The TNHS Excavations: Origin and Excavators

The TNHS excavations were overseen by a committee of the British Association for the Advancement of Science (Roberts, Reference Roberts, Davies and Charles1999), with TNHS directors supervising work in the cave. For the first three seasons (January 1926 to July 1928, the crucial period of excavations in the Vestibule during which the KC4 maxilla was recovered) work was co-directed by H.G. Dowie, Frederick Beynon, and Arthur H. Ogilvie (Beynon et al., Reference Beynon, Ogilvie and Dowie1926, Reference Beynon, Dowie and Ogilvie1929a, Reference Beynon, Dowie and Ogilvie1929b; Dowie & Ogilvie, Reference Dowie and Ogilvie1927). Dowie had some archaeological training, having worked with Dorothy Garrod at Torbryan in 1924 (Roberts, Reference Roberts, Davies and Charles1999). Beynon and Ogilvie acted as leaders in geology and palaeontology respectively (Dowie, Reference Dowie1928). From January 1929 to December 1935 work was directed by Beynon and Ogilvie, while from 1936 to 1941 the excavations were directed by Ogilvie assisted by B.N. Tebbs and E.H. Rogers (Beynon & Ogilvie, Reference Beynon and Ogilvie1930; Reference Beynon and Ogilvie1931; Reference Beynon and Ogilvie1932a; Reference Beynon and Ogilvie1932b; Reference Beynon and Ogilvie1933; Reference Beynon and Ogilvie1934; Reference Beynon and Ogilvie1935; Reference Beynon and Ogilvie1936; Ogilvie et al., Reference Ogilvie, Rogers and Tebbs1937; Ogilvie & Tebbs, Reference Ogilvie and Tebbs1938; Ogilvie, Reference Ogilvie1938–1941).

Before the start of the Kent's Cavern excavation, the TNHS members had gained experience over two seasons at Torcourt Cave (Dowie, Reference Dowie1925)—an excavation that Roberts (Reference Roberts, Davies and Charles1999) has described as having been conducted to a high standard. Excavation at Kent's Cavern was carried out by the directors and male members of the TNHS, assisted by a paid workman, the latter taking on heavy work such as breaking large rocks and removing sorted spoil. The material excavated was then sorted by team members, which included men and women (Beynon et al., Reference Beynon, Ogilvie and Dowie1926; Reference Beynon, Dowie and Ogilvie1929b). These early published reports make it clear that the directors were active in the excavation, and that a limited number of other volunteers were involved. We have no detailed day-by-day records of attendance at the excavation for the early years but the excavation journals list the persons present for its later stages (see Online Supplementary Materials). These confirm that even at this late stage, the directors (Beynon and Ogilvie in November 1935, Ogilvie from 1936 onwards) were present on every single day on which intact deposits were being dug. Furthermore, the other members present generally consisted of a small nucleus of regular attendees, who would have become skilled at their respective tasks.

Professional quarrymen were employed occasionally and, when needed, carried out blasting and heavy rock clearing work. On days when the quarrymen were in the cave, the journal records that no digging of the deposits took place.

Excavation Methods

The TNHS learnt their techniques at Torcourt Cave among the Torbryan caves, and the archives for the excavations at both Torcourt Cave and Kent's Cavern show considerable similarities in approach. Working methods at Torcourt Cave are described in a preliminary manuscript report in the Torquay Museum archive: ‘The method of excavation was the same throughout, the earth being taken out by means of small hand picks, and sent up to the entrance in buckets to be sorted in full daylight. Measurements were taken as the work proceeded’ (Interim report on the further Exploration of Cave A.i at Torbryan, Torquay Museum [hereafter TM]: AR4422).

Tool receipts in the Kent's Cavern excavation archive (TM: AR4300) include references to small picks. Several archive photographs also show small tools including small hand shovels and a small pick with a total blade length which can be estimated from the photograph as about 20 cm long (TM: PR20558; see Online Supplementary Materials). Other smaller tools are likely to have been used but if so they would probably have been owned by the excavators (as is common on archaeological excavations today) and so are not recorded. Confirmation that the fossiliferous deposits were dug with care is provided by the excellent state of the bones in the Torquay Museum. Their condition is inconsistent with removal by large, heavy tools, and we conclude that they must have been extracted from the matrix with considerable care (B. Chandler, pers. comm.).

In some parts of the deposit, the journal and early published accounts of the excavation make reference to the difficulty of work in heavy, bouldery ground and hard stalagmite. Accordingly, heavy tools and occasionally blasting were sometimes required to make any headway. Beynon et al. (Reference Beynon, Ogilvie and Dowie1926: 326) noted:

The reference to ‘quarrying’ is an honest acknowledgement of the difficulty of making progress in deposits with large boulders and massive, hard stalagmite floors and bosses. It is important to note that similar approaches are used, by necessity, on Palaeolithic sites to this day, to remove large boulders and break up rocks that would otherwise make proper excavation impossible.Footnote ¹ The work undertaken in Kent's Cavern was little different in its basic approach. Beynon et al. (Reference Beynon, Ogilvie and Dowie1926) noted that ‘quarrying’ was necessary over only half the area and in fact they rapidly abandoned attempts to excavate such ground in the 1926–1928 trench, which covered Trench B, and the adjacent areas, which were only excavated to a shallow depth. The deposits near the east wall are noted as ‘the best area’, implying that the heavy digging methods described would not have been necessary there. This covers the area of Trench C, which was later to yield the KC4 maxilla. Here, a coherent and easily excavated deposit of loamy Cave Earth was present, which they eventually excavated to a depth of 23 ft (7 m) below datum.

