Dendochronolgy, or the dating of wooden objects by analysis of their annual growth rings, is a technique of great potential which has only just begun to be exploited in Britain, and is therefore still in its tentative stages, particularly for certain archaeological periods. The method depends on four criteria: definition of the growth rings as annual; fluctuation in the amount of growth year by year; the availability of particular species from buildings or excavations; and finally the extension the sensitive growth pattern over a sufficient geographical area to allow correlation between sites. Oak is the major British wood failing into these four categories but many other species are suitable and some, such as ash, have been used with success in small areas. Conifers, though rarely found in excavations here, have been widely used in Europe and the USA, but tend to produce false or double rings and in some years no growth at all, causing difficulty in calendar dating.
The greatest potential of tree-ring analysis lies in the last two millennia. By reference to a unique master series of annual ring widths (which would start at the present day and extend back through overlapping earlier sample sequences) it will eventually be possible to date absolutely many wooden structures and artefacts, a feat no other dating system has yet achieved. Until such a master sequence is created, and in dealing with the prehistoric period when man was using wood, ‘floating’ tree-ring sequences - fixed at neither end but spanning a definable number of years - can be established. Existing dating methods (e.g. historical records of construction for medieval timber, and, for wood of any age, radiocarbon [C14] analysis, whose deviations are now well recognised and can be overcome to some extent) can then be used to give fairly accurate calendar dates to the sequences.
Oak, as has been indicated, is particularly suitable for tree-ring work. It is one of a group of ring-porous species (including ash and elm), so called because the spring wood consists of lines of large thin-walled vessels while the summer wood is a band of thick-walled small vessels (fig 1). The ring boundaries are invariably annual, though in extremely slow-grown timber they may be so close together as to be almost indistinguishable. Another great advantage is the presence of sapwood, the outer active area of the trunk which is paler in colour (photo [L]) and grows at a fairly uniform width representing about 25 years. Thus, if a number of sapwood rings are missing through trimming or damage, it is still possible to estimate the year of felling and probably of construction. However, it is mote liable to insect attack than heartwood and was often removed from structural timbers.
In sampling and measuring tree-rings, sections are sawn transversely across the widest points of beams and planks, avoiding knots and taking care not to damage any sapwood present. From soft timbers, cores may be removed with a conventional forestry borer. Waterlogged samples are deep-frozen for about 24 hours; all can then be cut along the edge of one radius with a sharp knife, which exposes the ring boundaries clearly. Where difficult rings are encountered, two radii are plotted as a check on each other. The rings may then be marked off every five years and are measured with a small x10 lens which can be focused and which contains a 1/10mm scale, against which the ring boundaries are aligned. Each value is plotted on translucent logarithm- mic recorder paper and the points are joined. The heartwood-sapwood boundary must be marked in its correct position, and any other notable features or difficulties, such as the crossing of a ray during measurement, on either side of which the rings do not always correspond (fig 1). Individual samples must contain at least 50 annual rings if any accuracy is to be achieved as correlation for dating, although younger samples may provide alternative information. Individual ring sequences are then correlated into a mean sequence for the particular sited which can be compared to other sequences of similar or overlapping date. Correlation is first carried out visually on a light-table by overlaying two curves and moving them along against each other until the best match, if any, is found. As some method of objectively expressing the level of similarity is necessary, the relative changes of the growth pattern are subsequently analysed by computer, which prints out a correlation coefficient in percentage form for the best matches under columns of statistical significance. This is random at 50 percent but increases as the level of agreement is improved, to acceptable values of over 50 percent. Agreement is based on the alignment in particular of narrow rings and of ‘signatures’, or groups of rings which fluctuate in a significant way and occur in the majority of samples (e.g. fig 2, years 155-160). The overall trend of rising and falling widths may sometimes be of help, but it may be indistinguishable from the natural trend of age, which is from wide rings to narrow ones.
Some of these points may be illustrated by analysis of waterlogged materials in York, such as oak planks and posts from the Anglo-Danish levels below the Coffee House in Pavement (INTERIM vol l nos 1 & 2) and worked timbers found in roadworks in cooperate, thought to belong to the same features found by Radley (Medieval Arehaeology, 1971). Thirty samples from the Pavement material (reduced to 14 after rejecting those with very vide and uniform rings) resulted, after processing, in ring sequences of between 37 and 151 years, with seven still retaining some sapwood (fig 3 lower half, and photo[R]). The best correlation occurred between three samples from Trench II with a value of 76.4% percent, providing a mean tree-ring sequence of 180 years - a sufficient length to allow radiocarbon dating of the level from which they came. Accordingly, three samples of 20 annual rings each were removed from years 40 to 60, 90 to 110 and 140 to 160 (i.e. at 50-year intervals), and submitted for analysis, the results of which are pending. They will provide three calculators of the final felling date of the timber, when various adjustments have been made to account for the 50-yr intervals and the loss of some sapwood; and will also give some indication of the deviations of radiocarbon years from calendar years.
Further samples from trenches I, II & IV have given information on the time which elapsed between the construction of the superimposed timber buildings, although the possibilities of reuse of old timber or the insertion of new must not be forgotten. The intervals are roughly as follows:
I level 12 to II level 21 – c. 85 yrs
II level 23 to II level 21 – c. 10-15 yrs
II level 30 to II level 23 – c. 45 yrs
II level 31 to II level 30 – c. 20yrs
II level 32 to II level 31 – c. 35yrs
IV level 32 to IV level 31 – c. 30-70yrs (4 samples of 31 span 40 yrs)
II level 32 to IV level 32 – 15-20 yrs
II level 23 to LV level 32 – c. 100yrs
These figures are based on a sapwood addition of about 25 yrs, which is accurate where some sapwood is preserved. In the cases where it is absent, a number of heartwood rings may also have been removed.
Nine sections from Coppergate could be correlated into a tree-ring sequence of 177 years (fig 3, upper), and several were evidently split from the same tree, as shown by their almost perfect agreement. Two have sapwood remaining and indicate at least three felling periods, the earliest roughly equivalent with II level 32, the second between II levels 23 and 30 and the latest prior to level 23. The mean curve agreed well with two Pavement samples with values of 66 percent, which has led to the interpretation that they are contemporary, and date from the Anglo-Scandinavian period.
The results of this analysis have provided a floating tree-ring sequence floating somewhere in the 8th to the 10th centuries, which it is hoped will be fixed by C14 dating, and will in the future make possible the dating of Saxon and Roman timbers in the Northeast.
Ruth A. Jones
University of Sheffield