a general comment on the figures: i know that space is tight but many
figures would display much more information if you showed the galaxy's
name right next to the respective symbol. that way it would be much
easier to follow which galaxy in the table ends up where in the plots.

--I think the plots are already way too busy.  I've even excluded
  the error bars from the archival data, to ease the interpretation.  I
  fear putting in the names would make Figure 3 really difficult to
  view.  However, in light of your comment I have added names to 
  Figure 2, where there is a bit of room to squeeze in names.

another general comment: you don't show the actual spectra for most of
the galaxies. i know that adding all spectra would add many pages -
but as this the (data) heart of this paper i wonder if you will get
away with the referee without showing this (and only presenting the
numbers in the table)?

--The SINGS IRS subteam has discussed this issue during our weekly
  telecons.  We prefer to wait until all the data are obtained and
  processed, and then publish a SINGS IRS atlas paper.

page 1, [,] missing before H$\alpha$

--Fixed

page 5, 3.3, the the -> remove one 'the'

--Fixed

page 6, 4.2 - perhaps it would be worthwhile starting this section
explaining why you don't us the 7.7 micron feature (which i guess
would be the feature people would traditionally use). perhaps start
with something like 'traditionally, the 7.7 micron feature (as being
the brightest of all pah features) has been used to quantify the pah
emission strength' - then you could stress that, in the case of the
irs observations, the 6.2 feature is easier to deal with for the
reasons you give. related comment: from the spectra you show it
actually looks like the 7.7 micron feature is quite prominent and
well--defined - if you already have derived numbers for this feature
you may consider adding them to the tables as well.

--The 7.7um feature would be tough to quantify as it spans two orders
  in the Short-Low spectrometer.  The continuum on the red side is
  also problematic to define.  Genzel was the first to pose a 
  diagnostic like these, but I wouldn't then say the 7.7um feature is
  the traditional choice.  The ISO diagnostics that followed Genzel's
  actually used the 6.2um feature (for many of the reasons I outline
  in Section 4.2).

page 6, 4.2 - also tables 1 and 2 and figure 3. here you only refer to
the equivalent widths of the 6.2 micron feature. but would it be also
possible to give the line (peak) flux (or the peak-flux--to--continuum
ratio)? after all, this could give you a different answer. i am only
thinking aloud here but, e.g. the original genzel plot showed the 7.7
micron flux and one may wonder how things would change if they had
made the same plot using the equivalent widths...

--I address this issue by actually transforming their 7.7um-based
  curved to one that uses the 6.2um feature (as you allude to below).
  I think the Genzel work, though visionary, is a bit cartoonish: they
  use mostly upper limits, and incorporate quantities like 'relative
  strength' instead of simply measuring the actual EW.  During our
  weekly telecons, the SINGS IRS team agreed that an important aspect
  of this paper was to incorporate more typical measures like EW (and
  be more quantitative and less conceptual).

page 8, 1st par, second but last sentence "Our goal is not to..." -
better remove this sentence as it may open a can of worms?

--Unfortunately, you raise a point that conflicts with prior 
  suggestions.  People wanted a statement like this, to emphasize that
  this paper wasn't about contrasting optical versus infrared 
  classifications.

page 8, section 5.2.1: to show my ignorance: you say the ionization
potential for O IV is 54.9eV and the one for Ne I is 21.6eV - i.e. you
need a photon with 21.6 eV to get Ne II and one of 54.9eV to get O V? 
But then you compare O IV to Ne II which may be confusing (at least it
confuses me and i am sorry if i have missed the point!)

--Good point.  I also find this ionization/excitation business 
  confusing.  I have reworded these sentences.

page 8, section 5.2.1: i would state earlier on that the genzel
diagram was based on the 7.7 micron feature.

page 9, second line: it may be useful to quote the Si II ionization
potential as well.

--Done

page 10, line 8, when refering to the Genzel mixing model, you could
explicitely refer to their Fig.~5, right. i think it would be
important if you could explain in a bit more detail how you
transformed their 7.7 micron strength to your 6.2 equivalent
width. after all, the 7.7/6.2 pah strengths vary significantly from
galaxy to galaxy (at least so i thought).

--Actually, if you look at JD's figure 7 in his most recent manuscript,
  you'll see that 6.2 tracks 7.7 fairly well.

page 10, section 5.2.2. you state that the neon excitation tracks
sulfur excitation. but when i look at the figure 4 it does not seem to
be a linear relation (i.e. you seem to get less neon excitation for an
increased sulfur excitation). i don't know how important this would be
but in any case it may be useful to give the least square fits of the
two lines in the text (with errors). that way one could see if the
slopes are consistent (within the errors) with slope unity.

--Good idea.  I added the slopes (and uncertainties) to the text.

page 11, section 5.2.3. how is the definition of the four regions
justified?. in particular, where does the factor ('slope') 8.4 in
equation 5 come from? i am only guessing but it almost looks like this
has been done to include as many 'star forming' squares in region iii
as possible. perhaps this can be justified a bit better.

--They are just empirical demarcations, and I chose to go with the same
  slope for all lines to simplify things.  I originally had only one
  line and no corresponding table.  I have added two more lines plus the
  table to help make this a more user-friendly diagnostic.  Since there is
  no theoretical input behind these lines, I'm not sure I can 'justify'
  them.  I state in the abstract that 'we identify' regions dominated
  by AGN or star-forming sources, implying they are empirical.

page 12. section 5.2.4. when talking about the linear fits you state
that they 'differ along the diagonal' - what exactly do you mean? that
they have different slopes? again it may be useful if you could give
the best fits in the text (and that way one could judge how
significant the differences in slopes really are). also, perhaps you
should at least comment on the three points that fall below the
'low-density limit' in figure 6.

--Previous reviewers didn't think the two fits were different, because
  they had approximately the same slopes.  I italicized "along the
  diagonal" to emphasize that, though they have similar slopes, the AGN
  and star-forming populations appear in different parts of the figure.
  I have added the slopes (and uncertainties) to the text.
  
  Bruce pointed out that three sources below the low density limit is
  o.k. for >60 sources overall, as long as the error bars are 1 sigma
  (which they are, as I now state in the caption).

page 13, last sentence in summary. 'this has proved useful' to me
sounds like an understatement. perhaps change to something along the
following lines: '...has proved critical...'

--Done

page 21, fig. 1: it would be good if you could label the spectral
lines/features discussed in this paper (or even more lines) in at
least one of the panel.

--Done

page 26, fig 6: dotted line is [a] linear fit  <- 'a' missing

--Fixed