When we want to store and play audio in a computer, we must convert the analogue sound into digital data first. This process is called digitalization. It converts an electronic signal into a sequence of digital values.
Digitalization of the electronic signal
The conversion can be understood as a repetitive measurement of the electronic signal's value at certain time, thus taking a sample of the signal. The result is then encoded as a digital value.
The sampling could be done in arbitrary distances or in constant intervals. The later method is much easier to handle, and thus it is normally used, with a constant rate - the so-called sample rate. Usual sample rates are 8000, 11025, 22050, and 44100 samples per second. In practice sample rates are also given as frequencies, in Hz or kHz.
The sample rate limits the highest frequency a digitized signal can represent. Due to Shannon's theorem the highest usable frequency is half of the sample rate, so with 44.1 kHz sample rate you cannot sample signals with more than 22 kHz. To avoid a violation of that half-sample rate rule, your soundcard already has built-in filters that filter away frequencies that are higher than half of the used sample rate.