Geological and geophysical data digitized by Fabbri (1981)are augmented by eleven lakesediment geochemical variables in a quantitative study of the Kasmere Lake—Whiskey Jack Lake area, NW Manitoba. Each of 33 geological map units (Archean and Aphebian igneous and metamorphic rocks)have a fairly distinct multivariate geochemical and geophysical signal, as shown by discriminant analysis which is able to “recover” the geological map from the geophysical and geochemical data. Autocorrelation analysis of the geochemical variables, after removing a quadratic trend, indicate that copper, nickel, uranium, and cobait have a zone of influence from 15 to 30 km or more, with a marked anisotropy parallel to geological, geophysical, and structural trends which run NE-SW, parallel to the Wollaston fold belt. The observed uranium autocorrelation coefficients are modeled as a two-dimensional exponential function with elliptical contours. A trend-signal-noise model is applied to the lake-sediment uranium using kriging on residuals from the quadratic trend. A map comparing the resulting lake-sediment uranium signal with radiometric uranium/thorium patterns indicates considerable overlap and a generally good prediction of known uranium occurrences. In areas with a strong lake-sediment uranium signal, Aphebian metasediments have a somewhat different multivariate geochemical character from Archean and Hudsonian igneous and metamorphic rocks, suggesting a method of removing areas from the signal underlain by granites that may be economically uninteresting. Logistic and stepwise regression experiments based on a control area which includes most of the presently reported occurrences are used to predict areas favorable for uranium exploration.