1. The only significant science point where I disagree is the discussion on the bottom paragraph of p13, where you write that the 24 micron emission is more tightly coupled to the TIR emission than the PAH flux. I do believe that is the case for this sample, because of the suppression of PAH's in the low-metallicity systems. But a number of studies, for example George Bendo's recent paper (Bendo et al. 2008, MNRAS, 389, 629) show that in normal metallicity systems the PAH emission may well be a better tracer of the cold dust emission than 24 microns. I think the confusion here arises because Daniela's studies were for HII regions, not galaxies as a whole. My suggestion would be to drop the sentences comparing the tightness of 8 microns vs 24 (apart from the discussions of weak PAH emission at low metallicity), and leave it at that. KINGFISH should be able to sort this out in a definitive way. --Fair enough. I have removed the direct comparisons between 8 and 24um, and have added references to work showing that PAH actually traces far-IR/submm well for normal metallicity galaxies. 2. Figure 12: I have been mulling this over too. As I mentioned in Long Beach Monica Relano was seeing similarly strange UV colors for some M33 HII regions, and she traced part of the problem down to an error in the AB magnitude zeropoint for FUV or NUV in some of the GALEX documentation. I forget the details but will ask her about it, and we can see whether there is a similar problem in our data. I suppose it is also possible that the colors are fine, and what we really are seeing is not a left-to-right shift but simply a vertical shift, reflecting the fact that most of the LVL galaxies have low extinction but also intermediate UV colors, due to more extended SF histories than starbursts. I agree this needs more thought. --I have tried to rework this section to better explain the possibilities. Back to comments in order in paper 3. p4, first par, line 18, "The total sample encompasses..." By total sample I think you are refering to the LVL sample. However the text immediately preceeding is describing the 11HUGS sample, so this is not completely clear. Since this is a way long paragraph anyway is there a way to separate the descriptions of the LVL and 11HUGS samples more clearly. Maybe simply starting a new paragraph with the sentence above would do the trick. --I have created a paragraph break here, and with Janice's help have clarified things quite a bit. 4. p4, bottom, reference to Table 1: It would be good to have a clear statement somewhere, either in the text or in Table 1 itself, of what the columns are and where the data were obtained. I realize there is a note at the bottom of Table 1 that includes some of this information, but it is incomplete. For example what are "2a" and "2b" on the right side of the table? Are they the measuring apertures, or the R25 dimensions, or something else? And how is the position angle defined, from north to east? --A more explicit/complete description is provided as a table note. 5. p5, middle par, line 3: Maybe put a tilde in front of 5 Mpc, as some galaxies with larger distances have distances from TRGB, etc. --Done 6. p5, middle par, last 4 lines: I can't recall whether this draft includes Janice's new text or not. Just to be sure, the current version does point out that there are 4 galaxies with d>11 Mpc in the sample, but we also need to say that there are some other galaxies with currently estimated d<11 Mpc that are not in the LVL sample. I think Janice gave you text for that but I mention just in case. --Done 7. p5, last 2 lines: "focuses more on" --> "includes more"? --Done 8. p7, end of top par: Is the meaning of the last sentence that short exposure corrections were only needed for these galaxies, or is it that even after the short exposures were applied there are problems with these galaxies in the nuclei? Would help to add a few words to clarify. --My intent was that the effects of IRAC saturation in LVL have been corrected. However, M82 is another story. --> I clarified the text so that the reader will better understand that saturation has been corrected. There is a caveat in the note to Table 2 that M82 photometry is less reliable. 9. p7, Sec 3.2, par 2: This is a really minor point-- the examples of large galaxies given all have archival data, so they really don't apply in this section. On the other hand you obviously don't want to repeat this discussion in Sec 3.3. Probably just ignore this point(!), or maybe add a line to this section and 3.1 to say that the same reduction processes were applied to the archival data (but not LMC, SMC, and M31, right?). --This was an insert suggested by Chad, and since I believe he and Karl led the processing of LMC/SMC/M31, he's the authority on whether or not our MIPS processing mimicked that of the large galaxies. As for IRAC, the processing of SINGS, large galaxies, and LVL is a bit different here or there, since different folks led those different efforts (e.g., I believe we do a better IRAC saturation correction). So it's hard to make a blanket statement that consistently addresses both IRAC and MIPS in Section 3.1. I ultimately decided to not change the text. 10. p8, end of top par, "is only a few percent." Any reason not to give the actual percentage range? --Done 11. p8, Sec 3.3: You might want to mention here that pipeline S18 was used for IRAC data, after all the fuss it has caused! --Done. I didn't bother before since I mention we do the same as for new data, and in the new data section I mention S18. But it definitely won't hurt to mention this again. 