Bias in the PAS Database: The Case of Annular Brooches

Over the course of my research into fifth- and sixth-century Anglo-Saxon brooches, I’ve noticed some substantial disparities between metal-detected and archaeological sources of data.   As the Portable Antiquities Scheme becomes an increasingly popular mode of studying the past (the PAS website currently lists 366 projects), we would do well to study the means by which these data are produced and their inherent biases.

The most signficant bias I have come across in the PAS database is its massive over-representation of cruciform brooches compared to the pitiful number of annular brooches it has recorded.   Cruciform brooches and annular brooches make a good comparison because they both have the same distribution in eastern England.  The vast majority of both archaeologically excavated and metal-detected brooches are also likely to originate from the same contexts: graves, disturbed or otherwise.

Figure 1: Cruciform brooche (left), annular brooch (right). Drawings by T. F. Martin.

Figure 1: Cruciform brooch (left), annular brooch (right). Drawings by T. F. Martin.

A simple search on the PAS database reveals 686 examples of cruciform brooch (plus a further 78 fragments that are likely to have originated from cruciform brooches, but which lack diagnostic features) and 101 annular brooches (including a handful that might be of a slightly later 7th century date).  I’ve provided detail on the searches I used at the bottom of the post, so that these results can be repeated and verified.

I compared these results from a sample of nine excavated and published cemeteries from eastern England: Castledyke South and Fonaby from Lincolnshire; Morning Thorpe, Westgarth Gardens and Bergh Apton from East Anglia; Empingham II and Broughton Lodge from the East Midlands; Norton and Sewerby from the Yorkshire and the North.  The excavated cemeteries produced 390 annular brooches and only 147 cruciform brooches.  As can be seen from the histogram (Figure 2), the proportions of these brooches are virtually reversed depending on whether they were excavated or metal-detected.

Figure 1: Comparing the frequency of cruciform and annular brooches in excavated and metal-detected samples.

Figure 2: Comparing the frequency of cruciform and annular brooches in excavated and metal-detected samples.

This massive difference might be partly explained by an over-representation of cruciform brooches: they are larger objects and tend to break up into more pieces in the plough soil.  Being more substantial, they might also produce a stronger signal to a metal detector.  However, even these factors cannot account for this degree of bias: much smaller items than fragments of annular brooch are frequently found on the PAS, including cruciform brooch knobs.  Additionally, annular brooches frequently break up into several pieces too, albeit less substantial ones.  The most promising explanation in my mind is that annular brooches are actually found far more regularly than the PAS database suggests, but they are only rarely reported to Finds Liaison Officers.  Annular brooches are distinctly plain objects and might easily be mistaken for a piece of bronze washer.  I imagine though, given the knowledge and experience of most metal-detectorists, that these object are actually regularly recognised, but just not deemed worthy of reporting given their dull appearance.

For the early Anglo-Saxon period, such biases are not devastating because we can check the excavated and metal-detected data against each other.  Nevertheless, this very basic analysis reveals the severe risks of working only from metal-detected data.  The impact of different taphonomic processes and finds reporting practices is considerable.  Research based on just one of these sources would produce the opposite findings.

Please do comment below if you have seen similar biases in other material. Any alternative explanations of this problem from either the perspective of the PAS, metal-detectorists or archaeologists are also most welcome.

Portable Antiquities Scheme Data Collection

The search was done using only finds whose records are publicly available.  Search terms were as follows:

  • Object type: “BROOCH”
  • Object description contains: “cruciform” / “annular”
  • Broad period: “EARLY MEDIEVAL”
  • Sub-period from: “early”
  • Sub-period to: “early”

The search results were then limited to just objects with accompanying images, so that the identification of the object could be checked.  Each sample was then scanned through manually to remove any stray items that also contained the term “cruciform” or “annular” in their object descriptions.  These totals represent substantially less than is actually present in the online database, but as the same standard search terms were used to obtain both samples, the data is unlikely to be skewed at this level.  Biases in PAS search terms is a subject I’m going to address in a future post.

Cemeteries Sampled

  • Cook, A. M. (1981) The Anglo-Saxon Cemetery at Fonaby, Lincolnshire. Sleaford: The Society for Lincolnshire History and Archaeology.
  • Drinkall, G. and Foreman, M. (1998) The Anglo-Saxon Cemetery at Castledyke South, Barton-on-Humber. Sheffield: Sheffield Academic Press.
  • Green, B. and Rogerson, A. (1978) The Anglo-Saxon Cemetery at Bergh Apton, Norfolk. Gressenhall: The Norfolk Archaeological Unit.
  • Green, B., Rogerson, A. and White, S. G. (1987) The Anglo-Saxon Cemetery at Morning Thorpe, Norfolk. Gressenhall: The Norfolk Archaeological Unit.
  • Hirst, S. (1985) An Anglo-Saxon Saxon Inhumation Cemetery at Sewerby East Yorkshire. York: York University Archaeological Publications.
  • Kinsley, A. G. (1993) Broughton Lodge: Excavations on the Romano-British Settlement and Anglo-Saxon Cemetery at Broughton Lodge, Willoughby-on-the-Wolds, Nottinghamshire, 1964-8. Nottingham: University of Nottingham, Department of Classical and Archaeological Studies.
  • Sherlock, S. J. and Welch, M. G. (1992) The Anglo-Saxon Cemetery at Norton, Cleveland. London: Council for British Archaeology.
  • Timby, J. R. (1996) The Anglo-Saxon Cemetery at Empingham II, Rutland. Oxford: Oxbow.
  • West, S. E. (1988) Westgarth Gardens Anglo-Saxon Cemetery Suffolk: Catalogue. Bury St Edmunds: Suffolk County Planning Department.
Advertisements

12 Comments on “Bias in the PAS Database: The Case of Annular Brooches”

  1. Rosie Weetch says:

    I have similar biases. The most obvious is the lack of iron brooches on PAS despite them being recovered in huge numbers during excavations at Flixborough. Most metal-detectors can be set so they don’t pick up iron, which I guess is what is happening rather than there being a very localised fashion for iron brooches at a single settlement in Lincolnshire!

