Nearest neighbour analysis is a statistical tool for measuring the degree of clustering, randomness or uniformity in a spread of geographical data. By measuring the nearest distance between every stone row it is possible to provide an accurate indication of the amount of clustering or otherwise whilst permitting comparisons between different areas. Analysis has indicated considerable variations between different regions and countries with some being very clustered whilst others have a random distribution. The most clustered region is Northern Scotland whilst at the other end of the spectrum is Bodmin Moor where the distribution is verging on uniform. Between these extremes is the North Yorkshire Moors where the distribution is perfectly random. At a national level England unsurprisingly has the most clustered distribution whilst Wales is less so although even here there is a strong tendency towards clustering. A graph showing the variation is presented below.
Figure 1. Graph showing the results of nearest neighbour analysis for the different regions and countries
Discussion
The distribution of the stone rows in the different regions and countries varies considerably. There are probably many reasons for this and whilst differences in the original distribution almost certainly contribute it is unfortunately not possible to entirely eradicate the fact that differential survival may have played a part. This said the differences are so marked that we are almost certainly seeing an abraded version of an original pattern and confirmation that the rows should not be viewed as a single site type with the same function. If this was the case one would expect the distributions to be similar. The prominence of clustering suggests that rows were deliberately positioned in certain areas and that perhaps they attracted others to those areas. The most marked clustering is in England as a whole and this reflects the known concentrations on Dartmoor and Exmoor together with the absence of rows from large swathes of the middle and east of the country. The region with the most clustering is Northern Scotland where most of the rows are found together in groups. Indeed in this region solitary rows are relatively uncommon and 18 of the 31 rows are within 500m of another one. The most remote row (Loch Rimsdale) is less than 7km from its nearest neighbour (Cnoc Molach). Given the small size of the stones, afforestation of the area and peat accumulation it is likely that Northern Scotland has seen greater loss of sites than many areas and it is not inconceivable that all of the rows originally formed discrete tight clusters and that the individual rows are more the result of survival than the original disposition. If this is the case then it is possible that we should see the rows in this area as forming part of ritual complexes each with more than one stone row. Northern Scotland also has the greatest number of multiple rows and it therefore no coincidence that multiple rows have the most clustered distribution (0.012) of all types of row (Figure 2).
Figure 2. Graph showing the results of nearest neighbour analysis for the different types of stone row. All types of row are clustered but it is clear that the single rows are less clustered than the more complex examples. It is perhaps significant that the most complex type also has the most clustered distribution.
At the other end of the scale the stone rows on Bodmin Moor have been laid out in a much more uniform pattern. In this region there is no clustering of rows and they are all positioned individually at some distance from the others. This distribution is so different from that found in other areas that there must almost certainly be a contemporary explanation. The stone row builders on Bodmin Moor appear to have placed their rows in a relatively uniform way across the landscape and on no occasion (as far as we can tell) was a second row ever built close to an existing one. The only other regions where closely spaced pairs of rows are not currently known are the North York Moors which Nearest Neighbour Analysis indicates a perfectly random distribution and Central Scotland which is close to random. On Bodmin Moor it is highly likely that the precise placement of the rows was determined by prominent landscape features such as Brown Gelly and Brown Willy and the more uniform pattern observed may be a direct consequence of this. Perhaps of more significance is the the observation that no two rows were built close to each other. Why on Bodmin Moor, the North Yorkshire Moors and Central Scotland were stone rows built in isolation whereas in other regions they are either frequently or sometimes erected in close proximity? This does imply an important difference between the rows in these two areas and combined with the fact that they are fairly evenly distributed they may be seen as denoting separate territories of some form.
The region with the second most clustered distribution is Dartmoor and this is in no small part due to the close proximity of the rows at Butterdon Hill, Drizzlecombe, Fernworthy, Merrivale, Sharpitor and Shoveldown as well as the frequent pairing of rows at places such as Collard Tor, Lakehead Hill, Laughter Tor, Tottiford, and Trowlesworthy. Only 34% of the Dartmoor rows are more than 500m from another row and half are less then 300m from their nearest neighbour. Dartmoor represents the densest concentrations of rows in Great Britain and they are also significantly clustered within the region. This compares dramatically with adjacent Bodmin Moor where the distribution is much more uniform. Another difference between Dartmoor and Bodmin Moor is that 94% of the cists on Dartmoor are orientated north west to south east whilst on Bodmin Moor the orientation is entirely variable (Butler, J., 1997, 176). It may therefore be that we are seeing evidence for two separate groups of people living close to each other, sharing many customs and rituals, whilst at the same time doing certain things very differently. You need look no further than differences between contemporary religions for an apposite parallel.
The Western Isles and Brecon Beacons both have similar degrees of clustering within their distributions. Both regions have a mixture of closely connected rows and isolated rows. On Exmoor and in Mid Wales whilst having the occasional cluster of rows they are fairly evenly distributed. In Argyll and Isles and Central Scotland the clustering present in the other regions is rare and the distribution is now nearly random.
The differences in distribution between the different regions are very marked and confirm that the stone rows were erected in a dynamic world where different communities had distinct relationships with their rows. Precisely why the distribution pattern varies so much will inevitably be the subject of speculation and for the moment it is worth emphasising that these differences occur at a regional level. The differences are however not just confined to regions and Figure 2 illustrates that there is also a variation between types of row. Double, triple and multiple rows are generally found with or close to other rows whilst single rows are more commonly found by themselves or at some distance from others. Clearly this is an over simplification as there are many examples of single rows being found close to other rows, but Nearest Neighbour Analysis does demonstrate a marked tendency for them to be further away from other rows than the other types. The reason for this is currently unclear but it does imply a difference which in turn may be cultural and or functional.
Figure 3 Graph showing the results of nearest neighbour analysis for the different sizes of stones within stone rows.
Stone rows are composed of stones of different sizes. The rows composed only of small stones (minilithic) have the most clustered distribution and significantly the rows composed only of large stones (megalithic) have the least clustered distribution (Figure 3). This indicates that megalithic rows were more commonly erected further away from other stone rows than those composed only of small stones. This is an interesting detail and provides further evidence of the considerable differences between megalithic and minilthic rows. As well as looking very different they were also positioned differently relative to other rows.
Figure 4 Graph showing the results of nearest neighbour analysis for the different numbers of stones within stone rows.
There is considerable variety in the number of stones found at stone rows. Rows composed of three stones only are much less likely to be close to another row whilst those with more than 50 stones are amongst the most clustered. The difference is so great that this conclusion is likely to reflect the contemporary situation and represent evidence for yet another difference between these types of row. In part it is likely to be a result of the known clustered character of the multiple rows which generally have a greater number of stones.
Conclusion
Nearest Neighbour Analysis has confirmed that the distribution pattern in the various regions, of the different types, stone size and numbers is different. The significance of this is that it provides further evidence to support the idea that the distinct forms of rows were probably built for slightly different purposes by separate groups of people. The observed differences in distribution confirm the observed variation in stone rows and emphasise the rich and varied character of Late Neolithic and Bronze Age life.
Methodology
References
Butler, J., 1997, Dartmoor Atlas of Antiquities – Volume 5 – The Second Millennium B.C., Devon Books
Royal Geographical Society – A Guide to Nearest Neighbour Analysis
FIRST PUBLISHED: 6th June 2020
LAST UPDATED: 7th June 2020
The Stone Rows of Great Britain 