Using waveform cross correlation, we have measured ~17,000 P and S wave differential travel times from earthquakes and receivers near a great circle path from Japan, across Alaska, to North America. We jointly invert the data for variations in shear and bulk sound speed and derived Poisson's ratio variations in the mantle beneath 1000 km depth. We explore the model space using different levels of regularization and our conclusions are based on a range of models that give acceptable data fits. The correlation between vS and vP variations is good to ~1500 km depth but gradually degrades in the bottom 1000 km, whereas the ratio R = ∂lnvS/∂lnvP is between 1.0 and 1.8 at most depths. Our data suggest that there is no significant correlation between bulk and shear wave speed, but they appear anticorrelated in several regions. We estimate the effect on the Poisson's ratio of changes in temperature, iron content, and the magnesiow¿stite (mw) to perovskite (pv) ratio. At all depths the effect of temperature is largest and that of the pv to mw ratio smallest. Our data cannot resolve the trade-off between thermal and compositional effects, but explaining the Poisson's ratio variability by temperature alone would require unrealistic perturbations and it would not explain the inferred anticorrelation between shear and bulk sound speed. Our results suggest that despite R being significantly smaller than 2.5 we need a combination of thermal and compositional effects in order to explain the elastic parameter variations in the deep mantle beneath Alaska, with realistic values of ΔT of ~300--800 K, variation in XFe of ~4%, and pv enrichment (or depletion) of up to ~10% just sufficient for explaining the data.