Abstract:

Might the PCA conclusions be taken to suggest that the transition time
from SF in obscured environments to a population of unobscured young
stars is very small, such that observing a galaxy composed of many
such sites (of different ages) yields a bimodal population of clumps in
terms of IR-to-UV ratio?  This seems to me what is going on, even
if (as I think it is) the small scale connection is overinterpreting
the PCA work.

--I don't know if I can say "yes" or "no" to your question, so I guess
  I am uncomfortable raising the point in the text.

I would recommend putting a bit more detail about the 24um morphology
results into the end of the abstract.  Right now, the current statement
leaves one wondering and it seems there is ample room to add another
few sentences.

--Done

Subject headings:

add "ultraviolet: galaxies" and "galaxies: photometry"?

--Done

Introduction:

Given that this is a "SINGS-GALEX" paper, I was surprised to find no
mention of GALEX in the Intro or even the Abstract.  It is really the
combination of the two observatories which makes the study possible.
So, even though the sample and ancillary data comes from SINGS, I'd
suggest adding a few words to acknowledge this fact.

--I've added a reference to the intro.  Let me know if you can think of
  another logical/easy place I could slip in a reference to GALEX.

Section 2:

In Table 1, it seems like \alpha_0 and \delta_0 would be more
appropriate than x(0) and y(0) -- though this is a matter of personal
preference.

--Done

Section 3:

Via collaboration with Pauline Barmby (on M31 IRAC observations), I've
become aware of work by Tom Jarrett suggesting slightly different
extended source calibration factors-- [0.91,0.94,0.66,0.74], in order
of increasing wavelength.  The only one that is significantly
different is the 5.8um value of 0.66 vs. 0.772.  Even this difference
is slight, but I was left wondering which calibration was thought to
be better.  Given that the data in your paper will be a reference for
generations, should this be addressed?  If not, it might be good to
acknowledge the remaining systematic uncertainty in this band.

--Tom has been keeping the SINGS team in the loop on his extended
  aperture correction factor work.  The values you give are appropriate
  in the radius->infinity limit (but they are close to what I'd be
  using, since our aps are very large).  I asked Tom if he'd like me
  to go ahead and publish his approach (not sure if he wanted to publish
  it first himself).  He told me to go ahead and use his new approach,
  and to publish it.  [ Note: he also suggests using 0.72 for 5.8um,
  not 0.66, which is more for a truly flat (zodiacal) case. ]

Table 2 states that "the data" have been corrected for Galactic
extinction when presenting the IRAC and MIPS fluxes, but by-and-large
isn't the correction neglible at wavelengths past the 2MASS bands?  In
contrast, the UV correction can be quite large and introduce some
(dust model-, reddenning curve-dependent) uncertainty into the IR/UV
ratio.

--Yes, there's basically no correction for Galactic extinction at long
  wavelengths.  Since I already mention this correction in the text, 
  I've gone ahead and removed this mildly-confusing remark from the
  table notes.
  
Section 3.1:

End of 1st para. -- add "At the median distance of the SINGS sample
(10 Mpc), the joint UV--IR data probe spatial scales of about \sim 300
pc."  before "This resolution..."

--Done 

Section 3.2:

If the optical-IR fluxes are only integrated within R25, why use the
asymptotic GALEX measurements?

--The apertures for optical photometry go out to *at least* R25, and
  usually beyond.  They were chosen to fully encompass all the emission
  from the optical and infrared data, and thus are a better match to
  the asymptotic values from GALEX photometry.  I have added a few words
  to clarify this.

I think it would be very useful to include the IR color criteria
employed for identifying stars, if it was done in a manner other than
visual inspection of the color images (eg. by thresholding).

--Done
 
Section 4.1:

Figs 1-8 would benefit from additional annotation of the IR/UV ratio
and the 24um morphology designation for each galaxy.  Along the same
line, perhaps the morphology (or nuclear-to-total value) should be
added to Table 1 in addition to the res/unres value?  Alternatively, a
Table 5 could be created to host all this value-added content, plus
the IR/UV ratio, the L(1500)/L(2300) value, and their uncertainty.

--I have added the nuc/total value to Table 1, and TIR/UV to Figs 1-8.


Also, the Vazquez & Leitherer (2005) curve is not corrected for even
_Galactic_ extinction?  Seems to me like it should be.

--The curve is not corrected for *internal* extinction.  I've changed
  the text accordingly.  Since I've already corrected the data for
  Galactic extinction, I don't need to modify the curve for that.

Section 4.3:

I'm certainly not well-versed in PCA work, but would it make any sense
to fit the eigenvector spectra in terms of a toy SFH and dust model?
That is, is the e1 SED consistent with a 100 Myr burst population
with minimal internal extinction?  Is the e2 SED matched by something
along the lines of a highly obscured zero-age burst?

--It's a good idea, and I have thought a bit about it.  But it seems
  that the PCA gives us what drives the variations in the SEDs, and not
  necessarily a basis set for representing a given spectrum in terms of
  a linear combination of the eigenspectra.  Though others may have used
  PCA in this sense, the second eigenspectrum spit out by my analysis 
  doesn't seem to match any observed SED.

