Internet Archaeol. 13. Symonds and Ling.

5. Travelling beneath crows

Road travel was calculated at 3.2km per hour and river travel was estimated to be approximately 6.4km per hour. The length of time that could be spent travelling per day was estimated to be eight hours. This time was chosen as a lower estimate of what could actually be accomplished. Longer periods of travel, especially in the summer, are certainly possible. It was considered important not to overestimate the average travel time per day, as the manner in which pottery was transported from kiln site to consumption site is not clear. An average day's travel was selected over one that was overly strenuous.

Measurements and calculations were done via a GIS (ARC/INFO) into which the pottery data and the transportation (road and river) were added. Rivers were mapped for their extent of navigability as determined by Edwards and Hindle (1991). The network of roads was compiled from Margary (1973, 192) and Vince and Oliver (in Sawyer 1998, 17). Roads and rivers were maintained in separate network covers. While it would be exciting to be able to map the use of roads and rivers operating as one combined transportation network, the data needed to perform this analysis are extensive and often not available. Written records are simply too sparse to be informative about local preference of routes.

In order to calculate the distance from the kiln site to various settlements in the landscape, the settlements needed to be connected to a road and/or river system. This calculation was made by selecting the nearest point on the transportation networks to the consumption sites and connecting the two. The result was a series of small offshoots from the primary road and river networks. It was important that the offshoot vertices were not excessively long. It does not make sense for someone to travel 20km to a river when a road is 5km away. Therefore, some sort of reasonable limitation had to be imposed. Five kilometres was chosen as it would take approximately an hour to walk. The proximity of late Anglo-Saxon pottery sites to roads and rivers was compared with that of the entire EMASPP database of post-Roman pottery. This demonstrated that 80% of the Anglo-Scandinavian sites were within 5km, or an hour's walk, of a Roman road, a greater percentage than the EMASPP dataset as a whole (Figure 5). Additionally, 25% of the Anglo-Scandinavian sites were within 5km of a river (Figure 6). This was a much smaller percentage than the overall EMASPP database, suggesting that the Roman roads were more commonly used than the rivers.

Figure 5: Proximity of the distribution of Lincoln wares to the Roman roads - EMASPP vs Late Saxon sites (cumulative percentages)
Figure 6: Proximity of the distribution of Lincoln wares to the navigable rivers - EMASPP vs Late Saxon sites (cumulative percentages)
Figure 7: Regression diagram of Lincoln wares as measured along Roman roads

Regression analysis on the distances measured via the roads displayed quite a different slope than the Euclidean measurements. The resulting r-squared value increased to 0.0125 from the previous Euclidean value of 0.0052 (Figure 7). This demonstrated an increased correlation between distance and pottery distribution in comparison with the Euclidean distance regression. From both regression plots it is apparent that most of the sites were associated with small amounts of pottery, illustrating the ubiquity of pottery throughout the region. However, fifteen sites were associated with large amounts of pottery, suggesting that different social practices were affecting the material distribution of pottery at these sites. When the regression plot included only those sites with a significantly larger number of pots (in this case, more than ten), the slope increased to 0.1595, indicating that distance was substantially more important to the elevated distribution of Lincoln wares (Figure 8).

Distance in the scatterplot was also calculated with regards to travel time. Half of the fifteen sites with large amounts of pottery were clustered approximately a day's travel from Lincoln. Included in this count were the few sites over eight hours travel away. Many of these sites had larger amounts of pottery than those scattered two to three days away from Lincoln, illustrating the reason for the increased r-squared value. This implies that distance, particularly that which could be travelled in one day, had an impact on the amount of pottery distributed from Lincoln.

Figure 8: Regression diagram of Lincoln wares as measured along Roman roads showing only those sites with more than ten vessels
Figure 9: Regression diagram of Lincoln wares as measured along navigable rivers

The regression analysis performed on distances calculated along the rivers produced a different pattern (Figure 9). Here, the r-squared value equalled zero, demonstrating that distance was not an important factor with regard to river travel. Indeed, the sites with large amounts of pottery were not clustered within a day's travel but instead ranged from two to three days away from Lincoln. It is more difficult to ascertain how quickly boats were able to move along the rivers, which may have resulted in an erroneous estimation of a day's travel. Nevertheless, distance does not seem to have been influential in the distribution of pottery along rivers.

While the scatterplots were useful in terms of discerning patterns in the data, they did not map these distances geographically, nor did they represent the distance in socially meaningful terms. In order to provide a geographical element to the data, it was decided to map the distance via the transportation networks on the original pottery distribution maps of each ware. It was considered important to demonstrate the difference between the Euclidean (as the crow flies) distance and the transportation distance. Thus, the Euclidean distance between kiln site and consumption site was measured and subtracted from the transportation distance. This differential distance was then mapped along the same compass orientation as the site.

Figure 10: Temporal distance map showing the difference between Euclidean distances and the distance measured along Roman roads
Figure 11: Temporal distance map showing the difference between Euclidean distances and the distance measured along navigable rivers

The map demonstrating the distortion between the geographical and temporal distances along roads (Figure 10) shows a significant difference between measurements. Indeed, while Roman roads are often considered to be direct routes from one place to another, this map indicates that the 'directness' of these routes is relative. Euclidean distances are clearly shorter than those measured along Roman roads. Conversely, in comparison with the river distances (Figure 11), the distance along roads was clearly shorter.

While variation between Euclidean and travel distance is apparent from the maps indicating the distortion between the two, they do not account for cognitive understanding of these distances. Perception of distance is structured by the time it takes to travel from one place to another. Travel along rivers was at least twice as fast as that along roads (see above). Thus, the distances along rivers is conflated in relation to the measurements along roads. In order to explore the cognitive perception of the distance between sites, the geographical background to these maps was removed. Instead, temporal measurements were added to reflect the number of days travelled along either river or road.