Dataset CreatorsJeremy Mennis, Temple UniversityContributorsLuke MathersMadison VoorheesThis geospatial dataset represents the city/town and county locations from which people escaped enslavement on the Philadelphia, Pennsylvania branch of the Underground Railroad between 1853 and 1861. These location data were developed by leveraging the valuable prior work of a series of other scholars of the Underground Railroad. Narrative descriptions of people escaping slavery along the eastern seaboard of the U. S. to Philadelphia were initially recorded by Black abolitionist William Still, who served as chairman of the Philadelphia Vigilance Committee of the Pennsylvania Anti-Slavery Society in the 1850s. Still received nearly 1, 000 people escaping slavery who arrived in Philadelphia traveling along Underground Railroad routes, primarily from the states immediately to the south of Pennsylvania, including Delaware, Virginia, and Maryland, as well as from Washington, D. C. , though a handful of escapees also came from states further south and west. Upon receiving these freedom seekers, Still wrote down brief narrative details about each person and their journey to freedom before assisting in arranging lodging and transportation for them, often to points further north in the United States and in Canada. Details on Still’s life and his important contributions to the Underground Railroad, the abolition of slavery, and African American economic advancement can be found in several recent books and articles. 1These narratives of individuals’ escapes from enslavement were recorded in two volumes written by Still, his 1872 published book The Underground Rail Road 2 and in “Journal C, ” an internal and unpublished document of the Philadelphia Vigilance Committee. 3 Still’s book is widely available in print and digitally. 4 Journal C is presently made publicly available by the Historical Society of Pennsylvania as a digital file. 5Scholars James McGowan and William C. Kashatus analyzed these escape narratives as described in Still’s published book and in Journal C to develop a table of people escaping enslavement, including several tabulated characteristics of each escapee, such as their name; alias; age; sex; month and year of arrival in Philadelphia; city, county, and state of the origin of their escape; and mode of transportation of the escape (e. g. by foot, horse and carriage, by steamship, etc. ). Encoding of certain characteristics, including the location of the origin of the escape, were supported by additional research on contemporaneous newspaper advertisements offering rewards for the return of people escaping enslavement. 6 This table was ultimately published as an extensive Appendix in Kashatus’s 2021 biography of Still. 7This Appendix was then transformed into a digital file in table format and published as a publicly available dataset by scholar Nick Sacco, who also made minor corrections for spellings of the names of the escapees and enslavers as well as the places of escape origin. This digital dataset was released publicly via Sacco’s website8 and was later published and disseminated by the Journal of Slavery and Data Preservation 9 in the public data repository Harvard Dataverse. 10The dataset presented here extends the important work of this prior scholarship by geocoding the place name data encoded in Sacco’s dataset to yield a geospatial (i. e. mapped) dataset of the specific cities or towns, as well as counties, from which each person escaping enslavement departed prior to arriving in to Still’s residence in Philadelphia. Information on the location origins was not available in Still’s original narrative entries for all escapees, so these data represent a subset of all the individuals escaping enslavement described in Still’s records. As identification of the county of origin was more prevalent than the specific city or town of origin, the present dataset provides separate geospatial data layers for the city/town location (in cases where such information was available) and the county location of origin for each escapee. The complete dataset consists of four separate geospatial data layers of point locations, each in the shapefile format, 11 a well-known and widely used format for geospatial data sharing: EscapeesbyCityₙ644. zipa. This shapefile represents the city or town of origin for each of the 644 escapees for whom such city or town origin location information could reliably be derived. Each escapee is represented as a point location. Citiesₙ92. zipa. This shapefile represents the 92 cities and towns from which the 644 escapees escaped. Each city/town is represented as a point location which also includes data on the number of escapees from each city and town. EscapeesbyCountyₙ776. zipa. This shapefile represents the county of origin for each of the 776 escapees for whom such county origin location information could reliably be derived. Each escapee is represented as a point location based on the county geometric centroid. Countiesₙ55. zipa. This shapefile represents the 55 counties from which the 644 escapees escaped. Each county is represented as a point location which also includes data on the number of escapees from each county. As with Sacco’s 2024 dataset publication, the present contribution does not necessarily aim to present new or previously unknown data, but rather to transform an existing dataset into a format that provides new opportunities for analysis and discovery regarding the operation of the Underground Railroad along the eastern seaboard of the U. S. The geospatial dataset presented here can be placed into a geographic information systems (GIS) software package to support geographic analyses regarding the spatial distribution of the places of escape origin among those escaping slavery on the Underground Railroad. It can also support analyses of the geographic distributions of other related characteristics of escapees and of escapes as represented in Still’s original records and in the datasets presented by Kashatus (2021) and Sacco (2024), such as whether and how demographic characteristics (e. g. age and sex), may have varied across different cities and counties. The present dataset supports research on how escapees from different regions may have utilized different modes of transportation and how the number of escapes from different places may have varied over the time period (1853-1861) of Still’s records. Further, GIS supports the integration of the geospatial data of escape origins presented here with other historical geospatial data, 12 such as transportation nodes and corridors (e. g. roads, railroads, river, canals, ferry landings, bridges), land cover (e. g. forest, agricultural fields, wetlands), the locations of settlements and key built infrastructure (e. g. cities, towns, court houses, post offices), and other important historical features of particular relevance to the Underground Railroad, such as the locations of African American churches, Quaker meeting houses, free Black communities, and known Underground Railroad stations. 