6.1 Introduction
Like all indigenous pottery from North-Holland in the Roman period, that of Uitgeest and Schagen was tempered with organic materials visible as the carbonized remains or negative impressions of vegetable fibres. Only a very small percentage of the pottery of Uitgeest is also tempered with crushed shell, but none is included in the samples; one vessel from Schagen is exclusively shell-tempered. The shell species are of a marine origin (identifications by mr. W. Kuyper, FAL). Shell-temper will therefore not be dis-cussed any further1.
The effects of organic, fibrous temper on the properties of fabrics were discussed in chapter 2.4-5. To repeat, it will in general improve the workability and plasticity of the clay. The use of this type of temper results in a relatively high thermal stress resistance and a high percentage apparent porosity (%AP) of the fabric. The two properties are influ-enced by (a) the type of organic matter, (b) the size of the fibres and (c) their quantity. The ‘ideal or standard recipes’ (DeBoer & Lathrap 1979) can vary for different functional groups, but the standard is often set by the most frequently made vessel. As argued before, if cooking pots set the stan-dard for the fabric composition, organic matter is a very suitable choice, even though it does not improve and proba-bly decreases the mechanical strength of a vessel. The pot-ters had therefore to decide on the ‘right’ type, size and amount of organic material to be added for specific use-categories of the pottery, in relation to the type of clay. The remains visible in the pottery suggest that fibres of plants were used. Potentially such remains were available to the potters in several forms: dung from sheep, horses and cows, ashes, the stems and leaves of grasses, reeds or culti-vated plants like barley, wheat, flax, etc. These kinds of temper are also mentioned in the few ethnographic accounts on the subject (David & David-Hennig 1971; Saraswati & Behura 1966; Thompson 1965). The type of material together with the way it is prepared will influence the size and shape of the remains found in the pottery.
The methods for identification and quantification of fibrous temper in archaeological material have received limited attention so far. Data on the botanical identification of organic filler present in prehistoric pottery are hardly
available. Stöckli (1979) mentioned dung as the most likely source for Iron Age pottery from Manching. At the onset of this study, neither data nor proper methods were available in the literature for the quantification of this type of temper. The development and testing of methods for identification and for quantification of the amount of temper, is therefore an important part of this research. The aim was to establish a relatively quick and easy way to estimate the amount of temper in a fabric and through this, the choices of basic recipes by the potters. The few examples from ethnographic studies (chapter 2.4) indicate that volumes of vegetable temper are not measured precisely, while the test tablets of this study show that even when they are, there is some variation in the density of temper. If the potters in North-Holland did indeed use an ideal or standard recipe, the standard amount of temper can expected to be a yardstick for the amount used for other functional groups, when and if variations on a basic recipe were applied. Such variations are an important criterion for the fabric classification on the one hand (see chapter 7) and for the relationship with the func-tion and use of the pottery on the other (see chapter 9). The hypothesis for the pottery studied here is that the amount of temper added will not have been measured very precisely, partly because of a low degree of functional differentiation; therefore, even within a standard recipe, some variation in the amount is to be expected between vessels (chapter 2.5). The quantification methods were chosen accordingly. They are rather imprecise by mathematical standards, but have the great advantage of speed in measurements and are still well within the degree of precision which is likely to have been exercised by the potters.
Three separate methods were used. Firstly, a series of test-tablets with different types, sizes and quantities of organic matter were made, using the same Holocene clays from North-Holland. The methods of preparation and the results are presented in the next section. These test-series were used mainly as a reference set for the volume percentage of tem-per and additionally for the identification of the type of organic matter. A small number of sherds (15) were studied for species identification of the fibrous remains by prof. dr. C.C. Bakels and mr. W. Kuyper, see below. Secondly, a specific method was developed for the quantification of the
amount and size of temper in the pottery. The test tablets were used as a control for this method. Thirdly, the %AP was measured for all tablets of the test-series and for a selection of the sherds from Uitgeest and Schagen.
Method of quantification
The method for quantification of the amount and size of temper was designed to meet three conditions. The method (a) should allow the processing of large samples, (b) should be reliable and precise enough to allow a classification of amounts of temper in the pottery, and (c) should be such that any archaeologist would be able to use it.
The development of suitable techniques for quantification has taken up much time and effort. Firstly, the visibility of the temper on the surfaces of the vessel is strongly influ-enced by the finishing techniques, which can result in a complete obliteration of the temper, for example when the surface has been polished. Furthermore, organic temper cannot be measured in any reliable way on fractures, as the fibres tend to have an orientation parallel to the vessel sur-faces and thus perpendicular to the fracture structure. Finally the coiling- and joints-technique results in an uneven distrib-ution of pores (Vandiver 1988, 142-4, fig. 4). The soldistrib-ution to these problems was sawing sherds lengthwise (see chapter 2.6.1), which resulted in core ‘surfaces’ in which the temper (fibres) shows up quite clearly (fig. 6.10). Such core ‘sur-faces’ offer great advantages for temper analysis(those for observations on clay composition were already discussed in chapter 2.5 and 5). A first one is that the fibrous remains are exposed in an optimal way, as most fibres are embedded horizontally in the matrix. This way, the amount as well as the length and width of fibres can be measured. The second one is the relatively large surface obtained in this manner, allowing a much greater control over variations in the amounts of temper, both within one sherd and between sherds. Moreover, the influence of surface treatment is excluded and measurements of the amount of temper can be compared for all sherds. Finally, the temper in the core is often still present in a carbonized form, whereas it is burnt out at the surfaces. This made the identification of temper, as opposed to pores caused by other factors such as knead-ing, much easier and more reliable. The specific counting method is explained in section 6.3.