12. pp8-9, Sec 4.1 (2MASS): What is missing here is a statement of how accurate the new fluxes are. I guess the reader could figure this out for themselves by scanning through Table 2, but a few general words would be useful. Since we show that the large galaxy catalog has errors of up to factors of hundreds, people will immediately wonder what the errors in our measurements are. --I moved the blurb on calibration uncertainty up to the Section beginning of Section 4, adding what I've done for 2MASS. I've also added a brief description on 2MASS uncertainties in Section 4.1. 13. Table 2: Maybe I am especially thick, but I became very very confused by the fluxes in this table, because I thought the exponent only applied to the error, and not the flux itself. So I could not understand why you could measure a flux of 5 Jy to less than 0.001 Jy accuracy. Eventually I figured it out, and probably not worth changing. But if someone else is tripped up maybe we have to explain this somewhere. --Good point. I added an explanation in the table footnotes that addresses this. 14. p9, Sec 4.2, par 1, next to last line: It might read more easily with spaces after the commas (for the numbers in brackets), but I realize this is pretty arbitrary. --Done 15. p9 general: I could not find anywhere an explanation of how the aperture sizes were chosen-- are they based on R_25, or based on growth curve measurements at some IR wavelength, or something else? --It was at the beginning of Section 4.2.2. To make it more prominent I have moved it to the top of Section 4. And I have added comparisons of our apertures to R_25. 16. p10, equation (3): I suspect that this equation tends to underestimate the photometric uncertainties, because there will be additional errors from spatially variable sky (from flatfielding errors, scattered light, cirrus, etc) that are not captured in the equation. Is this worth worrying about? --For most upper limits I used the sigma_sky listed in the manuscript. But I also raised the value of sigma_sky in many instances where the background spatial variability is higher. So in that sense I've tried to account for the additional concerns you list above. I've added such a note to the text. 17. p11, top par, lines 6-7: "as very few MIPS detections are available fainter than this magnitude limit" --> "because very few galaxies fainter than B = 15.5 were detected in MIPS."? --Thanks 18. p11, end of same par: Is someone on our team going to do a stacking analysis? If not it is fine to give the suggestion, but if someone is doing this you might want to cite that future paper or leave out the sentence, to prevent a reader from doing it right away. --I put in Fabian Walter et al. as the lead for our stacking analysis, since he plans to do this in concert with MIPS/HI comparisons. 19. p11, Sec 5.2, middle, and Figure 7: A bit more explanation here might not hurt. I have a few questions: a) Why are so few points plotted (<<100)? Did IRAS fail to detect most of our 258 galaxy sample? --There are a total of 70 LVL galaxies with 25micron detections, necessary for all three panels (and 126 detections at 60microns). I state in the text that a subset of 70 galaxies have secure IRAS and MIPS detections. b) The rms scatter must be something like +-20% for 24 vs 25 microns and >30% for TIR, much much higher than our quoted errors and probably more than the quoted IRAS uncertainties. Do we understand this? The text just says casually that IRAS data have large errors, but if so why is the scatter nearly independent of flux? --It's good that you pointed this out, b/c it needs to be addressed. It turns out that the 5 "worst" offenders in the lefthand panel are all problematic in their own ways: LMC & SMC - huge; not our LVL data M51a & b - pair proximity makes IRAS data suspect (baseline, etc) NGC 0253 - saturated at 24um, and I didn't catch that until your comment made me look closer at the data. Chad hasn't addressed it yet nor mentioned it in a data delivery document since this galaxy will only be delivered in DR4. I attempted a fix for the saturation by replacing the central hole with a scaled version of the theoretical 24um PSF. However, since the scaling is inherently uncertain, it is safer to simply set the 24um flux based on nu*fnu(24)=nu*fnu(25). I've added a note to the flux table concerning this. To give the reader more reassurance, I've added labels to the figure for these five galaxies, to emphasize that the discrepant high-flux data come from problematic sources. c) Some readers will be confused by the highly asymmetric scatter especially in the TIR comparisons. We know part of the explanation for this is the lack of long wavelength coverage in IRAS, as you explain in the last sentence. But the explanation is too terse to be understandable to most readers, and a few more words might be helpful. --More explanation has been added. The common theme in these points is that we don't want someone coming away from Figure 7 thinking that we can only do photometry to +-20-40%. 20. Figure 10 missing-- just a placeholder comment. --For some reason my multi-lambda montages of N5236 and U5829 do not display in my .pdf (but they do in the .ps version, which I've also posted). 21. p13, Sec 5.4, and Figure 11: I really like these comparisons, but do have some questions. Have you compared the colors of our galaxies to the dwarfs in Fabian's SINGS study of M81 dwarfs (Walter et al. 2007, ApJ, 661, 102)? As I recall he found unusual combinations of 24/70 and 70/160 micron colors, which I think you can reconcile with the low surface brightness and relatively high SFRs in the galaxies-- a combination different than you find in spirals. I realize this is not the place to discuss such things in detail, but you might find the comparison informative. I suspect Fabian will comment on this when the paper comes out for general team review. --I originally had a reference to Walter et al. (2007), but then I commented it out. If you look at the relevant figure in Fabian's paper, you'll see that the dwarfs stand out more in metallicity than in IR colors. Likewise, the dwarfs don't seem to systematically differ in our