[misnome]

I'd be very hesitant to make that conclusion. Much more likely is that more nuanced middle-ground questions are a lot more open to individual interpretation than the "extreme" statements, and thus give you a lot less "signal" in what you are trying to measure - even some of these could be seen as widely open to interpretation.

I'm guessing the scale (rather than agree/disagree) is attempting to compensate for this somewhat.

(Caveat: I'm a scientist but not with a background in anything close to Psychology; I've seen but not written these sorts of tests).