Beynon et al. (Reference Beynon, Ogilvie and Dowie1926: 326) describe the excavation method thus: ‘So far as it has been physically possible, the deposits have been taken out foot by foot in slices 3 ft [91 cm] long by 1 ft [30.5 cm] broad by 1 ft in vertical depth.’ This method is clearly based on that of Pengelly before them, who excavated in blocks of similar size (Lundberg & McFarlane, Reference Lundberg and McFarlane2005). Pengelly's working method was to start with a cut face 4 ft (1.21 m) deep, and to advance this face across the area to be excavated in vertical, 1-ft (30.5 cm) slices or ‘parallels’. One would therefore expect the TNHS excavators to have used a similar technique.

We have analysed the recorded find depths for the Vestibule excavation of 1926–1928 (see Online Supplementary Materials). Figure 2 illustrates the depths of finds in Trench C (the deepest trench and the findspot for the KC4 maxilla) plotted against the date the finds were made. This shows that the excavators must have dug down to a depth of 4 ft (1.21 m) at one spot and then widened the trench, consistent with a technique similar to Pengelly's. After the first 4 ft (1.21 m) had been excavated in Trench C, the method appears to have been modified slightly with the base of the trench taken down by 2 ft (0.61 m). Thereafter the pattern of find depths is consistent with the excavation being taken down in steps 2 ft deep (0.61 m), with no more than a 2-ft depth range being recorded for finds on any one day. An exception is 13 June 1927, on which date the trench was deepened by 3 ft (0.91 m) from 15 ft (4.57 m) to 18 ft (5.49 m) below datum. Over this depth range the deposit appears to have been almost barren (only one find was recorded in the whole trench), which doubtless accounts for the speed with which this part of the sequence was excavated.

Figure 2. Plot of find depths in the 1926–1928 Trench C against find date (after Beynon, et al., Reference Beynon, Dowie and Ogilvie1926–1932).

Further evidence is provided by the excavation journal, which refers to trench widening on 11 March 1926, 16 March 1926, and 21 October 1926, also wholly consistent with this excavation method (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932). As work proceeded to greater depths, a trench with a stepped bottom profile would have resulted. This is seen in the drawn sections (TM: AR4271, AR4273) which show regularly stepped profiles of the base of the excavation at the end of the 1926 and 1926–1927 seasons. The excavation journals, the drawn sections, and the published accounts therefore all indicate that the excavations proceeded in an organized and systematic manner (contra White & Pettitt, Reference White and Pettitt2012: 10).

After excavation, the deposits were spread out on sorting tables and sorted by hand. Sieving was not generally used, almost certainly because of the nature of the sediment. The Cave Earth is a very poorly sorted, damp deposit that is sticky with a marked tendency to form clay balls. Wet sieving of the very large volumes of sediment excavated during the TNHS excavation would have been impractical, so the system of hand sorting was a practical solution to a problem posed by the nature of the sediment. The recovery of small spalls resulting from scraper retouch shows that the recovery of finds was in fact very efficient (Jacobi & Higham, Reference Jacobi, Higham, Ashton, Lewis and Stringer2011).

Recording and Publications

A substantial archive survives from the 1926–1941 TNHS excavations (TM: AR4262–AR4304; see Online Supplementary Materials). It comprises the three excavation journals, which provide a detailed record of the entire excavation and the finds made (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932; Beynon & Ogilvie, Reference Beynon and Ogilvie1932–1938; Ogilvie, Reference Ogilvie1938–1941), five drawn sections (Figure 3), an inventory of flints, manuscript accounts of the excavation, and receipts, letters, and photographs.

Figure 3. AR4273 Longitudinal section of the Vestibule trench, June 1927. Note that on this section the datum line is drawn horizontal, though elsewhere (Section AR4271) the slope of the deposits and datum line is accurately recorded.

The excavation was nonetheless criticized by Campbell and Sampson (Reference Campbell and Sampson1971) and Tingley and Chandler (Reference Tingley and Chandler2008) as being poorly recorded. These views have been accepted by White and Pettitt (Reference White and Pettitt2012). Campbell and Sampson (Reference Campbell and Sampson1971: 6) said of the TNHS excavations that ‘for most of their finds one only knows that they come from Kent's Cavern, since no data on locality or even depth were recorded’. This shows that these authors cannot have seen the excavation journals, which contain extensive records of find locations and depths (see below). Indeed it is not clear whether Campbell and Sampson saw any of the archive. They were thus scarcely in a position to fairly judge the excavation. Tingley and Chandler (Reference Tingley and Chandler2008) did use the archive but found the journals confusing and cryptic. It is important to note, however, that Tingley and Chandler (Reference Tingley and Chandler2008) is an unrefereed and unpublished student undergraduate project report (C. Jones, B. Chandler, pers. comm.). Moreover, Tingley carried out no work in the cave, nor was he apparently aware of the British Association excavation reports (which are not referenced anywhere in Tingley & Chandler, Reference Tingley and Chandler2008); it is not surprising, therefore, that Tingley found the excavation journals difficult to comprehend. Again, our conclusion is that we can set these authors' work to one side.

