Amplitude preservation in seismic reflection data can be obtained by a relative true amplitude (RTA) processing technique in which the relative strength of reflection amplitudes is preserved vertically as well as horizontally, after compensating for amplitude distortion by near-surface effects and propagation effects. Quantitative analysis of relative true amplitudes of the Great Lakes International Multidisciplinary Program on Crustal Evolution seismic data is hampered by large uncertainties in estimates of the water bottom reflection coefficient and the vertical amplitude correction and by inadequate noise suppression. Processing techniques such as deconvolution, F-K filtering, and migration significantly change the overall shape of amplitude curves and hence calculation of reflection coefficients and average reflectance. Thus lithological interpretation of deep crustal seismic data based on the absolute value of estimated reflection strength alone is meaningless. The relative strength of individual events, however, is preserved on curves generated at different stages in the processing. We suggest that qualitative comparisons of relative strength, if used carefully, provide a meaningful measure of variations in reflectivity. Simple theoretical models indicate that peg-leg multiples rather than water bottom multiples are the most severe source of noise contamination. These multiples are extremely difficult to remove when the water bottom reflection coefficient is large (>0.6), a condition that exists beneath parts of Lake Superior and most of Lake Huron.