Not yet reviewed by Pith; the record is open.

The limiting magnitude of the HST data set used by Cochran et al. (1995) to detect small objects in the Kuiper belt is reevaluated, and the methods used are described in detail. It is shown, by implanting artificial objects in the original HST images, and re-reducing the images using our original algorithm, that the limiting magnitude of our images (as defined by the 50% detectability limit) is $V=28.4$. This value is statistically the same as the value found in the original analysis. We find that $\sim50%$ of the moving Kuiper belt objects with $V=27.9$ are detected when trailing losses are included. In the same data in which these faint objects are detected, we find that the number of false detections brighter than $V=28.8$ is less than one per WFPC2 image. We show that, primarily due to a zero-point calibration error, but partly due to inadequacies in modeling the HST'S data noise characteristics and Cochran et al.'s reduction techniques, Brown et al. 1997 underestimate the SNR of objects in the HST dataset by over a factor of 2, and their conclusions are therefore invalid.