between before and after conditions) of the jay and grackle study design tables in the column “Interpretation given different outcomes”:

“Alternatively, it could indicate that they might have used flexibility in the past when originally forming the association, but the need to maintain flexibility in their repertoire is no longer necessary, or that changes induced by the increase in flexibility do not persist for sufficiently long times to make a difference on the subsequent likelihood of success.”

COMMENT 17: 5. For the toutouwais, what if juvenile dispersion is more related to an individual’s ability to compete with others for a territory rather than their potential to occupy a territory in a more heterogeneous habitat? Also, it is not entirely clear how the research on that aspect is linked to individuals’ ability to adapt to a human-modified environment?

RESPONSE 17: Good point. We added a sentence in the toutouwai background section to more closely link the survival in human modified environments with flexible behavior:

Research Questions > Toutouwai > Background: “One potential reason for the failure to persist, is that toutouwai tend to forage on the ground and are thus at high risk from invasive mammalian predators, which they fail to recognize. It is possible that the more flexible individuals that disperse outside the sanctuary might forage at a diversity of heights and/or more readily learn to recognize novel predators, therefore having a higher likelihood of survival post-dispersal, but currently this hypothesis is untested.”

COMMENT 18: 6. One aspect I am curious about is the expected lasting effects of the manipulation. Is it expected that the flexibility training has immediate consequences and/or lasting ones? If so, at what scale of time? 

The reason I am asking this is to ensure that seasonal effects have been controlled for. For instance, individuals may use different microhabitats and/or consume different food items depending on the season (e.g., when food availability in the “natural” environment is low or during breeding season). Hence, comparisons of food items and microhabitat use before and after the manipulation need to be done when the environment is similar and when the energetic demands of an individual are similar.

RESPONSE 18: We know from our grackle research in Arizona that there are at least short term effects: we showed that increasing flexibility using serial reversals also improved performance on a multi-access box (innovativeness and flexibility) (Logan et al. 2022). The multi-access box (plastic) tests began within three weeks of finishing the serial reversals and ended within four weeks of finishing serial reversals. Beyond this time period, we do not know how long the effects of the flexibility manipulation might last. It would be really useful to find out, but we think it is beyond the scope of the current study where we are trying to identify which success variables are influenced by a flexibility manipulation. For the jays and grackles, we plan to collect the post-manipulation success data within a couple of months of their finishing the serial reversals, so the effects, if any, are likely to last for this period of time. For the toutouwai, where the success measures are collected several weeks after the manipulation, there is the possibility that the effect, if any, has worn off by this time. We listed this as an alternative in Research Questions > Can behavioral flexibility in individuals be increased by increasing environmental heterogeneity? > Prediction 1 alternative 1. We plan to measure the pre-manipulation success variables, conduct the experiment, and measure the post-manipulation success variables all within the same season (e.g., non-breeding season or breeding season). We now specify this in the text as follows:

Methods > Great-tailed grackles > Planned sample: “We will collect data on pre- and post-manipulation success measures and conduct the experiment within the non-breeding season to control for potential temporal differences in the environment and behavior.”

Methods > Jays > Planned sample: “We will collect data on pre- and post-manipulation success measures and conduct the experiment within the non-breeding season to control for potential temporal differences in the environment and behavior.”

Methods > Toutouwai > Planned sample: “We will conduct the experiment in the breeding season and collect data on post-manipulation success measures in the breeding season as well as the non-breeding season. In this case, we do not need to control for potential temporal differences in the environment and behavior because we are measuring juvenile dispersal behavior.”

Research Questions > Can behavioral flexibility in individuals be increased by increasing environmental heterogeneity? > Prediction 1 alternative 1: [changes in grackles were still present for four weeks after the manipulation and longer time periods were not attempted so the threshold is unknown @logan2022flexmanip]

COMMENT 19: 7. Another point on that subject is that if the effect is only lasting on the short term, then the timing of the manipulation might be differently affecting certain measures of success, such as breeding success or survival.

Of course, it might also be interesting to evaluate how long different species retain the result of this training! But this needs to be addressed in the report.

RESPONSE 19: It would be very interesting to evaluate the duration of time the manipulation has an effect on success behavior! It’s a great idea for a future study! Please see Response 18 for how we addressed this comment.

C. Soundness and feasibility of methodology and analyses pipeline (+ power analyses)

COMMENT 20: Again, the methodology for the proejcts and the proposed analyses are sound and most of my questions are more points of clarification that could be added in the protocols.

