some thoughts on art-historical dendrochronology

23
Journal of Archaeological Science 1984, 11, 371-393 Some Thoughts on Art-historical Dendrochronology M. G. L. Baillie” Imported timbers pose potential problems for the dendrochronologist. The problem is general and any particular dendrochronologist may run into difficulty when he attempts to correlate an exotic ring pattern against an indigenous chronology. Sufficient evidence exists to suspect that high quality oak boards and planks were extensively imported into England in medieval times. However, chronologies cons- tructed from just such boards, namely the art-historical chronologies, have been treated as indigenous without being proven to be so. Almost identical comments can be made with respect to art-historical chronologies constructed by German workers using supposedly indigenous Flemish oak panels. This article explores how failure to allow for importation may well call into question the dating of these,art-historical chronologies and in addition demonstrates how an art-historical dating system has been created wherein erroneous placement of the chronologies could escape detection (this latter aspectmay well make the argument of general interest to scientists). Keywords: ART-HISTORY, BALTIC IMPORTS, DENDROCHRONOLOGY, EXOTIC TIMBERS, OAK PANELS, PANEL PAINTINGS, SAPWOOD ESTIMATES, Introduction For the successful dating of timbers, dendrochronology relies on the existence of suitable reference chronologies applicable to the species and the climatic area from which the sample to be dated derives. To take an extreme example, there would be little point in attempting to date an Irish-oak ring pattern against the American bristlecone pine chronology as logic suggests no possible basis for consistent similarities in tree growth for either the two areas or the two species. Such lines of thought lead to the construction of chronologies for each area, where area is defined loosely as the locality within which any particular dendrochronologist is working (or from which he is drawing his material) and within which it is assumed that cross matching between ring patterns will apply. As an extension it has become common practice where a chronology is constructed using timbers from a particular locality to name the chronology after the area (e.g. the Hamburg chronology) or the workers (e.g. the Huber-Giertz chronology) in this case referring to the chronology for south-central Germany. Obviously from a dating point of view it does not matter where a timber came from originally as long as a suitable chronology is available against which its ring pattern can be dated. Unfortunately it may not be possible simply by looking at a piece of wood to tell where it is from and hence to know which chronology would be applicable to its dating. Failure to choose the correct chronology may lead to com- “Palaeoecology Centre, The Queen’sUniversity of Belfast, Belfast BT7 INN, Northern Ireland. 371 030~4403/84/050371+ 23 rSO3.00/0 0 1984 Academic Press Inc. (London) Limited

Upload: mgl

Post on 02-Jan-2017

216 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: Some thoughts on art-historical dendrochronology

Journal of Archaeological Science 1984, 11, 371-393

Some Thoughts on Art-historical Dendrochronology M. G. L. Baillie”

Imported timbers pose potential problems for the dendrochronologist. The problem is general and any particular dendrochronologist may run into difficulty when he attempts to correlate an exotic ring pattern against an indigenous chronology. Sufficient evidence exists to suspect that high quality oak boards and planks were extensively imported into England in medieval times. However, chronologies cons- tructed from just such boards, namely the art-historical chronologies, have been treated as indigenous without being proven to be so. Almost identical comments can be made with respect to art-historical chronologies constructed by German workers using supposedly indigenous Flemish oak panels. This article explores how failure to allow for importation may well call into question the dating of these,art-historical chronologies and in addition demonstrates how an art-historical dating system has been created wherein erroneous placement of the chronologies could escape detection (this latter aspect may well make the argument of general interest to scientists).

Keywords: ART-HISTORY, BALTIC IMPORTS, DENDROCHRONOLOGY, EXOTIC TIMBERS, OAK PANELS, PANEL PAINTINGS, SAPWOOD ESTIMATES,

Introduction For the successful dating of timbers, dendrochronology relies on the existence of suitable reference chronologies applicable to the species and the climatic area from which the sample to be dated derives. To take an extreme example, there would be little point in attempting to date an Irish-oak ring pattern against the American bristlecone pine chronology as logic suggests no possible basis for consistent similarities in tree growth for either the two areas or the two species.

Such lines of thought lead to the construction of chronologies for each area, where area is defined loosely as the locality within which any particular dendrochronologist is working (or from which he is drawing his material) and within which it is assumed that cross matching between ring patterns will apply. As an extension it has become common practice where a chronology is constructed using timbers from a particular locality to name the chronology after the area (e.g. the Hamburg chronology) or the workers (e.g. the Huber-Giertz chronology) in this case referring to the chronology for south-central Germany.

Obviously from a dating point of view it does not matter where a timber came from originally as long as a suitable chronology is available against which its ring pattern can be dated. Unfortunately it may not be possible simply by looking at a piece of wood to tell where it is from and hence to know which chronology would be applicable to its dating. Failure to choose the correct chronology may lead to com-

“Palaeoecology Centre, The Queen’s University of Belfast, Belfast BT7 INN, Northern Ireland.

371 030~4403/84/050371+ 23 rSO3.00/0 0 1984 Academic Press Inc. (London) Limited

Page 2: Some thoughts on art-historical dendrochronology

372 M. G. L. BAILLIE

parison between two totally unrelated ring patterns, the result being, by definition, spurious. To reinforce this concept let us consider the following hypothetical exercise. A group of the Californian grown sequoia discs are collected from museums and institutions around the British Isles. It would almost certainly be possible to cross date these and form a floating clironology. The re_lative end-years of the individual samples would reflect the felling dates of the trees. As most of these specimens were acquired in the late 19th or early 20th centuries it would be possible to place the chronology roughly in time. However, without specific documentation relating at least one felling date to a particular specimen it would be impossible to date the relative chronology precisely in real time. In the absence of such evidence the only way in which this hypothetical chronology could be precisely dated would be by comparison with the pre-existing Californian sequoia chronology constructed by Douglass ( 19 19). It would be impossible to date the floating chronology against the ring patterns of sequoias which have been growing in the British Isles since the 19th century. The species may be the same but the conditions controlling growth would have been totally different, making the area factor insuperable. This rather laboured example is given simply to reinforce the idea that when attempting to date a section of chronology there must be good grounds for choosing the reference chronology against which dating is to be attempted. There is no point in hoping that the two will be related merely because they are derived from timbers of the same species. This example sets the scene for the discussion of a group of oak chronologies constructed by workers in Germany and England, namely the Netherlands II (Eckstein et al., 1975) and the English Type A (Fletcher, 1978~) art-historical chronologies. In brief, workers in both Hamburg and Oxford have measured the ring patterns of large numbers of oak panel paintings and some other objects such as chests, cupboards and carvings and have constructed relative chronologies. Both sets of workers have assumed original sources for their material in the Netherlands and England respectively and accordingly have attempted to cross-date their art-historical chronologies against pre-existing absolutely dated chronologies in the Netherlands, Germany and the British Isles. In both cases attempts to establish the precise date range of these floating chronologies have run into difficulties and the question is raised whether any justification exists for attempting to date the art-historical chronologies in this way.

The Netherlands II chronology has never been published, for reasons given below. However, in 1977 a series of tree-ring chronologies was presented in this Journal (Fletcher, 1977). These reference chronologies, namely Refs 1-8, spanned the period from the 5th to the 16th century AD and could be divided into two clear units. The older chronologies, Ref 6 securely dated to 782-l 193 and Ref 8 subsequently dated to 416-737 (Baillie, 1980), were formed principally of archaeological timbers and could be assumed to represent indigenous British Isles material. The younger chronologies, Refs 1-5, were what have come to be called the Type A-art-historical-chronologies and were claimed to span 1120-1609.

The purpose of this paper is to draw attention to the continuing question which hangs over the provenance of the timbers used to form both the Type A and Netherlands II chronologies and as a logical extension the question which remains over their dating. Ideally the chronologies should be correctly sourced and correctly dated-nothing less is acceptable in the long term if deductions on past climate, ecological zones or dating are to be drawn from the ring patterns. If they are correctly dated but not indigenous then they can be used for dating but not for climatic/ecological inference. However, there is at least some suggestion that if they are not indigenous then by extension they may not be correctly dated. Such a possibility has not been seriously considered by those working in the field.

Page 3: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 313

1400 1500 1600 1im300 I900 1980 I I f

- t t t Beltast n t t I I I I I t=a43

Figure 1. Correlation values between modern and medieval Type H chronologies showing total consistency, i.e. each chronology can be dated against any other. Belfast (Baillie, 1977); Sherwood (Baillie & Pilcher, unpubl.); BritishIsles IS-site (Baillie, 1982); Giertz (Siebenlist-Kerner, 1978); Yorks (Hillam, 1980); Sheffield (Morgan, 1977); MC10 (Fletcher, pers. comm.).

