- what does the phrase "...compared to starbursting galaxies" in the  
abstract mean ?  That whole sentence is confusing to me.  The latter
term (starbursting) includes things like LIRGs and ULIRGs, and the 
SINGS data certainly do not show a larger dispersion in IR/UV than 
those objects.

--The UV folks love to make plots of IR/UV vs FUV/NUV; as you know,
  there have been lots of papers showing how starbursts follow a
  canonical, tight trend in such a plot.  In the abstract I refer
  to the last figure in this paper, which shows SINGS data plus
  archival starburst data plus the canonical trend for archival
  starburst data.  SINGS galaxies deviate from this trend.  I've
  reworded this sentence a bit for clarification.

- on page 3, 1st sentence, you bring up the morphology, color comparisons
for the IR/UV ratio variations, which is fine, but you need to explicitly
state what wavelengths you are talking about.  I think the main "gain" 
here is that for the 1st time we can compare the ratios to the MIR 
morphologies.  Certainly there have been other papers comparing the IR/UV
to optical morphologies.

--Done

- later in the same paragraph you mention that the SINGS data allows us 
to do this over entire galaxies, but certainly this has been done before
at other wavelengths (using IRAS, 2MASS, HST, etc.).  It seems that you
are selling this as the 1st time we can do this sort of panchromatic work
on galaxies, and I would argue that is not true.  What specifically is 
NEW from the Spitzer data (and SINGS) that provides a unique way to look
at this puzzle ?  I would guess it has to be the MIR spectra (which you 
don't use here) and the MIR imaging (which you do).

--Well, it's also true that the SINGS dataset is unique in that it has
  complete UV-IR images for such a diverse sample.  Many prior UV efforts
  focussed on starbursting systems, since those were more easily detected
  with pre-GALEX instruments.  One could argue that Gil de Paz has now
  scooped us with the GALEX Nearby Galaxy Atlas, which can detect normal
  and wimpy galaxies in the UV.  But that database still uses IRAS data,
  which doesn't provide IR fluxes for wimpy objects like SINGS swarfs. 
  So this paper still has its niche.  I have added a note that emphasizes
  the luminosity diversity in the SINGS sample.

- toss out the 1st sentence in section 4.2, since it reads like an  
excuse or a discussion of things we cannot do.  Just get to the point, as
GXN would say.  Start with "Fig.9 shows a stack..."

--I removed it.

- isn't always true that the spread in a distribution is often dominated
by a few objects?  Isn't that the definition of a normalized 
distribution?  Those first few sentences in the 3rd paragraph of section 
4.2 got my scratching my head as to what your point was.

- this may reflect my own ignorance, but while the eigenvector analysis
is neat, what exactly did we learn from it?  What is the bottom line of 
section 4.3?

--This section has been revamped in an attempt to be more illuminating.

- the 1st sentence of section 5 says that we will learn something about
the "geometry of the interstellar grains", but you probably mean the
relative distribution of the grains in the ISM relative to the heating
sources and not the grain geometry itself.

--Fixed

- I look at fig.13 as a plot of average global opacity (y-axis) vs.  
average global dust temperature.  The galaxies scatter all over the place
but, there are no galaxies with cool dust and high opacity (upper left). 
Why?  Is this a SINGS slection effect?   Probably not. So where are they?

--We talked about this on a telecon.  It's hard to overlay on this plot
  archival data that would fill in some of the missing parameter space,
  if this is partly a SINGS selection effect.  All the prior UV studies
  with IR data seemed to focus on starbursts, which means they reside in
  the right half of the plot.  I asked Eiichi about adding some of his
  SCUBA sources, to see if we can find cold and dusty sources.  He says:

  "Very few SCUBA galaxies have been detected at 70+160um with high S/N.
  I have one MAMBO (i.e., 1.1mm-selected) galaxy at z~1.4 strongly 
  detected at 70 and 160um, so I'll try to send you the K-corrected 
  70/160um flux densities within the next few days."
  
