The output of speaker drivers rolls off at certain frequencies. For example, a big bass speaker might do well between roughly 30Hz and 300Hz, but not produce much sound outside this range. The outside edges of this range need to be defined somehow, and are usually stated as being the frequency where the output of the driver is reduced by 3dB (3 decibels), compared to its strongest output. So if the output looked like this:

           /                     \
          /                       \
         /                         \
       40Hz       100Hz       200Hz

...if we played white noise through the speaker such that the volume at 100Hz was measured as a pretty loud 90dB, and the low-end output dropped to 87dB at 40Hz, while the high-end output dropped to 87dB at 240Hz, then these two frequencies would be the -3dB points.

Rather than saying "the driver covers the low bass", we could say "the driver has a bandwidth of 40 - 240Hz, defined by its -3dB points". Much more accurate :)

A lot of the effort that goes into designing loudspeaker enclosures is in an effort to shift the lower -3dB point down a few Hz.

Note (about decibels): a 3dB increase in sound level is a doubling of acoustical energy. Therefore a 90 dB sound is 4 times as loud as a 84 dB sound. In practice, a 3dB dip or peak in acoustic output is hardly noticable, so defining the bandwidth of a speaker by it's -3dB points is pretty conservative. Non-conservative definitions may use the -10dB points, or whatever the marketing division thinks they can get away with :)

Note (about this measurement): -3dB points are used to define the output capabilities of other things, such as amplifiers and transmitters. I just think in terms of speakers rather a lot.