The predictions of version 2.1E of the DOE-2 program for building energy analysis have been compared with measurements in the Pala test houses near San Diego, CA. This work was part of the California Institute for Energy Efficiency 'Alternatives to Compressor Cooling in California Transition Zones' project in which DOE-2 was used for parametric analysis of cooling strategies that reduce peak electrical demand. To establish the validity of DOE-2 for this kind of analysis the program was compared with room air temperature measurements in a 'low-mass' house with conventional insulated stud wall construction and a 'high-mass' house with insulated concrete walls. To test different aspects of the DOE-2 calculation, four different unoccupied, unconditioned thermal configurations of these houses were considered: unshaded windows, shaded windows, white exterior surfaces, and forced night ventilation. In all cases DOE-2 agreed well with the air temperature measurements, with a mean deviation between simulation and measurement ranging from 0.2 to 1.0 K depending on configuration and type of house. Comparisons with inside surface temperature measurements also showed good agreement. Agreement between predictions and measurements improved when a more accurate calculation of foundation heat transfer was used, the ground surface temperature was calculated, and the normal 7-day 'warm-up' period in DOE-2 was extended to 11 days for the high-mass house.