"Theory-testing in psychology and physics: a methodological paradox" (Meehl 1967; excerpts) makes an interesting argument: because NHST encourages psychologists to frame their predictions in directional terms (non-zero point estimates) and because everything is correlated with everything (see Cohen), the possible amount of confirmation for any particular psychology theory compared to a 'random theory' - which predicts the sign at random - is going to be very limited.