The epistemology of digital modelling

The basic question that arises with digital models of historic sites and events is the character of their outcomes. We are familiar with the truth-claims of conventional historical research. We are also familiar with models that tell us far more about our culture's perception of the past than about the past itself, such as Sleeping Beauty's castle at Disneyland, a building far more about the fantasies of Disney Studio's animators than it is about medieval castles. The Virtual Paul's Cross Project has sought to follow the principles set forth in the London Charter for the Computer-based Visualisation of Cultural Heritage (www.londoncharter.org) and to apply them to acoustic as well as to visual modelling, so that such methods "are applied with scholarly rigour, and that the outcomes of research that include [such tools for research] accurately convey to users the status of the knowledge that they represent, such as distinctions between evidence and hypothesis, and between different levels of probability" (2009, 2). In this regard as well, it is important to remember that the outcome of a project using digital modelling is not an exercise in time travel; the images one sees-and for that matter, the sounds one hears-are not mirrors of what was, no matter how convincing they may appear. They are instead constructions based on interpretations of existing data, assemblages of information made available for consideration through the use of digital tools.

Digital models are by nature "conjectural artefacts," simulations of things created according to principles of interpretation, organisation, and display (McCarty 2005, 47). They represent specific assumptions about what comes down to us about a historic event, and, in the process, make those assumptions available for our understanding. On the one hand, they enable us to integrate a wide variety of what we believe to be true about a past event into a coherent presentation; on the other hand, they help us clarify our assumptions, help us understand what it is that we believe to be true and how we have come to believe that. The process of modelling, therefore, helps us come to know the character, as well as the composition, of the information we draw on to make our models.

Digital modelling also helps us understand the limits of the various kinds of information we have about the place or event. This process also challenges us to identify and evaluate models for supplying what we lack. Just as one can fill in missing pieces of a visual model based on patterns exhibited elsewhere in the data, so too one can approximate events for which no specific information survives. In modelling the visual space of St. Paul's Cross we used representative rather than specific examples to supply architectural information for buildings we know were there but for which we have no visual record. Similarly, in modelling the acoustic space, we have supplied representative, randomly occurring sounds from the kinds of sources documented in the historic record to meet the need for the experience of ambient noise. The Project models an event that unfolds in real time, so we have been challenged to ask how, specifically, the Paul's Cross sermon took place, how people arrived at their places for the event, how the crowd was called to order, how ambient noise was accommodated, and how people responded to the unfolding of the sermon.

We must always be clear that any specific state of a digital model is at best provisional; change the assumptions on which it is based and the outcome changes as well. For example, the representational, approximate, and provisional character of our models are embodied in the specific details of the event we are modelling. We present on the website an experiential model of John Donne's sermon for Gunpowder Day, November 5, 1622, delivered at Paul's Cross, in Paul's Churchyard. We chose this sermon originally because our goal was at every step of the way to be as accurate as possible, as close to the reality of London, of Paul's Churchyard, and of the experience of the Paul's Cross sermon, as we possibly could be. This sermon helps us fulfil that goal of accuracy in a special way because it is unique among all of Donne's surviving sermons as the one set down the closest to the actual delivery of the sermon. This text–a manuscript prepared within a day or so of its delivery by a professional scribe and which contains annotations and corrections in Donne's own handwriting–is therefore likely to be the most accurate record we have of what Donne actually said while preaching any of his sermons (Donne 1996). Yet this sermon, while it meets the requirements of a Paul's Cross sermon in every way–prepared with every expectation of being delivered at Paul's Cross and exhibiting every distinctive characteristic of a Paul's Cross sermon, was "because of the weather" actually delivered inside the cathedral. So, in this project, we are not representing an event that actually happened, but an event that should have happened, but didn't. It now finally happens, in virtual space though within the real two hours of time originally allocated for it, almost 400 years later. That captures for me both the restorative and the speculative character of digital modelling, what it can do and what kind of reality it constitutes.

As Braxton Boren puts it, regarding his work with Venetian churches, "if a model is constructed using accurate geometrical and material data, its simulations can provide a powerful tool for recreating lost soundscapes" (2010, 9). Boren's work benefits from the fact that the actual structures he is modelling survive in today's Venice. However much they might differ today from their original architectural configurations, they are still available to provide "accurate geometrical and material data" to serve as a standard against which to evaluate the accuracy of his models. One major challenge we have faced with the Virtual Paul's Cross Project has been to compensate for the absence of similar external referents. The medieval cathedral and its churchyard and the Paul's Cross preaching station are gone forever, yet we still seek to provide "accurate geometrical and material data" so we can still speak in terms of accuracy for the acoustic model of the space as well as for the sounds one hears within the acoustic model.

Regarding the space itself, an extensive record survives of the visual appearance of St. Paul's and its environs, but this record is not complete for the entire area, nor were the images that survive created with modern photo-realistic standards in mind. Regarding the sounds, no one knows, or can ever know, the sound of John Donne's voice or the full experience of his preaching style. That said, the fact that we do not know everything does not mean that we know nothing. There is a great deal of information available about St. Paul's and Paul's Cross and about the Paul's Cross sermon, especially as delivered by John Donne. This evidence comes in a variety of forms, as we will see; some is visual, some is textual, some is derived inferentially from other evidence, and some is on the surface tangential to the events at hand. This project has required the assembly of all the evidence we have had the wit, diligence, and good fortune to be able to assemble. It has also required assessment of each piece of evidence to determine what, and in what way, this piece of evidence contributes to the larger picture. We have had to conclude that no evidence is valuable un-interpreted, or un-assessed as to its validity or as to the degree of its validity.

This essay explores the process we have followed as we have sought to create a research tool for exploring the phenomenon of the Paul's Cross sermon. The strength of this project lies not in its capacity to recreate a past event, but in its capacity to provide an experiential approach to conceptualising the visual and acoustic dimensions of the Paul's Cross sermon as an event that took place in this section of London in the early seventeenth century. To do this has required the integration of the historic record of St. Paul's appearance with modern archaeological research as well as the integration of the manuscript of Donne's sermon with historic accounts of John Donne's preaching style and with modern understandings of the sound of early modern London English. As an experiment in modelling methodology and its use in study of historical performance, this project also helps us clarify aspects of this event by requiring us to reconstruct the order of events for a Paul's Cross sermon, addressing basic questions of how things were done, under what conditions, and in what order.

The visual model

The Project combines a visual model (Figure 1) of the north east end of Paul's Churchyard, including St. Paul's Cathedral, the Paul's Cross Preaching Station, and the Churchyard's bookshops, with an acoustic model of the same space. Our goals for this model were threefold: first, to integrate into a single visual presentation all the information we could verify about the visual appearance of Paul's Churchyard between 10:00 am and noon on a day in London in early November of 1622; second, to provide complete transparency about the sources of this information; and third, to ensure that the visual model we created would serve as an appropriate source for the acoustic model to be based on it.

Figure 1: From the Visual Model of Paul's Churchyard, the Cross Yard. Constructed by Joshua Stephens and rendered by Jordan Grey.