13 Geographic analyses that analyze and integrate the dataset presented here with other historical geospatial datasets can contribute to a richer understanding of the interaction between individuals and the broader social, economic, political, and geographic forces that influenced the successes and limitations of the Underground Railroad. How such contextual forces interacted with the agency of the escapees and Underground Railroad conductors and stationmasters is a vital but largely unexplored aspect of the Underground Railroad which geographic analyses can help reveal. 14 The present dataset thus also aims to advance the use of geospatial data and geographic analyses for Black history15 and for historical and humanities research more broadly. 162024-2025EnglishAlabama, Delaware, Georgia, Kentucky, Louisiana, Maryland, Missouri, North Carolina, South Carolina, Virginia, Washington, D. C. 1853-1861LetterLife History or NarrativeMembership ListRunaway AdvertisementStill, William. The Underground Rail Road. . . . Philadelphia: Porter entries for West Norfolk, VA, Tanner’s Creek, VA, and Portsmouth, VA were renamed to aggregate with entries for Norfolk, VA). While the aim of such minor edits and transformations is, of course, to enhance geographic accuracy and consistency in geocoding, it should be acknowledged that these actions also have the potential to introduce bias or diminish place meanings embedded in the original, primary data source. Place names reflect power relations, and variations in particular names of places, and their spellings and pronunciations, often reflect significant histories and experiences of those places by different individuals or groups. 18 The use of geospatial data and cartographic representation has notable limitations in conveying such issues of power, history, and place meaning. 19 It should be emphasized that ethical considerations in the use, context, and integration of digital place name data that acknowledge these issues are important features of analyses of digital geographic data generally, 20 the present dataset included. Out of the 772 valid entries, geocoding yielded a total of 644 reliably geocoded locations to 92 unique cities/towns. Reasons for the inability to geocode 128 of the entries typically concerned entries listing cities or towns with names that did not correspond to any modern city or town names located in the historically encoded associated host states (even after accounting for changes in city names and county boundaries as noted above). Geocoding thus resulted in a total hit (success) rate=65% (644/995) and a valid hit rate=83% (644/772). While superficially it is understood that those non-geocoded entries were missing key geographic identifiers or did not contain place name data which matched to contemporary place identifiers, it is unclear precisely why or how these entries might otherwise differ from those that were geocoded. Such differences may be due simply to chance but may also reflect latent differences in the experiences or characteristics of certain individuals or groups. It may be the case that such location information was intentionally concealed by those escaping slavery or by Still himself, or that location “missingness” may be concentrated among those from a certain region or who used a certain mode of transportation for escape. Most importantly, researchers should consider the limitations of geographic representation embedded within geospatial data for understanding the actual experiences, and interactions with their geographic environments, of those escaping slavery. 21 Those entries which were not geocoded are not simply ‘missing’ but may also hold key information to understanding place meanings associated with the Underground Railroad. It may be helpful to contextualize the geocoding hit rates reported here among hit rates reported in other historical and contemporary research. Automated geocoding hit rates employing contemporary geospatial reference data to geocode late nineteenth- and early-twentieth-century British address data ranged from 37% to 77%. 22 These hit rates, at the upper end, are also broadly consistent with contemporary manually written, text-based address location data recorded by police departments and other organizations, 23 though, notably, geocoding of Still’s records in the present research focused on geocoding cities and towns as opposed to street addresses, a substantial difference in granularity. On the other hand, the historical and contemporary hit rates reported above are for addresses recorded in official governmental census and police administrative records. Broadly analogous hit rates for the present analysis can be considered particularly favorable given the historical setting of Still’s records, in which place name data were collected under illegal and clandestine circumstances, and were reported by individuals for whom place knowledge and naming conventions likely differed substantially from contemporary geospatial data representations. It is acknowledged that the use of contemporary digital geospatial reference data sources (i. e. ESRI World Geocoding Service) for geocoding historical place names, while often the only easily accessible source of digital geospatial reference data, 24 is problematic. Although an attempt was made in the present research to address obvious historical circumstances which may affect geocoding, such as a single county splitting into multiple counties over the historical period, unaccounted for official place name change over time likely inhibited geocoding for some entries. Progress has been made in the use of historical gazetteers for enhanced geocoding of historical data. 25 Although historical city and town data were not employed in the geocoding reported here, incorporating such data from historical digital gazetteers and other sources remains an important aim of future research for enhancing the geocoding hit rate and the positional accuracy of origins of escape from enslavement as recorded by Still. This issue of the use of appropriate historical geocoding reference data for the city/town level geocoding is addressed to some extent in the present research by the approach taken to geocode escapes by county. First, entries from the original 995 were excluded if the county name was missing (n=185), yielding a valid dataset of 810 entries (note that although Washington, D. C. is not a county, it was manually edited to be included in the county-level data so it could be mapped in the county-level dataset). County-level counts of entries were generated by joining these entries to an historical geospatial dataset of 1850 county names and GIS cartographic boundary lines. 26 After excluding entries for which Still’s recorded counties did not match any historical county names, or for which the proper county could not be identified due to mismatch or uncertainty of association with the correct state in the historical county dataset (e. g. certain counties, such as Kent County, appear in multiple states and thus could not be mapped if no state identifier were also included in the original data), geocoding yielded a total of 776 entries (total hit rate=78%, valid hit rate=96%) mapped to 55 different counties. Dataset Repository: Harvard Dataverse, https: //doi. org/10. 7910/DVN/1BLSRLCollege of Liberal Arts Undergraduate Research Award (LAURA), Temple UniversityCenter for the Humanities at Temple (CHAT) Fellowship, Temple University
Jeremy Mennis (Mon,) studied this question.