Our study of the journals has revealed a very different situation. We have established, by comparison with known samples of the handwriting of Beynon, Dowie, and Ogilvie, that the journals were written by the directors themselves (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932; Beynon & Ogilvie, Reference Beynon and Ogilvie1932–1938; Ogilvie, Reference Ogilvie1938–1941). Over most of the period of the Vestibule excavation the record provided by the journals is generally good, with detailed accounts of excavation by a series of trenches between 1926 and 1928, and 1932 and 1938. We have cross-referenced the journals with the sections, the publications of the excavations, and Lake's (Reference Lake1934) survey, which clarifies what the journal entries refer to. In addition, a number of measurements that were made in the early stages of the excavation are repeatable today and we have found them to be accurate. This gives us confidence that the measurements given in the journals and other sources are reliable.

Over the Vestibule as a whole, finds can be assigned to fourteen identifiable trenches. No single plan of the Vestibule trenches was produced, but the sections, plans in the excavation journal (Beynon & Ogilvie, Reference Beynon and Ogilvie1932–1938), and Lake's (Reference Lake1934) survey provide significant information, to which numerous measurements and descriptions of trench locations in the excavation journals and publications can be added. We have found that these sources are consistent, and they provide enough information to enable us to reconstruct a plan of the trench layout (Figure 4). In many cases precise measurements allow us to plot trench locations exactly: elsewhere uncertainties remain but these are generally not more than 1 ft (30.5 cm) to 2 ft (61 cm). Given that the horizontal dimensions of trenches were normally given to a precision of 1 ft this is an acceptable degree of uncertainty. The reconstructed trench layout for the Vestibule provides a coarse grid within the chamber. Given the dimensions of the largest trenches, 10 × 10 ft (c. 3 × 3 m) for Trench A (1926–1927) and 8 × 10 ft (c. 2.44 × 3 m) for Trench C (1926–1928) and Trench F (1934–1938), this grid is of fairly low resolution. Nonetheless it offers considerable potential for future work as it can be used to approximately reconstruct the sedimentary sequence and finds in three dimensions.

Figure 4. Plan of the Vestibule showing the location of the trenches.

The excavations in the Vestibule were carried out relative to datum lines that were set to obtain as closely as possible the depth below the original top of the Cave Earth (or the base of the overlying Granular Stalagmite), allowing for the 4 ft (1.21 m) of Cave Earth excavated by Pengelly. For the 1926–1928 trench along the north-western wall of the Vestibule, the TNHS excavators assumed the existing sediment surface to be the base of Pengelly's 4 ft deep excavation. Find depths were obtained by measuring the depth below this surface, and then adding 4 ft (1.21 m) to obtain the depth below datum, which was set to represent the original top of the Cave Earth (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–32: entry for 3 February 1926). The excavators recorded that this produced a datum line sloping down by 2 ft (0.61 m) over the 41 ft (12.5 m) length of trenches A-C, parallel with the slope of the existing sediment surface (Section drawing AR4271). We have surveyed sediment remnants on the Vestibule walls and located both the base of the Granular Stalagmite and features shown on the TNHS sections, confirming that the 1926 sediment surface was indeed the base of the Pengelly excavation and enabling us to accurately reconstruct the position of the TNHS datum line.

White and Pettitt (Reference White and Pettitt2012: 11) question the reliability of the 1926–1928 datum, citing an entry in the excavation journal for 2 January 1928 (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932), which reads: ‘Remeasurement, direct from base Gran. S. to lowest point reached in Trench C. shows over measurement last year = c. 1.’ The probable source of the excavators’ confusion lies with a second stalagmite remnant, which lies about 1 ft (30.5 cm) lower. This is in fact a thin skim of stalagmite that grew in a compaction void between the Cave Earth and the wall, which overhangs at this point. If the excavators mistook this feature for the true Granular Stalagmite on 2 January 1928 (at which time it would have been 18 ft (5.49 m) above the trench floor and impossible to examine closely), then that would account for the journal entry. Regardless of the reasons for the journal entry, this does not affect depth measurements made before that date. As noted above, this consisted not of measuring directly from an assumed position of the base of the Granular Stalagmite, but in measuring from the known level of the existing sediment surface, and then adding 4 ft to obtain the depth below datum. Thus, all measurements up to 2 January 1928, at which a depth of 19 ft (5.79 m) below datum had been reached, well below the 10 ft 6 in (3.20 m) depth of the KC4 maxilla, were internally consistent relative to each other.

Although it is clear that all measured depths made before 2 January 1928 are relative to the same datum and therefore directly comparable, the question remains of whether the excavators changed the datum after this date. This would introduce an uncertainty of 1 ft (30.5 cm) in depth measurements made from 2 January 1928 until the trench was bottomed on 7 May 1928. In fact this affects very few finds— no artefacts and only fifteen bones and teeth of animals were recorded from Trench C over this period. It affects only the deepest dated sample used in the Bayesian analysis reported by Jacobi and Higham (Reference Jacobi, Higham, Ashton, Lewis and Stringer2011) and Higham et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011), and in no way invalidates the analysis; indeed the sample concerned lies at least 4 ft (1.21 m) below the next sample higher in the sequence and it is simply modelled as being below the archaeology and at the bottom.

