Patterned peatlands of cool-temperate and subarctic regions have alternating low ridges (strings) and shallow pools (flarks) elongated normal to the direction of slope and water flow. Previous investigators have sought unsuccessfully to find one process, either biological or physical, to explain the features. I compared two sites in the Upper Peninsula of Michigan, both on sandy outwash, with similar water chemistry. Shingleton Fen is small ( 3 ha), with surface slope 0.3%, and basal date of 1680 +- 80 B.P. Peat stratigraphy and pollen analysis show that paludification replaced the original forest with the present open fen, comprising sparsely vegetated flarks and steep-sided, abruptly-margined strings with a moss-dominated ground layer and scattered trees. Strings and flarks average 1-2 m wide. Creighton Marsh is a large ( 5000 ha) fen, surface slope 0.7%, and basal date 3230 +- 140 B.P. Originating as shrub-carr with abundant Sphagnum, it developed into the present sedge-dominated fen with shrubs on the strings, which slope gradually into the moderately-vegetated flarks. Strings and flarks average 3-4 m wide. Fires were frequent in the early history of this site, but the vegetation of both sites has been floristically unchanged for about 1000 years. Freeze/thaw cycles may maintain or accentuate the patterns, but cannot explain their formation or alignment. Coalescence of hollows by oxidative peat degradation to form elongate pools may be a secondary process, but my radiocarbon dates of flark sediments indicate no measurable degradation. The gentle slopes are insufficient to constrain such coalescence to linear development. Formation of flarks by coalescence of hollows requires flark width to be inversely proportional to slope; I found the opposite to be true. Careful resurvey of markers implanted in the peat shows that downslope peat movement does occur. The requirement for water flow, the magnitude of seepage force even on these gentle slopes, and the low bulk density of the surface peat suggest that this force could cause shallow splitting of the surface layer normal to the direction of flow. Such splitting would not disturb peat stratigraphy, and the linear pools thus formed could be maintained and enlarged by decomposition or freeze/thaw cycles.