Are you using discontinuous gels (the kind with stacking and resolving portions)? I have heard an explanation for why glycine is used in the running buffer for those gels. The idea is to keep proteins from separating by size until they reach the resolving portion. To keep the protein in a tight band while it’s in the resolving gel, glycine is used to form one end of a "moving boundary."

In a typical Western blot, the sample buffer contains Tris-HCl and the running buffer contains glycine. When you have loaded your samples and applied a current, your proteins, the chloride ions, and the glycine start to flow throw the gel. Glycine is a zwitterion in the stacking portion of the gel (~pH 7). The chloride ions take the lead and the glycine in the running buffer takes up the rear: your protein sample gets trapped in between them in a "moving boundary." Once the pH 9 resolving gel is reached, glycine passes the protein sample because it ionizes and migrates faster. With the boundary gone, the protein can migrate by length/charge.

I’m no good at pchem so this explanation may be bunk, but I picture the negatively-charged (SDS-coated) protein being repulsed by the chloride ahead of it and the zwitterionic glycine behind it. Crummy explanations aside, the moving boundary is the reason why your proteins don’t separate by size in the stacking gel. In the resolving gel, glycine ions from the running buffer continue to pass the sample throughout the run, but no longer form a front that pushes the protein along.

I assume glycine does the same thing in the transfer buffer: push the protein forward onto the membrane through (assuming again) repulsive force. This explanation isn’t very satisfying though because the pH under transfer conditions is similar to that in the resolving gel itself – wouldn’t glycine ions just whiz by like in the stacking gel? Maybe it is just used along with Tris for buffering.