'This Does Not Compute': The All-American Pipeline Project Revisited

Article excerpt

In 1989, ANTIQUITY published the ambitions and methods of a remarkable project to automate and computerize the recording and analysis of artefacts from a very large salvage program in the western United States. Here is a follow-up that explains why the methods did not realize the ambitions, and questions whether those ambitions were well chosen.


I have been asked by the editor of ANTIQUITY to re-evaluate certain aspects of a project published earlier in this journal. With a limitation on size, this critique will be succinct (a longer article soon to appear is Ackerly in press).

The project examined here is the All-American Pipeline Project (AAPL). Inaugurated in 1986, this project pushed the limits to which electronic data capture, analyses, storage and retrieval technologies were applied to archaeological data. This paper provides an end-of-project assessment that differs somewhat from the evaluation presented in Plog & Carlson (1989). I hope that this re-examination may prove instructive as we continue to integrate computers into the treatment of archaeological data.

Archaeology and goals

The AAPL project resulted in the collection of artefacts from 450 diverse sites scattered from the California coast near Santa Barbara (CA) eastward to the plains of West Texas near McCamey (TX). It traversed more than 1400 miles across the western United States and crossed five major physiographic provinces. The project encountered remains from a cross-section of the archaeological variability typical of arid and semi-arid environments of the American West. Artefacts ranged in age from Palaeo-Indian (c. 10,000-12,000 BP) to recent historic times. Approximately 120,000 artefacts were recovered.

The artefact assemblages consisted primarily of chipped stone items and ceramics. Within the chipped stone assemblages were such diverse artefact classes as debitage, formal tools (e.g. scrapers), cores and projectile points. The raw materials included obsidian, various cherts, andesite, rhyolite, fine-grained basalt and slate. The ceramic assemblage consisted almost exclusively of sherds in plain, surface-textured (e.g. corrugated) and painted wares.

The goals of the system are outlined in Plog & Carlson (1989: 260). Conventional archaeological analysis procedures were subject to errors including

* mis-measurement or misidentification of artefact attributes leading to

* reduced inter-observer replicability that could be further compounded by

* data entry errors.

These difficulties would increase exponentially as the number of artefacts and analysts increased. The primary goal of the project was to develop a computer-assisted means for capturing, analysing, and storing information about large quantities of archaeological artefacts. The project was motivated by the belief that reliance on computer-based analyses would

* speed up the data-analysis process,

* reduce labour costs associated with large-scale laboratory analyses,

* remove or significantly reduce inter-observer biases in data recording and analyses,

* automate the storage of archaeological data, and

* allow for the long-term archival storage of data.

There were four components to the technological goals of the AAPL:

* a system was to be designed to capture artefact images;

* methods would be developed so that the images, rather than the artefacts themselves, could be analysed;

* the results of image-based analyses could be linked and automatically entered into a database containing numerical data and graphical images;

* methods were to be developed so that both of these data sets could be stored and retrieved.

Plog & Carlson (1989: 260-63) describe the system in detail. The base system consisted of a television camera connected via a multiplexer board to a SUN 160 work station. Artefact images were displayed on a 17[inches] monochrome monitor so operators could confirm that images were present; images were then captured and stored as raster files for later analyses. …