    I recall reading a study on the difference between excavated and metal-detected early Anglo-Saxon brooches in Kent. It suggested that this revealed a difference in dress in life (the rural metal-detected finds) and in death (excavated finds from cemeteries). Perhaps a similar thing is going on with your cruciform and annular brooches – Cruciform brooches are a common everyday item, but not necessary appropriate for burial, and the annular brooches were more like to be part of a costume for the deceased? Not sure how much this argument can hold up in reality though. Can dig out the reference if you haven’t come across it.

    • Toby Martin says:

      I have come across that idea before I think, although as far as know most evidence suggests that the vast majority of PAS early Anglo-Saxon jewellery originates from disturbed cemeteries – they tend to occur in small groups, and this jewellery is extremely rare from settlements. I imagine some bits of cruciform brooch represent casual losses though – they had a tendency to fall apart it seems, as quite often bits of them are missing in graves! I’d be interested to know that Kent reference though

      • Mary C-K says:

        As far as I know, it’s still unpublished, but I may be wrong. I was speaking at the same conference so remember it well.

        MCLEAN, L. & A. RICHARDSON. 2007. Early Anglo-Saxon brooches in southern England: the contribution of the Portable Antiquities Scheme. Paper given at the Portable Antiquities Scheme conference, 17-18 April 2007, London.

      • Rosie Weetch says:

        Yes, it is that talk, and was subsequently published in this volume:

        Worrell, S., Egan, G., Naylor, J., Leahy, K. and Lewis, M. J. (2010) A Decade of Discovery. Proceedings of the Portable Antiquities Scheme Conference 2007. British Archaeological Reports, British Series. Oxford: Archaeopress. 520

        It actually uses the difference between cruciform and annular brooches as a case study so well worth a look.

      • Toby Martin says:

        Thanks v much both! I’ve never come across that volume.

  2. Judith Plouviez, Suffolk Co Co says:

    The discrepancy also occurs with Roman brooches – penannulars are rarely found by detectorists, see especially Hacheston Suffolk (E Anglian Archaeol no 106, 2004). I did a little research with detector users a few years ago and there may be a problem with certain shapes being less easy to find with a detector, as well as the fragments being less identifiable. Hilary Cool also noted that Roman toiletry items (scoops, tweezers) are under-represented in detected collections, and there may be similar reasons for this.

    • Toby Martin says:

      Thanks for the comment – it would be quite interesting that different shapes alter the signal, and may go some way to explaining the discrepancy too I think. I wonder if there might also be an interesting study to be done along these lines comparing Roman vs AS brooches given their typically different archaeological contexts.

  3. Toby Martin says:

    Looking into this a bit more, I’ve just noticed Mary Chester-Kadwell presents very similar statistics with considerably more detail and insight in her book “Early Anglo-Saxon Communities in Landscape of Norfolk” (2009, BAR British Series 481). Here the discrepancy is explained by a higher rate of fragmentation for cruciforms, as well as sorting processes in the soil through which fragments of a particular shape and size are more likely to rise to the top of disturbed soil (p.73-78).

  4. […] pursuing this I must confess a debt of inspiration to an excellent post by Toby Martin on his These Fragments blog examining the contrasting recording of cruciform and annular brooches through P… – the sites’s gone a bit too quiet in the past couple of months, get back blogging […]

  5. Rob says:

    Penannular brooches are the easiest shape to find with a metal detector and far less likely to break up in the ground due to ploughing and the use of other farm machinery. Cruciform brooches I have found bent in many shapes but rarely broken unlike square headed brooches are nearly always broken due to being made thinner and flatter with chip carving applied. Modern farming today destroys most artefacts, saxon graves are not always deep but vulnerable.

  6. […] cruciform brooches are the most common.  I’ve explored the potential bias toward these items in a previous blog.  The sheer size of the gap between cruciform brooches and most other types, however, is still […]

  7. The first thing that occured to me was that some of the Anglo-Saxon brooches are being misclassified as later medieval not early medieval, but since I am interested in late medieval not early and don’t know what the early ones look like, that is likely wrong.

    I’ve had similar issues with copper alloy and lead alloy buttons and suchlike in the later medieval period, and of course lead spindle whorls. I posted about it here:
    http://distillatio.wordpress.com/2013/09/29/the-randomness-of-surviving-artefacts-and-re-enactors-having-the-wrong-stuff-medieval-buttons/
    Although re-reading it I see I didn’t mention lead spindle whorls – the point with them being that hundreds turn up on ebay, quite a few in the PAS, and as far as I have been able to find, one or two in all the towns excavated. So dating them is a bit hard, and the point there is that like with the buttons, there’s a difference between town and country, excavations and random metal detector finds.
    That’s what is so great about the PAS, it gives us large enough numbers to start making reasonable hypotheses, even if a full resolution is unlikely.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s