Nathan's suggestion of multypling the lrecs by the nrecs for "worst case" is quicker than what I've always done to estimate this; take the lrecs, 655, assume all the lines are 65 columns (more if you're a pessimist) so multiply by 65 to get the number of characters and divide by 8 (bits per character) so 655*65/8=5322 bytes, close to Nathan's method. I figure it's inflated, but if you're getting the stuff via email you have to allow for the header, and the inflation usually amply covers it. My reasoning may be faulty, but it works and the greater the lrecs the closer it seems to come to being quite accurate, in my experience. Douglas Douglas Winship Hays County, Texas [log in to unmask] Secondary AUTOCAT Listowner MEDLAB-L