The steady state response is a sustained oscillatory response to a stimulus that is varying in strength with some regularity. But it is not a true oscillation, it's closer to an ERP in nature: it is a stimulus-driven response to variations in the stimulation.
In auditory processing, which I am familiar with, people take a tone and then add an amplitude modulation to it. This means that the sound level increases and decreases regularly at a certain frequency. Increases and decreases in neural stimulation are clearly visible in the brain, so if you average the frequency decomposition of your signal over all trials, you get a response at the frequency of the amplitude modulation. If you then manage to change something about this frequency response (e.g. the power) over different experimental manipulations, then most likely you've found some true top-down effect.
In vision I assume it would be similar if you flicker a picture: the stimulation will increase and decrease regularly and you should get a response at the frequency of the flicker. I haven't got a clue what would happen if you would split the frequency rate for the two eyes though.
One thing that mystifies me is that in audition, the steady-state response is particularly well picked up if the amplitude modulation is around 20 or 40 Hz. I have no idea why this is. I also don't really understand what the advantages are over just recording brain activity and looking at the ERP. With the steady-state response, you're in danger of covering up genuinely interesting oscillatory activity with the frequency of the flicker. Also, I'd expect that you'd also pick up a flicker that matches the frequency of the refresh rate of your display. I don't know how that would influence the steady state response (we don't have that problem in audition).