Canadian Historic Sites: Occasional Papers in Archaeology and History No. 2

An Archaeological Study of Clay Pipes from the King's Bastion, Fortress of Louisbourg

by Iain C. Walker

Part I The King's Bastion and its Casemates: Discussion

The use of rouletted lines on stems is known in England during the second half of the 17th century. Luellin Evans of Bristol, who gained his freedom in 1661 and was still alive in 1691, used two bands on either side of his initials and a band of diamonds (Omwake 1946: 20-1; 1958: 10-11; 1963: 40-2, Pl. 11B), and identical decoration with the letters IF has also been found. As Evans was an apprentice of a James Fox of Bristol who gained his freedom in 1654 (Omwake 1957b: 6; 1962: 19), it is highly probable that the stems lettered IF belong to Fox. (Recent work on Bristol pipemakers by the writer indicates Fox was making pipes for some time before he was made a freeman, for he is noted as a pipemaker in 1651.) The fragments with rouletted decoration from the John Howland House at Rocky Nook, Kingston, Massachusetts, datable to the period 1650-80 may well have come from these two makers (Deetz 1960a: 9th page; 5th page, Fig. 1, 10; 1960b: 4).

21 Bowl fragment of English pipe with maker's initials ER on side facing smoker; probably Edward Randall, fl. 1699 to at least 1719. Context: ca. 1716-49/50.

22 Decorated stem from a Dutch pipe. Context: ca. 1716-49/50.

Similar if not identical decoration was used by the Dutch, however. Dunhill (1924: 222) and Brongers (1964b: 59) illustrate mid and late 17th-century Dutch pipes with this decoration, in one case combined with fleurs-de-lis; and at Fort Ticonderoga, New York, several Dutch and German stems show banding of different types around the stem, one with fairly close parallels being marked IN.GOUDA (Gifford 1940: 123, 130-1, Figs. 27-8, 36-9). Further, rouletted stem decoration occurs in the Chateau St. Louis (for example, 16C.4.344), in one case bordered with impressed triangles and occurring with a bowl bearing a Gouda mark, the crowned ES (Helbers and Goedewaagen 1942: 156; No. 134; cf. Walker 1966b: 747-8). Rouletted decoration without edging occurs on stems of bowls of the same appearance and shape but without marks (at least on the parts of the bowls that survive), and a stem has been found with rouletting bordered by PLENS at one end and GOUDA at what is probably the other end (4C.47.55). At Fort Michilimackinac (Omwake 1962: Fig. V, 17) the same decoration edged with C:D:ROOS and GOUDA occurred. This appears to have been a frequent decoration on pipes manufactured in Gouda, and also on those manufactured at Dunkirk (Duhamel du Monceau 1771: passim). Indeed, Duhamel du Monceau (1771: 4) says that one of the differences between "English style" and "Dutch style" pipes is that the former do not have decorated stems, although this is clearly not strictly true. A bowl of Dutch shape, unmarked, with a slightly different rouletting on the stem, edged with what appears to be a lattice pattern at one end, has come from the Chateau St. Louis (16D.3.236). Rouletting is included on an extremely baroque late 17th-century Dutch pipe probably from Gouda (Brongers 1964a; 46). Rouletting also occurs with edging similar to the circles, zig-zag lines, or triangles (cf. Fig. 30) on other Dutch stems (Douwes 1964: 365, 367). A decorated stem with these circles and triangles came from London: its suggested date was about 1660-70 (Atkinson 1965: 251, Fig. 5; 252), though the reasoning behind this suggestion is not clear. At Louisbourg, rouletted stems appear to be more common during the earlier part of its occupation, at least in this casemate and especially in the right face casemates; a fact which may be significant, for the French, who occupied the fortress between 1713 and 1745, would have been more likely to obtain pipes from the Dutch than the English. Further, the English, who occupied the fortress between 1745 and 1749 and again after 1758, certainly would not have obtained their pipes from the Dutch. Stems with the touching circles and rouletting were also found at Fort Michilimackinac (Petersen 1963: Fig. 27), occupied by the French from about 1714 to 1761 and by the English from 1761 until 1781; and some were found at Santa Rosa Pensacola (Omwake 1964: 15-6, 26), where most of the pipe material was Dutch. The latter site was occupied by the Spanish between 1722 and 1751. Examples came from a site in Ghana together with Dutch (as well as English) bowls apparently of the 17th and 18th centuries (Nunoo 1957: 16-7, Pl. III lower; additional information from the author) but no stems appear to have been found attached to bowls. An example found on an Eskimo site at Hopedale, Labrador, was identified as Dutch and dated to the 17th and 18th centuries (Bird 1945: 143) though unfortunately no source is given for this statement. The rouletted lines and touching circles (Fig. 29) also occur on a stem recently found in Amsterdam though not certainly Dutch (Bresnick, personal communication), and on a definitely Dutch pipe from the Pen site, New York, dating about 1685-96 (Pratt, personal communication). This latter pipe is typologically datable to the end of the 17th or the beginning of the 18th century (cf. Friederich 1964d: 4).

