correlation versus causation 1

If correlation does not imply causation, what does it imply? Are there ever any circumstances when a correlation can be interpreted as evidence for a causal connection between two variables? If yes, what circumstances?