Flat glass fragments on historic archaeological sites most commonly represent two major categories defined by their usage. By far the most common type is pane glass. The other is mirror glass. Mirror glass can be difficult to identify because the reflective backing has usually decayed over time, leaving only the base glass material. Mirrors would often be made from polished plate glass, which is usually two to four times thicker than other pane glass types (Jones and Sullivan 1985).
This chapter will demonstrate the results of the flat glass analysis from the Beaver Creek Trail Crossing Site (25SW49). The class of flat glass artifacts examined most intensively is window pane glass sherds.
Manufacture and Transportation
In the United States, pane glass was produced using two techniques in the early nineteenth century: crown and cylinder. Both methods of manufacture are off-hand, meaning hand blown, and required skilled craftsmen to blow the glass forms individually.
Crown glass production begins by gathering ten to fourteen pounds of molten glass, and then forming it into a hollow pear shape. It is attached to a metal rod, called a pontil, at the base, leaving the opposite end open. The form is then placed back into the furnace while the craftsman spins the rod. The combination of heat and centrifugal force opens the glass into a circular sheet (Roenke 1978). "The force wherewith this the glass flies open at the close of this process would be sufficient, if its brittleness were not removed by heat, to break it into innumerable fragments" (Lardner 1972 ). The result of this process is a disk of glass from fifty to sixty inches in diameter. These disks, called tables, were cut into smaller panes for specific use. Cutting regular panes from a curved sheet, as well as cutting around the pontil mark, or "bull's eye," results in many irregularly shaped scrap fragments.
More common in the United States was the production of cylinder glass, sometimes called broad glass. In this process, a gather of molten glass is blown and formed into a cylinder, which is scored along one side, then placed back into the furnace where the cylinder opens along the scored line to form a rectangular plate. As this form is already rectangular, and lacks the pontil mark, which was removed with the near end of the cylinder, less glass is lost as waste in cutting rectangular panes. Unrolling the cylinder, though, stresses the interior and causes waves or cockles in the final product. By the 1830s this process was used almost exclusively in the United States because of its economical production (Roenke 1978). Economy of production was additionally realized because the ingredients for broad glass could also be "inferior" due to the lowered stress on the sheets in manufacture (Lardner 1972). As in this type of manufacture the glass was not subjected to the intense centripetal force as in crown class production.
Steam power expanded in the early nineteenth century. Shipping times were reduced while increasing reliability and amount of traffic delivering from the manufacturer to customers. Land transportation from the glass houses to installation in Nebraska underwent a revolution in the early nineteenth century with the expansion of rail lines and canal completion. Railroads reached the Mississippi River during the 1850s, and the transcontinental railroad was completed in 1869. In 1865, Thomas Garfield reported almost all the glass merchants he had spoken with in Milwaukee, Chicago, St. Louis, and Cincinnati were buying Pittsburgh glass (Roenke 1978).
The Beaver Creek Trail Crossing Site's occupation ranged from 1862 until at least 1871, beginning in the middle range of the transportation evolution and ending as railroad systems virtually eliminated large scale wagon freight along the Nebraska City Cutoff. In 1871, rail lines reached the town of Beaver Crossing and Roland Reed closed his road ranch in favor of homesteading his own property several miles away.
Dating of Window Glass
Window glass has been demonstrated to increase in thickness over time throughout the nineteenth century (Roenke 1978; Schoen 1990). Roenke (1978) and Schoen (1990) have each developed a dating formula demonstrating there can be made close correlation between the actual thickness of flat glass fragments on a given site and the date of glass manufacture. These methods are based upon the premise that all the windowpanes from a given structure initially were ordered from a single glass house at a single point in time. While Schoen restricted his study to dwellings with a short period of occupancy (under 20 years), Roenke noted additional modal points in the thickness data could represent replacement of damaged panes.
While glass may resist decomposition of form for great amounts of time, early glass was known to patinate; especially the lower quality glass produced using the cylinder method. It came of such interest that Michael Faraday, an early nineteenth-century physicist, tested three pieces of flat glass, breaking each into two pieces. One half of each piece was protected from light and moisture, and the other half was left exposed to sun and rain. The halves left in the sun were distinctly altered in clarity after eight months (Lardner 1972 ). It can be inferred that older windowpanes would have become more and more patinated over time if left exposed in their frames.
It can be extrapolated that additional modal points could indicate transference of ownership from one person to another, with the new owner replacing damaged or older panes, especially panes showing extreme patination.
Schoen's dating method demonstrates regional differences in the thickness dating patterns. These differences reflect transportation times, common suppliers, and changes in manufacturing practices. In 1845, England repealed excise taxes restricting windowpane thickness, permitting thicker, more durable glass to be transported. Schoen identifies five criteria for inclusion in his regional pattern study (Schoen 1990). A site must demonstrate that it:
- occurs in a single geographic region.
- represents a similar socio-economic level and lifestyle.
- was occupied by a short period of time.
- was built between 1800 and 1900.
- has historical documentation describing the original structure and date of manufacture.
The Beaver Creek Trail Crossing Site meets these requirements in that:
- it is located within the Central and Northern Plains.
- the Roland Reed Ranch was a road ranch (as were the Chappell Site and Rock Creek Station which Schoen's study examines).
- its recorded occupation was over a period of nine years.
- it was built between 1862 and 1871.
- historical documentation on the site is available, including images naming the site.
Some archeologists are cautious about the utility of the dating formulae. In particular, Deiss, in discussing dating of pane glass noted the offhand production of each table of crown glass and each sheet of cylinder glass. Thickness and quality of the glass depended on individual batch composition and skill level of the craftsmen blowing the pieces. Deiss felt attributing provenance to the date ranges from artifact measurement would be, ". . . difficult for obvious reasons" (Deiss 1981). One of his major points opposing dating was a lack of comparative analysis.
