Experiments were conducted to quantify the effects of entrapped air on water infiltration into a loamy sand. Transparent three‐dimensional (3‐D) and 2‐D columns were used for experiments carried out for two infiltration conditions: (1) when air was free to move ahead of the wetting front and leave the bottom of the column (air draining) and (2) when air was confined ahead of the wetting front and hence could escape only through the soil surface (air confining). The measurement setup was composed of a tension‐pressure infiltrometer, an air flowmeter, water manometers, and video‐picture cameras. We applied both positive and negative water pressures at the soil surface and measured the simultaneous changes in the rates of water inflow and air outflow, the air pressure ahead of the wetting front, and the dynamic behavior and advance of the wetting front. The air pressure ahead of the wetting front for the air‐confining condition was generally found to increase with time rather than reaching a constant level, as observed in other studies by other researchers. The air pressure fluctuated locally because of air escaping from the soil surface. On the basis of an analysis of the results we present two empirical equations to predict the maximum air pressure at which air begins to erupt from the soil surface and to predict the minimum air pressure at which air eruption stops. We found that the infiltration rate was always equal to, and controlled by, the rate of air outflow. The infiltration rate varied inversely with the air pressure ahead of the wetting front and with the ponding depth at the soil surface. The infiltration rate fluctuated with time rather than undergoing changes in a three‐stage process, as is often characterized in the literature. The volume of residual entrapped air in the air‐confining condition increased 7% on average, and the infiltration rate decreased threefold to tenfold as compared to the air‐draining condition. Finally, it was shown that the air‐confining infiltration flow is fingered and unstable, consistent with the predictions of an existing theory.