Sea surface temperature (SST) estimates using the delta (super 18) O composition of fossil planktonic foraminifer calcite, within the time slice 3.12 to 3.05 Ma (Pliocene, Kaena Subchron-C2An1r) are assessed for nine Atlantic Ocean sites. These are compared with SST estimates from fossil assemblages for the "Time Slab" 3.29-2.97 Ma and with estimates from a fully coupled ocean-atmosphere General Circulation Model (GCM) for the same time interval. Most SST estimates derived from the delta (super 18) O data indicate a cooler ocean surface than at present, through the latitudinal range 69.25 degrees N to 46.88 degrees S. At some sites the temperature difference is greater than 5 degrees C (cooler than at present). This contrasts with SST estimates from fossil assemblages that give warmer than present temperatures at mid- to high latitudes, and similar temperatures in the tropics, and with the GCM, which predicts SSTs warmer than at present across all latitudes for this time interval. Difficulties interpreting the ecology of fossil foraminifer assemblages and inaccurate estimates of mid-Pliocene seawater delta (super 18) O composition (delta (super 18) O (sub sw) ) at some sites may partly produce the temperature discrepancy between isotope-based and fossil-based SST estimates, but do not adequately explain the cool signal of the former. We interpret the cool SST estimates from the delta (super 18) O data to be the product of: (a) calcite formed at a level deep within or below the ocean mixed-layer during the life-cycle of the foraminifera; (b) secondary calcite with higher delta (super 18) O formed in the planktonic foraminifer tests in sea bottom pore waters. Although these effects differ between sites, secular and temporal oceanographic trends are preserved in the primary calcite formed in the mixed-layer near the ocean surface, witnessed by the latitudinal variation in estimated SSTs. Reconstructing accurate mid-Pliocene SSTs with much of the existing published oxygen isotope data probably requires a detailed re-assessment of taphonomy, particularly at tropical sites. This study also indicates that methods for estimating Atlantic Pliocene delta (super 18) O (sub sw) need to be refined.