Exploring eastern Mediterranean Iron Age coastal connections through sheep astragali using geometric morphometrics

The paper has been significantly improved, particularly in its structure. Most of my comments have been addressed, except for the two remarks below, which I believe are important for this article. Apart from these, I think the article is ready for publication. Congratulations to the authors for this preliminary work.

Discussion: 
L400-401: "Our results demonstrate a significant pairwise difference between ABM and LTD (p=0.001), but not between Dor and Keisan." This result is not presented in the Results section, and I believe it should be before it is discussed and compared with other literature.

Supplementary data: 
Regarding the duplicate points, I understand what you have done. However, have you tried something like this: tps_op_total <- readland.tps(XXXXX, specID = "imageID", readcurves = TRUE) tps_op_total <- tps_op_total[-c(8, 17, 18, 27, 28, 37, 38, 47, 48, 57, 58, 67),,] # Delete landmarks number 8, 17, 18, 27, 28, 37 XXXX because they are duplicates, like your number 12 and 25.

The article's format has significantly improved, which is commendable. The discussion now presents stronger arguments compared to the previous version, offering a critical analysis of both the data and results, which is highly insightful. I'm pleased to see that most of my previous comments have been taken into account, and I extend my gratitude to the authors. Despite being preliminary, this work has the potential to pave the way for further research.

• Specific Comments:
  • Line 27: Could you elaborate more on the conclusions?
  • Line 49: Consider using "Antiquity" instead of "antiquity."
  • Line 128: It would be helpful to include a legend explaining the zone referred to in the table.
  • Lines 140-141: The mention of "IA1/IA2" appears for the first time without explanation. It would be beneficial to standardize this throughout the text.
  • Lines 205, 201, 301: Please ensure consistent spacing with double spaces.
  • Digitization Error & GPA: I suggest rearranging the order of the paragraphs discussing digitization error and GPA. Since analyses Procrustes are already being discussed for error testing, this rearrangement would avoid repetition.
  • Lines 276-278: Bibliographic references should be reserved for the discussion section. However, it's worth noting that the percentage of error is similar to other studies.
  • Line 411: This result seems new (unless I missed it previously). If so, it should be presented earlier. The paragraph discussing variability due to topography is particularly intriguing!
  • Supplementary Data: Regarding geometric morphometrics analysis, it's crucial to address duplicate points. For instance, in the sliding procedure, landmarks 1 & 3 are duplicated. I suggest removing slidings 12 & 25 to prevent double-counting points, which could introduce bias.

Further, it's worth noting that placing the end of one curve, the beginning of the next, and a fixed landmark in the same position results in the point being counted thrice, potentially leading to bias. Additionally, after digitization, consider removing two of these points before analysis,(given that the 3 landmarks are supposed to be in the same place, and therefore with 3 times more weight for this point than for another)

The paper submitted by Harding et al. "The contribution of Mediterranean connectivity to morphological variability in Iron Age sheep of the Eastern Mediterranean", addresses the question of the movement of flocks by sea around the Mediterranean during the Iron Age. It examines this question by analyzing the morphology of sheep talus from two coastal sites and two inland sites in Israel and Cyprus. The authors hypothesize that the intra-site morphological variability of this bone must be more significant on the coastal sites than the inland sites due to the greater genetic diversity of sheep populations on the coast exposed to commercial networks. To highlight this variability, the authors use a geometric morphometric approach, which should better analyze the phylogenetic signal by isolating the variables of shape and size. Their results support the initial hypothesis, as the data show more significant variability at coastal sites. 

The authors present the issues related to the studied assemblages and remain cautious in the interpretation of the results. Except for one site (ABM), the number of sites and bones per site is minimal, and the chronology is very broad, covering several centuries with the pooling of phases.  
The study supports the idea that maritime connectivity is related to morphological variability in the Iron Age. Later studies should verify these results by analyzing more Iron Age II assemblages.
Despite the limited number of sheep talus and archaeological sites, this study merits attention. It explores a new approach using GMM to investigate the role of domestic animals in trade and commerce in the Mediterranean during the Iron Age, providing promising results.

Please, note that the following references cited in the text are missing from the bibliography:

Aubet, 2014 
Bartosiewicz and Lisk, 2018 
Briend and Humbert, 1980
Gambash, 2015
Gilboa, 2015
Gilboa et al., 2015
Gilboa and Sharon, 2017
Lehmann, 2001; 2021
Lehmann and Peilstocker, 2012
Nitschke et al., 2011
Panitz-Cohen et al., 2013; 
Panitz-Cohen and Yahalom-Mack, 2019
Raban, 1981
Raban and Galili, 1985
Raban-Gerstel et al., 2008
Raveh and Kingsley, 1991
J. F. Rohlf, 2017 
Sapir-Hen et al., 2014
Stern et al., 1993a 
Stern et al., 1993b 
Wachsmann and Raveh, 1984
Yahalom-Mack et al., 2018
Yasur-Landau et al., 2018 

This paper discuses the morphological variability of sheep astralali bones from Levantine sites to test the hypothesis that coastal sites will have a higher varibility of livestock introduced by trade. The authors use GMM to undertake detailed analysis of shape of the bones and, despite the small sample sizes and equifinality of the underlying causes of variation, this paper discusses is a novel application of the method. The paper is primilinary as noted by the authors but will be a useful starting point for more detailed analyses with larger samples sizes, better dated material and a comparison with aDNA data in the future will make results more meaningful. However, the paper is a very useful publication as a starting point for more detailed analysis. 

