PROBLEMS WITH MANUSCRIPT: In reviewing the figures of your paper, we note that the following changes would be needed in order for your figures to conform to the style of the Physical Review. Please check all figures for the following problems and make appropriate changes in the text of the paper itself wherever needed for consistency. * Figure 2: If units are defined in terms of another quantity, please precede this quantity by "units of," e.g., (a) -> (units of a). This will avoid the misinterpretation that the plotted quantity is a function of "a" (i.e., "a" is a variable). ---------------------------------------------------------------------- Report of the Referee -- AG11709/Chin ---------------------------------------------------------------------- The manuscript provides a study of the temperature obtainable by polarization gradient cooling in a tightly focused optical dipole trap. As expected from the behavior of the light shift for the ground-state Zeeman levels, the minimum temperature is obtained for linear trap polarization. This behavior is also supported by model calculations of the cooling force by solving the master equation. The short article presents a detailed systematic evaluation of this kind of cooling in tightly focused traps that to my knowledge was not published before. I thus recommend publication after a few minor modifications. The introduction should also mention Raman sideband cooling, that was also employed in several of the references. The x, y, z orientation of the beams could also be indicated in Fig. 2. In several places the “orientation of the PGC field” (the E field?) is mentioned, where actually the direction of beam propagation is relevant. The residual magnetic fields and the corresponding shifts should be given. The physical reason for the missing tensorial shift in the case of the employed atomic states should be mentioned. From the observed significant sensitivity to the purity of the linear trap polarization, I wonder, if the strongly suppressed residual HFS induced vectorial shift might still be relevant. Giving an estimate of its magnitude could be helpful. Ref. 12 mentions a vectorial shift even with linear polarization due to the non-paraxial focus in the large-NA trap. In the present experiment the conditions seem to be similar. What would be the corresponding influence on the level structure and the temperature? Fig. 4: The error bars indicate statistical uncertainties. Is there any estimate of systematic uncertainties in the temperature measurement?