We conducted laboratory experiments using a setup consisting of a flask and a conduit, which reproduces cyclic eruptions and the time-predictability of natural geysers. We measured pressure and temperature in the flask and the eruptive mass of each eruption, varying three geometric parameters of the experimental system: (i) the flask volume, (ii) the inner diameter of the glass conduit and (iii) the initial water level in the conduit. During the heating stage of each eruption cycle, continuous pressure oscillations were observed in the flask. The pressure measurement and visual observation using normal-speed and high-speed cameras revealed that the pressure oscillations consisted of two components: (i) lower-frequency fluctuation related to the vertical displacements of the water column in the conduit and (ii) superimposing pulses caused by bubble nucleation in the flask. The dominant frequency of the pressure oscillations decreased systematically toward the next eruption. To explain the frequency of pressure oscillations, we present a mathematical model relating the pressure change in the flask to the vertical displacements of the water column in the conduit. The frequency calculated by our model well explained the observed frequency, which suggested that the oscillation frequency was controlled by the three geometric parameters and the effective bulk modulus of the fluid in the flask. The systematic decrease of dominant frequency toward the next eruption was explained by a decrease in the effective bulk modulus of the fluid in the flask due to an increase in gas volume fraction by heating. We also found that the erupted mass of an eruption controlled the initial frequency just after the eruption. The larger erupted mass causes a larger temperature drop in the flask, lowering the initial gas volume fraction to increase the initial oscillation frequency.