A high-energy electron-impact spectroscopy (HEEIS) apparatus has been constructed for high-precision Compton-scattering experiments. Electron-Compton-scattering experiments are performed by crossing a beam of high energy, but nonrelativistic, electrons with a beam of atoms or molecules and measuring the energy-loss spectrum of the scattered electrons over a range of scattering angles. The improvements of design and technique, the method of data analysis, and the theory used to convert cross sections to Compton profiles are discussed fully. It was found that the energy-loss spectra taken over a range of scattering angles do not reduce by means of the binary-encounter approximation (impulse approximation) to Compton profiles in agreement with theory. This disagreement is most apparent in a shift of the experimental Compton peak---the Compton defect---from the peak predicted by the binary-encounter theory. The Compton defect has been studied in detail for momentum transfers from 1.5-12 a.u. for both He and ${\mathrm{H}}_{2}$. Defect measurements for ${\mathrm{D}}_{2}$, ${\mathrm{N}}_{2}$, and Ne have also been made and it was found that the ${\mathrm{N}}_{2}$ and Ne defects were opposite in direction from the He and ${\mathrm{H}}_{2}$ defects. The ${\mathrm{D}}_{2}$ defect was identical to that for ${\mathrm{H}}_{2}$. The electron Compton defect is discussed in relation to other recent defect measurements using x-ray and ($e, 2e$) techniques as well as recent theoretical results. An evaluation of the theory used to convert cross sections to Compton profiles is presented and, on the basis of the defect measurements, it is suggested that, even when the binary-encounter conditions have been attained at large momentum transfers, the binary-encounter theory breaks down in the high accuracy (1%) limit. An explanation for this breakdown is given and recent theories, which at least qualitatively account for the Compton defect, are discussed.