For the first three field seasons in the Vestibule, from 1926 to 1928, the excavation journal provides detailed measurements of trench positions, depths, and details of the sediments. Artefacts and fauna were recorded by giving the trench in which they were found and their depth below datum. Beynon et al. (Reference Beynon, Ogilvie and Dowie1926: 326) reported: ‘Every find, whether of flints, bones or teeth, has been carefully measured in regard to its depth in the deposit, and entered in a Field Book on the spot, to be subsequently transferred to a Journal.’

A high proportion of the depth measurements are recorded to the nearest inch, which shows that the finds concerned must have been located and their depth measured in situ. Critically, this includes the KC4 maxilla, which was found in three fragments within an area of 2 × 2 ft (0.61 × 0.61 m), all at the same depth (10 ft 6 in; 3.20 m) (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932: entry for 14 March 1927). Other finds have a depth range recorded and may have been recovered on the sorting tables. The journal for this period is supplemented by the section drawings, all of which date from this phase of the excavation. On the resumption of excavations in the Vestibule in October 1932, the journal again describes the trenches and their depths, supplemented by diagrams of their layout and details of the sediments. The locations of finds are given by trench as before. However, the depths of faunal remains were no longer measured although their approximate find depth can be inferred from the trench depth measurements. In most cases, the depths of artefacts continued to be recorded individually. Late in the excavation the level of detail given in the journals declined significantly. From late 1934 onwards the recording of find and trench depths became erratic, although the journal continued to list finds of fauna and artefacts made on each day. The ultimate reason for this decline in recording standards in the 1930s must lie with the departure of Dowie in 1928 followed by Beynon in 1935.

The excavation archive is supplemented by a series of photographs (see Online Supplementary Materials). One of them (Torquay Library, Kent's Cavern A. Ogilvie 025, showing one of the trenches) is reproduced by White and Pettitt (Reference White and Pettitt2012: fig. 4) and Zilhão (Reference Zilhão, Akazawa, Nishiaki and Aoki2013: fig. 3.5) who claim that this picture was taken in the Vestibule in 1927. We have established, however, that the entire archive of photographs of the Kent's Cavern excavations are by professional or press photographers and date from the 1938–1939 and 1939–1940 seasons, very late in the excavation. While the photographs are useful for showing some of the tools and methods used, they are heavily biased towards ‘human interest’ subjects and are clearly posed (Figure 5). Such press manipulation of photographs is routine and must be taken into account in their interpretation.

Figure 5. Press photograph of members of the excavation team at the sorting tables in the Vestibule, dating from between 7 November 1938 and 6 February 1939. Torquay library photo A. Ogilvie 019; see Online Supplementary Materials for further details. Reproduced with permission of Torbay Libraries.

The excavations were published annually in the British Association reports (Beynon et al., Reference Beynon, Ogilvie and Dowie1926; Dowie & Ogilvie, Reference Dowie and Ogilvie1927; Beynon et al., Reference Beynon, Dowie and Ogilvie1929a; Beynon & Ogilvie, Reference Beynon and Ogilvie1930; Dowie, Reference Dowie1930; Beynon & Ogilvie, Reference Beynon and Ogilvie1931; Reference Beynon and Ogilvie1932a; Reference Beynon and Ogilvie1932b; Reference Beynon and Ogilvie1933; Reference Beynon and Ogilvie1934; Reference Beynon and Ogilvie1935; Reference Beynon and Ogilvie1936; Ogilvie et al., Reference Ogilvie, Rogers and Tebbs1937; Ogilvie & Tebbs, Reference Ogilvie and Tebbs1938). Several other publications describe the 1926–1928 excavation in the Vestibule (Dowie, Reference Dowie1928; Beynon et al., Reference Beynon, Dowie and Ogilvie1929b), and the KC4 maxilla and some of the artefacts (Keith, Reference Keith1927; Smith, Reference Smith1940). Lake's (Reference Lake1934) survey of the cave shows trenches and other features described in the journals. These publications add up to a substantial corpus, and have the great merit that they were published within months of the excavations they described and must therefore have been written immediately after each season was completed. This is a far more satisfactory situation than the common practice of delaying publication until a final report is produced, often many years after the excavation was completed and with all the attendant problems of lost data and fading memory.

The excavation reports are also very useful for reconstructing the sequence. Most of the reports (with the exception of Keith, Reference Keith1927; Dowie, Reference Dowie1928; and Smith, Reference Smith1940) are concerned primarily with describing the excavations. This includes the locations and depths of trenches, and the types of sediments encountered. The latter are crucial and include much that is not contained in the excavation journals, suggesting that Beynon (the excavation geologist) probably kept his own notes, which have not survived. By comparison, the relatively brief descriptions of fauna and artefacts are less important because very substantial museum collections from the excavation are still available for study.

Taken together the 1926–1941 TNHS excavation in Kent's Cavern was a competently conducted enterprise. Its poor reputation is owed to previous workers either not examining the excavation archive at all, or doing so in insufficient detail to understand it properly. Several other British caves were excavated around the same time as the TNHS excavations, including Gough's Cave in 1927–1931 (Donovan, Reference Donovan1985; Jacobi, Reference Jacobi1985), Aveline's Hole in 1919–1930 (Donovan, Reference Donovan2005), and the Creswell Crags caves (notably Pin Hole in 1924–1936) (Jacobi et al., Reference Jacobi, Rowe, Gilmour, Grun and Atkinson1998; Pettitt & Jacobi, Reference Pettitt, Jacobi, Bahn and Pettitt2009). While these excavations differed considerably in their standards and methods of recording the stratigraphy and finds, none used a grid system for recording; the horizontal position of finds were recorded, if at all, using a simple measurement of distance from the entrance. Thus, at none of these sites can the distribution of finds now be accurately reconstructed in three dimensions. Comparing the TNHS Kent's Cavern excavation with the contemporary excavations described above shows that it was in fact typical of its period in its execution and level of recording. This is to be expected: the Kent's Cavern excavation was made not by unsupervised amateurs but with the funding and guidance of a committee of the British Association and would thus have been expected to satisfy a certain standard.

