Speckle tracking are well known methods in ultrasound imaging to estimate the motion of each pixel from consecutive frames. Due to the low lateral resolution of ultrasound images, many methods of speckle tracking rely on interpolation to generate subsample accurate displacement. This interpolation is commonly performed on the correlation map generated by the sample level cross correlation calculation. This study applies interpolation to the correlation map, a standard speckle tracking practice, and compares this to data that instead interpolates the radio frequency (RF) data before correlation, which is shown to be beneficial. Estimation accuracy and the shape of the correlation map are compared for both interpolation of the correlation map and interpolation prior to correlation. These methods are tested on data sets generated by Field II to simulate an elastography phantom, with subsample lateral displacement. The RF interpolation method is more accurate than correlation interpolation with cubic spline fitting, showing that interpolation at the correlation level is not perfectly analogous to correlating an interpolated RF image.