Formation of ferrimagnets in well-drained, buffered, unpolluted soils appears to be related to climate, and especially rainfall. If robust, this magnetism/rainfall couple can be used to estimate past rainfall from buried soils, particularly the multiple soils of the Quaternary loess/soil sequences of Central Asia. However, dispute exists regarding the role of climate vs. dust flux for the magnetic properties of modern loessic soils. Here, we examine the mineralogical basis of the magnetism/rainfall link for a climate transect across the loess-mantled Russian steppe, where, critically, dust accumulation is minimal at the present day. Magnetic and independent mineralogical analyses identify in situ formation of ferrimagnets in these grassland soils; increased ferrimagnetic concentrations are associated with higher annual rainfall. XRD and electron microscopy show the soil-formed ferrimagnets are ultrafinegrained ( 6 V50 nm) and pure. Ferrimagnetic contributions to Mo « ssbauer spectra range from 17% in the parent loess to 42% for a subsoil sample from the highest rainfall area. Total iron content varies little but the systematic magnetic increases are accompanied by decreased Fe 2þ content, reflecting increased silicate weathering. For this region, parent materials are loessial deposits, topography is rolling to flat and duration of soil formation effectively constant. The variations in soil magnetic properties thus predominantly reflect climate (and its co-variant, organic activity)ŝ tatistical analysis identifies strongest relationships between rainfall and magnetic susceptibility and anhysteretic remanence. This magnetic response correlates with that of the modern soils across the Chinese Loess Plateau. Such correlation suggests that the rainfall component of the climate system, not dust flux, is a key influence on soil magnetic properties in both these regions. ß