The Background In the British Isles absolute oak chronologies were successfully constructed back from modern anchor dates through historic building timbers to archaeological timbers in the north of Ireland (Baillie, 1974, 1977a) and in Scotland (Baillie, 19773). Working within the small geographical unit of northern Ireland the potential problem of non- indigenous material was minimized for the Belfast chronology. Historically there was little reason to suspect substantial importation of timber before the 18th century- even then the vast bulk of imported wood was pine-but more importantly chrono- logies constructed from historic and archaeological timbers could be checked against the certainly indigenous sub-fossil oak chronologies. These naturally preserved oaks were found to range in age from at least the 6th millennium BC to as late as 1600 AD.

Within Britain other floating chronologies based on historic and archaeological timbers have been successfully cross dated against the Irish and Scottish chronologies and/or against the pre-existing German chronologies of Hollstein (1965) and Huber & Giertz (1969). Examples within the British Isles are given in Figure 1. All of these chronologies form a highly integrated group and the levels of consistent cross-dating are borne out in studies on recent oak chronologies within the British Isles (Baillie, 1982, 1983~) and earlier 1st millennium AD material (Baillie, 1980, 1982). In 1974 the first English art-historical chronology, MC18, was published for the period 1230 to 1546. Dating was claimed to be

“ . by region-to-region bridging from the Huber and Giertz-Siebenlist (1969) curve for central and southern Germany . . .”

In addition it was found to match directly with a pre-existing but unpublished art- historical chronology based on panel paintings by Rubens at Antwerp, constructed by Bauch (Fletcher et al., 1974).

As more chronologies became available it became apparent that MC18 did not fit comfortably into the scheme of British Isles chronologies constructed using building timbers (Baillie, 1978). For example, the only high correlation values between the Belfast chronology and MC18 were for end-dates of 1530 and 1550, the latter end-date

Page 4: Some thoughts on art-historical dendrochronology

374 M. G. L. BAILLIE

being used for a short time by the Oxford workers (Fletcher, 1975) before the 1546 end-date was reverted to in 1977 (Fletcher, 1977).

The inconsistency of MC18 within the family of British Isles chronologies was ameliorated to some extent by the statement that MC18 had

“been derived from oaks used between 1470 and 1570 in south-east England or for a few in Flanders or northern France” (Fletcher et al., 1974).

At least some element of non-British material was clearly allowed for in MC18 and this allowed other workers to assume that the differences between the art-historical chronology and the other British Isles chronologies might be due to the use of imported material by 15th and 16th century artists.

In 1975 publication of results of research in the Netherlands by the Hamburg group (Eckstein et al., 1975) threw some light on the art-historical chronologies. An absolute chronology, Netherlands I, constructed from living trees and historic timbers spanned 1385 to 1972 and showed strong similarity to the existing German chronologies. However, an art-historical chronology based on 151 paintings and two sculptures, Netherlands II, was noted as being independent of the Netherlands I chronology. Indeed Netherlands II was stated as having

“no marked similarity to those of the Dutch mills and paintings of (Netherlands) chrono- logy I, or to any of the neighbouring German chronologies”.

So here again workers were confronted with an art-historical chronology which did not sit comfortably with the pre-existing and widely applicable continental absolute chronologies. Some of the information supplied by Eckstein tit al. (1975) was important in grasping what was coming to be regarded as an art-historical problem. For example, the latest-known felling date of a Netherlands II type oak was around 1650. Paintings later than this date were consistent with Chronology I ; conversely, attempts to find Chronology II type wood in late-17th century or later buildings failed. ,There appears to be a cut-off date for the use of this Type II wood. In an attempt to explain the disappearance of Type II it was suggested that overexploitation of the source or perhaps complete denudation of the area (of trees) had taken place. The possibility that the Type II wood could have originated outside the Netherlands was discounted on the basis that

“continuous importation of wood of the same provenance for over 200 years is hard to imagine” (Eckstein et al., 1975).

So by 1975 art-historical analysis had yielded two chronologies, MC18 in England and Netherlands Type II. These chronologies showed good cross-agreement with each other but failed to provide convincing cross-agreement with any of their neighbours. The conclusion of the Hamburg workers was that “a coastal habitat on both sides of the North Sea is assumed”.

The Question of Importation When MC18 was recognized as being different from the other British Isles chrono- logies, one immediate possibility was to assume importation, i.e. could the timbers be exotic to the British Isles? (Baillie, 1978). By that criterion the Netherlands II chrono- logy could also be exotic to the Netherlands though that possibility had been dismissed by Eckstein et al. Also, Fletcher et al. had-made a strong claim for an English origin for most of the MC18 material. What are the possibilities? Either

(a) there are two separate sources in England and the Netherlands, or (b) there is one source either in England or the Netherlands, or (c) there is some general source exotic to both England and the Netherlands.

Page 5: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 375

Possibilities (b) and (c) are discounted by the Oxford and Hamburg groups. How- ever the implausibility of two separate sources matching each other yet neither matching any neighbouring chronology renders (a) unacceptable as a realistic possibility. For (a) to be true we have to imagine medieval woodsmen, on opposite sides of the North Sea, selecting trees which were exceptional within their own areas, in both areas these exceptional timbers were then further selected for use by artists and then subsequently it transpires that these doubly selected groups exhibit similar patterns of rings. The possibility (b) of one source in England or the Netherlands would still leave the basic problem of why that source fails to date decisively with any other European chronology. We are left therefore with (c).

What are the possibilities of(c) being the true situation? B.efore discussing this it is necessary to point out the dangers of (c) and the reason why it is not an answer desired by those working with art-historical material. If all the oak panels of MC18 and Netherlands II were exotics, in the same sense as the sequoia museum specimens exampled in the introduction, then the precise dating of these chronologies would have to be called into question, i.e. it might not be possible to establish the true dates of MC18 and Netherlands II by comparison with genuine German and British Isles chronologies. It might be necessary to find the source of the exotics and establish a suitable chronology to link them to modern times. As numerous art- historical and some climatic deductions have been drawn based on the precise place- ment of these chronologies, the question of origin with its “knock-on” implications for the dating must be given serious consideration.

The question returns therefore to the possibility of oak importation into the Netherlands and England in the later medieval period. Did such importation occur? The question is not so much about specific importation of oak panels for art-historical purposes as about a background of importation existing which should have been taken into account when devising the dating of the art-historical chronologies. The background for England is summarized by Rackham (1980) who notes that oak wainscot boards derived from Poland were imported in large quantities from the Baltic Ports to be used for panelling. In fact Rackham grasps the art-historical nettle by stating that the oak boards of furniture and panel paintings (as used in dendro- chronology in England)

“are seldom of known provenance and it is not always sufficiently realized that such boards are particularly likely to be imported”.

However, Rackham’s comments are after the event. Let us look briefly at a few sources from before the art-historical use of dendrochronology.

Symonds (1946) discussing the craft of the English joiner in the 16th century says that wainscot

“meant boards of quartered, sawn or riven oak of the type grown in Russia, Germany and Holland. Much of this foreign timber was imported to England . . because it was recognised, by reason of its straight grain and even texture, to be infinitely superior to the majority of English grown oak”. He also notes that “it would appear that the greater part of the quarter sawn oak used by the English joiner for wainscot and good quality furniture was imported”.

Abram (1909) notes that there was a great increase in English commerce with the countries around the Baltic in the 15th century and specifies trade with Prussia as “of great value because timber, which was much needed in England, could be obtained there”. Salzman (1931) talking about English trade in the middle ages, says of Hanseatic (Baltic) imports “Wainscots and Eastland boards are common, almost essential, features of medieval building accounts” and cites the goods aboard 11

Page 6: Some thoughts on art-historical dendrochronology

376 M. G. L. BAILLIE

German and Frisian ships stormbound in Scarborough in 1293 as including 20,600 boards. Doilinger (1970) states

“The Hanseatic east was also the west’s great source of’timber” and cites “the vast forest areas in the basin of the Vistula and in Lithuania which made Danzig the leading exporter of timber”. He goes on to be more specific, “the principal customers were England and Flanders who needed timber for their ships. In the 15th century Prussia was exporting oak beams and planks in thousands and boards of varying thicknesses of hundreds of thousands”.

In 1460 Danzig alone -exported 3161 “hundreds” of planks. But it is Dollinger’s constant re-iteration of England and Flanders as the recipients of timber which is the most telling.

“Prussia . . . supplied that most esteemed in England; timber and its products.” “Both Dutch and Hanseatic merchants brought timber and grain to Amsterdam.” “Due to density of population and industry the Low Countries absorbed large quantities of Eastern products. They made substantial purchases of timber. ” “The seabourne commerce of Prussia (Danzig) . . . was based primarily on relations with Lubeck, Flanders and England.”

