The hippocampus, the brain structure responsible for memory, contains many tiny subregions. These subregions are so tightly interconnected that a change in one almost always means a change in its neighbors. That's why it's so hard for scientists to figure out which specific subregion matters for thinking or remembering.
To test whether the right subregion can be found at all, the researchers took data from 638 people. They measured the volumes of 36 hippocampal subregions and compared them with the results of memory and attention tests. At first, no subregion showed a clear link to the test results. So the scientists added an artificial effect to the data, as if they knew exactly which subregion mattered. But even then, the analysis often failed to find it. When the effect was weak, the task became nearly impossible, even with perfect data.
It also turned out that the way the data are processed shapes the result more than the data themselves do. This means that different scientists using different methods can reach different conclusions about the very same thing. So pinpointing hippocampal subregions remains a tough problem for now.
The condition number of the design matrix was 1727, and the median variance inflation factor was 7.3, meaning that information about many subfields is almost entirely duplicated by their neighbors. At the same time, 32 of the 36 bivariate correlations with the cognitive factor were positive, so a negative coefficient in a multivariate model would have to be interpreted as suppression rather than an inverse relationship.
Not published on arXiv. Here are its page, the PDF and the formalization: you can check it yourself, not only read our account.