23 Dutch pipe with same mark as Figure 12 but without Gouda coat of arms and therefore probably earlier than 1739/40. Context: ca. 1716-49/50.

Hundreds of stems with the touching circles were found at the mission site of St. Francis Xavier IV, Caughnawaga, opposite Montreal, datable to 1696-1719, and Omwake (quoting in 1964: 26, 33), noting that hundreds of bowls with Bristol makers' marks were also found, suggested Bristol makers must have used such decoration. As for a French site using English pipes, he noted that French sites in Alabama and Louisiana have no Dutch pipe material but much English. Through the kindness of the excavator, Dr. Jury, and Fr. Béchard of the present St. Francis mission, however, I have been able to examine photographs of the site's pipe material (Béchard, personal communication), and while there are many English (some at least certainly Bristol-made) bowls there are also many Dutch. In addition to stems with rouletting and touching circles, there are those with rouletting edged with pendant impressed triangles. Unfortunately, no photograph showed stem fragments attached to bowl fragments large enough for identification, so that definite identification of these stems is still lacking. On previous evidence, however, it seems clear that the Dutch used both styles, whatever the English may have done. (During visits in 1969 to Belgian and Dutch museums the writer has observed not only this decoration but that of rouletted lines and impressed pendant triangles and also that of rouletted lines edged by impressed diamonds with a dot in the centre on Dutch and Dutch-style pipes, some of them certainly made in Gouda.)

24 Stem fragment with half of a rayed sun motif; bowl fragment with part of medallion with Setters HT separated by a star, surmounted by a pyramid of six very small, diamond-shaped dots, and enclosed in a circle cog-toothed inside and out; all raised, Context: ca. 1716-49/50.

In the 19th century at least, heavy ridged decoration moulded on the stem seems to have been popular among French, Dutch, and German makers, to judge from the way pipes with this type of stem made in northern France by Peter Dorni about 1850 (Omwake 1961: 12-15) were being copied at Gouda about 1880 (Sackett 1943: 77, 78, Fig. 2; Omwake 1961: 14; Fairbanks 1964: 48, 49); and at Höhr near Coblenz (Gifford 1940: 131, 123, Fig. 27), by the firm of Müllenbach and Thewald, which was founded in 1830 and ceased pipe manufacture in 1930 (Müllenbach and Thewald, personal communication).

An unresolved problem regarding pipes of the kind illustrated in Figure 30 — from Layer 9 — is their indifferent workmanship compared with other Dutch material. These pipes are certainly not English, and some of this type from other parts of Louisbourg bear marks known to have been in use at Gouda at this time. Further, the parts of the bowl that survive, as well as the style of decoration, indicate Dutch inspiration. Dutch pipes, however, are consistently of a more refined workmanship than English pipes. Figure 5 shows a Dutch pipe with the Gouda arms and the letter S on either side of the heel, indicating that it belongs to the lower classes of Gouda pipe; but in fact it is still superior in quality to English pipes. The types shown in Figure 30, however, are markedly inferior to the example of the lower classes and it is possible that they may have been either crude imitations of Gouda pipes made elsewhere in The Netherlands, or perhaps in the Pas-de-Calais area of France, or in modern Belgium whence the Dutch obtained some of their clays. The former area was where the French clay pipe industry was later very prominent (cf. Fresco-Corbu 1962: 1445; Lesur 1957-60; passim) and where pipes were already being made from early in the 18th century (Duhamel du Monceau 1771: passim; Diderot 1713-84: IV, 37S; Lesur 1957-60: passim; Jakowsky 1956: 23-5). Alternatively, these pipes may simply have been shoddy goods made, rather like glass beads, especially for selling to hapless overseas buyers.

25 Stem fragment with REUB/ENSI impressed, the mark of Reuben Sidney. hitherto unknown maker. (Now known to be from Southampton.) Context: ca. 1716-49/50.

