March 31, 2026: Reproducible Methods at Building s

By Corey Sattes and Derek Wheeler

March 31, 2026

Paper presented at the Middle Atlantic Archaeological Conference in Gettysburg, VA, March 11-15.

This paper is a part of an ongoing project to establish methods for identifying chronological phases at archaeological sites across the Monticello plantation once owned by Thomas Jefferson. A major goal – and challenge – is to apply a consistent analytical method across all projects, legacy or non-legacy, plowed or stratified. We have begun this work, so far testing the methods discussed here on two other sites: O'Connor and Neiman 2024; Wheeler and Sattes 2024. 

Looking at the forest through the trees, we aim for this work to establish site phases that can allow for inter-site comparisons. This will help us better understand change over time and space – something that becomes lost at a site-level analysis that in effect time averages sites.

We want to emphasize that this all builds off the amazing work by current and past colleagues at Monticello and the Digital Archaeological Archive of Comparative Slavery (Hill 2002a, 2002b, 2003; Neiman et al. 2001). This chronological phasing was established by Fraser Neiman and colleagues back in 2003 (Neiman et al. 2003; see also Bates et al. 2020). And the sites we discuss have undergone excellent analyses since their respective excavations (Kelso 1993, 1997; Kelso and Sanford 1983; Sanford 1995). So, this paper is a continuation of these past studies.

To organize our thoughts for this discussion, we will first walk through the chronological phasing method. That first step is an iterative process with understanding site formation. Finally, we peek at some teasers for how these methods can enhance our analyses. 

To set the stage…Monticello, located just east of Charlottesville, Virginia, was once home to Thomas Jefferson, his family, and hundreds of free and enslaved individuals. From 1770 until Jefferson’s death in 1826, the plantation was divided into four quarter farms (Figure 1). The site we will discuss today, Building s, was located at the Monticello home farm quarter, along Mulberry Row, a thousand-foot-long road just south of the mansion house (Figure 2). 

Once a bustling street of domestic and industrial activity where enslaved and free workers lived and labored, Mulberry Row was heavily excavated throughout the 1980s, informed by a plat Jefferson drew in 1796 as part of an insurance policy he took out with the Mutual Assurance Company of Richmond (Jefferson 1796) (Figure 3).

For the three cabins on the far-east side, Jefferson instructs his overseer Clarkson in 1792 to manage their construction (Jefferson 1792). Jefferson also instructs his son-in-law in 1793 to move enslaved workers from Building E into Buildings r and s (Jefferson 1793). 

Buildings r, s, and t are a trio of likely identical cabins, meant to replace an older underlying barracks-style dwelling (Figure 4). Therefore, a major caveat in this paper is that Building s really shouldn’t stand alone in an analysis. For the sake of time, we started with just Building s and plan to incorporate the other two cabins with future research.

Director William Kelso excavated Building s between 1983 and 1984, opening forty-two units, including several downslope towards the garden terrace (Figures 5-7). Excavators identified the partial remains of a stone foundation and hearth. They also uncovered a wood-lined subfloor pit (Figure 7a) and packed earth floor. Additionally, they noted an “Occupation Zone” within and exterior to the Building s footprint. 

Also of note are a series of postholes from an 1809 paling fence constructed related to the vegetable garden terrace (Figure 8a). Finally, the remains of two brick piers from a later structure overlie part of the Building s occupation zone (Figure 8b).

As many of us know, legacy excavation records can be challenging. It is important to balance a critical eye with the need to trust in the identifications made by the excavators. With that in mind, we aim to best identify the timeline for how these respective materials were deposited and the people they represent.

Chronological Phasing Methods

We won't go through all of the steps of how we phase sites at Monticello. Check out the 2020 chapter by Bates, Galle, and Neiman. In short, building these site chronologies relies on a frequency seriation model of ware types, adjusted mean ceramic dates, and correspondence analyses to identify groups of contexts with similar ceramics.

Correspondence analysis is a statistical method that calculates dimension scores for both our contexts and ware types. The dimensions represent variables that cause differences in our data. The first dimension accounts for the most variation, then the second dimension, and so on. The closer the dimension scores between contexts means that the respective ceramic assemblages are more similar than contexts with scores further apart. 

The two plots in Figure 10 show the Dimension 1 scores plotted against the Dimension 2 scores for contexts and ware types. The red crosshairs on each plot show you the estimated “typical” assemblage. We can examine which ware types significantly impact contexts based on their spatial association. 

Usually in archaeological assemblages, the primary dimension of change is time. This is true here, as we see older ware types on the left and younger ware types on the right. 

To independently check this, we plot those scores by each context’s Blue MCD (Figure 11). This is a mean ceramic date representing a weighted average based on the frequency of recovered ware types, and it also more heavily weights wares with narrower date ranges (Bates et al. 2020:66). So, the linear regression here supports that Dimension 1 represents time, and we use these scores to assign groupings of contexts that could represent different occupational phases.

