We describe the spectroscopic target selection for the Galaxy And Mass Assembly (GAMA) survey. The input catalogue is drawn from the Sloan Digital Sky Survey (SDSS) and UKIRT Infrared Deep Sky Survey (UKIDSS). The aim is to measure redshifts for galaxies in three 4x12 degree regions at 9h, 12h and 14.5h, on the celestial equator, with magnitude selections r < 19.4, z < 18.2 and K(AB) < 17.6 over all three regions, and r < 19.8 in the 12-h region. The target density is 1080 per sq. deg. in the 12-h region and 720 per sq. deg. in the other regions. The average GAMA target density and area are compared with completed and ongoing galaxy redshift surveys. The GAMA survey implements a highly complete star-galaxy separation that jointly uses an intensity-profile separator (delta(sg) = r-band psf mag - model mag) as per the SDSS) and a colour separator. The colour separator is defined as delta(sg,jk) = J - K - f(g-i), where f(g-i) is a quadratic fit to the J-K colour of the stellar locus over the range 0.3 < g-i < 2.3. All galaxy populations investigated are well separated with delta(sg,jk) > 0.2. From two years out of a three-year AAOmega program on the Anglo-Australian Telescope, we have obtained 79599 unique galaxy redshifts. Previously known redshifts in the GAMA region bring the total up to 98497. The median galaxy redshift is 0.2 with 99% at z < 0.5. We present some of the global statistical properties of the survey, including K-band galaxy counts, colour-redshift relations and preliminary n(z). Comment: 15 pages, 15 figures, accepted by MNRAS; v2 includes new figure with K-band galaxy counts, more description of UKIDSS reduction, and other minor changes