6.2 Test tablets with organic temper
Three of the clays used in the firing experiments (chapter 4.2) were also used as the basis for three sets of test tablets with different kinds of temper and of paste preparation, all with controlled quantities of temper. The aims of this experi-ment were the following:
– To obtain control data on the porosity of fibre-tempered pottery.
– To compare the impressions left by known tempering materials with those in the pottery and to identify the temper used by the prehistoric potters.
– To create a standard reference set to estimate the volume percentage of organic temper present in the sherds. If such a method would prove to be reasonable accurate, it could be a tremendous time-saving device for the study of pottery with organic temper from other sites.
– To provide a means of comparison for the method of counting the amount of temper as developed for this study.
The choice of tempering materials was based on the discus-sion above. Three series of tablets were made:
SET A
Clay 64 with hay*, cut to a size of 10 mm SET A 1
Clay 64 with hay, cut to 5 mm SET A 2
Clay 65 with dried horse dung SET A 3
Clay D3004 with the finer fractions
of sheep dung SET A 4.1
Clay D3004 with the coarse fractions
of sheep dung SET A 4.2
* dried grass, sold as rabbit-fodder
Volume of temper : Volume of clay (approximate:) 5%- 15%- 33%- 50%- 60%- 75%
SET B
Clay 63 with fine grass, hay cut to 0.5 cm. SET B 1 Clay 64 with:
– hay cut to 1 cm. SET B 2
– the coarser fractions of dried horse dung SET B 3.1 – the finer fractions of horse dung SET B 3.2
Volume of temper : Volume of clay
5% - 15% - 20% - 25% - 30% - 35% - 40% - 45% - 50% - 60%
SET C
Clay 64 with specially prepared hay temper Volume of temper : Volume of clay
5% - 10% - 15% - 20% - 25% - 30% - 35% - 40% - 45% - 50%
The volume of temper will be referred to as the volume %, hereafter abbreviated to vol%. In fig. 6.1, the cores of the test tablets of set B and C are shown for each vol%.
6.2.1 METHODS OF PREPARATION
fibres longer than 10 mm and 5 mm respectively. It was expected that the material would break up into smaller bits during the kneading of the paste. The horse and sheep dung was collected from pastured animals which were not given any additional fodder2.
The clays were dried and pounded to a fine dust. Clay and temper were mixed on the basis of their dry volumes. Both the horse- and sheep-dung proved to be very hard once dried and had to be broken up by pounding; the resulting fractions of ‘coarse’ and ‘fine’ dung were used separately. The fine dust measured less than 3 mm. The tablets with sheep dung were limited to 5 and 15 vol% for the coarse fraction (set A 4.1) and 5, 15, 33 and 50% for the finer fractions (set A 4.2). The temper was lightly pressed to exclude air. After mixing the dry clay and temper, water was added in quantities necessary and the paste was thoroughly kneaded. The tablets of set A were pressed into blocks of 33 ≈ 99 mm, allowing shrinkage up to 10 %. The tablets were fired in a gas oven. The maximum temperature of 750 °C was reached after circa 6-7 hours and maintained for one hour. During this hour, the atmosphere in the oven was changed to a more or less reducing one. To aid reduction, a bunch of wet grass was put in the chimney of the oven. The oven was allowed to cool off gradually over circa 16 hours to room temperature with the oven door and the chimney closed. For several reasons, a different method of shaping the tablets was used for the second series, set B. More control was sought over the influence of organic temper itself. The first set of grass-tempered tablets also lacked the most important characteristics of the pottery, the alignment of the fibres parallel to the wall. In the first series these fibres were ori-ented any odd way, resulting in many undefinable pores in the cut surface; consequently, the distributions of the fibres in the tablets and the sherds did not match very well. For set B, the paste was rolled into strips, copying as exactly as possible the making of a coil. Moreover, the ratios of temper : clay volumes were extended, as a finer division in the ranges between 15 and 40 vol% was desirable. The tablets were fired in the same manner as the first series. This way of preparing the tablets clearly increased comparability between tablets and sherds; the distribution of the fibres became indeed much more ‘realistic’, especially in hay-tempered tablets (fig. 6.1). This experiment proved that the construction technique of coiling does indeed cause a spe-cific arrangement and distribution of organic fibres in the vessel wall3. The similarity of the tablets with hay fibres was, however, still not as good as aimed for. The specific form of the cavities, the length, but especially the width of the fibres were on average much larger than in the majority of the sherds, even for set B 1. Apparently, the intensive kneading did not result in a further breaking up of fibres into smaller sizes. On the other hand, the fibres in the finer horse
dung temper left cavities that were too short, although the width was more comparable to the fibres in the pottery4. The experience gained from set A and B was used for the third set of test tablets, set C, to create a reference set with as much similarity to the characteristics of the fibres in the pottery as possible. The same type of hay as had been used in the previous sets was prepared in a different way. Instead of cutting it, the hay was rubbed by hand. This indeed caused it to break up both across and along the veins into small fragments. These fragments were sieved to remove most of the fibres still larger than 5 mm as well as most of the very fine ‘dust’ (<1 mm) that resulted from the rubbing. In this way the specific length and width (variation) of the fibres in the pottery was simulated. The same methods as for the second series were used in the preparation of the paste (with clay 64) and the forming and firing of the tablets. Tablets with 60 vol% temper were not made as those of set B tended to crumble; instead, a tablet with 10 vol% temper was added to allow a better comparison in the lower range. Set C indeed showed the desired effect, a very good fit with the majority of the sherds. The length and width of the temper as well as its distribution pattern through the core of the tablets were quite similar to those in the pottery.
