I came across a remarkable study that could open up some avenues of empirical research into socionic phenomena.
Some British researchers learned to predict with 94% accuracy whether a new couple would get divorced or not in the next four years based on data from just 15 minutes of conversation.
I highly recommend this radio discussion transcript from 2004. Basically, they would film newlyweds discussing a contentious topic such as family finances and attach point values to various displays of positive and negative emotion, such as humor or contempt. "Marriage success" was defined simply as remaining together after four years. The more positive a couple's points, the greater their chances of staying together. Successful couples, they found, had up to five times more positives than negatives, whereas unsuccessful couples had an equal number or more negatives than positives.
Furthermore, by looking at the data and observing couples, they were able to distinguish five general patterns of interaction: (source)
1. "the validating couple" -- calm, intimate, companionable, like to back each other up
2. "the avoiders"-- do their best to eschew conflict.
1. "the hostiles" -- neither person wants to talk about contentious issues
2. "the hostile-detached couple" -- one is fiery and argumentative, and the other isn't.
Borderline couples (can go either way)
1. "the volatile couple" -- romantic and passionate, but have heated arguments
If you're interested in learning about their methodology and findings in depth, their book The Mathematics of Marriage is available online at Google books.