Fully-contained neutrino interactions provide the greatest amount of information. An event is fully contained if the neutrino interacts within the volume of the inner detector, and no particles escape. To avoid the possibility of contamination from entering particles, only events originating greater than two meters from the surface of the PMT's are accepted. This makes the effective target volume 22,500 tons.
For the neutrino-oscillation analysis, only events with a single Cherenkov ring are considered. In such events, electron-neutrino interactions almost always result in a visible electron track, and muon-neutrino interactions produce a muon. The task of identifying the type of neutrino then simplifies to task of determining whether an electron or a muon was observed.
The Cherenkov ring patterns produced by electrons and muons are significantly different. Because electrons are very light, they do not travel far in the water before being knocked off course, often radiating a high-energy gamma ray (photon). Each photon likewise travels only a short distance before splitting into an electron/anti-electron pair, with each secondary particle heading in a slightly different direction than the original electron. This process is known as an electromagnetic "shower". As a result a single high-energy electron eventually generates hundreds of low-energy electrons, each of which travels only a short distance and makes its own faint Cherenkov ring. It is the sum of the rings from these many short, divergent tracks which is recorded by the light sensors, and hence a blurred, diffuse Cherenkov pattern is observed, rather than the ideal case illustrated above. On the other hand, muons are much heavier than electrons, and hence do not get knocked off course. Instead, they maintain their initial direction, and travel a distance roughly proportional to their initial energy before stopping. The Cherenkov pattern of a muon corresponds much more closely to the ideal case. Tests with real electrons and muons, as well as simulated data, show that the probability of mistaking an electron for a muon (or vice versa) is less than 2%; hence the number of identified electrons and muons should accurately track the number of electron- and muon-neutrino interactions.
Based on comparisons between the number of interactions predicted and the actual data, some 35% of the expected muon-neutrino interactions are missing in the sample of data in which the total visible energy is less than about 1.4 times the mass of a proton (this particular energy value is chosen for historical reasons, to compare with previous experiments in which an identical choice was made). This sample of fully-contained events below the within that energy region is called the "sub-GeV sample", a GeV being a unit of energy.