Dear Drs Werner Ulrich, Richard N. Holdaway, and Tim Coulson,
I am extremely grateful for the opportunity to revise my manuscript. Thank you all for providing thoughtful feedback. In revising my article following your comments, I believe that the manuscript has been improved greatly. Accordingly, I have added your names to the Acknowledgements section of the article in recognition of your contributions. Please find below my detailed responses and descriptions of revisions.
Review by Dr Tim Coulson
1. “As with any extinct species, reports of sightings can often be unreliable, and assigning confidence in each alleged sighting can be a challenge. What I would suggest is removing each sighting in turn or altering confidence in the sighting, and see how this impacts results. It will give an alternative estimate of confidence in the likelihood of extinction in each year.”
Response: Thanks, the jackknife resampling you describe (i.e. removing each sighting in turn) is an excellent suggestion. I’ve updated the Methods and Results sections of the manuscript to implement this method, among other sensitivity tests that were recommended by Dr Richard N. Holdaway in their review below.
2. “A second comment is that a few more details on the methods could be given so that readers do not need to read Brooke.”
Response: Sorry for this; I was aiming for brevity, but I realise I should have sympathised with the reader a bit more in describing the methods. In this revision, I’ve completely restructured the Methods section into three distinct subsections (Data; Model; Conservative Measures and Sensitivity Tests), and I have provided a mathematical description of the model that was implemented, including equations. I hope this is much more clear now. Sorry again.
Review by Dr Richard N. Holdaway
1. “In the Introduction, the author develops the case for reanalysis of moa extinction with a brief survey of moa characteristics and biology. But they rely on secondary, peripheral, and obviously incorrect literature. The best example of the last is the “3.6 metre tall moa” of Widrig & Field 2022). Where those authors obtained that number is puzzling: the largest female giant moa (Dinornis) were 2 m high at the rump and their necks would have allowed them to reach perhaps 2.5 m. And the same reference is derivative with respect to moa biology, for which there is an extensive and easily located literature. Scofield & Ashwell (2009) is hardly the most appropriate reference for the interactions between Haast’s eagle and moa: there are more authoritative publications on the subject… on reproduction rates and age of maturity when there are primary literature references to growth rates, age at maturity, and reproductive rates, such as Turvey & Holdaway (2005) Postnatal ontogeny, population structure and extinction of the giant moa Dinornis. Journal of morphology 265: 70-86 and Turvey et al. (2005) Cortical growth marks reveal extended juvenile development in New Zealand moa. Nature 435: 940-943”
Response: I am both grateful and lucky to have had my manuscript assessed by Dr Holdaway, who I recognise is perhaps the world’s foremost expert on the Moa. Dr Holdaway’s experience with this literature vastly exceeds my own, and so I appreciate these suggestions. Accordingly, I have:
- Deleted the comment on the height of Dinornis robustus and reference to Widrig & Field from the second paragraph of the Introduction.
- Replaced the reference for Moa as megaherbivorous browsers in the second paragraph of the Introduction with Wood et al. (2020), which represents a recent synthesis of current knowledge of Moa feeding habits and contains many comments on Moa browsing.
- Replaced the reference for Moa as Haast’s eagle prey in the second paragraph of the Introduction with Brathwaite (1992) and Worthy & Holdaway (2002).
- Replaced the reference for reproduction rates and sexual maturity in the third paragraph of the Introduction with Turvey & Holdaway (2005) and Turvey et al. (2005).
- Replaced other citations with Worthy & Holdaway (2002).
2. “While acknowledging the rapid extinction model in Holdaway et al. (2014), the author then cites Perry et al. (2014) as reporting high quality radiocarbon ages. The earlier cited paper and Allentoft et al. (2014) reported much larger, and just as precise, suites of ages from throughout the South Island, and assessed the chronology of extinction.”
Response: Thanks, I have added citations to Holdaway et al. (2014) and Allentoft et al. (2014) to this sentence in the third paragraph of the Introduction, as requested. I’ve also added Holdaway and Jacomb (2000) in acknowledging the rapid extinction model.
3. “In discussing alternatives to the overkill, the author cites Diamond (2000). Diamond (pers. comm. 2002) pointed that his view had changed after he had seen how even sporadic hunting trips by the indigenous inhabitants severely limit cassowary populations in terrain far more challenging than anything inhabited by moa in New Zealand.”