There is agreement among the authors of these works that over time, the tendency was a gradual increase in average thickness. Even lacking specific dates, a thicker fragment of windowpane glass would most likely indicate a later date of manufacture than a thinner one, and similar thicknesses of fragments on a given site would indicate a close relationship in time.
The goal of dating the glass samples is not to provide a specific date, but instead to compare the approximate dating of the pane glass artifacts to the historical records. It is a statistical tool, not an absolute measurement. In particular, the identification of the major modal points, and the ranges between them, could be compared to the historical record for periods of repair, renovation, and transfer.
Recovery and Distribution
Glass recovered from the Beaver Creek Trail Crossing Site was sampled first removing all obvious colored, curved, and heat altered fragments. Each fragment was then measured to 0.005 inches in a minimum of three locations to calculate an average thickness accurate to 0.01 inches. In order to restrict the sample from including non-windowpane fragments, calculations for mass and dating were restricted to 90 percent of the range, removing the highest and lowest values from the data, five percent from each the thinnest and the thickest. This was the procedure indicated in Schoen's work (Schoen 1990).
Sizes of individual fragments ranged from ½ in. to 8 ½ in. along the long axis. When counting numbers of fragments recovered to quantify the distribution across the site, these two fragments would each be considered the same quantity. In order to visualize the actual amount being recovered, the flat glass from each provenience was weighed to the nearest 0.01 g. Each test unit volume was calculated from excavation records to the nearest 0.01 m3. These data could then express the density of flat glass being recovered from each test unit.
A contour map was used in order to highlight the relationship between test units. The three greatest densities were found in Feature 5, considered a fire pit, and Features 4, 7, and 8, which were concentrations of mortar, possibly wall indications. The heaviest density of flat glass was recovered in and around a large mortar concentration related to Features 7 and 8 in the northeast corner of the site. In addition, Figure 10.2 shows the distribution of sherd by number in each test unit.
The 90 percent sample of flat glass used to represent pane glass at the Beaver Creek Trail Crossing Site consisted of 1,627 fragments totaling 1429.96 g. distributed through 96 proveniences. Each provenience was weighed separately, and then each fragment in the provenience was measured for thickness, using the method mentioned above, which is the accuracy recommended by both Roenke and Schoen. Schoen's formula developed for mean thickness dating with a 90 percent sample to yield a mean date of occupation is:
Date=1731.461 + (1777.117 * Thickness in inches) (Schoen 1990:Table 10)
Schoen tested this formula against historical data in the Northern and Central Plains and demonstrated a 98 percent correlation between the calculated date and the historical record, with a standard error of only 4.358 years. This formula was used to calculate an estimated date for the average thickness of each provenience. Each individual date was used two ways in order to recover as much information as possible.
First, the date determined by each provenience was multiplied by the mass ratio that provenience contributed to the overall mass from the site. The result was a single date, that date most commonly indicated from the mass of glass overall. According to the calculations, the Beaver Creek Trail Crossing Site would have initially been occupied in 1860.5 +/- 4.348 years.
Second, the calculated dates by provenience were assembled in a histogram to identify the major modal points.
As shown in Figure 10.3, there are three major modal points represented. The calculated dates for these three points are 1854, 1864, and 1871. This would indicate three major purchases of pane glass.
The Beaver Creek Trail Crossing Site's historical documents indicated this was the site of the Roland Reed Ranch, a business designed to interact with passing traffic along the Nebraska City Cut-off Trail. John Leonard built his home on this location in 1862. Roland Reed took possession of the property from Leonard in 1866, though it was owned by his brother Amos Reed. In 1871, with the arrival of the railroads virtually eliminating freight wagon traffic along the Nebraska City Cut-off, he moved to his own homestead elsewhere. (Waterman 1914:33–34). There is a reference to the Reed property being split and sold to Chris Klem and George Geis (Smith 1937:25), though no record could be located to indicate usage, destruction, or removal of the buildings of the Reed Ranch.
The modal data indicate three manufacturing points, and the calculated date ranges for the site overall as well as the three modal points include the dates from the historical records. While the precise measurements of the structures remains to be determined, this provides excellent supporting evidence that the Beaver Creek Trail Crossing Site could be the site of the Roland Reed Ranch, as the initial excavation sought to demonstrate.
1982 The Development and Application of a Chronology for American Glass. Midwestern Archeological Research Center, New Holland, Illinois.
Jones, Olive, and Catherine Sullivan
1985 Parks Canada Glass Glossary, Minister of the Environment, Ontario, Canada.
1972  The Manufacture of Porcelain and Glass, History of Technology, Vol. IV, London. 1832. 1972 reprint ed. Noyes Press, Park Ridge, New Jersey.
1978 Flat Glass: Its Use as a Dating Tool for Nineteenth Century Archaeological Sites in the Pacific Northwest, Northwest Anthropological Research Notes, 12(2), Part 2:1-128.
1990 Window Glass on the Plains: An Analysis of Flat Glass Samples From Ten Nineteenth Century Plains Historic Sites, Central Plains Archaeology, 2(1).
Smith, Wm H.
1937. Early Days in Seward County Nebraska. Seward, Nebraska.
1993 Substance and Practice. Building Technology of the Royal Engineers in Canada. Minister of the Environment, Ottawa, Ontario.
Waterman, John H.
1914 Revised General History of Seward County Nebraska. Beaver Crossing, Nebraska.
1972 New England Glass and Glassmaking. Old Sturbridge, New York.