APPENDIX G: MARINE MOLLUSCS by Greg Campbell

Introduction

This section presents the analysis and conclusions based on the 28,980 identifiable items of marine invertebrate organism recovered from 49 samples of the 22 deposits recognized as containing this type of remains. Of the identifiable items, 3400 were quantified by their dimensions, producing a dataset of approximately 10,000 measurements. While not all the assemblage was weighed, quantification by weight to the nearest gram was undertaken for 16.5 kg of these remains and of associated material, producing a dataset of approximately 320 weights.

Chronology of the samples

Of those deposits with samples containing shell, the earliest is from the Iron Age layer 644 in the south part of Trench 41, from the La Tène final period (c.120–c.50 BC). Shell-bearing samples from the early Roman Empire (c.50 BC–AD 260) came from two successive shell- rich layers (643 followed by 646) in the south corner of the building formed by walls F726 and F731 in the south part of Trench 41. Shell-bearing samples from the Gallic Empire (AD 260–AD 600) came from the base fill 523 of feature F600 pre-dating the gate structure in Trench 38, and from the very large shallow pit F1081 in the plot of land to the rear of post- built structure F1000 in the central part of Trench 41. Samples were taken from the pit’s initial fill 700 (rich in sea urchins), later fill 685, and final fill 681 (Sample 682 from its north part, and Sample 678 from its southern part and probably later, as the pit seems to have been filled in from the north side). The sole shell-bearing sample from the early medieval came from layer 708, the final activity within the complex of walls of possible ecclesiastical structures in the north part of Trench 41. This layer has been radiocarbon dated to 666– 775 AD (95% probability, calibrated).

Most shell-bearing samples came from the high medieval period (AD 1000–1500). Within the south-east part of Trench 41, deposit 634 was one of two shell-rich spreads which lay in the patches of rubble E of the postulated oval building defined by wall F623, and was limited to the south-west by curb F725. Shell-bearing layers 557 and 558 lay in the patches of rubble south-west of the curb and south of the oval building. Further to the south, on the flat area defined by earlier (Roman) rubble spread 552 and wall F726, were the two successive deposits 586 and 560 separated by a rubble layer 578 associated with the walls F676, F677, F629. In the same area lay the extensive spread of cockles and mussels 569 (referred to by this author as ‘the East midden’). These were all relatively early in the high medieval period, as was the earlier part of a long sequence of shell-rich deposits (beginning with layer 581, followed by 585, then 561 and ending with 543) in the extreme southern part of Trench 41 in the area between medieval structures and the remains of the rampart. This sequence formed what this author called ‘the South midden’.

To the south of Trench 41, in Trench 6 in Parcelle 20, shell-bearing high medieval samples were taken from layer 219 within Building 4, from layer 217 within the succeeding Building 6, and from external layer 204. Also in Parcelle 20, a shell-bearing sample was taken from late medieval deposit 451 in Trench 34.

Research aims

During the first seasons of excavation, several deposits from a range of periods were observed to be rich in marine shell. In 1998 four shell-rich medieval deposits were sampled (contexts 204, 217, 219, 451), and analysis showed that shells were preserved well enough at the site for identifications and measurements to be useful. From these results a number of questions emerged for later years in the excavation programme. The aims were

• to study the changes through time in the broad strategies used to collect shellfish (such as parts of the sea exploited), and to study the changes through time in the role of shellfish (such as changes in culinary preferences or the overall contribution of shellfish to the diet); • to study changes through time in the tactics used by collectors during specific episodes of collecting shellfish (the methods of extraction, the selection criteria for the various types of shellfish, or the tidal zone exploited for a particular type); • to determine whether the shell-rich deposits are tips or middens. At Le Yaudet shells are preserved only in shell-rich deposits. Typical deposits containing normal or average types and quantities of rubbish have lost their shell content through leaching. So the typical types and quantities of shellfish being used at various periods must be inferred from the rich deposits rather than observed directly. A deposit is rich in shell for one of two reasons. It is either a discrete tip (containing shell discarded as a mass from a single distinct event in the past, such as a feast), or it is a midden (containing shell discarded at the same place from a large number of successive events in the past). Discrete tips contain data directly relevant to tactics and indirectly to strategy. Middens contain data directly relevant to broad strategy, but must be dissected into something approximating individual discard events to investigate tactics.

Sampling methods

During the excavation programme large samples of the usual size (30 litres where available) were taken from the bulk of almost any deposit clearly containing shell. Such bulk samples were taken to recover data regarding past strategies by collecting a broad enough range of marine remains to include those that are quite rare. The aim was the recovery of about 2400 items in each large sample, so that the actual proportion of any particular shell type in the deposit would lie within 2 percentage points either side of the value estimated for that proportion by the sample, in 19 of every 20 samples (Veen and Fieller 1982, 296).

During the 2000 excavations extensive and thick shell layers were exposed and bulk sampled, and these were more finely sampled in 2001. A column 10 cm square in plan was excavated from the late medieval shell accumulation at the south baulk of Trench 41 (contexts 581, 565, 561, 543), with each 5 cm increment of depth collected as a sample. A distinct lens of shell within 581 but pre-dating the column was also collected. A similar incremental column 10 cm square in plan was excavated in late medieval shell accumulation 569 in the east baulk, where the deposit had been protected from contamination by the stone footing 618 and rubble layer 552. The late medieval shell spread 634 was sampled by excavating 15 cm squares 50 cm apart along a N-S line at its widest point. It had been hoped that sampling would include at least one 15 x 15 x 5 cm incremental unit from all shell-rich deposits revealed in the 2001 and 2002 seasons. Similar small samples similar in weight to a typical incremental sample (3.0 kg) were taken from several bulk samples where it was clear the deposit had not been incrementally sampled. These small samples were taken to study the nature of the deposits by recovering data on

• non-shell material: the amounts of soil, artefacts and stones being incorporated will vary during the accumulation of a midden, and therefore will vary with depth within the deposit. As middens are exposed for longer than discrete tips, the amount of non- shell material in an incremental sample will be relatively high in a midden and low in a tip, except for the upper surface where such incorporation will be similar; • shell richness: this varying in the incorporation of non-shell material also means that the proportion of shells to non-shell material will vary during the accumulation of a midden, and therefore will vary with depth within the deposit. As middens are exposed for longer than discrete tips, the proportion of shell to non-shell material in an incremental sample will be relatively low in a midden and high in a tip, except for the upper surface where such incorporation will be similar; • shell fragmentation: Shells exposed to the elements will have been eroded by leaching and been broken up by forces such as frost action, roots, and trampling. As shells in a midden are exposed for longer during its accumulation, they should be recognizably fragmented throughout its depth. For a particular kind of shell within a particular incremental sample, the ratio of weight of shell in the 10–6 mm size range to the weight in the >10mm range will be relatively large. In contrast, a discrete tip will have been exposed only on its upper surface, so fragmentation (as measured by the same weight ratio) will be relatively low, and highest at the surface.

These small samples were also taken to recover data regarding tactics of collection by statistical analysis of measurements taken from each kind of shell. Meat yield is proportional to shell size, so it will be possible to trace changes through time in the residents’ trade-off between palatability and meat yield within a type of shellfish, and between types of shellfish. Differences in shell form revealed by these shell measurements can be related to differences in habitat, and therefore to position in the tidal range.

Such samples should be large enough to contain statistically usable amounts of marine remains, while being small enough to relate to individual collection events. The aim was the recovery of about 200 items in each small sample, so that the actual proportion of a shell type in the part of the deposit being sampled would lie within 7 percentage points either side of the value estimated for that proportion by the sample, in 19 of every 20 samples (Veen and Fieller 1982, 295). The relationship between a past collection event and any small sample is indirect, because it can never be certain that a sample has collected a single collection event. In middens, which are accumulations of rubbish over time, any sample is bound to collect the margins of two or more collection events. Even a discrete tip may be the amalgamated rubbish from several teams of collectors working at several points on the shore.

Methods of extraction

Incremental samples

Each incremental sample was weighed to ± 0.1 kg without air-drying, as this correlates more closely to conditions within the deposit while in the ground. The soil volume of unconsolidated sample was measured to ± 0.5 litre. The samples were all of a consistent volume (2.0 ± 0.2 litres). Each sample was then gently agitated in a sieve suspended in water, first through a 10 mm (2/5th inch) and then through a 6 mm (about ¼ inch) aperture. In the one case where sea urchins were observed in the original sample (Sample 681 of deposit 700), the finer portion of the sample was further wet-sieved, through 2 mm (about 1/16th inch) and 1 mm (1/25th inch) mesh. The resulting two incremental sample fractions (>10 mm, 10–6 mm) or four fractions for sub-sample of deposit 700 (the additional 6–2 mm and 2–1 mm fractions) were rinsed and air-dried, after the larger shells were scanned for cleanliness and washed with a soft toothbrush if that was necessary for identifiability. The whole cockles recovered on site from within the incremental samples were washed, air-dried and weighed separately from the samples.

Both coarse fractions from each incremental sample were sorted into types of material: marine invertebrate remains whether whole or fragmentary (which consisted of limpets, cockles, mussels, sea urchin tests, and other marine molluscs); vertebrate remains (all of which were pieces of large vertebrate); pottery fragments; and stones. Each type of material recovered from each fraction was then weighed to ± 0.1g. These weights included the whole cockles from the relevant sample in the relevant material type.

In the case of the sub-sample from urchin-rich Sample 681, the 4–2 mm and 2–1 mm fractions were sorted for urchin remains, and these were separated into the various anatomical elements present within each fraction. Many of the elements were recovered only in these finer fractions.

Bulk samples

Marine shells were retrieved as part of the extraction of charred plant remains by flotation. Each bulk sample had its volume measured to the nearest litre, and was wet-sieved through a 10 mm (2/5th inch) mesh into a mechanical flotation machine. Following flotation (retaining the sample on a 0.5 mm mesh) the residue was wet-sieved through 4 mm (approx. 5/32nd inch) and 2 mm (about 1/16th inch) mesh, or 1 mm (1/25th inch) mesh if sea urchin remains had been recovered from the 3.0 kg sub-samples treated as incremental samples. The resulting coarse (>10 mm), medium (10–4 mm) and fine (4–2 mm or 4–1 mm) fractions of that bulk sample were air-dried if any animal remains were observed on any mesh during the wet-sieving. During the 1998 processing (cxts 204, 217, 219, 451) the finest fraction was not kept, since at that early stage animal remains were not expected to be well preserved.

When fully dry the >10 mm and 10–4 mm bulk sample fraction were sorted to extract identifiable remains of the shells commonly found (any piece of cockle or mussel shell with a valve, and any piece of limpet with an apex), and any pieces of any other organism. Unidentifiable fragments of cockles, mussels and limpets were extracted if there were no identifiable remains; this occurred in samples with a low density of remains. The fine fraction was sorted to extract any piece of any animal except pieces of the larger types of marine shell (fragments of this size being difficult to identify reliably). Shells or fragments of terrestrial snails were extracted from the >10 mm fraction only.

It had been observed during the sorting of the incremental sub-sample of Sample 681 that anatomical elements of sea urchins are retained in the finer fraction. Therefore, the finer fraction (4–1 mm) of any sample in which sea urchins were known (from the incremental sample processing) was dry-sieved through a 2 mm mesh, and the 4–2 mm fraction sorted to extract urchin elements other than spines (which are difficult to assign reliably to a type of urchin). The 2–1 mm fraction was sorted only for the urchin-rich deposit 700 and was discarded for the other samples (see the warning in the section on sea urchins below).

Methods of identification

Shells were identified by comparing their morphology and surface sculpture to specimens in the author’s own collection and to the identified shell collection held at the English Heritage Centre for Archaeology, Fort Cumberland, Portsmouth. The types used in the classification are not always equivalent to species, since not all shells are identifiable to species. Separating the rough winkles Littorina saxatilis (Olivi), L. rudis (Maton) and L. arcana Hannaford Ellis is not reliable on the basis of shell sculpture even on intact modern shells (Hayward, Nelson-Smith and Shields 1996, 188), so these were all identified as L. saxatilis agg. Any ormer was identified as ‘Haliotis sp.’, without attempting to distinguish between the two European species. The limpets were separated into Patella species based on surface morphology using Bowman (1981, 649) as a guide and with modern specimens of the Atlantic species including the Portuguese limpet (P. rustica) present. This approach (while not absolutely accurate) has already given reliable results on archaeological limpets (Cunliffe and Hawkins 1988, 37; Southward et al. 1995, 147). A limpet apex was identified as ‘Patella sp.’, if it was too small for the surface morphology to be clear, if the external surface sculpture had been worn away during life, or if it was only the nacreous inner layer (having been separated from the outer layer after discard). There is therefore a possibility that this type includes some unrecognized badly preserved examples of slit limpets and tortoiseshell limpets. Distinguishing between species of mussels (Mytilus) is problematic even with whole modern shells (Gosling 1992, 4–6). Taking Hawkins and Jones (1992, 53) and Gosling (1992, 3) as a guide, only those with distinctly down-turned umbones were recorded as French mussels (M. galloprovincialis). Therefore, shells of any hybrid animals were included with those of common mussel (M. edulis); those umbonal pieces too small to observe the umbones clearly (any with less than about 8 mm of margin) were classified as Mytilus sp. The laminated nature of oyster shells makes them particularly subject to degradation; where the oysters in a sample were fragments without umbones these were identified as ‘cf. Ostrea’ rather than the more technically correct ‘Family Ostreidae’. Taking Tebble (1966,105) as a guide, only those cockle (Cerastoderma) valves with furrows distinct on the internal surface between the muscle-scars or further into the umbones were identified as lagoon cockle (C. glaucum). Those that lacked distinct internal furrows between the muscle scars were identified as common cockle (C. edule); those valve fragments which were so small that they lacked the muscle-scars were classified as Cerastoderma sp. Those razor- shells bearing the distinct groove behind the anterior margin were identified as ‘Solen marginatus’ while those whose anterior margin lacked the groove were all grouped as ‘Ensis sp.’, without separation into species; fragmentary razor-shells were identified as ‘Solenidae’. Identification of the sea urchins became a distinct project in itself, and the methods developed to produce reliable identifications are described in the analysis of the sea urchins, elsewhere in this report.

For each sample (whether a bulk, an incremental or a 3.0 kg sub-sample), the number of each type of gastropod identified was found by counting each whole shell and apex of that type, except for the ormers (where each shell bearing an intact columella was counted) and common dog-whelks (where siphonal canals were more common than apexes). The shells of each type of bivalve in each sample were separated into left and right valves and the valves counted separately. These counts for the bulk samples are presented in Table 1. Nomenclature and taxonomic order follow Hayward, Nelson-Smith and Shields (1996). Due to the large numbers of limpets in Sample 680 of deposit 685 and Sample 679 of deposit 708, the limpets >10 mm were halved by heaping and dividing prior to identifying, and the count for each identified type of limpet was doubled to provide the count presented in Table 1.

Methods of measurement

Quantification by weight

Each type of material (limpets, cockles, mussels, sea urchin tests, other marine molluscs, vertebrate remains, pottery fragments, and stones) recovered from each fraction from each small sample (incremental or 3.0 kg sub-sample) was weighed to ±0.1 g. These weights included the whole cockles and intact lenses of limpets from the relevant sample in the relevant material type. These weight data were used to calculate

• the percentage of the small sample weight made up by the weight of each type of material over 6 mm in size. For pottery, the weight percent was calculated for fragments over 10 mm; any pottery weight percentage less than 0.01% was treated as zero; • the density in ground by weight: the dimensions of the ground extracted for each incremental sample were recorded, so the volume in the ground of each incremental sample was calculated by multiplying together its known dimensions and expressing this as multiples of 100 ml. The ‘density in ground by weight’ for each type of material was calculated by dividing the air-dry weight of that type over 6 mm in size by the number of 100 ml multiples in that incremental sample; • the ‘crush ratio’: the weight of a type of shell in the 10–6 mm fraction, divided by the weight of the same type of shell over 10 mm, for each of the three main types of shell (limpets, mussels and cockles), for any small sample where the weight of that type of shell in both fractions was at least 10 g. This takes the ‘fragmentation ratio’ for oyster-rich sites of Claasen (1998, 114), segregates it by type (since types of shell differ in their resistance to mechanical stresses) and inverts the calculation (so a higher value indicates a more broken-up assemblage).

Counts of identified material from small samples were used to calculate

• ‘species ratio’: The ratio of the number of identified valves of Mytilus galloprovincialis to M. edulis and of Cerastoderma glaucum to C. edule were calculated for each small sample where there were 20 or more identifiable valves of that genus; • the density in ground by number: the ‘density in ground by number’ for limpets, cockles and mussels was calculated for each incremental sample, by dividing the count of identified shells over 6 mm of that type in a sample by the number of 100 ml multiples in that sample (as calculated for ‘density in ground by weight’ above); • disinterred item density: the total number of shells identified was divided by the volume of soil after its excavation, recorded just prior to processing. This value is similar to that of density for bulk samples in Table 1, but is likely to be reduced (as the bulk sample includes common shells to 4 mm rather than 6, and uncommon shells down to 2 mm). Quantification by dimension

Following identification to species, measurements were attempted on the complete examples of each species of the three types of shell commonly recovered (limpets, cockles and mussels) on most of the samples. Preference was given to measuring incremental samples. If an incremental sample did not contain over 100 identifiable complete shells of a species, or if a deposit did not have an incremental sample, the bulk sample (or the portion identified) was used.

Most dimensions were measured to the nearest 1 mm with vernier calipers. If a dimension was less than 10 mm it was measured to the nearest 0.1 mm (to keep precision to 2 significant figures). All shells were eroded to some extent, so more precise measurements could be no better than an estimate of the original dimensions of the shell at deposition.

