I'm pretty sure that you would actually use two different frequencies: One to represent a zero, one to represent a mark/one.
That would prevent a loss of modulation to be interpreted as a zero.
Most frequencies, when assigned as "channels" (like FM radio) is actually a chunk of spectrum that represents the fundamental center, plus the sum of fundamental + max modulating signal frequency and the difference of the fundamental and the maximum modulating signal frequency.
When you listen to FM radio, the audio at any given instance in time represents a number that is the difference, the delta, the change of frequency from the fundamental.
By assigning the "frequencies" as "channels," it prevents an interfering overlap.
If you look at the channel assignments for 802.11b or 802.11g, you'll see that each channel uses 25MHz ... but the channels are only 5 MHz apart. That's why channels 1, 6, and 11 are called "non-overlapping" channels. They don't overlap each other, but they will overlap adjacent channels.
More common modulation techniques for data are Phased-Shift Keying (PSK), Quadrature Phase-Shift Keying (QPSK) and variants.
PSK-style modulation shifts the phase of the signal to represent a one or zero, QPSK can represent multiple bits, four positions give you 00, 01, 10, and 11 ... so you are essentially sending two bits with one phase shift ... (kinda) like doubling your throughput.
http://www.howstuffworks.com probably has a pretty decent explanation, give it a look.
Good Luck
Scott