Salzman (1952) states that “in addition to native timber (in England), a large amount was imported mostly in the form of boards, from the ports of the Baltic and the North Sea. From the later years of the 13th century onwards, this trade, which was mostly in the hands of the Hanseatic League, became of increasing importance and its products found their way all over England”. More importantly he specifies that “while the wainscots were definitely, and the other varieties probably, of oak, there was also a considerable import of fir and deal boards” and in talking of wainscot he even cites preference “It may be noted in passing that this imported timber was favoured for such work as doors and screens, where it was important to avoid warping, as being better seasoned than the local stuff ‘.

Postan (1972) states that “Prices of timber invariably fell in the 15th century (in London) after heavy shipments of timber arrived from the Baltic”. It is clear that the literature of late medieval trade links England and Flanders not only with the Baltic, but with timber, often specifically oak. Moreover the nature of Hanseatic trade suggests that it is in fact possible to postulate a source area which could have lasted for several hundred years. The Hanseatic League lasted from the later 13th century until the 17th century. So we can link Baltic oak with England and Flanders and cite preference for imported boards on account of seasoning and even texture. It would be fortuitous if we could find a specific reference to panels for paintings being imported. Dollinger was not writing with any knowledge of the dendrochronological question which would emerge during the 1970s. He does however give one or two tantalizing snippets of information. For example,

“It was not only in Germany that the works of German-painters and sculptors were highly thought of. Commissions came from all the Baltic countries and as a result there was a regular trade in works of art, whose importance is difficult to gauge” (1970).

Elsewhere he cites a source which estimates that in the 15th century some 300 carved and painted panels were exported from Lubeck to the Baltic countries and in case that trade seems in the wrong direction there is the telling case of an altar piece intended for a church at Reval in the east Baltic

“commissioned in 1419 the carvings were executed by a local artist. Sent by sea to Lubeck’ it was taken overland to Hamburg (to be painted and gilded) . . the work took seven years and it was not until 1436 (?) that the finished piece was taken back to Reval”.

Perhaps the best link of timber trade to art comes in the following statement “But the finest surviving portraits of Hanseatics were the work of Hans Holbein of Augsburg, who painted them in 1532 while he was in England. The most evocative is that of Georg Giese of Danzig, a merchant at the London Steelyard” (Dollinger, 1970).

Page 7: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 317

It is interesting to speculate what sort of panel Holbein would have used when painting a Baltic trader in England! Finally Symonds (1946) notes that even as early as the 14th century coffers and chests from Flanders occur in (English) inventories and makes the important point that middlemen must have existed to distribute such wares. So, apart from panels one of the other sources of art-historical ring patterns-chests- are associated with importation. In summary therefore, evidence exists to postulate at least the possibility that oak boards used in England and Flanders, in the correct period, could have shared a common exotic origin. This evidence appears to have been largely ignored by workers in both England and Germany. When the subject was raised in an informal Tree-Ring Society Newsletter by Brett of London University in March 1979, Fletcher’s reply in the same issue included the. statement

“until the relative volumes of home-grown and imported timber can be assessed, it would be wrong to assume that imported timber played a major role in medieval England apart from specialised imports--such as boards (this author’s italics) and linenfold panelling . . .”

Since it is specifically boards we are considering this reply does not assist with answering the question of import.

In the same Newsletter Fletcher sets the scene for this discussion by stating “By all means let us try to obtain evidence about the ring width patterns of oak imports from the Baltic and elsewhere, and be cautious about tree-growth and climatic provinces”.

The Question of the Dating In 1977 some of the first clear evidence for difficulty in establishing the absolute date of the art-historical chronologies appeared. Fletcher (1977) stated

“for a number of years researchers at Oxford and Hamburg have had difficulty in finding convincing evidence for the exact dating of Refs 1 to 5 (the upgraded and extended art- historical chronologies based on MC18) and for the Netherlands II chronology . . . of the accuracy of the dating to within one year, or at the most two years, there has been little doubt”.

While in 1978 Bauch informs us that “Because at present the master chart (for Netherlands II) can only be proved with a precision of + 1 year, the curves for this chronology have not been published” (Bauch, 1978a). Here we have two art-historical chronologies which match each other, yet neither matches definitively with the neigh- bouring absolute chronologies. Worse still each laboratory favoured a dzferent place- ment

“the Oxford laboratory has favoured and used dates two years earlier than those favoured by the Hamburg laboratory” (Fletcher, 1978~).

To take the Oxford dating through its whole sequence, the original MC18 was claimed to end in 1546. In 1975 the chronology was moved forward in time by 4 years and climatological deductions were drawn (Fletcher, 1975). In 1977 the original dating was returned to and was used again in 1978 though it was conceded that dating was possible in either the original (1546) position or one or even two years further back in time (Fletcher, 1978~). In 1980 we have the statement that

“The uncertainty of one or two years in pre-1979 versions of Refs 1 to 5 has now been resolved and moves each date previously used to a date one year earlier” (Fletcher, 1980~).

Such a move was equivalent to placing the original MC18 end-year at 1545. Un- fortunately this apparently final placement of the art-historical chronologies has been undermined by a further reversion to the original 1546 dating (Fletcher, 1983). In this latter case, as in 1975, the author failed to state that the chronologies had been redated. However, sufficient information was given to allow the moves to be deduced.

Page 8: Some thoughts on art-historical dendrochronology

378 M. G. L. BAILLIE

The 1980 redating was even more confusing in that the 1545 dating was used for some ring patterns while the 1546 dating was retained for others (Fletcher, 198Ob).

Is such inconsistency allowable in terms of what we know about normal chronology behaviour? The answer has to be no. Tree-ring chronologies belong to unique sets of years, inability to satisfactorily specify the years covered suggests that the chronologies, against which dating is being attempted, are incompatible. The concept of a chrono- logy dendrochronologically dated to “within one or two years” is unacceptable to most dendrochronologists.

To understand fully the anomalous nature of the art-historical chronologies it is necessary to view them against the background of chronologies constructed using building timbers and archaeological samples. Fletcher has christened all non-art- historical oak chronologies, in northern Europe and the British Isles, as Type H (after Huber) to distinguish them from the art-historical Type A (Fletcher, 1978~). All of the Type H chronologies, from Ireland to central Germany, are precisely dated to unique calendar dates. Consistent cross-correlations link these Type H chronologies into a highly integrated group, see for example Figure 1.

How then do the art-historical dendrochronologists seek to explain the observed differences? The German workers are content to suggest that

“the wood for Dutch and Flemish panels prior to 1650 (was) derived from coastal and local forests in the Netherlands or Flanders that were exhausted completely by 1650 . .” (Bauch, 1978u)

i.e. this “different” source existed but was exhausted and no modern analogues exist- modern trees from such coastal areas are observed by Bauch to be highly consistent Type H. The German workers have therefore left the question open and hope that at some stage suitable Type A timbers will turn up to allow direct linking to the present day. As already stated they held back from publishing their chronology until any question relating to its precise dating was resolved. In England Fletcher took a rather different approach. First, he published the chronologies with precise date ranges despite any remaining doubts. Second, he postulated that between 1200 and 1600 conditions were such that Type H and Type A both grew in England. No-one is contesting Type H so it is Type A on which we must concentrate. There is plenty of information.

“Type A occurs in lowland areas or in rain shadows where precipitation is relatively low or soil moisture becomes depleted to levels which play an important role”. Type A can also occur “intermingled with Type H in areas of altitude from 100 to 200 metres with 3C40 ins rainfall, such as coastal parts of southern and western England, the Midlands and S.W. Scotland”; in addition species, soils and other biological factors are suggested (Fletcher, 1978u).

All such arguments pre-suppose that the timbers used in the Type A chronologies are English and indeed that they can be quite accurately localized within England. Unfortunately when it comes to discussing the provenance of the art-historical timbers there is little consistency. As would seem logical to any outside observer, given the portable nature of panels or boards and the historical background detailed above, there is the clear statement “. . . this research is based on pattern matching for trees from unknown areas . . .” (Fletcher, 19806). However, in the same article we are told that before the 17th century panels were made from local wood. Both statements can hardly be true but the latter has to be if the Type A hypotheses are to have any validity.

In an attempt to test for Type A ring patterns in recent times (i.e. the identification of well replicated chronologies which showed no agreement with any other British Isles

Page 9: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 319

chronology), 18 modern oak site chronologies were investigated in the late 1970s (Baillie, 1982). The results were quite surprising: not only could no inconsistent modern chronology be identified, but it appeared that there was only one underlying oak “signal” within the British Isles. When this work was extended to a larger area of northern Europe no inconsistent modern oak chronology could be identified (Baillie, 1983~). These results are backed up by independent observations by Bauch (1978~) and Bridge (1983). Having established the apparent non-existence of any Type A chronologies for modern times, the Oxford response has been to rule the results out of court by claiming that they are “unlikely to apply to many periods earlier than the 185tK1969 to which they refer” (Fletcher, 1983). So the most objective test for the existence or non-existence of Type A, i.e. a test using precisely provenanced modern trees, is ruled inadmissable.