The histogram in Figure 12 plots the number of occurrences of each Dimension 1 score across Building s – in short, this theoretically shows us the peaks of different chronological groups. We identified five phases– and it is worth saying that we did a series of splitting and lumping to determine the most reasonable groupings.

From here, we assign contexts to those specific phases based on their Dimension 1 scores. Figure 13 is the same plot in Figure 11 of context Dimension 1 scores by their respective Blue MCDs, now demarcated by phase assignments and relative chronology. A TPQ90 (terminus post quem) is based on the beginning manufacturing date of the latest occurring ceramic and calculates from that the 90th percentile – thereby removing the accidental inclusion of later ceramics from excavation biases. Based on documentary records, Phase 1 best represents the contexts from the occupation of Building s.

Stratigraphic Groups

We now focus on the chronological span of stratigraphic groups, a term we use to group contexts identified by the excavators as part of the same deposit. 

There are a handful of early stratigraphic groups that appear relatively homogenous in Phase 1, such as those interpreted as the Building s occupation zone, the subfloor pit inside the structure, the terrace fill, and deposits from deconstructing Building s (see Figure 14).

But then there are a handful that have contexts with a wide range of Blue MCDs, suggesting that although these deposits may have been lithologically distinct to excavators, their ceramic assemblages vary by context. Most of these layers are associated with destruction layers and one associated with the parking lot fill (see Figure 15).

Issue with Time Averaging

These observations bring into question how much time averaging is impacting our data here, and we need to ask whether that time averaging is: 

1. Happening on the stratigraphic group level, in which case individual contexts have tight time ranges but stratigraphic groups contain contexts from a variety of periods OR
2. Happening at the context level, in which case it is the context itself that has a chronologically wide range of ceramics. 

Most likely, it’s a combination of these two scenarios. As this is important to our understanding of the site formation processes, we want to investigate this, particularly figuring out how to view this at the context level. 

We first look at a battleship plot of the seriation of ware types by contexts (Figure 16). Ideally, a seriation would show for each ware a gradual increase in use, then a peak of popularity, and then a slow decrease. We see this with later ware types.

But, not with earlier wares. Creamware and pearlware, most notably, are consistently recovered across most contexts, including the later ones. This means that within that earlier group of contexts, there is a significant amount of time averaging. Therefore, when thinking about site formation, something or someone is mixing up those earlier contexts and causing them to appear more homogenous than the later contexts.

Returning to the original CA of contexts and ware types (Figure 10), notice how the scores are so densely clustered on the left, spreading out as we get later in time. The CA recognizes and therefore demarcates the later contexts because they are unique in the appearance of later wares.  

Therefore, we a need a way to parse the early contexts to get a more accurate understanding of the occupation period of Building s. We are now currently working on how to calculate an independent measurement of time averaging. So, stay tuned…   

That’s not to say that the CA isn’t correct – it still shows a reliable relative chronology, but most likely there is greater variability in that early cluster than what the CA can recognize. Right now, the nuances of that early period are obscured.

So What?

Now for the most important question we can ask: So What? Is this all math for math’s sake? No, of course not! We can put this information to good use, especially as it concerns taphonomic processes. We have already discussed a caveat in measuring and then mitigating the impact of time averaging. But, even just identifying where that time averaging is happening, we are getting a picture of how the site was formed, and just as importantly, how to interpret the 1980s legacy records. The identified stratigraphic groups provide an understanding of events that took place at the site over a roughly 150-year period, and the CA and seriation indicate important differences in those earlier deposits that we need to investigate further. And, ultimately, we will have established a better baseline for subsequent analyses. 

A Few Teasers

Because we mentioned subsequent analyses, here are a few examples of just the beginnings of future research.

Building s Subfloor Pit

We can zoom into the Building s-period to examine the relative chronology of those stratigraphic groups (Figure 17). One quick rabbit hole would be to look at just the subfloor pit fill, which was excavated in arbitrary levels with number 4 being the topmost layer. The CA suggests that the subfloor pit filled in gradually during the occupation of Building s, but then a mass filling episode topped it off – in effect redepositing earlier sediment on top.

Sherd Size

A cursory examination of the faunal assemblage revealed smaller cuts of meat recovered from the Building s occupation zone, with a possible theory being that the occupants did not have access to better cuts. However, another possibility involves larger bones rolling down the hill or redeposited there as cleanup. To test this, we look at ceramic sherd size (Figure 18). A wash layer, the terrace fill, and an erosion deposit – all downslope from Building s – have larger ceramic fragments than the subfloor pit and occupation zone. This supports the second hypothesis that gravity or cleanup is to blame, instead of Building s occupants having less access to quality meat cuts.

