In the early 1970s, Edward Harris, excavating at the pluri-stratified site of Winchester in England, added a valuable refinement to the schematic drawings that record stratigraphic sequence. Within the logic of an excavation, the purpose of a section drawing was to show where a layer came in the sequence. There are three possible ways to express the relationship between two layers: we can say that an SU either has no relationship to another SU; that an SU is earlier than an SU or that an SU is later than an SU; or that an SU is the same as/ equals another SU. Harris reasoned that a schematic diagram would express this better than a profile drawing, because it could include all of the elements of a sequence. He called this diagram a Matrix, and published his thinking about matrices in a fundamental book. Principles of Archaeological Stratigraphy (London, 1979, 2nd ed. 1989: see also http://www.harrismatrix.com/). Others have suggested refinements, but the basic principle is Harris’s.

Every single SU in the site can then be positioned in a single matrix showing its stratigraphic relationships to everything else. In any matrix the SUs (and SSUs and HRUs) form stratigraphic sequences, but often a lot of these chains do not intersect. Walls, erosion, or other events interrupt the overlapping and cutting of the layers. We can nonetheless divide the SUs into phases, using their relative chronology (their position within the sequences) or absolute chronology (determined by diagnostic materials or other archeometric techniques, like C14).

We generate mini-matrices each time we fill in a scheda, noting the relationships of each adjoining SU. When inputting data into ARK, too, we record the stratigraphic relationships between SUs in the matrix diagram.

In theory, we could do away with section drawings altogether and simply express all stratigraphic information through Harris matrices, so that floors, occupation debris, and construction debris can be distinguished. In fact, there are computer programs designed to create the matrix for you. In practice, however, we do not abandon section drawing and we do not use computer programs to make our matrices, for two reasons. First, the matrix still conveys less information than a section drawing: it is much more abstract than a profile, which actually shows you graphically what the stratigraphy looks like. A good section drawing has power to communicate the character of the archaeology. Secondly, making a good section drawing is an analytical process in itself. It is the last thing to be done to process the stratigraphic data. The same thing holds for a matrix.

The construction of a matrix begins by showing all units covered by ploughsoil that are not covered by anything else. This is because the only relevant relationship that exists is that between the latest deposit and those that immediately precede it. Thus a layer from which a pit was cut will be found directly under the SU of the cut, but not any earlier deposits cut by the pit or any physical contact which exists between the pit fills and the earlier deposits.