Studies on the deeper past often rely on relative methods of dating. Ancient objects are organized by some trait, known as typology, and placed in some order that may reflect some kind of development. Seriation is a closely related method, which aims to align multiple typologies to arrive at the best possible outcome. Seriation was pioneered in 1899-1901 by Flinders Petrie, a scholar who spent most of his career directing archaeological excavations in Egypt, and who set himself the goal to bridge history and prehistory via seriation. Petrie named his system Sequence Dating and effectively aimed to map chronological development of ancient Egyptian culture before the first historical evidence, around 3,100 BCE. To make his system more widely applicable, Petrie extensively published on Sequence Date values he had derived from prehistoric artifacts, where these artifacts fell in his overall system of sequences. Generally, low Sequence Date values correlate to older material than those with higher Sequence Date values. Many of the values are provided in Petrie's publications on the Predynastic period (Petrie & Mace 1901; Petrie 1920; 1921).
Petrie's goals were simple, to put prehistoric stuff in chronological order until it matched up with historical stuff, but he unknowingly introduced a problem that up to this day proves difficult to solve. The problem is akin to the Traveling Salesman: how does one determine the most optimal 'path' (order) in a large dataset? The answer to this seemingly simple question is deceptively difficult. Firstly, datasets are often not meant to be put in a particular order; while some order can always be defined, the meaning of that order is difficult to ascertain. Secondly, order can be defined in many ways and how one measures order is wholly dependent on the researcher and their research questions. Accounting for all variables and considering every potential possible influence on the process of organzing are notoriously difficult to control.
My PhD dissertation at Brown University examines the archaeological remains from Mesa'eed, a burial grounds from Egypt's late prehistoric period that was excavated over a century ago. The excavations were never properly published, but most of the documentation and artifacts that resulted from these excavations have over the last century ended up in the Museum of Fine Arts in Boston. I collated all documentation and artifacts in Boston to arrive at an overall picture of the excavations' results. In short, over 3,000 artifacts were uncovered in c. 739 burials. To put these burials in chronological order, and study social developments at Mesa'eed, I designed this small program to bulk batch process some of my data to put the burials at Mesa'eed in chronological order according to Petrie's original Sequence Dating system.
This program expects a .csv-file as input with artifact data organized in rows and columns. Each row represents an artifact with the following information on the headers:
- Site.string() : Name of the site
- Grave.string() : Number of the burial
- SD_MIN : Sequence Date low value
- SD_MAX : Sequence Date max value
The program calculates quartiles for each burial based on the Sequence Date values provided for all associated artifacts. The program will output a new CSV-file named SD.csv, with six columns, named 'Context', 'min', 'max', 'Q1', 'Q2', and 'Q3' (see SD_example.csv). In addition, for each burial/context, the program can output boxplots, based on the quartiles, mapping them onto (1) the three major periods defined by in his Sequence Date range (Amratian, Gerzean and Semainean) and (2) the three somewhat similar phases defined by Werner Kaiser (1956; 1957) of Naqada I, II and III. At the moment, this program only works with Petrie's Sequence Dates rather than individual types of artifacts. Kaiser defined his phases partially on the same types that Petrie used and therefore similar Sequence Dates. However, Kaiser moved some types to different Sequence Dates.
This file is included as an example input file. Running the program (see below) will result in the following four images.
NOTE: The gradients represent the different stages defined by Petrie and Kaiser. Dark gray is the earliest phase, the Amratian according to Petrie and Naqada I according to Kaiser. The subsequent phases are Gerzean, Semainean and the First Dynasty, according to Petrie, or two stages of Naqada II, and Naqada III (overlapping with the First Dynasty) according to Kaiser.
The program can be run with the command "python main.py Artifacts_sample.csv" from the command line. This should generate the images and CSV-file described above.
I have no plans at the moment to take this program much further than what it is. A number of much more sophisticated software packages exist to do original seriation work though I have not found any that produce easily comparable boxplots (in batch) such as this program does. I might rewrite parts of the program to focus on specific artifact types rather than a Sequence Date range.