Converting a sample or other digital signal to a lower sample rate. When a digital signal is downsampled, it is necessary to apply a low pass filter to the signal to reduce the signal's bandwidth to less than the Nyquist rate of the new sample rate; otherwise, aliasing will result. It is sometimes necessary to downsample in order to transfer a sample or other signal to a different system that records or stores at a different sample rate. Additionally, if the old sampling rate is not a multiple of the new one, a process called interpolation is needed in order to produce the sample words at the new rate that are "in between" words at the old rate.
Performers might sometimes downsample a signal as an effect, to achieve a low-fi sound. The anti-aliasing filtering will reduce the bandwidth of the signal and attenuate the high frequencies, and the interpolation process, depending on how it is done, can add noise or "grittienss" to the sound. Some samplers allow the performer to bypass the anti-aliasing filter so that aliasing is produced intentionally.
In the past, downsampling was often performed on samplers in order to save memory. If a sample didn't have much high frequency content, it could be, for instance, downsampled from a 48 kHz sampling rate to 24 kHz, cutting the amount of memory needed to store the sample by half. However, with the advent of software samplers running on computers, available memory is almost infinite, and so there is seldom any need to downsample to conserve memory.
Compare with bit crushing.