GENERAL *** Your description of the principal component analysis (or at least the language that you use to describe it) is confusing, and I don't think many readers will understand it. Moreover, I don't think it's accurate; epsilon 1 and epsilon 2 do not describe SEDs by themselves. You should discuss eigenvectors as though they are components that you add together to build the SED. In this light, I would expect that the zeroth eigenvector (which does not appear to be shown here) describes the typical SED of the galaxies in the sample. (You must have an eigenvector like this; you cannot add epsilon 1 and epsilon 2 together in any way to produce flux in all the wavebands between 3.6 and 160 microns.) If your analysis produces such an eigenvector, you should show it. The first eigenvector should be described as the principle way in which a galaxy's SED will change, and you can probably attribute a physical explanation to the variation. In this case, it looks like the first eigenvector describes how an increase in the (unobscured) starlight creates variations in the SED. The second eigenvector should be described as how another physical process will chnage the sum of the zeroth and first eigenvector. In this case, the second eigenvector probably describes increases in dust mass that lead to dust extinction in the UV, the reddening of starlight, and the increase of FIR emission. --Thanks for helping me redo this section. *** The morphological parameters used in this paper seem like they would suffer from severe bias related to distance. Given that many of your "point-like" sources are very distant and many of the "clumpy" sources are nearby, I would bet that the parameters you are using are biased (especially the resolved/unresolved flux ratio). You need to seriously demonstrate that the parameters are unbiased if you want to use them in this analysis. --I have added distance to Fig 16, which shows that distance is not an effect, at least for the SINGS sample. *** The 24 micron band still samples starlight. This should not be a problem for most of your sources, yet it could be a problem for some of the early-type galaxies (including but not limited to NGC 4594, NGC 584, and NGC 1404). The results based on 24 micron morphology for these galaxies would be suspect. --If the stellar contribution to 24um is 0.032*f(3.6), as Helou and others have claimed, then the stellar contribution to the three galaxies you mention above is 18%, 24%, and 26%. In n4125, n4552, n3265, n0855, the numbers are 26%, 28%, 0.3%, and 1.6%. So you're right - stars can play a role some of the early-type galaxies. I have added a caveat to the text. *** If you have not done so, this would be a good opportunity to measure and publish flux densities in the most recently processed MIPS data (where the noise levels are lower than previous versions of the data). The IRAC data have also improved since the publication of Dale et al. I strongly advise you to measure new flux densities rather than recycle the old flux densities. (If you have remeasured the flux densities, then you need to clearly need to state this in the text. Sec 3.4 would be the most appropriate place.) --I have indeed redone the MIPS flux extractions using v4-0. Thanks for pushing me to do this. *** Some of the galaxies are problematic and should be treated as exceptional: 1. DDO 53 and NGC 6822 (and possibly a few other dwarf galaxies) produce infrared emission outside their ontical disks. You should increase your apertures to include this emission if you do not do so already. 2. NGC 4594 has an abnormal SED in the submillimeter range (and it looks like it's roughly flat between 850 and 1300 microns). Bendo et al. (2006) might be worth mentioning somewhere (possibly a footnote in Table 3 and a note in Fig 6?). 3. The far-infrared emission from NGC 1404 is flagged as potentially coming from a background source (which I like), but it is still used in the analysis in Sec 5. Given the dubious nature of the emission, it should be excluded. 4. Some of the elliptical and irregular galaxies are weak detections or non-detections in the FIR, yet they are still included in the analysis of Sec 5. For example, I noticed NGC 584 in Fig 13. You should not include these galaxies. (Maybe NGC 584 was an exception?) --I do use very large apertures for ddo053 and n6822, about twice the area of the optical apertures you provided me. They are listed in Table 1. n584 and n1404 are particularly weak at both 70 & 160, so I have removed them from the relevant plots. You should probably read Lehnert & Heckman 1996 ApJ 472:546, which discusses a correlation between the color temperature, spatial distribution, and luminosity of infrared light in nearby galaxies. This seems like it may be relevant to your discussion. --Thanks. They discuss quite a bit starburst sizescales, rotation curves, starburst timescales, but it is somewhat related. Man, did they produce some crappy-looking figures! ABSTRACT: *** You should clarify what you consider to be "early-type galaxies". --Sentence re-worded. *** A few of the dwarf galaxies (for example, Mrk 33) are also single bright knots of infrared emission. Do these have the same SEDs as the ellipticals? --No, they tend to be comparatively UV-bright with warmer FIR. Sent 4: This sentence begins with too many cluases. Move "Similar to previous findings on normal star-forming galaxies" to the end of the sentence (or place it in a new sentence). --Sentence has been re-worded. SEC 1 INTRODUCTION Par 1: This paragraph contains very little useful content and could easily be deleted. --But it's so fun! SEC 2 SAMPLE *** The information in this section seems generally jumbled together. It seems like sentences were cut and pasted to make this paragraph. If you moved the first sentence to the end of the