Water content sensors used in access tubes for sensing profile water content were compared for two seasons in Texas and one season in California on clayey soils. Sensors included the neutron moisture meter (NMM) and four electromagnetic (EM) devices: three different capacitance sensors, and a device based on a short-rise-time pulse traveling along a transmission line. Measurements were compared with direct soil sampling. Transects of ten, twenty, and twelve access tubes or sampling sites were used in the three seasons, 2003, 2004, and 2005, respectively. Variability varied greatly among the devices, with standard deviations of profile water contents <0.7 cm/m for the NMM and direct sampling methods, and ranging from 0.8 to 12 cm/m for EM methods. Variability was different in different soil horizons, and it increased as the soils dried, though less so for the NMM and direct sampling methods. Larger variabilities for the four EM devices were due to their relatively small measurement volumes, which were typically smaller than a representative elemental volume (REV) for soil water content, and the relatively large variability of soil water content on scales less than the REV. Results indicate that unreasonably large numbers of access tubes are required for the EM devices to be used to determine differences in profile water content to reasonable precision at probility levels commonly desired in research; calling into question their suitability both for plot research and for studies of spatial variation of soil water content.