The visual model was made by Joshua Stephens, a graduate student in architecture, using Google Sketch-Up, a widely-used architectural modelling software. Working under the supervision of David Hill, Associate Professor of architecture at NC State University, Stephens was assisted by Jordan Gray, another NC State graduate student in architecture, who rendered Stephens' images to incorporate meteorological and climate data and the signs of aging in a building already several hundred years old. The website itself was developed by Craig Johnson, yet another graduate student in architecture at NC State. This model integrates the surviving visual record of this part of London in the sixteenth and seventeenth centuries, especially the work of John Gipkin (Figure 2) and Wenceslaus Hollar (Figure 3), with a collection of measurements that locate the sixteenth- and seventeenth-century visual imagery on a grounding of verifiable data.

Figure 2: John Gipkin. Paul's Cross. 1616. Painting. Image courtesy of Bridgeman Art Library, New York, and the Society of Antiquaries, London.

Figure 3: Wenceslaus Hollar. St Paul's Cathedral, the east end. 1650's. Drawing. Image courtesy of Sotheby's, London.

These sets of measurements include the results of a survey of the Cathedral done by Christopher Wren in the early 1660's (Figure 4), measurements of the foundations of houses surrounding the Cross yard taken by surveyors after the Great Fire of 1666 (Blayney 1990), measurements of the foundation of the Paul's Cross preaching station done in the 1880's, and measurements of the foundations of the medieval cathedral made by archaeologists (Schofield 2011) working over the past century in Paul's churchyard (Figure 5).

Figure 4: Christopher Wren. 1662. Scale Drawing of pre-Fire St. Paul's Showing Proposed Dome. Image courtesy of Warden and Fellows, All Souls College, Oxford.

Figure 5: St. Paul's Churchyard around 1450. Image courtesy of John Schofield (2011).

The visual record of St. Paul's Cathedral and its environs is especially rich because St. Paul's was the cathedral of the Diocese of London and thus the centre of religious life in the largest city in early modern England. Visual images of St. Paul's made before the Great Fire range in media from a drawing scratched into the wall of Ashwell Church in Hertfordshire to a host of engravings and paintings, images included in maps (Figure 6) and panoramas of London, and a collection of seals and badges (Schofield 2011; Keene, Burns, and Saint 2004).

Figure 6: St. Paul's Cathedral. 1550(?). Detail from the Copperplate Map. Image courtesy of the Museum of London.

These all agree as to the general appearance of the building, although they do vary in the scale of depiction, the specificity of the details, and the angle of view from which the building is seen in the image. Visscher's View of London (1616, Figure 7) for example, shows St. Paul's as a much taller building than it appears to be in Wenceslaus Hollar's Panorama of London (1655, Figure 8).

Figure 7: Claes Jansz Visscher. Detail from View of London, Panoramic View. Ca. 1625. Engraving. Image courtesy of the Museum of London.

Figure 8: Wenceslaus Hollar. Detail from Panorama of London. 1647. Engraving. Image courtesy of the Wenceslaus Hollar Digital Collection, University of Toronto.

Contemporary evidence suggests that Hollar's view is the more accurate and that Visscher's view reflects his sense of the importance of the building rather than the way it actually looked. Conventions for depiction of objects in the early modern period clearly allow some artists to omit or expand details when their depiction would pose a challenge to the composition or to the space available for depiction. John Gipkin's painting of a Paul's Cross sermon (Figure 2) shows the east end of the cathedral as well as the crossing, but to do so it must omit seven of the eleven bays of the cathedral's choir. Gipkin also depicts the important people sitting in the sermon house to be physically larger than the (less socially prominent) people sitting on the ground in front of the preacher, even though they are some twenty-five feet further away from the imagined position of the viewer.

The most valuable, detailed, and at least relatively accurate depictions of St. Paul's Cathedral itself are those done by Wenceslaus Hollar in the 1650's for William Dugdale's History of St. Paul's Cathedral (1658). Hollar made drawings of the cathedral's exterior and interior, from which he developed a set of engravings that appear in Dugdale's book. Two of the drawings survive, one (Figure 3) now at the Metropolitan Museum of Art in New York City and the other (Figure 9) at Oxford's Ashmolean Museum.

Figure 9: Wenceslaus Hollar. St. Paul's Cathedral, the north side. 1650's. Drawing. Image courtesy of Ashmolean Museum, Oxford.

Hollar's engravings usually but not always follow his drawings. Figure 10 is Hollar's engraving of the north side of St. Paul's. Note in comparison the overall similarities, but also note differences in detail in the two images, especially in the number of doors Hollar shows along the north side of the choir as well as differences in detail between the two views of the north front of the north transept.

Figure 10: Wenceslaus Hollar. St Paul's Cathedral, north side. 1658. Engraving. In Dugdale 1658. Image courtesy of the Wenceslaus Hollar Digital Collection, University of Toronto.

Our major conclusion from a review of the visual record of the cathedral is that early modern standards for accurate depiction are different from ours, that no image should be accepted as a direct representation of the object being depicted, and thus, regardless of how carefully we work to be accurate, every image we use to provide information for the visual model of St. Paul's turns out to involve some degree of approximation. The Gipkin painting of a Paul's Cross sermon in the process of being delivered, for example, blithely truncates both the north transept and the choir of the cathedral to make them fit into the space available on the panel Gipkin chose for his painting. Other images of Paul's Cross (see Figures 11 and 14) show a building significantly smaller than either the Gipkin image, or Penrose's survey data would support.

Figure 11: John Speed. Paul's Cross. 1611. Detail from Map of Middlesex. In Speed 1611. Image courtesy of the British Library.

The model we constructed began with the incorporation of John Schofield's survey of St. Paul's original foundations still in the ground in London and the boundaries of the other buildings in Paul's Churchyard into Sketch-Up (Figure 12). Gradually, more and more of the information available to us was incorporated into the model (Figure 13). One can follow the course of the visual model's development by reviewing entries on the Project's blog (thevirtualpaulscrossproject.blogspot.com).

Figure 12: Paul's Churchyard, the Cross Yard. From the Visual Model constructed by Joshua Stephens.

As we developed the visual model, we of course ran into a long series of questions about the character and accuracy of the evidence we were working with, as well as what to do when evidence was sketchy, ambiguous, or simply non-existent. In such cases, we were forced to consider the extent to which we could proceed on the basis of evidence that was approximate or representative.

Figure 13: Churchyard, the Cross Yard. From the Visual Model constructed by Joshua Stephens.

Indeed, in the process of creating the visual model of Paul's Cross, we have come to think in terms of degrees of approximation. At the highest level, even the most accurate of our representations will lack fine detail, because we can model more detail than Hollar gives us in his drawings and engravings. There is also the issue that any surviving image of St. Paul's records the look of the building at a specific time; for example, the appearance of the north end of the north transept in Hollar's images is based on the way this part of the building looked after Inigo Jones' neo-classical remodelling efforts in the 1630's. We have restored the medieval state of the north transept's façade, but our work in this part of the model has had to be based on images far less detailed than those of Hollar (see for example Figures 6 and 11).

