Experiments in psychology and sociology are plagued by uncontrolled (and perhaps uncontrollable) variables and difficulties in replication. For any engineer or scientist used to the somewhat firmer ground of the “hard” sciences, this makes investigations into human behavior seem like something other than science and closer to opinion. It doesn't help that facets of these softer sciences that are believed to be true are routinely and frequently overturned.
On that theme, here's an article exploring one of the more recent such overturnings. My main reactions after reading this: (1) don't believe anything is actually “known” in those fields, and (2) I'm sure glad my own areas of interest aren't quite so ambiguous and difficult to measure – my oscilloscope and voltmeter work every time, with perfectly repeatable results :)