On my desk I keep one reference work within easy reach: the New Historical Atlas of Religion in America. It is not just that I am fascinated by maps. When writing or teaching about religion, I try to remember the dictum that I first learned from Jane Kamensky: “events take place.” To bring place into my work, I need to remind myself constantly of the where and when of American religious history.

But I also keep in mind Laurie Maffly-Kipp’s memorably titled essay “If It’s South Dakota You Must Be Episcopalian.” In that essay Maffly-Kipp reviews several atlases of American religion and makes a number of telling critiques of the whole enterprise of mapping religion, and indeed mapping in the humanities generally. Those critiques can be lumped into two categories: sources and definitions. First, sources: “We are still limited by data that is partial, ambiguous, and clearly slanted toward things that can be counted and people that traditionally have been seen as significant.” In other words, our sources for making maps of religion aren’t very good, and the best sources let us make maps of groups that we already know a lot about. Second, definitions: Maffly-Kipp observes that our “our dazzling new technologies and spatial theories” might only have “brought us back to much more circumscribed definitions of religious experience.” In other words, map-making requires us to use denominational categories when the whole trend of religious history and religious studies is towards more complex, layered understandings of religion.

The scholars who created the leading atlases of American religion often raised the same questions. When I look at the New Historical Atlas or Bret Carroll's Routledge Historical Atlas of Religion in America, I see their great ingenuity in dealing with these questions within the limits of mapmaking in print. But it seems to me that digital mapmaking permits us to escape some of the constraints of maps in print atlases and monographs. So my questions today are these: Can digital mapmaking let us make maps which address these theoretical critiques? How can we take the next step from the work of Gaustad and others, while taking into account Maffly-Kipp's concerns?

I'll try to answer those questions by showing you two projects that I have worked on. The first is a class I taught on mapping Boston's religions from the American revolution to the 1880s. And—since I've done more interesting work since I submitted my abstract a year ago—I will also sketch out the sources and methods for digital maps of religion on a much larger scale.

My general argument is that there are large sources of data on American religion after the colonial period and before Word War II which historians have not used to make maps. Scholars have not passed over these sources because they are unaware of them, but because they could not meaningfully represent them in print maps. The problem is one of resolution. Print atlases could convey relatively few data points. Furthermore, because atlases can contain only so many maps, they have often been forced to set their chronological or geographic scope very large. By using these more detailed sources we are able to make maps which better approximate the sophisticated thinking about religious categories that we expect from our prose. These richer maps can tell us not just more, but more humanistic, things about religious history. To take advantage of these more comprehensive sources we need digital maps. To be sure, digital history has had more than its share of hubris, more than we have time to repent of today. But digital maps do offer the possibility for working at different scales, for displaying change over time, for integrating maps with our sources, and for crafting narratives with maps. While none of these advantages entirely solves with the problem of mapping humanistically, they do permit us to at least start to address these theoretical concerns.

My first work to show you is a deep map of the history of religion in Boston. In the spring of 2014, I taught a course at Brandeis University on mapping the religions of Boston in the nineteenth century. In this project-based course, students did original research to locate congregations in Boston and created a map of the religious landscape of the city as it changed over time. Students looked at nineteenth-century maps of Boston, city directories, and denominational records to find the locations of about 235 congregations. They then developed a metadata schema to handle the information they had researched. In other words, they had to design the categories for their maps and figure out how to represent change over time and ambiguity in the sources. Using Omeka and Neatline, students created records for the congregations, including information such as the date a congregation was formed and the date it disbanded, its street location, denomination, and any other relevant information, including citations. Many of the records included photographs or engravings of the church buildings. Churches were pegged to a timeline, so that changes in congregations could be shown dynamically. Students also georectified original maps to put beneath their congregations to show the changing landscape of the city. This was especially important for Boston, since large parts of the city were reclaimed from the harbor or the Back Bay during the middle of the nineteenth century. Finally, students created exhibits (online essays, really) about particular aspects of what they had learned from creating the map.

I wish to be cautious about about making historical claims from the map. The students were all excellent, but the class was very small; too small, probably, for the size of the project that we attempted. I can't claim that we identified all the congregations in the Boston for that eighty year period, which would have taken a team of graduate students much longer than a semester.

But I can point out some of the methodological advantages of this kind of deep mapping. First, by animating the map we were able to see change over time. One student who was interested in the Methodist congregations was able to chart their spread throughout the city. Methodists, like everybody else, started in the downtown section of the city. But as new land opened because of the Back Bay landfill, churches moved to the west. We identified photos of churches in the Back Bay where they were literally the only buildings in sight. By observing movements connected to photographs, we were able to discuss the significance of landscape—an observation only available through this kind of deep mapping.

Second, students learned from creating this map to think of religions as being part of an ecosystem. The map led to useful questions about how religious groups related to one another and to their environment. Another student wrote about the disputes between the “Bayer” (i.e., German) and the “Polander” (i.e., Polish) synagogues. She was able to identify many of the homes of the Jews who made up the membership of Congregation Ohabei Shalom. The split between these congregations over matters of culture and ritual later influenced the settlement of Jews, who lived in different neighborhoods of the city at different times. A third student examined racial segregation in housing and churches along with interaction between black and white Baptist churches on Beacon Hill and in the North End.

