Abstract Experiments on intermittent gas lift were conducted in a field scale test loop to measure: a) the effect of injection pressure, liquid column length, formation gas-liquid ratio, crude API and gas flow rate injected per cycle on the liquid fallback and b) the dynamic behavior of gas lift pilot valves. The test loop used consisted of an 8,000-foot deep well with surface storage capacity of 3,000 Br of fluids, gas-liquid separators, surface pumps able to handle up to 12,000 Br of fluid per day, natural gas supply of 3 MMSCFD at 1,500 psig, surface and downhole pressure and temperature sensors, data acquisition system and surface control equipment. The choice of liquid column length, gas injection pressure, size and type of pilot valves, point of injection depth and type of crude used were made based on a first set of limited tests performed at the same test site in 1996 and pressure and temperature surveys performed in more than 50 real wells in Lake Maracaibo in 1997. The survey procedure used for this work was such that the liquid fallback, along with an important number of other engineering parameters, could be obtained. Injection pressures from 450 to 900 psi, and liquid columns from 200 to 800 ft were studied in the test loop. Two different types of crude's API were used to measure the effect of fluid viscosity on the liquid fallback for the first time. Different liquid column gradients were obtained for a particular type of crude by changing the amount of free gas present in the liquid column. In this way, the effect of the formation gas present in the liquid column on the fallback was also studied for the first time. The points of gas injection were located at 4,839 and 2,516 feet. The dynamic behavior of both, spring loaded and nitrogen charged, gas lift pilot valves was studied using different port to bellow area ratios. As it was found in the first experimental work [1] conducted in 1996 for a limited set of conditions, the behavior of spring loaded and nitrogen charged gas lift pilot valves deviates from currently accepted theoretical behavior in opposite ways. The dynamic behavior of the gas lift pilot valves used is presented for a wide variety of operating conditions and a general analysis of the deviations found is discussed.