Previous investigations of membrane penetration mainly focused on uniform sandy soils, and few studied gravelly soils. A common problem in most previous investigations was that unverified assumptions were made to evaluate or mitigate the membrane penetration of a cylindrical specimen in triaxial test. In this study, a simple measurement of the membrane penetration of gravelly soils was proposed, based on small-amplitude cyclic confining pressure Δσ3 loading, in which the cyclic membrane penetration amplitude ΔVm is recoverable, and the cyclic soil skeleton volume deformation amplitude ΔVk gradually decreases with increasing number of cycles and becomes elastic. The accuracy in the proposed method is related to the accuracy in measuring ΔVk. Because of the natural limitations in measurements of the radial strain of a cylindrical specimen, ΔVk is calculated based on the isotropic assumption. The proposed method provides sufficiently accurate ΔVm values for gravelly soils because of the small portion of ΔVk in the total cyclic volume change amplitude ΔVc (=ΔVm+ΔVk), and the accuracy increases as the coarse fraction increases. Furthermore, the method can be easily implemented for individual specimens to remove inconsistences in specimen preparation. The test results indicate that ΔVm values corresponding to the same Δσ3 exponentially decrease with the confining pressure σ3. Membrane penetration increases with increases in the void ratio and coarse fraction, but the trend is becoming weak with decreasing coarse fraction and void ratio, respectively. It also demonstrated that there exists a clear difference between the results of Nicholson’s empirical formula and the test results.