The University of Bath put out a press release about some research suggesting that motorists leave less berth for cyclists with helmets than those without, which presumably increases the chances of a crash. "This study suggests wearing a helmet might make a collision more likely in the first place," says the researcher, traffic psychologist Ian Walker, who was in two crashes while doing the research, both during the half when he was wearing a helmet. Helmet boosters didn't like these dirty lies from Bath, and predictably -- perhaps accurately -- said the things are still worth it.
As for why drivers give the helmeted less room, Walker says it's because they think bicyclists wearing helmets are more experienced, skilled, and predictable, so they can get closer without hitting them. At the risk of crossing a real traffic psychologist (I've heard they can trick people into driving into trees), I'd suggest a different explanation: drivers are consuming the safety that bicyclists gain by wearing helmets. I.e., they feel that the bicyclists with helmets are relatively safe in the event of an accident, so they don't need to worry much about hitting them.
This "risk compensation" is often seen in people who get access to new safety measures and then start doing more dangerous things, e.g., better skydiving gear -> more dangerous dives, seat belts -> more dangerous driving, better AIDS drugs -> more unsafesex, sunscreen -> more sun exposure and potentially more skin cancer, etc. The bicycle example is a little different in that in the other cases, people "consume" their own safety benefits, whereas motorists are consuming someone else's increased safety -- a rare case of risk compensation/moral hazard. (Which is particularly galling considering that decent road bicyclists only need helmets to protect themselves from cars.) Of course, judging by how people sometimes change when they get into a driver's seat, it might not be that surprising.