This is a summary of the major corrections made on the article after Andy's first set of comments. (See files andy_comm.txt and corrections.txt in this directory). Some comments are added whenever I deviated from what Andy's requested. General comment: --------------- The paper now tries to put it clear that the ideas motivating this work are purely empirical. The title was modified to reflect this. Andy had very fair worries about the lack of physical implications of the analysis, but the paper is aimed at something else. I think the modified text should convince the reader of what our goals are. Modifications: ------------- 1- Figure 1 now shows feature definitions at different epochs. This provides a much clearer hint on what's been measured. 2- Section 3.3 added. It describes some tests done to see the possible effects of (a) reddening, (b) resolution, and (c) noise. Andy requested me to add tables with resolution and S/N for each spectrum. I decided just to mention in this part of the text the range of values these quantities take. A table would be cumbersome to make and its use is questionable. Additionally, I measured the same EW's using IRAF as a sanity check. Results coincide. I gain reliable error-bars using my code. 3- The abstract has been modified to include more concrete conclusions. 4- The beggining of the introduction was changed, trying to make it clearer, more concise and with references to past work. 5- More attention is payed to line identification when describing the conformation of features and their evolution. For example, the gross oversight of the contribution from Ti II in subluminous SNe is fixed. Also, probable identification of "Ca II H&K" as Ni II (Fisher et al. 1999) is mentioned. Andy asked me to add a table with rest and observed wavelengths of those lines appearing in SNe Ia spectra. Again, I don't see much of a point in adding such a table if things are mentioned in the text. On top of that, observed wavelengths of individual lines in a complex feature should come from model fitting. I might have missed Andy's point here. 6- The fact that classification of high-z SNe Ia relies on Si II 4000 is now explicitly mentioned. 7- The special comment on SN 1999bm showing a remarkably broad "CaII H&K" absorption around maximum light may seem minor but I think it is worth mentioning, even before some serious study is done. Andy suggests it might present a high velocity Ca II component, as SN 2001el does. Unfortunately, the study of SN 2001el is based on CaII IR triplet which our spectra of SN 1999bm barely cover. Vice versa, early spectra of SN 2001el don't cover CaII H&K. I wonder if there's more work on this HV component of CaII H&K in particular. 8- As suggested by Andy, every mention to "over-luminous SNe" was changed to the more appropriate "1991T/1999aa-like SNe". This is because the distinction is based on spectroscopic peculiarities. The term "under-luminous" is kept for SN 1986G, SN 1991bg and SN 1999by, because they are proved to be so. Before, I used the phrase "luminosity sequence" to speak about what's seen in Figures 4, 6 and 8. This has been changed to "a trend of increasing \Delta m_{15}(B) with decreasing EW". The doubtful assumption of equating high luminosity and low \Delta m_{15}(B) (or high stretch) is thus rectified. Final comment (References): -------------------------- I added some references all over the text. I don't consider this issue completed at all, but I may need the input from other people to finish it. In some places I added bold-faced signs where I know I need references I didn't find. This is especially the case of the sources for the public spectra. I'll see if I can gather all that info by myself but I suspect that some people in the collaboration must have this info at hand. I should say I took some of the published spectra from Peter and Andy. The rest come from the Asiago Catalog. Gaston Folatelli.