Perplexity: when looking at your measurements of the few dwarfs we have,
I started to scratch my head, since my own recollection of the SPIRE images
for many of these is of virtually `empty' fields.
So I went back and looked at the images of, e.g., HoI, HoII, and M81DwB.
For all of those, you plot detections in the SPIRE bands, but, for instance,
HoI appears undetected by SPIRE in all bands, while HoII and M81DwB appear
undetected at 500 micron. I am perplexed at how small the error bars
are in both Table 2 and Figure 3.
I fully understand how you estimate the error bars (page 8), and that you
get a S/N advantage from integrating over large apertures, but somehow the
values reported in Table 2 clash with what I would have quoted based on
the quality/depth of the images.

I have looked in detail to the 500 micron image of M81DwB: my flux measurement
for that galaxy (using your aperture and your center coordinates) varies by
about one order of magnitude, depending on how I average the sky values.
If using my most reliable sky measure, obtained from the MSKY routine of
Mark Dickinson (which uses as many of the pixels in an image as possible), I
get that the 500 micron flux of M81DwB is about 1 sigma of the sky noise
within the same aperture. Much less than the ~3 sigma value you list in Table 2.
You state in your summary section that the sky subtraction is the most
problematic part for some of these galaxies. However, I would maintain that
this `problem' should be reflected in the photometric uncertainties of Table
2 and Figure 3.

*** Thanks for taking such a close look.  I like to tell people that these dwarfs drive me nuts, and that they take as much time as the rest of the sample combined.  Because they are so faint and thus susceptible to large percentage-wise measurement errors, I've looked at them again and again.  Your comments encouraged me to check them once more.  

Thought #1: Yes, M81DwB, HoI, and HoII are barely detected, if at all, by Herschel.  However, when you look at them in conjunction with HST optical, IRAC 3.6, MIPS 24 and 70, and most importantly, H-alpha imaging, the small features seen by Herschel match up nicely with the detections at other wavelengths.  In fact, for an astounding confirmation that MIPS is more sensitive than PACS, you should try blinking the two types of imaging for HoII.  Thus, though the formal uncertainties may seem underestimated, I claim that our ancillary data gives us more confidence in our photometry compared to what we'd otherwise assume if we'd considered each Herschel image by itself.

Thought #2: Your point is not just about the extracted flux, but also about the quoted uncertainty given the relatively large sky fluctuations compared to the source level.  Fair enough.  I have increased the uncertainties for HoII and M81dwB at 500um so the S/N is now 2.


Minor comments:

Abstract: when you remark the presence of sub-millimeter excess in
dwarf galaxies, may be worth adding: `, as already noted by other authors.'
(since we are not the first ones to see it).

*** Done

Page 3, when you list upcoming papers: would you mind changing
Calzetti et al. to Li et al.? (My graduate student Yiming Li will be
leading the paper).

*** Done

Section 2, sample: OK, this is nit-picking, but I suggest to move
the reference to Kennicutt et al. 2011 to later in this brief
section, at the end of the sentence starting with: `The 61 galaxies\u2026'.
Where it is now seems a reference to SINGS, rather than KINGFISH.

*** Good point.

Section 3.1: I guess it will depend on the order Kennicutt et al. and
this paper come out, but most of what described in section 3.1 is
already in Rob's paper. Of course, since you have already written
the section, the path of least resistance is to leave it as is. But I
wonder how useful is to have a repeat of the same data observation and
reduction steps.
My personal suggestion is to reference Rob's paper, and simply give a
very brief summary of the products you use in your section 3.1.

Ditto for section 3.2.

*** If I get similar comments from the larger team, I'll consider it.

Page 7: Background galaxies removal. Maybe a few more details here would
not hurt. I understand this is relevant only for a few galaxies, but in those
cases, the flux contributed by bck galaxies is, as you say, significant.
How is the removal performed? Do you lay an aperture on top of the bck galaxy
and measure its flux at each wavelength? How big is the aperture? Do you apply
aperture corrections also to the flux of the bck galaxies?
Also, can you briefly quantify how significant is the contribution of
bck galaxies to the dwarfs?

*** I have added more explanations, including the impact on the dwarfs.  Plus, a new figure has been added showing an example case of sky aperture selection.

Page 14: Would it make sense to include two histograms: one showing the
distribution of T_d and the other showing the distribution of beta? Just
out of curiosity, I would be interested to know if there is a peak (or
preferred value) somewhere within each distribution.

*** Done.  Nothing surprising here.

Page 16: for the 500 micron excess of the dwarf galaxies: an alternate explanation
to the dust emissivity change is the presence of `funny' dust components (see
one of the recent papers by Caroline Bot on the Magellanic Clouds).

*** Bot et al., which I reference in the Introduction, review the possible reasons for the excess, including CMB fluctuations, spinning dust, very cold dust, and an unusual emissivity.  I didn't mention the first two since they seem to be more applicable to longer wavelengths, but I do now point to this paper in the Summary as a reference to learn about other possible solutions.

Page 16, last paragraph: perhaps I would remove this `publicity' stint. Not
needed here, and breaks the science flow of the paper. (and you have already
expressed the same concept in the Introduction).

*** ok