I have an interesting history with the MMPI. I've taken the inventory...twice when in-patient for a year. Then 3 years later I took a graduate school course on some of its design features in a biostatistics class. What I can tell you about those random odd questions, my favorite was "I like cheese", Is that the MMPI has been administered to literally 10s of millions of people since it's development 50 or so years ago. That's a huge sample size. What they were able to do was statistically correlate answers to seemingly random questions to certain mental illnesses. It's an internal check to see if someone might be trying to trick the exam. Lets say you want the examiner to believe you have a particular mental illness. So you answer all the obvious questions in a way you think someone with that illness might answer. What you don't know is built into the test are various questions that aren't obvious. These are questions someone who really does have the illness has a high probability of answering only one way. There are many of these questions. Statistically, if you have the illness you are likely answer them in the same way others with your illness have answered them. If you don't, that clues the scorer into the fact that you may be trying to manipulate the test. Some people try to hide an illness they have. Some people want a certain diagnosis. Either way, answers come up as untruthful.

There is also a lie scare built into the test. It also statistically measures whether your answers were truthful based on how answers to different questions correlate with each other.

The first time I took the MMPI I was severely psychotic and in a deep trauma state. My in-patient psychologist told me my scores made no sense and were not interpretable. She said I was too psychotic to have been administered the test. Why they ever bothered I don't know. When the psychotic episode and trauma resolved, they got what turned out to be an uncomfortably accurate measure. I found it creepy that a paper and pencil test could predict with great accuracy my diagnoses and if I was telling the truth. I got curious about how the test was designed and when I had a chance to take a class that included MMPI design features, I took it.

The bottom line. It's not a test you can easily get away with manipulating.