On yet another level, some details of size and scale have had to be approximated. As we will see when we get to a discussion of the booksellers' shops that ring the churchyard, we have no contemporary images of these structures and have had to use as representative models the appearance of buildings that survive from the early modern period in cathedral towns and urban areas of England. For the acoustic model to represent usefully the way sound behaved in Paul's Churchyard, however, we needed not just an accurate model of the geometric forms of the buildings inside the churchyard, but an accurate measurement of the relationships among them–the distances between structures and the scale and dimensions of the structures themselves. We have been able to ground our reconstruction of the buildings of Paul's Churchyard on direct surveys of their foundations and, in the case of the cathedral, a late seventeenth-century survey of the height of the arches, the windows, and the ceiling. The most basic of these are the measurements of the foundations of the various structures in the churchyard.

The foundations of pre-Fire St. Paul's Cathedral remain in the ground in London. They have recently been surveyed (Figure 5) by Cathedral archaeologist John Schofield and made available in his St. Paul's Cathedral Before Wren (2011). Regarding the cathedral itself, our work incorporates measurements of the internal dimensions of pre-Fire St. Paul's Cathedral made in the early 1660s by Christopher Wren, who had been hired by the cathedral to devise a plan for renovation of the building.  His measurements survive in a drawing he made showing the building with a neoclassical dome on top (Figure 4).

The foundation of the Paul's Cross preaching station similarly remains in the ground in London. It was excavated and measured in 1878 by Francis Penrose, who published his findings in Archaeologia in 1883. The foundations of the buildings around the Churchyard were surveyed by the City of London after the Great Fire of 1666 destroyed the Cathedral and all its surrounding buildings. The results of these surveys have been gathered by Peter Blayney into his The Bookshops in Paul's Cross Churchyard (1990). In addition, the dimensions of the buildings around the Churchyard are sometimes, though not invariably, indicated by the City of London surveyors whose records have been assembled by Blayney. They indicate that all these buildings were at least three stories high, while some were four stories high. We have followed their guidance in modelling the scale of these buildings. As a result of our work with historic and contemporary evidence, with stylised, impressionistic, and representational images, and with numerical data about lengths and heights and widths, we believe that we have created a visual model that is highly accurate in essentials and at least representationally accurate in its details.

The challenges in integrating measurements of actual objects with the historic visual record and in using representational or approximate models where the historic record is only suggestive rather than informative especially come into play in our modelling of the Paul's Cross preaching station. This structure was a wooden building with a lead roof that stood on a stone base that raised it some feet above the crowd. Our sources for its appearance are three-fold: John Gipkin's painting of 1616 (Figure 2), an engraving from Speed's Theatre of the Empire of Great Britain, from 1611 (Figure 11), and an engraving of 1621 that appeared in the seventeenth-century in at least two different states (Figure 14 and Figure 15). All these visual depictions agree in basic terms about the design of the building, but they disagree about the dimensions and proportions of the structure.

Figure 14: View of St. Paul's Cross as it appeared in the year 1621. Woodcut. Image courtesy of Bridgeman Art Library.

Figure 15: View of Paul's Cross from 1625. Woodcut. Image courtesy of Bridgeman Art Library.

Francis Penrose's excavation of the foundations of Paul's Cross provides us with solid evidence for the scale of the overall structure. Penrose (1883) found that the stone base of the structure was thirty-four feet across and that the structure in which he stood to address his congregation was seventeen feet across (Figure 14). These measurements suggest a building much larger than the structure shown in contemporary woodcuts (Figures 14 and 15). The overall look of the structure in Gipkin's painting (Figure 2)–a structure tall enough to stand inside comfortably and large enough to accommodate several people–is much more in keeping with the measurements provided by Penrose's excavations (Figure 16).

Figure 16: Francis Penrose. 1883. Scale Drawing of the Foundation of Paul's Cross. In Penrose 1883.

None of this evidence, however, helps us with either the height of the overall structure or, perhaps more important, how tall the opening was in which the preacher stood to deliver his sermon. The preaching station we have in the visual model incorporates the seventeen-foot wide base of the original structure and approximates the proportions of Gipkin's structure regarding the opening around the preacher, rather than the height of the structure overall, on the grounds that the size of this opening was more important to the audibility of the preacher than the height of the roof and its cross above his head. The result is a building (Figure 17) that is differently proportioned from the one shown in Gipkin's painting, chiefly in that the resulting structure is not as tall as the Gipkin image (see detail, Figure 18) would suggest, relative to its width. Thus must one kind of data (the proportional look of the structure) be accommodated to another kind (the measured width of the base).

Figure 17: Paul's Cross, from the Visual Model of Paul's Churchyard, the Cross Yard. Constructed by Joshua Stephens and rendered by Jordan Grey.

Figure 18: John Gipkin. Detail of Paul's Cross. 1616. Painting. Image courtesy of the Bridgeman Art Library, New York, and the Society of Antiquaries, London.

Another concern regarding the preaching station was to determine in what direction the pulpit faced. In Gipkin's painting, the pulpit faces northwest, which is curious because this orientation turns the preacher not only away from the "people of quality" seated in the sermon house to his left, but in fact away from the bulk of the crowd sitting in front of him (Figure 18). The image in Speed's Theatre of Empire (Figure 11), on the other hand, shows the pulpit facing generally southwestward toward the point of intersection between the choir and the north transept, a perhaps more reasonable direction, since it turns the preacher more toward the sermon house.  Adding to the confusion is the seventeenth-century woodcut we have already examined (Figures 14 and Figure 15), which shows the preacher facing southward toward the spot along the choir where the sermon house was located, although it does not show the sermon house in its place against the north wall of the choir.

This confusion was clarified somewhat by Pamela Tudor-Craig's comparison of this woodcut with the Gipkin painting in her 'Old St. Paul's': the Society of Antiquaries' Diptych, 1616 (2004). Tudor-Craig points out that the horse and groom on the right side of the engraving is the same as the horse and groom in the left side of the Gipkin painting (Figure 19). 

Figure 19: John Gipkin. Detail of Paul's Cross. 1616. Painting. Image courtesy of the Bridgeman Art Library, New York, and the Society of Antiquaries, London.

Then it occurred to me that an engraving is a mirror image of the original image produced by the artist on the plate. I opened a file of the 1621 image in Photoshop, flipped it horizontally, and got Figure 20. Suddenly, lots of other echoes of the Gipkin painting stood out in this engraving. Note the people standing behind the crowd and between the crowd and the cathedral, for example, and the facade of the cathedral itself, which now looks a lot more like the facade of the north transept than it does like the facade of the east end of the choir. It has a door in it, for example, which the east end of the cathedral did not.

Figure 20: View of St. Paul's Cross as it appeared in the year 1621. Reversed image by the author. Woodcut. Image courtesy of the Bridgeman Art Library.