6.2.2 THE APPARENT POROSITY IN THE TABLETS OF SETS A,B,AND C
Legend:
Set B2 Clay sample 64 with hay cut to < 1cm. Set B1 Clay sample 63 with hay cut to < 0.5 cm
35 40 45 50 60 35 40 45 50 60 B.3.2 B.3.2 Legend:
%AP from 0 to 35 vol% of temper is 2.3% in set C. The average is based on the values for each vol% in fig. 6.1. For the finer dung temper (set B 3.2), the average increase in %AP from 0 to 45 vol% is only 1.4%. In this set there is a sudden increase in %AP from 25 to 30 vol%. It is possible that this is caused by the preparation of the tablets: the paste with dung was difficult to mix and roll without breaking and tearing it (pers. comm. E. Mulder). The tablet with 30 vol% might have contained a large air pocket. The porosity in tablets with the coarser fractions of horse dung (B 3.1) is the lowest of all, with an average increase of 1.3% from 5 to 40 vol% of temper. For set A 3, with unselected horse dung, the %AP was measured for 25, 35, 50 and 60 vol.% only. Compared to set B 3, the porosity is higher up to 25 vol%, but above that value the differences become negligi-ble.
Secondly, the size of temper is obviously an important factor. The finer fractions of the same temper result in a slightly higher %AP than coarser material for both types of temper, but especially for dung. This result is consistent with that of Bronitsky & Hamer (1986) and can be traced back to differ-ences in the connectedness between pores (Reid 1984, 63). However, the effects of fibre size seem to be contradictory; for a single type of material finer temper results in a higher %AP, but between types the opposite is the case. All hay temper can be regarded as coarse in comparison with dung. This contradiction is more apparent than real and can be explained by the nature of the temper and the resulting pore structure. Dung is only partly made up of the remains of plants. Other organic components perhaps evaporize quickly, leaving no cavities in the matrix of the pottery. Finer frac-tions of the horse dung may result in a higher channel
5 10 0 30 35 40 45 50 55 60 15 20 25 30 35 40 45 50 55 60 % AP (7 00 °C ) Vol. % temper 1: B 3.1 2: B 3.2 3: A 3 4: B 2 5: C 6: B 1 1 3 2 4 6 5
Fig. 6.2 The %AP of the test tablets of set A3, set B 1-3 and set C for each vol% of temper. Legend:
porosity and therefore in a higher %AP than the coarser ones. At the same time the much smaller pores of dung compared to hay temper result in a much lower connectedness between pore spaces at the same vol%, and vice versa. The minimal differences between the three sets with hay temper, at least up to 35 vol% of temper, suggest a much higher connected-ness between pores, due to the size. Such an explanation is supported by the differences in the quantity of fibres visible on the cut surfaces in set B 1 and 2 (hay) and B 3 (dung) for the same vol% as well as by the firing ‘behaviour’ (see below). However, it cannot be excluded that the preparation method itself is of some influence; it is conceivable that the same volume-unit will contain slightly more of the finer material than of the coarser, especially for dung. Thirdly, the interaction between the kind of clay and the kind of material seems to have little effect on the resulting %AP (chapter 4, fig. 2), at least for clays 63, 64 and 65 (set A 3). The slight differences between clay 63 and 64 with the same type of temper is caused by the difference between the %AP of the clays themselves in an untempered state (see fig. 4.2).
6.2.3 FIRING PROPERTIES OF TEMPERED TEST TABLETS The attempt to fire the tablets in a reducing atmosphere, to prevent the complete combustion of the temper, failed; no carbonized remains were present after firing (fig. 6.1). Nev-ertheless, the different sets show an interesting pattern in the degree of oxidation of the clay itself. As all tablets were fired under exactly the same conditions, the differences in firing ‘behaviour’ are due exclusively to the size and amount of temper and the resulting pore structures.