Independent Attempts to Date the Type A Chronologies In order to attempt an independent dating of the art-historical chronologies, in 1981 a series of modern cores were taken at Sherwood Forest. The samples were supple- mented for chronology building with some 25 slices from fallen trunks and stumps. These samples allowed the construction of a Sherwood chronology from 1450 to 1981. The site conditions, in a rain shadow area to the east of the Pennines, with rainfall of around 20 inches in modern times and with mixed species, would have tended to suggest a Type A rather than a Type H woodland on the basis of the parameters hypothesized by Fletcher (1978~). In practice the Sherwood mean chronology bears a remarkable resemblance to the British Isles 18-site chronology (Baillie, 1982) (t = 9.3 for the period 185Cr1969). It also matches extremely well with the original Belfast modern chronology made from 30 trees from throughout the north of Ireland (t = 7.5 for the period 1850-1969, t = 8.8 for the period 1450-1970) see Figure 1. In British Isles terms Sherwood could be classed as a highly typical site. More importantly, it extends precisely provenanced English tree-rings back into the 15th century. Reference to Figure 1 shows how this indigenous chronology ties perfectly into the coherent Type H picture. The four medieval chronologies constructed from English building timbers can be dated directly against the Sherwood chronology. Since the modern study shows that Sherwood is an archetypal British Isles chronology this coherence in medieval times points again to the anomalous character of the art-historical chrono- logies which show no significant correlations against Sherwood at their stated end-dates. Moreover, the good agreement between the chronology from the substantially wetter Belfast area and Sherwood with its drier easterly aspect tends to argue against the idea of substantially different dendro-ecological zones within the British Isles. Both of these factors, modern observations and good medieval consistency, argue strongly against the concept of medieval workers being able to select arbitrarily trees which were so different that their ring patterns would be mutually exclusive to the normal Type H. [Fletcher points to the absurdity of this selective process when he states that “the pattern of MC18 is therefore derived from a small and unrepresentative fraction (perhaps no more than 0.1%) of the oak timber grown at that time in the region” (Fletcher, 1977. Here are timbers, claimed not to be exotic imports on the one hand, being claimed to be significantly different from the majority in the region on the other.

This Sherwood exercise made use of trees which were definitely growing in England, from AD 1450, on a site with broadly Type A site conditions (as hypothesized). Yet these trees turn out to be Type H and to be capable of dating numerous pre-existing Type H chronologies. Even had we not had the historical evidence for importation the inconsistency of the Type A chronologies in this exercise would have suggested the possibility of their having exotic origins.

Page 10: Some thoughts on art-historical dendrochronology

380 M. G. L. BAILLIE

How werC the Art-historical Chronologies Dated? Ifwe take the first art-historical chronology constructed at Oxford, MC18, and compare it with the various chronologies in Figure 1 and with some other European and generalized chronologies, using the Belfast CROS program (Baillie & Pilcher, 1973) we find that there are no significant ‘t’ values at the suggested 1546 endyear (nor indeed at the alternative 1545 position). However, in the immediate vicinity of 1546 we find that some consistently high values do occur. For clarity the various Reference chronologies l-5 (Fletcher, 1977) which were developed subsequent to MC18 are related to this 1546 position. Some of the correlations are listed in Table 1.

Table I

t

At an endyear equivalent to 1550, i.e. “4 years forward” MC18 v. Belfast MC18 V. Hamburg 440 to 1446 Ref 2/3 v. Belfast Ref 4 v. Hamburg 440 to 1446 Ref 4 v. Scotland MC18 v. Eurochron 9

At an endyear equivalent to 1530, i.e. “16 years back” MC18 v. Belfast MC18 Y. Scotland Ref 1 v. Eurochron 9 MC18 and Ref 1 Y. Sherwood

3.0 2.9 4.1 5.1 3.1 3.0

3.2 3.2 3.4 2.5

Now it is not being suggested that either of these positions for the Type A complex is correct but, in conventional dating terms, i.e. using the procedures which have led to successful chronology construction and unambiguous dating for Type H chronologies, these positions show as much or more consistency than the 15456 endyear positions. This was at least partially recognized by the Oxford workers who, as noted above, used the “4 years forward” position for a short time (Fletcher, 1975).

What we may be seeing here is a case where none of the proposed positions is correct and any apparent similarities are the result of comparing unrelated chronologies. It is possible that the Type A chronologies cannot be uniquely dated against the Type H chronologies. This is certainly the inference to be drawn from the difficulties experienced in both Hamburg and Oxford in establishing the exact dating position and in particular in the fact that the two laboratories cannot agree on a consensus dating.

The preceding discussion leads to a series of intermediate conclusions as follows. (1) There is evidence which allows importation of the art-historical timbers to be

suspected. (2) The failure to establish a unique dating for the chronologies and the anomalous

character of the chronologies in this respect lends support to (1) and suggests a mutually exclusive source to the indigenous Type H chronologies.

(3) If the chronologies are from a mutually exclusive source with Type H, for example the far Baltic, then there may be no basis for attempting to date the Type A chronologies against Type H. If this were to be the case it would lead to the inevitable conclusion:

(4) that the Type A chronologies (Netherlands II and Refs l-5) may be undated and may be completely floating in dendrochronological terms.

Page 11: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 381

It becomes imperative to enquire just how the proposed dating of the art-historical chronologies was achieved. Obviously it was not by direct, consistent correlations against Type H.

In 1974 Fletcher et al. claimed that the “1546” dating was by agreements between sub-groups of MC18 and the Munich Curve and similar but slightly lower values with Hollstein’s chronology but, “there were higher values with an unpublished curve derived from paintings by Rubens at Antwerp” supplied by Dr Bauch. This information has now to be viewed in the light of subsequent statements, first, from the Hamburg workers that the dating of Bauch’s chronology was only to + 1 year (Bauch, 1978a) and from the Oxford workers that the two laboratories favoured different absolute positions (Fletcher, 1978a).

The Hamburg workers had placed their Netherlands II chronology in time on the basis of a weak match with the pre-existing German chronologies (D. Eckstein, pers. comm.). Bauch notes that despite the hope of finding examples which would link the two type chronologies “only two panels, one from each, have been tentatively syn- chronised” (Bauch, 1978a). It is worth noting that these dating attempts were carried out before it became obvious how “different” the Type II chronology was.

Given the almost total lack of direct agreement between MC18 and the German Type H chronologies it becomes clear that the positioning of MC18 in time was influenced by other factors, and we are told what these were,

“because of the dates on the paintings themselves, it was possible, allowing for the annual rings of sapwood cut off the boards, to conclude that the . curve . ended between AD 1546 and 1.554” (Fletcher et al., 1974).

So, in fact the chronology was dated by looking for a matching position some 15 to 23 years (an approximate sapwood width) back in time from the latest signed and dated painting then available. Unfortunately this brings the discussion back to the whole question of the number of sapwood rings to be expected on mature oaks.

Sapwood If a transverse section is cut through an oak trunk two clear zones are apparent. The inner consolidated heartwood has blocked vessels and is dark in colour due to waxes, tannins, etc. Beneath the bark is a zone of light coloured sapwood with hollow vessels, living cells and food reserves. The presence of sapwood tells the dendro- chronologist that his sample runs close to the outside of the tree and allows the estimation of a felling range for incomplete samples.

The most comprehensive study of sapwood in English oaks was published by Hughes et al. in this Journal (1981). From their results an “estimated average of 30 sapwood rings” would appear to be normal with “only a one in 20 chance that the true value was outside the range 19 to 50”. This is in accord with a rough and ready estimate of 32 f 9 used at Belfast since the early 1970s (Baillie, 1973, 1982). The reason for the continued use of this rough estimate was the recognition that any more refined estimate was pointless given the range of observed sapwood numbers from 12 to 50 in mature oaks, coupled with the knowledge that extreme examples with up to 90 sapwood rings were known and could not be accounted for by any of the published estimates.

In Europe the situation with respect to sapwood rings has always been rather different. Bauch & Eckstein (1970) state that “In the case of old oak trees the number of sapwood growth rings normally varies between 15 and 25” and indeed the generally smaller number of sapwood rings in German oaks is borne out to some extent by Hollstein’s studies (1965, 1980), e.g. 20.4 + 6.2 for oaks between 100 and 200 years, 25.9 + 7.7 for oaks greater than 200 years. It will however be immediately

Page 12: Some thoughts on art-historical dendrochronology

382 M. G. L. BAILLIE

12 20 30 40 50

Sopwood numbers

60

Figure 2. Frequency of sapwood numbers for oaks older than 120 years (a) Hughes et al., 1981. (b) Norwich & Ludlow, (c) north Ireland; (d) Combina- tion of (a), (b), (c): (e) Hollstein, 1980. The absolute range 20 f5 is indicated for comparison.

apparent that even within Germany the assumed 20 ? 5 estimate for old oak trees and the observed sapwood numbers for trees greater than 200 years are not compatible. In fact for trees between 200 and 300 years old Hollstein found a range of 14-43 sapwood rings with a median of 24.5 (Hollstein, 1980). However when we look closely at the art-historical literature a pattern begins to emerge. In the late 1960s there was a tendency for the prevailing sapwood estimates to be “around 20 years” (for example Huber & Giertz, 1970).

