The formation of many arctic wetlands is associated with the occurrence of polygon-patterned permafrost. Existing scenarios to describe and explain surface landforms in arctic wetlands (low-center and high-center polygons and polygon ponds) invoke competing hypotheses: a cyclic succession (the thaw-lake hypothesis) or a linear succession (terrestrialization). Both hypotheses infer the predictable development of polygon-patterned wetlands over millennia. However, very few studies have applied paleoecological techniques to reconstruct long-term succession in tundra wetlands and thereby test the validity of existing hypotheses. This paper uses the paleoecological record of diatoms to investigate long-term development of individual polygons in a High Arctic wetland. Two landform processes were examined: (1) the millennial-scale development of a polygon-pond, and (2) the transition from low-center to erosive high-center polygons. Diatom assemblages were quantified from habitats associated with contrasting landforms in the present-day landscape, and used as an analog to reconstruct past transitions between polygon types. On the basis of this evidence, the paleoecological record does not support either of the existing models describing the predictable succession of polygon landforms in an arctic wetland. Our results indicate a need for greater paleoecological understanding, in combination with in situ observations in present-day geomorphology, in order to identify patterns of polygon wetland development and elucidate the long-term drivers of these landform transitions.