In recent years, the study of rock glaciers has increased remarkably. Substantive progress has been made, particularly in understanding the formation of rock glaciers that have developed adjacent to existing or former valley or cirque glaciers, However, our understanding of valley-wall rock glaciers that are located at the base of talus slopes remains scant. Published work exhibits little consensus on the formation of valley-wall rock glaciers and several hypotheses remain under vigorous debate. The major objective of the research reported in this thesis has been to test the generality and feasibility of seven major models of valley-wall rock glacier formation using both empirical and theoretical evidence. The primary conclusion is that only one of these models, the segregation ice model, emerges as a general model of valley-wall rock glacier genesis. The model assumes that a thin layer or several thin layers of segregated ice are overlain by interstitially frozen sediments and an unfrozen mantle of coarse debris. A wide range of empirical and theoretical findings are shown to be consistent with the implications of the segregation ice model. Detailed observations on the morphology, sedimentology and distribution of active, inactive and relict valley-wall rock glaciers studied in Switzerland, northern Norway and Scotland provided a range of findings that support this model. Theoretical evidence was obtained by modelling a number of different density models that reflect different distribution of internal ice by applying a simple laminar flow equation to field measurements. Although only the segregation ice model appears to be valid at a general level, the possibility cannot be excluded of alternative modes of valley-wall rock glacier formation under particular circumstances. Snow avalanching, deformation of snowbank or matrix ice, and basal sliding under conditions of high hydrostatic pressure all constitute possible contributing mechanisms of formation and movement in particular cases.