Table 1 lists the ten marks which occur in more than one layer of the casemate. Dated with relative accuracy are the first and last listed pipes, those marked TD and those made by the Robert Tippet family. The first seems to be no earlier than the mid-1750s; the latter no later than the late 1750s. Of the eleven pipes marked TD, ten came from Layer 5 and above, and the eleventh from an undifferentiated deposit of Layers 5, 6 and 7. All five Robert Tippet pipes (plus one J[o]ane Tippet pipe) came from Layer 7 and below. An initial examination of the chart suggests, therefore, a break in the history of the fill around Layer 7 (no significant pipe material came from Layer 6).

There are only two cases where similar marks are found above and below this level. A pipe with the letters FS, each crowned, came from each of Layers 2, 3, 7 and 9, and some with the encircled letters EC came from the deposit containing material from Intrusion 1, Layer 5/6/7, Layer 8 (one each), and Layer 9 (two).

Disregarding the possibility of contamination for the moment, it is quite possible, in fact probable, that the makers of the pipes bearing the marks FS and EC spanned the difference in time between the postulated two periods of fill, for only an unusual coincidence or a considerable gap in time would have two deposits so mutually exclusive as to have no pipe-makers' manufacturing lifetimes overlapping. Layer 8, with only one significant mark, and that one common both in the earlier and later deposits, could be placed in either group. A firm terminus post quem for Layers 2 and 3 and Intrusion 1 is given by the appearance in these layers of Dutch bowls bearing the Gouda coat of arms, which dates the deposition of these layers, if not all the material in them, to post-1740. Of the other material, the TD pipes offer the best dating material, for on evidence from other sites, it is difficult to date their deposition here to earlier than the mid 1750s. The appearance of the WM marks at Williamsburg almost exclusively in the period 1750-65 is further indication of a late dating for these layers, for this mark occurs four times. If the dating of the mermaid mark at Gouda were certain, and, more especially, if the suggested dating of the bowl with the lion guardant were correct, then it would prove that these layers were deposited late in Louisbourg's history. Indeed, the only material which specifically cannot be fitted into a 1750 and later dating is the group of ornate stems that came from Layer 4, although as their suggested maker was still alive in 1758 they need not necessarily be confined to the period in which they were found at Chester.

26 Stem fragment with CAR/TER raised from round cornered depresed square on top of stem. Probably a Southampton maker of first half of 18th century. Context: ca. 1716-49/50.

The evidence that Layer 9 was early comes principally from the Tippet pipes, but none of the other pipe material from this layer controverts this and some in fact supports it. None of the three Dutch pipe bowls found below Layer 7 (all from Layer 9) carry the Gouda arms. In the case of the crowned 6, this is probably significant, since two bowls with this mark plus the coat of arms came from the upper layers. However, the addition of the arms to a maker's mark was probably entirely permissive, and makers whose marks were not being plagiarized may not have added the coat of arms. Thus no definite significance can be attached to the absence of Gouda arms from the other two Dutch marks in this layer.

The occurrence of two bowls with the encircled EC, however, unless they represent an unknown maker, suggests that this layer was not deposited until the 1740s at the earliest, for Evans Cheever, the earlier of the two known makers with these initials, did not become a freeman until 1741. In view of the demonstrably earlier material in this layer, especially the pipe ascribed to J(o)ane Tippet, which must be dated about 40 years before Cheever gained his freedom, we must either conclude (assuming that Cheever is the maker of the pipes with the letters EC) that Layer 9 was deposited in the 1740s and comprised a great deal of much earlier refuse, or that it was not, in reality, one deposition.

Although the fact that early clay pipes had shorter stems and wider bores than later ones had been noted previously (Calver 1931: 97), Harrington (1954) was the first to realize that this might afford a means of statistically dating otherwise uninformative stem fragments. He produced a series of charts giving the percentages of 4/64 in. to 9/64 in. bore diameters covering the 17th and 18th centuries and showing a steady reduction in average bore diameter from the beginning of the 17th century until towards the end of the 18th.

Largely because of the dating of the various sites from which the original dating material came, Harrington's figures applied to periods of between 30 and 50 years. These percentages were adapted by Binford (Maxwell and Binford 1961: 107-9; Binford 1962: 19-21) to a straight-line regression formula which resulted in an equation Y = 1931.85 — 38.26X, Y being the desired date, 1931.85 being the theoretical date at which the bore diameter would reach zero by this formula, 38.26 being the slope of the line (the number of years between each 1/64 in: decrease), and X being the mean bore diameter for the sample to be dated. The result is a single date, theoretically the median figure for the occupation time of the material under examination.

