You're not missing anything. As user joncrocks comments in this subthread, the actual study[1] does not do anything other than measure the structural changes. The "why" of the grey matter alterations (which appear to primarily occur in the prefrontal cortex and hippocampus regions) and their resulting impacts are not quantified at all beyond offering hypotheses for future studies.
It is disingenuous to call these alterations "improvements" and any attempt at extrapolating the results to real-world improvements in e.g. perceiving threats in strangers is complete speculation. There simply isn't any material in the research to justify labeling these changes as positive or negative.
The authors do give a few good-faith remarks about how these changes could improve e.g. facial recognition (for a certain subset of defined metrics to satisfy "improve", of course), but there isn't actually evidence to support this. It's conjecture based on what we know of the brain.
What we can reasonably say based on this research is that women experience grey matter reductions after pregnancy that may have an impact on their executive and social functioning skills (if we extrapolate from the specific areas impacted in the brain). Obviously, this depends upon the statistical rigor involved in the study's interpretation as well, but I'm not going to comment on that.
As a meta point, this study is another disappointing demonstration of the way in which the media frames legitimate studies and primes scientific discourse. It is excellent research that stands on its own, and a great contribution to neuroscience at that. There is no need to color the results with speculation, which is cleverly couched between actual results in an optimal context. My partner brought this study up with me at dinner last night, as it had already started making the rounds. Naturally, when I asked her for concrete examples of these "improvements" and how they were supported by the study, she couldn't offer any, and was annoyed by my questioning. Not only did the articles do an intellectual disservice to her (and others!) by priming her expectations with illegitimate results, they encouraged a "pre-chewed" version of the study that would not hold up to real challenge or inquiry. I subscribe to Nature (and many of its subjournals), and I absolutely refuse to read scientific literature from secondary sources if I can find the study. But this isn't a reasonable expectation for everyone; in a world where most people aren't going to click through to read a 4 - 20 page study, this sort of coverage diminishes the general public's appreciation of scientific results at best, and encourages dangerous interpretations of legitimate data at worst.
________________________
1. http://www.nature.com/neuro/journal/vaop/ncurrent/pdf/nn.445...