Experimental studies were performed on a pulsating heat pipe (PHP), consisting of a heating section, an adiabatic section, and a condensation section incorporating a heat sink. The capillary tube used in this study has an inside diameter of 1.18 mm and a wall thickness of 0.41 mm. The experiments were conducted under the condition of pure natural convection, for heating powers from 5 to 60 W, fill ratios from 60% to 90%. Three working fluids-FC-72, ethanol, and deionized water--were used. The thermal oscillation of the thin wall surface was recorded by a high-speed data acquisition system. Such thermal oscillation waves are random for some run cases due to the randomly distributed vapor plug and liquid slugs inside the PHP. The thermal oscillation amplitude is much smaller for FC-72, due to its lower surface tension, than for ethanol and water, while the oscillation cycle period for FC-72 is shorter than for the other two fluids, indicating the faster oscillation movement in the channels, possibly due to the lower latent heat of evaporation for FC-72. The unlooped PHP is not helpful for the fluid circulation and the PHP does not work. For the looped PHP, there is a minimum heating power that initiates the PHP working. Such minimum heating power is strongly dependent on the working fluid, and is considerably smaller for FC-72 compared with water. The optimal filling ratio is around 70% for all three working fluids. The looped PHP with water as working fluid provides better overall thermal performance once the heating power is greater than a minimum value. However, FC-72 is suggested to be used for low-heat-flux situations, due to its lower minimum heating power.