The goal of this project was to conduct a spatial analysis of archaeologically-recovered materials from the 17th-century fishing station on Smuttynose Island, one of the Isles of Shoals archipelago which is currently part of the town of Kittery, Maine. The artifacts in question are North Devon ceramics (Figure 1) and West Country smoking pipe bowls (Figure 2). These artifacts were chosen for two reasons. First, both are distinct and readily identifiable by a trained archaeologist. Second, when the artifact collection from Smuttynose Island was analyzed as part of the author’s doctoral dissertation, it was theorized that these two artifact types could be used as an indication of a migratory fisheries assemblage, at both types appeared to largely disappear from the assemblage at the same time that the migratory fishery largely disappears from the Gulf of Maine (Clausnitzer 2018: 233-234; 308). A later analysis of the smoking pipe assemblage from the same site demonstrated that West Country smoking pipe bowls and North Devon ceramics co-occurred vertically in the archaeological record; in other words, they appeared during the same chronological periods (Clausnitzer n.d.).
Figure 1. North Devon Ceramic Sherds
Figure 2. West Country Smoking Pipe Bowl
With this in mind, the question asked was “do these artifacts co-occur spatially?” Perhaps more accurately, the question actually became “how strongly do these artifacts co-occur spatially on Smuttynose Island?”, as it was already known they were recovered from the same excavation units. This is an important question, as it provides additional evidence for the hypothesis that these artifact classes can be used as distinct chronological markers to separate migratory and residential fisheries sites in the Gulf of Maine, a task which has proven difficult in the past. Being able to answer this question can help identify migratory fisheries sites, which in turn can help answer questions about the early European settlement of northern New England and what role extraction industries played in the development of colonial New England.
The data used for this analysis was drawn from a subset of data originally used in the author’s doctoral dissertation; in total 42 artifacts were selected for inclusion in this study. Eight of these artifacts are pipe bowls, and the other 34 are sherds from North Devon ceramic vessels. While relatively small sample size, all of the artifacts chosen were diagnostically significant and included the provenience information necessary to conduct the spatial analysis. Because this data was drawn from the author’s own records, it required very little in terms of wrangling and clean-up. An Excel file was created which contained the artifact number, a description of the artifact (Pipe or ND Ceramic), the excavation unit it was recovered from, and the geographic coordinates of the artifact.
The last piece of information proved to be the most difficult data to generate, due to the field recording techniques used. Excavation units were assigned a number in the format of XRY, where X was the number of meters west and Y the number of meters north of a 0R0 point, with the south-west corner used as the datum point. Each catalogue entry then also contained piece-plot coordinates in centimeters, again measured as X-Y from the south-west corner datum point, which represents the location of the artifact within the excavation unit. GPS coordinates were not taken for each excavation unit, so this data had to be reconstructed in ArcMap. Further complicating this procedure is the fact that “north” and “east” did not refer to geographic directions, but rather to “grid” directions; grid north, in this instance, is actually closer to a north-northeast direction. By using field notes and recorded measurements, along with readily-identifiable features which were used to establish the original baseline, unit coordinates, and piece-plots were converted to UTM coordinates in the Maine East state plane system.
Finally, for the actual spatial analysis, which used a cross-K analysis, the geographic extent was limited to actual limits of the archaeological excavation. This was an area of around 55 meters by 20 meters; compared to the 25-acre size of Smuttynose Island, this is but a tiny fraction of the total area, as illustrated by the map below.
As mentioned previously, the spatial analysis was undertaken using a K function, which tests the spatial relationship between point patterns. Ripley’s K functions were run on each artifact type, and then a cross-K function was run on both types together. The code used to run this functions are included in the code chunk below.
