Item response theory is the idea behind most adaptive standardized tests like the GRE. The basic premise is that an individual's underlying ability (as approximated by their GRE score) is being estimated after each question, and subsequent questions can test the soundness of that estimate. Each question is associated with a probability that a person with a certain underlying ability X will answer correctly, and so getting a hard question wrong is evidence that your ability is lower than that needed to answer the hard question correctly. As the test goes on, the estimate of the underlying ability becomes more precise (it narrows in on a smaller range of possible ability levels), and that is why the swings in difficulty get smaller as the test goes on. Usually, the duration of the test is set so that the precision of the estimate lies within a certain acceptable range. Most people have an incorrect understanding of how this works, and so they get overly anxious about the hard questions (as if they carried more weight). It's true that getting them wrong leads to a lower estimate of their ability, but that's precisely the point-- if the test-taker truly had the ability to answer the question, they would have gotten it right.

Hallucinations, by definition, are perceptual experiences that aren't caused by an external stimulus. They too are associated with brain activity in the appropriate sensory areas. For example, see this paper by Ffytche et al. (1998), which describes activity in various parts of visual cortex during visual hallucinations. So even if you do detect activity in some sensory area, you wouldn't be able to rule out an internally-generated cause. Detecting activity would merely confirm that the perceiver's brain is acting as if it is perceiving something...

Skim on OS X. Does forms, embedded notes, highlighting, the works. It's much more powerful than Preview and only takes a tiny bit longer to start up. It seems to be updated very frequently.