The dimensions recorded were

Limpets: Overall length, maximum width, and height of shell at the apex (rather than maximum height) (Fig. 1a). Squatness index: ratio of height to length. This takes the inverse of the length/height ratio of Jones (1984, 150), so a higher value indicates a more pointed shell.

Cockles: number of concentric growth-check rings, overall length, maximum height. Maximum half-breadth for complete right valves, or full breadth for intact cockles (Fig. 1b) was recorded to the nearest 0.1 mm. Many of the valves bearing three rings had a very small band of shell (less than 2 mm wide) beyond the final ring while some had a distinctly wider band of growth, so those with the wider growth band were recorded as having a ring count of 3.5. Shell shape: following Stanley (1970, 50), degree of elongation (length/height ratio) and obesity (height/breadth ratio) was calculated for the complete right valves with three rings in each incremental sample (and some selected cockle-rich bulk samples), for both species of cockle. Calculations were performed for a species only for those samples where there were at least three such valves of a species in a 3-ring cohort. The calculation was restricted to valves of a similar length and ring-count, since shell shape may change within a population with changes in age and size (Seed 1980, 42, fig. 4).

Mussels: only one valve (from South midden deposit 561 bulk sample 301 of 8.8 mm length) was complete enough to be measured.

Nature of the deposits: results from the small samples

Data ranges, trends and classification

Weight data for the small samples (incremental samples from the South and East middens and the 3.0 kg sub-samples for deposits not incrementally sampled on site) are available in archive. Values derived from these weight data (weight percent, weight density in the ground, and crush ratio) are presented in Table 2. For the middens, the increments are depth in cm from the upper surface of the midden. The results are presented in order of decreasing depth, corresponding to chronological sequence of accumulation at the point where the midden was sampled. For any midden such a column sample is unlikely to represent the entire history of the accumulation of the deposits sampled, and it is also unlikely that each increment represents an equal interval of the time-span of the accumulation. For the spread 634 (Sample nos. 447–50), the increments are distances in cm from the north edge of the deposit to the north edge of the 15 x 15 cm square. Weight percent and density for Sample 448 are underestimates because the 10–6 mm fraction was lost during processing, causing the author some distress (as he was the processor responsible).

Counts for the identified remains from the small samples are available in archive. Values calculated from these identification data (number density in the ground and species ratios) are presented in Table 3.

The per-increment weight percent values of Table 2 are plotted in Figure 2, the in-ground weight density values are plotted in Figure 3 and the crush ratios are plotted in Figure 4. For the incremental samples through the middens the results are plotted so that depth in midden decreases (and time increases) from left to right. The 3.0 kg sub-samples are plotted in chronological order from left to right.

The various categories of calculated values have a wide range. The narrowest range is that of pottery incremental percent and sea urchin incremental percent. Other categories range from quite small to very large, especially in the cases of the in-ground shell densities in the East midden (Table 2). It is difficult to imagine getting almost half a kilo of cockle shell into 100 ml (a small wine glass). Ranges are broadly similar for incremental weight percent for the main shell types and stone, and for in-ground densities. Crush ratios differ between the shell types, reflecting the different mechanical strength of each shell type (Fig. 4). Mussels are much more broken up than the other two shell types, even within the same deposits. Cockles tend to have higher crush ratios than limpets especially within the same deposit, although limpets have a greater range of ratios.

There is some co-variance between categories of calculated values. The per-increment weight percent (grams per 100 g of deposit) follows the same trend as the in-ground weight density (grams per 100 ml of deposit) in both middens, except for three samples. In the East midden the weight density of cockles in Sample 455 is lower than would be expected from Sample 456, and the weight density for mussels in 452 is unexpectedly high compared to Samples 453 and 454. In the South midden the weight density of limpets in Sample 465 is lower than would be expected for its weight percent compared to Sample 443. These deposits are made up of shells and voids between shells, so these differences indicate significant differences in the air content (or density of packing). In deposits made up principally of shells, weight percent varies less than weight per volume for the same change in density of packing. As density of shell increases (and air pockets between shells decrease) there will be some increase in the number of grams of shell per 100 g of deposit, but a far greater increase in the grams of shell per 100 ml of deposit. So Sample 455 is less densely packed with cockle shells than 456, Sample 452 is more densely packed with mussel shells than 453 and 454, and Sample 465 is less densely packed with limpets than 443.

The in-ground number density values of Table 3 broadly follow the same trends as the in- ground weight density values, differences between them coinciding with high values of ‘crush ratio’.

Classifying the ranges of the categories of calculated values into ‘low’ ‘moderate or typical’ and ‘high’ was not straightforward. Distributions were not normal; in some cases the standard deviation was larger than the mean, making it impossible to separate the low values from the moderate by using a cut-off of one standard deviation below the mean. A crude classification method based on the median of that category was used. The ranked values were inspected at about half the median for an arithmetic gap (a sudden increase in the difference between consecutive ranked values), and the value at which this sudden increase occurred was used as the upper limit of the ‘low’ class, and the lower limit of the ‘moderate’ class. A similar sudden increase in difference between consecutive values was sought about twice the median value, and used as the upper limit of the ‘moderate’ class. If a similar gap was noted in the upper quarter of the ranked values, this was used as the upper limit of the ‘high’ class, separating it from the ‘very high’ class. The reader is warned that this classification method has no statistical justification or comparable values from other sites known to the author.

The middens

The per-increment weight percent values and the in-ground density values for the South midden (Figs. 2a and 3a) both show this midden has two distinct stages. During the earlier stage (Samples 465–444) mussels are present but limpets predominate; densities of shell, stone and pottery are highly variable and do not coincide. There is no real difference in coarse inclusions between deposit 565 (Sample 444) and the rest of the earlier stage. During the later stage (Samples 443–439) stone density is moderate to high and tends to fall over time. Overall shell density is high and rises before gradual falling off over time. Pottery density coincides with limpet density, and these two fall off fairly rapidly, as mussels and then cockles alternate as the principal shell type. Crush ratio (Fig. 4) shows a similar pattern. Limpet fragmentation is very variable and occasionally extreme in the earlier stage prior to Sample 443, then becomes low and gradually increasing in the later stage. In the earlier stage cockles and mussels are too rare for ‘crush ratio’ to be valid. In the later stage breakage is moderate or high for cockles and consistently high for mussels compared to the site as a whole. In Sample 442 cockles are relatively well preserved for this stage but mussels are at their most fragmented for the site. Mussels are again highly fragmented in the last sample. The coarse inclusions are somewhat less dense and the mussels more crushed in the last deposit (543) than in the rest of the later stage.

In the East midden deposit 569 (Figs. 2b and 3b) shells are very rich until the last sample, stone is low after the initial sample, and pottery is almost absent. In the early stage (Samples 457–455) cockles grossly predominate and are at their least crushed state for the site (Fig. 4). Species ratio shows that cockles are fairly consistent throughout this stage, and estuary cockles are much less common than in the South midden (Table 3). Sample 454 marks a transition between the two stages. In this sample limpets are at their most rich in this midden and are well preserved compared to the rest of the site, stone content increases notably, mussels become grossly predominant but are quite fragmented (slightly less than the average crush ratio for the South midden). In the later stage (Samples 453–451) mussels grossly predominate, and are moderately crushed until the top of the midden. Species ratio shows that mussels are fairly consistent in mussel-dominated samples of this stage, and the proportion of French mussels to common mussels is about the same as that in the South midden. Comparison of the disinterred item density between the incremental samples (Table 3) and the bulk sample 304 (Table 1) shows that the bulk sample is of the same density as the more concentrated parts of the midden. The contribution to the total identified items by the finest fraction was negligible (seven items). The bulk sample was therefore taken from a point with a shell density similar to that of the richer increments. Preservation of remains differs between the bulk and incremental samples in this midden. While their disinterred item density is similar, incremental Sample 456 contained 121 measurable cockles in 4 disinterred litres (30 measurable valves per litre), while bulk sample 304 contained 108 measurable cockles in 15 litres (7.2 measurable valves per disinterred litre).

The interpretation of the incremental samples across spread 634 (Samples 447–50) is plagued by the loss of 10–6 mm fraction of Sample 448, the deepest increment and that with the largest number of shells. For this sample the crush ratio cannot be calculated, and weight percent, in-ground density and number density are serious underestimates. It is still clear that limpets are grossly predominant over mussels, and cockles are at least rare, and probably absent. Weight percent and in-ground weight density follow the same trend across the deposit; density of shell (as measured by either of these values, for either shell type) is high at the north edge, peaks at 50 cm S (Sample 448) and drops off further south. Pottery and stone density follow the opposite trend to shell density. Limpet fragmentation is low to average for the site (Fig. 4), while mussel content is so low that valid crush ratios could not be calculated.

The 3.0 kg sub-samples

Within Gallic Empire pit F1081 (deposits 700, 685, 681 north and 681 south in Figure 2c, respectively sampled as 681, 680, 682 and 678 of Figure 4) stone content is fairly consistent and is similar in range to that in the later stage of the South midden. Shell content and the dominant shell type vary between samples; sea urchins predominate only in deposit 700. Where preservation allows a valid calculation, mussel and limpet crush ratios are moderate or more commonly low in comparison to the site as a whole.

Dark Age deposit 708 (Sample 679) has a low stone content and high shell content, in which limpets and cockles have high crush ratios. The content of shell, stone and pottery of deposit 560 is comparable to that of the Gallic Empire pit fills and the more eroded portions of tip 634, with limpet and mussel fragmentation comparable to that in the mussel-rich portions of the East midden.

Pottery was a noteworthy component only in one of the sub-samples, Iron Age 644 (Sample 316), in which the limpet crush ratio is moderate despite the stone content being high.

Conclusions: midden or tip?

The South midden formed in two stages. In the early stage a soil (layer 581) was accreting, with occasional inputs of fine gravel eroding from waste from stone building construction or from upcast from diggings to bedrock. Occasionally limpets were discarded, sometimes as a tip and sometimes mixed with domestic refuse (as shown by pottery). The later stage was a midden in its usual sense. The area was covered by a series of tips, perhaps firstly a mix of domestic refuse and shells dominated by limpets, and then mixtures of cockles and mussels with either dominating. Between episodes of tipping soil continued to accrete, fragmenting the shell and homogenizing the shell content with further inputs of stone and gravel. The comparatively high level of fragmentation and low level of stone in 543 may indicate an old top-soil formed on the midden’s surface after abandonment.

The East midden was not really a midden in the usual sense; the shells were consistently rich and unfragmented, and the stone content was consistently low. First a large heap of cockles including rare limpets and mussels was discarded. There was a hiatus in which limpets were tipped and the shells in the top of the heap slightly fragmented. As there was no evidence of soil accretion or inputs of gravel, the hiatus was probably brief and the fragmentation caused by trampling. The heap of cockles was covered by a heap of mussels with rare cockles and limpets. The main difference between the incremental samples (preserved under a stone pad) and the bulk sample (presumably taken from the bulk of the deposit) was not density of shell (as measured by disinterred item density), but preservation of individual shells (measurable cockles per disinterred litre in the bulk sample is one quarter of that in a comparably dense incremental sample).

Medieval deposit 634 had a high shell content and little stone indicating it was a tip of limpet shells. Its low crush ratio indicated it suffered little trampling and was buried soon after it was discarded. Pottery and stone was richest in the shell-poor and fragmented margins due to soil accretion from subsequent waste disposal and erosion.

Interpretation of the nature of the other deposits must rely on comparing the content of their small samples with the incremental samples. Comparison of the incremental samples of the East midden with its bulk sample indicates high density of shell remains and low density of gravel and stone characterizes a tip, with the fragmentation of the shells being a less reliable guide. The marginal samples of spread 634 show that low pottery content in the presence of moderate to high gravel content may be the result of pottery being part of the erosive input rather than being discarded with the shells; therefore, only high to moderate pottery content can be taken to indicate shells are a component in mixed domestic waste. These criteria are applied to the 3 kg sub-sampled deposits below.

Iron Age layer 644 was a tip of domestic waste including some shellfish. The low crush ratios in conjunction with high stone content indicated rapid re-filling of the feature. Gallic Empire pit F1081 was filled with a series of tips of shell-rich debris with some domestic waste, and of rubble. The consistent moderate gravel and shell content shows that soil was accreting in the pit between discard episodes, but this soil accretion incorporated the shells into the deposits, preventing them from suffering fragmentation to the same extent as the South midden. Dark Age deposit 708 had a higher shell density and low stone content, indicating it was a tip. The low stone content is especially indicative here. The stone walls near the deposit must have been stable. Had these walls been undergoing construction, demolition or erosion following abandonment they would have injected high volumes of gravel and stone into this and other neighbouring deposits. The moderate to high crush ratios for all shell types indicate damage following discard was considerable.

Of the medieval shell-rich deposits with 3 kg sub-samples, layer 558 with its high shell content and low stone density was probably a tip, despite its relatively high crush ratios. Deposit 586’s high shell content and low fragmentation indicate it was a tip, its moderate stone density caused by incorporation of stone from overlying rubble 578. The deposit of shells 560 resting on rubble 578 was ambiguous in its content; similar to the Gallic Empire pit fills and the eroded parts of spread 634, it seemed most likely that this was a tip which has suffered more erosion and exposure than its well-sealed predecessor 586.

Discriminating between tip and midden for a medieval shell-rich deposit without 3 kg sub- sample was less clear, since data on stone content and shell fragmentation were not available for comparison. Disinterred shell density (items/litre in Table 1) was not a reliable guide, as this measure for midden deposit 561 was similar to that from tip 569. Deposit 557 was probably a tip because it was in the same area and stratigraphic position as tips 634 and 558. The remaining shell-rich deposits 204, 217, 219 and 451 are best thought of as middens, as these had low shell densities and most had large assemblages of domestic rubbish including medieval potsherds, all with similar sherd weights (14–18 g).

Composition of the assemblage

The count percentages of the main types of shell found at Le Yaudet (limpets, mussels, cockles, urchins and dog-whelks), calculated for the interpretable bulk samples from the data in Table 1, are shown in Figure 5. This figure includes count percentages from the data in Table 3 for some incremental samples selected because the only sample for the deposit was incremental (such as medieval tip 634 or base of South midden layer 581) or because there is a considerable difference in composition within a deposit (such as East midden 569 or South midden 561). This figure shows that the great majority of any sample was made up by these five main shell types, and that over 90% of any sample was made up of just three types: limpets, mussels and cockles. The samples can be separated into two groups:

1. Those deposits in which the shells were mostly limpets. Some were almost entirely limpets (79% or more), such as Iron Age deposit 644, the Early Roman tips, Gallic Empire pit fill 685 and the south part of 681, medieval tips 634, 557, 558, and 581 and probable middens 204, 219 and 451. Other deposits had shells that were mostly limpets (50–75%) with some mussels (10–40%). These include Gallic Empire pit fill 681’s north part, Dark Age tip 708, the medieval tips 560 and 586.

2. Those few deposits which are not mostly limpets and mussels. Gallic Empire pit fill 700 was an example, mostly sea urchins with a mix of limpets and mussels. Mussels predominated in likely midden 217. The East and South midden bulk samples were made up of a mixture of cockles and mussels, but the incremental samples show that the middens were dominated by a single type varying with depth, as shown in Figures 2 and 3.

These various shell assemblages are the product of the strategy employed by the gatherers when collecting the shells on the shore. For the first type of deposit the main shell sought was limpets, or limpets with mussels. For the second type shells other than limpets were sought. In some cases these were mussels, or sea urchins. Cockles and many of the other bivalves recovered (razor shells, carpet shells, tellins) are animals of sandy or muddy (‘soft’) shores, so deposits containing sizable amounts of both rocky and soft shores must represent gathering episodes on different shores. For those deposits interpreted as tips, the shells were probably gathered at about the same time, possibly on the same day, and those from similar types of shore (rocky or soft) were probably gathered during a single expedition to that shore. For those deposits interpreted as middens, this association in time and space cannot be assumed.

The mussels

Results

Of the 6233 mussel shells recovered at Le Yaudet, only one was intact, a single M. edulis right valve from South midden deposit 561 bulk sample 301 of 8.8 mm length. This is considerably smaller than the majority by weight of the fragments of mussel shell recovered, and smaller than many of the mussel shell fragments which could be assigned to species.

Bulk samples of three medieval deposits, all with considerable mussel shell content, contained fragments of tubes of keelworms. Common colonizers of hard surfaces such as rocks and shells in silt-free water, there are two species in Atlantic Europe, one with a three- ridged tube (Pomatoceros lamarckii Quatrefages) and one with a single ridge (P. triqueter L.). The former is predominantly a low inter-tidal animal, and the latter lives at low-tide fringe and below (Hayward, Nelson-Smith and Shields 1996, 112); both are common amongst mussel shells where they are a recognized pest of modern cultivated mussels (e.g. Cotter et al. 2003, 41).

East midden deposit 569 contained P. lamarckii tubes on a mussel shell and an elliptical dark rock about 40 mm long. Tip 560 contained a few fragments of P. lamarckii and a tube of P. triqueter attached to a worn cockle shell. Tip 586 (which preceded 560 in the same spot) contained loose fragments of both species and a P. triqueter tube attached to the inner surface of a mussel shell. Another mussel shell interior had a mat of bryozooans, another encrusting marine organism. This deposit was the only one in which barnacles (another colonizer of hard surfaces) were notable, as loose plates.

Conclusions

Due to the poor preservation of the mussel shells nothing can be inferred about the ideal size or acceptable size range at any period from these excavations. The association with keelworms in medieval deposits 560, 586 and 569 indicates that the mussels grew at very low tide. The collecting of mussel-shaped keelworm-covered stones and empty mussels with organisms on their inner surfaces is consistent with ripping up patches from mussel mats and sorting after transporting these home, rather than careful selecting by size during gathering. This would also explain the small mussel from medieval deposit 561, since such mats include mussels with a wide range of ages and sizes.

