On January 25, 1997, at 1038 hours Pacific standard time, an amateur built, experimental Crane Thorp T-18, N40918, lost engine power during the takeoff initial climb from runway 26R at Brown Field, San Diego, California, and impacted a commercial building during the ensuing off-airport forced landing. The aircraft was destroyed. The commercial pilot/flight instructor and private pilot dual student were fatally injured. Visual meteorological conditions prevailed. The local area instructional flight originated at Brown Field on the morning of the accident and had completed two touch-and-go landings prior to the accident. According to acquaintances of the aircraft owner (the dual student) who were familiar with the accident aircraft, the engine had a history of power loss dating back several months prior to the accident flight. In a telephone conversation with the Safety Board, a pilot who had flown the accident aircraft between June and September, 1996, reported that he had experienced complete engine power loss on two occasions while flying the accident aircraft, and stopped flying it for that reason. The first instance occurred while he was performing calibration tests on a GPS system installed in the aircraft. While conducting maximum power, maximum rpm, level flight tests at 8,000 feet msl, with the fuel tank approximately 1/2 full, the engine suddenly quit. He stated that the "Engine obviously had fuel starvation at full power -- [it] had to do with fuel pressure." In response to the power loss he "Retarded power to 1/2 throttle and the engine restarted - it ran OK at 2100 RPM." After returning to the airport, he told the owner about the power loss and that the aircraft had a fuel flow problem. According to this pilot, the aircraft owner subsequently experienced "carburetor problems" during climb phases of flight operations and the owner had the aircraft examined by an A&P mechanic who found a bad carburetor jet which was replaced; however, this failed to correct the power loss condition. The same pilot reported that the second instance of engine power loss he experienced occurred after 1 hour of flight, while performing a touch-and-go landing. During the full power takeoff segment, at about 100 feet agl, the engine quit. The power loss was corrected by retarding the throttle to 2,100 rpm. The engine would constantly misfire at power settings over 2,100 rpm. The fuel tank was approximately 1/2 full of fuel at the time of the power loss. He indicated to the aircraft owner that he believed that the fuel flow problem might be caused by a fuel line obstruction or vent problem. He said that the owner had attempted to correct the problem with a number of potential fixes, including: installing an auxiliary fuel boost pump with a "manual one-way bypass valve," and complying with a carburetor airworthiness directive, which had not been complied with previously. None of these corrective actions fixed the power loss problem. According to this pilot, the aircraft owner also thought the problem might be caused by a fuel line obstruction or fuel tank venting problem. The fuel tank was located in the fuselage behind the firewall and forward of the instrument panel. The owner subsequently vented the fuel tank from the top of the fuel cell with a tube which was routed behind the firewall, through the bottom of the fuselage, terminating with a screen over the end of the tube. The vent tube was beveled, facing forward into the airstream to pressurize the fuel tank and avoid vapor lock. In a telephone conversation, a friend (a certificated pilot) of the aircraft owner, indicated that he was familiar with the maintenance and operation of the accident aircraft. He also stated that he was aware of a "minor" problem with the accident aircraft during the 2 weeks prior to the accident flight. Specifically, the problem manifested itself as less fuel flow than the required which caused the engine to misfire. The problem resulted in a reduction of engine rpm by 200 or 300 rpm. The engine would not exceed 2,300 rpm when the fuel flow problem was occurring. This problem could be corrected by the use of an auxiliary fuel pump which would be used as needed. The same friend indicated that in the process of working on the fuel flow problem, the owner had examined numerous components and solutions. Specifically, the owner checked the fuel flow from the fuel pump and found it to flow at a rate of 35 gallons per hour. The friend also indicated that the owner had the carburetor examined by "experts somewhere." Aircraft records indicate that the carburetor was worked on by a licensed repair facility and that an outstanding airworthiness directive for the carburetor was complied with at that time. The friend said that the owner told him that he thought the problem was a "backflow" problem with the fuel tank. The owner then vented the tank with a tube from either the fuel cap or top of the tank, which was looped down, in back of the firewall, through the bottom of the aircraft, and pointed into the airstream to force air into the tube and into the fuel tank. The friend said that the owner had been receiving dual flight instruction from the flight instructor to increase his proficiency in the higher performance, tail wheel, Thorp T-18. The friend believed that the accident flight was an instructional flight and that was why the aircraft was making touch-and-go landings. According to the engine logbook, the engine installed on the accident aircraft was originally installed on a Cessna 172 aircraft in 1977. The engine was removed from the Cessna on March 1, 1985, after accumulating 1,518 total hours. The location and disposition of the engine is unknown until January 14, 1995, when the owner of the accident aircraft noted in the engine logbook, "Started engine 1st time since installation with all systems operational." According to the engine logbook, at the time of installation on the accident aircraft, the engine had accumulated 1,518 total hours since factory new. When the accident occurred, the engine had accumulated approximately 1,550 total hours since factory new. Subsequent postcrash engine examination was performed by a Textron Lycoming investigator, who was a party to the investigation. He reported that the fuel system consisted of a vented fuel cell, which flowed fuel to the carburetor either by gravity or through a pilot activated electric fuel pump. A manual fuel shutoff valve was located at the fuel tank outlet and was accessible from the cockpit. Once the fuel line entered the engine compartment, it reached a "T" fitting which allowed the fuel to flow in two directions; one direction to the electric fuel pump after passing through a gascolator, and the other direction directly to the carburetor after passing through a check valve. The gascolator was only utilized if fuel was drawn through the electric fuel pump. A schematic diagram of the fuel system, prepared by the Textron Lycoming investigator, is attached. When the "check" valve was disassembled, it was determined that the valve contained no internal parts, or evidence of any part(s) which might have been destroyed by the accident or postcrash fire. According to the Lycoming representative: "The fuel system was configured in such a way . . . that when fuel was delivered by gravity during normal operation, the gaskolator [sic] was by-passed, thus allowing unfiltered fuel to flow to the carburetor (there was no filter in the carburetor). Fuel would only pass through the gaskolator [sic] when the electric fuel pump was on. In addition, the absence of a functioning check valve could possibly have an adverse effect on fuel delivery to the carburetor when the electric pump was operating."