Response: Because I did not have this personal communication with Diamond, I don’t think it’s appropriate for me to cite Diamond directly, however, I have revised this fourth paragraph in the Introduction to refer to the past tense, i.e. changing “is not a view held by all” to “was not a view held by all”; “critics point” to “critics pointed”; and “critics opt” to “critics opted” to reflect a change in opinion. I have also added to the end of the paragraph “Some (previous) critics now support the overkills hypothesis (Holdaway 2023)” where the latter citation is private communication with the reviewer (i.e., the above comment number 3.).
4. “With reference to the obvious possibility of moa survival after 1450 CE (which the analyses in Holdaway et al. (2014) do not preclude), it is interesting that a “sighting” in 1993 should be referenced to a 1973 publication.”
Response: I am slightly surprised to hear the reviewer say that the possibility of moa survival after 1450 CE is “obvious”; although the upper 95% HPD for moa extinction in the South Island is given as 1480 CE in Holdaway et al. (2014), the abstract of this article states that the Moa were “exterminated” in “a brief (<150 years) period”, which would coincide with 1400-1450 CE if the first settlers arrived circa 1250-1300 CE, and this wording suggested to me that the authors of that study did not believe later survival was likely (or indeed “obvious”). Regardless, I have added “which the model by Holdaway et al. (2014) does not completely preclude” to the fifth paragraph of the Introduction. On the 1973 publication, that citation was meant for the latter comment on reports having originated from Maori and European peoples, rather than the former comment on purported sightings having been as recent as the 1990s. Sorry this wasn’t clear. I’ve now cited Spittle (2010) after the 90s sighting comment. Hope this helps.
5. “whether or not Mackal (1983) devoted a chapter to moa is irrelevant: people’s interest in a subject is not evidence for or against the object of the interest.”
Response: Sorry. My intention here by citing Heuvelmans and Mackal was simply to convey that this study (Foxon 2023) is not the first written piece on the possibility of late Moa survival. These citations were not indended to provide evidence one way or another. Accordingly, I have re-written this part of the fifth paragraph of the Introduction to say more precisely that “Other works have discussed this possibility, including Heuvelmans (1986) which lists “A surviving species of the Moa family... known to some as roa-roa” in an ‘Annotated Checklist of Apparently Unknown Animals,’ as well as Mackal’s (1983) book on ‘hidden animals’.”
6. “In addition to the perhaps unwarranted extension of the method in time and including accounts at second and often greater hand, there are issues in the data included. I feel that the events the author has harvested from a compendium of such reports must be far more critically assessed before they are accepted for analysis. For example, many of the records included are by apparent eye witnesses. “Eye witnesses” are regarded as being less reliable in general than other forms of evidence in courts of law because too many conscious and unconscious factors can bias the observation, but obviously not all can be rejected solely on that basis. Having said that, the author includes in the category of eye witness accounts those occurring several generations earlier than the actual report. That approaches the category of hearsay, to which many of the other events can be assigned. The author is well aware of that, given the annotations to the records in the database, but the events were included in the analysis nonetheless.”
Response: First, I have added an acknowledgement of the “unwarranted extension of the method in time” in the revised manuscript, citing the referee (see final paragraph of the Model subsection of the Methods section). Second, to address the referee’s requests for ‘separation of wheat from chaff’, I have added an additional sensitivity test removing all ‘second and often greater hand’ sightings from the model (see Methods and Results). Predictably, the probability of Moa persistence remains low in this sensitivity test, comparable to the pessimistic model with all sightings.
7. “most if not all the observations of bones on the surface or in caves and swamps, can be dismissed as undateable regardless of the observer’s views of the possibilities of their recent deposition. All field experience shows that such remains can appear “recent” after centuries of exposure.”
Response: I agree with this point, which is why I previously made it in the ‘Notes’ column for each surface bone ‘sighting’ in the original NMSRD dataset I provided with the submission (I previously stated in the ‘Notes’ column of the dataset for such records that “bones may have been deceptively old and well preserved”). Whereas I tried to take the ‘ambivalent’ approach (Brook et al.’s wording) of assigning a very low but non-zero probability to this type of sighting, the reviewer would prefer that I took the ‘extreme pessimisim’ approach (also Brook et al.’s wording) by removing them from the dataset altogether (i.e., assigning probability zero). Accordingly, just like with the ‘second and often greater hand’ category of sightings addressed in the response above, I have added another sensitivity test to the manuscript in which all bone ‘sightings’ are removed to estimate their impact on the model. Once again, the probability of Moa persistence remains low in this sensitivity test, as in the main model, as would necessarily be the case when removing a number of sightings from a model of this kind.
