I'll try to answer without getting too involved and boring you to tears with the sort of gibberish that is commonplace around Silicon Valley.

In many wireless adapators/routers/etc today, there are really at least two standards: 802.11b, which had a top speed of 11Mbps, and 802.11g, which has a top speed of 54Mbps.

These two modes of wireless networking use two different modulation methods.

The faster (802.11g) modulation is more "complex" and needs a better signal quality (less interference, less noise, more strength) because it has less "redundancy" in it. These wireless networking products have redundancy built into their modulation methods in the event there is a signal drop-out, multipath fading/bounce, interference (eg, you have a 2.4 GHz wireless network... and you have a 2.4 Ghz wireless phone right next to it -- that can interfere with each other, but you don't hear it as interference -- what you see is a decrease in speed or range). The slower modulation methods (eg, 802.11b) have more redundancy built into the protocol, which means that more of the signal bandwidth is devoted to guaranteeing data quality than at the higher speeds (802.11g).

When the signal quality degrades to a point where the 802.11g modulation methods won't transmit the signal with sufficient quality, the wireless networking gear "falls back" to a slower, more reliable transmission method, much as modems with adaptive rate modulation would fall back to a slower speed in the face of phone line noise. It does this in two ways -- first, it will vary the "coding rate" -- the manner in which your actual payload bits are encoded into the signal along with "error correction" bits. As the signal quality goes down, more error correction bits are encoded into the signal. There's a finite amount of data that can be transmitted in the given bandwidth at the prevailing gain, so as more error correction bits are added, fewer payload bits can be transmitted.

Then, at some point, all the error correction in the world won't keep the link worthy, so the actual modulation modulation method is changed to one less susceptible to interference, noise, low strength, etc. Typically, this means that you have to give up bandwidth. You can think of this in ham radio terms, if you'd like. You can hear a CW signal when you can barely hear a AM or SSB signal -- but CW has a lower bandwidth than AM or SSB. Same sort of thing is going on here, only we're now talking of the difference between 64-QAM (high bandwidth) vs. 16-QAM modulation (lower bandwidth).