Preprint strengths:

Preprint weaknesses:

GENERAL COMMENTS:

Title

Abstract

Introduction, Materials & Methods:

1. Present the corpus, adding information on the temporal resolution of the data + the origin of the assemblages (paragraph 2 of the material and method section).
2. Present the GMM method + statistical tests and justify it.

Regarding the methodological presentation, in my mind it lacks many justifications for the choice of
statistical tests used and bibliographical references. I didn't find any references on GMM. These need
to be added.
A methodological presentation of the various statistical tests that are used in the results section to
discuss the data are missing: centroid size, ANOVA, permutation tests, allometry tests, visualisation of
unrooted morphological proximity trees.
Similarly, the method used to produce the average conformation patterns per site and their
visualisations is not explained.
The sites are not presented in the same terms. In introduction view, I recommend that you provide
additional information describing the topographical characteristics of the environment in order to
justify your choice of your samples when you write « we chose samples that derive from sites in a
similar topographic setting ».
I would also advise you to add information on the main findings of the archaeozoological studies to
give an overview of animal economy identified.
Furthermore, if it is possible, I advise you to add as co-authors of the publication your colleagues who
participated in the acquisition of the morphometric data on which your results are based.
Discussion:
Overall, interpretations need to be nuanced due to corpus heterogeneity (bones from totally different
contexts). Even though these arguments are set out in the third paragraph, I suggest that you detail
your hypotheses on morphotypical ovine diversity, taking these limitations into account.
I would also advise you to write some details of the interpretative problems associated with taxonomic
identification that might suggest the variability of the data within sites. Although it is doubtful that the
outliers from the LTD site were wild animals, this hypothesis cannot be refuted due to the anatomical
identifications.
Figure 5: explain in the text.

DETAILS COMMENTS:

Introduction

« to minimize possible functional morphological differences » = justify

« Our results, which represent the first study of animal mobility in the southern Levant using geometric
morphometric methods, should be somewhat liberally interpreted due to the primacy of this study »:
I think the term 'liberally interpreted' is completely inappropriate. On the contrary, as the data set
is heterogeneous and limited, and as no comparative study has yet been carried out on the influence
of maritime mobility on the introduction of new sheep varieties thanks to GMM, we need to be
extremely cautious about the results.

Methods

« osteologically mature at time of death » = to be justified due to the age identification problems
based on talus ossification.
« Ovis sp. » = explain why the identification remained at the gender level and why this may be
problematic in relation to the hypotheses formulated in the article.
Better justify the use of CVA.
Can you include the results of the PCA in the additional data?

Results

Overall, in this section, you should remove the comparisons with other studies and describe the results
objectively.
Digitization error
Can you also indicate the result obtained on landmarks alone (without taking sliding into account).

Centroid size
« The smaller size of the Dor specimens also resonates well with the results from other Levantine
coastal sites in the Iron Age (Chahoud et al., 2023). The specimens from Cyprus are especially small.
We do not know how universal this pattern may be on the island during the Iron Age, and this
observation requires further investigation » = for discussion
« The effect of the interaction between group and size on shape is not significant » = indicate the pvalue.
You should make the results of the allometric tests more obvious. I recommend adding this information
to the subtitle « centroid size » and “allometric pattern”.

Ordination:
In view of the CVA results, I doubt the need to add the tree calculated from the Mahalanobis distances
from the CVA. I advise you to add this graph and this information to the "supplementary" section.
« The group-based CVA ordination of the first ten principal components explains almost all the
variability in that dataset » pay attention to the formula. The first two axes of the CVA performed on
the first 10 dimensions of the PCA explain 98% of the variability.

Disparity:
Be careful with the formulation: here the data are calculated on the basis of Procrustes distances using
which method (PCA)?

« The similarity between the results of two different analyses—direct measurements of disparity on
landmark data and on eigen-ordinated coordinates—support the statistical results »: add the raw
data.

Figure 3: I advise you to divide up this figure for greater clarity and visibility. You could, for example,
present the boxplots and then present the conformation data in a second figure. In addition, I would
suggest that you remove the individual numbers from the CVA biplot and, if possible, use colour plates
to bring the groups together.

Figure 4B is not explained in the text: either explain it and therefore add the method that enabled it
to be carried out in the method section, or remove it. In my own view, this description is very important
in order to show the differences in shape that exist between the 'average' sheep astragalus at each of
the sites.

REFERENCES

All the references mentioned are appropriate and accurate.
However, GMM references are missing. You have to add some references concerning the origin of the
method like (Rohlf and Marcus, 1993, Bookstein, 1991…)
If you want, you can add too the PhD thesis of Manon Vuillien next to the reference of Colominas et
al., 2019 and Haruda et al., 2019 : https://www.theses.fr/2020COAZ2020
She works on the morphological variability of Late Neolithic and Iron Age sheep in Provence combining
GMM and more traditional approaches in archaeozoology.