All surfaces are oxidized after firing, but in the cores of the tablets a remarkable difference between types and vol% of temper can be seen. The cores of tablets with fine and pre-pared hay (B 1, C) are all oxidized and more so with higher vol% of temper; only in set B 2 (with coarse hay) the core is still slightly reduced in tablets with 5-30 vol% temper. In contrast, the cores of all tablets with dung temper (B 3) show little or no oxidation and the ‘reduced’ state is more intense with higher amounts of temper. These differences clearly show the influence of the kind and number of pores on the resulting connectedness (the pore structure) and through this on the degree of combustion. Hay temper results in a larger number of pores than dung with a higher channel porosity and more so when finer fractions are used; the pores allow oxygen to pass easily from the surface to the core of the tablet (Rye 1976, 110-1). With lower vol% of temper, up to 25%, the coarser fibres will mainly lead to pocket porosity as there are only limited connection between the separate pores; the cores therefore remained slightly reduced in set B 1. For the same reason the %AP is slightly lower as well. The prepared temper of set C, on the other
hand, leads to an increase in channel porosity, even at low vol%. The resulting porosity is also slightly higher and the difference is larger with higher vol%.
Dung temper, on the other hand, results in a different pore structure, in which the number and size of pores are much smaller and have a lower degree of connectedness at the same vol% than that of hay. Another important difference with hay temper is the fact that the pore structure does not change significantly with higher vol%; the channel porosity remains low. This affects the amount of oxygen that can penetrate to the cores of the tablets in the same length of firing time, resulting in the lack of oxidation of the core as well as to a significantly lower %AP.
The experiments definitely prove that the available oxygen in a fabric is first used up for the combustion of organic matter and only in second instance for the oxidation of other elements, such as iron, in the clay.
6.2.4 SUMMARY RESULTS
The following important conclusions can be drawn from these experiments:
– Increasing amounts of temper cause an increase in the %AP at a more or less regular rate, but this is mediated by the type as well as the size of organic temper. A coarser type of material results in a higher %AP for the same quan-tity of temper. The overall %AP, as well as the rate of increase is much lower for horse-dung than for dried grass. Each type of temper obviously creates a specific pore struc-ture which influences the %AP as well as the amount of oxygen available for the combustion of the temper and the oxidization of the clay itself in a given time span. Within one type of material, finer fractions will result in higher porosity than coarser fractions. The results prove that precise definitions of ‘coarse’ and ‘fine’ for specific types of temper are basic to any comparison and interpretation of fabric properties.
– The first and second series of tablets, set A and B, espe-cially those with hay temper, proved that the coiling tech-nique does indeed cause a specific arrangement and distribu-tion of organic fibres in the vessel wall. This technique has an influence on the amount and size (distribution) of the temper as well as on the pore structure, and through this on the resulting %AP.
– The influence of the amount and size of temper seems to be more important than the composition of the clay itself. More experiments are needed, however, to warrant this conclusion, as the clays used here are rather similar in their %AP curves for untempered tablets.
test-tablets hardly showed any sign of the temper. The series of test tablets also provide a good basis for comparing the measurements of the amount of organic temper and possibly on the type of temper in the pottery itself. Of all sets, the pore structure in the tablets of set C, with the specially prepared hay, was the most similar to that in the sherds from Uitgeest and Schagen. Hence, the %AP of the pottery should also show a good correspondence with those of set C and this set is used mainly for the comparison of quantities of temper and %AP in the pottery.
6.3 The type of organic temper
6.3.1 COMPARISON OF POTTERY AND TEST TABLETS The length and width of fibres and their distributions can be an indication of the type of material. In a recent experimen-tal study on vegetable temper, it was found that grass and hay cut at circa 5 mm resulted in an average length in the fired test tablets of 5.36-6.81 mm (Piena 1991, 9). The use of dung resulted in much shorter average lengths. The smaller the animal, the smaller the size of fibres in the dung. The average length of horse dung fibres in the experimental tablets was 2.32 mm and 1.53 mm for a horse fed on fresh grass and on hay respectively, when the dung was added in a fresh state. When the dung was dried first, the average size was much shorter for the horse feeding on fresh grass, less than 1 mm. The dried dung from the horse fed with hay, on the contrary, resulted in an average length of 1.5 mm5(Piena 1991).
Most of the core surfaces of sherds from Uitgeest and Scha-gen show the best visual fit with the tablets of set C with the specially prepared hay; the form and size of cavities and the size distribution are quite similar, although the tablets still contain coarser material than most sherds. The width of the fibres in the sherds varies, but is usually less than .5 mm. This width would match the size of fibres left by horse dung, as found by Piena (1991). But larger widths do occur in most sherds and often complete bits of stems (in a car-bonized state) still show the original round and hollow form (see fig. 6.9). In virtually all pottery from both sites fine organic dust <1 mm was present and fibres with a length of 1-3 mm were seen in all sherds, whereas the presence of fibres >3 mm varied. For both sites a small group of sherds containing more coarse temper show a better fit with the B 1 tablets (Uitgeest: n=15; Schagen n=11) and a few sherds are more comparable to the patterns caused by the finer horse dung. It is clear that sheep-dung and untreated horse dung was not used at all. The coarser particles of both types of dung leave forms and sizes of cavities that are quite different than those in the sherds.
