Each year the Institute for Advanced Technology in the Humanities awards a two- year fellowships for computer-assisted research in the humanities to a member of the humanities faculty at the University of Virginia. The selection committee evaluates proposals based on the merit of content and on innovative use of computer technology in humanities research. The Salem Witch Trials and Boston's Back Bay Fens projects were awarded fellowships in 1999, under the Institute's early policy of offering two one-year fellowships each year. The Salem Witch Trial project is using computer technology to document, research, and interpret the witch trials that took place between 1692 and 1693 in Salem, Massachusetts. The Boston's Back Bay Fens project is using computer technology to document, research, and interpret the history of the infrastructure of Boston's Back Bay Fens park and surrounding area from 1878 until the present. While the two projects differ significantly in the subjects being investigated, we can characterize each as an historical, social, and cultural study. The similarities in methodology and the related focus on social and cultural phenomena in a bounded space and time lead to similar computer data representations and architectures. At the outset of each project, the faculty researchers leading each project began a series of meetings with Institute faculty and staff, and graduate students assisting the projects. The goals of these meetings are to analyze the objectives of each project, to identify the resources and relations among them required to meet the objectives, to identify the appropriate computer architectures for representing the resources and meeting the research objectives, and, finally, to develop appropriate, detailed representational schemas. Both projects utilize a wide variety of both primary and secondary resources. While many of these resources are textual in nature, many represent a variety of other forms: geographic information, pictorial and other graphic materials, as well as social and cultural data extracted from primary resources and represented. A central challenge in all projects at the Institute is identifying the central "objects of interests." Objects of interest comprise both the key artifacts and resources that constitute the core evidence under study, as well as the people, places, things, and events documented in these resources. While the resources and the various social, cultural, and political entities they document are of primary importance, the relations among them are equally important. Both projects employ relational databases at the center of their architectures. The Salem Witch Trials database has primary tables representing people (accusers, accused, witnesses, and the like), organizations or corporate bodies (including courts, juries, households, and town, church, and legal officials), legal and judicial events (accusations, arrests, trials, convictions, executions, and the like), public and private buildings and structures (court houses, homes, and so on), and bibliographic resources (court records and other documentary evidence, critical early and modern secondary resources, pictorial materials, maps, derived and original digital materials, and so on). Many of the materials described in bibliographic records in the database are the sources for describing the individuals, organizations, events, and structures represented in the databases. Each primary table defines a set of descriptive attributes for each object. In turn, each object represented in a primary table is related to each of the others in a variety of ways through secondary tables. People, for example, are related through secondary tables from which family structures can be dynamically derived. Families, in turn, are related to households, which encompass broader relations than families. Families, further, are related to houses, which are also related to specific geographic information locating them in space, as well as in time. In this manner, people, organizations, structures, events, and resources function as primitives that, when interrelated, comprise more complex structures serving particular research and analytic objectives. Boston's Back Bay Fens has five primary tables: people, organizations or corporate bodies, events, objects or structures, and bibliographic resources. While the primary tables are remarkably similar to the tables in Salem Witch Trials tables, there is only some overlap in the attributes associated with each object. The differences in attributes primarily reflect differences in the nature of the disciplines (religious studies and history of architecture) and the nature of the objects and relations being investigated. The primary objects identified and the attributes identified and associated with them reflect the disciplinary methodology employed, and assumptions about which phenomenon will lead to furthering our understanding. In essence, the entities represented in each primary table and the relations between them in secondary tables reflect the hypothetical assumptions of each researcher. The information represented in the databases is extracted from a variety of primary and secondary resources. The latter category includes both second and first generation digital materials. Decisions concerning the appropriate architectures and representations of the primary and secondary resources required additional analysis, though not independent of the analysis and design of the databases. The entities represented in the database are necessarily related to the documentary and secondary sources, and these relations must necessarily be maintained. Maintaining these relations requires identifying and representing in both the database representations and the representations of the resources machine-readable data facilitating correlations. Given the historical nature of each project, spatial and temporal orientation of the people and organizations that play significant roles, the natural and cultural objects that coexist with them, and the events that reveal significant relations among them, become of critical importance. Geographic and temporal information thus plays a critical role in each of these projects. We can characterize existing Geographic Information Systems (GIS) as generally dealing with change over time through a series of static maps linked to relatively flat structures representing the data being plotted through time and space. Both of these projects, however, wanted to be able to display (visualize?) people, objects, and events in geographic space and through time dynamically. This presented the challenge of not merely re-presenting maps as images or even GIS, but representing geographic information in relation to entities represented in the respective databases. The data in the databases is used "to drive" dynamic temporal-geographic displays. Each project also had to consider the representation of texts, pictorial and other graphic material. Given past and extensive experience with markup and imaging technologies, texts were not a difficult challenge. Neither project had significant interest in text bearing artifacts as such, or in exploring text or editing theory. Both projects were primarily interested in the information content of texts: reading, searching. In addition to reading and searching, individual texts and people, objects, and events documented in them also had to be represented in a machine-readable manner that would facilitate correlations with entities in the relational databases, and, through them, maps. TEI Lite was chosen as the method for representing texts. Some texts from primary resources are also represented as images, and some texts, where creating machine-readable texts was not deemed necessary or currently economically feasible, are only represented as images. The Salem Witch Trials and the Boston's Back Bay Fens projects are socio- cultural historical studies. While the subject matter in each differs, the similarities in methodology lead to similarities in the phenomenon represented and the methods for representing them. Both projects employ relational databases as axis around which are correlated sophisticated dynamic GIS systems, and resources represented using XML and XML systems, as well other technologies appropriate to the nature of the data and research objectives. While the databases play a central role in the analysis and navigation of the data and resources, XML plays the role of representing texts, as well as integration of the various resources and resource systems employed in the two projects.

