If you had a graph of phenomenon A and it matched exactly phenomenon B and the sampling was pretty large, you'd at least want to see if there was some causal relationship.
Definitely, you'd want to see -- and that would take more than a graph showing CORRELATION. That's because it's prefectly possible that phenomenon A causes a completely unmeasured phenomenon C, and it is phenomenon C that causes phenomenon B. You don't want to go around waving your correlation and raving about how A causes B, because you look kind of silly when phenomenon D shows up and independently wipes out C. Because then you've still got your A, but B doesn't come to the party, and you get discredited and loose your grant. It's also possible that A & B are results of some other cause C, and when D shows up and crushes A, you look silly again when B is still hanging around. Also, B might be the cause. And, even though these guys happen at the same place and time... they really might have absolutely nothing to do with each other.
So... let's repeat. Correlation does not imply Causality. Good.
Machines have less problems. I'd like to be a machine. -- Andy Warhol