Many theories, old and new, of landscape and earth-surface system development involve concepts of self-organization. There are at least eleven distinct definitions of self-organization in the literature that are relevant to landscapes. Some have profoundly different implications with respect to the nature and trajectories of landscape evolution and earth-surface system behavior, including whether evolution is convergent or divergent, whether entropy or energy dissipation is maximized or minimized, the role of chaos, and the mechanisms by which self-organized patterns are generated. Despite these differences, most self-organization concepts can be broadly aggregated into two categories: those concerned with the evolution of order and regularity in the aggregate or ensemble properties of the landscape, and those concerned with the differentiation of landscapes into more diverse spatial units. This paper presents a theory of spatially divergent self-organization related to the latter, showing that autogenic differentiation is directly linked to dynamical instability and chaos. The determination of the self-organization properties of a landscape should be a starting point rather than a goal of geographic explanation. The extent to which field-testable hypotheses are generated, or explanations provided based on process mechanics or landscape history, will ultimately determine the utility of self-organization concepts and methods in physical geography.