Modern historiography, in whatever discipline it might occur, is primarily an attempt at multi-leveled contextualization--a weaving together of anecdotal evidence and "micro-narrative" with broader, and perhaps more data-centric quantitative evidence. The modern historian, in other words, cares as much about letters, photographs, and snippets of conversation as about census data, geographical information, and tax records. The former has traditionally resided in digital archives; the latter, in relational databases and GIS systems. Effecting the union of the various technical approaches is not by itself a difficult matter. RDBMS, GIS, and digital collections all participate in a similar quest for order and organization. But with the rise of humanities computing has come opportunities for thinking about how technology can be brought to bear on the two-pronged historicism now current. How can the traditional forms of computer-assisted data representation be used to draw forth unforeseen connections, micro-narratives, and anecdotal anomalies in the large and diverse datasets available to the modern historian? The Institute for Advanced Technology in the Humanities confronted this problem recently with two historiographic projects: "The Salem Witchcraft Papers" and "Evolutionary Infrastructures: Boston's Back Bay Fens." Both of the scholars initiating the projects wanted to capture large datasets concerning events in a sizeable geographic area over a long period of time. They both envisioned databases that would contain the usual large-dataset materials, like GIS data, building records, and population census materials. In both cases, however, the usefulness of the historical record was thought to depend equally upon individual images, biographical information, and individual documents. The database schema we designed (discussed in a previous paper by D. Pitti) organizes the data into a deliberately heterogeneous ontology: people, places, significant events, objects, corporate bodies, and bibliographical records. The data in each of these major tables is organized according to its own internal logic, but these five tables are related in an overall schema that in both systems reaches to almost a hundred separate relational junctions and subtables. The schema, therefore, consists mainly of many-to-many relationships of the form "person/corporate_body.owner()," "object/event.role()," "bibliographical_record/person.about()," and so forth. The data entry mechanisms for the projects involve a combination of both large-scale import, as well as more finely grained data entry. The former method is employed for entering the main records themselves, but the relationships are hand-drawn and classified by members of each respective team. The knowledge possessed by any particular team member is necessarily partial and imperfect. A scholar might have evidence of a connection between the builder of the Boston Conservatory and the building itself, but not about the relationship between that builder and other buildings, or about the relationship between that building and what stood before it. However, the database is all the time watching and recording these connections. We developed a generic Java class library (called the Sibelius libraries) that can search through the junction tables for any particular record looking for these relationships. Whenever the user arrives at a particular record, all of the relationships with other tables are then aggregated and presented to the user. We believe this function is significant for two reasons. On the one hand, it creates synergies out of the partial ontologies possessed by the individual team members; what was partial before may now reveal itself in full. More significant, however, is the fact that these "unwound" connections often have the character of micro-narratives. A typical search result for buildings in the Muddy River Watershed in Boston, for example, might yield information about the New England Conservatory. The linking mechanism will in turn yield further connections to various architects, places, buildings, and events, each of which will yield further connections. One is therefore able to create statements like, "The New England Conservatory, which was once the Harvard Medical Library, was built by Henry Hunnewell, who also built the student house at 96 Fenway, which was once owned by T. Frothingham . . ." At no point was this narrative explicitly entered into the database in this form, and yet the system can use the local connections to build the larger narrative. This paper will discuss our method from both a philosophical and technical standpoint, and will go on to suggest some directions for how research might proceed on the visualization of these "unwound" narratives. We have been experimenting, in particular, with graph visualization software, which we hope eventually to apply to the ontological submaps created by the Sibelius engine.