Slightly later, it was also realized that proper bias subtraction occasionally requires the use of separate bias values for odd and even columns. WFPC2 data taken before May 4, 1994 did not use this form of bias subtraction. As a result a striated pattern with a typical amplitude of a few electrons (a fraction of a DN) may remain in these images. These data will benefit from recalibration with calwp2, version 1.3.0.6 or later, unless the signal of the observation is so large that the noise statistics are dominated by Poisson noise. A recent study of the properties of bias frames throughout WFPC2 operations can be found in the WFPC2 ISR 97-04. From their results, it appears that large differences between the bias level in the even and odd columns have been rare after 1995, thus it is unlikely that recent WFPC2 data suffer from this problem to any measurable degree.