Figure 2 shows some observed sapwood numbers, for trees with more than 120 rings in Germany (Hollstein, 1980) England (Hughes et al., 1981; Baillie & Pilcher, unpubl.) and the north of Ireland (Baillie, 1982). It is clear that the main body of sapwood numbers in England and Ireland are significantly greater than those observed in Germany. It is also clear that the somewhat arbitrary 20 k 5 figure is not compatible with either the German or English observations for trees greater than 120 years. This art-historical figure accounts (in round figures) for only 50% of the German and only 25% of the English observations. Fletcher’s suggestion of 20-30 sapwood rings again accounts for substantially less than 50% of the English observations (Fletcher, 1978b). Can we conclude therefore that the placement of the art-historical MC18 chronology may have been made using an erroneously small sapwood estimate?

In the first art-historical dating exercise based on panels painted by Wouwerman, six samples with definite heartwood/sapwood boundaries showed that allowance of

Page 13: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 383

20 f 5 missing sapwood rings placed the estimated felling years within Wouwerman’s creative period (Bauch & Eckstein, 1970). These trees were however of Netherlands Z Type, i.e. consistent material matching directly with Germany. Unfortunately when it came to placing the Type II art-historical chronology it was assumed that the amount of sapwood was the same. This was coupled with the assumption that the Type II trees grew in the Netherlands (and the Type A in England). It must now be remembered that:

(1) observations show that sapwood numbers vary from Germany to England or Ireland. Any sapwood estimate used must therefore be compatible with the area of timber origin. As doubt exists over the provenance of the art-historical material no specific sapwood estimate can be assumed ;

(2) no population of Type II/Type A specimens with complete sapwood exists on which to base an empirical estimate ;

(3) the failure to specify absolutely the date of the Type II/Type A chronologies removes the only other way to provide an estimate, i.e. a “last existing heartwood ring to use” estimate (henceforward LEHR-use estimate). (Note: art-historical dendrochronologists now claim to have a lot of evidence to support a short sapwood estimate but since this information is derived from LEHR-use intervals in turn derived from chronologies which may be placed in time using a short sapwood estimate this evidence is circular and cannot be used to support the original assumption.)

So, if Type II/Type A chronologies are placed in time assuming local wood origins and assuming an amount of sapwood around 20 f 5 years, then it is possible, if Type II/Type A are exotics as postulated above, and if the sapwood numbers for this material are in reality greater or more variable than in the German/Netherlands area (as is demonstrably the case with the indigenous British Isles material above), that the placement of these chronologies may be in error. In particular the chronologies may be placed too close to the present.

MC18 and its dating In any analysis of the dating of the art-historical chronologies the dating of MC18 is fundamental. It was the first Type A art-historical chronology to be specified and its dating (although it has been moved at different times) is still adhered to (Fletcher et al., 1974; Fletcher, 1983).

Of the panels used to produce MC18 11 were from paintings of known date. However, from the point of view of logical discussion, as can be seen below, only two of these are really relevant. These are the two panels whose ring patterns extend furthest forward towards the present, namely 1st Viscount Montague (1569) and Thomas 2nd Baron Wentworth (1568) (Fletcher et al., 1974). The Wentworth panel is the most important as it retained about 11 sapwood rings and therefore its all- important heartwood/sapwood boundary. None of the other panels retained sapwood and for reasons given below are of very dubious merit in the critical exercise of attempting to place the chronology. The Montague panel is considered purely as an example.

Let us consider the problem confronting the dendrochronologists as they attempted to decide where the end of the chronology might come. Had the ring patterns mostly run out to definite heartwood/sapwood boundaries then the chronology could have been placed roughly in time by setting the average heartwood/sapwood boundary an average sapwood amount back in time from the average known painting date (or

Page 14: Some thoughts on art-historical dendrochronology

384 M. G. L. BAILLIE

possibly a bit further to allow for other factors such as transport, seasoning and storage). Unfortunately the presence of definite sapwood on only one sample forced the assumption that all of the others ran close to the heartwood/sapwood boundary. This of course does not have to be the case. It is conceivable that in many cases a considerable number of heartwood rings could be missing, after all in the absence of sapwood it is impossible to be certain. In fact we already know that missing heartwood will almost always represent a finite number of years. The justification for this comes from cases where the Oxford and Hamburg workers have positively identified panels as deriving from the same parent tree (Fletcher, 1978h, 198Ou, b; Bauch, 1978a, b). It can be deduced that on average aroundfive rings of heartwood appear to be missing from any particular panel. This is a minimum figure derived assuming that, of any group of panels, from the same tree, the panel running closest to the present comes within 1 year of the heartwood/sapwood boundary for each tree. In fact it is quite possible in the complete absence of sapwood that in any given case even this youngest panel might have a considerable number of heartwood rings missing. Since we have this empirical evidence for a finite number of missing heartwood rings, little weight can be given to the non-sapwood panels-which leaves Wentworth.

MC18 as published ends 2 years after the heartwood/sapwood boundary on the Wentworth panel. Using the Hamburg 20 f 5 sapwood estimate we would expect the end of MC18 to lie in the range 15451555, basically the same as suggested by Fletcher et al. (1974). This of course ignores any additional factors such as transport or seasoning and relies on the assumptions (a) that the Hamburg sapwood estimate is realistic, and (b) that that estimate applies to the material-under study. If we were to repeat the exercise using the empirically derived English sapwood estimate of Hughes et al. (1981) we would suggest an end-date for MC18 (on the basis of missing sapwood only) in the range 1520-1551 with only a one in 20 chance that the true value was outside this range.

Had a tree-ring match been found for the chronology even as far -back as 1520 it would have been possible to “explain” the generally rather long LEHR-use intervals for the other MC18 panels as due to seasoning etc., i.e. an end date as far back as 1520 is certainly technically possible. Searching for a tree-ring match in the restricted 1546-l 554 range was potentially hazardous. (Note: Had a really good tree-ring match allowed the chronology to be exactly specified this discussion would be unnecessary- it is the fact that no such match appears to exist which makes these theoretical considerations necessary.)

If we now consider the Montague panel whose heartwood runs out 6 years beyond the Wentworth heartwood/sapwood boundary, its LEHR-use interval is only 19 years if MC18 ends in 1546. As already noted there is no guarantee that the last heartwood ring on this panel is at the sapwood boundary. Had it retained even a few more heartwood rings it would have immediately rendered the 1546 MC18 position un- tenable as insufficient time would have been available for sapwood plus seasoning etc., i.e. the non-sapwood examples offer little comfort or support to the positioning of MC18 and the LEHR-use intervals could be regarded as embarrassingly short.

An alternative approach and the legacy of the 1546 dating The initial attempt to date MC18 was made at an early stage in the development of dendrochronology in the British Isles. At that time it was not known whether consistent matching was to be expected, nor whether a whole series of what might be called dendroecological zones would exist. For all anyone knew dendrochronology might have turned out to be too complex (in the sense of large numbers of mutually

Page 15: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 385

exclusive tree-ring areas) to form a realistic dating method in these climatically complex islands. So in 1974 there was little reason to regard MC18 as abnormal.. The situation has changed considerably over the last decade and it is now clear that consistency is the order of the day. Let us imagine that we were trying to date MC18 today. On discovering that no consistent dating could be obtained against the available Type H chronologies, the approach to further dating could be as follows. On purely theoretical grounds we could choose an arbitrary position for the floating chronology at a distance back in time from the known use dates greater than any likely amount of missing sapwood. The interval LEHR-use could then be measured for each sample and a mean and standard deviation calculated. Obviously the variability would be the same irrespective of the distance back in time. When any further ring pattern from a painting was matched into the arbitrarily dated chronology, the mean LEHR-use interval would be added, plus some measure of that quantity’s variability, and a likely use range established. This purely observational approach would by-pass any need to know the actual amount of missing sapwood or the other variable factors discussed above (it would of course rely on the authenticity of the known-age paintings used to establish the LEHR-use estimate).