In view of the relatively close dating obtained from the study of the pipe material, it was decided to apply the Binford formula to the individual layers (cf. Walker 1965; 1968). A. Noël Hume (1963: 22-5) has shown that while it took a minimum of 900 to 1,000 fragments to obtain a consistently stable date (plus or minus six months or less) by the Binford method — increasing percentages were taken at random from a deposit of over 12,000 fragments — it was found that sites in Virginia which had terminal dates before about 1760 gave good results whether the number of stems was 17 Cr190; while later sites, whether with 31 stems or 485, gave inaccurate results, and increased in their inaccuracy (by giving dates too early) into the 19th century. This merely confirmed in detail what both Harrington and Binford had said; namely, that their formulas broke down towards the end of the 18th century. (The effect of Dutch stems on the calculation is uncertain: Harrington [1954: third page] specifically excluded such stems as they had smaller bores, and were also shorter than English pipes of the same period, but Omwake [1957a: 2; 1965 letter: 15-6] has suggested there was no appreciable difference. However, the Binford dates for the deposits in Casemates 13-15 Right were over ten years later than the median for their occupation dates, 1720 to about 1732; and as these layers contained a large amount, perhaps as much as 50 per cent, of Dutch material, it seems likely that this explains the Binford dates for the material. If Dutch material did have narrower bores than English material of the same period, then the former would, when used in the Binford formula, give a later date.)

The following table gives the Binford dates for the layers discussed above.

In the preceding table, the bore diameters were measured with the ungrooved end of drill bits. Omwake (in a letter dated 16 March 1965: 14) noted that this is the only accurate way of measuring the bores, as the grooved end permits a certain amount of play which allows that end to penetrate a short distance when the butt end would not. As an experiment, measurements were also taken with the grooved end and it was found that the dates obtained, even from taking the drills that fitted most comfortably, were between two and five years earlier than those obtained by using the blunt end, while the tightest fit — that is when a bit entered only part way into a stem — were between 11 and 18 years earlier than those obtained from using the blunt end. Three examples taken from the above layers are given in Table 3.

From the right face casemates, however, as will be shown in the second half of this paper, in the light of strong historical evidence and some indirect evidence from the pipes themselves, the only acceptable dates came from the tightest fit. In other words, the dates that were ten or more years earlier than those obtained by using the most accurate measuring appeared to be the correct ones, which indicates that statistical analyses per se are not always reliable.

Taking into account the evidence already produced by makers' marks, the dates obtained from using the butt end of the drill seem to give the only reasonable dates for this casemate. The marks suggested there should be a break in the occupation history in the area of Layers 7 and 8. Judging from the Binford dates, the break occurs between Layers 8 and 9, and Layers 7 and 8 belong to the same group as those above them.

The date for Layer 9 adds some evidence to the idea suggested earlier that either this layer may not have been one deposition or that it includes the redeposition of earlier material, for in view of the amount of early dating suggested by many of the marks, a rather greater difference might have been expected between its median date and those of the upper layers. No reliable inference can be drawn from the date of Layer 10, as it unfortunately yielded only three fragments.

However, there is also a suggestion that Layers 2, 3 and 4 may be closer in dating to each other and a few years later than Layers 5, 6, 7 and 8 which may themselves be closely contemporaneous. This, however, can be no more than a suggestion, for the application of statistical analyses to this material is by no means foolproof, and in this case the numbers of stem fragments from Layers 5, 6, 7 and 8 are so much smaller than those from Layers 2, 3 and 4, that the chance of inaccuracy in the result is increased.

In summing up the history of the fill of this casemate as deduced from the clay pipe evidence alone — analyses of makers' marks and statistical analyses — the sequence would appear to be as follows. Layer 9 was deposited in the 1740s, possibly as a deliberate fill of rubbish from various areas, for it contains material that appears to range over a period of some 40 years. (Layer 10 is mortar detritus spill along the foot of the walls. It is separated from Layer 12 which is the same material mixed with mud, by Layer 11, which comprises mainly large pieces of wood, some of it decomposed, scattered mortar detritus and mud, but all three are clearly connected with the initial construction. However, although the Binford date agrees with this, a date from three fragments in Layer 10 cannot be accepted statistically.) The remaining layers appear to have been added in the 1750s and 1760s. The occurrence of so many marks common to these latter leaves no doubt as to their substantial contemporaneity, and the 11 TD marks in the upper layers make it difficult to date Layers 2-5 earlier than about 1755 on the present evidence. Viewing this evidence in relation to the Binford dates, it may be that these layers include earlier material, too; but if so, the range seems to be too small to be represented in a study of the makers' marks. It is possible that the upper half of these layers may be a little later than the lower half, but this is not certain. In any case, the dating evidence from the pipes in these layers precludes a difference of more than a very few years.

