Returning to this, I did the calculations and here is the margin of error for each group: 11%, 9%, 9%. Their common error (the overlapping region of area) is the interval 87.79-104.95. The IQ scores from each group (94.91, 96.29, 96.47) are well within that margin of error. If you perform a t-test on 3-Ca vs. 3-placibo you'll see that the difference is not statistically significant.

How is this study concluding these results? What math are they using?