A reference model for the electrical conductivity structure beneath the deep seafloor is proposed and justified using a variety of geophysical evidence. The model consists of relatively conductive sediment and crustal layers of 6.5 km extent overlying a resistive (≈10−5 S/m) subcrustal channel of 30 km thickness and terminated in a deeper conductive layer and half-space. Its seafloor-to-seafloor response to a horizontal electric dipole source is explored as a function of frequency and range, showing that, compared with the response for a half-space with the lowest conductivity in the reference model, significant enhancement of the field amplitude can occur at long ranges (>100 km) and low frequencies (<1 Hz). At the same time, marked attenuation relative to the half-space response is seen at higher frequencies. The field enhancement is due to trapping of electromagnetic energy in a leaky subcrustal waveguide, as demonstrated by computing the complex Poynting vector. The attenuation occurs in the relatively conductive sedimentary and crustal layers overlying the lithospheric waveguide when their electrical thickness exceeds a skin depth. The results indicate that attempts either to model controlled electromagnetic sources or to interpret controlled source data using half-space models for the Earth can be badly misleading. The practicality of lithospheric communications in the real Earth is also investigated. Using measured receiver noise figures and the reference model, the receiver bandwidth necessary to achieve a given signal-to-noise ratio as a function of range and frequency is estimated for a seafloor horizontal source of strength 105 A-m. The results indicate that significant (≈100 km) ranges can be achieved only around 1 Hz with a bandwidth of ≈1 Hz at a SNR of 10, yielding a very low data rate of <3.5 bits/s. Longer ranges and higher frequencies are precluded by attenuation in the sediment and crustal layers and because the conductivity in the resistive channel is too large.