Jon - I think this question is directed towards you. I have implemented a
version of your broken y-axis scheme. My data looks like the following:

sample average avedev
negative control 0.36252361 0.029793573
sample 1 3.265795907 2.878954305
sample 2 0.149275395 0.035304333
sample 3 73.57998574 2.237144009
sample 4 0.346937718 0.02837823
sample 5 4.636386221 0.089703534
sample 6 709.7039626 16.33742397
sample 7 1.157628293 0.667009325
negative control 0.494505996 0.051731733
RNase A 1pg/ml 1.604749215 0.106664119

As you can see, most of the sample are in the 0-5 range, but nos. 3 & 6 are
way out. The question I have is how do I plot the avedev as error bars when
I have broken the y-axis? I need to adjust these values to reflect the
change I have made to the average series. I have the y-axis broken twice,
with the sample #3 bar broken once and the sample #6 bar broken twice. I
also did not use the method you use to calculate the midpoint of the bar for
placement of the break; rather, I eyeballed it.