Unfortunately, while the difficulty in exactly specifying the date of the Type II/Type A chronologies means that their placement is essentially arbitrary, a pure LEHR-use estimate is not used. At Hamburg, if any trace of sapwood exists the figure 20 f 5 is used as an (apparently) absolute estimate of felling, and seasoning etc. is taken to be very short. Alternatively if no sapwood is present a terminus post is used with 15 years

‘being added to the date of the last existing heartwood ring (LEHR). This latter approach at least makes minimum assumptions and is valid provided again that the chronology placement is far enough back in time to allow adequately for missing sapwood.

Workers at Oxford, while following the Hamburg approach to some extent, place much greater emphasis on estimating the sapwood missing from indivi&aZ samples (Fletcher, 1978b, 19806). The procedure used is to assume circa-one-inch width of sapwood and derive the likely number of missing sapwood rings by dividing the average width of the terminal heartwood rings into one inch. Such an approach, attempting to tailor the sapwood estimate to the individual sample, can only be valid if

(1) such an estimation procedure is valid, and (2) if the chronology is independently placed in time purely by dendrochronological

matches. The procedure can be criticized on a number of counts. For example the assumption

that sapwood occupies some standard width close to 1 inch is not borne out by the observations of someone who had spent a lifetime working with homegrown English oak “. . . there is a definite ring of 2-3 inches in width of sapwood all round the tree” (Venables, 1975). Similarly the oak cores used to derive the data for Figure 3 had sapwood ranging from 1.8 to 9.5 cm in width. In addition the relationship between heartwood ring-width and sapwood numbers has been shown to be ill-founded by Hughes et al. (1981). Figure 3 provides independent confirmation of that finding and illustrates the actual relationship between the width of the terminal heartwood-rings and the number of sapwood rings in a sample of 30 modern English oaks. It is clear that no predictive relationship exists. For terminal heartwood rings between 1 and 2 mm the number of sapwood rings varied between 16 and 39 with one outlying value of 62 (this latter sample which has been credited elsewhere with having only 55 sapwood rings (Baillie, 1983b) has a sloping heartwood/sapwood boundary, the maximum being more appropriate for this argument). The heavy black line in Figure 3 shows the relationship predicted by the Fletcher assumptions

Page 16: Some thoughts on art-historical dendrochronology

386 M. G. L. BAILLIE

15

I I I I I I

25 35 45 55 f Sapwood numbers

5

Figure 3. Sapwood numbers for 30 English oaks from Norwich (0) and Ludlow (a) plotted against the mean ring-width of the last 40 heartwood rings for each tree. The dotted rectangle indicates the range of sapwood numbers (with one outlier) for trees with terminal heartwood rings between 1 and 2mm. The heavy black line indicates the relationship predicted by Fletcher (19786).

It appears therefore that the procedures used by the Hamburg and Oxford workers may be a recipe for falling into the trap left by.this whole dating’problem, i.e. what happens if the arbitrary placement is too far forward in time (towards the present)? Because of the “slop” in the dating introduced by the variable “missing heartwood/ missing sapwood/transport/seasoning/storage” quantity, it will be possible to explain most of the dendrochronological date ranges derived for known age samples (obviously dates derived for panels of unknown age will not be contested). For example, if the derived date range comes rather far back in time, it can be claimed that “some heart- wood is missing or the panel was stored before use”. If on the other hand the date range is rather far forward in time then “the tree had very few sapwood rings”. Such a scheme works well with one exception and that is when a panel turns up whose last existing heartwood ring (LEHR) is so far forward in time that even a minimum sapwood estimate would cause the dendrochronological date to fall after the known date. In such cases the art-historical dendrochronologists claim that the painting is a copy.

The foregoing paragraph lies at the heart of the art-historical problem. It is clear that the dating of the art-historical chronologies themselves can never be checked or con$rmed by the dating of known age samples. Because “too late” dates can be explained as copies, any misplacement of the chronologies may not even be suspected.

Moreover the too-far-forward placement reinforces the idea of short sapwood and has the added effect of “proving” that there is little time involved in the other factors such as seasoning, e.g. “This assumption contradicts a widespread old opinion that decades for drying were necessary” (Bauch & Eckstein, 1981). Interestingly it also runs contrary to what we know about late-17th century paintings on Netherlands I panels where there must have been some considerable interval for seasoning or storage (Bauch, 1978b), and recently Klein (1983) has noted that for Flemish paintings of the 15th

Page 17: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 387

century limited studies “indicate a storage time of l&15 years, which corresponds to analyses performed on 15th century panels of the School of Cologne”.

Again in the too-far-forward situation the addition of a short sapwood interval with a low f figure will often give estimated felling dates very close to the known date. These results when they do occur serve to reinforce the belief that the chronology must be correctly dated. In fact this apparent accuracy could simply be an artifact ofplacing the chronology too far forward and using a short sapwood estimate. Consider the examples in Table 2 and see Appendix I.

Table 2

Painting Rings Dendro date Signed

Br486 Tobias and Anna BR2 Rembrandt self p. BRl6l/H BRl57/P BR’Il/W (Bauch & Eckstein, 1981)

250 1622 f 5 1626 150 1630 f 5 1629 200 1631 k 5 1632 150 1632 f 5 1632 200 1636 + 5 1639

The LEHR-use intervals are 24, 19, 21, 20 and 23 years for these old trees and that includes any transport, seasoning and storage. These results appear so accurate that it becomes impossible to believe that the dating of the chronologies could be wrong unless one realizes that any arbitrary placement of the chronologies plus a suitable LEHR-use estimate would give some equally good results.

The problems come with pictures such as BR250 (Portrait of a young Jew) and BR624A (Christ) where in each case the earliest possible felling date is 1650 and in each case the paintings were assumed to have been produced before that date (Bauch & Eckstein, 1981). Or in the case of the paired paintings of the Archduke Albert and Infanta Isobella “definitely attributed to 1609” which have their last heartwood ring dated 1603, making 1609 impossible if the dendrochronological dating is correct. So these “definitely attributed” paintings become copies (Bauch, 19786). The same phenomenon is seen in the Oxford results. For example in the case of the portrait of Jane Seymour at Woburn Abbey. Here dendrochronology suggests that the panel was made after 1537 (the year of Jane’s death) and probably a decade later when her son came to the throne (Fletcher, 1976~). These apparently known-age samples fail to provide an objective test of whether or not the chronologies are correctly placed in time. Appendix 2 shows how some paintings of unknown date also have to be explained away as copies.

In summary therefore it is possible, given the flexibility in the sapwood allowance, to explain the results achieved for most known-age samples, i.e. in the eventuality of the date range being too-far-forward in time to be compatible with a painting of known date the painting becomes a copy. The art-historical dendrochronologists have thereby created a dating system which cannot be wrong, even if the chronologies are placed too-far-forward in time. Moreover, because it relies on the “scientific” basis of dendrochronology, it is unlikely for the results to be contested by art-historians whose own subject is sufficiently flexible to fit in with any “hard” facts.

So these workers by failing to recognize the possibility of an exotic source and by making assumptions about the amount of sapwood missing from their samples have created a closed system which, were it not for the noticeable inconsistency of their chronologies within the family of precisely dated European Type H chronologies, might go unchallenged indefinitely.

Page 18: Some thoughts on art-historical dendrochronology

388 M. G. L. BAILLIE

Conclusion This article does not in itself prove that the dating of the art-historical English Type A and Netherlands II chronologies is in error. It sets out to show that those working with art-historical materials in England and Flanders, by failing to take adequate account of the possibility of their material being-imported, have left their results open to question on the following series of counts:

(1) that their raw material, in the form of oak boards, may have been imported from a source dendrochronologically incompatible with the established chronologies of north western Europe;

(2) if importation from a mutually exclusive tree-ring area is the case, then the chronologies cannot be dated definitively against the established Type H chronologies and hence must be classed as floating. The observed difficulty in establishing a certain dating for the art-historical chronologies is a strong clue that importation is the cause;

(3) that the sapwood estimates used in conjunction with art-historical material are not compatible with what is known about sapwood numbers in England and Germany and the only support for the sapwood estimates used comes from the placement of the chronologies themselves;

(4) that the placement of the chronologies in time is not by definitive repeatable criteria, i.e. placement is not by straightforward dendrochronological matching but, for the English Type A chronologies at least, may be heavily constrained by an expected placement based on a “sapwood width” back from known use dates. Hence the apparent confirmation ofthe sapwood estimates for art-historical material may be based on a circular argument;

(5) that if chronology placement is constrained by a short sapwood estimate (20 + 5) and short sapwood estimates are used in turn to establish the likely felling/use range of known age timbers, the results will be largely self-confirmatory;

(6) that even when impossibly late dates are produced for particular panels these dates can be explained away by citing the paintings as copies thus producing a system which cannot be checked by the results it produces;

(7) that by ignoring the possibility of importation and by claiming that the manufac- turers of art-historical panels in England and Flanders somehow chose trees which were dzfirent from the Type H norm they have created a system whereby comments on the obvious peculiarity of the Type A/Netherlands II chronologies and questions relating to their dating can always be countered merely by citing the differences.

