From the perspective of Phanerozoic time, coal balls are rare, apparently limited to a 24 m.y. interval (323-299 Ma) in the Pennsylvanian and earliest Permian. Yet within this interval, coal balls occur in many coals. Approximately 82 transgressive-regressive sedimentary cycles have been described for the Midcontinent, Illinois and Appalachian basins of North America during the mid-to-late Pennsylvanian. One third (27/82) have coal balls, including 57% of major cycles, 36% of intermediate cycles and 16% of minor cycles. Coal-ball occurrence in the Donets Basin is similar: over an interval of about 4 m.y. (similar to 315-311 Ma, latest Bashkirian to mid-Moscovian), 39% (11/28) of transgressive-regressive cycles have coal balls. As North American paleoclimate became drier, tree ferns replaced lycopsids as the dominant plant in peat swamps at the Desmoinesian/Missourian boundary, and coal-ball occurrence declined. Overall, 47% of cycles (19/40) with lycopsid or cordaitean dominance have coal balls, whereas 19% (8/42) of cycles with tree fern dominance have coal balls (p < .004). While 32% (6/19) of minor cycles with lycopsid or cordaitean dominance have coal balls, no minor cycle with tree fern dominance has coal balls. This pattern may reflect the relative abundance of coal in the Late Pennsylvanian, with drier paleotropical climates in the Missourian-Virgilian leading to less paleotropical coal, fewer mines and a lower probability of discovering Missourian and Virgilian coal balls. However, average cycle duration decreased from 150 kyr in the Desmoinesian to similar to 100 kyr in the Missourian and Virgilian, and faster rates of relative sea-level rise could have affected coal-ball abundance, particularly if coal balls formed in marine swamps, with more rapid transgression leading to thinner coals, or to freshwater coals directly overlain by marine sediments. Rygel et al. (2008) suggested that Missourian-Virgilian cycles recorded more erosional relief than Desmoinesian cycles, consistent with increased amplitude and rates of sea-level rise during each glacial-eustatic cycle. Changes in Missourian-Virgilian cycles may have been augmented by stratigraphic attenuation as sediment filled the antecedent topography.