section (and edited it appropriately), the paragraph would flow in a more logical order. --Done You should probably mention that the sample includes both field and Virgo cluster galaxies. As the section is currently written, it sounds like the SINGS sample only contains galaxies within the Local and M81 groups. --Done SEC 3 DATA Sent 3: The effects of the submillimeter equivalent of "airmass" are also removed from the submillimeter data. --Sentence changed SEC 3.4 INFRARED DATA *** See the general comments. Par 3 Sent 2: I would write "convolving the R band image of the galaxy with the MIPS PSFs" rather than what you have written, but that is my personal taste. --Done SEC 5.2 HUBBLE TYPE Par 1 Last Sent: This sentence mentions "strong ultraviolet emitters". I am unfamiliar with such objects. Could you explain what they are? --Yi et al. (2005) describe them as F(NUV)/F(r)>0.08 and/or F(FUV)/F(r)>0.16. *** Par 2: Some of the Im and I0 galaxies (for example, Mrk 33) are compact sources, wheras other objects (e.g. NGC 6822) have large spatial extents. If you compare the objects that scatter from the trend to the 24 micron morphology can you explain the scatter in Fig 12? Please add comments on this if this is the case. --As you know, the Im and I0 classes are distinct; Im galaxies tend to be very amorphous/clumpy, whereas I0 are not. I looked at each of these, and Mrk33 seems to be a special case of a 'very' distant Im. All the other Im galaxies are within a few Mpc, whereas Mrk33 is over 20 Mpc away. So perhaps its comparatively smooth appearance is a distance effect in this one special case. SEC 5.3 FAR-INFRARED COLOR *** Par 3: Is the nuclear flux measured in the original 24 micron image or in the "point sources" image? In either case, the method may have problems. First, sources near the nucleus, particularly star formation rings (as in NGC 1097, NGC 1512, and NGC 7331) will contaminate the 12" region. Second, the physical size of the central region varies as a function of distance. Other characteristics of the SINGS sample vary with distance, too (such as Hubble type). Therefore, the "nuclear flux" given here may only be a crude measurement of the true nuclear flux. You should at least clarify in which images the nuclear fluxes are measured and state the problems with these measurements. You may also want to potentially modify the measurement of the nuclear fluxes. (Also see the next comments and the comments on Table 1.) --The nuclear flux is from the original image. That is noted in the last sentence of the paragraph. I have added the distance to Fig 16, to help the reader understand that, though distance plays a role (larger dispersion?), it does not drive any of the observed trends. *** Par 3: The diffraction spikes of the model PSF need to be aligned with the diffraction spikes of the data. Otherwise, the residuals may contain flux from the improperly-subtracted diffraction spikes, as is the case for NGC 1266 in Fig 14. This would case some resolved/unresolved flux ratios into doubt. --I will ask Karl about what he did in such cases. *** Par 3: How does the detection of unresolved sources depend on distance? Most of the galaxies described as "clumpy" in Fig 15 are also very nearby compared to the sample as a whole. If you simulate what some of the nearby galaxies look like at 30 Mpc, do they still appear to be "clumpy"? I suspect that NGC 2915, NGC 1705, and DDO 53 would look like point sources. In contrast, NGC 2798 (which is part of an interacting system and which may have a complex morphology) and NGC 1482 (which shows some spatial extent in the 24 micron image) may not appear point-like when seen at 3.5 Mpc. You need to address this problem in the text and potentially alter your analysis and conclusions accordingly. --I have added distance to Fig 16, which shows that distance is not an effect, at least for the SINGS sample. SEC 5.4 SPECIFIC STAR FORMATION RATE *** Par 3: I disagree with the assessment that the early-type galaxies have higher resolved/unresolved ratios than the late-type galaxies (although this partly depends on what is "early-type" and "late-type"). According to Table 1, the following galaxies have resolved/unresolved ratios greater than 2: NGC 584, NGC 1291, NGC 1512, NGC 2841, NGC 3521, NGC 4254, NGC 4321, NGC 4594, NGC 4631, NGC 4826, NGC 5033, NGC 5055, and NGC 7331. 6 of those 13 sources are Sbc or later, and a few are Sb (which is roughly in the middle). You either need to statistically prove this statement or remove it. --Yeah, I had to trash that argument and go with a better one posed by David Hollenbach. TABLES: Table 1: I could not see the crosses in the second part of the table. --Fixed. Thanks. *** Table 1: I find some of the nuclear/total flux ratios suspicious. A few of these galaxies (for example, NGC 1377, NGC 3773, NGC 2798) are unresolved sources, but they are not reported as having nuclear/total flux ratios of 1. --Remember that 'nuclear' here is defined as the flux within only 12". Table 1: Some of the sources are missing resolved/unresolved annd nuclear/total ratios. Some are understandably non detections so a footnote in the last two columns of the data should say as much. Why are other galaxies not listed? --OK, footnote added. FYI, Karl is still working on n6822 and n5915. *** Table 1: Can you validly report resolved/unresolved flux ratios or nuclear/total flux ratios for soome of these galaxies, such as DDO 165? --I've changed the entry to a limit for a few cases. FIGURES: Fig 9: This figure needs error bars (or a comment that the error bars are invisible). It's not clear as to whether the highest and lowest measurements for a given data point are statistically different or not. --It would be nice to put in "1sigma dispersion" bars, but I just think it is already too crowded.