ANT 412 - Historical Archaeology

Uses of Ceramics (and other domestic artifacts!)

Course Schedule Evaluation FAQ Assignments ARH Prog Anthro Events

Introduction Rural and Domestic Life: In progress


Last update: 9 November 2004

 19th-century Farm Life

Early and mid-19th centuries were time of pioneering in Ontario

Although there are some written descriptions of life on the frontier, archaeology is a key source

Increasing Farm Prosperity

Immigrants took land grants, cleared forest, and built log houses. After their farms became well established, they replaced their houses with frame ones. By 1880s, many had built brick houses.

Fairs and advertising reflect investment in farm technology, including, by the late 19th-century, steam tractors.


The fact that southern Ontario was still largely forest-covered in the mid-ninetheenth century, and that farmland had to be cleared, led to a major lumber industry, with numerous saw-mills along rivers, railways and ports to export the lumber, and eventually also a prosperous furniture industry. These industries died out after most of the forest had been destroyed, by the early 20th century.

Use of Ceramics  

One of the most important classes of artifact in historical archaeology is ceramics. Of the many important uses of ceramics, here I'll only summarize a few points with regard to the following:

  • Uses of ceramic smoking pipes in chronology of historic sites
  • Uses of stoneware and porcelain in dating historic sites
  • Uses of dinnerware in determining social status of households
  • Uses of dinnerware in determining ethnicity of households

Clay Tobacco Pipes

Long-stemmed, clay tobacco pipes are common on sites of the 17th, 18th and 19th centuries, in part because they are so fragile and were considered disposal.

Not only do they provide good evidence for an important leisure activity, pipes' design changes over time provide a useful basis for site chronology. In addition, indications of their manufacturers, including makers' marks, also document trade in imported pipes and the rise of local pipe-making industries.

Even in the 1860s, there were antiquarian studies of pipe chronology based on shapes of pipe bowls and makers' marks.

One method for dating sites is based on the observation that the diameter of pipestems changed over time, probably as the stem itself became longer to provide a cooler smoke as pipebowls got larger and tobacco got cheaper. This takes advantage of the most commonly found pipe fragments. In the 1950s, J.C. Harrington found that pipestem bores changed in samples from Jamestown and other Virginia sites:

 Dates  Diameter (inches)
 1590-1620  9/64
 1620-1650  8/64
 1650-1680  7/64
 1680-1720  6/64
 1720-1750  5/64
 1750-1800  4/64

Archaeologists began to use tables such as these, in conjunction with the average bore diameter of pipestems found in excavations, to date sites and deposits. In this dataset, for example:

 Bore (inches)  No. stems
 9/64  3
 8/64  14
 7/64  37
 6/54  7
 5/54  2
The Average of 7.14/64 suggests a date of ca. 1760, according to Harrington's table.

Later, Lewis Binford introduced use of regression analysis to estimate dates based on mean stem bore. This involves plotting a line through the scatter of points that results from plotting date (y-axis) against bore diameter (x-axis) for pipestems of known date. For example, in one more recent version of the Binford regression,

Date = 2073.98 - 50.57d

where d is the diameter in 64ths of an inch. This allows, in principle, fairly precise dates on collections of pipestems.

We do need to be cautious about the assumptions and problems of this technique, however. Generally, we need to construct a graph of the pipestem relationship based on diameters of pipestems found at dated sites (e.g., shipwrecks), and good pipestem collections of precisely known date are not as common as we might like. In addition, the method is not really applicable to deposits after about 1800. And, finally, there are many examples where the pipestems have given inaccurate dates, presumably because the deposit's site-formation history is more complicate than assumed, or the pipestems were manufactured by a maker who was somewhat anomalous.

Dating and identifying pipes turns out to be a lot more complicated than had been hoped, and there's a huge literature on pipes.

"Mean Ceramic Date"

Stanley South (1971; 1977) pioneered a measure of central tendency in assemblages of datable ceramics from historical sites in eastern United States. He intended this to give an idea of most likely occupation date of site or context. The method uses the range of dates for each ceramic type as basic data.

South attempted to estimate central tendency in the date of site occupation by taking the average of these range's midpoints (which he called "median dates"):
MCD = S(xifi)/ S fi
xi = "median date" of type i,
fi = number of artifacts of type i, and
Sfi = n = total number of artifacts

For this method to work well, we need to make some assumptions:

  • Little or no time lag from manufacture to distribution
  • Time lag from distribution to discard is inconsequential or can be ignored
  • Acquisition, breakage, and discard rates reasonably uniform
  • Manufacture distribution is symmetrical, so that median = midrange (in reality, South simply misued the term "median", as the middle of the time range would not normally be the median in the usual sense of dividing the pots so that half are earlier, and half later, than that date)

Most likely, both variation in manufacturers' outputs and delayed breakage, loss, and deposit would likely lead to skewed distributions, with a long tail toward the more recent end of the distribution. Even if we only take into account the production range, as South does, it is likely that more than half of the pottery was made in the first half of the range, so that the true median age of the pottery type (not to mention mean age) would be earlier.

If we model the way South's method works, we treat each pottery production range as a uniform distribution, with equal probability that any given pot was made in any year of its complete production range (e.g., 1% probability of being made in each year from 1701 to 1800). South's method, by only counting the middle of the range (e.g., 1750), essentially lumps all the probabiliy into one year instead of spreading it out across its range, thus ignoring most of the distribution.

Critiques of the Mean Ceramic Date have mostly concentrated on the fact that it gives insufficient attention to discard lags. Another problem is the arbitrary substitution of mid-century dates for types that last for a very long time spread over two or three centuries, instead of using the "real" midpoints of the production range. There has been little notice that the "median" is not really a median, that most of the information is being ignored in favor of the mid-ranges, that it gives more weight to common, long-lived types over rarer but more time-sensitive, short-lived types, or that the method does not result in confidence intervals. There have been some attempts to substitute the mode (peak of popularity) for the mid-range in the formula (e.g., Jacobs 1983; see also Burke 1982), which usually results in older dates. Steponaitas and Kintigh (1993) and Banning (1995) have suggested an alternative that takes the whole uniform distributions into account, rather than only the midranges.

Cumulative Uniform Distributions: Stacking the uniform distributions of different types results in a probability distribution (probability density function) and uses more of the information preserved in the ranges. Summing the uniform distributions of several artifact types produces a probability distribution that does not give undue weight to types with long production ranges. the peak in the distribution suggests the most likely period of the assemblage's composition. Some of the advantages of this approach are:

  • It gives weight to types that are chronologically sensitive
  • It accounts for length (although not shape) of production distribution
  • It allows use of confidence intervals and Bayesian statistics
  • It is less sensitive to errors in the extremes of the distributions
  • It visually aids in identification of multiple occupations

Recent and Possible Future Improvements:

  • Computer software that makes density plots by time (Kintigh's ARRANGE module)
  • Modelling shape of manufacture distribution and not just its length (replace uniform distributions)
  • Modelling discard and deposition as well as production
  • Bayesian analysis to narrow the distribution and estimate actual periods of occupation

Ceramics as Evidence for Social and Economic Status
In progress...

 Use of Nails   TBA