I think the distinction is between hard and soft sciences. Hard sciences like chemistry, physics, mathematics are very reliable. Barring some very special cases (where an explanation hasn't been found yet) identical conditions will produce identical results. So they're not only reproducible but you can use one to deduce the other.

Soft science like economics, psychology, sociology will have none of the 2: conditions vary, results vary, and even if you could ensure you start with absolutely identical conditions it's very implausible that the result won't be heavily influenced later on by random factors.