Figure 3 is a great illustration of why the p-value is not a useful metric in a poorly powered study like this: The confidence intervals overlap, indicating that the true average population IQ could be anywhere inbetween (to some degree of certainty) , yet one would think from the p-value of 0.03 that the effect would be confirmed and obvious. Past that easy criticism, I stay skeptical of most longitudinal studies because they usually fail to account for very likely biases in the data, as your link #2 seems to indicate. Show me a RCT and I'll reconsider their conclusion. Bonus challenge to other readers: I've yet to find any study that meaningfully convinces me that intelligence (above the norm) correlates with anything other than education. Convince me that the entire field of study isn't garbage. Edit: Erm, I am dumb: not correlates with education, is positively effected by education, a healthy lifestyle, and general socioeconomic factors.