Limpets

Introduction

Limpets are the principal grazers of temperate rocky sea-shores between low and high tide, where the extent of their grazing is a major force keeping rocks bare of encrusting organisms such as seaweed, barnacles and mussels (Hawkins and Hartnoll 1983). Common limpets (Patella vulgata) are cold-water animals and are the main limpet of the north-east Atlantic, where they are common from shores in moderate shelter from the waves to those of moderately high exposure (Graham 1988, 78). In temperate waters china limpets (P. ulyssiponensis) co-exist with common limpets low on moderately wave-exposed shores and replace them throughout the shore with increasing exposure (Graham 1988, 80; Thompson 1980, 204). In warmer water the blackfoot limpet (P. depressa) becomes common on moderately exposed shores from middle to high tide level (Hawkins and Jones 1992, 17; Thompson 1980, 204–5), becoming common throughout the middle shore with increasing temperature (Hawkins and Jones 1992, 21). In the present sea temperature conditions these three species co-exist on the north Breton coast (Hawkins and Jones 1992, 20). Once removed from the rock the three are told apart easily by the colour of the large clearly visible foot: dark grey on the blackfoot, pale orange on the china and strongly tinted olive-green on the common (Hayward, Nelson-Smith and Shields 1996, 180).

Of the limpets intact enough to measure from the samples at Le Yaudet, only common limpets were regularly recovered in numbers useful for comparison. The statistics for the measurable common limpets (Patella vulgata) from the samples are presented in Table 4.

Results

Length

The mean length of almost any sample fell within the range of any other sample. Mean lengths fell in two groups, those below 28 mm and those, the majority, in the range 29–35 mm. The range of lengths was narrowest for the sample of the tip in Iron Age layer 644. Gallic Empire samples were quite scattered, even though the sampled deposits were all from the same pit (the last two from the same layer).

Mean and range was consistent for all samples from medieval tip 634, with a slight decrease for the most disturbed sample. Mean and range for spread 634 was very similar to that of the South midden, including its bulk sample. The samples from the East midden had some increase in size over time; the average limpet was distinctly smaller for the earliest incremental sample than for others. Almost all medieval samples were similar in size and broadly similar in range. The exception was Sample 302 from deposit 560, quite a bit smaller than the adjacent sample from deposit 586, the earlier tip in the same spot.

Length-frequency curves were prepared for a number of samples with over 100 limpets selected to cover the periods and deposit types recognized; the tip in Iron Age layer 644 was included (even though it contained only 19 limpets with measurable width). These profiles used the percentage of a sample’s measurable limpets in an interval (rather than the actual count), and were distributed by length (in 1 mm intervals), as usual for limpets, rather than the preferable age, since limpets do not produce clear annual growth-rings (e.g. Lewis and Bowman 1975). From these curves (Fig. 5) it can be seen that most tips are fairly symmetrical or slightly favouring the higher lengths (positively skewed), and bulk samples have a broader range than small samples. Those samples with limpets less than 17 mm long come from deposits with at least moderate mussel content, according to Table 1 or Table 3. Most examples exhibited a slight peak between 25 and 35 mm length; early Roman tip 646 was exceptional in having such a very small peak in this range.

Width

Plotting shell width against length produced consistent linear relationships for all the samples bearing more than ten limpets (Table 4); as the limpets grew in length they widened at a rate consistent for each sample. Significant differences in this rate of widening between the samples may have been caused by differences in the living conditions of the limpets between samples. For example, Craighead (in Bailey and Craighead 2003, 193) found that limpets on shores exposed to the full force of the sea were slightly narrower than those on more sheltered shores.

To test if the length-width relationship was really different between the samples, an analysis of covariance (ancova) was undertaken on the 2114 limpets in the 25 samples with over ten limpets, using only one sample (the largest, SS 448) from medieval spread 634 to prevent overemphasis on this deposit (Sokal and Rolph 1995, 504). The regression of width on length was highly significant for all 25 samples. For example, the F-statistic (Fs, the ratio of mean square due to regression to unexplained mean square) was smallest for Sample 316 of the tip in Iron Age layer 644. At 81.0, it was much greater than the expected value for 0.1% significance F.001[1,19-2] = 15.7. This accorded with the high values of regression coefficient R2 for the samples, which were 0.91 or greater (Table 4). The pooled regression of length on width was also highly significant (Fs within slopes = 31,511; F.001[1,2088] = 11). The least- squares regression slopes of length on width were very unlikely to be the same for all samples (Fs among least-squares regression slopes = 1.57, F.001[25-1, 2114-2(25)] = 1.15).

However, the analysis of variance (anova) on width for the samples showed that the variance of the width was very unlikely to have been similar for each sample (Fs = 19.7, F.001[25-1, 2114- 25] = 2.1), so the data do not conform to an assumption fundamental for an ancova. Also, the test recommended for unplanned comparisons between regression coefficients when the sample sizes are quite different, Gabriel’s approximate method for the GT2 test (Sokal and Rolph 1995, 499 and 244–9), could identify no significantly different sample at a 5% error rate. The slopes of width on length of Table 4 had a mean of 0.818 and a standard deviation of 0.046. Throughout the period under study common limpets have been collected from points on the shore where conditions allowed them to grow with a consistent width of 82 ± 5% of length.

Spirorbid encrustation

Some of the samples contained limpets with outer surfaces bearing the coiled tubes of polychaete worms of the Family Spirorbidae. North Atlantic inter-tidal spirorbids live in damp places such as among seaweeds or in rock crevices; while their upper tidal limit is high water of neap tides, only one species (Janua pagenstecheri) can live above middle shore outside permanent rock-pools (Knight-Jones and Knight-Jones 1977). No limpet bore a spirorbid worm tube on its inner surface. Such encrustation was usually sparse tubes of Janua pagenstecheri, Spirorbis spirorbis or S. tridentatus. About one-fourth of the encrusted shells bore colonies of tightly packed tubes, about half-way between the shell edge and apex and on one side, covering up to a third of the shell surface. The typically colonial spirorbid worm S. rupestris seemed absent as virtually no tubes bore its characteristic transverse growth rings. One of the better examples of these colonies (that from the tip in Iron Age layer 644) was examined by Professor E.W. Knight-Jones at Swansea, who identified the tubes as most likely those of S. inornatus, often an associate of tufted red algae and kelps.

As the figures on spirorbid-bearing limpets of Table 4 show, the proportion of limpets bearing spirorbids was always small, even in those samples with such limpets. It was reasonable to suspect that some samples included or lacked spirorbid-bearing limpets simply by chance. Determining whether a set of samples could be drawn from a population with a particular proportion requires the use of a statistical goodness-of-fit test, but the two most common of these tests (the G-test and the chi-squared (X2) goodness-of-fit test) were not readily applicable to these samples, since several samples included one or no spirorbid- bearing limpets. The G-test is unreliable where observations in a sample are less than 2 (Sokal and Rolph 1995, 702), and the chi-squared test is unreliable where expectations are less than 1 in any sample in a set and less than 5 in 20% of the samples in a set (Baxter 2003, 129). Therefore the nine samples in which two or more spirorbid-bearing limpets were observed were subjected to a X2 test to see whether these were homogeneous (could have all come from a population with the same proportion of clean to spirorbid-bearing limpets). Only the largest of the four samples of medieval tip 634 was included, to prevent overemphasis on this one tip. The results of the X2 test were then used to test whether the samples with one or no spirorbid-bearing limpets were also homogeneous.

A X2 2x9 contingency test was carried out on the set of the nine spirorbid-rich samples (those with two or more spirorbid-bearing limpets: SS 300, 301, 302, 305, 306, 448, 465, 680), using the ratio of all spirorbid-bearing limpets to all measurable limpets in the set (0.0506) to generate the expected numbers of observations for each sample. This gave an X2 value of 2 29.74, considerably more than the critical value expected once in 20 repeats (X .05[(2-1)(9-1)] = 15.51), making it unlikely that these samples were homogeneous. The source of the lack of homogeneity was sought by means of the un-planned tests of homogeneity of replicates outlined by Sokal and Rolph (1995, 722–3). The main contributions to this X2 value came from Sample 448 (16.71), so a X2 2x8 contingency test was carried out on the eight other samples, using the ratio of total spirorbid-bearing to total limpets in these eight samples (0.0394) to generate the expected observations. This gave an X2 value of 9.92, less than the critical value for the set of eight samples and less than the expected value for this set of seven 2 (X .05[(2-1)(8-1)] = 12.59), making these seven samples homogeneous.

For each sample, the probability of recovering the spirorbid-bearing limpets observed, if that sample was part of the population identified by the eight homogeneous samples, was calculated by expanding the formula for the binomial distribution (Sokal and Rolph 1995, 72) to the relevant number of observed spirorbid-bearing limpets for that sample, and adding up the probability of that and all less likely observations for that sample. The proportion used in these expansions was 0.0394, that for the homogeneous spirorbid-rich samples. These probabilities of recovery are presented in Table 4. Recovering one spirorbid-bearing limpet was likely in a group of 25 samples for Samples 161, 316, 443 and 454. Recovering no spirorbid-bearing limpets was unlikely or highly unlikely for SS 163, 304, 315, 678, 679; therefore there was a group of samples in which limpets were genuinely very sparse (absent or with a proportion of spirorbid-bearing limpets significantly less than that for the homogeneous samples). The numbers of spirorbid-bearing limpets was improbably unlikely for SS 679 of Dark Age tip 708, and just possible for Sample 680 of Gallic Empire pit fill 685. Results were inconclusive for SS 34, 160, 317, 442, 453, 456, 681; due to the small number of limpets recovered in each of these samples it was not possible to tell if these samples belonged to the homogeneous or to the very sparse group. Spirorbid-bearing limpets were improbably rich in Sample 448 of medieval tip 634 and Sample 465 of the limpet lens in deposit 581 at the base of the South midden. Therefore the samples can be differentiated into four groups: those rich in spirorbid-bearing limpets (SS 448 and 465, with about 12% of the limpets bearing spirorbids), those with a moderate content of about 4% (SS 300, 301, 302, 305, 306, 316, 443, 454, 680, 682), those where spirorbid-bearing limpets were very sparse or absent (SS 163, 304, 315, 678, 679) and those which are inconclusive due to small limpet numbers (SS 34, 160, 161, 317, 442, 453, 456, 681).

Height, and its relation to length

The percentage frequency of common limpet squatness ratio for those samples with at least 40 shells (and for the chronologically important La Tène Iron Age sample) is presented in Figure 7. The intervals are the same as those used by Jones (1984, 104 and 150), converted from Jones’ length/height ratio to squatness (height/length). These histograms came in three forms. Most were dominated by limpets in the 0.34–0.40 squatness interval, with less than 30% in the flanking intervals. Among these ‘moderate’ distributions were Iron Age and early Roman deposits, the Dark Age tip 708 and the medieval middens 204 and 219. Several had a sizeable proportion (over 30%) of flatter limpets, in the lower 0.29–0.33 squatness interval. These ‘flatter’ distributions included the earliest South midden incremental sample 465 from the lens in layer 581, the earliest East midden incremental sample 456 the cockle-rich base part of deposit 569, and medieval tip 560 (in which this interval was dominant). ‘Sharper’ distributions (in which sharper limpets, in the 0.41–0.50 squatness interval, contributed over 30%) included the limpets from urchin-rich base fill 700 of the Gallic Empire pit, and the samples from layer 561 in the South midden (bulk sample 301 and increment 443).

The distributions of squatness plotted against length for a selection of samples are presented in Figure 8 in order to observe any inter-relationships. These distributions came in two forms. Some samples had an even coverage of limpets across the range of length and squatness; among these ‘even’ samples were early Roman tip 646, Gallic Empire pit fill 700, medieval midden 204. The other form had limpets ‘concentrated’ in part of the ranges of length and squatness for the sample, almost always small (21–35 mm long) and flat (squatness index 0.26–0.40) compared to the full range of the sample. Some had concentrations of very small limpets (21–29 mm, such as Gallic Empire pit fills 685 and 681 Sample 678, Dark Age tip 708, medieval tip 560), while others had limpets concentrated in a slightly larger size range (25–35 mm, such as medieval tips 634, 557, 558, 586).

East midden 569 in bulk (SS 304) was a ‘concentrated small’ form, with a few very small limpets less than 17 mm. While numbers were small from the incremental samples, the earliest of these (SS 456) seemed possibly a ‘concentrated’ form, with the earliest mussel-rich sample (SS 454) possibly a ‘concentrated’ form and the later sample (SS 453) of ‘even’ form. The bulk sample seemed a product of all these incremental samples, with SS 454 predominating. All samples of South midden deposit 561 (SS 301) were weak ‘concentrated’ forms; the early lens of limpets in layer 581 (SS 465) was the same form, with a slightly narrower squatness range (0.29–0.33).

Several other deposits were also difficult to assign due to the small number of measurable limpets in their samples. Iron Age tip 644, early Roman deposit 643, Gallic Empire pit fill 681’s north part (SS 682) and medieval middens 219, 451 may have been ‘concentrated small’ forms.

Those common limpets with spirorbid worm tubes attached were almost always outside the concentration of small flat shells, in the large conical forms. This was especially distinct in medieval tip 634.

Discussion

Length

There is a considerable literature on the population structure of common limpets on modern European Atlantic shores, in which length-frequency histograms have been prepared for a wide range of shore conditions. The population structure of the common limpet is more dependent on conditions in the immediate vicinity than it is on mean sea or air temperature. Cold shores in the Northern Isles of Scotland (Baxter and Jones 1987, 39) and cool shores of the north-east English coast (Lewis and Bowman 1975) were comparable to those in the more temperate conditions of south-west Ireland (Thompson 1980, 203). However tempting, the length distributions of the Le Yaudet samples cannot be compared directly to those in the literature, because the archaeological assemblages were biased as the result of selection by a predator (humans gathering limpets). Gatherers may have favoured a particular size range and selected these from a limpet population, rather than sampling a population evenly according to length. Limpets may have been gathered from many locations on the shore during a single gathering event, conflating groups of limpets from populations with differing length distributions. Limpet size distributions at any one location also change throughout the year as the limpets grow. Taking a bulk sample from a midden also conflates groups of limpets gathered separately, compounding selection bias and conflation within gathering events; this is the cause of the wide ranges in curves for bulk samples (Fig. 6).

The mean length and standard deviations of Table 4 indicate the gatherers’ idea of an ‘acceptable’ limpet was 19–50 mm. The length-frequency distributions of Figure 6 show that in most cases the gatherers favoured a smaller size of 20–35 mm while accepting larger animals. These concepts of acceptability seem to have remained basically the same throughout the occupation of Le Yaudet. The gatherers for the La Tène deposit produced the smallest range, making them the most discriminating gatherers; early and late Roman gatherers seemed distinctly less fussy than the earlier or later stages. The limpets discarded with cockles at the start of the East midden were smaller and more selectively gathered than those discarded later with the mussels. Very small limpets (less than about 17 mm) were gathered accidentally with mussels, since they were found in samples with at least moderate mussel content.

Lewis and Bowman (1975, 178) found in cool seas an average length of about that usual for the samples from Le Yaudet (30 mm) can be achieved by common limpets 3–4 years old in good conditions of growth (low or mid-tide with at least some bare rock for grazing), but may never be achieved in poor growth conditions (middle or high shore with extensive barnacle or mussel cover). The ranges observed in the samples can be regarded as ‘typical’ of mature common limpets in good to fair growth conditions (e.g. Lewis and Bowman 1975, 195–7). In seas warm enough to support blackfoot limpets growth ranges could be expected to be slightly higher, as food would be available for more weeks in the year.

The average length decreased slightly with increasing disturbance in similar deposits (especially for medieval tip 634), implying that such disturbance damages large limpets more than smaller ones. The archaeologists’ argument that decreases in size over time within a midden demonstrates over-exploitation of the limpets (e.g. Jones 1984, 195) may therefore be simplistic.

Width

The consistent nature of the relationship between width and length in common limpets was difficult to put into context. While it is common to measure width in studies of modern limpets (e.g. Baxter 1983, 150), the results are seldom discussed. It is best to assume until it is demonstrated otherwise that a width-length ratio of 0.82 ± 0.05 is normal for the common limpet, and there was not enough variation between the conditions of growth of the various samples to alter this ratio significantly.

Spirorbid worm tube encrustation

There were two statistically significantly different groups of samples: those samples with significant numbers of spirorbid worm-tubes (the majority), and those with no or insignificant numbers of such tubes. Spirorbid worms live in damp sheltered places often among specific seaweeds. Therefore, samples with significant spirorbids had limpets from damp sheltered shore areas probably with some weed cover, while samples without significant spirorbids had limpets from drier more open conditions. Therefore, the occupants of Le Yaudet generally sought out limpets from sheltered areas of the shore. However, they sought out limpets in predominantly drier and fairly weed-free areas at times in the Gallic Empire (south part of fill 681), at times in the Dark Age occupation (tip 708), and occasionally during the medieval period (midden 204).

The nine of the 11 interpretable medieval samples had limpets from at least some shelter, some from considerable shelter (tip 634, early South midden deposit 581). While limpets from shelter were preferred throughout the occupation of Le Yaudet, their predominance in the medieval samples may indicate sheltered shores were the common type of medieval inter- tidal rocky shore. This may be a reflection of the climate at the time; cool damp climates favour inter-tidal seaweed growth while warm periods favour open rock (Southward et al. 1995, 131).

