Nothing in science can ever fully be proven or disproven, but the more studies that have found significant or non-significant results on a certain topic, the more sure we are of the results' validity. For example, many studies have been done about vaccines being linked to autism and none has found a link of vaccines to autism. Maybe a very small number of children already prewired for ASDs have a traumatic reaction to the vaccine and have their symptoms start to show shortly afterwards, but from a scientific standpoint, the vaccine claim is moot. But this is really only because NO study on vaccines has found significance.
Most neuropsych disorders out there, ASDs especially, have the research literature filled with mixed, often contradictory results/findings. So, in these instances, finding significance sometimes but not others means that there is no clear-cut answer. Were the contradictory results brought about by type-I error or type-II error in one or more studies? Does a specific abnormality occur in only a small amount of participants, and sampling error turned out to have a large amount of these participants all be in the same test group? That's why it's particularly hard to sort out the neuropsych literature, because many of the claims may play an important factor in the disorder in a small subgroup of individuals. So, often, what we're seeing is that we tend to have to make a choice: either sacrifice symptom homogeneity to get increased power from a larger sample size or sacrifice power by using small groups with symptom homogeneity. And, of course, the first option may cause type-I error, whereas the second option may cause type-II error. It's very frustrating, because it's like you can't win!
_________________
Helinger: Now, what do you see, John?
Nash: Recognition...
Helinger: Well, try seeing accomplishment!
Nash: Is there a difference?