The selection of tau-pair candidates at SLD (as given in Ref. ) is based on the multiplicity, momentum and direction of tracks in the central drift chamber, and on properties of electromagnetic showers in the calorimeter. Candidates are required to have at least two but fewer than seven tracks. Each event is divided into hemispheres by the plane normal to the track with the highest momentum. Tracks in each hemisphere must fall within cone of the net momentum vector in the hemisphere, and the cone invariant mass in each hemisphere is required to be less than 2.3 GeV/c. Furthermore, the cone axes in the two hemispheres must be back-to-back within . These criteria discriminate strongly against background from multi-hadron final states. The polar angle of the missing momentum in each event is required to satisfy to discriminate against two-photon interactions and Bhabha events. Two-prong events are required to have a minimum acolinearity of 10 mrad. The scalar sum of the momenta of the two stiffest tracks in any event must be less than 65 GeV/c. These cuts primarily reject Bhabha events and muon-pair final states.
The total visible energy in an event is required to be at least 12%of the center-of-mass energy () to reject two-photon interactions. To discriminate against Bhabha events, the total energy deposited in the electromagnetic section of the calorimeter is required to be less than , and the most energetic EM cluster must be less than . In addition, the total calorimeter energy not included in a cone is required to be less than 5 GeV, and there must be fewer than six energy clusters not included in the cones.
These criteria resulted in a sample of 4522 tau-pair candidates selected from the 1993-1995 data. The event selection efficiency and background contamination were estimated using Monte Carlo. The production of tau-pair events at the resonance was simulated using the KORALZ  Monte Carlo generator. The same program was used to generate muon-pair events, while wide-angle Bhabha scattering, two-photon interactions, and final states were produced using the generators described in Refs. , , and , respectively. All these Monte Carlo data samples were subjected to the SLD detector simulation based on the GEANT  program and to the above event selection. The SLD trigger was also simulated in the Monte Carlo.