Altogether, the evidence indicates that both types of temper, selected parts of horse dung and the finer bits of stems and leaves from plants, could have been used. If the results of
Piena (1991) are taken as a criterium, dung and more specif-ically horse dung seems to have been the most frequently used tempering material. However, these results were based on the average size of fibres and it is questionable whether this is a suitable criterium, considering the large variations found in the pottery and in the test-tablets. Moreover, varia-tions in the size of fragments also result from the preparation methods, as shown by Piena's as well as our own experi-ments. The test tablets demonstrate that, if stems and/or leaves of grasses of cultivated plants were used as temper, the material must have been treated in a different manner than was done here for set B 1 and 2. The resulting fibres are far too coarse, i.e. too broad and too long compared to most of the pottery. More likely the potters used ‘dust’ from hay lying around the settlement or they used the same method as for set C, by rubbing the stalks of hay. The good match between pottery and set C makes such a method a viable option.
6.3.2 BOTANICAL IDENTIFICATION
A few sherds (n=15) with well-preserved fibres were studied by prof. dr. C.C. Bakels and mr. W. Kuyper (FAL) in an attempt to determine the nature of the temper. This sample covers the variety of organic material in the total pottery assemblages, as far as shape, length and thickness are con-cerned and as far as variability in possible kinds of organic temper could be established from the carbonized remains. Again, the core surface proved helpful in determining this variation. Most of these fragments were identified as stems of graminae. No further specification could be given as to the kind of grasses or cultivated plants nor was it established whether these were a part of dung. The remains of cultivated plants (seeds, nodes and/or chaff) were also present in a number of sherds, though certainly not in all the pottery. Among them are Triticum (identification W. Kuyper), Hordeum and an impression of a (stekelnoot*), horse-bean and possibly reed (identification Dr. J. Buurman, ROB). Such rather sporadic remains always occur together with the ubiquitous fibres and they may have been incorporated by chance. Cultivated plant remains can however also be a part of dung6.
at 700 °C) show a distribution and pattern, that is very simi-lar to that of the horse dung used here.
6.4 The amount and size of organic temper in the pottery
The analyses consisted of measuring the amount of temper (%areal density, see below), the size of temper and the vol%, estimated by comparing the density of the temper in the sherds with that in the test-tablets of set C. For all mea-surements the core surfaces were used. The sample of Uit-geest consisted of 128 sherds from the vessels in sample I and an additional sample of 60 sherds, mainly from the creek-fill in trench 20/34 (chapter 3, fig. 7). The sample for Schagen consists of 97 vessels, including one vessel with shell-temper only. For a selection of the sherds from both sites the apparent porosity was measured.
6.4.1 METHOD OF QUANTIFICATION
As described above, the core surfaces created by sawing sherds through the core are the basis for measuring the amount and size of temper. The next step, finding a reliable and accu-rate way to quantify the amount of organic temper from these surfaces proved to be less simple. The most important aim was to establish relative differences in the amount of temper within a sample as the basis for defining significant and mean-ingful variations between different types of vessels. As the level of precision of the potters was probably rather low, the method is also a compromise between the amount of time involved and the accuracy of the count. Several methods were tried out in a pilot study for a subsample of sherds from Uit-geest7. The chosen procedure is the following (fig. 6.10): – In a piece of clear perspex a grid of 3 mm was engraved
within a 3 ≈ 3 cm square, thus creating a hundred cells; these were subdivided into squares of 1 mm. This piece of perspex was put on the cut surface of the sherd and both were placed under a binocular. The core surface area was minimally 9 cm2. For each 3 ≈ 3 mm cell, the presence or absence of fibres ≥ 1 mm was noted, aided by the scale-bar in the binocular and the 1 mm grid on the perspex. The actual number of such fibres within each cell was not counted. The standard enlargement used for the counts was 10≈, but whenever necessary a 20≈ magnification was used. Where possible, or necessary due to the specific distribution of the temper, the count was repeated over a different area. – The total number of squares containing cavities of organic
remains ≥1 mm was expressed as a percentage of the total number of cells and therefore ranges from 0 to 100%. Because not all fibres ≥ 1 mm within a cell are counted, this method is an estimation of the minimal surface area taken up by temper, the areal density (abbreviated to %AD). – For each sherd an estimation of the relative amounts of
organic remains was noted for three size classes: <1mm,
1-3 mm, >3 mm8. As a control of the counting method, the actual number of fibres ≥3 mm were counted within the area of 3 ≈ 3 cm.
the counting takes circa 15 minutes per sherd. Moreover, the accuracy of the counts is likely to be much higher than the accuracy used by the potters to measure clay and temper volumes, while the data on temper size provide a correction for the counting method.