Figure 11, except in 8um emission, as mentioned. But since they do seem a bit higher in 70/24 (Walter et al.) and 70/160 (Walter et al.; this work), I'll put back in my comment and reference to Walter et al. 21. p13, just below equation (4): Is there a reason to write the constant as 0.232 X 3.6/8.0? I guess I see what you are getting at, but it did trip me up for a moment. --I admit that I'm being overly particular, but I like to separate the scale factor from the wavelength ratio. It probably made more sense to do this in Dale et al. (2005), when I was comparing both fnu and nu*fnu (one has just the scale factor 0.232, and the other has 0.232*3.6/8.0). I'd prefer to keep the expression the way I have it. 22. p13, bottom par: See comment (1) --Addressed above at comment #1 23. p13 general: Not that you need more work to do, but did you ever try plotting the various IR flux ratios as functions of total optical (or IR) luminosity? Of course this was done in the early days of IRAS, but I don't think anyone has ever added really faint galaxies to see if the same trends continue. Might be worth a look, even if the results don't end up in the paper. --Good question, but a tough one to address robustly. Take a peek at Figure 1 of Chapman et al. (2003 ApJ 588 186), which shows 60/100 vs L_TIR and L_FIR. Note how the trends have significant slope only above 10^10.5 Lsun, and note how this trend is less steep for TIR. It doesn't make sense to plot 60/100 vs IR for LVL, since there are so many IRAS non-detects at the faint end. A plot of 70/160 vs L_TIR for LVL can be found at http://physics.uwyo.edu/~ddale/research/FIRcolor_TIR.ps Given that the known IRAS relation is flat at 60/100~0.42 (which converts to 70/160~0.52), does LVL follow the IRAS relation at the faint end? i. Does LVL show a flat 70/160~0.5 below 10^8 Lsun? Yes, but that's not too interesting. --One might also ask about how LVL's slope at the bright end copmares, but clearly there aren't enough data to address that question. 24. p14, last sentence of Sec 5.4: Are you really sure that the deviation of the Dale et al. models reflects a paucity of galaxies with high SFRs in your original sample?? If anything I would have thought that your ISO sample had more galaxies with high SFRs and high IR surface brightness. The other aspect of the bottom two panels that puzzles is the fact that the models under-predict both 70 and 160 microns-- how can that be? The only explanation that seems to make sense is that the models grossly over-predicted the amount of PAH and 24 micron emission-- as can clearly be seen in the upper 4 panels of Figure 11. That has nothing to do with high SFR galaxies in the Dale sample-- if anything the SFRs in LVL are lower on average (lower 24/TIR ratio). So my best guess for the model deviations are: (1) lower PAH contribution in LVL, due to lower metallicity, etc.; (2) lower radiation field strength in LVL relative to ISO KP galaxies, which lowers dust temperatures and 24/TIR ratios. Together these put more of the TIR in the cold dust emission, as seen in the bottom two panels. I would not bet my house on that explanation but it makes more sense to me than what is currently written. --Great questions. I agree that LVL has lower average SFRs, and I've changed the text along the lines of your description above. But in my mind the ratio 70/160 maps to SFR (per area), and so why should LVL lie above the model for a given 70/160 (SFR)? I must be partly correct: my models simply were not well-constrained at the highest SFRs. Figure 10 in Dale et al. (2001) shows only three data points above 60/100=1.0 (or 70/160~1.83); the ISO Key Project on Normal Galaxies didn't have a large number of very IR-warm galaxies. This is an argument we can continue outside of the paper! 25. p14, Sec 5.5: Somewhere you will want to indicate where the UV data came from. --There's a footnote in Section 5.3, and I reference Gil de Paz in the Figure 10 caption. 26. p14, bottom par, line 8: "birthrate parameter" will be unfamiliar to many readers, maybe say instead "the star formation rate per unit mass" or something similar? --Janice suggested the current phrasing, so I've added yours to be more complete. 27. p15: I hope Ben Johnson will comment on this. Several other groups have attempted to confirm Kong's results and found no trends with galaxy type, SFR, etc. It would be interesting to understand why we have seen something. And we probably need to cite some of those earlier studies if they are published-- Ben can comment. 28. Figure 14 does not seem to have any text with it? One caution-- I suspect that Buat calibrated her extinction formula using a sample of galaxies that has a different IRX-beta relation than ours. If you want to go ahead with this what we really should be doing is using our, offset IRX-beta relation to calibrate an equivalent to the Buat curve. I suspect that her curve over-estimates the dust attenuation for our sample. 29. Figure 12: Related point, if we do end up using Buat's relation later in the paper, you probably should add her IRX-beta relation to the curves plotted here. --Buat et al. do not provide a new IRX-beta relation (they overplot Kong et al.'s). What Buat et al. provide is a relation between A_FUV and TIR/FUV using PEGASE models and various dust attenuation configurations (screen, mixed, etc). I have rearranged things and added more description on Buat et al. in an attempt to clarify. 30. Figure 6: I have a B/W printout and the 4 curves in the bottom panels can't be readily distinguished. No problem if the figure is in color, but IoP charges $300 per color figure in the print journal, so I am guessing you will want to produce a B/W version for that anyway. --Daniela asked me to switch from color to B/W, for easier viewing! I've offset a couple of the histograms in magnitude for clarity. Let me know if that doesn't work well.