Height and its relation to length

It is well known that common limpets are more conical with increasing tidal level (e.g. Graham 1988, 77). Archaeologists studying limpets have used this trend to reconstruct shore position from which limpets have been gathered, e.g. Jones (1984, 151) and Russell et al. (1995, 282). Jones (1984, 192) amalgamated low-shore and high-shore groups of modern limpet populations from an island in the Inner Hebrides (south-west Scotland), producing a ‘low-shore’ distribution curve with a peak in the 0.29–0.33 interval slightly stronger than the ‘flatter’ distribution defined here, and a ‘high-shore’ curve with the peak in the 0.41–0.50 interval slightly greater than in the adjacent smaller interval. Only one distribution (medieval tip 560) conformed to that of Jones’ ‘low-shore’ curve, with the others intermediate between Jones’ curves; none of the distributions of Le Yaudet samples conformed to Jones’ ‘high- shore’ curve.

Russell et al. (1995, 287) also amalgamated modern Inner Hebridean island limpet populations into three distributions: a ‘low-shore’ curve similar to the ‘flatter’ curve as defined here; a ‘mid-shore’ curve equivalent to the ‘moderate’ distribution defined here; and a ‘high-shore’ curve with the same shape as the ‘moderate’ distribution defined here but with the peak in the 0.41–0.50 interval. Using this scheme the Le Yaudet distributions are principally ‘mid-shore’ and ‘low-shore’ with intermediate distributions between these two. ‘High-shore’ type limpets were included in urchin-rich Gallic Empire fill 700, and less so in increment and bulk samples of South midden deposit 561.

Clearly such classification schemes are highly dependent on the method of amalgamation of results. In amalgamating their results both Jones and Russell et al. must have assumed that increasing time spent uncovered by the tide is more important for forcing limpets to grow pointed shells than the local conditions. However, local conditions have a profound effect on limpet populations. Cover (rock crevices, seaweed, mussel clumps and rock-pools) increases local dampness, reducing stress by dessication or over-heating. Lewis and Bowman (1975, 187) state, with respect to a limpet population’s mean height: ‘In the interplay between the influence of local biological environment and of tidal level it is the former that is again the stronger.’

In those samples where highly pointed limpets include those with spirorbid encrustation, this increased pointedness cannot be due to dry open conditions because spirorbids are animals of damp sheltered conditions. Dry conditions need not be the only cause of increased pointedness. Any chronic stress on the edge of a limpet’s shell would force the animal to reduce the exposure of its mantle (the living tissue along the shell edge which builds up the shell), thereby reducing the relative proportion of shell that can be built up outwards and away from the animal. In a study of limpet shell hydrodynamics Denny (2000, 2620) demonstrated that wave force was not the main determinant of limpet shell shape and summarized five other forces which may be involved. Most of the spirorbids observed on the shells from Le Yaudet are characteristic of some seaweed cover. Therefore it would seem that increased seaweed cover causes increased pointedness in the same manner as increasingly dry conditions. The seeming ‘high shore’ limpets in Samples 301 and 443 of South midden deposit 561 cannot be high-shore, because these samples include significant numbers of spirorbids. The spirorbid evidence is inconclusive for urchin-rich Gallic Empire fill 700.

It was reasonable to suspect that the spirorbid-encrusted limpets were different, part of a group collected from a more sheltered part of the shore than the bulk of the limpets found in a deposit. Limpets from bare rock would be expected to be smaller and more conical on average than those from shelter, yet the squatness-length scattergrams of the limpet-rich samples (Fig. 8) showed the smaller limpets tended to be flatter than the spirorbid-bearing limpets in any sample. Therefore most of the limpets within any sample with significant numbers of spirorbid-bearing limpets were collected from sheltered shores, regardless of size.

Conclusions

Comparing shell height-length distributions demonstrates only that most Le Yaudet limpets were growing in conditions commonly found in the middle and lower shore. Gallic Empire urchin-rich fill 700 and South midden deposit 561 include some limpets from conditions more common in the upper shore.

Examining either length distribution or height-length relationship alone without reference to the other obscures the differences in height-length distribution between limpets of different sizes. While larger limpets did tend to be more conical in the limpets recovered from Le Yaudet, the squatness-length scattergrams of Figure 8 showed the relationship was far from straightforward, in contrast to the relationship between length and width.

The patterns recognized in the scattergrams of Figure 8 reflected the tactics employed by the gatherers. Samples in which the squatness-length scattergrams are ‘even’ reflected an even effort of those gathering limpets across shore positions, without concentrating effort on any particular part of the shore. Only three limpet distributions were ‘even’ (early Roman tip 646, Gallic Empire urchin-rich pit fill 700, medieval midden 204), so gatherers did not often range across the shore for limpets. Deposits in which the squatness-length scattergrams were ‘concentrated’ reflect concentration of effort by the gatherers on that size range in shore positions where that level of squatness was common. No samples had a distribution in which limpet sizes were concentrated in the large size range. Therefore there is no evidence that gatherers sought out large limpets at the expense of smaller sizes. The occupants of Le Yaudet seldom if ever went out with the aim of getting limpets with the highest individual meat yield.

Usually the gatherers employed the tactic of concentrating on small flat limpets from points on the shore where they would grow to this size relatively quickly (mid- or low-shore usually in shelter), probably because smaller flatter younger animals were less tough or better tasting than larger older meatier limpets. This was the common tactic regardless of period, and the more common practice within any period with a number of sampled deposits. Commonly the gatherers concentrated on limpets between 21 and 35 mm long, while also collecting up larger limpets available nearby. Limpets were gathered from open dry areas in the middle to low shore for the early Roman layer 646, and the south part of Gallic Empire fill 681 (SS 678). Limpets were gathered from open dry areas on the middle and upper shore for Dark Age tip 708 and medieval midden 204, and from the middle and upper shore in unknown conditions for urchin-rich pit fill 700. Limpets were gathered from damp sheltered conditions probably with some seaweed cover on the middle to low shore for the Iron Age tip within layer 644, Gallic Empire pit fill 681’s north part (SS 682), all the medieval tips including 581 at the base of the South midden, the earliest increment in South midden layer 561, the transition between cockles and mussels in the East midden (SS 454).

The cockles

Statistics

The principal statistics for the measurements on the common cockles (C. edule) from the samples at Le Yaudet are presented in Table 5. With average lengths of 26.5–31.3 mm and average ages of 3.1–4.6 years old, these were large and old in comparison to modern cockles where averages of 18.8–24.8 mm and 2 years old are regarded as harvestable (Hancock 1967, 139). The lengths of cockles were similar to ranges in which lengths of limpets were concentrated in limpet-rich samples at Le Yaudet. In the South midden the numbers of cockles and their average size and ages were quite variable between samples, reinforcing the view that this was a midden rather than a single tip of shells. The East midden had a more consistent average size and age, reinforcing the view that this was essentially a single tip. The cockles from both middens were similar in the proportion found intact (articulated and closed in the ground). In the East Midden 569 the proportion of right valves in intact shells to all right valves over 6 mm was about 3% (2.7% in incremental sample 455, 3.2% in Sample 456, 4.2% in Sample 457), and 6% in South Midden incremental sample 442. These cockles were probably discarded before use because they were partly open since such ‘gapers’ are more likely to be inferior (diseased or dead) to those shut tight.

Length-frequency histograms

Percentage length-frequency histograms of the various species of cockles were constructed using 1 mm intervals of length for a selection of samples rich in cockles. These histograms are presented as Figure 9. In the South midden the distribution was very different between the only incremental sample with reliable numbers of cockles (Sample 442) and the bulk sample, which was dominated by cockles over 29 mm. The distribution of the two species of cockle was very similar within Sample 442, the common cockles having a small number of large individuals which were not found in the lagoon cockles. In the East midden the length distribution was quite similar for all the samples, and was more evenly distributed than the South midden increment. However, the bottom incremental sample 457 had fewer small cockles (less than 27 mm long) than the others.

Length in relation to age

Plotting the length of C. edule against the ring count of the same shells was undertaken for all the incremental samples where there were more than three shells in every cohort (all shells with the same ring count). The results of this are presented graphically as Figure 10; the error bars are the standard deviations for Sample 456. For the East midden there were four usable samples (incremental samples 455, 456, 457 and bulk sample 304). The average length of each age cohort within these samples was very similar. Overall, these average lengths followed the anticipated growth curve of von Bertalanffy, gradually approaching an upper size limit over time. This indicated that even at this very broad level the sub-samples are all drawn from the same population in the wild. However, in the single usable sample from the South midden (Sample 442 from deposit 561) the cohorts did not follow the expected growth curve, indicating that even within this small sample the cockles are not likely to have come from a single site in the wild. The average length of shells given the ring count of 3.5 was virtually the same as those with the 3-ring cohort in the same incremental sample, and the 3.5-ring cohorts were omitted from this analysis.

Von Bertalanffy’s equation relates some measurable quantity of an animal (here, length) to its time since arriving in its adult conditions (t-to) and to a maximum asymptotic value of that quantity that can be achieved in the circumstances in which it lives (here, an asymptotic length L∞):

–K(t-to) Lt = L∞ (1 – e )

For comparison with other more recent cockle populations the parameters for the von Bertalanffy equation were estimated from average length for each of the six cohorts in

Sample 456. Visual inspection of the figure gives an estimate of L∞ of 33 mm. Re-arranging the above equation yields

K = (-1) ln (1 - Lt / L∞ ) / (t-to)

Substituting the average length for a cohort and that cohort’s ring-count for t-to into the equation above gave an average value of K of 0.53 ± 0.06.

Common cockles from the East midden were relatively small. Cockles in northern Scotland had asymptotic heights H∞ typically greater than 34.8 mm (Jones and Jones 1981, 210), showing that they regularly grew taller than the East midden shells managed to grow in length. Cockles in the Dutch Wadden Sea in populations where tidal uncovering did not affect growth achieved lengths of 28.3 ± 2.0 mm in two years (Kristensen 1957, 441); another nearby population achieved L∞ of 40 mm (Ramon 2003, 429). Despite being relatively small, the cockles from the East midden had a relatively high K value, indicating they achieved this size quite quickly. K values calculated for common cockles in northern Scotland were seldom over 0.6 (Jones 1995, 205), and were about 0.40 for the Dutch Wadden Sea (Ramon 2003, 429).

There are some relationships between K, L∞ and conditions on the shore, especially time submerged due to the tide (Kristensen 1957, 425). The values for the East midden cockles are very similar to K of 0.53 , L∞ of 35.1 mm for low-shore cockles in north-east Wales (Sanchez-Salazar et al. 1987, 251). However, local conditions cannot be reconstructed directly since these relationships are similar at different sites, and differ from year to year at a given site (Jones 1995, 212). The von Bertalanffy parameters in the East midden remaining consistent for several cohorts showed that shore conditions and the density of the cockle population remained undisturbed over the several years in which the East midden cockles grew. In particular, there seems to have been no episodes of cockle harvesting. The South midden annual growth evidence showed that two other groups were being harvested, one of large and rapidly-growing cockles dominated by the two-year-old cohort, and another group more stunted and slow-growing than the East midden and dominated by the three- and four- year-old cockles. The lagoon cockles having a small average length and a length distribution lacking large individuals showed that the lagoon cockles were probably harvested principally with the smaller three- and four-year-old common cockles.

Species proportions

The ratio of the two species of cockles was noticeably different between the two middens (Table 3), the proportion of lagoon cockles (C. glaucum) higher in the East midden than in the South. To discover whether this was likely to be a chance variation, the average proportion for all the incremental samples in the East midden (104 C. glaucum in a total of 994 valves) was compared to the incremental samples with usable number of cockle shells as a series of 2x2 contingency tables, the X2 value calculated, and the associated probability determined for one degree of freedom (P[X2,1], since the table was 2x2).

For the East midden, Sample 455 (33 lagoon cockles in 310 valves, instead of the expected 32) was virtually indistinguishable from the average for the East midden (X2 = 0.035, associated probability P[0.035,1] = 85%). Sample 456 (63 lagoon cockles in 516 valves) was well within chance variation from the East midden average (X2 = 1.68, P[1.68,1] = 19.5%). However, Sample 457 (8 lagoon cockles in 168 valves) had too few lagoon cockles compared to the East midden average to be the product of chance alone (X2 = 5.83, P[5.83,1] = 1.6%). The South midden Sample 442 (61 lagoon cockles in 155 valves) was found to be incredibly unlikely to be from a population like the East midden average (X2 = 138, P[138,1] = 6.9 x 10- 32). Therefore, there was a profound difference in the species make-up between the populations being harvested for the two middens. It was also likely that there was some difference in species make-up between the upper two-thirds of the East midden and its base.

The difference in species make-up can be explained by differences in the conditions in which the two species of cockle thrive. While the areas in which the two species of cockles live overlap, C. glaucum generally lives nearer river mouths (Hayward, Nelson-Smith and Shields 1996, 244) where the mud-flats are sheltered from currents and wave action which therefore have finer more muddy sediments, and where the water can become even more brackish than C. edule can tolerate (e.g. Mariani et al. 2002, 483 and 488–9). The South midden samples having a considerable component of lagoon cockles showed they were harvested from part of the shore where estuarine conditions had some influence but were not dominant. The significantly lower lagoon cockle component in the bulk of the East midden samples showed they were harvested from part of the shore where marine conditions were more dominant. The significantly smaller lagoon cockle component in the earliest East midden incremental sample showed that these were harvested from a more fully marine part of the shore than the bulk of the East midden, yet so close that their growth rates and asymptotic sizes were the same (Fig. 10: cockle length v age).

Shell shape

Cockle shell obesity (height/breadth ratio) is plotted against degree of elongation (length/height ratio) as Figure 11 for each cockle species from each incremental sample with a usable number of cockles. The 2-ring cohorts from the South midden Sample 442 were compared with the 3-ring cohorts of the East midden since their sizes and numbers were similar. The error bars shown in the figure are for the three-year-old common cockles from East midden Sample 456 and the two-year-old common cockles from South midden Sample 442. The common cockle shells from Sample 442 with three to four rings were included, since age-height data indicated these may have come from a different population. All lagoon cockles from Sample 442 were also included, since age-height data and length distribution indicated that these may have been harvested from the same population as the three-year-old common cockles from the same sample.

As it lives within the sediment near the surface, a cockle is subject to disinterment by wave- scour or a predator’s probing; it must then rebury itself at speed if it is to avoid death by being eaten or (if disinterred when uncovered by the tide) by drying out or overheating. Stanley (1970, 51) observed that most bivalves that are nearly circular in the plane of valve closure (those with length/height ratios near 1.0, such as the cockle family) rebury by using a reciprocal rocking motion which saws the shell edge into the sediment. The pronounced radial ribs and surface ornamentation of some cockles give them a rasp-like serrated valve edge which cuts through the sediment, allowing these species to have high burrowing rates despite having quite spherical shells for sediment-living bivalves (Stanley 1970, 62–3). A fatter shell would be more difficult to rebury using a reciprocal sawing action than a shell with a larger ratio of height/breadth. An elongated shell would be more difficult to rebury using a reciprocal sawing action than a shell with a smaller ratio of length/height. Therefore, a population of cockles living in a coarser sediment of more fully marine conditions should have a higher average height/breadth ratio and a lower length/height ratio than a population in a muddier more easily burrowed sediment of more sheltered conditions such as lagoons and estuaries. This is not the current argument among developmental biologists (e.g. Mariani et al. 2002, 489) but is known to palaeontologists, among whom Stanley (1970, 49) points out that members of the cockle family often have a shell height greater than length; it is possible that its living consistently in undisturbed muddy sediments means that C. glaucum has begun to go down the path taken by its Laevicardium cousins.

The height/breadth ratio was similar for all common cockles, typically around 1.28. The average length/height ratio was similar for all the three-year-old common cockles from the East midden (1.14–1.15), and distinct from that of South midden two-year-old common cockles (1.19). The 3–4-year-old common cockles from the South midden sample were distinct from the two-year-olds from the same sample, and were more comparable to that of the East midden.

Overall the ratios were not very different for the various different groups of cockles. For a common cockle of typical length for the assemblages (27 mm) its height in Sample 442 (based on its length/height ratio of 1.19) would be 22.7 mm while in Sample 456 (based on that sample’s length/height ratio of 1.15) its height would be 23.5 mm, a difference of 0.8 mm. Its breadth in Sample 456 (using the typical height calculated above, and height/breadth ratio of 1.27) would be 17.9 mm, while in Sample 455 (using the same height, and the height/breadth ratio of 1.25) it would be 18.2 mm broad, a difference of 0.3 mm. It is fair to suspect that these differences were simply due to chance variation. To determine if the difference in length/height ratio between two-year-old common cockles in South midden Sample 442 (1.19 ± 0.023, based on six shells) and three-year-old common cockles in East midden Sample 456 (1.15 ± 0.042, based on 51 shells), a t-test was employed. The preliminary F-test recommended Sokal and Rolph (1995, 189) demonstrated that the variances were likely to be the same (Fs = 3.3; F[α/2,50,5] = 3.3 for a probability α of 18%). The standard t-test assuming the variances to be similar (Sokal and Rolph 1995, 223) returned a t-statistic ts of 2.28, making the difference in length/height ratio significantly different at the 3% confidence level (two-tailed probability P[ts,n1+n2 - 2] of finding this t- statistic was P[2.28,55] = 0.026). The difference was unlikely to be due to chance alone.

The ratios for the two-year-old common cockles (C. edule) from the South midden Sample 442 seemed typical of more sheltered conditions. Kristensen (1957, 411–12) noted that in sheltered Dutch Wadden Sea mud-flats the length/height ratio was a consistent 1.19 ± 0.04 for C. edule over 20 mm. While he did not examine the height/breadth ratio Kristensen reported a consistent length/breadth ratio of 1.56 ± 0.05, so it is possible to calculate the height/breadth ratio by dividing the latter value by the length/height average, giving a height/breadth ratio of about 1.3 ± 0.1. The 3–4-year-old common cockles from the same sample were marginally more spherical, so these were probably harvested from slightly coarser (and therefore less sheltered) sediment. The samples from the East midden were less elongated, so these were from a more coarse (and therefore more fully marine) sediment.