This change in perspective clarifies the relationship between the Gipkin painting and the woodcut image. The woodcut showing the preacher facing southward, toward the north side of the choir, is actually an image derivative of the Gipkin painting, but is reversed in the engraving process. This image is therefore not a primary source of information about Paul's Cross.  Nevertheless, these engravings can help us understanding the design of Paul's Cross. Gipkin, it turns out, chose a position from which to make his painting that, had he made his painting geometrically correct, would have made the preacher invisible to the viewer of the painting. Gipkin's goal was to show all the elements of the scene, so in his composition he turned the Cross preaching station to the right, bringing the preacher into view for the observer of the painting. So, we've concluded that the engraving is right about the orientation of the preacher, and about his position a step forward of the roof of the Cross structure, but that the Gipkin painting is right in other, though complementary, ways. Hence, we decided that the preacher at Paul's Cross faced westward, toward the east wall of the north transept, and stood in a small pulpit a step out from under the roof line of the Cross structure itself.

Another matter we had to come to terms with is the colour of the cathedral. Again, the evidence is contradictory. Gipkin's painting–the only colour rendition of the building I know of–shows it in a deep yellow (Figure 2) while the actual stones that survive from the building, on display in today's St. Paul's Cathedral on shelves just outside the door to the cathedral library, turn out to be grey in colour. We tried both colours in early versions of the model but ultimately settled on a light grayish yellow, a colour choice based on the fact that the choir end of the cathedral was made of Caen stone imported from France, the same stone used to construct the Tower of London. Thus, the colour of St. Paul's Cathedral in our visual model is based on the colour of the Tower of London as it appears today. In addition, we have been able to bring into the visual model some traces of urban life in a world lit only by fire. Heating and cooking was done in London of the early seventeenth-century by burning an especially sulfurous kind of coal called seacoal, which was burned in such large amounts that "a seemingly perpetual cloud of sulfurous smoke [hung] over London" (Hiltner 2009). William Dugdale blamed the use of this particular form of coal for some of the damage that time had caused to St. Paul's Cathedral. We reflect this practice in the generally smoky atmosphere and the blackened colour of buildings.

We have also incorporated into the model weather conditions representative of London on November 5, 1622 on the Julian calendar (November 15 on the Gregorian calendar). We know the weather was bad that day, since the sermon we are using was actually "Preached in the Church . . . by reason of the weather." But, according to the website weatherspark.com, which gives weather data for every day of the year in London, the weather in mid-November is often damp, dark, and chilly, even when it isn't actually raining. The days are short and nights are long: the sun rises at 7:20 am and sets at 4:12 pm.  Between 10:00 am and noon, the hours of a Paul's Cross sermon, the sun rises from about 18 degrees of elevation above the horizon to 20 degrees of elevation, casting, even at noon, a long shadow across the Churchyard. The temperature typically varies from 44°F (6.7°C) to 50°F (10°C), and the relative humidity typically ranges from 76% (humid) to 90% (very humid). The median cloud cover is 87% (mostly cloudy) and does not vary substantially over the course of the day. The sky is clear or mostly clear only 24% of the time, partly cloudy 15% of the time, and mostly cloudy or overcast 56% of the time. There is a 70% chance that precipitation will be observed at some point during the day and that it will be in the form of moderate rain (a 71% chance of that). The wind speed is 8 mph (gentle breeze) on average.

When the angle of the sun and other weather conditions are incorporated into the visual model, rendering images of the model in Photoshop, we get the look of Figures 1 and 21, a damp, chilly, overcast day, with the sun low on the horizon casting long shadows across the Churchyard. There is a light breeze. Because of the chill in the air, people in Paul's Churchyard have bundled up against the cold. Those in the surrounding buildings have built fires; plumes of smoke rising from their chimneys bend with the breeze. The overall atmosphere is grey, dark, gloomy, the sun pale if it peeks through the clouds at all. Surviving Paul's Cross sermons date from every month of the calendar year. One of the most striking things about the phenomenon of the Paul's Cross sermon is that it persisted as a central feature of London life for nearly a hundred years, through the fall and winter as well as through the spring and summer.

Figure 21: Paul's Cross in Paul's Churchyard, looking east. From the Visual Model constructed by Joshua Stephens and rendered by Jordan Grey.

We have thus been able to create a visual model of St. Paul's Cathedral and Paul's Churchyard that incorporates a wide range of data about the look and feel of St. Paul's Churchyard on November 5, 1622. Nevertheless, this model has its limits. We have only been able to include representative sounds (Smith 1999). Nor have we been able to include the smell, surely a ripe and heady mixture of rotten cabbage, unwashed bodies, and horse manure (Cockayne 2007; Jenner 2011). Within these limits, however, this model brings the surviving visual record of Paul's Churchyard together with contemporary archaeological data as well as data about weather and climate conditions. The archaeologists and surveyors whose work we have incorporated into this model to establish the scale and dimensions of individual structures also provide us the information we need to establish with high accuracy the physical positions of these structures in relationship to each other. We have thus been able to create a visual model that possesses what Boren calls the "accurate geometrical and material data" required to produce an accurate acoustic model.

The acoustic model

The acoustic model developed by Ben Markham and Matt Azevedo (2013), acoustic engineers at Acentech, Inc. in Cambridge, MA, may well be the most consistently accurate feature of the whole project. For an accurate model of the acoustic properties of a space, one needs only the dimensions of the space, the arrangement of basic geometric forms in the space, and the materials from which objects in the space are made. Sound, when it encounters an object, is either (and to some degree) absorbed, reflected, or dispersed. The acoustic modelling program incorporates the absorptive, reflective, or dispersive qualities of the forms and materials in the space–in this case, the space of Paul's Churchyard–into the acoustic model it produces. When one introduces sounds into the model, the output from the model enables us to experience the source sound as though we were hearing it from inside the acoustic space. To experience this effect, of course, the sound to be heard through the virtual acoustic model must be recorded in an anechoic recording studio which yields only pure source sound, without any of the ambient qualities of the room in which the sound was recorded. We delivered a simplified version (Figure 22) of our visual model, together with a list of the materials out of which the objects would have been constructed in 1622–stone, wood, plaster, brick, dirt, and the bodies and clothing of the congregation–to our acoustic engineers Markham and Azevedo who then imported it into the CATT-Acoustic auralisation program.

Figure 22: Wireframe Image of the Acoustic Model Paul's Churchyard, the Cross Yard. Wireframe from the Visual Model constructed by Joshua Stephens.

The resulting acoustic model of the space of Paul's Churchyard has brought surprises with it, for, as Markham has pointed out, the relationship of the buildings in Paul's Churchyard to the Paul's Cross preaching station created a space that amplified the sound of the preacher's voice (n.d.). Markham notes that in an open field sound "decays by 6 decibels per doubling of distance from the source" (n.d.). Hence, "if John Donne's speech was 75 dB at the ears of a listener a mere three feet away," then a "listener ninety-six feet from the speaker would receive only 45 dB of sound from speech in an open field, and would have struggled to hear" (Markham n.d.). But because Paul's Churchyard "was no open field" but a "courtyard surrounded with buildings," a listener could be as far from the speaker as "140 feet or more from the speaker" and still hear the words of the speaker, "thanks to sound reflections from the nearby buildings" (Markham n.d.).  Interestingly, this effect would have been increased as the years went by. Paul's Cross had been a station for outdoor sermons in Paul's Churchyard since the late fifteenth-century, well before the development of the English book trade over the next century led to the building of the structures that housed the booksellers' shops to the northeast of Paul's Churchyard (see Azevedo, Markham, Wall 2013).