6.4.2 QUANTITIES AND SIZES OF TEMPER IN THE POTTERY OF UITGEEST
% Areal Density
The frequency distribution of the %AD in fig. 6.4a shows a more or less unimodal distribution. There is no indication for the presence of distinct groups in the amount of temper. The average amount of temper is 45 %AD, the maximum fre-quency is also between 40-50 %AD. Percentages higher than 85 and lower than 10 hardly occur. There is a sharp increase in frequencies from 10 to 30 %AD, but a much more grad-ual decrease after 50 %AD. As a control on the composition, the data of sample I and the sample of sherds were split (table 6.2a). The average of %AD of the sample of sherds is slightly higher, but the frequency distribution for the two samples is very much the same. Whether this slight differ-ence is due to chronological variations is not clear9. The few closed contexts, such as the wells 18-1 and 31-1 suggest that this influence is negligible. The %AD varies from 10 to 90% in the pottery of well 18-1, but from 30 to 50% only in well 31-110, while the datings of the wells are similar. Within the group of possibly older wells (19-1; 7-1; 14-1), the %AD shows the same variation as the pottery from well 18-1 (see chapter 8.15 and 9 for further analyses).
The %AD was classified into 3 classes, 0-30, 30-60, >60 %AD (table 6.2), based on the frequency distribution and on the comparison with set C (see below). In view of the method, a variation of +/- 5 %AD in measurements should be regarded as a minimum error11. Most of the sherds would then contain 30-60 %AD. The distribution (fig. 6.2) also suggests that the potters had this one basic recipe in mind and varied the amount of temper around it. The relation with the types of clay are discussed in chapter 7 and those with pottery forms and function in chapter 9.
Size (distribution)
For all sherds the approximate amounts of size-fractions in the temper was noted. All sherds contained temper smaller than 1 mm and temper of 1-3 mm. The amount of ‘dust’ (<1 mm) in a sherd is usually quite large and seems to con-tribute to the dark colour of the core (see fig. 6.1 and 6.10). There is, however, a clear variation in the presence and amount of fibres larger than 3 mm, defined as ‘coarse’ tem-per. The actual number of such fibres was counted over an area of 3 ≈ 3 cm; the frequency distribution, which includes sherds with no coarse fibres present, is shown in fig. 6.4b. In 22% of the sample, (n= 42), no coarse material was present,
50% of the sherds contained 1-5 fibres >3 mm, and 28% more than 5. The average frequency of coarse temper is 5.4 for sample 1.
There is a clear relation between the presence of coarse material and the measured %AD (table 6.3). Sherds with little or no coarse temper mostly have a lower %AD and vice versa. Of the sherds with 30 %AD or more (n= 141) only 15% contained no coarse fibres, while of those with less than 30 %AD, 43% had no coarse temper. This relation is at least to some extent the result of the counting method itself, because coarser fibres will always be counted. As remarked above, this ‘skew’ is regarded as a correction because the amount of coarse temper does represent a larger area within the matrix.
6.4.3 QUANTITIES AND SIZES OF TEMPER IN THE POTTERY OF SCHAGEN
% Areal Density
The frequency distribution in fig. 6.4a of the %AD of tem-per in the Schagen sherds is basically the same as for Uit-geest, although the 5% intervals suggest a possible bimodal distribution. The distribution is skewed with the highest frequency in the class of 30-40 %AD; the majority of the sherds (60%) had 20-50% of temper. The average %AD is 36%. These values are clearly lower than for the pottery of Uitgeest, perhaps indicating a lower standard amount of temper used by the potters of Schagen. Based on this distrib-ution the %AD was classified into the classes
0-25, 25-50 and >50 %AD (table 6.4a). As all of the pottery belongs to one period of occupation a chronological influ-ence on the amount of temper is excluded for Schagen.
Size (distribution)
discussed in chapter 7 and 9. The size variation could also indicate the use of different kinds of temper, but more research is necessary before such a conclusion can be drawn.
6.5 The volume of temper (vol%) in the pottery The core surfaces of sherds were compared with those of the tablets to estimate the vol% of temper. The main purpose of this comparison was to check and compare the measured %AD with the known vol% of the tablets. If there is a high correlation between the two, the test tablets can be used as an even quicker way to estimate the amount of organic material in pottery.
6.5.1 METHOD OF COMPARISON
For all sherds from Uitgeest (n= 187) and Schagen (n= 97), the vol% of temper was estimated by visual comparison with the vol% in the test tablets, placing the reference set and the sherd under a strong light. The sherd was placed next to the set and assigned to the tablet(s) to which the amount of temper was most comparable, for example 10% or between 10 and 15%; the latter were coded as 12.5%. Only in diffi-cult cases a magnifying glass was used12. Although the visual comparison is difficult to check in mathematical terms, it provides a quick and reasonably reliable way of obtaining data on relative differences. Checks can be made by letting more persons carry out the comparison indepen-dently and compare the results. The differences in the assignments between, in this case, mr. F. Wiegmans and myself were restricted to a small number of sherds and were of a magnitude of 5 vol% at most. The use of a binocular was discarded as it obscured rather than enhanced the impressions of temper over the total surface of sherds and tablets. Good lighting is a precondition as is demonstrated by the set of photographs. The excellent lighting obtained by the photographer J. Pauptit (Faculty of Archaeology, Leiden) shows up every cavity in the tablets very clearly. Using the photographs even resulted in a slighter lower vol% than using the tablets themselves. The relative ranking of the sherds is not affected by this difference and both methods are therefore suitable.