8. “There are far more recent and higher precision ages on moa eggshell than were av” [sic]
Response: Unfortunately, it appears that the rest of this comment (after “av”) was cut off in the version of the reviewer report I received. I think the gist of the comment is that, in the original NMSRD dataset I collected, the first few records correspond to carbon-dated Moa remains. I think the reviewer was hoping for “more recent and higher precision” carbon-dated Moa remains than those from Higham (1994), which is the source I used previously. Accordingly, in this revision I consulted Supplemental Table 1 of Holdaway et al. (2014), which contains the conventional radiocarbon ages of 96 Moa eggshells in archaeological contexts. I replaced the data from Higham (1994) with these 96 more recent and higher precision estimates, and I have updated the manuscript to reflect this upgrade. The impact of swapping the original data from Higham (1994) with the more recent and numerous data from Holdaway et al. (2014) was negligible, for example the cumulative persistence probability was 0.21 for 1770 CE in the main model both for the Higham data (in the previous version of this article) and the Holdaway et al. data (in this revision), and to be clear, that’s without making any of the other changes to the data described elsewhere in this document. The reason for this is because all that really matters is the last (most recent) eggshell, which was 1390 CE in Higham and 1408 CE in Holdaway. So while I assure you that I have made the requested change (and I agree that these newer data are more appropriate), the dates for the last eggshells (1390 CE and 1408 CE) are so similar that there’s no impact whatever on the results or conclusion. I do appreciate this suggestion, however.
9. “The bases for other “sightings” seem not to have been thought through. A few examples will suffice… [the referee then provides a number of examples].”
Response: I sympathise (and agree with) this comment by the referee. My question, however, is that given the results of the model already provide very low probabilities of recent Moa persistence even when giving proponents the ‘benefit of the doubt’ and including all sightings in the model, regardless of how dubious they may be, what exactly is the point of applying any additional filter on the data? The referee asserts that removing this selection of sightings “must have implications for the results of the analysis”. This is true, and the only possible impact on the model from removing a selection of sightings would be to decrease the probability of persistence. If we were in a situtation in which the main or even pessimistic models were providing ‘borderline’ persistence probabilities of ~0.5, it would be informative to see if this extra filtering of the data had any impact on which side of the fence the model results land (i.e., more likely than not that Moa persisted, or more likely than not that they were extinct). But the fact is, we are already in a situation with the complete sighting record that the results are pretty firmly in the extinction camp. Given that in addressing previous comments as part of this revision I have already added sensitivity tests for (1) jackknife resampling; (2) removing all surface bone accounts; and (3) removing all second or greater hand accounts, and given the results of these and the main model(s) favour extinction, I think it is sufficient simply to remove the sentence “but give probabilities of Moa survival post-1450 CE that cannot be dismissed entirely” from the Abstract of the article, and to add an additional paragraph to the Discussion section of the article acknowledging the uncertainty of the sightings highlighted in the referee’s comment above, and stating what I have said in this response. Again, I agree with the referee- it’s just not clear to me what more can be added.
10. “While the author is to be applauded for trying to take anecdotes from subjectivity to objectivity, the manuscript in its present form does not achieve that objective in my view. At the end of the process, the results echo the situation before the analysis. Are we further ahead? The Bayesian extinction analyses reported elsewhere have long tails extending to the present, but with vanishingly small probabilities after 1750 CE.”
Response: I think there’s a point to be made here about the fact that the results of the sighting model support and supplement the results of the Bayesian extinction analyses. I don’t see the model in the present study as replacing or surpassing the previous Bayesian model (that was certainly not my intention!), but as complementing it using different kinds of method and data. I’ve added a new paragraph to the Discussion section of the manuscript stating the above.
11. “Given that on present evidence moa had large territories and most species physically could not negotiate steep terrain is beyond possibility that such viable populations of moa survived into the 20th century. All potential surviving habitat (which excludes the precipitous rain-soaked valleys of Fiordland) has been traversed continuously first by gold prospectors (who had visited most valleys by 1880) and since by hunters, hikers, and scientists.”
Response: I like this quotation so much that I have included it in the Discussion section of the manuscript with citation to you. I hope you don’t mind…
Thanks again for your consideration of my manuscript. I hope that these revisions are satisfactory.
Floe Foxon - Folk Zoology Society