Markham and Azevedo's acoustic model now functions like Stephens' visual model, allowing us to gather what we know or can surmise, in this case about sounds of preaching, about the sound of spoken English, and about the ambient sounds of urban life in early modern London (in our model, the sounds of the bells, horses, dogs, and birds) in a specific place and over a specific period of time–in this case on the morning of November 5, 1622, between the hours of 10:00 a.m. and noon. We can now evaluate that knowledge in terms of what can be repeated and what is lost, and to integrate this knowledge into a coherent approximation of Donne's Gunpowder Day sermon for November 5, 1622. Through the Virtual Paul's Cross Project, we can hear the full text of a performance of John Donne's sermon experienced as a public event unfolding in real time delivered in early modern London pronunciation, in the acoustic space in which Paul's Cross sermons were delivered, in the presence of large crowds of people, and in a space filled with the ambient noises of birds, horses, dogs, church bells, and the sounds of the crowd. This project encourages us to explore the audibility of the sermon by allowing us to hear the preacher's voice at different distances from the preacher and with different sizes of crowd, ranging from about 250 to 5,000 people. This project also allows us to explore questions about style of delivery, the congregation's response to different kinds of passages in the sermon, and the preacher's use of the time of delivery in making his points.

Getting to this point has required for the acoustic model, as it did for the visual model, a gathering and assessment of existing data and a willingness to accept the value of representational approximations when necessary. This is especially true for the recording of the sermon featured in the model, for of course we have no recording of John Donne's voice to measure our efforts against. Here, the accuracy we are dealing with is clearly at a very general, rather than a specific, level intended far more to raise possibilities and introduce new questions than it is to provide final answers. For me, especially, there has been a shift in thinking about the texts of Donne's sermons that come down to us, from their being textual objects of study to the recognition that they are at best memorial reconstructions of the actual, delivered text. This is the result of being able to experience the text of Donne's sermon not as a theological essay to be experienced with the eye, which can proceed word by word, or can skip over, move around, go back, pause, and resume its activity, but as a performance that unfolds in real time, word by word, an event of the ear (Wall 2007). That in itself represents a fundamental shift in our thinking about the early modern sermon, some compensation, perhaps, for all the aspects of Donne's preaching that are still beyond our reach (McCullough, Adlington, and Rhatigan 2011).

The process of getting to this place has involved careful review of what we do know and what we can make of it. To review this process, let's begin with the ambient noise. During an event of at least two hours' duration in the open air in the middle of a city of 200,000 people, there definitely would have been ambient noise randomly occurring from a variety of sources and in a variety of intensities. Rather than try to take a full inventory of all possible kinds of ambient noise and try to assess the degree of impact for each of these sounds, we have chosen to represent this noise by concentrating on three sources of noise–the birds, the horses, and the dogs (Figures 2, 19, and 23) - all documented as sound sources in Paul's Churchyard through their appearance in John Gipkin's painting, all occurring randomly during the delivery of Donne's sermon. Creaturely sounds were obtained from the creative-commons licensed sound library website freesound.org. These sounds, again, do not represent a literal reconstruction of the ambient noise audible in Paul's Churchyard in 1622, but serve as representatives of that ambient noise, reminders that when we consider the experience of the Paul's Cross sermon we need to include in our thinking the presence of such sounds as part of that experience.

Figure 23: John Gipkin. Detail of Paul's Cross. 1616. Painting. Image courtesy of the Bridgeman Art Library, New York, and the Society of Antiquaries, London.

The world of early modern London–noisy though it was–was still quieter than our post-industrial world. Markham suggests that:

instead of [today's] steady mid-day background noise level of perhaps forty-five to fifty decibels or more, the steady-state ambient noise of the day was perhaps closer to thirty-five decibels at the critical frequencies when people were quiet and listening.  In other words, background sound in Paul's Cross in 1622 might have been more than ten decibels quieter than it is now–less than half as loud. (n.d.)

We have added to the random sounds of birds, horses, and dogs, the sound of a bell occurring regularly to tell the time. This bell is not like the early modern bells we have known about for some time, the bells that rung in peals and changes and also rung as a form of public notice such as the bell that Donne describes in his Devotions Upon Emergent Occasions as ringing to announce the death of a member of the cathedral's congregation. The practice of change ringing in English churches developed in the seventeenth-century; one of the earliest records of the practice is from 1612 at Lincoln Cathedral. One assumes that St. Paul's would be a leader in the practice. Nevertheless, we have not included the sound of these bells because their ringing was discretionary and unlikely to take place during the preaching of a Paul's Cross sermon.

Thanks to Tiffany Stern's recent research, we know that part of the soundscape of early modern London was the sound of a bell, activated by a clock mechanism, that rang on the quarter-hours and hours of each passing day (Stern 2012). English cathedrals in the sixteenth-century are known to have mechanical clocks that recorded time by striking bells; one of these survives at Salisbury Cathedral (Maltin and Dannemann 2013). The clock at Salisbury had no face; the ringing of a bell was its sole means of communicating the time. The sounding of this clock/bell was not discretionary. It could not be stopped and started again without disrupting its accuracy as a timepiece. Stern's research into the ringing of church bells in early modern London has taught us that St. Paul's Cathedral had such a clock and that the clock at St. Paul's Cathedral rang on the quarter hour as well as on the hour, marking the passage of time in fifteen-minute increments. We have incorporated into the acoustic model the sound of a struck bell ringing on the hour and the quarter hour.

This addition to the soundscape of the acoustic model has raised, however, the question of how the preacher dealt with the bells. He either talked over them (or tried to) or paused when they rang. The bells must have been loud enough to be heard over a good bit of London; that would have been part of the cathedral's role as center of city life, as focus of the community's attention, as marker, and organiser, of the passage of time in human affairs. It seems unlikely, therefore, that the preacher could have talked over them. So, he paused. The pause at fifteen minutes past the hour would have been fairly brief, but it would have gotten longer with each passing quarter hour. The pause on the hour, at 11:00 a.m., would have been of a significant length, about a minute.

The length of the pause required to prevent the sound of his voice, hence the content of his sermon, from being obscured by the sound of the bell, however, means that Donne had options about how to deal with this inevitable but also predictable interruption. The sound of the bell would have been an interruption, coming at some otherwise random moment in the unfolding of the sermon. It would have been a pause to be dealt with in some fashion, otherwise Donne's congregation's attention could wander, perhaps causing them to lose track of where the preacher was in the unfolding of his argument when the interruption began. Donne could try to ignore this pause, resuming his delivery of content when the interruption was over where he left off when the interruption began; he could also do something when beginning again to refocus the congregation's attention, calling them back to the task at hand and perhaps reminding them what was being said at the beginning of the tolling.

Traces of such accommodations of time's passage in the texts of Donne's sermons are well worth looking for, as are signs of yet another possibility, that the predictable, and hence anticipatable, interruption by the bell every fifteen minutes could be regarded by Donne not just as an annoyance to be accommodated but an opportunity to be incorporated into his delivery. With planning, the tolling of the bell could be an opportunity for Donne to complete a thought, then treat the bell as an underlining of that point, then treat the pause created by the bell as a chance to catch his breath, perhaps take a sip of the wine we are told preacher's kept in the pulpit at Paul's Cross, and then begin the next section of his sermon afresh.

