Two separate experiments in Colorado compare the frequency of emissions from atoms or ions to an uncertainty of 10^-16 or better. Earlier atomic clocks operated by reading out the movements of internal transitions from one quantum state to another in cesium atoms; the light emitted was in the microwave range. With frequency comb techniques (http://www.aip.org/pnu/2008/split/853-1.html) measurement of optical-range frequencies can also be made with high accuracy.
In the 28 March 2008 issue of Science two groups reported new superb levels of precision. One experiment, carried out by a JILA/Colorado/ NIST-Boulder team (Ludlow et al.), gauges the uncertainty of a clock based on neutral strontium atoms to a level of 10^-16 by comparing it to a clock using calcium atoms and located a kilometer away. The other experiment, carried out at NIST-Boulder (Rosenband et al), looks at two clocks 100 meters apart.
The clocks contain respectively a single aluminum and a single mercury ion. The fractional uncertainty in the ratio of the frequency outputs for the clocks was determined to be 5.2 x 10^-17. (NIST information at http://www.nist.gov/public_affairs/releases/logic_clock/logic_clock.html and http://www.nist.gov/public_affairs/clock/clock.html )