The lagoon cockles (C. glaucum) were somewhat more spherical than the two-year-old common cockles in the South midden (Sample 442). However the lagoon cockles were slightly taller in relation to their width than the common cockles in the East midden. This is unexpected as C. glaucum is generally more spherical than C. edule (e.g. Mariani et al. 2002, 488). The ratios for the lagoon cockles from the East midden were not as reliable as those for the common cockles, since they were based on measurements of small numbers of valves (five valves for Sample 465, eight for 456.

To determine if the difference in height/breadth ratio between three-year-old common cockles in East midden Sample 456 (1.27 ± 0.064, based on 51 shells) and three-year-old lagoon cockles in the same sample (1.31 ± 0.058, based on eight shells) was due to chance, a t-test was employed. The F-test of Sokal and Rolph (1995, 189) demonstrated that the variances were likely to be the same (Fs = 1.22; F[α/2,50,5] = 1.22 for a probability α of 85%). The standard t-test assuming the variances to be similar (Sokal and Rolph 1995, 223) returned a t-statistic ts of 0.03, making the height/breadth ratios very similar (two-tailed probability P[ts,n1+n2 - 2] of finding this t-statistic was P[0.03,57] = 0.98). The difference in height/breadth ratio between the common and lagoon cockles was very likely due to chance alone.

Season collected

The proportion of the cockles which had less than 2 mm of growth from their third ring (assigned the ring count of 3) to all those with a third ring (those with a 3 and a 3.5 ring- count) was calculated for each sample with a usable number of shells. Within the East midden samples those just beginning to grow were always in the majority, forming 67–70% (50 of 75 for Sample 456, 36 of 52 for bulk sample 304 and 20 of 28 for Sample 455) or more (all five in Sample 454, and 39 of 40 in base midden increment 457). In South midden increment 442, those just beginning to grow were in the minority (one of six) of the 3 and 3.5-ring cohorts. To determine if this difference was an illusion due to the small number of shells involved, the usual 2x2 contingency table compared the observed values for Sample 442 to those expected from the East midden results (70% of the 3 and 3.5 ring cohorts, or four of the six). This returned a X2 value of 6.75, associated with a probability (at two degrees of freedom) of only 3.4%. It was therefore very unlikely that the value from Sample 442 was due to chance alone.

The great bulk of the three-ring shells in the East midden had just begun to grow in the following year, while a few had been growing for some time. This would be consistent with the shells being collected quite early in the year, some time in the spring (Hancock and Franklin 1972, 570). The bulk of the shells in East midden increment 442 being well into their growth season would be consistent with their being harvested later in the year.

Conclusions

The cockles from the incremental samples follow the same trend in both species make-up and shell form. Cockles in the South midden incremental sample 442 were from the most sheltered and estuarine shore harvested, because the proportion of lagoon cockles to common cockles was highest. This sample included two groups harvested from two distinct parts of the shore, one of young (predominantly two years old) large and fast-growing cockles from highly estuarine sediments (since the two-year-old common cockles had the same size distribution as the lagoon cockles), the other small, old (3–4 years old) and very slow- growing from a somewhat less sheltered shore (being more spherical than the two-year-olds from the same sample).

The cockles from the bulk of the East midden (the bulk sample and the upper incremental samples) were harvested from a more fully marine shore (since the proportion of lagoon cockles was lower and the shells more circular than in the South midden), where growth was moderately slow (intermediate between the two groups in the South midden) and the cockle beds seldom disturbed. The cockles from the base of the East midden were collected from nearly the same part of the shore as the bulk of the deposit (since both groups had the same growth rate and asymptotic size), but where this shore was slightly more marine (since this sample had the lowest proportion of lagoon cockles). A small proportion of the cockles in all the cockle-rich incremental samples in both middens were discarded intact, probably because they were dead and gaping.

The sea urchins

Introduction

Urchin remains were extracted as part of the processing of the bulk samples and 3.0 kg sub- samples; the methods used are described in the ‘Methods of extraction’ section above. Deposit 700, the base fill of Gallic Empire pit F1081, was recognized on site as being rich in urchin remains. Urchin remains were also recovered from the other sampled deposits in the same Gallic Empire pit (limpet-rich tip 685, and the north and south parts of final fill 681), the other Gallic Empire deposit (pit fill 523), the two early Roman limpet-rich tips 643 and 646, the tip in Iron Age layer 644, the Dark Age shell tip 708, medieval limpet-rich tips 586 and 560, and the bulk sample of the medieval East midden, 569.

Number of animals represented

Anatomical terms

The anatomy of the sea urchin is surprisingly complex for such a superficially simple-looking animal. Full descriptions and detailed illustrations of the arrangement of the parts are available in many zoological textbooks, and in The Echinoid Directory, a very informative website supported by the Natural History Museum, London (Smith 2003).

Some knowledge of the anatomical terms used for the various elements of the urchin is required to understand the analysis undertaken here. The regular sea urchin, generally a flattened sphere in shape, does not have a shell (layers of calcium carbonate crystals glued together by proteins, surrounding the animal) but a test, made up of plates (single crystals filled with microscopic pores containing much of the animal’s living tissue), the plates held together at joints or sutures (much like the epiphyses are held to the shafts in immature vertebrate long bones). The anal space or periproct (in the centre of the top or apical area of the animal) contains five small heart-shaped plates with holes opposite the point (the ocular plates) and five irregularly pentagonal plates each with a hole near one apex (the genital plates), one of which is enlarged and visibly porous (the madreporite).

The bulk of the test, lying below the periproct at the top and peristome (the hole in the centre of the base containing the jaws), is formed of alternating vertical zones, the ambulacra and the interambulacra. Ambulacral plates bear tubercles (bosses which form the joints for the spine-bases) and pore-pairs for the tube-feet. The interambulacral plates carry only tubercles. In most living sea urchins (including those discussed here), the edge of the peristome is a rigid ring (the perignathic girdle) made by the bottom interambulacral plates having an internal lip (the interambulacral ridge). Adjacent ambulacral plates in the girdle also have internal projections often jointed together to form a loop with a central hole (the auricles). The space between the jaw and the test is bridged by a tough membrane which in most sea urchins contains five pairs of buccal plates, small and D-shaped with a small hole.

The urchin jaw or Aristotle’s lantern, so called because it was said by that philosopher to look like a five-sided bone lantern (Historia Animalium IV.5), is a complex structure containing:

• five gently curving teeth. Most urchins have a tooth with a T-shaped cross-section, making them look under magnification like a piece of railway track bent up from the rail-bed; • ten hemipyramids, five clockwise of a tooth as viewed from above, within the animal, the other five counterclockwise. At a glance a hemipyramid resembles an odd jawbone from a small rodent, or a tiny shoulder blade. A thin flat plate in the shape of a right triangle with a short leg about half the length of the longer; one face of the plate (furthest from the tooth, held vertically in the radial plane in the living animal) is covered with parallel fine weakly sinuous grooves. On the face nearest the tooth (opposite the grooved face), set slightly back from the hypotenuse and parallel with it, is a thin triangular ridge which widens towards the short leg and projects beyond it (the supra-alveolar process); the ridge thickens and curves towards the point where the hypotenuse meets the long leg, projecting beyond the plane of the outer face as a blunt hoof-like tip. This thicker curved part of the ridge bears a neat curved groove (the dental slide) which locks around the tooth in life; • ten epiphyses, small elements resembling a tiny axe or a Roman letter V with the thicker arm shortened; • five rotuli, the cartoon ‘dog-bone’ of Green (1999, 145), similar to a miniature pig’s metatarsal; • five compasses, T-shaped elements with the cross-member slightly below the shaft, rather like a ‘Colchester’ Iron Age brooch (Hattatt 1985, 30).

Sorting of the recovered remains

For the >10 mm, 10–4 mm, 4–2 mm of each bulk sample, the urchin remains were extracted and sorted into the various elements identified above. For the larger two fractions, the fragments of the bulk of the test were separated into ambulacral, interambulacral and perignathic fragments. Ambulacral fragments were separated according to the number of pore-pairs on a plate. Perignathic fragments were separated into pieces bearing interambulacral ridges and ambulacral pieces, often bearing auricles. Spines and fragments of the bulk of the test were extracted from a sample’s 4–2 mm sample fraction if none of these had been recovered from the coarser fractions. All other elements (including perignathic girdle fragments) were collected from the 4–2 mm fraction from all samples except those four taken early in the project (medieval deposits 219, 217, 204 and 451) where these fractions were discarded after scanning. No bulk sample was collected from earlier medieval limpet spread 634, and the fine fractions were discarded from its increments.

Elements were found to be broken in substantial numbers, so decisions had to be made regarding which fragment of an element should be counted as representing a whole. Madreporites were usually complete, occasionally found with a corner broken. Genital, ocular and buccal plates were only found whole, since once broken they were too small to be recognizable or to have been retained on a 2 mm mesh. Auricles were counted as either whole, or as a fragment originally clockwise or counterclockwise from the central hole (as viewed from above, within the animal). Teeth were always found broken; any tooth fragment was counted as one (thus overestimating the number of teeth recovered). Clockwise and counterclockwise hemipyramids were counted separately. The supra-alveolar process, the hoof-like tip, and in some cases the outer plate could all be lost. A hemipyramid was counted if the portion where the dental slide met the base of the supra-alveolar process was present. The V-shaped portion of each epiphysis was counted, whether one or both arms were still attached. Most rotuli were found whole; all rotuli or fragments with the broad V-notch (the end nearest the centre of the animal in life) were counted. Compasses were usually broken, with the narrow long shaft lost; a compass fragment was counted if the junction of the cross- piece with the shaft survived.

MAU, MNI and urchin richness

For each bulk sample, the counts of those elements with a consistent number in an urchin were added together for all fractions, and divided by the number present in an urchin, to produce a minimum number of animal units (MAU). Whole auricle count was added to the clockwise fragment count to produce a clockwise auricle MAU, and to the counterclockwise fragment count to produce a counterclockwise auricle MAU. The minimum number of individual urchins (MNI) in a sample is the largest MAU for that sample, rounded up to the nearest integer (Lyman 1994, 104–6).

The total element counts, element MAU and percentage of the element MAU of a sample’s MNI are presented in Table 6 for the samples from which items other than spines or simple test plates were recovered. The MNI (the largest MAU) for each sample is highlighted in bold. The minimum urchin density (as minimum number of urchins per litre of excavated soil) is also presented. As anticipated, many urchins were found in Gallic Empire fill 700 (at least 85 urchins), where they were very rich (7.7 urchins per litre). More typically, a few urchins were discarded along with other more common shellfish, resulting in 0.5–0.25 urchins per litre. Urchin MNI and urchin density is somewhat misleading, over-estimating the importance of deposits in which remains are very sparse. The handful of general test and spine fragments recovered from the six samples not on Table 6 still generated a MNI of one for each of those samples; as most samples are about 10 or 20 litres, any sample with urchin remains would have an urchin density similar to that of Sample 288 of deposit 523.

General test fragments and spines were the only urchin remains recovered from Iron Age tip 644, the early Roman tips 643 and 646, and medieval tips 586 and 560 and East midden 569. The test fragments were loose individual test plates except for that from deposit 569 (an articulated ambulacrum fragment) and 586 (an articulated interambulacrum fragment and loose ambulacram plates).

The size bias in element recovery: a warning

The 2–1 mm fraction of bulk sample 682 from the urchin-rich base fill 700 of Gallic Empire pit F1081 was also sorted for urchin elements other than spines, ambulacra and interambulacra, a process taking some days. To determine if this effort was truly needed, the counts of elements recovered from this fine fraction were compared to those for the fractions over 2 mm. While sorting only down to 2 mm missed all the ocular and buccal plates, nearly all the compass fragments (94%), over three-quarters of the tooth fragments (78%), and a fifth of the genital plates (22%), it missed only one in twenty epiphyses (5.1%), fewer rotuli (4.0%), and less than one percent of the madreporites, hemipyramids and auricles. The author judged at the time that MNI values would be little affected by these biases (because he assumed MNI would be based on hemipyramids, auricles and rotuli), and proceeded to sort the other bulk samples only to 2 mm, discarding the finest 2–1 mm fraction. In retrospect, this was a mistake: genital plates can be critical for determining MNI (e.g. Sample 682 of deposit 681, see Table 6), and data critical for reconstructing butchery, animal size, growth rate, and demographics (see below) would be lost if naturally small species or populations of small animals in a larger species were being exploited. Sorting down to 1 mm is required, however painful.

Methods of identification

Kinds of sea urchin in the region, and their identification

There are seven species of Atlantic European urchin which could be expected at the low tide line or within easy diving depth below it (Mortensen 1927; Hayward et al. 1996, 292–4). Strongylocentrotus droebachiensis (the northern sea urchin, called ‘green sea urchin’ in North America), an Arctic animal, ranges south into the North Sea but is rare south of Shetland. Two forms of more temperate water, Echinus esculentus (the common sea urchin) and Psammechinus miliaris (the ‘green sea urchin’ of Europe) are found from north Norway to Portugal. Two principally Mediterranean species, Paracentrotus lividus (the violet sea urchin) and Sphaerechinus graularis (the purple sea urchin), have the north Brittany coast as their present northern limit. Two Mediterranean forms now found in Atlantic Portugal, Psammechinus microtuberculatus and Arbacia lixula (the black sea urchin) may have been more common further north in warmer climates.

Generally the tests of the various species are told apart by their surface patterns, especially the number of pore-pairs on their ambulacral plates (Mortensen 1927; Smith 2003). Arbacia has three pore-pairs on small pentagonal ambulacral plates usually with only one tubercle. Echinus and Psammechinus have pore-pairs arranged in rows of three on a distinctly rectangular ambulacral plate, the latter commonly with tubercles interspersed with the pore- pairs. Sphaerechinus has an arc of four pore-pairs, occasionally six. Paracentrotus has 5–7 pore-pairs, in arcs curving round one or more tubercles, while Strongylocentrotus also has 5– 7, in rows separated by rows of small tubercles.

Certain problems complicate identification of fragmentary urchin tests. The different genera can have overlapping counts of pore-pairs. Ambulacral plates near the peristome often are reduced in pore-pair number, often down to three (Mortensen 1927, 314), while those near the periproct often have larger number, up to seven or eight for Paracentrotus and Strongylocentrotus. Younger Echinus tests (and therefore the bases of older tests) occasionally have tubercles within the rows of pore-pairs, making them difficult to distinguish from Psammechinus ambulacra (Mortensen 1927, 314; Smith, 2003). Mortensen (1943, 167) noted that Paracentrotus lividus and Strongylocentrotus droebachiensis tests are difficult to distinguish, but can be separated by the shape of their auricles. This comment led the author to compare the seven species potentially present on Breton shores in the collections held by the Oxford University Museum of Natural History and the Natural History Museum, London, and the author’s own small collection. This survey revealed that the auricles are quite distinct for each genus (Fig. 12).

Identification of the auricles

Auricles and their fragments from each sample were compared with the results of the auricle survey outlined above. Overall, 10.7% of the auricle or auricle fragments (as counted in the manner used to generate MAU for Table 6) could not be identified because the upper parts and edges of the auricles had broken away. Two (from Dark Age tip 708) were probably Psammechinus. Most auricles and fragments (83.5%) were clear examples of Paracentrotus. Some auricles and fragments (5.5% of the total found) which could not be distinguished immediately between Paracentrotus and Strongylocentrotus were found on comparison with modern examples of Paracentrotus to fall within the range of shapes for that species. The sample from deposit 700 contained the great majority of the auricle fragments (751 of the 867). In this sample 6.4% were unidentifiable, while the remaining 93.6% were Paracentrotus. Breakage was a serious problem only in Dark Age tip 708, in which 32 of 45 auricle fragments were too damaged to identify; two of the identifiables were probably Psammechinus, with the remainder Paracentrotus. The portion too damaged to identify was half from Gallic Empire pit fill 685, and none in the other samples with auricles; in these samples, all the identifiable auricle pieces were from Paracentrotus.

Ambulacral plate pore-pair counts

For each sample the ambulacral fragments were separated into ambital (at point of greatest girth), apical (near the top) and aboral (near the mouth) types, or loose (unarticulated) plates if no fragments were found articulated. For each fragment the number of pore-pairs in a row was counted, and the overall pattern compared with photographs of identified specimens (Smith 2003). Those with three or four pore-pairs per plate were compared to modern reference material in the author’s collection; a potential form of Echinus was submitted to Dr Andrew Smith at the Natural History Museum, London.

The overwhelming majority of fragments bearing articulated ambulacral plates had five pore- pairs per plate, with patterns of pore-pairs and tubercles like that found in Paracentrotus and Stronglyocentrotus. Ambital plates had a consistently higher pore-pair count (up to eight) while those near the mouth had consistently lower numbers (4–5). All fragments with four pore-pairs per plate had Paracentrotus-Strongylocentrotus patterns and were usually articulated with plates with five pore-pairs; therefore ambulacra of Sphaerechinus were not identified. This was consistent with the results of examining the perignathic elements, none of which bore the characteristic deep buccal notches of Sphaerechinus. An ambulacral fragment with only three pore-pairs per plate from the urchin-rich Gallic Empire pit fill 700 was identified by Dr Smith as an aboral piece of Echinus esculentus on the basis of its tubercle pattern. The fragment of articulated ambulacral plates from medieval East midden 569 had three pore-pairs per plate; comparing it with modern reference material identified it as Psammechinus. The loose ambulacral plates from medieval tip 586 (retrieved with an articulated interambulacrum fragment) and from Gallic Empire fill 523 had patterns of pore- pairs and tubercles consistent with their being Paracentrotus or Strongylocentrotus; these fragments were assigned to the former. The very few loose ambulacral plates from Iron Age deposit 644, the early Roman tips 643 and 646, and medieval tip 560 were too broken to have their pore-pairs counted accurately.