1. Design 1 (i.e., for the individual training of flexibility, more applicable in a laboratory context where individuals released after training?) is included in the report. Yet, none of the proposed research projects actually use that design (they all use Design 2). As such, description of Design 1 seems out of topic. I suspect that the authors wanted to include it as an alternative design option for the ManyIndividuals project, and it might be worth adding a statement explicitly stating that.

RESPONSE 20: Sorry for this oversight! We added a phrase clarifying this in Methods > Design 1: “Although we do not use this design in our planned studies, we present this design as an option for researchers interested in using the ManyIndividuals framework.”

COMMENT 21: 2. I am not convinced by the focal follows and would like to have more details on it. While I appreciate the amount of work that they require, would 10 minutes be a sufficient amount of time and four times be a sufficient number of repetitions to determine which micro-habitats individuals are using? Similarly, would it be enough to know which food items individuals are consuming?

RESPONSE 21: For the justification about conducting a minimum of 4 follows before and 4 follows after the manipulation and whether these are sufficient for detecting microhabitats and food types, please see Response 8. Because of the confusion around the focal follows, we added more details in a new section: 

Methods > Assessment of the likelihood of success in human modified environments with regard to the flexibility manipulation > Observational methods:

“We will use slightly different observational methods for grackles and jays to collect data on foraging and microhabitat use. However, the categorization of food and habitat types will be the same:

Microhabitat types in the suburban habitat (<100m from human structure) include: vertical human structure (e.g. building, bench), native vegetation, non-native vegetation, grass, cement, and dirt. In the natural habitat (>100m from human structure), microhabitat types include all previous categories, but not human structure or cement. All categories can be further defined by whether the subject was high (>3m) or low (<3m).

Food types are broken down into plant (seed, fruit, human-provided, or unknown plant) and animal (insect larva, adult insect, amphibian, reptile, mammal, bird, egg, human-provided, or unknown animal).

During follows we will record each microhabitat the individual is present in and all food items consumed. Before data analysis, to ensure that we are only including the microhabitats individuals use (rather than just pass through), we will filter the data to include only microhabitats that account for at least 5% of their data points. Although we may not observe every possible microhabitat or food item the individual may use, by equally sampling before and after the manipulation we can detect changes in habitat use.

Additionally, for all observational methods we use binoculars so that we can always attempt to stay far enough from the focal individual that our presence is not affecting their behavior. Because that distance can be different for each individual and species, we hesitate to give a specific number. However, if at any point the focal individual shows it is affected by our presence by looking directly at the researcher, alarm calling or startling, then we end the focal immediately, drop all data from that follow and attempt the follow again the next day.

We will attempt to balance follows for each observational method in the morning and afternoon for all individuals in the study. In this way, we will collect a random sample of data from active and inactive time periods for all individuals. Additionally, we will use automated tracking technology on most of the species, which records daily movement patterns. This will tell us whether there is temporal or spatial variation in bird behavior.

Focal follows

We will collect data on foraging and microhabitat use of grackles during 10 minute focal follows and our minimum sample size will be 20 individuals. We will do 4 focal follows per individual before the flexibility manipulation and 4 focal follows per individual after the manipulation (8 total 10-min follows for 800 minutes of follow per bird). To minimize the temporal and spatial autocorrelation of behavior, we believe it is better to do the shorter follows of 10 minutes and space sequential follows apart by at least 1 week. 

All occurrences sampling

We will collect data on foraging and microhabitat use of jays during spatial movement tracking (as part of another investigation). We will follow jays for 60 minutes and record at 1-minute intervals the spatial location (GPS coordinates), any food items consumed, the microhabitat the jay is present in, and breeding behaviors if it is the breeding season. We will do 4 tracks per individual before, and 4 tracks per individual after the flexibility manipulation.”

COMMENT 22: 3. Related to the later point, following individuals to evaluate which food items they consume need to be controlled in terms of distance. Following individuals would require getting physically close to individuals and potentially distracting/disturbing them? How would the authors control for that? Also, is there a specific time of day that these focal follows will be performed? Birds usually have a specific activity pattern depending on the time of day.

RESPONSE 22: As in all studies in which data are collected through observations, there is a trade-off between quality of data and distance. We describe in more detail our efforts to avoid affecting the bird’s behavior during observations in a new section describing the focal follows: Methods > Assessment of the likelihood of success in human modified environments with regard to the flexibility manipulation > Observational methods. See Response 21 above for the content of this section.

