STATS MAKE ME CRY

Hi,

I am running a study to determine the effect of technology-based collaboration on learning outcomes of older and younger adults. In the control condition, participants are individually reviewing training material and in the experimental condition, participants are discussing training material with other participants in a chatroom. I am also looking at turn-taking in the chatroom as a way of shrinking the age-related learning performance gap. Essentially, the design is a 2 (old, young) X 2 (individual, collaboration) with turn-taking nested within the collaboration condition. So, if I had 80 subjects, cell sample sizes would be:

old, collaboration + turn taking (10)
old, collaboration w/out turn taking (10)
young, collaboration + turn taking (10)
young, collaboration w/out turn taking (10)

Control:
old, individual (20)
young individual (20)

How would I do a power analysis to determine the necessary sample size?