The purpose of the hourglass was, after all, to make the preacher accountable to his congregation for the length of his sermons; it served as a trace of the passage of the time allocated to a sermon visible both to the preacher and to the congregation. Robert Chambers, in his Book of Days, tells us that the "Hour-glass that hangeth by the pulpit where the preacher doth make a sermon" was there so that the preacher "may know how the hour passeth away" (n.d.). Congregations assumed an unspoken contract between them and the preacher, when "having named the text [he] turned up the glass"; "if the sermon did not last till the sand had run down, it was said by the congregation that the preacher was lazy; but if, on the other hand, he exceeded this limit, they would yawn and stretch themselves till he had finished" (Chambers n.d.).

For Donne to be able to incorporate the cathedral clock's record of the passage of time into the structure of his sermon, he would have needed to be very good at keeping track of time while preaching. He would need to be able to interpret the level of sand in the hourglass that tracked the progress of his allotted time sufficiently well to anticipate the approaching end of a fifteen-minute time period, conclude a section of his sermon just in time for the bell, then be ready to launch into the next section as the echo of the bell's tolling faded away. We have evidence from other sermons that Donne was in fact, able to organise the structure of his sermon around the movement of time so that the content of his sermon and the time of its delivery coincide in ways that reinforce the experience of his preaching.

In his sermons delivered at the Chapel Royal on February 11, 1627, and again on February 29, 1628, for example, Donne planned his delivery so that he ran out of his allotted time just before he ran out of content (Wall 2007). In 1627, describing the everlastingness of God's justice, Donne asserts that the news is grim: "as long as his eternity lasts . . . God shall never see that soul, whom he hath accurst, delivered from that curse, nor eased in it." And just at that moment, he punctuates his point by drawing attention to the hourglass that has been marking the passage of time during the course of his preaching and informing his congregation that he is out of time:

But we are now in the work of an houre, and no more. If there be a minute of sand left, (There is not) If there be a minute of patience left, heare me say, This minute that is left, is that eternitie which we speake of; upon this minute dependeth that eternity. (Donne 1953-62, 7: 368)

Donne says, in effect, that he is now living on borrowed time as a preacher, able to continue only in hope that there is left for him "a minute of" his congregation's "patience." This precarious position is, however, one he shares with his congregation, for they, too may be out of time as well, for "this minute makes up your Century, your hundred yeares, your eternity, because it may be your last minute" (1953-62, 7: 368; See also Wall 2007). Donne's congregation, presumably, agreed to an extension of Donne's contract, for the text of this sermon continues for three more pages, and the moment of patience he has successfully negotiated from his congregation becomes an opportunity for Donne to bring the sermon to a dramatic–and now positive, and hopeful, and celebratory–close. "[B]e this your Peace," Donne says, "[T]o know [that] God hath laid the whole curse belonging to us upon him. . . this glorious Sonne of God . . . that hangs upon the Crosse" (1953-62 7: 369).

In 1628, he again used the same device. Approaching the end of a sermon on the relationship between sleep and death, Donne planned his conclusion, once more, to involve running out of time, out of sand, and needing to ask for an extension of his contracted allotment of time:

Now of this dying Man, that dies in Christ, that dies the Death of the Rghteous, that embraces Death as a Sleepe, must wee give you a Picture too. There is not a minute left to do it; not a minutes sand; Is there a minutes patience? (1953-62, 8: 190)

In both these cases Donne was able to incorporate the mutually agreed-upon conventions of time for the duration of a sermon into his sermon's argument, entwining his theological argument into his interactions with his congregation around the question of time, sand, and the passage of time. He shifts the news from bad to good only after asking for and receiving from them the gift of more time, putting himself at their mercy even as they are at God's mercy, thus demonstrating Donne's skill at keeping time while preaching, since the whole point of both these sermons depends on Donne's ability to bring his hour-long argument to just the right place at the precise moment when the sand has run through the glass. But the effectiveness of his argument also depends on the responsiveness of his congregation, on their willingness after an hour of preaching to grant him mercy, to extend to him more time to shift from bad news to good. In other words, the impact of this sermon is dependent not just on Donne's performance, delivering the word to a passive audience, but to active collaboration between preacher and congregation.

The evidence of Donne's practice in these two sermons suggests that Donne may well have been able to track the passage of time, presumably by monitoring the movement of sand through the hourglass standing next to him in the pulpit (Figure 24) with sufficient accuracy to work the sound of the clock bell striking the quarter-hours and the hours into the progress of his sermon. Donne would have been aided in doing this by the fact that he was preaching this two-hour-long sermon from notes rather than from a fully-written-out text and was able, therefore, to adjust the time of delivery to fit the time available to him. Either by reducing or expanding the number of words he used, or speeding up or slowing down the pace of his delivery, he could bring a major point to closure just in time for the bell to ring, and to reinforce his point, then move on to his next point when the tolling of the bell had passed.

Figure 24: John Gipkin. Detail of Paul's Cross. 1616. Painting. Image courtesy of the Bridgeman Art Library, New York, and the Society of Antiquaries, London.

I concluded that Donne's management of time and the clock bell was worth exploring only after noting repeatedly, while incorporating the sound of the clock bell into the progress of Donne's sermon, the number of times that transitional moments in Donne's sermon for Gunpowder Day 1622 came close to the beginning or end of fifteen minute time segments. To enable this exploration, the full-sermon recordings on the Virtual Paul's Cross website have Donne fit the organisational structure of his sermon within the fifteen minute intervals established by the ringing of the cathedral clock bell. The alternative would be, of course, that the bell came as an interruption in the exposition of his argument and a distraction for his congregation.

A word about the process of developing the audio for the VPCP website is in order. This has been an unfolding process. Because of time constraints on the NEH grant, this recording (described more fully below) had to be made early in the funding process, before we had finished the acoustic model and had become aware of the acoustic properties of Paul's Churchyard. Ben Crystal, the professional actor performing Donne's sermon, had studied contemporary accounts of Donne's preaching style and had been told that his ability to be heard in a space the size of Paul's Churchyard was a major concern of the project. The time his performance took was not a consideration during the recording session, yet it in fact comes in at just a few minutes over two hours. Subsequently, Markham and Azevedo, our acoustic engineers, told me that Crystal's pacing is almost perfect for taking maximum advantage of the acoustic properties of Paul's Churchyard. Then, when I set out to incorporate the sound of the clock bell, I decided not to allow the sound of the bell to obscure the delivery of the sermon, so I looked for the fifteen-minute intervals in the sermon to use as stopping points to accommodate the sound of the bell. The more I looked at the relationship between the rhetorical structure of the sermon and the passage of time, the more I saw how close to fifteen-minute intervals the major breaks in Donne's organisation happened to fall. In the recordings on the website, we have bent time to accommodate this interpretation of what Donne is doing in this sermon, but I am absolutely convinced that if we were to record the sermon over again, we could make the rhetorical structure of the sermon fit neatly into the fifteen-minute intervals marked by the sound of the clock bell and make the whole sermon come in right at two total hours, with minimal adjustments in the pace of Crystal's delivery. This is the kind of evidence I am finding to be increasingly convincing as we try to recover the conditions of performance in the early modern period.

