The performance of a 16-channel receive-only RF coil for brain imaging at 3.0 Tesla was investigated using a custom-built 16-channel receiver. Both the image signal-to-noise ratio (SNR) and the noise amplification (g-factor) in sensitivity-encoding (SENSE) parallel imaging applications were quantitatively evaluated. Furthermore, the performance was compared with that of hypothetical coils with one, two, four, and eight elements (n) by combining channels in software during image reconstruction. As expected, both the g-factor and SNR improved substantially with n. Compared to an equivalent (simulated) single-element coil, the 16-channel coil showed a 1.87-fold average increase in brain SNR. This was mainly due to an increase in SNR in the peripheral brain (an up to threefold SNR increase), whereas the SNR increase in the center of the brain was 4%. The incremental SNR gains became relatively small at large n, with a 9% gain observed when n was increased from 8 to 16. Compared to the (larger) product birdcage head coil, SNR increased by close to a factor of 2 in the center, and by up to a factor of 6 in the periphery of the brain. For low SENSE acceleration (rate-2), g-factors leveled off for n>4, and improved only slightly (1.4% averaged over brain) going from n=8 to n=16. Improvements in g for n>8 were larger for higher acceleration rates, with the improvement for rate-3 averaging 12.0%.