The dataset includes stable oxygen (δ¹⁸O) and carbon (δ¹³C) isotope ratios and magnesium to calcium (Mg/Ca) ratios in Globigerinoides ruber (white), reconstructed sea surface temperatures and salinities and planktic foraminifera assemblages of Caribbean sediment core M35003-6. δ¹⁸O, δ¹³C and Mg/Ca data were obtained using 60 specimens of G. ruber (white) sensu stricto picked from the >150 µm fraction. δ¹⁸O and δ¹³C values in G. ruber (white) were measured using a Finnigan MAT 253plus mass spectrometer coupled to an automated carbonate preparation ("Kiel") device at Isotope Laboratory at Faculty of Geosciences, University of Bremen. δ¹⁸O and δ¹³C values are expressed relative to the VPDB standard notation (VPDB). The long-term 1σ analytical uncertainty of the standard analyses was better than ±0.07 ‰ for oxygen and ±0.04 ‰ for carbon. Mg/Ca values were measured using a Perkin Elmer Optima 3300 R inductively coupled plasma optical emission spectrophotometer (ICP-OES) at MARUM, University of Bremen. The analysis was performed following the cleaning procedure described in Mohtadi et al. (2014, doi:10.1038/nature13196) and Barker et al. (2003, doi:10.1029/2003GC000559). Mg/Ca values are expressed in mmol/mol. The average 1σ analytical precision is ±0.01 mmol/mol. Replicate samples (n=3) revealed an average 2σ standard deviation of 0.11 mmol/mol. SST, SSS and their uncertainties were calculated in PSU-Solver (Thirumalai et al., 2016; doi:10.1002/2016PA002970) from coupled δ¹⁸O-Mg/Ca values in G. ruber (white) using equations of Kısakürek et al. (2008, doi:10.1016/j.epsl.2008.06.026; Bemis et al., 1998, doi:10.1029/98PA00070; LeGrande and Schmidt, 2006, doi:10.1029/2006GL026011). The average 1σ (2σ) uncertainties are about ±0.5 (0.8) °C for SST and ±0.3 (0.6) units for SSS. These errors include calibration errors, 2σ combined analytical and sampling uncertainty for Mg/Ca (±0.11 mmol/mol) and 2σ analytical uncertainty for δ¹⁸O (±0.14 ‰). Planktic foraminifera species and lithic grains were counted in the size fraction >150 µm; minimum of 600 specimens has been identified per sample. Large specimens of N. dutertrei were counted together with smaller juvenile specimens referred to as N. dutertrei-N. pachyderma (dextral)-intergrades.