Stratigraphy of the Vestibule Cave Earth

Beynon et al. (1929) describe the Cave Earth in the Vestibule as an unstratified deposit of soil containing fragments of limestone and stalagmite. It is a good description of a typical colluvial cave earth, but the excavators also suggest that the deposit was ravined and overturned by heavy floods (Dowie & Ogilvie, Reference Dowie and Ogilvie1927: 304; Beynon et al., Reference Beynon, Dowie and Ogilvie1929b: 238), a proposal repeated by White and Pettitt (Reference White and Pettitt2012: 11). Zilhão (Reference Zilhão, Akazawa, Nishiaki and Aoki2013: 32) has gone further, to state that part or all of the Vestibule Cave Earth was accumulated by torrential flooding, and that this will have resulted in material of different ages becoming mixed with no internal stratigraphic order. In the light of this claim, the original excavators’ proposal needs reassessment.

The flooding hypothesis can be traced back to two observations, neither of which stands up to modern scrutiny. On 7 March 1927 the excavators discovered a void within the sediment, roofed by cemented boulders. Although it was broken into from the side of Trench C, the void itself wholly underlies the adjacent trenches (Figure 4). The journal entry for that day suggests that the void had been eroded by running water coming from the direction of the NE Gallery (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932: entry for 7 March 1927), an explanation elaborated by Dowie & Ogilvie (Reference Dowie and Ogilvie1927) and Beynon et al. (Reference Beynon, Dowie and Ogilvie1929b) where it was described as due to ‘torrential flooding’ or ‘heavy floods’. However, we now know that the void directly overlies a large choked passage opening into the Cave of Rodentia below (Figure 4). This passage is clearly the source for the Cave Earth that filled the Cave of Rodentia, and the sediment choke itself shows clear signs of suffosion (C.J. Proctor, unpublished data), providing good evidence that the void was formed by suffosion and collapse of sediment down into the Cave of Rodentia.

The second observation concerns the lithic assemblage. Dorothy Garrod (Reference Garrod1926) had identified material from Pengelly's 4 ft (1.21 m) level as Middle Aurignacian, whereas material from the TNHS excavations, supposedly from a deeper level, had been assigned by the Abbé Breuil to the Upper Aurignacian, thus implying an inverted stratigraphy (Beynon et al., Reference Beynon, Dowie and Ogilvie1929a; Reference Beynon, Dowie and Ogilvie1929b). However modern study has not supported the attribution of the Kent's Cavern Aurignacian assemblage to more than one industry (Jacobi & Higham, Reference Jacobi, Higham, Ashton, Lewis and Stringer2011). Moreover the only lithic artefacts illustrated by Garrod (Reference Garrod1926) from the Vestibule are culturally undiagnostic (R. Dinnis, pers. comm.).

In fact, the nature of the sediment is well known. The Vestibule Cave Earth forms part of a larger body of sediment, the proximal and distal ends of which have been investigated by modern excavations in the NE Gallery entrance and Wolf's Cave, respectively (Proctor, Reference Proctor1995; Proctor, unpublished data; Dinnis & Proctor, Reference Dinnis and Proctor2015). It is a typical colluvial cave earth gradually deposited in subaerial conditions by a combination of processes including cryoturbation, soil creep, minor hill-wash, and collapse of the cave walls and roof. Remnants of the top 6 ft (1.82 m) of the Vestibule Cave Earth are visible today in wall pockets in the position of Trench C and the sediment closely resembles the well-studied Cave Earth of the Wolf's Cave. The sediment in the Cave of Rodentia, derived from deeper in the Vestibule sequence, is broadly similar. The radiocarbon-dated sequence (see below) supports the hypothesis of a gradual deposition of sediment. The most rapid accumulation was between 7 and 11 ft (2.13–3.35 m), a depth range over which the Bayesian model implies deposition over around 1000–2000 years. Even assuming deposition over 1000 years, this equates to 4 ft (1.21 m) over this period, or 1.2 mm/year, a rate of accumulation entirely consistent with gradual deposition by the colluvial processes described above. Thus, there is no reason to believe that any of the Cave Earth in this part of the cave was any different in terms of its mode of emplacement. There is certainly no evidence for emplacement or overturning of the Vestibule Cave Earth by torrential flooding, contra Zilhão (Reference Zilhão, Akazawa, Nishiaki and Aoki2013: 32). Research into the sedimentology of the Cave Earth in this part of the cave continues and should further inform our knowledge of the Vestibule sequence.