The possibilities of contamination of the layers may be grouped under three main headings: (1) contamination during or after excavation; (2) contamination by intrusion after the main period of occupation but before present excavation; or (3) contamination caused by redeposition of older, together with contemporary, material during the main period of occupation.

The excavations and cataloguing were carefully controlled, and contamination during and after excavation may be regarded as minimal. The second possibility is more difficult to assess: a pit was dug below the present (restored) doorway for approximately three feet out from this wall and extending halfway towards each side wall, during unrecorded work in 1962 (Intrusion 3), and its remains were visible when controlled excavation commenced in February, 1964. As excavation proceeded, however, an earlier intrusion (Intrusion 2) at this end, shallower but larger than the 1962 pit which cut through it, was uncovered. This earlier intrusion, which extended the full width of the casemate, reached out about eight feet from the doorway wall. There was evidence that this had been dug earlier than the restoration work in the 1930s, but the fill included the base of a glass bottle which Gerald Stevens of the Royal Ontario Museum, Toronto, dates to about 1915 (information from D. G. MacLeod). It may date to the work done by Kennelly in the first decade of this century, but we have no detailed description of his work. Finally, in 1962 or 1963, virtually all the rubble, which was the back wall of the casemate collapsed inwards, was removed without recording or collection of artifacts. As noted earlier, there is little historical evidence to date the collapse of the back walls of any of these casemates, though it appears to have happened by the earlier part of the 19th century or before. In Casemate 4 Right, an intrusive pit sealed by the wall collapse contained a stem with -TON LIVERPOOL moulded on it which cannot be dated much before about 1790 at the earliest (Omwake 1965: 33-4), indicating that, in this particular case at least, the rear wall collapsed sometime after that date.

27 Upper row: three typical North American export-type English bowls; lower row: three typical Dutch bowls. Diagnostic features are the shape; larger size of English bowls; plane of rim of bowl of English pipes being parallel to line of stem; smooth glossy surface of Dutch material; and milling of edge of mouth of Dutch bowls.

The two intrusions noted at the front of Casemate 1 Right do not appear to have materially affected the pipe evidence offered in the preceding pages. The earlier intrusion cut into most, it not all, of the previously described layers; the later one certainly did, penetrating the earlier intrusion and definitely into Layer 9. This means that there were opportunities for material from one layer to fall into a lower, and for intruded material to come in contact with that of the layers under discussion. Furthermore, once the back wall had collapsed, the logical place to deposit rubbish would have been at the entrance, so any 19th-century material there would have been effectively dug into the lower fill by either intrusion. There is no specific evidence of intrusive pipe material, however, among the nearly 1,400 fragments studied; for example, none of the ornate bowls or stems with moulded names that are typical of clay pipes from about 1780 onwards.

As noted earlier, the number 8 impressed on the base of three bowl-and-stem fragments from Layer 4 and from Intrusion 1 (the deposit containing parts of Layers 2, 3 and 4) has certain parallels in the 19th century, where numbers, usually moulded on the stem or on the spur, designate makers' types or styles. A similar bowl-and-stem fragment with a flat-headed figure 3 came from relatively high up in the material used to build the right face rampart over Casemate 14 Right (Fig. 46); and while the most likely explanation is that it was deposited with the material used to build the body of the rampart, the possibility of it being later or even post-French in date cannot be entirely ruled out, There is archaeological evidence, however, to be dealt with later, which places Intrusion 1 (which in any case was entirely sealed by the fall of the back wall, as it lay directly beside this wall) in an historically datable context.