So instead of saying “these chronologies are different and hence may not be datable against the Type H chronologies” they have chosen to say “these chronologies are different but they can be dated against Type H to within a year or two”. It is in fact the internal contradiction of that statement, in terms of what is known about the dating of all Type H chronologies, which argues most strongly for the Type A/Netherlands II chronologies being exotics. It is also interesting to note that the only panels which definitely match known (Type H) English chronologies, e.g. Lady Butts and Henry IV (Fletcher, 1978a), are among the few not datable against the Tye A chronologies. The logic of that finding is clear.

In the light of the above, this author’s recommendations are that the art-historical chronologies be considered floating, i.e. undated until either

(a) the art-historical dendrochronologists prove the dating of their chronologies to the satisfaction of their colleagues in the field or

(b) oak ring patterns from the eastern Baltic, specifically Poland and Lithuania, be acquired and the importation hypothesis be proven or abandoned.

Page 19: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 389

In addition, until the question of importation is finally resolved the hypothesis that the Type A timbers are English in origin must be set aside. Given the historical information on importation of oak into Flanders and England, the burden of proof rests not with those who suggest Type A are exotics but with those who would see these timbers as indigenous. \

Were it to be proven that the origin of the Type A/Netherlands II chronologies was indeed within the Baltic area, then the dating of the chronologies might have to be changed with important implications for the art-historical interpretations derived so far. Alternatively, should the chronologies prove to be Netherlandish/English then dendrochronology itself will be proven to be a more complex subject than modern indications suggest. In particular the assignment of Type A chronologies to an English origin would mean that during the middle ages mutually exclusive tree-ring areas did exist within the British Isles, with all that implies for spurious between-Type dating (Baillie, 1983).

Finally it is necessary to look at potentially the most serious consequence of this paper. Let us assume that the “worst possible situation” comes to pass and the true placement of the art-historical chronologies turns out to be some years back in time from the 1545/6 position. It might seem that a new “last existing heartwood ring to use” estimate could be derived and dating could proceed as before. Unfortunately that need not be the case. If a true LEHR-use interval was found to be longer and more variable than that suggested in all the existing Type A work, then it might not be possible to rescue the apparent accuracy of the method. It is a combination of the shortness of the art-historical sapwood estimates and their small assumed ranges together with the “proven” lack of seasoning or storage which make the results so attractive. As indicated in Appendix 2 a longer LEHR-use interval combining a wider sapwood range with a variable seasoning or storage quantity would make it difficult, if not impossible, to separate early from late works by a particular artist or indeed to distinguish originals from copies made shortly after the artist’s death. So not only is correct placement of this chronology complex necessary for purely scientific reasons, it is equally necessary in order that the true potential of the method be understood.

References Abram. A. (1909). Social England in the 15th Century. London: Routledge. Baillie, M. G. L. (1973). A Dendrochronological Study in Ireland. PhD thesis, Queen’s University,

Belfast. Baillie, M. G. L. (1974). A tree-ring chronology for the dating of Irish post-medieval timbers.

Ulster Folklife 20, 1-23. Baillie, M. G. L. (1977a). The Belfast oak chronology to AD 1001. Tree-Ring Bulletin 37, 1-12. Baillie, M. G. L. (19776). An oak chronology for south-central Scotland. Tree-Ring BulIetin 37,

33-44. Baillie, M. G. L. (1978). Dendrochronology for the Irish Sea province. BAR (British Series)

51.27-37. Baillie, M. G. L. (1980). Dendrochronology-the Irishview. Current Archaeology 7,61-63. Baillie, M. G. L. (1982). Tree-Ring Dating and Archaeology. London: Croom-Helm. Bailiie, M. G. L. (1983a). Is there a single British Isles oak tree-ring signal? Proceedings ofthe 22nd

Symposium on Archueometry (A. Aspinall8c S. E. Warren, Eds), pp. 73-82. University of Bradford.

Baillie, M. G. L. (1983b). Dendrochronology; Accuracy and Limitations. In Dendrochronology and Archaeology in Europe (D. Eckstein, S. Wrobel & R. W. Aniol, Eds), pp. 49-60. (Mitteilungen der Bundesforschunganstalt fur Forst- und Holzwirtschaft 141).

Baillie, M. G. L. & Pilcher, J. R. (1973). A simple cross-dating program for tree-ring research. Tree-Ring Bulletin 33, 7-14.

Page 20: Some thoughts on art-historical dendrochronology

390 M. G. L. BAlLLlE

Bauch, J. (1978~). Tree-ring chronologies for the Netherlands. BAR (International Series) 51, 133137.

Bauch, J. (19786). Dating of panel paintings. BAR (International Series) 51, 307-314. Bauch, J. & Eckstein, D. (1970). Dendrochronological dating of oak panels of Dutch 17th century

paintings. Studies in Conservation 15, 455.50. Bauch, J. & Eckstein, D. (1981). Woodbiological investigations on panels of Rembrandt paintings.

Wood Science and Technology 15,251-263. Bridge, M. C. (1983). The use of Tree-Ring Widths as a means of Dating Timbers from Historical Sites.

PhD thesis, Portsmouth Polytechnic. Dollinger, P. (1970). The German Hansa. London: Macmillan. Douglass, A. E. (1919). Climatic Cycles and Tree Growth. Washington. Eckstein, D., Brongers, J. A. & Bauch, J. (,1975) Tree-ring research in the Netherlands.

Tree-Ring Bulletin 35, I-13. Fletcher, J. M., Tapper, M. C. & Walker, F. S. (1974). Dendrochronology-A Reference curve

for slow-grown oaks AD 123c1.546. Archaeometry 16, 3140. Fletcher, J. M. (1974). Tree-ring dates for some panel paintings in England. The Burlington Magazine

116,250-258. Fletcher, J. M. (1975). Relation of abnormal earlywood in oaks to dendrochronology and

climatology. Nature 254, 506507. Fletcher, J. M. (1976~). Oak antiquestree-ring analysis. Antique Collecting and Antique Finder

Oct. 1976, 913. Fletcher, J. M. (1976b). A group of royal English portraits painted soon after 1513-a

dendrochronological study. Studies in Conservation 21, 171-178. Fletcher, J. M. (1977). Tree-ring chronologies for the 6th to 16th centuries for oaks of southern

and eastern England. Journal of Archaeological Science 4, 335352. Fletcher, J. M. (1978a). Oak chronologies-England. BAR (International Series) 51, 139-156. Fletcher, J. M. (1978b). Tree-ring analysis of panel paintings. BAR (International Series) 51, 30%

306. Fletcher, J. M. (198&z). A list of tree-ring dates for building timber in southern England and

Wales. Vernacular Architecture 11, 32-38. Fletcher, J. M. (198Ob). Tree-ring dating of Tudor portraits. Proceedings of the Royal Institution

of Great Britain 52, 8 l- 104. Fletcher, J. M. (1983). Dendrochronology-fallacies or research. In Proceedings of the 22nd

Symposium on Archaeometry (A. Aspinall & S. E. Warren, Eds), pp. 8387. University of Bradford.

Hollstein, E. (1965). Jahrringchrcmologische Datierung von Eichenholzern ohne Waldkante. Bonner Jahrbucher 165, 12-27.

Hollstein, E. (1980). MittelEuropaische Eichenchronologie. Mainz am Rhein : Philipp von Zabern.

Huber, B. & Giertz, V. (1969). Our lOOO-year oak chronology. Conference Report to the Austrian Academy of Science 178, 3242.

Huber, B. & Giertz, V. (1970). Central European dendrochronology in the Middle Ages. In Scientific Methods in MedievaI Archaeology (R. Berger, ed), pp. 201-212. Berkeley.

Hughes, M. J., Milsom, S. J. & Leggett, P. A. (1981). Sapwood estimates in the interpretation of tree-ring dates. Journal of Archaeological Scknce 8, 381-390.

Klein, P. (1983). Dating of art-historical objects. In Dendrochronology and Archaeology in Europe (D. Eckstein, S. Wrobel& R. W. Aniol, Eds), 209-222. (Mitteilungen der Bundesforschunganstalt fur Forst- und Holzwirtschaft 141).

Morgan, R. A. (1977). Dendrochronological dating of a Yorkshire timber building. Vernacular Architecture 8, 809814.

Postan, M. M. (1972). The Medieval Economy and Sockty. London: Weidenfeld and Nicolson. Rackham, 0. (1980). Ancient Woodlands. London : Arnold. Salzman, L. F. (1931). English Tra& in the Middle Ages. Oxford: Clarendon Press. Salzman, L. F. (1952). Building in England down to 1.540. Oxford: Clarendon Press. Siebenlist Kerner. V. (1978). The chronology, 1341-1636, for certain hillside oaks from Western

England and Wales. BAR (International Series) 51, 157-l 61.