Several distinct stratigraphic horizons can be distinguished within the Vestibule Cave Earth. These include the Late Magdalenian Black Band within the top 1 ft (30.5 cm) (Pengelly, Reference Pengelly1884; Jacobi & Higham, Reference Jacobi, Higham, Burdukiewicz, Cyrek, Dyczek and Szymczak2009); the top of the middle Devensian faunal assemblage in Pengelly's 4 ft (1.21 m) level (Pengelly, Reference Pengelly1864–1879); the main scatter of Aurignacian artefacts recovered by the TNHS from the 4–7 ft (1.21–2.13 m) level (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932); and a lower stalagmite floor recorded by the excavators in Trench C and the adjacent Entrance to the NE Gallery (ENEG) trench at a depth of 8 ft (2.43 m) (Beynon, et al., Reference Beynon, Dowie and Ogilvie1926–1932; Dowie & Ogilvie, Reference Dowie and Ogilvie1927; Beynon et al., Reference Beynon, Dowie and Ogilvie1929b; section drawings AR4271, AR4273 (Figure 3); see Online Supplementary Materials). All these features parallel the top of the Cave Earth and demonstrate that it is reasonably well stratified.

New Radiocarbon Dating

Work undertaken over the last decade has confirmed the widely held suspicion that radiocarbon determinations from Middle and Early Upper Palaeolithic sites in Europe often underestimate the true age, sometimes significantly (>10,000 years offset) (Higham, Reference Higham2011). This is due to the low concentrations of ¹⁴C in bones of this age and the fact that even a trace contaminant can create problems. The contaminant must be modern in origin for it to be significant, however. If the contaminant is ancient and ¹⁴C-free then it must be present in a significant proportion to shift the age to make it older. For this reason radiocarbon specialists usually only have to concern themselves with modern C contamination at this age. Previously, we identified animal-based glues on the surface of KC4 (Higham et al., Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011) that indicate a potential modern contaminant which if unremoved, would make the date younger than it would be in reality. In addition, the original determination on KC4 of 30,900 ± 900 bp may be an underestimate because of the lack of correction made at the time for pretreatment background. This correction is applied to take into account the routine uptake of trace contaminants derived from sample handling and pretreatment chemistry in the laboratory. It is important to make corrections for this, based on the dating of standards of known age that date to before the radiocarbon limit (see Wood, et al., Reference Wood, Bronk Ramsey and Higham2010). This all suggests that the age of KC4 must be at least 35,000 cal bp in age and not younger. It is, as Higham, et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011) concluded, very probably older than this, based on the regular underestimation of radiocarbon ages produced 20–30 years ago for this period, the presence of modern animal glue on the specimen, and the fact that the maxilla is surrounded by samples that are 5000–7000 years older. By this logic, KC4 must date to between 35,000 cal bp and 43,000 cal bp, and not later. White and Pettitt (Reference White and Pettitt2012: 22) make much of previous dating experiences wherein samples previously assumed on stratigraphic grounds to be Pleistocene in age turned out to be Holocene (see Street, et al., Reference Street, Terberger and Orschiedt2006). As is clear from the discussion above, this is not relevant to the Kent's Cavern case since the maxilla is already demonstrably old by virtue of the initial AMS date.

It is important to point to the severe underestimates of age demonstrated at the same site on several previously dated bones. When redated using ultrafiltration methods they became significantly older. OxA-6108, a direct date on a Coelodonta (rhinoceros) cranial fragment found 1 ft (30.5 cm) above the Homo maxilla initially yielded an age of 30,220 ± 460 bp. A new ultrafiltered determination on the same cranium gave an age some 7000 years older (OxA-13965: 37,200 ± 550 Bp) (Higham et al., Reference Higham, Jacobi and Bronk Ramsey2006). Determinations of other rhinoceros bones above this find spot were also underestimates, by some 2000 years.

White and Pettitt (Reference Tingley and Chandler2012: 22) criticize the selection of samples by Jacobi and Higham (Reference Jacobi, Higham, Ashton, Lewis and Stringer2011) to date the Kent's Cavern sequence, saying that the approach contradicts their written recommendations for selecting good samples for radiocarbon dating. At Kent's Cavern it was sadly not possible to select cut-marked or humanly modified bones, because apart from one bone pin there were none, probably due to the activities of hyenas (Jacobi, Reference Jacobi2007). Instead, samples were obtained to attempt to date the excavated sequence surrounding the archaeological finds, with careful attention to the depths recorded by the TNHS excavators.

Bayesian modelling helps to provide a probabilistic estimate of the degree to which the original age of KC4 is aberrant. With a Bayesian model it is possible to incorporate the prior information reflected in the relative sequence of finds made by the TNHS during their excavations. White and Pettitt (Reference White and Pettitt2012: 17) criticize this aspect of the work, saying that the ‘ordered stratigraphic sequence’ ‘critical to any Bayesian statistical program’ is not available. As we have shown, the stratigraphy and excavation is not as problematic as they claim. In terms of the Bayesian modelling, of course there is no right model and all models are to a degree wrong. However, several pieces of evidence allowed Higham et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011) to have some confidence about the model constructed:

1. 1) Articulated woolly rhinoceros bones provide support for inferring that at least some of the material in the sequence is in its original location on the site and has not been subject to significant post-depositional movement. These remains provide three indistinguishable radiocarbon ages from two bones. The bones in question were found together in ‘pocket 3’, 2 ft 3 in (68.5 cm) above the maxilla (Beynon, et al., Reference Beynon, Dowie and Ogilvie1926–1932: entry for 20 December 1926; Jacobi & Higham, Reference Jacobi, Higham, Ashton, Lewis and Stringer2011) and do not appear to have moved in prehistory. They ought to constrain the age of material below to be older.