28 Classic example of English pipe bowl of Oswald's type 9c. Context: ca. 1716-49/50.

The third possibility of contamination — redisposition of older together with contemporary material during the main occupation period of the fortress — can be fully discussed only with reference to other artifact evidence. However, the pipe evidence from Layer 9, covering as it does some 40 years, indicates either that this layer is of more than one deposition or that the fill used comprised rubbish from various dumps which contained material dating from the earliest French settlement. Statistical dating evidence of stem bore diameters, as already noted, suggests that the upper layers, 2 to 8, may have some redeposited pipe material in them but certainly cover a much shorter period than Layer 9. Nevertheless, the very nature of Layers 3 and 4 indicates that they comprise material redeposited from some other area, for Layer 3 contained much burnt organic material and Layer 4 was actual ash, while almost none of the artifacts from either layer indicated that they had been burned. In addition, as can be seen from the section, Layers 2 to 8, in marked contradistinction to Layer 9, lay in thin horizontal bands, one atop the other, strongly suggesting they were deliberately laid till rather than dumped fill or the result of natural accretion.

Turning to the stratigraphy of the casemate as shown in its longitudinal section (Fig. 31), the basic difference in deposition between Layer 9 and the layers above it is immediately obvious, confirming the suggestions made earlier from the pipe evidence that Layer 8 marks the beginning of quite a different period in the history of the casemate, it is also clear that Layer 9 represents a major fill of the lower half of the casemate, thus strengthening the contention that this material was rubbish and debris collected from various sources and simply dumped in. The slope of this fill indicates that it was thrown in through the doorway and progressively filled towards the rear. Historical evidence indicates that in 1749 or 1750 one casemate and part of another on the right flank were filled with refuse from the ditch of the Chateau St. Louis: Casemate 1 Right would have been a logical place for dumping, being the nearest casemate.

Subsequently, the top of this deposit (which may have reached as high as the threshold level) was removed, and a deposit of wood scraps (Layer 8) laid, apparently sloping up towards the doorway, although this vital area had been destroyed by the earlier intrusions, Why such unusual material should have been used is difficult to explain, but its occurrence over the entire area and the removal of the top of Layer 9 apparently to receive it, suggests its use was deliberate. On top of this was laid a deposit of mortar detritus (Layer 7) and on top of this a wooden floor (Fig. 34). The use of mortar detritus at Louisbourg as a base for a floor is quite common, and appears to have been designed to absorb moisture. Layer 6 was composed of areas of mortar detritus apparently identical to Layer 7, but lying atop the floor. On top of this lay Layer 5, an intermittent band of mud.

29 Stem fragments showing rouletting and runs of circles, probably Dutch. Context: ca. 1716-49/50.

30 Stem fragments showing rouletting bordered by a line of impressed triangles, probably Dutch, Contest: ca. 1716-49/50.

At this point in the history of the casemate, the material at the back wall was cut back for approximately a foot along its entire width and depth. Sections through the right flank (Fig. 35) show that this wall had been entirely rebuilt on the base of the original and bonded to the side walls of the casemate. As this work would have been done mainly from the outside, the entire fill of this casemate had been exposed and trimmed back to facilitate reconstruction. Some work, such as pointing and the insertion of three timbers lengthwise for bracing, must have been done from inside the casemate, and Layers 6 and 5 probably represent evidence of this work, for they are thickest along the centre of the Casemate, which is where one would have expected workmen to walk and bring in material for their repairs. On historical evidence, this rebuilding of the entire right flank took place in 1755, so this date is a firm terminus ante quem for the deposition of Layers 6 and 5, and their associated floor. In fact, Layers 6 and 5 were deposited in 1755. By the same means, a terminus post quem is given for the deposition of Layers 4, 3 and 2.

After the deposition of these three layers, however, a trench was dug through 2 and 3 and into 4 for the width of this wall. Marks in the mortar indicate that two timbers had at one time been set into the wall here and a third timber still in place was found below this level, from which it seems reasonable to deduce that this intrusion was made to remove these timbers. It is known that the English sappers who mined the defences of the fortress in 1760 encountered great difficulty in digging their tunnels because of continuing collapse, partly caused by the unstable nature of the soil, partly because of the extremely wet summer of that year. So much timber had to be used for shoring the mines that some of the houses in the town were dismantled to provide further supplies. It is therefore quite possible that this intrusion represents the removal of the timbers here in the course of the mining in 1760. It this is so, then Layers 2, 3 and 4 must have been deposited between 1755 and 1760. What they represent is not obvious, but their regularity certainly suggests a deliberate and careful laying.

Returning to the dates given for Layers 8 and upwards by the Binford formula, it can be seen that the tentative division of these layers into a later group — Layers 2, 3, 4, 5 and 6 (1755 and later) — and a slightly earlier group — Layers 7 and 8 (pre-1755 but post-1749/50) — is substantiated by the historical and archaeological evidence.