What is happening is that you can think of a personality profile as a vector in very high-dimensional space. MBTI presents a basis in a 4-dimensional subspace of that. How well it works for a particular person depends on how her personality aligns with that basis.
A lot people will test near the middle on the MBTI characteristics, and say that the test does not work for them. If most of the population is near the middle, statistical studies won't show much validation for MBTI either.
Now, none of that means that the test has little value -- it actually has a lot of value, but only for people who test at relatively high scores. For example, I generally test as either ENTP or INTP -- and since I'm near the middle of E/I scale, it does not tell much either way. my N, T, and P, however, are quite extreme (I've done quite a few of these, and not once have I had a case when my N or T score was not at the end of the range, P nearly there too) -- so the descriptions and advice for these types that I've seen have been quite relevant and helpful.
Exactly... there are hundreds of systems to evaluate people. All of them are abstractions which are unable to capture the full depth of reality, and so will work well under some conditions and bad under others. It's a matter of design choices and tradeoffs. Choose the system that better fits your needs and be aware of its limitations.
That said, I like MBTI's (or Keirsey's) simplicity. It gives you a lot of data for such a simple system, and is easy to apply. You have to be aware of its limitations, of course. But I've been playing with it for almost 10 years (in my personal life, not so much for work) and it rarely fails me.
I think it's the same principle as the Fibonacci scale for planning poker... you will always have some error in your system, don't fool yourself. So the ROI of a simple system might pay off...
PS: I agree, MBTI seems VERY accurate for people that are not in the middle of the scales, and not-so-accurate otherwise.
The classification would be more meaningful, I think, if there was a good way to characterize certainty/degree.
In my case, I and T are both generally very strong, while N has been weaker.
The most interesting part is the J/P has gradually shifted. I used to test as a very strong J, but that has been changing and today I finally flipped and tested P.
What is happening is that you can think of a personality profile as a vector in very high-dimensional space. MBTI presents a basis in a 4-dimensional subspace of that. How well it works for a particular person depends on how her personality aligns with that basis.
A lot people will test near the middle on the MBTI characteristics, and say that the test does not work for them. If most of the population is near the middle, statistical studies won't show much validation for MBTI either.
Now, none of that means that the test has little value -- it actually has a lot of value, but only for people who test at relatively high scores. For example, I generally test as either ENTP or INTP -- and since I'm near the middle of E/I scale, it does not tell much either way. my N, T, and P, however, are quite extreme (I've done quite a few of these, and not once have I had a case when my N or T score was not at the end of the range, P nearly there too) -- so the descriptions and advice for these types that I've seen have been quite relevant and helpful.