Other sounds essential to our acoustic model include the sounds of the crowd assembled to hear the sermon and the sounds of the preacher himself. We know that some of the folks in attendance sat or stood quietly, even taking notes (Hunt 2011). Yet Donne himself says that sometimes congregations take a vocal role in the process of the sermon, describing "those often periodicall murmurings, and noises, which you make, when the Preacher concludeth any point; for those impertinent Interjections swallowe up one quarter of his houre" (Donne 1953-62, 10:132). Here, Donne says, the behaviour of congregations in early modern London recalls the behaviour of congregations in the Patristic age, when, Donne says,

all that had been formerly used in Theaters, Acclamations and Plaudites, was brought into the Church, and not only the vulgar people, but learned hearers were as loud, and as profuse in those declarations, those vocal acclamations, and those plaudites in the passages, and transitions, in Sermons, as ever they had been at the Stage, or other recitations of their Poets, or Orators. . . . (Donne 1953-62, 10: 132).

Donne's account of crowd behaviour here assumes an active and vocally responsive congregation for his sermons, precisely the kind of crowd behaviour we have come to understand characterised crowd behaviour in the theater of Donne's day. As Andrew Gurr and others have taught us, early modern audiences–and those who performed for them–were accustomed to verbal interaction during the course of the event. We have not tried to model such extensive–or potentially disruptive–crowd response, but have included in the behaviour of the crowd enough verbal engagement with the preacher's delivery to suggest the effect of such response on the overall experience of the sermon, for, as Donne says in the same sermon, sometimes crowd response to a sermon becomes an indication of the quality of the sermon for members of the congregation who are unable to hear the preacher: "many that were not within distance of hearing the Sermon, will give a censure upon it, according to the frequencie, or paucitie of these acclamations" (Donne, 133-134). The actual crowd sounds were recorded by members of my university's faculty and graduate students in linguistics, on the theory that they could make mumbled sounds of speech using early modern London English pronunciation. These sounds have been integrated into the acoustic model so that as the preacher's voice increases in energy and volume, theirs does too, again not as a reproduction of actual events but as a reminder that there was a congregation, and a vocally active one, for these sermons.

Finally, there is the matter of the sound of the preacher's voice. This aspect of the Project has so far drawn the most extensive commentary, especially from listeners who find the sound and style of delivery not to comply with their assumptions about how Donne would have sounded in his preaching. Some background is therefore in order. The first thing that people notice is that the accent employed in the recording is an accent unfamiliar to contemporary speakers of English, regardless of their home country. The second is that the pace of delivery is deliberate, with frequent pauses between words; some listeners have called it annoyingly slow, even monotonous, not the style of delivery they have imagined for Donne's preaching. Both of these features are the result of deliberate choices made either for the sake of authenticity or as a result of the Project's goals and objectives.

The distinctive accent employed in this recording is a result of the fact that the actor performing the sermon is using a script of Donne's sermon that incorporates early modern London pronunciation prepared by the linguist David Crystal. The decision to use an original pronunciation script was prompted by the fact that Donne was a Londoner by birth (Crystal 2011; Meyer 2013). Actor Ben Crystal, who specialises in early modern English pronunciation, recorded the sermon in an anechoic recording studio at the University of Salford, in Manchester, UK, a step necessary to prevent introduction of modern ambient noise or the acoustic properties of the recording studio into the model of seventeenth-century acoustic space. Crystal delivered his performance after considering two kinds of information about this project, the first having to do with general principles of delivery with special regard for audibility in large reverberant spaces and the second having to do with what we do know about Donne's preaching style. In regard to the former of these considerations, Crystal was told that evaluating the audibility of a preacher delivering a sermon without amplification in a large open space was a major concern of this project. He thus paid special attention to the pacing of his delivery, knowing that his performance would be presented in a space with significant reverberation and that his performance would be accompanied by the noise of a large crowd. The sermon was recorded from multiple angles, and Crystal was encouraged to visualise the crowd around him and direct his speech to various people, particularly the nobility who would have been seated in the sheltered seats along the cathedral wall. Crystal delivered the sermon in a manner consistent with a practiced orator delivering a speech to a large outdoor crowd: the voice is strong, the cadence measured, a reasonable and appropriate strategy for improving intelligibility. Had he spoken more quickly, the effect of the reverberation in this space would have run together the sounds of his speech, mudding his articulation and making aural comprehension much more difficult.

This is a case, I believe, in which the recreation of original performance conditions in Paul's Churchyard has given us access to Donne's original style of delivery in a particular space. Listeners' surprise at Crystal's choice of pacing is an artefact of our lack of familiarity with the sound of the unamplified human voice in large open spaces; amplified sound, together with the effects of acoustic engineers' ability to shape sound delivered through speakers, means that we have grown accustomed to hearing almost everyone, in any setting, speaking at a conversational pace. I would expect Donne to use a faster pace when audibility was not so much of a concern, in a smaller venue or before a smaller crowd, such as in a parish church or the Chapel Royal, for example, or the choir of St. Paul's Cathedral, where he preached most frequently in the 1620's. Nevertheless, because of the response to this feature of Crystal's performance, and additional grant funding permitting, we will record him delivering the same sermon at a faster, more conversational pace and see how that would sound played through our acoustic model of Paul's Churchyard.

In regard to the second of these considerations, Crystal was asked to remember that a sermon is not a theological essay to be read in one's study or a formal lecture to be attended to in silence, but a performance intended to entertain as well as to inform, to move his hearers emotionally as well as to educate them about the content of their faith. Crystal was given a full account of Donne's own comments on preaching: that, for example, the effective preacher sought to engage his hearers, to change his hearers, to bring them–through both cognitive and emotional means–to amend their lives in directions set out by Donne in the sermon. Donne once described the performance of an effective preacher in terms of a coordinated effort of body, feeling, and ideas, uniting "matter and manner," the quality of the voice ("pleasant") and personal manner ("acceptably, seasonably, with a spiritual delight"), with "a holy delight," toward the goal of "profit" for his congregation (Donne 1953-62, 2: 167). 

Crystal was also given every available account of how Donne's contemporaries described his preaching style, noting especially their emphasis on Donne's wit, his eloquence, and his capacity to express and arouse feeling, to elevate, to captivate, and to motivate. These texts, chiefly elegies written shortly after Donne's death and gathered together for the 1633 edition of Donne's poems, present their own interpretive challenges because they employ a formal and often stylised tradition of compliment that does not translate easily into the specific details of an actual performance. Nevertheless they emphasise a number of common elements in Donne's preaching style that can form the basis for a recreation of that style. Donne's sermons, his contemporaries claimed, were delivered in an empathetic, emotionally expressive, responsive, evocative speaking style, or, better, range of styles, always aspiring to intimacy and engagement, sometimes confrontational, sometimes laudatory, always lively, personable, and connecting with his congregation. Izaak Walton, in his biography of Donne, especially remarks on Donne's capacity to communicate with the minds and feelings of his listeners, to teach and to move them:

carrying some, as St. Paul was, to Heaven in holy raptures, and inticing others by a sacred art and Courtship to amend their lives; here picturing a vice so as to make it ugly to those that practised it; and a vertue so, as to make it be loved even by those that lov'd it not, and all this with a most particular grace and an unexpressable addition of comelinesse. (1658, 47-48)

As an example of Donne's ability to evoke emotional responses in his hearers, Walton cites Donne's sermon preached shortly after his wife's death, in which, Walton says, Donne's performance "did so work upon the affections of his hearers, as melted and moulded them into a companionable sadness" (1658, 55).

