Brought in from another thread....

Originally Posted by
MysticSonic
"Why do you think so?"
Because then inter-type relations are left as a rather impractical bumper to the theory rather than leave it as some sort of predictive tool.
It also undermines the concept that the situation can bring about different effects depending upon a person's functional order, for if the functional order itself is dependant upon the situation, then there could be no correlation---no simply one, anyway.
There are 16 types, which would evenly divided give an amount of 6,75% of your time to be in a single type. But this idea seems silly. It seems probable that most people have a dominant type that they spend at least 50% of their time in, leaving 3,33% of their time per each other cathegory, though most probably even the rest of their time is spent in a type that is close to the original one, maybe more something like this ISFp 55%, ISTp 30%, INFp 11%, others 4%.
Now would there remain significant predictive value? Yes, there would. Do we find that our type descriptions are 100% predictive? No, we don't.
Considering the other point you raise... Something has to create the functions themselves. Type creates personal experience but in some way at some point of life "the situation" decides the type. Humans are biological constructs, constantly in flux, it's very difficult to find any trait that isn't mutable to an extent. If type did not change at all, ever, it would be a most exceptional thing.
Of course, a suggestion that there might be a slow slide from one state to another during a person's lifetime is a far cry from Kamensky's suggestion that a person might be first and ISTp for 5 minutes and then an ISTj for the next 5, but even he himself admits that's an exagerration.
If a dominant type is the norm, which it seems to be according to most evidence, there's little need to know for certain the factors that might make a given person change his type. Socionics that speculates a constant type is reasonably good predictor of relations in the long term if this is true.
Is there any reason to speculate on subtypes if socionics works without them?
Well... that depends... Is there ever a reason to strive to make a useful tool even more useful?