Classification Presupposes Type Definitions

Andreas Oldeberg (1892-1980) is rumoured to have had some pretty ugly political leanings. But just because you like cheese, you needn't socialise with cows. If you're into Late Neolithic and Early Bronze Age metalwork from Sweden, there is absolutely no getting around Oldeberg's huge illustrated catalogue from 1974.

I'm currently grabbing data out of the catalogue for my sacrificial sites project. And I've come across a funny detail that shows that old Oldeberg was not up to speed with his day's archaeological methodology.

Whenever Oldeberg describes a spearhead, he classifies it according to a fairly new piece of work at the time, Gernot Jacob-Friesen 1967. This scholar named his types for find spots, such as Valsømagle, Bagterp etc. But quite often, Oldeberg isn't sure what type a certain spearhead belongs to. He'll say wahrscheinlich Typ Bagterp, "probably Bagterp type", for instance. This is fully understandable if you only have part of a spearhead: the distinctive characteristics of various possible types may not be extant on the bit you've got. But Oldeberg does this all the time with complete objects. And that makes no sense whatsoever after 1962.

In his 1962 dissertation, Jungneolithische Studien, Mats P. Malmer established that an object type's identity rests entirely upon a verbal definition. Material, dimensions, proportions, decorative details: a scholar must tell her readers clearly what the rules are for inclusion and exclusion in a type, or it isn't a type. Do feel free to illustrate the definition with pictures of objects that belong to the type in question, but don't ever think that it's enough to say (as Jacob-Friesen did) that "Type Valsømagle is spearheads like figs. 1-3". Because that doesn't tell the reader what characteristics specifically make those spearheads members of type Valsømagle. And it doesn't tell the reader what sort of variation is permissible within the confines of the type you're proposing.

So the reason that Andreas Oldeberg often couldn't tell what type a well-preserved spearhead belonged to was that Jacob-Friesen's classification scheme is completely flawed and contains no stringent type definitions. Oldeberg could see that a given spearhead looked kind of like the pictures of Jacob-Friesen's "type" Bagterp, but he had no way of telling whether Jacob-Friesen would have accepted it as a member of the type. Because Jacob-Friesen's work does not contain instructions for how to make that call.

More like this

Mats P. Malmer in 1989, holding a miniature replica of a Bronze Age sword. Photograph by Dr Rune Edberg, published with kind permission. Yesterday, 18 October, was Swedish archaeology professor Mats P. Malmer's 86th birthday. Sadly he passed away on 3 October. I wrote a brief appreciation when I…
Spring is late this year in Sweden, and the weather has been dreary. But now things have perked up, and suddenly I felt the itch to get out and check out some sites before the leaves and grass sprout in earnest and ruin visibility. So Sunday night I hurriedly checked through my database of Bronze…
Professors tend to have a few pet issues that they emphasise time and again over their careers as researchers and supervisors. This is quite clear with two 1960s-70s professors in my field. In Bertil Almgren's case, one such pet issue was the source-critical quality of archaeological information.…
In the Lake Mälaren area of Sweden, you rarely find any large pieces of Bronze Age metalwork in graves or at settlement sites. When the beautiful larger objects occur - axe heads, spear heads, swords, neck rings, belt ornaments - they almost exclusively come from odd find contexts that I for one…

Computers?
Scan a photo of a spearhead, and let a program measure the salient proportions.
Scan a hundred typical spearheads to capture the morphological spread.
Distill into a couple of paragraphs.
--- --- --- ---
This is how it might work in practice: use a laptop with attached GPS during excavation. When a spearhed is uncovered, log exact position into laptop (along with positions for all other items).
When the site is dated, compare it wth other sites in region and get a map showing (for instance) spearheads in space and time.
For each site, you have a diagram showing variations of spearhead morphology, and can see if there are clusters related to different styles.
In theory, you should be able to spot the migration of new spearhead styles across the region, if you have enough data points across the region and across the time periods.
Yes, I know that in practice it will be much more complicated.

By Birger Johansson (not verified) on 30 Jun 2011 #permalink

Automatic pattern recognition has not to my knowledge been implemented. But multivariate stats are often used these days to find clustering in the data. If you record a bunch of parameters for those spearheads and look at the data with correspondence analysis you will see which ones clearly belong together as a type and which ones are outliers. You may also see gradation, which may be due to typological drift over time.

I always thought scanning would be a good way to identify clusters in Native American prehistoric projectile point types. But in CRM archaeology in the southeastern U.S. "type cluster" is usually defined by looking at them and recognizing difference in appearance from experience. Some people will even go by co-occurence with easily established types being used to date an archaeological context and other points being determined to be a type based on their assumed age. This strikes me as analyzing the artifact based on the site when it should be the other way round. However, one does have to demonstrate one's expertise by naming as many artifacts as classic types in CRM archaeology.

Some people will even go by co-occurence with easily established types being used to date an archaeological context and other points being determined to be a type based on their assumed age.

Haha, that's just priceless! "I don't know how this arrowhead type is defined or even if it has a definition, and I have no real idea what it looks like, but I do know that it's supposed to go together with the easily identified Hacklebush Trapezoid". Soooo bad.

Clustering spearheads with computers will not generate a one true set of different groups, but rather a n-dimensional set of possible clusters/groups, with varying usefulness. The act of re-creating a sensory/classification system highlights of course the problem of there not actually being any real categories in the real physical world, to the extent of what we can speculate about it anyway.

Anyhow, this clearly calls for some data mining and the inevitable phone-app for instant classification! Sounds like a perfect project.

I like that there's an age axis to the problem, and not just the clustering btw - that decreases the level of navel gazing should a detection system be devised.

There is whole "sub-discipline" devoted to this. Morphometrics. Google this with archeology!!!!

By John P. Cook (not verified) on 14 Jul 2011 #permalink