Finally students thought deeply about the categories we had to use. So far from mapping obscuring the categories we used, it was the central means of discussing them. One day we read Ralph Waldo Emerson's “Divinity School Address,” with its call for his audience to be “yourself a newborn bard of the Holy Ghost,—cast behind you all conformity.” One student observed that Emerson would not have found much use in our class project, because we were mapping congregations rather than individuals, institutions rather than prophets. Our map featured a city, when Emerson would have called us to “nature.” It was a brilliant observation, because it used our reading of Emerson’s text to interrogate the assumptions about religion and society that underlay our class project: assumptions that I had made, and also assumption inherent to mapmaking. But our discussion also led us to conclude that though Emerson had a point about individualism, many people in Boston would not have so easily discarded the congregations of which they were members.

While the map by no means fully resolves the critiques that I cited earlier, it does start to address them. In particular, this map is based on very different sources than the maps in Gaustad, Barlow, and Dishno's New Historical Atlas of Religion. Where they used data from the Census Bureau for the majority of their national maps, we were able to use just about any kind of source and to capture many more details about it within the Omeka records. And though our map was still beholden to more or less the same denominational categories as the New Historical Atlas, we were able to turn our attention to ecosystems and interactions.

All of our methodological differences from the New Historical Atlas depended on looking at a city rather than a nation. Is there a way to extend these kind of methods to a larger scale?

Most of the maps in the Gaustad atlas depict a particular denomination in a given year using aggregated data from the U.S. Census Bureau about the number of congregations in a county. These maps are thus able to give an overview of the regional prominence of particular religious groups county by county. The map of Congregationalists in the 1850 census is typical.

Thanks to the National Historic Geographic Information System (NHGIS) at the University of Minnesota, it is easy to reproduce any of the maps in the Gaustad atlas which are based on the census. NHGIS has digitized the aggregate-level census data and keyed it to shapefiles of United States in each decade. We can use that data to create web (or print, for that matter) maps which are visually similar to Gaustad's and which depict the exact same data.

The mere translation of the Gaustad maps from print to digital gains us nothing. Beyond the strong presence of Congregationalism in New England and its spread to parts of Ohio and Illinois, we cannot learn much from county-level maps. Were we to zoom into Massachusetts, we would learn that each county has 25 churches or more (how many more?) but we would not know where those churches were located, how close they were to other Congregationalist churches (let alone churches of other denominations) or how many congregants they had. The problem is that the county is too coarse for meaningful maps.

But there is much more detailed information available to us. I mean most obviously the records of many denominations, which often contain detailed information about membership, and usually at a minimum contain long lists of congregations. Often they even include information about Canada. Once these lists are digitized and geocoded, they can provide maps of much greater resolution than the aggregate census data.

Leaving aside for a moment the ability to pan and zoom the map, to access additional information by clicking on the congregations, and to link out to primary source, how much more information does this map contain than the equivalent map in the New Historical Atlas? The Atlas has 137 counties with at least one Congregationalist church: it thus makes 137 meaningful marks on the page. This map from denominational records has 2,212 Congregationalist churches in the United States and Canada. Also, the map is able to depict not just the existence of a congregation, but also its size in members. This is a 32-fold increase in the amount of information conveyed in the map. For comparison, that is the difference between an 800 x 600 pixel CRT monitor and two 4K monitors.

This map really does let us see things that we could not in the map based on aggregate census data. For one, we can see how New England's cities were complex ecosystem of small and large Congregationalist churches. Only a few miles separated the pastor-less sixteen members of a church in North Chelsea from the 603 members of Park Street Church. Second, there is a sharper line between the Massachusetts border with New York than between churches in Michigan and Canada. And third, we can see the effects of personality: Oberlin's eleven hundred Congregationalists drawn by Charles Finney far outweigh the membership of any other church west of Massachusetts.

Another use of these maps is the ability to set them in motion over time. This effectively lets us make scores of maps where before we only had one.

One other observation about sources. Denominational records for the nineteenth century vary widely from denomination to denomination, and can even vary in detail from year to year within the same denomination. So another source of information available to historians are the published census data up through 1926 (or for some denominations, through 1936). You will recall that I criticized the county-level data from the census as obscuring as much as it reveals. In order to create those aggregate counts, the Census Bureau had to keep track of data at the level of congregations. The published census reports contain counts of churches and members in cities with more than 25,000 people. Compared to the attention paid to cities like Boston, few of us have done work on Council Bluffs, Iowa, or Woonsocket, RI, but it is possible to know with some specificity about those cities. For example, in the published 1926 census report there are 292 cities containing records for 215 religions or denominations.

Though I have not yet tracked them down, it is also possible to access the raw returns for the 1926 Census. For an extraordinary number of congregations in the United States, there are census returns giving addresses and often counts of members. The Census Bureau used these in compiling their aggregate reports, but they can in turn be sources for detailed maps.

In conclusion, the scholars who created print atlases did extraordinary work, not least given the constraints of the medium of print. But print could only show a limited amount of data. Given how little data we’ve actually had to work with, it's not surprising that the maps we have are open to criticisms. Some of these problems are inherent in the nature of mapping—you'll never find Emerson's "newborn bard of the Holy Ghost" on a map like this. But at least some of the critiques can begin to be solved by producing far more detailed maps of American religion, which in turn let us ask more complex questions about ecosystems, interactions, and the layering of space and identities.