COMMENT 23: 4. Similarly, individuals living in urban environments are known to have different activity patterns than the ones living in rural environments: for example, urban individuals sing earlier to reduce noise interference. They also tend to develop a more nocturnal activity pattern to avoid human interference. Will that be taken into account when performing the focal follows?

RESPONSE 23: Please see Response 21, above.

COMMENT 24: 5. As a minor comment, for the grackles, was there any pilot work done on how to place the radio nodes within the city to detect individuals? And are there permits needed to place the nodes?

RESPONSE 24: We are tentatively planning on going back to our Santa Barbara, California field site to conduct the ManyIndividuals experiment on the grackles exactly for this reason - we have already studied this species in this location and already have permissions for land use and studying this species there. We also know where the grackles live there and where best to place the nodes. We previously had permission from the City of Santa Barbara and the Santa Barbara Zoo to conduct research on their land and it is their land where we would want to place the nodes. We are confident they would give us permission for node placement as well. The only reason we didn’t mention in the RR that this is the planned field site for the grackles is because we will not begin data collection on this study for another couple of years and we know how fast plans can change - we wanted to be able to adapt accordingly.

COMMENT 25: 6. Is there a control at the feeders to prevent other species/non-identified individuals to access the food during habituation? Also, was there any pilot work done/planned on where to best place the feeders (e.g., known foraging grounds)?

RESPONSE 25: Significant previous research on toutoutwai, grackles, and Florida scrub-jays will inform the best areas to place feeders to prevent learning from non-target individuals and access by non-target species. Surveying for known foraging grounds and territory boundaries is the first step in attempting to trap and tag the study subjects. The length and location of the feeder habituation period is described in our species-specific protocols (links: jays, grackles, toutouwai), where we state that habituation does not end until tagged individuals are eating from the feeders. If this does not occur, we will move the feeder to a different location and try habituation again. To prevent non-tagged individuals/species from accessing the feeders, the RFID enabled feeders can be programmed to open only for individuals with specifically coded tags.

COMMENT 26: 7. The power analyses are very detailed in general. However, I got confused with the number of trials required for the passing criterion for reversal learning. More specifically, I understood that, at the end of the section on the passing criterion, the authors will use 10/12 trials for their project. Yet, on the description of the methods of Design 1, the criterion is said to be 17/20. Maybe the authors can clarify this?

RESPONSE 26: Thank you for catching this! It was a typo - we meant 10/12 and we now corrected it.

COMMENT 27: 8. For both the change in microhabitats and in food items, how would change/switch of microhabitats and food items rather than increase of new ones will be controlled in the analyses? Will there be an assumption that the microhabitats/food items that have been seen used/consumed before the manipulation but not after is just due to missing observations and assumed to be still used/consumes after the manipulation?

RESPONSE 27: Our prediction is that flexibility will specifically be linked to the number of different microhabitats/food items an individual uses, not which they choose. Accordingly, we did not predict that birds might switch to new microhabitats/food items after the manipulation, but that, if the manipulation made them more flexible, they would use more. We only count the microhabitats/food items that have been observed before and observed after, and make no assumption about missing observations. 

D. Clarity and degree of methodological details sufficient for replication

COMMENT 28: Generally, the methods are clear and detailed enough to be replicated, while being general enough to de adapted for other species within the ManyIndividuals project. 

1. As a minor comment, it was not clear whether the session of 10 trials is handled when the individual is reaching the criterion. Does the session immediately ended as soon as the individual pass criterion or the session ends up once the 10 trials are done?

RESPONSE 28: We are trying to automate the feeders such that, when an individual reaches criterion, that feeder type becomes immediately inaccessible and the other feeder type becomes accessible, all within the same session. If it is not possible to program the feeders to automatically switch the feeder type, then we will have to check the data sheets that are automatically generated by the feeders at the end of each experimental session to determine whether an individual has passed criterion. This means that we will likely not be able to switch an individual as soon as it passes, but rather in the next test session because we would have to end the session for the whole group rather than just for that individual and we want the rest of the individuals to continue participating. We added to the RR:

Methods > Design 2: “Ideally, feeders will be programmed to automatically switch which feeder type is rewarded as soon as an individual passes criterion in the middle of a test session. If automation is not possible, then the data sheets will be checked at the end of each test session to determine which individuals have passed criterion, and their rewarded feeder type will be changed in the next test session.”