Species recovered from Le Yaudet, and season of harvest

Almost all the Le Yaudet remains were of violet sea urchins (Paracentrotus). Ambulacra of this form or Strongylocentrotus were by far the most common, but the auricles were almost exclusively Paracentrotus, with no Strongylocentrotus auricle pieces found. The close examination of both ambulacra and auricles was required to identify all the forms present in a sample. Within urchin-rich Gallic Empire pit fill 700 the counterclockwise hemipyramids gave a MNI of 85. The auricles indicated 93.6% of the assemblage was Paracentrotus, giving a violet urchin MNI of 79. The remainder of the MNI (six individuals) was unidentifiable on the auricle evidence but a single sample of Echinus was found in the ambulacra. Therefore the MNI of common urchins must be 1.0, with the remaining five unidentifiable. These are the values reported for this sample in Table 1. Similar calculations were used to generate the values for Table 1 for this and the other urchin-bearing samples. Other Gallic Empire deposits contained only remains of Paracentrotus or unidentifiable auricles.

Dark Age tip 708 while badly broken was principally Paracentrotus, with some green urchins (Psammechinus). Sparse remains from the medieval limpet tip 586 included Paracentrotus, and from the East midden 569 included Psammechinus. The urchin remains from the Iron Age (layer 644), the early Roman period (tips 643 and 646) and medieval tip 560 could not be identified to species.

Only the roe (egg mass) within the urchin test is eaten, and the build-up of roe is seasonal in most Atlantic urchins. Allain (1975, 196) found that Paracentrotus in the Baie de Lannion were ripe from September to April; the official fishing season ran mid-October to mid-April. Psammechinus miliaris spawn in summer beginning May to July (Kelly and Cook 2001, 220) and Echinus esculentus also spawn in early summer (Mortensen 1927, 298) so these are likely to be filled or filling with roe beginning in autumn and increasing until spring. These deposits with urchin remains very likely were the product of the colder part of the year, probably in spring.

Quantification by weight

Weights of the various elements sorted from the >10 mm and 10–6 mm fractions in the 3.0 kg sub-samples were recorded to the nearest 0.1 g, as a basis for interpreting fragmentation and relative urchin content of the deposits. Weights for urchin remains from the finest fractions sorted (6–2 and 2–1 mm fractions) were recorded only for the 3.0 kg sub-sample of urchin-rich Gallic Empire pit fill 700. For those samples where elements of the jaw, peristome or periproct have been recovered (allowing an estimation of MNI), the weight percent made up of the relevant 3.0 kg sub-sample by urchin test fragments over 6 mm is presented in Table 6 with the urchin MNI. The weights of urchin remains from the bulk samples were not recorded, since a method for determining MNI had been found.

Finding large pieces was a rare event, even in deposits rich in urchins. Of the dozen samples with urchin remains only five had remains over 6 mm in size, and only three (Gallic Empire pit fills 700, 685 and 681 north part) had any over 10 mm. In urchin-rich base fill 700 less than a third of the non-spine elements over 6 mm were over 10 mm (34.2 g of 115.6 g), and that over 6 mm was less than half the total weight of non-spine elements (246 g over 2 mm). This still grossly underestimated the spines: over half by weight came from the very fine 2–1 mm fraction. On this evidence urchins fall apart readily under terrestrial conditions, the tests separating mainly into loose plates or occasionally into small groups of plates. Urchin fragments bigger than 10 mm from a sample simply indicates exceptionally well-preserved urchins, but a sample lacking them does not show urchins are absent from the sample. Only sorting the 4–2 mm fraction shows whether urchin remains are absent, or are too sparse to be interpreted.

Differences in minimum animal units, and butchery methods

While the periproctal minimum animal units (MAU) and hemipyramid MAU were similar in most of the samples, these were quite different for Sample 682 of final fill 681 (Table 6). A 2x2 contingency test comparing the number of genital plates to clockwise hemipyramids gave Χ2 value of 6.77, a probability at one degree of freedom of 0.93%, or once in every 107 attempts. It is therefore quite likely that more urchin tops than bases were discarded in the north part of fill 681.

The periproctal MAU and hemipyramid MAU based on the >2 mm material were also quite different for the urchin-rich Sample 681 from pit fill 700. A series of Χ2 tests on 2x2 contingency tables were employed to discover if this was simply due to chance. Comparing the base of the test to the jaw elements, the counterclockwise hemipyramids should be equivalent to the 400 clockwise auricles, not the 424 found. The contingency table produced a Χ2 value of 0.699, equivalent at one degree of freedom to a probability of 40%. Therefore the difference between the base of the urchin tests and the jaws is very likely due to chance alone. For the periproctal elements, a test of departure of the madreporites and genital plates from the expected ratio of 1:4 produced a Χ2 value of 0.352, equivalent at one degree of freedom to a probability of 55%. Therefore, the difference between periproctal elements was probably due to chance alone.

Comparing the urchin bases with the periproctal elements over 2 mm, for the counterclockwise hemipyramids the number of madreporites expected is 85, not the 35 recovered. A test for the departure of the observed number of counterclockwise hemipyramids and madreporites from the expected ratio of 5:1 produced a Χ2 value of 27.0, equivalent at one degree of freedom to a probability of 0.000002%. Recognizing that some of the periproctal elements are under-represented in the material greater than 2 mm, the material greater than 1 mm was also tested. The genital plates gave the periproctal maximum MAU for periproctal elements (40), and the maximum MAU for the urchin bases was again given by the counterclockwise hemipyramids (85.4). The 2x2 contingency table testing the departure of the observed number of genital plates (160) and counterclockwise hemipyramids (427) from the expected ratio of 4:5 produced a Χ2 value of 70.2, equivalent at one degree of freedom to a probability of 5.3 x 10-15 %. It is therefore vastly unlikely that the top parts of the sea urchins were discarded along with the bases and jaws into this pit to form fill 700.

The type of break on the fragments also showed that breakage of the test was more common around the mouth. Urchin tests tend to break across test plates when live or freshly dead, but tend to separate along plate margins once decay has set in (Smith 1984, 17–19). Of the 56 peristomal fragments over 10 mm with auricles, 27 had one or more edges made by an old break (one in which soil particles had penetrated the pores of the test plates, giving a dirty look to the broken face) which crossed plates, while the others had breaks along joints between test plates, or more rarely breaks across plates that were recent (with a clean look to the broken face). Of the 33 fragments over 10 mm with periproctal margins, ten had old breaks across plates. A X2 test on the 2x2 contingency table was used to determine if the departure of the observed numbers of auricle-bearing and aboral fragments, from that expected from the proportion of the fragments with old breaks in the total of all the fragments, was due to chance alone. This test produced a X2 value of 3.17, equivalent at one degree of freedom to a probability of 7.5%. The peristomal fragments were more likely to be broken than the periproctal fragments. The result was significant because the peristome is made of larger and thicker plates than the periproct, and therefore peristomal fragments are inherently less likely to break than periproctal fragments.

The differences between periproctal and basal elements cannot be due to periproctal elements being destroyed more easily than basal elements after discard. Periproctal and basal elements were preserved in roughly equivalent amounts in some deposits in this pit, and periproctal elements were more common in the north part of 681. However, both shell breakage (as measured by mussel crush ratio, Fig. 4) and stone content (Fig. 2c) were moderate and similar in all the fills of this pit, so they had all experienced approximately the same amount of erosion and disturbance following deposition. Therefore physical erosive forces cannot have had a more destructive effect on the periproctal plates. Carbonate content (as measured by shell weight percent, Fig. 2c) was similar in the deposit with under-represented periproctal plates (fill 700) to that in which they were over-represented (fill 681 north part), and lower than that with a balanced representation of the elements (fill 685). Therefore chemical dissolution cannot have had a more destructive effect on periproctal elements.

Early in the use of the Gallic Empire pit as a place of discard, the bases and jaws of the sea urchins were being discarded as waste, while the urchin tops were being retained. Towards the end of its use as a discard area, more tops than bases were being discarded. This difference can be explained by the use of the urchin tops as natural bowls for the egg masses (the edible part of the urchin), a practice familiar in modern cuisine. During the preparation of the urchins that formed fill 700, the jaw and peristome were carefully removed and discarded from some of the urchins, and the space topped up with egg masses from other urchins opened with less care, probably by cleaving (as shown by the large test fragment complete from peristome to periproctal articulation). The excess jaws and peristomes and the surplus urchin tests were discarded as ‘processing-’ or ‘kitchen-waste’ to form fill 700. Later on, in some cases the tests used as bowls would be discarded along with the tests and jaws of the urchins just processed for food, as part of a mixture of ‘consumption-’ or ‘table-waste’ and ‘kitchen-waste’, to form fill 681.

Urchin populations exploited, and sizes

Introduction

The range and distribution of urchin sizes could not be measured directly, as no tests survived complete. The sizes could not be reconstructed from the surviving fragments by measuring the distance between adjacent ambulacral or interambulacral sutures, since this distance was retained on only a handful of fragments. Fortunately most urchins grow their jaws (in particular, the length of the hemipyramid) roughly in proportion to their test diameter, the proportion being roughly linear for a number of species (examples in Ebert 2001, 90–4) including P. lividus (Gruet in press).

The range and distribution of jaw (hemipyramid) sizes would therefore stand as proxy for the urchin diameters for a group heavily dominated by a single species. Therefore, the clockwise hemipyramids sorted from the bulk sample of urchin-rich pit fill 700 were examined for completeness, and the 205 which were complete had their maximum dimension (from top of the super-alveolar process to the oral tip of the dental slide) measured to the nearest 0.02 mm with vernier calipers. The statistics for these lengths are shown in Table 7, and their distribution histogram (grouped into 0.1 mm intervals) is shown as Figure 13.

While the overall statistics showed little deviation from that of a normal curve (with low values of skewness and kurtosis), the histogram appeared to be made up of several groups. Inspired by the method of Cerrato (1980, 427–39) in discriminating age classes in marine shelled organisms in the absence of clear annual growth rings, the length data were converted into normal equivalent deviates by the method of Sokal and Rolph (1995, 118), and graphed as a normal quantile plot. The same method was applied to the 0.1 mm frequency distribution and graphed as a normal quantile plot (the line on Figure 13). This method re- scales distributions so that points in a normal distribution lie along a straight line, while groups of normal distributions appear as a series of flattened S-shaped ‘steps’ separated by changes in slope or inflections, each inflection point marking the separation between groups (Sokal and Rolph 1995, 117; Cerrato 1980, 431).

Identification of groups, and their statistics

The two normal quantile plots were very similar; the line was slightly smoother in the plot derived from the 0.1 mm distribution, especially towards the extreme values of the quantiles. Inflection points were used to separate the distribution into nine groups. These inflection points corresponded with distinct gaps in the length data of at least 0.1 mm; since these gaps were larger than the 0.2 mm measurement error, these groups were probably genuine.

The defining length ranges, percentage contribution to the 205 measured hemipyramids, and associated length statistics for the groups are presented in Table 7. The percentage contribution and mean length for each group are plotted against increasing group size as Figure 14. According to Cerrato (1980, 428), these groups should correspond to succeeding year-classes in a population. The typical form for such a diagram as Figure 14 for most marine molluscs (Cerrato 1980, 421 and 428, e.g. Craig and Hallam 1963, 747) and for many urchins (Ebert 2001, 95–6) would have a single peak in the percentage contributions with a gradual drop-off in percentages with increasing size, corresponding to a population in which one or two age-classes form the bulk of the population and animals become increasingly rare with age. The typical mean length curve for most marine molluscs (examples in Hallam 1967, 32) and many urchins (Ebert 2001, 89) has a single flattened S-shape indicating growth falls off gradually with increasing age, like that for the cockles from the East midden at Le Yaudet (Fig. 10). Figure 14 is therefore unusual; growth rate and percentage both had two peaks and these peaks coincided, in the 11.3–11.9 mm group and the 13.26–14.4 mm group.

Populations exploited

The coincident peaks in growth rate and percentage were most simply explained by the jaw assemblage coming from two populations of urchins, one slightly smaller and somewhat more slow-growing producing the bulk of the first four jaw groups, and the other slightly larger and somewhat faster-growing producing the bulk of the last five groups.

To estimate the composition of the two populations the first four jaw groups were assigned to the ‘small’ population, and the larger groups assigned to the ‘large’ population. This assignment therefore assumed that the contribution of the ‘large’ population to the smaller groups, and that of the ‘small’ population to the larger groups was negligible. This assumption was shown later to be untrue for the third group (see ‘age of groups’ below), and required a refinement in estimating the population parameters, growth rate and maximum jaw size (see immediately below).

Growth rate and maximum jaw size for the two populations

A plot of the Ford-Walford points (mean length of adjacent larger group, HLt+1, for each group mean length HLt) was used to reconstruct the von Bertalanffy growth parameters of the two populations, by using mean-squares regression to fit a linear relationship with the formula

HLt+1 = (m)HLt + (HLto) with slope (m) and y-intercept (HLto) to the points for each population. Von Bertalanffy’s formula is already presented in the analysis of the cockles from this site. If a population’s growth is estimated well by the von Bertalanffy relationship, points plotted in this manner will lie along a straight line. The slope of that line (m) approximates the von Bertalanffy growth coefficient K by the relationship

m = e-K , or K = (-1) ln (m)

The asymptotic value (the theoretical maximum which the animal gradually approaches over time) of the measured parameter (in this case, maximum jaw length HL∞) is estimated by using the formula for the linear relationship generated by the least-squares regression to find the value of mean length HLt for which HLt and HLt+1 are equal.

The regression on the large population (using the Ford-Walford points of the sixth, seventh and eighth groups) gave a surprisingly precise line for biological data (a regression 2 coefficient R = 0.997) for the formula HLt+1 = 0.689 HLt + 5.48, giving an asymptotic length (HL∞) of 17.64 mm and a growth coefficient K of 0.372.

The initial regression for the small population (smallest four groups, where the third group had an interval of 11.32–11.92) gave a precise line (regression coefficient R2 = 0.990) for the formula HLt+1 = 0.8372 HLt + 2.571, giving an asymptotic length (HL∞) of 15.79 mm and a growth coefficient K of 0.178. Back-calculations showed that the three-year-old cohort of the large population would average about 12.07 mm (see ‘Ages of groups’, below). This coincided with a small peak in the population distribution at 11.92–12.12 mm. By assuming that this small population peak was the large population’s three-year-old cohort, and removing them from the calculation of mean length for the third group, the precision of the fit in the linear regression of the resulting formula and the new Ford-Walford points for the small population was improved: the regression coefficient R2 increased to 0.999. The improved Ford-Walford formula, HLt+1 = 0.830 HLt + 2.65, gave an asymptotic length (HL∞) of 15.60 mm and a growth coefficient K of 0.186.

Ages of groups

The age of a group can be estimated by reworking the von Bertalanffy formula and solving for years from emergence (t-to):

(t-to) = ln(1 – HLt / HL∞ ) / (-K)

Substituting the values of HL∞ and K derived from the Ford-Walford points for the large population into the above equation and solving for the ages of the groups in the large population showed that the ages for this population ranged from 4.1 to 7.1 years. It is also possible to substitute the small population parameters into the von Bertalanffy equation and solve it for the jaw length of any notional age of large-population urchin. Solving the equation for a three-year-old large population urchin showed it would have had a jaw length of 12.07 mm. This coincided with a small peak in the jaw population distribution at the 11.92–12.12 mm range. The assumption initially made that the large population made little contribution to the small population groups was not true, so the small population parameters (growth rate, maximum jaw size and age at emergence) were recalculated with this peak removed (see ‘Growth rate and maximum jaw size’ above). ‘Large’ population urchins younger than three years seemed to be very unlikely to have been harvested: solving the equation for a two-year-old large urchin showed it would have had a jaw length of 9.56 mm, smaller than the smallest recovered.

Substituting the parameters for the ‘small’ population into the von Bertalanffy equation and solving for the ages of the groups in this population showed that the ages of this population ranged from 5.4–9.4 years. There was very little contribution of older urchins of the ‘small’ population to the jaws. Solving the von Bertalanffy equation for 10.4-year-old urchin using the small population parameters gave a length of 13.38 mm, for which there was no clear peak in the length data.

The time since emergence into the conditions in which the urchins grew as mature adults (t- to) derived by the above formula for the various groups within the two populations are presented in Table 7.

Make-up of the two populations

The percentage contribution of each group to the total 205 intact hemipyramids, of the two populations identified, and the whole of the sample is presented in Table 7, and the percentage and mean length data for the groups from the two populations (‘small’ population on the left, ‘large’ population on the right) are plotted in Figure 14. Size played a role in urchin selection, since the ‘small’ population contributed markedly less than the ‘large’, and the very small urchins (the smallest two groups in the ‘small’ population) made up very little; the very smallest group was so small and so similar in length that it was probably from a single urchin. The minimum size of this group (9.84 mm) was a good approximation for the minimum acceptable size regardless of population, since jaws of two-year-olds from the large population (with an average size smaller than this) were not found in the sample. A jaw length of about 12.5 mm seemed to have been the ‘minimum optimal’. While smaller jaws were recovered, urchins with jaw lengths greater than this formed the bulk of those found regardless of whether they were in the ‘small’ or ‘large’ population. The percentage of the 12.52–13.16 mm group was large compared to the rest of the ‘small’ population, even though this group was oldest of all the groups in the sample and therefore would be expected to be the rarest in the wild. In the ‘large’ population the youngest cohort (3.1 years, 11.92–12.12 mm) made up a very small part of that population (probably only two or three urchins in the sample), even though this was the youngest group in the sample and therefore would be expected to be the most common in the wild.

