Over the years archaeologists have developed a number of different techniques for seriating artifacts and assemblages. The authors of OptiPath have developed a method of seriation, optimal path seriation, that subsumes most of these techniques. By setting the parameters appropriately, a user can effectively implement a number of popular techniques. It can be quite complicated to set the parameters correctly for each technique, especially for someone new to seriating (and even experts will disagree), so we have created a number of default settings for a few techniques. Technique can be set in the Seriations table.
If you choose a default setting, such as Occurrence seriation, OptiPath will automatically set the seriation and item parameters accordingly. You will not be able to change these setting as long as you retain the Occurrence seriation setting for Technique. However, we cannot recommend that you use these default settings for all situations of occurrence seriation.
There is a way to have your cake and eat it too. Set Technique in the Seriations table to Occurrence. This will set parameters to their default values for occurrence seriation. Then set Technique to Custom. The default settings for occurrence seriation (or whatever technique you chose) will be retained but they are now able to be modified.
Optimal path seriation is not a specific technique; it is more a methodology that allows us to implement any one of a number of techniques. What they must have in common is that we are seeking the path through the items in feature space that minimizes a chosen objective. It is possible to conceive of additional objectives than the ones included in OptiPath (for example, the objectives used in component analysis).
Shortest Path Seriation
Shortest path seriation is a seriation technique based on minimizing the path length in order to maximize the gradualness of change (rather than unimodality) from one item to the next in the seriation. This is appropriate when the user has not restricted the dates on any individual item. When dates on individual items are restricted, an alternative objective, such as minimizing the average or maximum rate of change, is more appropriate.
Although most traditional seriation techniques maximize unimodality rather than maximize gradualness, the two are often the same (for example, occurrence seriation). An advantage of shortest path seriation is that it can work with object level data. That is, it can work with objective measurements of individual features of individual items, rather than with the often subjective assignment of styles and collections (assemblages).
For more information see Shortest Path Seriation.
Occurrence Seriation
The simplest implementation of unimodality is to keep track of when a feature or style is present and when it is not. It is assumed that once a style disappears, it never reappears. This requires defining styles in a way that ensures non-reappearance - namely they must be complex enough and non-functional enough that they are not likely to be reinvented by those unaware of their previous existence. This generally means that defining styles is somewhat objective and restrictive (because of the non-functional requirement). (The authors feel this is a weakness of occurrence seriation and an argument for working with elemental and objectively measurable features rather than complex and somewhat arbitrarily chosen styles).
We consider occurrence seriation to be a special case of shortest path seriation. A seriation that minimizes path length would be an optimal occurrence seriation, for any break in the occurrence of a style followed by a later reappearance would necessarily increase the length of the path. Note that a perfectly unimodal path would also be optimal for discrete seriation.
For more information see Occurrence Seriation.
Nominal Seriation
Nominal seriation is a generalization of occurrence seriation. Instead of categorizing each feature into two classes (present or absent), the technique allows any number of classes. If the classes are numbered, it does not imply that items should be seriated in order of their class number, only that items within the same class should be consecutive in the seriation (see Classed Data).
Nominal seriation is a special case of shortest path seriation.
For more information see Nominal Seriation.
Discrete Seriation
Discrete seriation can be considered as an extension of nominal seriation. Rather than an unchanging style that is present for an interval and then disappears, never to reappear, we can consider a style that itself evolves over time, yet once it disappears, it still will not reappear. Rather than a continuous evolution of style, discrete seriation, as its name implies, considers a discrete evolution of a style from one stage to another.
If those discrete stages are numbered integrally in the order in which they are presumed to appear, then an optimal seriation would be one in which all of the stage 1 items are followed immediately by the stage 2 items, and then the stage 3, and so on.
We see discrete seriation as a special case of shortest path seriation. In this case the path that minimizes path length would be an optimal discrete seriation, for any item out of order would necessarily increase the length of the path. Note that a perfectly unimodal path would not necessarily be optimal for discrete seriation as the stages could increase for a while and then decrease.
For more information see Discrete Seriation.
Similarity Seriation
Similarity, or proximity, seriation is a technique in which a measure of similarity is assigned to each pair of items. The items are then seriated so as to keep consecutive items as similar as possible. The algorithms that are used are often restrictive and do not lead to as globally optimal a solution as shortest path seriation.
We consider similarity seriation to be essentially equivalent to shortest path seriation, the differences being the method of assigning degrees of similarity (not necessarily using a metric in feature space) and the algorithm for finding a best possible seriation. OptiPath does not provide default settings for similarity seriation beyond using Shortest Path seriation.
Frequency Seriation
Frequency seriation is often used when working with assemblages. Rather than measuring feature attributes directly, one measures the frequency with which a particular style appears in each assemblage. For example, if 30% of the artifacts in an assemblage were style 1, then feature 1 for the assemblage would have a value of 0.3.
The usual goal in frequency seriation is that a style will appear in the archaeologicalrecord, its frequency will increase continuously for a while and then decrease and finally disappear. In other words, it is expected that the frequencies will be for each unimodal (feature). However, one could also argue that we would expect the frequency to change gradually over time without necessarily being perfectly unimodal.
Frequency seriation is a special case of shortest path seriation.
For more information see Frequency Seriation.
Custom Seriation
Custom seriation is not a well defined seriation technique. It is simply a chance for the user to define his or her own seriation technique by adjusting the available parameters in OptiPath. Regardless of those settings, OptiPath will find a best seriation by looking for an ordering that optimizes the chosen objective.
For more information see Custom Seriation.