I am seeing some inconsistencies with regards to finding the upper and lower quartiles when you you have an odd number in your data set.

Which is correct, counting the median in both your upper and lower set, or ignoring it?