  However, I have added the GALEX Atlas of Nearby Galaxies to this figure.
  It shows a very similar distribution.

- when you are discussing the nuclear/total 24-micron ratios things get 
confusing because you first talk about the nuclear/total ratio, then list
in Fig.15 the ratio of resolved/unresolved flux (which is the inverse). 
Also, the largest symbols are the ones that are the most compact 
galaxies.  Please always stick to one ratio (I suggest compactness as 
defined by nuclear/total) and then make everything refer to that - 
including having the smaller symbols be the most compact objects.

--Based on your comment below, I think you misinterpreted my analysis of
  resolved-to-unresolved 24um emission.
  
  I understand your comment about plotting the symbol according to the
  compactness.  Since the unresolved-to-resolved ratio has more diagnostic
  power here than the nuc-to-total, JD has convinced me to switch the
  approach to symbols here.  Check it out.

- I guess I don't know why the nuclear compactness should define the  
corrrelation (hotter and more opaque) for the SINGS galaxies.  For ULIRGs
and LIRGs I completely understand this, since the dust is shoved into the
circum-nuclear regions and a SB or AGN heats it all up.  THey are warm 
and opaque.  For the SINGS galaxies however, I would think we would
want to use something besides nuclear compactness since we are trying to 
talk about the distribution of the dust and the stars throughout the
disk, and nuclear compactness only uses two global numbers.  Have you 
looked at a more general definition of compactness, say the amount of 
flux throughout the disk which is unresolved vs. the amount which is
resolved?  That might be really telling.  In many systems, this still 
might be dominated by the nucleus, but in others, not.  This would then
be a global measure of the relative amount of light at 24microns which is
compact (everywhere in the disk) to that which is diffuse.  It would be
like running a sort of deconvolution on the MIPS data, but not to make
the resolution better, but simply to quantify the amount of unresolved
vs. resolved emission.  I bet Eric has done this.

--Good suggestion.  Such an analysis is actually already present.

- I think you need to be more explicit about the timescales involved in
your specific star formation rate ratio.  Just saying current to 
historically averaged is not enough.  What timescales are reflected by 
the numerator?  This is what we were talking about last week on the 
phone.  Also, couldn't the SSFR be large in a system with a lot of cold 
dust (heated by the general ISRF) even if there were no UV flux ?  Or one
where there was a lot of diffuse dust (cold), blocking the faint UV 
light?  Then you might infer a large SFR compared to the past average, 
when in fact this was not true.

--Alessandro Boselli and Veronique Buat both raised the issue of using a
  SFR indicator that accounts for heating by the older stellar 
  population, among other caveats.  I made a comparison between their 
  more sophisticated equation and mine, and the results were similar (and
  the plot looked qualitatively almost the same).
  
  I have added a blurb about applicable timescales.

- in section 5.5 you use both star-forming and starburst without defining
them or clearly saying what the difference is.  In my mind a starburst is
a galaxy experiencing prodigious, recent star-formation (perhaps 
triggered by and encounter) that is occurring at a rate that cannot be 
sustained over the lifetime of the galaxy.  In other words, the rate is 
so high that the fuel will run out in much less than a Hubble time.  In
the most extreme cases, the ULIRGs, the rates suggest the SF will run 
through the available fuel in less than 10^9 yrs.

--I have added some text for clarification (plagiarizing the above).

- need to make the Discussion more specific and less broad-brush.  There
are a couple of findings that are not even mentioned in there (e.g. the
break in fig.17).  Your last statement seems obvious, and one that we 
could have made before Spitzer.  Namely that the star formation histories
controls the colors of galaxies.  No kidding.  Let's try and focus the
end again on something that is new from the Spitzer data - say the MIR  
compactness or the lack of opaque systems with cold dust.

--I've added some words about the lack of opaque systems with cold dust.