Intro: I think it's fair to argue that the range of spectral shapes may be "different" but not necessarily narrower, than we probe. The metal content must at some point be lower, the current to past average star-formation rates should be much higher, etc. There is very likely a large swath of parameter space galaxies occupy there that they simply don't locally. --OK 3. Observations: suggest "tweaked" => "altered". --OK You might make clear a priori that Eq. 2 only applies to a very small fraction of the flux in a small number of sources. It currently reads like a large correction, until later in the paragraph. --I've 'tweaked' the wording a bit. I think the last sentence here should help to explain most such questions. 4.1 Global broadband energy distributions ========================================= para 1: The UV discrepancy model from data is large here. We discussed this on the call, but the figures make it appear that this is a fit which is failing abysmally. The text makes it clear these are only "representative", but that's not the impression you get. The problem is mixing a tight fit to the mid-IR - submm, but then appending a scaled single stellar model. This should be mentioned more prominently and right away here. The UV model is always above the true fluxes. I might suggest another approach: redden this model by a single value of screen extinction, chosen so that the UV points fall equally above and below the model curve. Probably A_V~1-2 would do it. Then at least you can see which are more extincted and/or weaker UV emitters than the mean, and the discrepancy will not be quite as shocking. --You're not the first to bark about the stellar curve. However, such an analysis has been reserved for a Paper II in this series of SINGS+GALEX data, by agreement between the SINGS + GALEX teams. 4.2 SEDs Binned =============== para 1: you might say "distribution of spatially integrated, limited wavelength spectral energy distributions", since it's not clear how "global SED" and "panchromatic survey" here differ without reading RR05. para 2: One very interesting point you don't mention: only high IR/UV sources show a significant near-IR OH- opacity bump. This bump was sold as the perfect photometric redshift indicator, yet it clearly doesn't appear in all sources. I think quantifying the strength of this bump as a function of galaxy IR properties, or at least mentioning this, would be very useful for those who use it in phot-z codes. --Done para 3 (and elsewhere): since you've normalized in K, you need to couch the discussion of all variations among the sample with respect to the K luminosity. E.g. the last sentence might read "by the range in the sample galaxies' star formation properties, relative to their total stellar content", or similar. --Done 4.3 PCA ======= My biggest quandary on the PCA was what effect changing the location of the SED normalization would have on the PCA eigenvectors. Of course, if you pin things in the middle, you'll have some going up, and some going down. What if you pinned at 160um, say? I expect you'd get very different components. They may be physically related (e.g. steep UV, shallow UV slope sources, etc.) Obviously, you had to normalize somewhere, but I think this is a valid concern that should at least be mentioned. --OK, added another caveat 5.1 Inclination =============== Instead of "arbitrarily normalized" why not say "normalized to IR/ UV=1 at zero inclination". --OK 5.2 Hubble Type =============== "excess to an excess" => "to an excess" --Thanks You don't mention "leakiness" here, but I presume inefficient UV -> IR conversion could be an equal explanation for the UV bright ellipticals. I didn't understand how low star formation levels could make a UV bright galaxy. Dust doesn't care what sort of star warms it, but only the energetics of that star. If you buried a small amount of star-formation inside of 1 brick wall of dust, the dust may be colder, but the UV will be determined entirely by dust geometry (any holes, etc.). --I've reworded this section a bit. 5.3 FIR color ============= para 1: "The warmer FIR colors for SINGS dwarf/irregular are shown...": The warmest galaxy is an early type. --Yeah, but my point is that the typical irregular is warm. para 2: The parenthetical statement "or their UV emission does not come from a temporally singular..." is a bit awkward. Do you mean a single burst vs. multiple or continuous SF? Perhaps you could just say that. --Text modified The geometry argument here is one of the key points, but I found it somewhat difficult to follow. I see three separate arguments: 1. Unresolved nuclear emission: high extinction, high IR/UV 2. Unresolved clumpy emission: low extinction, low IR/UV 3. Resolved, extended emission: medium extinction, medium IR/UV, cool FIR colors. Part of the confusion is that you mention the scenario, but then later describe the test for it using 24um morphology. I think the other way around might be clearer: describe finding stars, calculating resolved vs. unresolved in the galaxies. Then codify the 3 geometric arguments above with those result in hand. See also a comment in the Figures section below. --I guess I prefer the present flow of introducing the idea, sampling the data, and then reconciling the idea and the data. But I've tried to streamline things a bit more, and