I'd thought direct observation of the photo would mean physically interact with it (and thus changing its behaviour). My understanding is that there are also other experiment designs which allow identifying which path a photon went after it passed the slits; in such cases the interference pattern disappears too.