Simple similarity comparison over time

I've seen that Pearson's r is typically used for evaluation between non-experimental conditions. However, I've also seen Pearson's r used to indicate effect size - is this acceptable? Is it a useful descriptive statistic at least?

For example, there are two groups of students. In one group students are put into pairs to do a task, in the other they aren't. (This is a task whereby improvement is to be expected over time.) There is a moderate effect size between time and similarity between partners after being grouped. Is this enough: r = 0.25 for paired students over time, and 0.67, 0.4, 0.3 after grouping by pairs. (It seems almost beside the point to indicate that nonpaired individuals showed no significant similarity to everyone else over time.)

-
R is a measure of correlation. If your effect is correlation, then it is appropriate. I have no idea what your example is about. –  Peter Flom Oct 27 '12 at 14:46
I think correlation is a meaningful assessment of the effect. The test is to examine a few things. Primarily, I'm interested in seeing if task similarity among partners increases meaningfully after being placed in pairs. The data is logged whenever a task component is completed - not in fixed increments (though there are several per minute in all instances). –  Donnied Oct 27 '12 at 14:49
If you are investigating differences between groups, correlation is probably not what you want. Perhaps t-test, perhaps regression. –  Peter Flom Oct 27 '12 at 14:53
Yeah t-test between groups and repeated measures (lme or lmer (or aov?)) in group seems the way to go. Though I have the sense I should look at some interrupted time series analyses. –  Donnied Oct 31 '12 at 13:31