1. Introduction �1, line 1: "Dust has always presented a challenge to 
astronomers". I would say that this might be better than "... to astronomy"

--Agreed.  Fixed.

1. Introduction �2, end: as I wrote in a previous email, we tried to 
quantitatively estimate the origin of the dispersion in Burgarella et 
al. (2005, MNRAS 360, 1413) on the sample of normal star-forming 
galaxies from Buat et al. (2005). We found that the star formation 
history (already identified by Kong et al.) and above all the shape of 
the dust attenuation laws (mainly the slope but also the strength of the 
bump) seem to be the main parameters influencing it. Of course, if we 
assume that the dust attenuation law is related to stellar populations 
through, for instance, the evolution of their environment (e.g. clumpy 
vs. diffuse emission), we can reach the conclusion that the age is an 
important parameter.

--I have added references to your 2005 paper.

3.2 Optical Data �1, end: "... at a signal to noise of ~ 10". It might 
be useful to write whether is it SNR /pixel, /resolution element, or 
integrated.

--Done

4.2 SED binned by the IR-to-UV ratio, �3, end "The 24 um emission from 
galaxies is known to be sensitive to the SFR ...". OK but what about 70 
and 160 um ?

--Good question.  Some of those same Spitzer papers referenced here have
  also shown that the SFR is also sensitive to the 70um data, when the
  galaxy is close enough to enable detailed, resolved studies of 70um
  and canonical SFR indicators.  But resolving the IR emission isn't
  really an issue here, since the plot deals with integrated values.
  I'm not sure why the integrated 24um emission shows more dispersion 
  than the 70um emission.  Perhaps it's because the 70um emission has
  more contributions ('contamination') from heating by older stellar 
  populations.

5.4 UV spectral slope �1, idem that in the introduction.
Fig. 16: blue and red star are used to represent Kong and Calzetti 
points. However, we cannot make the difference in black and white. Maybe 
could you use different symbols that can be clearly identified in black 
and white.

--I too usually try to make every plot accessible in black & white, in
  case someone is like me and only uses black & white printers.  The 
  implicit assumption with my adopted scheme here is that the B&W reader 
  will only care about distinguishing between SINGS and archival galaxies.
  In other words, even in black and white we can distinguish between 
  archival and SINGS 'star' galaxies since the SINGS 'star' galaxies have 
  error bars.  For this plot I thought it would be ok to go with a color 
  plot for the online and print editions of the journal, as an option in
  case readers wanted to distinguish between Kong et al. and Calzetti et al.
  starbursts.

In Sect. 5.4 on the Ultraviolet Spectral Slope, you quote Meurer et 
al.'s (99) result (IRX vs. beta) and checked it on your data. However, 
with Veronique and Samuel, we discussed about the few galaxies that were 
in the initial sample of Meurer et al. (the one from IUE) and I have 
plotted using GALEX data and Spitzer data (L_TIR calibrated via Dale and 
Helou 2002). I found 5 galaxies (NGC1705, UGC5720, NGC7793, NGC2798 and 
NGC7552) which are plotted in the two attached diagrams as big orange 
dots. Very interestingly, although the sample is very small, they fall 
quite close to Meurer et al.'s (99) law shifted to plot L_TIR/L_FUV vs. 
beta_GLX in the first plot and L_1500/L_2300 in the second plot (blue 
solid line) with a small dispersion. Above all, they follow the same 
general trend (actually, they would almost perfectly be fitted by the 
L_FIR/L_FUV law by Meurer et al. 99. Warning: I do mean  L_FIR and not 
L_TIR). A small re-calibration would do a very good job on these 5 
objects. I also overplotted four of them with the data directly taken 
from Meurer et al. (green large crosses). Another interesting point is 
that these five galaxies fall in the same area than the objects from 
Calzetti et al. (95) in your Fig. 18: just a little bit below the solid 
line. The dispersion is smaller than for Kong et al. objects.

It seems that these five points already give a nice hint on the behavior 
of the original starburst sample that should be confirmed.

--I have added some text that explains that, though the SINGS sample shows
  very large dispersion and does not as a whole match the original starburst
  trend, the starburst subset of SINGS does indeed match the original 
  starburst trend.