I like the pulse idea quite a lot and i have a few ideas about solving things (however my electronics skills aren't the most elaborate)...

here's the deal, you can do mostly ANYTHING with a computer and you don't have to worry about programming chips (PIC being the easiest to use, and it's still a pain in the ass to get going)... so, if you take one of those heartbeat monitors from a watch and you figure out how to connect it to a port, you've got yourself a chance of having the rest of the work software based (which a lot more people can do than electronics).

i would assume an lpt port would be able to both power the sensor and record the feedback. all you have to do is setup a program much like the dream maker (is that what it's called? the rem goggles with IR detection and dos-operated software?) to record the rate of the pulses and show a graph at the end of the night... this would show you when individual people have rem's (assuming you'd see it in a graph) AND you'd know what the critical value for signaling is... from there on you can either have a pair of goggles stuck to your face (using the same lpt, built following instructions as in the link that was posted above) or you could make the computer signal to you.

i think light signaling would also be the best option simply because it's been proven to work better than any other stimuli.