Page 21: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 391

Symonds, R. W. (1946). The craft of the joiner in Medieval England I and II. The Connoisseur 109, 17-23,9%104.

Venables, C. J. (1974). Uses of oak, past and present. In The British Oak (M. G. Morris & F. H. Perring, Eds), pp. 1 l&122. Faringdon.

Appendix 1 Figure 2 showed sapwood numbers for Germany and the British Isles for trees greater than 120 years. It is possible therefore to look at how an assumed 20 k 5 sapwood figure would behave-in accounting for the actual felling dates of the trees if their sapwood was missing and if the chronologies were misplaced from their true positions.

(1) The 20 + 5 figure works reasonably well for Hollstein’s German material with the chronology at the correct position (as might be expected), see Table la. However, even here a wider interval would be more realistic, reflecting the fact that Hollstein finds larger sapwood numbers than predicted by 20 k 5.

(2) It is clear that using a sapwood estimate of 20 f5 we would date the largest number of British Isles trees correctly if the chronologies were some 10 to 15 years too far forward.

Table la

Percentageoftreescorrectlydatedby 20 f 5

Chronology position Holstein British Isles

5 back Correct 5 forward 10 forward 15 forward 20 forward 25 forward 30 forward

33 8 51 21 48* 32 27 36* 16 37* 9 32 3 20 1 8

The conclusion which can be drawn from this exercise is this : if the art-historical material used in the Netherlands II or Type A chronologies derives from an area where the sapwood numbers are similar to those in the British Isles, then the use of a sapwood estimate of 20 k 5 will only work if the chronology is placed some 10 to 15 years too far forward. The problem is a general one which applies to any underestimation of the true amount of sapwood.

Appendix 2 To fully understand the effect of dendrochronological dating on art-historical problems we can look in detail at a specific case. The information used here derives from four publications by Fletcher (1974, 19766, 19786, 19806) and relates to three royal portraits, namely Henry VII (Society Antiquaries 22), “A Prince” (private collection) and Henry VIII bearded (National Trust). All datings are based on the “1546” placement of the Type A chronologies including those dates derived from the 1980 publication in which a “1545” dating was used for all other purposes.

The panels ofthese three paintings are stated to have derived from the same parent tree and the last heartwood rings are,

Page 22: Some thoughts on art-historical dendrochronology

392 M. G. L. BAILLIE

Henry VII 216 rings ends 1487 A Prince 2 18 rings ends 1493 Henry VIII 199 rings ends 1488 In the latter case allowance for missing rings (varying estimates of “about 1.6cm” trimmed from each side accounting for about 10 rings and “about 1 cm” accounting for 5-7rings have been used) suggests that the original ring pattern extended to c. 1498. It is hypothesized that some of these missing rings may have been sapwood (subsequent insect damage being the assumed reason for trimming). In order to obtain the likely felling date, a sapwood estimate has to be added and as the parent tree must have been around 240 years old we are told that the amount of sapwood most comparable with this age/ring- width class is “about 20 rings of sapwood” thus “it was likely to have been felled in the period 1513-20” (Fletcher, 1976b). It should be noted that this estimate is very different from either the range suggested by Hughes et al. (198 1) or the 16-39 year range indicated above for English oaks greater than 120 years and with ring widths between 1 and 2 mm.

However, we are also told that comparable material, i.e. paintings of known use dates, exhibit 22-31 years between latest ring and inscribed use-date (Fletcher, 19766). On that basis we might expect use between 1515-24 (for “A Prince” ending in 1493) or between 1520 and 29 assuming the latest heartwood/sapwood boundary hypothesized for Henry VIII. Yet in applying these rules which suggest a possible range of 1515-29 or later (remember the presence of sapwood is only hypothesized) Fletcher says “the tree was probably felled between 1513-l 8 and the portraits probably painted between 15 15-21” (1976b) and elsewhere “the felling date of the tree can be estimated as about 1512 or soon after” (1978b) later still we have this group “produced within a year or two of 15 15” and “between 1513-19” (1980b).

The reason for the truncation ofthe date range at 1519 lies in the historical information that Henry VIII decided to grow a beard in that year. Clearly if this was the case the portrait could not be earlier than 1519. However, X-ray analysis shows that this bearded portrait overlies a sketch very similar to the “A Prince” portrait. The primary use of the panel must therefore be before 1519 (198Ob).

Unfortunately this date range is no longer an independent dendrochronological estimate (which can be no better than 1515-29 or later). Use is now restricted to a’ period of a few years at the earliest end of the range suggested by dendrochronology. Note that had the “four year later” (towards the present) position of the overall art-historical chronology favoured in 1975 (Fletcher, 1975) been used here, we would have had the dilemma of original use and overpainting in the earliest possible use year-1519! Moreover further problems arise as the portraits are interpreted in the light of the dendrochronological dates. To understand these it is necessary to take a closer look at the information relating to each of the three portraits.

(1) Henry VII. Before it was known that this picture formed a “same tree” group with the other two, it had been dated separately. Its last ring was for 1487 and its likely period of use 150612 (allowing 19-25 sapwood rings and clearly not allowing for missing heartwood which was only discovered later) (Fletcher, 1974). On the basis of this dating in 1974 it was stated that this portrait “could be a copy made shortly after, rather than before his death (in 1509)“. Clearly as far as art history was concerned the portrait could have been from life, only the dendrochronological estimate making for a possible copy.

(2) A Prince. Art history is equivocal on this portrait. It could be either a youthful Henry VII or Prince Arthur (died 1502) or a young Henry VIII (19766). However, when dated, last ring 1493 and earliest use 1515 or later, it became more likely that this was a youthful Henry VIII. This portrait had by 1980 become “Henry VIII recently crowned”, i.e. a likeness of c. 1509, see below.

Page 23: Some thoughts on art-historical dendrochronology

ART-HISTORICAL DENDROCHRONOLOGY 393

(3) Henry VIII bearded. The original use of this panel must be before 1519, see above. It overlies a youthful portrait similar to (2).

Here a dilemma arises. The two youthful portraits, “A Prince” and the underpainting, represent a very youthful individual. Too youthful to be Henry VIII around 1515 (when he was about 24). So in order to explain these youthful portraits on panels “used around 1515” it is suggested that both are copies of “an earlier likeness, probably one made at the time of his marriage and coronation in 1509 when he was nearly 18” (Fletcher, 19806).

To summarize, what are we seeing here? Art-historical dendrochronology suggests an earliest use around 1515 or later (possibly some considerable time later). Historical information truncates that range somewhat uncomfortably at 1519. So between about 1515 and 1519 we have

(1) a posthumous portrait of Henry VII from an unknown original, (2) two youthful copies c. 1515 of a hypothetical portrait of Henry VIII, c. 1509.

Art-historical information is clearly not capable of resolving the problems and art historians would have been just as happy with a life portrait of Henry VII and youthful portraits of either Arthur or Hem-y VIII (one overpainted later). The conclusions which we can draw from this dating exercise are as follows. Current dendrochronological dating against the art-historical chronologies tends to yield dates which are embarrassingly far forward (towards the present). In order to explain away the lateness of the use dates for this group of panels they all have to be copies. So this apparently interesting dating exercise works only if a series of hypotheses, about the presence of a sapwood boundary and the copying of unknown originals, is accepted.

In such a situation the dating of chronologies to “within a year or two” is not acceptable. Any doubt about the independence of the placement of the art-historical chronologies and/or the validity of the sapwood estimate used renders these art- historical deductions suspect.

If for the sake of argument we were to move the chronologies 16 years back from the “1546” position (at least some evidence exists to suggest this position) and we were to apply this dating to the three panels, then the last definite surviving ring would be for the year 1477 and using the empirical English sapwood estimate of 1639 years (above) we could start to think of felling in the range 1493-1516 or later. Such a range would com- pletely change the art-historical interpretation as follows. It would still allow all the possibilities suggested by Fletcher but it would widen the likely felling/use period to an extent which would allow the Henry VII portrait to be a painting from life [interestingly Fletcher points out that Henry VII looks younger in this portrait than he does in a documented portrait of 1505 by Sittow “a copy being made of a likeness made some 10 or more years before 1505, to judge by his appearance at that date” (Fletcher, 1976b)]. In addition the range 1493-l 516 would allow the “A Prince” and the Henry VIII under- painting to be original portraits of either Arthur or Henry VIII. In fact moving the chronologies by as little as 16 years would render precise interpretation impossible simply because it would allow almost any possibility. Unfortunately in the absence of sapwood this loss ofdating resolution is the dendrochronological norm. It appears it is only in art- historical analyses of panels without sapwood that the accuracy of dendrochronology can be restored to ranges such as “within a year-or-two of 1515”.