The method involves recording interactions of the couple, and then scoring verbal and non-verbal signals numerically for a range of factors (such as 'whining' and 'affection'). These numerical values are then fed into the modelling process, and Hey Presto.
I don't have a problem with the conclusions of the report, which are basically that couples can survive many things, so long as they continue to be 'influenced' by each other. That matches my intuition, the world is more likely to end in ice than in fire.
"He has argued, for example, that a crucial predictor of
divorce is a husband's inability or unwillingness to be
influenced by his wife's suggestions and emotional
expressions. This husbandly tone-deafness, he says, augurs
divorce much more strongly than, say, the general level of
anger expressed in the marriage. "
I'm just amazed at the given success-rate. 90% predictability is amazingly high for a social research project. Possible explanations that occur is that the sample was very highly skewed towards one outcome or another (ie perhaps nearly all the couples stayed together) or that the researchers were unconsciously picking up other strong predictors of marital stability (such as ethnic or religious characteristics) and allowing this to influence their 'scoring' of the couple's interaction.
I don't want to be too cynical though.