The Pennsylvanian (323.2–298.9 million years ago, Ma) is best-known for its peat-forming swamps. This unique, no-analog biome repeatedly dominated equatorial Euramerica during the glacial intervals of the glacial-interglacial cycles of the Late Paleozoic Ice Age (LPIA, ∼335 to 260 Ma). During this period, Earth’s landmasses had just recently aggregated into the supercontinent Pangea, and Euramerica was a mostly flat craton spread across the equator. Glacial intervals saw massive amounts of ice covering the south pole. Interglacials saw increases in eustatic sea level coincident with melting ice. In Euramerica, glacials meant everwet rather than monsoonal conditions in the intertropics, which allowed peat swamp vegetation to expand beyond wet basins and dominate the flat landscape. By contrast, interglacials came with more seasonal and therefore drier overall conditions. These changes in sea level and precipitation led to shrinkage of the biome, and survival of communities in refugia. Each subsequent glacial interval saw re-expansion of the swamps. Similarities and changes in swamp vegetation can be well-tracked from one glacial to the next, due to the exceptional paleobotanical record of this period. Despite the wet conditions that gave rise to Pennsylvanian peat swamps, this biome is known to have regularly experienced wildfires. Although these paleowildfires are well documented in the sedimentary record as fossilized charcoal, our knowledge of their ecological role and evolutionary impacts is limited. In this dissertation, I utilize one of the largest paleobotanical datasets in existence to investigate patterns in peat swamp plant diversity and community composition in response to changes in climate, and to investigate the role that paleowildfires played in this unique biome.In chapter one, I examine the research history and structure of the Pennsylvanian plant fossil record. Reconstructions of Pennsylvanian peat-forming communities are, for a large part, based on quantitative microscopic analysis of acetate peels of carbonate nodules—so-called “coal balls”—which three-dimensionally preserve plant litter from these swamps. Extensive coal ball analysis was carried out by the former Phillips lab of the University of Illinois, Urbana-Champaign, primarily during the 1970s and 80s. The origin of the Phillips Coal Ball Collection (PCBC) dataset is closely tied to the early days of scientific computing. Until recently, most of the collection data only existed in formats rarely used in the present day, specifically, continuous-form IBM 360 mainframe computer printouts and first-generation 8-inch floppy disks. In this chapter, I outline the process of “re-digitizing” these data, including the complex task of reading the floppy disks, attempts at automated transcription of the paper datasheets with Optical Character Recognition, and the ultimate setup for high-throughput photography and transcription. I describe the integration of undergraduate and community scientist researchers, outline recommended practices and potential pitfalls, and highlight the need for coordinated efforts to preserve scientific data from the early computational era in the near future. This work yielded an “occurrence-type” dataset with over 450,000 plant fossil observations. Spatially, this dataset spans modern-day North America, Europe, North Africa, and East Asia. Temporally, it covers the 24-million-year Pennsylvanian period and the latest Permian. My re-digitization work with these data paved the way for all of the analyses carried out in chapters two and three.Chapter two focuses on diversity dynamics in the Pennsylvanian, with a focus on the interval around the Middle-Late Pennsylvanian transition. This transition is well known for the Carboniferous Rainforest Collapse (CRC), a major floral turnover evidenced by and best known for the loss of all but one genus of arborescent lycopods. However, my analyses of the PCBC data indicate the CRC should not be considered a strict boundary phenomenon. I recover increased per-capita extinction rates and decreased per-capita origination rates for plant genera from all major taxonomic groups across multiple glacial-interglacial cycles starting before the Middle-Late Pennsylvanian boundary. Changes in ecological structuring—as indicated by dominance-diversity curves, evenness and ecological dissimilarity values, and Detrended Correspondence Analyses based on genus abundance at individual coal ball peels—also begin prior to the boundary. Furthermore, forward survivorship curves additionally hint at a more protracted decline in genus diversity. Transition to Late Pennsylvanian floral regimes beginning in the late Middle Pennsylvanian has been previously commented upon, and it has been suggested that the Late Pennsylvanian aridification started in the late Middle Pennsylvanian. If increased aridification and concomitant shifts in CO2 and O2 levels were drivers of the CRC, an earlier start would follow logically. Although changes in diversity dynamics predate the Middle-Late Pennsylvanian transition, the most dramatic decrease in lycopod domination—and with that a shortening of the ecological gradient and physical structure of the plant communities—occurs at the boundary. Lastly, in this chapter I discuss a potential ecosystem engineering effect by which arborescent lycopods may have helped create their own suitable habitat. I propose a hypothesis: that a breakdown of this effect, via the crossing of some ecological threshold, may have been a driver of the CRC.For my final chapter, I further investigate the ecology of this time period, focusing on paleofire. In particular, I use the PCBC dataset to explore the influence of wildfires on peat swamp communities. The Pennsylvanian has long been of interest to the niche field of paleofire ecology, due to its abundant charcoal record, as well the fact that the numerous paleoatmospheric oxygen reconstructions recover it as a “hyperoxic” interval. Hyperoxic (or “superambient”) oxygen levels refer to oxygen levels above the Present Atmospheric Level of 21%. Experimental work has shown that fire behaves very differently under such conditions, with potentially increased burn probability, burn duration, and spread rate. I set out to apply methods from neontological wildfire ecology to the PCBC dataset. Specifically, I investigated the distribution and relative abundance of paleofuel sources to determine which tissues, organs, and taxa contributed to paleowildfires, and to study how those fires behaved. I find that between-coal-seam charcoal variability is high, and that any long-term temporal shifts—if present—are therefore obfuscated. However, some patterns in charcoal abundance clearly hold across space and time. Wood, cortex, and periderm tissues are consistently over-represented as charcoal, when compared to other tissues and organs, while reproductive parts and foliage are consistently under-represented. Pteridosperms, which include the proposedly fire-prone medullosans, are found to be no more frequently charcoalified than any of the other major taxonomic groups present. The tree ferns—who dominate peat floras after the CRC—are an exception: their charred remains do not contain substantially fewer reproductive structures (which are leaf-born) and foliage seems overrepresented in tree fern charcoal rather wood, cortex, and periderm tissues, as in the other major plant groups. I use these results to try to make inferences about the types of fires present in these swamps, with a particular goal of evaluating the presence/absence and frequency of crown fires—intense fires that burn through the canopies of plant communities. Finally, I explore how complex interactions between hyperoxic combustion, no-analog plant communities, and under-explored taphonomic factors complicate attempts to reconstruct Pennsylvanian fire types and regimes.