It helps to think of it as a pawn capturing a pawn "in passing", the literal interpreatation of "en passant". In other words, the capturing pawn takes the pawn just moved in the middle of it's journey forward, during a brief moment of vulnerability.
But do read the faqs to get the specific details on the setup and execution of "en passant" captures.
Fortunately, it is almost always optional, and occurs infrequently, which means your game won't suffer too much if you decided to ignore the rule for sake of simplicity, so long as you can accept the rare occasion where it negates the effectiveness of someone's initial pawn thrust (yours or theirs).