Intersite Analytical Units

And finally, because we talked a big game about our forest-through-the-trees goal: we briefly compared Building s with the first site we ran this assessment on, Building o, at the project-phase level (Wheeler and Sattes 2024) (Figure 19). Building o is slightly earlier, but there is an overlapping period, as seen a the cross-hairs on the plot. Dr. Kelso observed that there were clear signs of economic stratification between the different dwellings, and he noted discrepancies between Building o and then r, s, and t as an example (Kelso 1993:21). Having comparative units of analyses like these is the first step in asking the question about how we can measure the economic differences between these two households. A question for future research.

We hope those teasers drive home the benefit of applying the chronological phasing method and subsequent analyses. Additionally, a major goal is for these methods to be accessible and replicable. For example, everything produced here was done in R Studio, which is free.

Acknowledgements

We would like to say a huge thank you to Fraser Neiman for being the brains behind this project. We also acknowledge the original 1980s archaeologists who excavated Building s, as well as the DAACS team who analyzed it in the early 2000s.

References

Bates, Lynsey E., Jillian E. Galle, and Fraser D. Neiman. 2020. Building an Archaeological Chronology for Morne Patate. In Archaeology in Dominica: Everyday Ecologies and Economics at Morne Patate, edited by Mark W. Hauser and Diane Wallman, pp. 64-87. University Press of Florida.

Hill, Martha 2002a. Summary of Archaeological Excavations by Site. Mulberry Row Project. Unpublished report on file at the Jefferson Library, Thomas Jefferson Foundation, Charlottesville, Virginia.

Hill, Martha 2002b. Summary of Archaeological and Documentary Evidence for Excavated and Standing Buildings on Mulberry Row. Mulberry Row Project. Unpublished report on file at the Jefferson Library, Thomas Jefferson Foundation, Charlottesville, Virginia. 

Hill, Martha. 2003. Building s. The Digital Archaeological Archive of Comparative Slavery, October 2003. https://www.daacs.org/sites/building-s/#background.

Jefferson, Thomas. 1792. Memorandums for Manoah Clarkson [23 September 1792], The Papers of Thomas Jefferson, vol. 24, 1 June-31 December 1792, edited by John Catanzariti. Princeton: Princeton University Press, 1990, pp. 412-414. 

Jefferson, Thomas. 1793. Thomas Jefferson to Thomas Mann Randolph, Jr., 19 May 1793. The Papers of Thomas Jefferson, vol. 26, 11 May – 31 August 1793, ed. John Catanzariti. Princeton: Princeton University Press, 1995, pp. 64-66.

Jefferson, Thomas. 1796. Monticello: building insurance, recto. N133; K136 [electronic edition]. Thomas Jefferson Papers: An Electronic Archive. Boston, Mass.: Massachusetts Historical Society, 2003. http://www.thomasjeffersonpapers.org/.

Kelso, William M. 1993. Archaeology along Monticello’s Mulberry Row. Unpublished report produced for the Monticello Department of Archaeology. On file in the Monticello Department of Archaeology records.

Kelso, William M. 1997. Archaeology at Monticello: Artifacts of Everyday Life in the Plantation Community. Thomas Jefferson Memorial Foundation: Monticello Monograph Series.

Kelso, William M., and Douglas W. Sanford. With contributions from Dinah Crader Johnson, Sandy Sanford, and Anna Gruber. A Report on the Archaeological Excavations at Monticello Charlottesville, Virginia 1982-1983. Unpublished report produced for the Monticello Department of Archaeology. On file in the Monticello Department of Archaeology records.

Neiman, Fraser D., Jillian E. Galle, and Derek Wheeler. 2003. Chronological Inference and DAACS. Society for Historical Archaeology Conference, Providence, RI.

Neiman, Fraser D., Kim Zawacki, and Jillian Galle. 2001. Notes on the Chronology for the Building s Site.” Unpublished report produced for the Monticello Department of Archaeology. On file in the Monticello Department of Archaeology records.

O’Connor, Crystal, and Fraser D. Neiman. 2024. Distinguishing Boundaries Among Households on Plowzone Sites: An Example from Monticello. . Southeastern Archaeological Conference, Williamsburg, VA November 13-16, 2024. Available online: https://www.monticello.org/monticello-archaeology-news/seac2024-oconnorneiman.

Sanford, Douglas W. 1995. The Archaeology of Plantation Slavery at Thomas Jefferson’s Monticello: Context and Process in an American Slave Society. Unpublished Ph.D. dissertation, Department of Anthropology, University of Virginia, Charlottesville, VA.

Wheeler, Derek, and Corey Sattes. 2024. Reassessing the Chronological Boundaries of Monticello’s Mulberry Row, Charlottesville, Virginia. Southeastern Archaeological Conference, Williamsburg, VA November 13-16, 2024. Available online: https://www.monticello.org/monticello-archaeology-news/seac24-wheelersattes.