E. Control and other quality checks

COMMENT 29: 1. One interesting research question raised by the authors is whether disturbance-resilient species are more flexible than disturbance-resistant species. Yet, the authors aim to only compare two jay species (one disturbance-resilient and one disturbance-resistant), which is not enough to draw any conclusion in terms of species resilience and flexibility. I would like to have more details on how they will avoid overstating the results, if they are still aiming to only compare the two species.

RESPONSE 29: We recognize that, by comparing only two jay species, our results will be most applicable to answering conservation questions in that specific system. We think this is already an ambitious research project to undertake and hope others will follow the lead of the ManyIndividuals project to study other disturbance resilient and disturbance sensitive species (in an example with U.S. bird species, there are several other ideal candidate pairs from different taxa like wrens, woodpeckers, raptors, etc.). 

This project includes two disturbance resilient (great-tailed grackles, and California scrub-jays/blue jays) and two disturbance sensitive (toutouwai and Florida scrub-jays) species, so we will be able to assess general trends in how the flexibility manipulation affects divergent species. But, to make it more evident that we want to avoid overstating our results, we added to the last paragraph of the Introduction the following:

“To increase the generalizability of the conclusions from the ManyIndividuals project, we here also provide multiple methodological options that other researchers can use to test these questions in additional species.“

COMMENT 30: 2. In case of a social species (such as the grackles and the jays presented here), how do the authors suggest to account for social learning if individuals from a similar social group have had a different training? Especially in fission-fusion species? That might even be a more relevant question for Design 1, if individuals that have been trained separately form a social group after the manipulation.

I hope my comments are relevant and can help improve the registered report!

Dr. Alizée Vernouillet

RESPONSE 30: Your comments have been super useful, thank you so much! Your social learning comment is a really good one and sparked lots of discussion amongst the team. We believe we have landed on a solid way forward. The post-manipulation data collection will begin for each individual as soon as that individual passes their manipulation (or control). In the case of the territorial species (jays and toutouwai), only one individual will be tested per territory, which means it is not likely that tested individuals would learn from each other after the experiment. However, the multiple grackles will be tested at once in their natural groups, therefore they would have the opportunity to learn about success behaviors from each other, and particularly from the manipulated individuals, who we predict will change their success behaviors after the manipulation. We are ultimately interested in determining whether we can change success behaviors as a result of the flexibility manipulation. If part of this change is the result of social learning from some of the manipulated individuals, it will still result in a change even if we do not quantify what percentage of the mechanism comes from individual learning after the manipulation or social learning after the manipulation. If there is a change in success measures between the before and after manipulation conditions, the manipulation will have been the cause in either case. 

There is a way that we could start to tease apart whether social learning is involved in the spread of success behaviors after the manipulation, but it would involve more data collection, which is not feasible for the grackle study. It would involve comparing post-manipulation success measures between individuals who did not pass the manipulation and those who did pass the manipulation. If the two groups have similar success measures to each other, and there is a difference between the before and after measures, then this would indicate that social learning played a role in the spread of success behaviors from manipulated individuals to other individuals. We added a note about this to the Methods so that researchers who are interested in exploring the mechanism in a separate article, would be keyed into the fact that they will need to collect additional data at this stage for later analysis.

Regarding both paragraphs above, we added the following to the RR:

Methods > Assessment of the likelihood of success in human modified environments with regard to the flexibility manipulation: 

“Begin collecting post-manipulation data on an individual as soon as it passes the manipulation because it is unknown for how long any potential effects of the manipulation will last. For the more social species and for populations who participate in the experiment in groups, there is the potential that individuals who were not in the manipulation condition or who have not yet passed the manipulation condition will learn about post-manipulation success behaviors from the manipulated individuals who passed the experiment. We are ultimately interested in determining whether we can change success behaviors as a result of the flexibility manipulation. If part of this change is the result of social learning from some of the manipulated individuals, it will still result in a change even if we do not quantify what percentage of the mechanism comes from individual learning after the manipulation or social learning after the manipulation. If there is a change in success measures between before and after the manipulation, the manipulation will have been the cause in either case. 

If other researchers are interested in beginning to quantify whether social learning is involved in the spread of success behaviors, then they could collect data on post-manipulation success measures from the individuals who did not pass the manipulation and compare them with individuals who passed the manipulation. If there is no difference between both groups’ post-manipulation success measures, this indicates that social learning was involved in the spread of success behaviors. They could then use this data in a future registered report examining the role of social learning in the spread of success behaviors. If researchers are not planning on a social learning component in future research and/or do not have the time or resources to collect more data, they can refrain from collecting post-manipulation data on success measures on the individuals who did not pass the manipulation criterion.”