Since the late 1980s, electromagnetic (EM) sensors for determination of soil water content from within nonmetallic access tubes have been marketed as replacements for the neutron moisture meter (NMM); however, the accuracy, variability and physical significance of EM sensor field measurements have been questioned. We studied the accuracy and variability of four EM sensors and the NMM, compared with gravimetric measurements, in transects of 10 to 20 access tubes during three field seasons, using soil-specific calibrations. The three capacitance EM sensors produced water content readings for which SD values were up to an order of magnitude larger than those from the NMM. The EM sensor based on travel time (waveguide) principles produced SD values up to six times larger than those of the NMM or gravimetric sampling. The EM sensors would require from two to 72 times as many access tubes to obtain a mean profile water content to a given precision than would the NMM or gravimetric sampling, with more tubes required for drier conditions. The NMM exhibited spatial variation of similar magnitude and pattern as that of gravimetrically sampled profile water contents. The EM methods poorly reproduced the spatial and temporal behavior of NMM and gravimetric sampling and implied spatial variability of profile water content that was not evident in either the NMM or gravimetric data, even though EM sensing volumes were larger than the ∼75-cm3 volume of the gravimetric samples. We infer that EM sensors were influenced not only by the mean water content in the sampling volume but by the smaller scale structure of soil electrical properties.