#second order properties of point patterns
#k-function compares presented point pattern to expected pattern
#looks at density and computes an estimated random distribution
#create ppp object for spatstat analysis
smuttynose_sp <- as_Spatial(smuttynose)
#Next, create ppp object
smuttynose_ppp <- as.ppp(st_coordinates(smuttydata_sf), W=as.owin(smuttynose_sp))
#add attributes/marks back to point pattern
marks(smuttynose_ppp) <- as.factor(smuttydata_sf$Description)
#subset out data
ceramics_ppp <- subset(smuttynose_ppp, marks=="ND Ceramic")
pipes_ppp <- subset(smuttynose_ppp, marks=="Pipe")
kceramics <- Kest(ceramics_ppp, correction = "border")
kpipes <- Kest(pipes_ppp, correction = "border")
#test relationship between two point patterns.
#recombine both point patterns by creating subset of mark types
artifacts_ppp <- subset(smuttynose_ppp, marks=="ND Ceramic" | marks=="Pipe")
ck.artifacts<- Kcross(artifacts_ppp, i="ND Ceramic", j="Pipe", correction = "border")
The results of the K function are presented in the three charts below. In all three cases, the slope of the Kbord(R) lines is much greater than the slope of the Kpois(R) line. In practical terms, this means that that North Devon ceramics and West Country pipe bowls are not only tightly clustered amongst themselves, but also with each other, suggesting a high level of spatial correlation between the two artifacts types. Even when using a Monte Carlo test to generate a simulation envelope (the fourth chart below), the curve of the Kbord(R) line is well above the bounds of the envelope. This further supports that the distribution of these artifacts is not random, but is tightly clustered and spatial correlated to each other.
Density plots were also calculated for these point patterns, the results of which are included below. Unsurprisingly, the density for the ceramic pattern is much greater than that for the pipes, given the significantly large sample for the ceramics. However, it is also interesting to note that while the locus of the density plot for the pipes is the same as for the ceramics, there is a noticeable secondary locus in the northern part of the site. While limited by the small sample size, this may be evidence for a much wider distribution of the smoking pipes compared to the ceramics.
The results of these tests show that not only are North Devon ceramics and West Country pipes chronologically correlated, as shown by prior research but that they are very strongly spatially correlated. This strengthens the larger argument that these two artifacts are potential chronological markers that can be used when separating a migratory/early settlement context from a later residential context.
This analysis sought to use a Ripley’s K-function, or more specifically a cross-K function, to determine if there was a strong spatial correlation between North Devon ceramics and West County smoking pipe bowls in the archaeological investigations on Smuttynose Island, Kittery, Maine. The results of these functions showed that the distribution of these artifacts was not random, but rather there was a tight clustering of each artifact type and that the artifact types were tightly clustered with each other. Thus, it can be said that the artifacts are in fact spatially correlated. This, combined with a previously-established chronological correlation, lends credence to the theory that these artifacts can be used as chronological markers.
This analysis is not perfect, and in fact, has several drawbacks. Most significant is the small sample size, which means that addressing the significance of the results is difficult. Compounding this is the fact that the sample used is a subset of a larger sample. As mentioned before, the archaeological site on Smuttynose Island is approximately 55 meters by 20 meters in size; the sample presented here is drawn from only 22 square meters of that site. This means that parts of this analysis, in particular, the density maps, are biased toward the center of the excavation area. This is in addition to the usual biases present in archaeological research and analysis; these include preservation biases, collection biases, unrecognized inherent assumptions on the part of researchers, and other such issues. Still, it does help to shed light on the research question at hand and shows that such studies may have a place in the larger archaeological discipline.
Other potential questions which could be addressed with this type of analysis include looking at the distribution of ceramic vessel forms (are drinking vessels more or less concentrated by the meetinghouse/tippling house?), the distribution of faunal remains (does the distribution of identifiable body parts of the codfish help to identify fish processing areas?), and the distribution of ceramic ware types (are higher-status ceramics concentrated in certain areas, or distributed evenly across the site?).
Clausnitzer, Arthur R., Jr. 2018 The Seventeenth-Century English Cod Fisheries of Newfoundland and New England, circa 1600-1713: An Archaeological and Historical Comparison. Doctoral dissertation, Department of Archaeology, Memorial University of Newfoundland, St. John’s, NL.
n.d. The Use of Tobacco Pipes in Identifying and Separating Contexts on Smuttynose Island, Maine. Northeast Historical Archaeology. Accepted for Publication.