An important test of the functions of competition in elephant grouping dynamics.
Abstract: Not sure what this means: “social variation in female-bonded species”
35: we found… (where) unclear relative to neighbouring forests?
Intro:
There are two separate questions posed in the introduction: evolution of sociality (and I would probably cite Jarman 1974 especially for a non-primate perspective rather than Terborg or Crook) and why some species are female-bonded, vs male-bonded or no bonds. It appears that there is an assumption that all species that are part of the EMFSR model applied here are female kin bonded? Wrangham’s original model was to explain differences between Ceropithecoids and apes (male bonded)… Strier (2008) does an excellent job discussing female dispersal and fission-fusion. The theoretical introduction could be made much clearer and focused on the questions you address. Perhaps start with elephants rather than the ancient history of social evolution.
“Elephants, both Asian and African, live in female-kin units (families or “groups” in Asian elephants; refs) with extremes of fission-fusion in larger aggregations (called “clans” in Asian elephants: refs). This makes elephants an excellent species for testing underlying hypotheses about the nature of female-kin bonded groups (e.g. refs). Understanding the dynamics of competition among females in kin units will contribute to refining our understanding of the processes that result in the evolution and maintenance of female-bonded social groups (e.g. refs)
Or something a bit more relevant to your study. And note that the literature on ungulates is replete with examples of female-bonded groups, while Strier would argue that even in primates, this model is more myth than evidenced (Strier, K. B. (1994). Myth of the typical primate. American Journal of Physical Anthropology, 37(S19), 233-271.) Also cetaceans, macropods etc.
Recent tests of Jarman 1974= Szemán, K., Liker, A., & Székely, T. (2021). Social organization in ungulates: revisiting Jarman’s hypotheses. Journal of Evolutionary Biology, 34(4), 604-613.
Basically my suggestion is to focus the introduction on the really interesting tests that you can carry out rather than so much on theory.
Also while a bit dubious the work of Dublin on within family competition in African elephants is relevant, as is the work by Archie et al. on dominance. The point being that elephants show more about elephant social evolution than primates do?
127: previously, rare agonism does this mean that the agonism was unexpectedly low? Meaning is not quite clear.
131: it’s clearly noted in Archie et al. 2006 that the differences in rates of within-family competition in savannah elephants were associated with large, contestable resources. And since Wittmeyer et al did assess resources and dominance, the statement is not valid. Your point is that you are uniquely placed to enable a detailed exploration of resource quality and distribution.
140: please ensure that you have controlled for opportunities for interactions to occur. E.g. clan sizes and parties / groups are similar.
250: replacing missing values is a bit dubious. Why not just run the models with the existing data? These are comparisons of variance in Anova? Not MANOVA? Or use GLM which doesn’t require matched samples?
255: please note your individual ID process and how clans etc were “known”. All animals present had ID? Age, sex etc? Details are lacking for crucial information relevant to your hypotheses.
279: refer to ethogram or some behavioural category definitions. Was a head raise a threat? No ear threats? No avoid after approach? Clarify how you defined “agonism”. Maybe provide a table of definitions in supplementary materials.
291: please refer to pp starting line 340 to specify how rates were corrected for differential observation time of clans and for the observation time for each clan.
297: Some explanation of a “non-independent” interaction would be helpful. Why not simply assume that all interactions were important in between clan dominance? Why use the cut-off? Was there an analysis that showed that if there was one interaction, it was followed by another within 2.5 hours? In the analysis (413) it appears that elephants engage in bouts of aggression? These distinctions make the use of a ratio (NI/I) problematic. What does this ratio actually mean?
If you present an R value, you don’t really need the R2 as it is R squared….
501: please see my comments re introduction. This statement is not accurate.
504: nice, hence elephants use fission-fusion!
525: obviously food patches, and quality, will affect rates of competition, but so will opportunities to compete. Again, elephants use fission-fusion to minimise (but not entirely remove) competition. Elephants also group! Which leads to competition.
533: agonism is NOT common in savannah elephants. It is very rare compared to all other types of interaction. And the dominance hierarchies within families are age-graded (not steep), those between families are of unknown source. Savannah elephants may use specific and very subtle cues avoid overt interaction (head lift, head turn) which could be sufficient to initiate a move away. Thus we really have no idea of rates other than overt aggression.
Note that Archie et al showed that Tarangire elephants were competing over point resources (trees, etc), while grass-eating Amboseli elephants were less competitive.
546: high density compared to what?
548: once every three times groups were together is hardly frequent!
585: agonism between females “from different clans” “compared to that between females from the same clan” (there are too many betweens and withins to make sense here).
628: but you showed LOW agonism within clans, so this argument doesn’t hold.
And furthermore, the tactic of avoidance of others may be sufficient to reduce the actual rate of interaction even when resources are patchy, but the patches are widely distributed.

Thank you for your attention to reviewers' comments. You raise some very interesting issues about our lack of clear information on patch distribution, quality and temporal availability in the previous literature, for primates and elephants. It might be worth a collaboration / comparative study. I enjoyed the paper and it should make an important contribution to our understanding of the dynamics of sociality.
You might consider using GLM in the future with larger samples so that you can more clearly rule out an effects of repeated individuals, families and clans.