As the photographs (fig.6.1) show, an obvious problem in estimating the vol% of temper in the sherds is the substan-tial difference in visual density for the same amount of temper between different temper materials. The tablets with dung clearly show far less fibrous cavities than those with hay (especially set B 1,2 and C) at the same vol%. The volume% observed in the sherds will therefore vary, depending on the set they are compared to. For this reason the sherds were first compared to the set with the best ‘fit’. For a small number of sherds of both sites the temper pat-terns are more comparable to those in set B 1 (fine grass, but ‘coarse in relation to set C) or B 3.2 (fine horse dung),
but the majority shows the best fit with tablets of set C. The data in the tables 6.6-6.9 and figures 6.8-9 are all based on the comparison with set C13
. As a check, the amount of temper in the tablets of set C was counted in the same manner as in the sherds (table 6.1). This proved to be more difficult, because the tablets contained no carbonized remains and the distinction between cavities left by temper or other factors, like preparation, was not as clear as in the sherds. Moreover, the fragmentation of the temper into fragments <1 mm increased considerably with increasing vol% in the test tablets and the fibres were often cut at an angle. This resulted in a %AD slightly lower than expected for the higher vol%; vice versa, fibres >3 mm are slightly overrepresented in the lower vol%. Set C is therefore still not the ‘ideal’ reference set, although it was sufficient for the explorative purposes of this study.
6.5.2 AREAL DENSITY AND VOL%IN THE SHERDS Uitgeest
The frequency distribution of the vol%, based on set C, in fig. 6.5 shows that the majority of the sherds were assigned to the range of 10-25 vol% (with an average of 15 vol%). The maximum vol% in the sherds is 30-35%, but more than 25 vol% occurred only sporadically. This is not surprising as the test tablets become increasingly brittle from circa 30 vol% upwards. The highest frequency occurs at circa 15 vol%. As the areal density varies from 0 to 98, the corre-spondence between the areal density (%AD) and the vol% of temper is approximately 3:1 for set C; in other words, an increase of 5 vol% equals an increase of 15 %AD. The most frequently measured %AD was 30-60% (with an average of 45%), which also is consistent with a ratio of 1:3. These ratios are of course averages and there obviously are excep-tions (table 6.6a), but they are supported by measurements of the %AD in the tablets of set C (table 6.1). The vol% was classified accordingly in three classes of 0-10, 10-20 and >20. In fig. 6.10a a series of sherds with increasing vol% are shown. As expected, there is also a clear correlation with the amount of coarse material (table 6.6b).
A small number of sherds from Uitgeest contained either quite a lot of coarse temper or only very fine temper (fig. 6.10b). Of the 15 sherds with coarse temper, 14 contained more than 40 %AD of temper corresponding to 15-30 vol%, when compared to set B 1. In the few sherds with extremely fine temper, the %AD is approximately twice as high as the vol%, when compared to set B 3.1. An increase of 5 vol% equals circa 10 %AD for these sherds.
Schagen
B Nr of fibres > 3 mm per 3x3 cm•0 2 4 6 8 10 12 14 16 50 40 30 20 10 0 Std. Dev = 3,73 Mean = 4 N = 187 % areal density of temper5 15 25 35 45 55 65 75 85 95
N 25 20 15 10 5 0 Std. Dev = 19,36 Mean = 45 N = 188 A
% areal density of temper0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 80 85 16 14 12 10 8 6 4 2 0 Std. Dev = 17,73 Mean = 35 N = 97 Temper > 3mm per 3x3 cm21 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 30 20 10 0 Std. Dev = 4,18 Mean = 4 N = 97 A B
Fig. 6.3 Uitgeest-Gr.D. Frequency distribution of (a) the amount of temper (%AD) and (b) the amount of coarse fibres in the pottery of sample 1 and an additional sample of sherds.
Fig. 6.4 Schagen-M1. Frequency distribution of (a) the amount of temper (%AD) and (b) the amount of coarse fibres.
N
N
B % AP29 - 3131 - 3333 - 3535 - 3737 - 3939 - 4141 - 4343 - 4545 - 47 30 20 10 0 Std. Dev = 3,27 Mean = 37 N = 88 A Std. Dev = 6.75 Mean = 14.9 Vol.% N = 188 40 35 30 25 20 15 10 5 0 37.5 32.5 27.5 22.5 17.5 12.5 7.5 2.5 B A
Vol. % of temper, based on test tablets0 5 10 15 20 25 30 35 30 20 10 0 Std. Dev = 5,95 Mean = 11 N = 97 % AP26 28 30 32 34 36 38 40 42 44 46 48 16 14 12 10 8 6 4 2 0 Std. Dev = 4,04 Mean = 39 N = 57
Fig. 6.5 Uitgeest-Gr.D. Frequency distribution of (a) the vol% of temper based on comparison with set C and (b) the %AP.
