Both reviewers have commented favourably on the value of the very detailed, multi-proxy and high-resolution study provided in this publication. It is clear that there could be some very valuable new information revealed by the data, but at present the relevance and support for those linkages is masked by a number of factors: (1) a lack of clarity around several components of the age control; (2) a lack of clarity on how the productivity proxies are interpreted, and what processes drive each of these; (3) difficulties in following the discussion where links are made between the East Asian Monsoon, Greenland, and Antarctica. I agree with the reviewers, and think that they make a compelling set of arguments for the corrections required for the manuscript text. Due to the density of data being presented here, it is also likely that very detailed graphics will result. But, as the reviewers also pointed out, some of these become overwhelming to negotiate. The authors have replied to the comments provided by the reviewers, in places clarifying their interpretations and/or methods. However, I still found several components of the discussion to be confusing, or not clarified to the extent that was sought by the reviewers. I have outlined below what is required to make a convincing case that a robust set of arguments is put forward in this manuscript, which will require more clarification that the information currently given in the authors replies to the reviewers. As a result, I recommend that the manuscript is reconsidered only after major revisions are made.

1. Age model construction
I agree with both reviewers that there are some serious concerns around the construction of the age model and how it is outlined in the main text. It is difficult to follow exactly what was done, where points were transferred or not, why certain offsets were considered, and how the tuning of productivity cycles was tested. The replies given to the reviewers are not totally convincing: a revised submission needs to very carefully lead the reader through each step, and justify the approach. There is a discussion of benthic-planktonic offsets and whether they vary through time – the authors reply that above 15 ka this offset was constant in the core from which they transfer their age model. But there are fluctuations shown in the Max et al. (2014) paper which also become pronounced before 15 ka – and this new manuscript includes older samples which might have been affected by these intervals. Furthermore, if there are large changes in productivity (which might reflect changes in mixing and/or carbon export) could these not also have affected these planktonic-benthic offsets?

A revised submission needs to more carefully outline the steps taken to generate the age model and the assumptions which have gone into it. There is confusion about where tuning to Greenland and/or East Asian monsoon records might have occurred: this needs to be clarified. I agree with the reviewers that the construction of the age model would be better placed before discussion the productivity records, because then the text can flow into the timings of the productivity changes not just the overall pattern. This could be achieved if there was an outline of what the different productivity proxies represent and so we might expect that e.g. b* can be used to tune between sites. At the moment, the text talks about the timings of changes and cycles in the productivity data BEFORE the age control has been justified.

2. Productivity proxies and their interpretation
A number of different proxies for productivity are presented. The paper does not review the benefits and limitations of these different proxies, to make a clear statement of how they should be interpreted. This means that, where the reviewer comments that different proxies do different things at different times, it is then difficult to know which record to accept, and when, and why. A link needs to be made at the site first between what these proxies mean, and then how the oscillations are interpreted. The reviewers may have been able to understand the tuning approaches and the descriptions of productivity trends if there was a stronger rationale of the data interpretation. In a revised submission I recommend that the interpretations of the proxies are outlined with more detail so that the reader knows how you are finding a productivity argument (e.g. preservation can affect several of your indicators but is not discussed).

I am not totally convinced by the use of the ‘stack’, but that is because without a revised manuscript to see what you did it is not clear to me what this represents. Is this an average of normalised data from several proxies? Is it a weighted mean? And is this even an appropriate way to create a kind of summary of the data, when you note in the text that sometimes the proxies behave in different ways? Given the complexity of the data, did the authors try any form of additional statistical analysis to determine whether there are 'groups' of productivity indicators or patterns?

In a revised submission you will need to very clearly explain what went into the stack, why those parameters were chosen, whether there is susceptibility in the stack to one or two of the records etc.

3. Teleconnections
Reviewer 1 makes valuable comments about the way that the text links the monsoon and Greenland records and/or with Antarctica. I agree with their comment that a slight restructuring or separating of the sections across lines 243-285 would be valuable, to first set the NW Pacific data into a hemispheric context, and only then discuss links with the south. This doesn’t have to be a problem for the figure (as the authors suggest): more than one section of the text can cite the figure. I also agree with reviewer 1 that explaining the longer term trends first, in your interpretation, would allow you to move from the long term into the short term variability. The structure might then flow better for the comparisons with wider teleconnections and their timescales.

I also agree with reviewer 1 that when making the cross-correlations for teleconnections (see below), what would make a stronger discussion of the significance of your data is a cross-correlation of this new data with the monsoon/Greenland records (whichever is appropriate). First, find the mechanisms and links with your own data, and then (possibly) make the bigger hemisphere/polar comparisons.

