[B]etween close relatives, a variety of neuropsychiatric disorders covary strongly with intellectual interests. We surveyed an entire class of high-functioning young adults at an elite university for prospective major, familial incidence of neuropsychiatric disorders, and demographic and attitudinal questions.
Students aspiring to technical majors (science/mathematics/engineering) were more likely than other students to report a sibling with an autism spectrum disorder (p = 0.037). Conversely, students interested in the humanities were more likely to report a family member with major depressive disorder (p = 8.8×10−4), bipolar disorder (p = 0.027), or substance abuse problems (p = 1.9×10−6). A combined PREdisposition for Subject MattEr (PRESUME) score based on these disorders was strongly predictive of subject matter interests (p = 9.6×10−8). Our results suggest that shared genetic (and perhaps environmental) factors may both predispose for heritable neuropsychiatric disorders and influence the development of intellectual interests.
From here (Campbell BC, Wang SS-H "Familial Linkage between Neuropsychiatric Disorders and Intellectual Interests." (2012) PLoS ONE 7(1): e30405. doi:10.1371/journal.pone.0030405).
The usual cutoff for statistical significance in this kind of study is p=.05, with lower p values indicating greater statistical significance and some studies tolerating and giving some weight to less significant results. These results are considerably better and the associations observed are the ones that you would generally expect, guessing naiively.
An extremely significant connection between humanities and a family history of substance abuse that is much stronger than the link to bipolar disorder which is well developed in the psychological literature is particularly surprising. Perhaps this could be because full blown clinical bipolar disorder is sufficiently rare that even a large sample isn't big enough to establish much statistical significant due to the small number of instances with familial bipolar disorder relative to then number of instances with substance abuse or unipolar major depressive disorder, which are much more common and hence produce larger and most statistically significant samples. This analysis seems to be supported by this description of the sample from the Princeton Class of 2014 (citations omitted):
We received 1077 responses from 1313 students, a response rate of 82%. 527 respondents indicated a technical major (natural sciences, engineering, or mathematics), 394 indicated non-technical majors (245 in social sciences, 149 in humanities), and 156 students were undecided. A follow-up survey determined the mean number of siblings per student to be 1.5.
We began by looking for a previously reported relation between ASDs and technical interests. Twenty-four freshmen (2.2%, 1 in 45) reported having a sibling with an ASD. This included 16 of 527 (3.0%, 1 in 33) aspiring technical majors compared with 4 of 394 (1.0%, 1 in 99) nontechnical students, for an odds ratio of 3.05 (χ2 = 4.33, p = 0.037). Thus the incidence of ASDs amongst siblings of technical majors was significantly higher than that of non-technical majors and roughly twice (after correction based on number of siblings) the US average of 1 in 160. . . . The two other disorders in this group were also most frequent in the families of prospective humanities majors but did not reach significance (ADHD, p = 0.10; PTSD, p = 0.31). From the second group, Alzheimer's approached significance but did not reach it (p = 0.069). The incidence of Parkinson's, stroke, and traumatic brain injury was roughly constant across subject matter interest.
The incidence of substance abuse was 167 cases, of major depression was 150, of bipolar disorder was 51, and of autism spectrum disorder was 20. There were 85 ADHD cases, 24 for PTSD, 145 for Alzheimer's, 53 for Parkinson's, 26 for traumatic brain injury, and 228 for stroke.
Since the relationship between sample size and statistical power is non-linear, a sample size of 51 has much less than a third of the statistical power of a sample size of 167 or 150 cases. The ADHD statistics also suggest that an adjustment for statistical power resulting from low incidence size that is related in a non-linear way to statistical power may be in order. Put another way, it is reasonable to expect that there is a good chance that a larger sample size might increase the statistical significance of the ADHD and bipolar associations, both of which are known to have a strong heritable component.
The PTSD sample was very small, is the poster child for a non-genetic condition, and could be attributed to a hidden missing variable, which is family involvement in undertakings that could give rise to exposure to traumatic events, rather than a subclinical disposition towards the condition itself.
Of course, it also bears considering that this is a classic "WEIRD" sample that may be atypical of the general population. Still, it is pretty impressive that such an elementary kind of study is among the low hanging fruit of obvious to conduct and interesting studies that had not previously been conducted. The amount of very elementary research that is left to be done in pscyhology is much greater than in the "hard sciences" where there is very little low hanging fruit easy but valuable experiments that illuminate core issues in the discipline left to be done.