2. 2) The excavators recorded that a lower stalagmitic floor at 8 ft (2.44 m) below datum covered about half the surface of Trench C and the adjacent ENEG trench, the remainder of the area being covered by a pavement of brecciated limestone blocks at the same level (Beynon et al., Reference Beynon, Dowie and Ogilvie1929b). This surface demonstrates the basic integrity of the sedimentary sequence in this area of the site.

3. 3) A sequence of age and depth consistent with the radiocarbon-dated samples suggests that the stratigraphic succession of material is not mixed or grossly distorted. This is supported by the Bayesian modelling, which showed a low number of outliers in the corpus of results, and a steady increase with depth. This provides evidence in support of a lack of mixing in the key sediments within which the maxilla was found.

White and Pettitt (Reference White and Pettitt2012: 15) suggest that the age estimate for KC4 ‘relies heavily on Bayesian modelling’, inferring, given their criticisms of the site and the excavation of the dated material, that it cannot be reliable. This is not the case. Comparing the likelihoods and the posterior distributions, as one can see in Higham et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011: fig. 2), is to compare virtually like with like. The Bayesian modelling does not have a significant effect on the posterior age ranges; we would know almost as much without the modelling than with it. The model does help to generate a probability distribution function (PDF) for the maxilla, but it does not dramatically alter the picture one would obtain simply by looking at the age-depth characteristics of the chronology and inferring the likely age that way. Bayesian modelling simply allows us to provide a probabilistic estimate of its age. We can examine the effect of the priors by sensitivity testing and varying the parameters of the model, and in doing so quantify the effect of the priors upon the posterior distributions, thereby testing the suggestion of White and Pettitt (Reference White and Pettitt2012: 15). Higham, et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011) undertook these tests and modelled determinations from the upper part of the sequence with and without constraining the results to be in absolute sequential order (see Higham, et al., Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011). We have undertaken other modelling with several different scenarios. In one of the least constrained models we merely included the determinations from 10 ft 6 in (3.20 m) up to the Lower Stalagmite as an unordered phase, and similarly made another unordered phase of the determinations above the stalagmite. The deeper determinations (below 12–13 ft, or 3.66–3.96 m) were also placed in a phase. Under these constraints the effect on the inferred age of KC4 is to move it to 43,070–40,120 cal bp (at 95.4% probability) rather than 44,180–41,530 cal bp. Other scenarios have very similar and overlapping results that suggest the model is insensitive to the priors applied and therefore that the age estimate is not likely to be significantly affected by these changes.

New AMS Results

We have also undertaken new AMS dating to test the previous model and determine how robust the Higham et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011) age estimate for KC4 is. We obtained nineteen new samples for AMS dating, using the journals of the TNHS excavations to guide our selection. We attempted to identify bone or teeth with a secure depth assignation with respect to the stalagmite levels and depths recorded. We were particularly interested in sampling material from around the find spot of the maxilla, but also above the lower stalagmite and in the Upper Palaeolithic sections of the site above 10 ft (3.05 m) below datum. The samples were AMS-dated at the Oxford Radiocarbon Accelerator Unit (ORAU) using the same laboratory methods as previously reported (see Higham et al., Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011). The new results are shown in Table 1. There were seven samples that failed to produce a good enough yield of collagen, or any collagen in some instances. These samples could therefore not be dated. In all other cases the collagen yields and other analytical data were acceptable and twelve new AMS dates were obtained.

We incorporated the results into the Bayesian model of Higham et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011). The model shows a similarly good level of agreement by age and depth (Figure 6). There are twenty-eight AMS determinations in the model and three appear as outliers (see Online Supplementary Materials). This is around twice as many outliers as would be expected on the basis of statistics alone. The outliers occur in the section of the sequence in Trench C below the Lower Stalagmite, within which the maxilla is found. Here, there are eighteen AMS determinations in total, fifteen of which are consistent and sit within the expected stratigraphic sequence.

Figure 6. Bayesian model for the Kent's Cavern sequence. The two new outlying determinations are shown in red. Model produced using OxCal 4.2 (Bronk Ramsey, Reference Bronk Ramsey2009) and the calibration curve (INTCAL13) of Reimer et al. (Reference Reimer, Bard, Bayliss, Beck, Blackwell, Bronk Ramsey, Grootes, Guilderson, Haflidason, Hajdas, Hatté, Heaton, Hoffmann, Hogg, Hughen, Kaiser, Kromer, Manning, Niu, Reimer, Richards, Scott, Southon, Staff, Turney and van der Plicht.2013). See Online Supplementary Materials for details of outliers and CQL code.

Two of the new results that produce ages at odds with the model are teeth samples. One of these is OxA-27442 (32,800 ± 500 bp), catalogued as ‘Megaceros’ and selected because it matched a journal entry for a deer tooth found at 12 ft 8 in (3.86 m), excavated on the day it was recorded. We have now re-identified this as the tooth of a Bovid, but finding as clear a match in the journal discloses uncertainty as to its proper position (see Online Supplementary Materials). Our assessment is that the specimen probably comes from between 10 ft 9 in and 12 ft 8 in (3.28–3.86 m). Regardless, the result does not fit within the expected sequence and is an outlier.

