Oceanic foraminiferal oxygen isotopic values provide proxies for Cenozoic temperature change of high temporal resolution, but combine records of both ice volume and temperature. Leaf margin and other analyses of fossil plants provide estimates of paleotemperature on land, but not at temporal resolution comparable with deep-sea cores. Paleosols now provide long term paleoclimatic records with resolution comparable with deep-sea cores. Mean annual paleoprecipitation can be inferred from paleosol depth to calcic (Bk) horizon, corrected for compaction and calibrated by transfer functions for soils. Paleoprecipitation and mean annual paleotemperature can be estimated from chemical composition of argillic (Bt) horizons. Paleosol records of paleoclimate for the past 45 Ma have now been obtained from Oregon, Montana and Nebraska-South Dakota. Each area shows such landmarks of deep-sea records as terminal Eocene warm-wet conditions followed by earliest Oligocene cold-dry conditions, and a middle Miocene (15 Ma) warm-wet spike. Milankovitch-scale paleoclimatic oscillations are clear in exceptional paleosol sequences: late Oligocene (28.7-23.6 Ma) on Longview Ranch, Oregon, and late Eocene (40.0-40.5 Ma) in Douglass Draw, early Oligocene (32.2-33.0 Ma) on Cook Ranch, and late Miocene (10.4-11.0 Ma) in Barton Gulch, Montana. Paleosols thus record global climate rather than local tectonic effects. Paleosol records do not show the large, long-term, stepwise decline seen in deep-sea oxygen isotopic records. Paleosol records of precipitation and temperature are most like deep-sea foraminiferal carbon-isotopic records in showing strong perturbations with ramps of slight but persistent long-term change. Teleconnection between marine and continental paleoclimatic records on time scales from tens of thousands to tens of millions of years is suggestive of a role for greenhouse gases (carbon dioxide, methane) and the terrestrial carbon cycle in Cenozoic global climate change.