Your 24um morphology analysis shows that the IR/UV ratio is tied to
the structure of each galaxy on local (sub-kpc) scales?  Might this be
alluded to already in this section, emphasizing the advantages we have
over high-z workers?

--I'm not sure I understand your point - do you have a specific
  recommendation (and location) for some additional text?

Section 5.1:

The Tufts et al. '04 reference should also be given here, not just in
the figure caption.

--Done

Fig 11 -- One can vaguely imagine that there is agreement with the
Tufts model, merely masked by huge variations in IR/UV driven by other
dominant params?  That is, the Tufts model isn't really disproved, but
has been shown to be irrelevant given the other factors.  Also, the
error bars all are the same or very similar?  Is this true in fact?
(Same for Fig 12 and following.)

--When I add systematics to UV (0.15 mag) and TIR (30%), they basically 
  wash out the statistical uncertainties.  I agree that the Tufts model
  is not ruled out; I've changed the wording a bit.

Section 5.2:

I suggest rephrasing the section title and nomenclature within to
reference "Hubble type" rather than optical morphology given that in
Section 5.3 you make substantial use of a very different variety of
"morphology" classification at 24um.

--Done

In Fig 12, have you thought about fitting a relation between IR/UV and
T-type (at least for the spirals) and reporting that in the text/fig?
Similar relations are given (eg. for FUV-K vs T) in Gil de Paz et al. '06,
the GALEX Atlas.

--I have added a fit to the figure, and briefly mentioned it in the 
  caption.

Section 5.3:

70/160 vs. incl. is _not_ shown in Fig 11, contrary to the 3rd sentence.
Perhaps do include such information in the plot, by color-coding the
points of Fig 11 according to f_nu(70)/f_nu(160)?

--Thanks for catching this typo!

Middle of 1st para. -- The clumpy galaxies having many low optical
depth sightlines for UV photons to escape might also be characterized
as systems in which the UV does not come from a temporally singular
event (nuclear burst), but is rather more continuous or
multi-generational in nature?

--Note added

Can the "single point-like blob nuclear emission" galaxies be
considered as "starburst" systems?

--The H-alpha-based SFRs (Msun/yr) for the four in the top right corner 
  are 1.5, 2.0, <0.1, and "?" in Kennicutt et al. (2003).  So maybe not
  systematically starbursts, but I should check into this further.

2nd para-- Where you state that the 24um emission is spatially closely
associated with HII regions, this is true only in part.  There is a
substantive component of diffuse 24um emission, as your own
decomposition into resolved and unresolved shows.  I would recommend
slight modification of the sentence beginning "In fact..." to reflect
this issue.

--Done

End of 2nd para. -- I think it would be useful to show a comparison of
the observed, point-source, and difference (resolved) images for each
type of galaxy (clumpy, point-like, and smooth).  This could be done
in a nine (3x3) panel figure easily generated by ds9, and would really
aid the reader in understanding the point-source extraction method.

--I put in a 3x3 figure with some nice-looking examples; let me know what
  you think.

Further, why use aperture photometry for the nuclear source, when it
must also be represented in the point-source image?  In the present
method, do you use an aperture correction and remove underlying
diffuse bkgd?  Is the issue that the nuclear sources are "blob-like"
and not perfectly matched by the PSF?

--Karl Gordon carried out this data processing.  What you say in the last
  question is basically correct.  We could repeat the analysis by
  removing the underlying diffuse emission, but it can be somewhat 
  problematic to define what exactly is the diffuse emission level in
  many of the galaxies.

This may not be a concern for the SINGS sample, given the relatively
nearby galaxies being studied, but it seems best to equalize the
spatial resolution of the images (to the worst value ~ 870pc) before
accomplishing the res/unres decomposition.  Nevertheless, there is
still a factor ~10x range in distance amongst the sample.  I'd at
least have a quick look for a possible effect by plotting
f(res)/f(unres) vs. dist.

--There's definitely no trend in the ratio with distance (I checked).
  I've also incorporated distance (with symbol size) into Figure 16.

Fig 14 and the text at the top of pg.10 are very convincing, making me
wonder if the res/unres ratio should be recast into \eta (such as used
by Bell (2003) and Iglesias-Paramo et al. (2006) to represent the
fraction of dust luminosity associated with evolved stars).  I'd be
especially curious to see what the typical implied fraction of diffuse
emission is, and if it varies with galaxy type?  (even if not recast
into eta)

--Don't the y-axis data in Fig 15 provide the fraction of diffuse 
  emission?  I've added Hubble types to this figure, to help answer your
  other question.

Section 6:

You mention the specific star formation rate in the summary, but
haven't computed or used it earlier.  Could it be added fairly easily
given the present compilation of K-band and UV+IR data?

--Good point.  I have added a new section on the specific star formation 
  rate, and linked the text in Section 6 to this analysis.