Abstract-Through ecological classification, researchers both (1) identify and (2) describe naturally-occurring, ecologically-distinct, spatial units from a holistic perspective. An ecological river classification involves the identification of structurally homogeneous spatial units which emerge along the channel network as a result of catchment processes interacting with local physiographic features. Our observations of Michigan rivers suggest that the natural ecological unit, as defined by the spatial scales of riverine physical and biological processes, is most closely approximated by the physical channel unit termed the valley segment. Valley segments are generally quite large, and characterized by relative homogeneity in hydrologic, limnologic, channel morphology, and riparian dynamics. Valley segment characteristics often change sharply at stream junctions, slope breaks, and boundaries of local landforms. We followed several steps in developing an ecological classification for the rivers of lower Michigan. Step 1 - We first selected catchment size, hydrology, water chemistry, water temperature, valley character, channel character, and fish assemblages as fundamental attributes to describe ecological character of river valley segments. Steps 2-3 - Two experienced aquatic ecologists worked together, interpreting map information on catchment and valley characteristics from a GIS, using their combined knowledge of ecological processes and interactions. We initially examined several key maps to become familiar with the general landscape patterns of a particular catchment; and to then identify initial valley segment units as defined by catchment and valley characteristics, and fish assemblages. Boundary definition required the integration of terrain features observed on several thematic maps (e.g., major stream network junctions, slope breaks, boundaries of major physiographic units or land cover units; or changes in stream sinuousity and meander wavelength patterns, riparian wetlands, or valley shape), combined with knowledge of fish distributions. We next developed categorizations for each component attribute and assigned category values for attributes to each segment unit. Assignments were based on map-interpretation rules drawn from modeling, survey data, and field experiences. Step 4 - our results were stored as a map and a table in ArcView 3.0 format. In all, we partitioned and classified the 19 largest river systems in lower Michigan. Summaries of the attributes assigned to over 270 river valley segments (covering mainstems and major tributaries) provided an initial description of the river resources of lower Michigan. Managers of lower Michigan rivers will be able to develop many of their thoughts and activities within this framework of ecological units. Development of this system is intended to be ongoing; with the extension of coverage to upper Michigan, the continued validation of attribute codings, and the addition of new attributes.