Ocean Drilling Program Site 983 (Gardar Drift) is receiving volcanic and glacial sediment from Iceland, the Iceland-Faeroe Ridge and Europe, and has high sedimentation rates averaging over 15 cm/Ky. The high-resolution records from these sediments show millennial cycles in the magnetic susceptibility and natural gamma data. The climatic significance of these millennial cycles is of particular interest since the mechanism that causes these cycles is not fully understood. The first step toward understanding what sedimentological parameters are controlling the magnetic susceptibility and natural gamma signals is to find out how the signals correlate to individual sediment parameters such as mineralogy and grain size. Mineralogy data are useful for understanding source areas, whereas grain size data provide information about sediment transport. Unfortunately, determining mineralogy and grain size data is time and labor intensive. The susceptibility and natural gamma signals are a function of the grain size and/or mineralogy of the sediment and can be obtained rapidly at high resolution (<10 cm). If mineralogy and grain size can be predicted accurately from natural gamma and susceptibility variations, then high resolution time series of mineralogy and grain size data can be obtained by proxy. This study relates continuous, nonintrusively-assessed sediment proxies (natural gamma and susceptibility) to laboratory measurements of sedimentological parameters (calcium carbonate concentration, grain size and mineralogy). Sediment analyses were performed on two sequences, approximately 220-293 Ka ( approximately isotope stages 7-8), and approximately 845-925 Ka ( approximately isotope stages 20-24). These laboratory analyses include grain size, calcium carbonate content, and mineralogy (smectite, illite, chlorite, kaolinite, quartz and plagioclase) of the fine silt (20-2mu m) and clay (<2mu m) fractions. Both intervals show lithologic variations related to glacial versus interglacial climate regimes. Forward multivariate regression was used to find the correlation between mineralogy, calcium carbonate, and grain size variations and natural gamma and susceptibility variations. Only regressions that were significant at the 0.9 level and had r-squared values >0.3 were used to predict the above sedimentological parameters from the natural gamma and susceptibility. The models were able to successfully predict only a fraction of the 22 total sedimentological variables. For those variables that were predicted, however, the models did a fairly good job of predicting low amplitude variations, but poorly predicted high amplitude variations. R-mode factor analysis was also performed in an effort to reduce the variables to several new sedimentological factors, which could then be compared to susceptibility and natural gamma. Since no small group of factors explained a statistically significant proportion of the data set's variance, factor analysis could not be used to reduce the data in this study.