A three-dimensional array of 20 current meters, temperature sensors, and vertical temperature gradient sensors was successfully deployed for 40 days in late 1973 in the main thermocline over the Hatteras Abyssal Plain southeast of Bermuda. Sensor spacings in the main array were 1.4--1600 m in the horizontal, 2.1--1447 m in the vertical. The minimum sampling interval was 225 s. The ultimate purpose of the experiment was to estimate a vector wave number-frequency spectrum of internal waves without the usual assumptions of simple modal structure, horizontal isotropy, and linearity. The purpose of this paper is to describe some of the early results. Autospectra from the array normalize quite well in depth according to the WKBJ 'high-mode' solutions. Spectra of vertical displacements show a significant contribution from the internal semidiurnal tide. Samples of 1760 cross spectra calculated (based on a 40-day averaging interval) suggest horizontal isotropy, vertical homogeneity, and a possible degradation of current coherences because of fine structure in the velocity profile. Coherence of vertical displacements (i.e., temperaure fluctuations) for measurements separated horizontally decays with increasing separation according to f1/2X=330 m⋅cph, where f1/2(cph) is the frequency at which the coherence falls to one half and X (m) is the horizontal separation. This empirical rule is based on 1600 m>X>140 m; for smaller X1/2 exceeds the local buoyancy frequency. Autospectra and cross spectra of vertical displacements sometimes show peaks at frequencies just less than the local buoyancy frequency; current spectra do not show such peaks. Inverse modeling of the internal wave field is in progress; expected results are a vector wave number-frequency spectrum and a description in parameter space that hopefully will permit future experiments to be less elaborate.