I've read lots of Kai's work and in yeast counting experiments his uncertainty bars come from the uncertainty in counting the cells* - which is by far the largest source of uncertainty in these experiments. So in this case, the uncertainty bars mean that he is absolutely certain that the real value lies somewhere between the upper and lower bars. They mean nothing more and nothing less. The whole point of uncertainty bars is that you really don't know where the data point lies, but you do know the bounds in which it will lie. So in this case, my statement about the uncertainty bars overlapping is correct (unless Kai comes along and tells us that in this case his uncertainty bars represent something different to all his other yeast growth experiments).
Also, when I said "significant difference" I didn't say, or mean, statistically significant.
* This uncertainty is calculated from the variation in cell denisty observed in different grid squares on the haemocytometer. So in a way it is a sort of statistcal analysis, but a basic one, and the bars represent the analogy of 100 % confidence interval.
EDIT: I will concede that it's quite likely that the real value for high speed is higher than that for medium speed, but this data doesn't show that conclusively.