to have a bit better explanation. I'm suffering from reviewer fatigue, so this is a bit of a cop-out: I'm happy, though, to consider any revised text/paragraph structure you could provide for this section. Does the "resolved vs. unresolved" measurement specifically exclude the nucleus? It might be a more powerful measure if so. --I added the nuc-to-total in attempt to account for this. I like your idea, but then it would throw off the metrics for clumpy galaxies that have a significant-but-not-dominant nuclear emission. last para: "leading to lower IR-UV" => "lowering their IR-UV" "exhibit lower FIR colors" => "exhibit cooler FIR colors". --Done 5.4 SSFR ======== The units of Eq 5. aren't correct. I know you're turning TIR and nu L_nu at 1500 into M_sun/yr, and (somehow) nu L_nu (Ks) into M_sun, but it would be clearer to include the constants used directly. E.g., what happens to the factor of 0.8 from Bell's sloan calibration? Is that deemed close enough to 1 to ignore? Since you're not just calling it "relative SSFR", but giving it in an explicit inverse timescale, I think you may want to clarify a bit. --If you look at Kennicutt's formulae, you'll see that he also simply goes from Watts -> solar masses per year. So I think it will be ok if I do the same. Now as for your question on the factor 0.8 from Bell: do you think I really should include it? I guess your point is that since I use two significant digits in both the TIR and UV coefficients, I might as well toss in that additional significant digit for the denominator. OK, I did it. para 2: You don't mention the "fork" in Fig. 17. You mention later Hubble types having larger SSFR and smaller IR/UV, but this isn't true at 10^-10 yr-1, where it forks up to high IR/UV for later types. --Those are the oddball nuclear 24um sources, so I dismiss them with a magic wave of my hand by saying "With the exception of a handful of nuclear 24um sources with high IR/UV, the SINGS sample shows a general trend..." last para: Again I'm confused by the "clumpy=UV escapes easier argument". Implicit in this argument, I believe, is that a constant amount of UV escapes from HII regions, and that, therefore, if a UV photon encounters multiple regions on it's way out of the galaxy, it's more likely to get absorbed. You could argue it the other way: clumpy galaxies consist of tightly packed HII regions or conglomerates of HII regions, from which UV escape into the ambient medium is more difficult. This may not be the case (in fact your results may argue against it), but it might be good to mention this implicit assumption, ideally earlier on where you lay out the whole "clumpy, nuclear, diffuse" geometries more clearly. --Another way to argue it would be: Galaxy A has a clumpy dust distribution, Galaxy B a nuclear dust distribution. Otherwise they are identical (dust mass, etc). It should be easier for UV light to escape A, since less dust surrounds a given HII region. Anyway, I've slightly modified my argument, to emphasize that enhanced SSFR in spirals implies a higher density of holes (blowouts, whatever) through which UV can escape. So the clumps themselves are clumpier (holey) as you jack up the SSFR. I suppose this doesn't help, and my argument is just as 'opaque' as before ... 5.5 UV slope ============ para 1: "from UV spectral data" => "from UV spectral data alone" --Thanks "increased diversity ... increased dispersion" => "increased diversity ... larger dispersion". --OK last para: it would be reassuring to mention the SINGS points which *are* consistent with the SB curve. I hope they're SB-like. --Done Figures ======= Figs. 1-8: As mentioned above, a stellar curve with an average extinction applied might be better. Fig. 9: I'm sure you've heard this 100 times, and rejected it as many, but connecting the dots would really help improve the appearance of these curves. I understand why you don't want to: you don't know anything about the shape of the SED in between the points: e.g. the 4000 angstrom break would look dumb. So here's another thought. What about showing the 1-sigma variation within the 10 or so galaxies in each curve as "bands" surrounding the points? Even if you opt to keep them disconnected, 1-sigma bars might be very useful. --Your wildest dreams have come true - I've connected the dots for you. Fig. 15: I tend to think resolved/unresolved is more important to nuclear/total, from the conclusions you draw. I know you need both to explain the range or IR/UV at warm colors. One possibility it to reverse them: let resolved/unresolved proportional to symbol size. Then you could bin into bins of "nuclearity", and use a different symbol (e.g. increasing sides: triangle, square, pentagon, etc.) or, shade the symbols white->black to indicate how "nuclear concentrated" they are (you'd have to give a minimum symbol size so the shading could be seen). The little text labels are a bit distracting, and hard to get the "flow" of the thing. If you do something like this, you'll also be able to appreciate that IR/UV has an upper envelope which changes for different resolved/unresolved ratios, obviating to a degree the one- off plot you sent me. --I've removed the little text labels - I agree, they can be distracting, and the info is in the next figure anyway. And I've switched the symbol format according to your suggestions.