4. Graphics
A positive aspect of this publication is the number of detailed new data sets, showing multiple indicators of productivity at high temporal resolution. The reviewers made some suggestions for how to simplify these – the use of a ‘stack’ in a revised figure 3 was one response by the authors. I agree with the reviewers that the graphics are still complex. One way to try and simplify e.g. figure 3, might be to plot the data sets that you are tuning side-by-side e.g. plot both RPI records together, then both b* together, or whatever the combination might be. Or in fact was it the ‘stack’ that was used for the tuning.

The revised figure 4 is more clear. Figure 5 is a bit strange/messy with simply an addition of the Bond 2001 hematite grains overlying Antarctica. I’m not convinced that Figure 6 is needed, unless you include (as one reviewer suggests) the cross-correlation of your own data into this. Discussion of monsoon/Greenland climate phasing is an important question, but to better embed this into your new findings you need to include a component of your data here. Or, remove the detailed discussion of the wider teleconnections from the manuscript and focus more strongly on your new data and its relationships to e.g. monsoon, Greenland, Antarctica.

The reviewers noted that there is other data on e.g. SST, sea ice, available which could help you to make the links between the mechanisms you outline to change productivity and the potential data which could support these. In your reply to reviewers you say that you will state the relationships: I don’t think that this is what the reviewers are asking for. I, like them, would prefer to see you plot the SST and/or sea ice evidence alongside your own.

5. Written style
Both reviewers comment that there are grammatical errors, leading to confusion and/or a lack of clarity in the arguments and presentation of data. I agree with this comment. A thorough and careful reading of the manuscript is required to remove these errors, to ensure that your arguments are being made with improved precision.

The authors have replied to both reviewers, and submitted a revised version of the text. The main issues raised by the reviewers and the editor centred on:

1. Age model construction. This is generally improved in the revised submission, but the authors still resist moving the construction of the age model before discussion of the proxies. I agree that it is difficult to identify the correct ordering here, but since there is no discussion of what the productivity records mean before they are tuned (this comes later), the structure of the text is very difficult to follow. If the authors wish to keep the age model after the discussion of the productivity proxies and what they might mean, so that there is some process-based rationale for why you might expect different records to align with other sites and with the monsoon. This is particularly key when one considers that the different proxies don’t always show the same patterns, so that a reason why some peaks were tuned and not others needs to be stated. A revised submission needs to either (1) place the age model construction first, before discussion of the productivity signals; (2) explain first what the proxy signals record, to make the link to why tuning could or should be appropriate (see next point)

2. Productivity proxies and their interpretation. The authors do now include some additional information on how the productivity proxies could and should be used, but this needs to come earlier in the discussion, to justify why these different proxies were used and how they have been interpreted. This is particularly important for the generation of the stack, and what it means. I don’t agree with the authors that this is the subject of a different paper: you are using these data to make bold statements about NW Pacific productivity changes, but as Reviewer 2 points out, the selection of which peaks are being used seems to vary through time and be quite arbitrary. It needs justification. This would also address the concerns of point 3 below, where the focus on the NW Pacific would be a better demonstration of the value of the next records, but give detail in how we should interpret them, especially where they may disagree or show variable responses.

3. Teleconnections. I still agree with reviewer 1 that you should undertake the cross-correlations for your own, new data, rather than shifting to an analysis of previously published records of monsoon variability. Yes, you have tuned your data to the monsoon data, but what would be most interesting here is to learn whether the productivity in the NW Pacific Ocean could be linked to millennial scale events elsewhere (see also comment by Reviewer 2 that section 5.1 is largely tangential – I agree). You note that not all of the same proxies show the same patterns: would it not be interesting to detail those relationships, which may evolve through time and give some interesting insights into what controls NW Pacific Ocean circulation and productivity? There must be factors beyond monsoon intensity which affect the regional primary productivity, which could be interesting to learn. Here is where the SST data could be very useful (Reviewer 1), rather than the repeated assumption throughout the revised text that productive equals warmth (cold oceans can be productive, depending upon the driver). Reviewer 2 also raises concern about the circular logic of tuning to the monsoon and then describing the synchroneity of the events between the two regions. The focus on the NW Pacific data and its variability has not been addressed in the revised manuscript, and I continue to be concerned that the main data analysis and discussion in the document is not about the new data which has been presented here. A revised version needs to address this concern by focussing more on the new data and less on the monsoon-Greenland teleconnections.

4. Graphics. These are better in the revised manuscript.

5. Written style. Whilst some of the corrections made have clarified the text, there remain areas of grammatical errors which require careful editing. I have highlighted a number of these on the tracked-changes document, but there is work to be done here.

Dear authors,
Thank you for your careful consideration of the detailed comments provided by the two referees and myself. The restructuring of the methods and the discussion have made this a much easier manuscript to read, and the novelty of your new data and new findings are much stronger now. Some minor grammatical corrections are noted on the attached pdf, which should be quick to rectify.
Best wishes,
Erin McClymont