A second outlier in the model is OxA-27443 (32,200 ± 450 bp) a Cervus tooth from a depth of 13 ft (3.96 m). The sample was found on 29 April 1927. On this date the excavation journal lists a piece of deer jaw with three teeth, but the dated sample was a single deer tooth. For this reason we cannot definitely relate the specimen to a journal entry. The best indication of depth might be to take the range from which finds were recorded on that day—12 ft 9 in–13 ft 8 in (3.89–4.17 m) in this case. Again, however, the determination is not in agreement with the sequence established previously by radiocarbon.

We have tried to explain these two aberrant determinations by assessing, and rejecting, the possibility that the bones were picked out from the side of the excavation trench or that the results were due to problematic AMS determinations. Although the precise context remains elusive, it is likely that within the margins of uncertainty, the most plausible reason for these two results is post-depositional intrusion of younger material (see Online Supplementary Materials).

Further evidence of post-depositional intrusion is provided by refitting artefacts. Most of the artefacts found in the 1926–1928 trench consist of a Middle Aurignacian assemblage found at depths of between 4 and 7 ft (1.21–2.13 m), above the lower stalagmite floor (Jacobi & Higham, Reference Jacobi, Higham, Ashton, Lewis and Stringer2011). A number of break refits are present in this assemblage. While most lie within the 4–7 ft (1.21–2.13 m) depth range of the main assemblage, two include fragments which were found at much greater depths. The first of these is a blade (Jacobi & Higham, Reference Jacobi, Higham, Ashton, Lewis and Stringer2011: fig. 11.7.1) comprising three fragments found at depths of 6 ft 6 in and 7 ft 3 in (1.98 m and 2.21 m), and a fourth found at a depth of 9 ft 6 in (2.90 m), all in Trench C. The second is an oblique scraper (Jacobi & Higham, Reference Jacobi, Higham, Ashton, Lewis and Stringer2011: fig. 11.13.5), one fragment of which was found in Trench C at a depth of 7 ft 3 in (2.21 m), the other at 11 ft 6 in (3.50 m) in the adjacent ENEG trench. The occurrence of fragments of both these refits within the main Middle Aurignacian assemblage strongly suggests that the deeper fragments are derived from material of the same assemblage that has intruded into the older underlying sediments.

The probable explanation of this intrusion of younger sediment into the older part of the sequence is the presence of pipes within the sequence, within which sediment has moved down into underlying passages. The presence of one such structure is attested: the void beneath a cemented mass of boulders in Trench B, which directly overlies a substantial choked passage opening into the Cave of Rodentia below and was formed by suffosion and collapse of sediment into the underlying cavity. We have identified a similar feature in a rift to the northeast of the Cave of Rodentia (Figure 4). The southeast end of this rift opens into the Vestibule beneath the Trench C/ENEG trench boundary at a depth of approximately 22 ft (6.70 m), and is filled with suffosed sediment similar to that in the choked passage underlying the Trench B void, implying that a second pipe is almost certainly present within the Vestibule sequence above. Further evidence for piping is provided by the condition of the artefacts. Overall, the Aurignacian assemblage is in good condition; however, some fragments from the refitting artefacts found at depth show deep edge-notching which probably result from sediment shearing within the pipe (R. Dinnis, pers. comm.).

Although these pipes are deep, they must be of very restricted horizontal extent. As already noted, the well-ordered sequence of AMS determinations clearly implies that the bulk of the sequence is intact and undisturbed. Furthermore the excavators noted that the lower stalagmite floor and its lateral equivalent, the ‘pavement’ of limestone blocks, formed a continuous surface over trench C (Beynon et al., Reference Beynon, Dowie and Ogilvie1929b). Any extensive disruption of the sequence would have led to large-scale foundering of the lower stalagmite floor that would have undoubtedly been recorded by the excavators. Of the main Aurignacian lithic assemblage overlying the floor, only very few examples have intruded to a greater depth, with the same implication.

Incorporating the new AMS dates within the Bayesian model allows us to generate a PDF for the maxilla. We obtained a range of 42,350–40,760 cal bp (at 95.4% probability) for the PDF (Figure 7). This compares with the initial estimate of 44,180–41,530 cal bp produced by Higham et al. (Reference Higham, Compton, Stringer, Jacobi, Shapiro, Trinkaus, Chandler, Groning, Collins, Hillson, O'Higgins, Fitzgerald and Fagan2011). The estimates are not statistically significantly different but it is clear that the additional AMS determinations we have obtained enable a greater precision in the PDF.

Figure 7. Age estimate probability distribution function (PDF) for the KC4 maxilla based on the model shown in Figure 6.

Given that artefacts have intruded into the lower part of the sequence from above, the possibility that the KC4 maxilla has also been so affected must be considered. While this possibility cannot be eliminated, support for it being in situ is provided by its recorded location. It was not found immediately adjacent to the void, as this part of Trench C was not cleared until 31 March 1927, two weeks after the maxilla's discovery (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932: entry for 31 March 1927). Nor is its find date close to that of any of the known intrusive artefacts or fauna. The maxilla was found in three fragments, all at the same depth, and scattered horizontally over an area some 2 ft (61 cm) across (Beynon et al., Reference Beynon, Dowie and Ogilvie1926–1932: entry for 14 March 1927). This distribution is consistent with its being in situ. It is in marked contrast to the artefacts which have demonstrably been affected by post-depositional disturbance and intrusion, where different fragments were found scattered over a range of depths as would be expected when resulting from shear forces within a pipe of restricted horizontal extent.