I guess so the sound will fade in and fade out so it's a more gentle beginning and end to the signal. Without windowing it's possible the sound could start abruptly and very loudly, or suddenly stop which could sounds weird. That's all in the time domain.
It could also reduce spurious high frequencies in the spectral domain which can happen if the left side and right side of the signal don't match up because the FFT assumes a periodic signal. Mismatch at right and left edge of the signal will give a sharp edge in the periodic extension of the signal which would create spurious high frequencies.
