The authors have provided a great deal of feedback to the reviewers and have outlined changes that will be made to the manuscript that address, for the most part, the issues raised by the reviewers.

I will not got through the list of changes that will be made as these are numerous, but the revision needs to more clearly set out the exploratory nature of this work and highlight associated caveats more clearly.

I am in agreement with the comments of Dowsett and Lunt on the use of the interpolated SST field from the PRISM data set. The real data are provided by the point based SST estimates and these should be used, I accept that this may introduce other uncertainties. I recommend retaining the whole-globe approach already included in the paper, but also to include some new analysis where the authors just compare with the actual SSTs at the sites of the data (and including discussion of the weaknesses of that approach too, e.g. spatial inhomogeneity and clustering of data).

There has been substantial discussion of tropical temperature stability within the data community over the years and recently. The recent review by Haywood, Dowsett and Dolan in Nature Communications may provide a useful place to start in order to incorporate an enhanced discussion on data uncertainty in the tropics.

The complications of the SSTs being an average of a time slab rather than a discrete time slice are touched upon but I think could be developed further. The work of Prescott et al. 2014 in EPSL has demonstrated the degree of surface temperature variability around discrete interglacial events within the mPWP. This can be substantial. In light of those results I think it would be useful to see an enhanced discussion on the degree to which the approached used here (models with time specific forcings compared against data time averaged over 300 Kyr) can actually inform us about climate sensitivity in the Pliocene.

Dear Julia,

Many thanks for the revised manuscript.

I am happy to accept the paper for publication but could you please update the reference to Haywood et al. 2015 to 2016 as the full paper in CP is now published.

Many thanks,

Alan.