The Item-based Approach

OCHRE's hierarchical, item-based approach differs from the tabular, class-based approach commonly employed in archaeological databases. The usual approach has been to define a set of classes (e.g., debris layers, architectural features, ceramic artifacts, metal artifacts, stone artifacts, faunal remains, botanical remains, etc.) and to create a table for each class. An individual item is represented as a row in a particular table. The table columns represent properties that describe the items (e.g., "type," "color," "length," and "weight" in the simple example shown here).

Data tables are easy to work with using commercial relational database software, which helps to explain the popularity of this approach. (It is worth noting that the relational data model does not actually require a class-based organization of data, although it is well suited to it. A hierarchical, item-based approach can be implemented in a relational database system, as was done in an earlier Windows application that serves as a prototype for OCHRE. But, in practice, relational databases usually employ a class-based organization of the data.)

The class-based approach is familiar and straightforward, but it requires that each observed entity be placed within a predefined class; moreover, the properties available to describe the entity are limited to the columns predefined in that class's table. It is not easy to add or delete a property for a particular entity because this would affect an entire table column and hence an entire class of entities. As a result, it is difficult to capture the variability and complexity of the entities encountered in academic disciplines such as archaeology, in which there are few common standards for how entities should be described. Observations are forced into a rigid tabular mold, relegating information about idiosyncrasies or unusual properties to unstructured (or inconsistently structured) free-form notes.

From a researcher's perspective, the fundamental problem with a tabular, class-based organization of data is that the classification of entities is determined in advance by the computer recording system instead of allowing for multiple overlapping classifications that might emerge from the ongoing analysis and comparison of the units of observation. In an item-based database, on the other hand, the basic structural element is not the class but the individual unit of observation, on whatever spatial or temporal scale that unit is defined. Any number of properties can be used to describe a particular item and new properties can be added as needed without affecting the structure of the database. Classes are not built into the database ahead of time but are generated by queries on the properties of individual items, which produce sets of items that are considered (at least by the person performing the query) to be in some way similar or worthy of being grouped together. Thus all classifications are determined by researchers themselves in the process of describing and querying the data—no classification is treated as absolute or universal.