Comments by others of Donne's contemporaries expand on Walton's basic claims. I'll cite a few examples (all of these are collected here: http://vpcp.chass.ncsu.edu/donne-preaching/). Jasper Mayne, writing while a student at Christ Church, Oxford, says that Donne, with his words, "could [so] charme [his] audience,/ That at [his] sermons, eare was all our sense" and with his "looke, and hand" and "speaking action" could give them "More Sermon, then some teachers used to say" (1912, 55-60). John Chudleigh wrote that Donne "did not banish" his wit when he took orders, "but transplanted it;/Taught it his place and use, and brought it home/To Pietie, which it doth best become" (1912, 14-16). Daniel Darneley's Latin elegy for Donne, here translated by my colleague Zola Packman, says that he employed "his eyes, his hands, his face" (1912, 41) and held "hearts, eyes, and men" (1912, 45) while he "reveal[ed] to the commoners/ Mysteries not earlier entrusted to them and not yet understood" (1912, 47-48) and "with changed manner and form of speaking/ Treated of sorrowful things/ . . . . /Then you would see them all give a groan" (1912, 50-53).

These accounts of Donne's style of delivery support a characterisation of Donne's preaching style as multi-vocal, varying in mode of delivery from section to section of the sermon, if not moment to moment of his delivery, depending on the content of each passage and its relationship to the overall structure and argument of the sermon as a whole Donne could, we gather from his contemporaries' accounts, be analytic, discursive, informal, witty, joking, declamatory, and affective, depending on what part of the sermon he happened to be in at any moment of delivery, using a style of verbal delivery appropriate for the kind of material he was covering at that moment in the section of the sermon he happened to be in and appropriate for the kind of effect he hoped to have on the audience at that particular point in the sermon. Of course, when we imagine Donne's preaching style, we must consider, along with voice, other aspects of performance, including gesture and posture as elements of Donne's performance, as Brent Nelson appropriately points out (2012). When all these elements of performance are taken together, we must imagine Donne' styles of delivery as enabling him to connect with his congregations on emotional as well as cognitive levels.

We also know that Donne preached for an audience well-experienced in sermon-going and with a high regard for the quality of performance. Holding their attention must have been a major concern for one performing the roles of priest, prophet, spiritual guide, interpreter, model and enabler of transformation. The occasion of the early modern sermon had the potential, Donne tells us, to change its participants, as well as to provide an entertaining way to spend time on a Sunday morning, create an occasion for a large social gathering with one's neighbors, or advance a clerical career (Hunt 2011, 60-116).  This occasion could also provide–from a theological perspective–an occasion that could change lives, advance the general welfare, promote social cohesion (especially promote support for the monarchy), and open the way to eternal life.

The fact that Donne did not read his sermons as one would a formal lecture, but performed them as one would a public oration, or, perhaps better, as one would perform a character or a role in a play, reminds us that Donne's preaching took place within the context of a tradition of practice, informed by manuals of oratory, that define the role of the preacher as a social and professional role in relationship to other roles played by members of his congregation. As I noted earlier, his sermons were performed from notes that guided his delivery, so, while the structure of the sermon was worked out in advance, the actual words were improvised in process, responsive to external elements like the regular sounding of the clock bell or the irregular responses of the congregation to this or that point, or gesture, or distinctive passage.

The Virtual Paul's Cross Project permits the user to sample Donne's sermon from eight different locations in the courtyard and with four different sizes of crowd. Especially interesting is to compare the sound of the preacher's voice when one listens from Position Two, the dignitaries' box in the sermon house to the preacher's left, with other positions on the ground level of the Churchyard. The preacher's voice is much clearer from the dignitaries' box, although more clearly reverberant, than it is from other listening positions. Clearly here as in other circumstances, rank has its privileges.

The account of Donne's preaching one can hear through the Virtual Paul's Cross Project does not, however, pretend to offer a definitive recreation of Donne's Gunpowder Day sermon for 1622. Far too much about that event is irretrievably lost to us to pretend to do so. It has, however, enabled us to integrate all the information we have and all we might infer about the place, the space, and the physical circumstances of a Paul's Cross sermon, including the light and sound, together with the words of Donne's text, the manner of his preaching style, and the behaviour of his congregation into a single experiential and interactive model. The goal of this project has been to make available for study our assumptions about the conditions of sermon delivery and reception, reminding us that these sermons were, originally, performances during which preacher and congregation interacted to shape their mutual experience of the occasion and of the sermon itself. We can then test the consequences of our assumptions as they are realisd in the visual model and played out in the acoustic model, always being aware that we can revise the model as we develop our understanding, incorporating new research into an unfolding process of development. Another aspect of this project has been to assemble what we do know and to explore this knowledge in progress, to reconsider what was involved in actually staging a Paul's Cross sermon, including such considerations as the need to gain and sustain a congregation's attention, the need to accommodate into the performance the realities of ambient noise, and the need to deal with problems of audibility and crowd response. This process has, over time, opened up new areas of inquiry, raising questions, for example, about how the preacher dealt with the sound of the bell tolling the hour and the quarter-hour, how the preacher sought congregational engagement, how he sustained the congregation's attention over the two-hour duration of his sermon, and how we might come to a fuller sense of his congregation's responses to his preaching.

The Virtual Paul's Cross Project will ultimately constitute Stage I of a larger project that will extend our effort to recreate worship and preaching in early modern London. Stage II of the project will revisit the original project, exploring aspects of the recreation, such as the question of Donne's pace of delivery, which we discovered in the process of completing Stage I but were unable to explore because of financial and other constraints. Stage II will also complete our visual model of Paul's Churchyard and the interior of St. Paul's Cathedral. It will then create an acoustic model of the acoustic properties of the Cathedral's interior so that we may restage a full liturgical day, bringing to life the liturgies of Matins, the Great Litany, Holy Communion, and Evensong from the Book of Common Prayer of 1604 in both their daily and their festival forms. We will also include sermons for both morning and afternoon. Stage II will also involve an even larger collaborative effort, this time bringing into the project musicians and singers. The daily services at St. Paul's were sung by a choir of men and boys augmented by the sound of the Cathedral's organ and by wind instruments on festivals and other special occasions in the church year. This phase of the project will also draw on the large body of music that survives from this period composed by Adrian Batten, John Barnard, and other musicians on the staff of St. Paul's. We hope to be able to hear this music again, for the first time in 400 years, in the space for which it was originally composed.