The preference of the gatherers for bigger urchins whenever available, and for the bigger urchins even among groups of big urchins, shows they were concerned with maximizing meat yield from individual urchins. Had they been maximizing yield for minimum effort, they would have found a part of the shore rich in urchins bigger than the minimum acceptable size, and concentrated their efforts on that space, almost clearing it of the urchins above that minimum acceptable size; there would have been no difference between a minimum acceptable size and a minimum optimal size.

Approximate size of urchins

The average test sizes for the various groups and populations were identified by linear interpolation, using the linear relationship between jaw height and test diameter assumed by Gruet (in press) for modern examples of this species from western Brittany as a guide to size (11 mm jaw for a 43 mm test, 14 mm for a 52 mm test, and 16.5 mm for a 60 mm test). This is clearly an approximation, since the actual proportion varies between populations of the same species of animal (Ebert 2001, 88). Also, jaws are proportionately large for the test diameter in violet urchins with a poor food supply (Boudouresque and Verlaque 2001, 198), so the assumption of a linear relationship may overestimate the diameter for the ‘small’ population. These approximate sizes are presented in Table 7.

The overall size range represented was small. The minimum acceptable size for an urchin would appear to have been about 40 mm, with the urchins commonly available typically about 43–55 mm ranging up to about 60 mm. The difference in average size between the two populations exploited (roughly 45 ± 2 mm vs. roughly 53 ± 3 mm) was so small that it was probably not apparent to the gatherers. The ‘minimum optimal’ jaw length of 12.5 mm would have come from an urchin of approximately 48 mm; urchins larger than this seemed to have been heavily preferred regardless of the population.

These size ranges and distributions would be within those expected for urchins gathered from rocky shores exposed at low tides. Allain (1975, 181) recovered a modern population with a similar size distribution from north Breton rocks at spring ebb tide. Modern Breton practice rejects urchins less than 37.5 mm, in the Baie de Lannion populations of urchins of a typically larger size (60–85 mm) were recovered by diving or sub-littoral traps (Allain 1975, 202, 205).

Interpretation

The jaw length evidence showed that the urchins in deposit 700 were gathered from two different populations. The smaller part of the urchins were gathered from a group which was small, slow-growing and dominated by older urchins, a population that could be expected between spring and neap tide lines or in large rock-pools. The bulk was gathered from a faster-growing population of larger but not large urchins, a population that could be exposed around the low water of spring tide level. Estimates of size from jaw-diameter relationships showed that urchins about 40 mm at least were collected, but larger animals (over about 50 mm) were much preferred. Natural forces account for the dynamics of the populations for several years prior to harvest, so human harvesting was a rare event for these populations.

Summary

Sea urchins were gathered during every period of occupation at Le Yaudet. Urchin remains were sparse and unidentifiable in samples from the Iron Age (layer 644) and early Roman periods (tips 643 and 646). At Le Yaudet it is likely that these and other sparse urchin remains were contemporary with the other remains in the deposit. The urchin skeleton easily breaks up into single porous delicate crystals. Such loose urchin remains are more likely to be leached away than mollusc shells which tend to remain intact. It might be that the slightly more soluble aragonite form of carbonate in mollusc shells acts as a buffer for mollusc calcite carbonate in the soil immediately surrounding mollusc shells (Claasen 1998, 59); high- magnesium calcite of echinoderms is known to be more soluble than mollusc carbonates in terrestrial conditions (Claasen 1998, 60). The soils at Le Yaudet were poor at preserving carbonates; mollusc remains were rare unless concentrated. In such acidic soil conditions urchin remains would have had a poor chance of surviving episodes of disturbance and transportation, so there was little chance urchin remains were intrusive or residual.

Gallic Empire pit F1081 had a base fill 700 in which violet urchin (Paracentrotus lividus) remains were rich, about eight animals per litre of sample. A small number were harvested from a population of small slow-growing urchins (probably from inter-tidal shore or rock- pool) and a larger number from a population of larger, faster-growing urchins (probably at the top of the inter-tidal). Small numbers of common urchin (Echinus esculentus) were also gathered. The minimum size gathered was about 40 mm, but those over 50 mm were much preferred; meat yield per individual urchin seemed more important than yield per hour spent gathering. These (and urchin remains in other deposits) were likely gathered in the cold seasons, probably spring, since that is when urchins are ripe. Many were prepared by breaking open the base and extracting the mouthparts, with the rest of that urchin used as a natural bowl. Every other sampled deposit from this pit also contained urchin remains dominated by violet urchin, one other with evidence for use of the test as a natural bowl. The earliest fill of this pit F1081 contained predominantly food preparation waste, the final fill contained a mixture of preparation and consumption waste. Since each fill contained some sea urchins, it is possible that the pit filled during a single season. The statistically reliable differences between the urchin content of the different deposits within the pit make it very unlikely that the urchins in the later features were reworked from the first fill. The other sampled deposit from this period (fill 523) also had sparse remains including some of violet urchin.

For Dark Age shell tip 708 green urchins (Psammechinus) were collected along with the violet urchins, therefore probably from somewhat less wave-battered or cooler shores than one occupied by violet urchins alone (Kelly and Cook 2001, 217).

Sparse remains from earlier medieval tip 586 included some of violet urchins, and from the successive tip 560 were unidentifiable. Those from the later medieval East midden 569 included some of green urchin. The earlier medieval tips very rich in limpets (layers 557, 558) had no urchin remains. The number of medieval deposits with urchins may be underestimated because the finer fractions of many of their samples (specifically contexts 634, 219, 217, 204 and 451) were not sorted.

The dog-whelks from the Dark Age

The informed reader, having scanned Table 1 (results from the bulk samples) and observed the large number of shells of dog-whelk (Nucella lapillus) recovered from the Dark Age shell-rich deposit 708 (Sample 679), would here expect a consideration of the exposure experienced by the shore upon which they dwelt based on a statistical examination of shell aperture to shell length, since this is a technique employed by both marine biologists (e.g. Crothers 2003) and archaeologists (Andrews et al. 1985).

Such a consideration is impossible with this deposit, as all but one of the dog-whelks were broken. The only whole shell was from a relatively small individual of 19 mm overall length. The well-preserved elements of the shells came in four forms:

1. The aperture is entire, with the final 20–40o of the final whorl still attached and bearing semi-circular margin to the break roughly parallel to the growth bands. (Nine examples of this form were recovered from the sample.) 2. The siphonal canal is intact and includes the base of the columella and the lower half or two-thirds of the aperture. The break on the reverse of the final whorl is partly semi- circular (as above) and partly along a weak point formed by one of the outer grooves. (Fifteen examples.) 3. The outer edge of the aperture is complete from the top of the suture around to include at least part of the siphonal canal, but with no more than the stub of the columella. The break runs across growth lines and sculptural grooves. (Twenty-four examples.) 4. The apex is complete for the earliest five or six whorls grown, often including the suture at the top of the final whorl. (Nine examples.)

Finding virtually all the dog-whelk shells broken in deposit 708 was highly unusual, and raised the possibility that they were broken before being discarded. Unfortunately, shells in deposit 708 were not well preserved; results from the 3.0 kg sub-sample gave a limpet fragmentation ratio of 0.66 and a mussel fragmentation ratio of 4.1, some of the highest values from the site (Table 2). The data recovered from the identifiable shells offered some guidance in deciding whether the dog-whelks were broken prior to discard. Of the 1312 Patella vulgata over 10 mm recovered from the half of the Sample 679 sorted and identified, 288 were complete enough to be measured, giving a probability of being found intact of 288/1312 = 21.95%. It seems reasonable to accept that these limpets have broken following discard; those few that were discarded broken were very likely to have broken into fragments smaller than 10 mm. It was possible that the dog-whelks were broken in a similar manner to the limpets, and the assemblage recovered (one whole dog-whelk in 69 shells over 10 mm) was different from the limpet assemblage only by chance. Testing this by means of a 2x2 contingency table gave a X2 value of 16.93, equivalent to a probability of 0.000039. In other words, the odds of finding only one limpet intact in any 69 limpets randomly selected from the over 10 mm fraction of deposit 708 were 0.0039%, or about once in every 25,600 attempts. If dog-whelks had been broken up following discard in the same manner as limpets, finding the dog-whelks as they were found (one in 69 intact) was equally unlikely. It is therefore extremely likely that a high percentage of the dog-whelks were broken before they were discarded.

One possible explanation for the breaking of shells was the use of the animal for food, and especially for bait, where the inclusion of unpalatable shell fragments in the animal would be less of a hindrance (Deith 1989). The author’s experience in preparing animals for his own collection shows that enough of a dog-whelk to be suitable for fish bait or for eating can be removed with a narrow flexible piece of wood or bent fish-hook without the need to break the shell. Therefore, it can be expected that assemblages of dog-whelks used as food or bait will have low or moderate percentages of the shells broken intentionally. This was the condition of the dog-whelk component of analysed British Mesolithic middens, where such shells are interpreted as food remains (Russell et al 1995, 287; Jones 1984, 180; Thomas and Mannino 1999, 97). As the dog-whelk shells in deposit 708 were heavily broken intentionally, it was unlikely that they are food or bait refuse.

The pattern in the broken forms 1 to 4, and the absence of fragments with the principal part of the main (final) whorl, indicate that the shells were broken by a blow to the outside of the main whorl opposite the aperture, the blow being strong enough to break the shell but not so strong as to crush it. The author reproduced forms 2 and 4 by holding a modern dog-whelk shell aperture downwards on a hard surface and delivering a sharp blow with a small hammer to the uppermost surface of the shell (the back of the main whorl). The intent would appear to have been to extract all the animal while keeping it intact. Another explanation for breaking open dog-whelk shells while keeping the animals whole is that this was the first stage in producing the dyestuff called ‘Tyrian purple’, also called ‘royal-’ or ‘imperial purple’. Once a common industry in Mediterranean antiquity, with its origins in Minoan Crete (Reese 1987, 203), the raw material for the dye is secreted by a small gland in several species of marine gastropod, of which three or four were exploited for dye production. In contrast to the Mesolithic British middens noted above, the shells from deposit 708 were broken similar to those in deposits in the Mediterranean interpreted as waste from Tyrian purple production, which usually were described as very heavily broken or ‘crushed’ (e.g. Reese 1987, 204). Pliny the Elder (Hist. Nat. IX.38) described the breaking of the shell, the removal of the intact animal, the dissecting out of the tiny gland and its complex subsequent treatment to extract the dye. There is increasing evidence for Roman-era production of Tyrian purple in Atlantic Europe. Intentionally broken dog-whelk shells were recovered from Roman-period sites in Cornwall (Light 2003, 54), where fragments of forms 1 and 2 are illustrated (Light 1995, 149 and fig. 34). There is also evidence of Dark-Age production of Tyrian purple. According to Bede writing about AD 730 (Hist. Eccl. I.1), it was a notable industry in England; as this was outside the range of the Mediterranean species of shellfish used to make purple, the animal providing the raw material must have been Nucella lapillus. Intentionally broken dog-whelk shells were recovered from ecclesiastical settlements of c.8th–9th century in western Ireland (Murray 1999, 363–413). The principal use of Tyrian purple in the Mediterranean world was the dyeing of cloth, but deposit 708 was too small for the production of Tyrian purple on the scale required. However, good quality pigments would be needed in just such small amounts for that quintessential ecclesiastical activity, the illumination of manuscripts. Recently, Tyrian purple has been found by a new application of non-destructive testing in a manuscript of the late 8th century AD (Porter et al. forthcoming). It therefore seems likely that occupants of Le Yaudet were making sophisticated use of local resources to produce illustrated manuscripts at the time when the shells of deposit 708 were discarded.

Interpretation of the remains from Le Yaudet

Strategies employed

Although the number of shell-rich deposits encountered and sampled at the site is small and unevenly distributed through time, the story of shellfish gathering at Le Yaudet can be outlined. That story is one of continuity through time in the face of cultural change. Gatherers concentrated on the same four shellfish (limpets, mussels, cockles and urchins), regardless of the period in which they undertook the gathering. Limpets were the main animal sought in every period, and for the most part a particular size of limpet (consistently 20–35 mm long) was selected from all those available on the shore and gathered mainly from moderately wave-beaten shores around or below mid-tide level. It seems that people familiar with limpets (whether La Tène, Roman, Dark Age or medieval) came to the same view on what constituted a limpet of good quality, and the places on the shore most likely to have them. Some samples had limpets gathered from a wider range of shore conditions but none had a concentration of large-sized limpets, so maximizing meat yield for each animal gathered was never part of the strategy. The view that falling average shell length over time is evidence for limpet over-exploitation is unsupported, and the opposite would appear to be true. Spatial analysis of the pattern of breakage for one of the medieval limpet tips revealed that limpets had a smaller average size with increasing fragmentation, so damage may affect larger limpets more than smaller ones.

There was no evidence that shellfish were over-exploited. No samples with groups of limpets with exceptionally high or low mean lengths were found, so gathering of limpets was not shown to have been so intensive that the good limpets were almost used up and less than ideal limpets had to be accepted. Cockle population structure was only interpretable for the medieval period. One group grew at a consistent rate for five years prior to harvest. The cockles gathered were larger and older than cockles now gathered commercially. One population of sea urchins in the tip from the Gallic Empire period had been growing for the preceding nine years without an episode where a substantial number of the larger animals were lost. Amount of food provided

These shell-rich deposits were sizable, but not the enormous features sometimes built up in early prehistory. Medieval limpet-rich tip 634 seemed roughly typical of the size of shell- rich deposits found, and was the only example completely exposed by the excavation which also had in-ground shell density data. A rough cylinder about 230 cm across and 10 cm deep (based on the average depth of its incremental samples), it therefore had an approximate volume of about 400 litres. With an average in-ground density (from the data in Table 2) of 9.3 shells per 100 ml, tip 634 would have contained in the order of 40,000 shells. With a shell typically 32 mm long for the deposit (from Table 4) and about 2 mm thick, a typical limpet would have been about 2.8 cm long. As shell width is typically 82% of length, the animal would have been 2.4 cm wide, or on average 2.6 cm across. The typical shell height/length ratio was 0.36 in this deposit, so an animal 2.8 cm long would have been about 1.0 cm high. Using the formula for the volume of a cone gives an estimate of the average meat volume of 1.8 ml per limpet, so the shells in tip 634 would have provided about 70 litres of meat. Assuming an average serving of about 300 ml, the shells of tip 634 would have provided a hearty seafood lunch for about 200 people. The shell-rich tips were therefore too large to be the daily meal of any typical agrarian household. They were the food remains for some sizable event (a public feast day, or a communal labour such as the spring crop sowing) or for preserving.

Season

Throughout the occupation of Le Yaudet gathering of shellfish on a scale large enough to produce shell-rich deposits seems to have been principally a spring activity. Sea urchin remains were recovered from most samples and from every period of occupation, and Atlantic European coastal urchins are ripe (filled with roe) only in the colder part of the year, mainly in spring. Smaller-scale gathering of shellfish in other seasons may have taken place, but may not have been preserved if the shells were discarded along with other general waste or table waste, not as tips. Medieval limpet-shell tips 557 and 558 contained no urchins even in the finer fractions where urchin remains were concentrated, so the medieval occupants may have gathered limpets intensively in the warmer months from time to time. Medieval cockles were gathered early in spring for the tip 569 (the East midden), and somewhat later for the South midden, according to growth-ring evidence.

Formation of the shell-rich deposits

The shell-rich deposits, whether middens or tips, were primarily waste from the processing of shellfish prior to consumption (‘kitchen waste’), with few examples of waste from consumption (‘table waste’). The main shells recovered from Le Yaudet are those which commonly undergo some processing prior to serving. Typically shellfish are eaten cooked (by boiling, steaming, or roasting) rather than raw. Cooking usually opens the shell of bivalves such as cockles, mussels, razor shells and carpet shells and frequently frees the animal from the shell, so often the flesh is extracted before serving and the shell discarded as ‘kitchen waste’. Intact cockles from the South and East middens were probably discarded as dead ‘gapers’ prior to cooking. The author’s own experience with limpets showed that these are one of the gastropods that can be removed from their shells easily prior to cooking (by putting the shell-tip in the fingers and pushing against the head with the thumbnail). Limpets invariably fall out once cooked, leaving a mass of shells to be separated from the flesh and discarded. Within Gallic Empire pit F1081, the first fill was principally waste from preparing sea urchins prior to consumption, while the later fills contained a mixture of kitchen waste and waste from urchin consumption.

Some of the shells recovered in small numbers can be explained as kitchen waste. Ormers are tough and require extraction from their shells and tenderizing to be palatable (Davidson 1999, 2). The blue-rayed limpet and miscellaneous small winkles occasionally recovered from Le Yaudet live in seaweeds or rocky cracks. These were probably collected accidentally along with the main catch of limpets or mussels, or with coarse seaweeds used to keep the catch from drying out until needed. The mussel evidence indicated that these had been gathered in mats. Prior to cooking the usable mussels would have been extracted and cleaned, and the remainder discarded as kitchen waste. These discarded mats would have been the source of small mussels, dead open mussel shells (with encrusting organisms on the inside), associated organisms (such as barnacles, stray yearling oysters such as those in Sample 304 in East midden 569, and limpets less than 17 mm), and the occasional common dog-whelk or sting-winkle (predators of sessile organisms, predominantly barnacles and mussels).

