(total tow time was 72.6 hours). Information was obtained on 185 PIT-tagged fish. Evaluating the Separation-by-Code system at Lower Granite Dam in 1995 was useful in revealing how the computer program needed to be modified to add the necessary flexibility to make it possible for multiple researchers to use the system simultaneously. These changes will be completed in 1996. Separation-by-Code System: Diversion Gates In 1994, NMFS started to address the need to route fish in multiple directions and to construct fish diversion gates for pipes. NMFS developed two-way and three-way rotational gates and side-to-side gates. General descriptions of the two types of diversion gates and how they operate are presented in the report. Evaluations showed that the side-to-side design has several advantages over the rotational design: it can be operated with the pipe at any degree of fullness, it causes less elevation loss, its fabrication is less costly because it requires fewer custom parts, and it is more easily maintained. v Separation-by-Code System: An Evaluation Tool Once the basic Separation-by-Code System was working, NMFS recognized that the computer program and test facility located at NMFS Manchester Research Station could be used to evaluate modifications being considered for installation at PIT-tag facilities in the CRB. To determine what modifications would be acceptable, the following comparisons were evaluated during 1994: 1) performance of single-read firmware versus double-read firmware at a water velocity of 4 m/second; 2) reading and separation efficiencies based on two versus four coils; 3) separation efficiencies at water velocities of 3 versus 4 m/second; and 4) separation efficiencies for two distances between last coil and diversion gate. Tests were conducted with PIT-tagged sticks and coho salmon diverted by a slide gate. Reading efficiency (RE) was calculated by determining the percentage of tagged sticks or tagged fish read by at least one coil out of all possible PIT tags used in that trial. Separation efficiency (SE) for each trial was calculated using the theoretical and actual distributions of tagged sticks or fish within the two terminal holding areas based on which tags had been read. Thus, SE represented the percentage of correct actions for each trial. Results for stick and fish trials using the four-coil arrangement at 4 m/second demonstrated that the RE and SE performance for double-read firmware was equivalent to that of single-read firmware. In the stick trials for both firmwares, all sticks were read and only one stick was not diverted successfully. Although more fish than sticks were missed, there were still no significant differences in REs or SEs between single-read and double-read firmware. Furthermore, the double-read firmware did not produce a single erroneous tag code. Thus, to avoid potentially harmful erroneous tag codes, NMFS supports incorporating double-read firmware into the interrogation systems at the CRB dams. However, after NMFS finished its tests, Destron-Fearing produc...