The absence and unexpected rarity of manifestly edible shellfish can also be explained by the shell-rich deposits being kitchen waste. An examination of Table 1 shows no crab fragments, even as 2 mm fragments. This lack cannot be explained by differential preservation since fish scales, which are chemically similar, were preserved in many deposits. Edible whelks (Buccinum undatum) were also absent even though other palatable but principally sub-littoral animals (ormers, sea urchins) were well represented. Other edible sea creatures are curiously rare from the sampled deposits. Periwinkles (Littorina littorea) were usually quite small and (apart from Iron Age layer 644 and the bulk sample of South midden layer 561) they were outnumbered by other miscellaneous gastropods too small to eat. The large top-shells (Monodonta lineata) were, like the periwinkles, uncommon and small even though the Mediterranean species is eaten (Davidson 1999, 802) and the Atlantic species found at Le Yaudet formed a significant part of some Atlantic shell middens (e.g. Thomas and Mannino 1999, 96). Oysters were also rare at Le Yaudet, even though they were often the principal shell in middens. Those not fragmented were misshapen and thin (urchin-rich Gallic Empire fill 700) or were too small to eat (those in bulk sample 304 of East midden 569 were all less than 20 mm).

Typically these unexpectedly absent or rare shellfish are served in their shells, with their shells forming part of the post-consumption waste. The periwinkles, large top-shells and oysters from shell-rich deposits were often unusually small or misshaped, so they were rejected as sub-standard and discarded with the rest of the kitchen waste. Most of the waste from consumption appears to have been disposed of in a different manner and in a different place from shell-rich kitchen waste, throughout the occupation of the site. Such ‘table waste’ was recognized as an element in only one deposit, the later fills in Gallic Empire pit F1081, when this feature was so full it would have been almost indistinguishable from the surrounding midden deposits. Post-consumption waste including fragments of crabs, whelks, periwinkles and oysters may have been discarded in more general spreads peripheral to structures, or as manuring scatters in fields.

Sea temperature

Based on the temperature tolerances of the animals recovered, average sea temperature seems to have been about the same as that on the modern shore from at least the early Roman period onwards. Common limpets (Patella vulgata) and periwinkles (Littorina littorea) are typical of cool or cold seas, with a distribution at present north of south Portugal (Southward et al. 1995, 134 and 135). Limpets of temperate seas (Patella ulyssiponensis and P. depressa) from early Roman layer 646 and later deposits showed sea temperatures were at most only slightly colder than at present. Since gathering was consistently an activity of the low and middle moderately exposed shores concentrating on limpets about 30 mm, the Portuguese limpet Patella rustica (a warm water animal, but small and restricted to high wave-beaten shores) could not be used as an indicator of a warm sea, since it would have been missed by the gatherers even if it was present. Sea urchins found were those of temperate seas, and types found now on the Breton coast (Paracentrotus from Gallic Empire onwards, Psammechinus from Dark Age and medieval periods), with no evidence that the sea became cold enough for the north Atlantic urchin Strongylocentrotus. The top-shells (Gibbula from the Gallic Empire onwards, Monodonta from the medieval) are temperate-ocean animals found on the modern coast of northern Brittany, where they are near their present north limit (Hawkins and Jones 1992, 20). The ormers (Haliotis, from Gallic Empire onwards) are also north Breton shellfish near their present north limit (Hayward et al. 1996, 178). Ormers and top-shells were found only rarely, so their absence from some deposits does not demonstrate these were absent from the shore at that period. Common mussels (the more widely distributed form) and French mussels (the more temperate form) were recovered in about the same proportion from the early Roman period onwards. The assemblage from the Late Iron Age was small, so sea temperature cannot be reconstructed accurately by this type of interpretation because the species critical for it may have been missed by chance. However, those species found were completely consistent with a sea temperature like the present. Precise estimation of sea temperature for this period would require more advanced techniques (such as oxygen-isotope analysis) or the discovery and sampling of Iron Age shell-rich deposits during further excavations.

Chronology

One sample of a shell-rich deposit was collected from the prehistoric period, from within La Tène layer 644. Based on the stone, artefact and shell content compared to those of other deposits, this sample was from a tip of domestic waste including some shellfish. All four of the shellfish that make up the bulk of the types used for food throughout the occupation of Le Yaudet (mussels, cockles, sea urchins and especially limpets) were already being exploited in this period. Soft shores (for cockles and razor-shells) were exploited, but almost all the gathering was on rocky shores for limpets, a strategy typical for later shell-rich deposits. For this group of limpets the gatherers worked the middle to lower shore in damp sheltered conditions, returning with limpets smaller than any of the groups recovered, making them the most discriminating gatherers.

The early Roman period samples came from the successive layers 643 and 646. Based on the stone, artefact and shell content compared to those of other deposits, these layers were tips of shellfish processing waste. All four of the shellfish that make up the bulk of the types used for food throughout the occupation of Le Yaudet (mussels, cockles, sea urchins and especially limpets) were exploited in the period. Both rocky shores (for limpets, mussels and sea urchins) and mud-flats (for cockles) were exploited, but mud-flats seemed to be very little used. Gathering was principally on the rocks for limpets, a strategy typical for the site’s shell-rich deposits. Gatherers concentrated on limpets of moderate size for layer 643, but used the unusual strategy of ranging across the tidal range gathering limpets of any useful size for tip 646. Early Roman gatherers tended to be less discriminating about their limpets than gatherers of other periods.

Five Gallic Empire deposits were sampled. There were too few shells from the base fill 523 of feature F600 to interpret. Marine invertebrates were rich in several of the fills in pit F1081. Its base fill 700 was a tip of kitchen waste rich in urchin remains (about eight urchins per litre of sample) dominated by the violet sea urchin, with the occasional common urchin. This foray for urchins took place in the winter or spring (since that is when urchins are ripe), probably during a monthly ‘spring’ low tide since the urchins came from two populations, one whose aged, small, slow-growing nature was consistent with life between low-water neap and low-water spring tide levels or in rock-pools, and another larger and faster-growing group consistent with life around low-water spring tide level. Urchins of about 40 mm were the minimum acceptable, with those over 50 mm much preferred; the strategy was to maximize meat yield per individual urchin. The urchins were prepared for the table by removing the base and mouthparts and using the curved shells as a bowl, a common method for serving fresh urchins today. Gatherers used the unusual strategy of collecting any limpets of acceptable size, while ranging throughout the tidal range. A few mussels and oysters were also gathered, probably along with the limpets while the gatherers followed the ebbing tide down until it exposed the urchins.

Limpets were the main quarry for the gatherers in later fills of this pit, where they were more discriminating in both part of the shore and the type of limpet sought. They concentrated their efforts on limpets of a moderate size. For middle fill 685 they collected on the middle and low shore, with shelter and with moderate-strong exposure to wave action, exposed enough for china limpets, blackfoot limpets and mussels (and for dog-whelks and sting- winkles, who prey on them and on barnacles).

The final fill, 681 was quite similar to that of a midden, with soil accreting and the shells degrading. Along with some mussels, limpets were gathered from damp sheltered conditions probably with some seaweed cover for fill 681’s north part (SS 682). However, fewer mussels and limpets were collected from more dry and open middle and low shores for the south part (SS 678), when rocky shores at monthly extreme low tide (‘spring’ tide) were explored for ormers. The later fills also accumulated in the winter or spring, as shown by the sea urchins, whose small and decreasing numbers may show the time of accumulation was near to the end of the urchin season. It is possible that the pit filled up during a single year.

As in earlier periods, shells of soft shores (such as cockles) were seldom brought back. Therefore, these flats were present in some form at that time and their shellfish were gathered, but without the intensity required to produce shell-rich deposits.

Dark Age tip 708 was unusual because it was the only one in which dog-whelks made up a significant part of the shells. Limpet gathering concentrated on small low-shore animals, in conditions relatively free of cover and algae (shown by the near absence of spirorbid-bearing forms in Figure 8). Mussels were collected in levels low enough to indicate these were gathered opportunistically while collecting limpets, with the very small limpets (less than 18 mm) entangled in the mats of mussels. The work of gathering seems to have concentrated on middle-shore open areas with patches of mussels, a good area for finding dog-whelks in quantity. The dog-whelks were broken up in a way which argues that they were exploited for pigment extraction, so it seems probable that the gatherers sought out an area where they knew dog-whelks would be available, and opportunistically gathered limpets and some mussels for food while there.

Most samples came from the medieval period. The size and nature of the shell-rich deposits were similar to those already discussed, although the lack of small samples of known volume in the ground hampered comparison of deposits and forced the assumption that many deposits were middens rather than tips. The incremental sampling of the South midden sequence (deposits 581, 585, 561 and 543) demonstrated that this sequence was a midden. The incremental sampling of East midden 569 demonstrated that it was not a midden, but a very large tip of cockles with some limpets rapidly followed by a very large tip of mussels. While not fully excavated it was clear that deposit 569 was much larger than other shell-rich deposits at the site, and probably was the waste from shellfish harvesting on a cottage industry scale.

The medieval tips relatively early in the medieval stratigraphic sequence (deposits 634, 557, 558, and 581) formed a distinct group, being almost entirely made up of limpets from low- shore damp places, with gathering of mussels opportunistic (possibly as the tide receded to the point that the limpets were sought, and again after the tide rose and covered the limpets) and gathering of soft-shore animals minimal, as in earlier periods. Slightly later in the medieval period tips (deposits 586, 560 and the early increment sample 443 of South midden layer 561) had a higher proportion of mussels and soft-shore shellfish gathered with the limpets, with the soft-shore shellfish (razor shells, carpet shells, tellins) tending to be animals of sand rather than mud. The possible middens 217, 219, 451 and 204 were similar to this slightly later medieval group of tips, with limpets gathered low on shore for the first three of these, and across a shore generally open and dry for 204. In the latest medieval samples, cockles had replaced limpets as the principal animals gathered, both in tips (East midden 569) and middens (the later increments of South midden 561). These cockles included a high proportion of lagoon cockles, of more muddy ground in calm water.

This gradual change in the dominant element from rocky shore to soft mud (possibly via soft but predominantly sand) is most simply explained by the bay silting up, gradually changing from mainly rocky to mainly soft ground. Cultural change is implausible because we know the culture of the occupants of Le Yaudet changed over time but the range of shellfish they were willing to gather remained the same throughout. The change seems to be one of changing availability rather than changing acceptability. Sudden sea-level change is also implausible, because sea level has been rising gradually in Brittany for the last six millennia (Lambeck 1997, 17–18). The most likely cause is the gradual deterioration of the medieval climate which culminated in the Little Ice Age of about AD 1500–1750, a period of deepening cold, intensifying winter storms and coastal change, such as the onset and growth of coastal dunes around the North Sea (Knight et al. 1998). Throughout the medieval period the limpet evidence favours rocky shorelines of damp weed rather than bare rock, a shoreline consistent with a cool damp climate. Temperate-water animals were still gathered from the shore throughout this change, so the silting up of the bay was advanced without a serious drop in mean sea temperature. The present view of the bay from Le Yaudet at low tide may have been familiar to its late medieval occupants, but not to any of their predecessors, whose shoreline lies sealed under medieval silts. Acknowledgments

Thanks are due to many people for the information and advice given for this analysis. The Oxford Archaeological Unit provided facilites and manpower for the first season’s sample processing, and Gill Campbell kindly provided sample processing facilities at The Centre for Archaeology, Fort Cumberland, Portsmouth for later seasons. Clive Orton, Professor of Quantitative Archaeology at University College London, provided guidance on some of the statistical methods. Dr Allan Jones of the University of Dundee gave advice on cockle ecology. Dr Maeve Kelly at the Scottish Association for Marine Science, Oban advised on sea urchins and contributed comparative material. Ms Jenny Mallinson, keeper of live collections at the Southampton Oceanographic Centre, also donated comparative material. Dr Katherine Szabo of the Australian National University, Dr R.G. Matson and Joanne Green of the University of British Columbia, Professor David Black of the University of New Brunswick, Dr Bruce Bourque of the Maine State Museum and Yves Gruet of Nantes all contributed advice on sea urchins in archaeology, and some granted access to research prior to publication. Dr Emily Murray of Queen’s University Belfast provided information on dog- whelk use in Dark Age Ireland and access to unpublished material; Dr David Reese of the Peabody Museum of Natural History at Yale University advised on the archaeology of Tyrian purple; and conservator Dr Cheryl Porter of Cambridge advised on early medieval Tyrian purple and granted access to research prior to publication.

Particular thanks go to Professor E. Wyn Knight-Jones, until recently of University College Swansea, for advice on spirorbid worm ecology and identification of some specimens; to Dr Andrew Smith at the Natural History Museum, London for advice on sea urchins, access to museum collections, identification of some material, and reading a draft of the sea urchin section of this work; and to Professor S.J. Hawkins, Director of the Marine Biological Association, for advice on rocky shore ecology, help and guidance with limpets and their identification, continued interest, and reading a draft of this work. Errors in the text remain the responsibility of the author.

Context number 644 643 646 523 700 685 Sample Number 316 317 315 288 681 680 La E E Gal Gal Gal common name and element Biological name Tene Rom Rom Emp Emp Emp Ormer Haliotis sp. 1 Common Limpet Patella vulgata L. 39 53 274 69 928 China Limpet P. ulyssiponensis Gmelin 5 4 24 Blackfoot limpet P. depressa Pennant 6 Limpet indet Patella sp. 11 46 86 1 7 28 Blue-rayed limpet Helcion pellucidum (L.) toothed top shell Monodonta lineata (da Costa) grey top shell Gibbula cineraria (L.) flat top shell Gibbula umbilicalis (da Costa) top shell type cf Gibbula 1 1 Periwinkle Littorina littorea (L.) 1 flat winkle Littorina obtusata (L.) 1 1 rock winkle Littorina mariae Sacchi & Rastelli 1 rough winkle Littorina saxatilis agg. 1 1 common dog whelk Nucella lapillus (L.) 1 1 2 sting winkle Ocenebra erinacea (L.) 1 netted dog whelk Hinia reticulata (L.) pygmy dog whelk Hinia pygmaea (Lamarck) large land snail Helicidae slug plates cf. Limax common mussel: left valves Mytilus edulis L. 1 5 14 14 2 common mussel: right valves Mytilus edulis L. 1 3 10 13 3 French mussel: left valves Mytilus galloprovincialis (Lamarck) 3 French mussel: right valves Mytilus galloprovincialis (Lamarck) 1 2 2 mussel (indet): left valves Mytilus sp. 2 11 3 mussel (indet): right valves Mytilus sp. 1 16 8 mussel: either valve Mytilus sp. 1 common oyster base Ostrea edulis L. 1 common oyster top Ostrea edulis L. 1 Oyster cf. Ostrea edulis edible cockle: left valves Cerastoderma edule (L.) edible cockle: right valves Cerastoderma edule (L.) 1 lagoon cockle: left valves Cerastoderma glaucum (Poiret) lagoon cockle: right valves Cerastoderma glaucum (Poiret) cockle (indet): left valve Cerastoderma sp. 2 cockle (indet): right valve Cerastoderma sp. cockle type: shell fragment cf. Cerastoderma 1 1 2 gold carpet shell: L valve Paphia aurea (Gmelin) gold carpet shell: R valve Paphia aurea (Gmelin) chequered carpet shell: L valve Tapes decussatus (L.) chequered carpet shell: R valve Tapes decussatus (L.) chequered carpet shell: fragmnt cf. Tapes decussatus 1 Thin tellin: L valve Angulus tenuis (da Costa) grooved razor shell: L valve Solen marginatus Montague grooved razor shell: R valve Solen marginatus Montague Common razor shell: L valve Ensis sp. Common razor shell: R valve Ensis sp. Razor shell type Solenidae family 1 violet sea urchin MNI Paracentrotus lividus Lamarck 1 79 3 green sea urchin MNI Psammechinus sp. common sea urchin MNI Echinus esculentus L. 1 indet sea urchin MNI Family Echinidae 5 2 sea urchin rare test fragment Family Echinidae 1 2 4 sea urchin rare spine Family Echinidae 1 1 total items 58 111 400 7 227 1019 volume processed (litres) 4 28 7 10 11 10 density (items/litre) 15 4 57 0.7 21 102

Table 1. Taxa recovered from the bulk samples at Le Yaudet

681 (N) 681 (S) 708 557 558 586 560 569 561 219 217 204 451 682 678 679 300 305 306 302 304 301 161 160 163 34 Gal Gal Dk med'l med'l med'l med'l med'l med'l med'l med'l med'l med'l Emp Emp Age 2 3 1 1 5 1 1 5 1 6 2 144 239 2131 875 1147 1217 870 117 299 168 16 343 55 9 2 1 8 2 1 43 19 2 6 3 4 16 47 471 51 148 93 87 28 46 60 43 12 1 1 1 1 1 1 3 2 2 1 1 1 1 1 1 5 4 4 10 3 1 1 7 1 1 1 1 1 3 4 1 3 3 1 4 1 1 1 4 160 1 1 2 1 1 2 1 1 3 1 4 2 6 13 10 2 59 2 121 37 78 71 2 13 15 2 65 3 109 31 57 78 7 10 1 14 1 34 11 12 49 13 10 1 1 12 29 13 10 26 4 15 28 12 160 3 5 377 145 505 293 4 18 16 4 34 10 134 2 8 394 117 476 292 5 17 18 3 2 9 4 1 5 1 1 1 1 21 3 1 188 185 11 5 4 1 16 3 1 202 177 1 1 9 10 57 1 94 177 59 107 157 1 1

1 2 1 2 8 1 1 5 6 2 2 1 32 8 1 2 19 5 2 2 1 1 1 5 1 4 2 1 1 1 1

1 1 4 3 1 255 333 3412 940 1322 2483 1352 1933 1906 298 88 489 81 10 10 23 14 15 22 27 15 15 10 5 10 25 26 33 148 67 88 113 50 129 127 30 18 49 3

Table 1. Taxa recovered from the bulk samples at Le Yaudet