Fig. 6.6 Schagen-M1. Frequency distribution of (a) the vol% of temper based on comparison with set C and (b) the %AP).
N N
majority of the sherds have less than 15 vol% of temper. As the %AD varied from 0-85% with a maximum between 30 and 40%, the ratio of the vol% and the %AD also is approximately 10:30 in the Schagen sample. The vol% were classified in accordance with that of the %AD into 3 classes and there is a high correlation between the two variables (table 6.7a). The same holds true for the vol% and the amount of coarse temper (table 6.7b). The cluster of cases with a %AD >50, as well as a vol% >15 contains 11 sherds with very coarse temper, comparable to that in the test tablets of set B 1 (fig. 6.9). These sherds suggest that a second recipe with a much higher quantity of temper and/or temper of a different kind may have been used. The sample is, however, too small to permit firm conclusions.
6.6 Apparent Porosity measurements (%AP) The %AP is an indirect measurement for the total fabric composition. The method for measuring the %AP was described in chapter 4.2.1. The %AP in sherds was measured to establish if there is a correlation with the amount and/or size of temper, which could point to manipulation by the potters of the total fabric properties (see chapter 2.5). The data are discussed in this chapter because, as the test tablets showed, the effect of the organic temper, its quantity and size, on the porosity of sherds should be clearly noticeable. In all series, there is a clear increase in the %AP with increasing vol% of temper. At the same time, the type and size of temper resulted in variations for the vol%; a finer sized temper resulted in a slightly higher %AP, but horse dung resulted in a much lower %AP than hay. These relations between vol% and %AP in the test tablets were used as an independent reference for the expected %AP in the sherds. In doing so, two impor-tant differences between the test series and the sherds had to be largely ignored. Firstly, the test tablets were fired at 700-750 °C, while the firing temperature of the pottery is sup-posed to have been around 850-900 °C (chapter 5). Secondly, the influence of the clay itself is not taken into account here (but see chapter 7). Although clay 64, used for most series, is quite common in the pottery of both sites, the %AP is clearly higher at the expected firing temperatures (compare fig. 4.2). On the other hand, this difference should not affect the rela-tive influence of the amounts of temper on the %AP; the range and increase in the %AP in the test tablets can therefore be a useful indication for the expected %AP in the pottery in relation to the amounts or volumes of temper. As the temper in most of the sherds from both sites is slightly finer than that of set C, but coarser than that of set B 3.1, it was also expected that the %AP would be slightly lower than that of set C and higher than that of set B 3.1 for corresponding amounts or volumes of temper.
The porosity was measured in 90 sherds from vessels of Uitgeest and 57 sherds from vessels of Schagen, which is
70% of sample I from Uitgeest and 58% of the sample from Schagen. The reason not to measure the %AP for all pottery from both sites was the limited result of the first series.
6.6.1 RESULTS UITGEEST
The %AP values vary from 29.7-45.2%, a range of circa 15% (fig. 6.6b). The average %AP is 36.6%, but more than half the sherds have a lower %AP. The distribution in fig. 6.8a,b shows that there is no straightforward relationship between the amount of temper and the apparent porosity. Although there is indeed a tendency for the %AP to increase with higher %AD, a substantial number of sherds with 0-60 %AD (comparable to 0-20 vol%) fall within the same range of circa 34-39 %AP. The average values of the %AP for each class of the %AD are 34.9, 35.9 and 39.8 respectively (table 6.8a). The result suggests that the increase in the amount of temper from 0-60% areal density is causing only slight differences in the %AP in the pottery, while higher amounts result in a considerable increase in the %AP. Underlying this relationship is the influence of the amount of coarse temper, as is clear in fig. 6.8b: sherds with high amounts of temper also contain higher amounts of coarse material and the combination results in a higher apparent porosity. The average %AP for sherds without any fibres >3 mm is 35.1, for those with 1-5 and more than 5 of such fibres it is 37% and 37.4 respectively. This outcome corre-sponds well with that of the test tablets.
In comparison with the test tablets with 0-30 vol%, set C, the range in the %AP in the sherds is larger, the minimum values are clearly lower, but the maximum value
corresponds to that of the tablet with 30 vol%. Despite the restrictions mentioned above, the vol% in the test tablets is used to classify the %AP at 10 vol% and 20 vol% (being 36.5 and 39.5 respectively; fig. 6.2), based on the 3 : 1 correspondence between the %AD and the estimated vol% in the sherds. The relation between the three classes of %AD and the %AP in this classification is shown in table 6.8b. Although there clearly is a trend for the %AP to increase with higher amounts of temper, this relation could not be tested statistically, using X2
values for the sherds with more than 60 %AD. That this correlation is influenced by the temper size, is quite clear in table 6.8d; there is a significant correlation between these variables. Sherds with higher amounts of coarse material tend to have a higher %AP and vice versa. These data seem to contradict the results of the test tablets, where finer tem-per results in a higher porosity, but there are two factors which can explain this difference. The first is that higher quantities of material, whether coarse or fine, also increase the porosity through increasing the interconnection between the pores; the second factor is that the amount of coarse temper is quite low in most sherds, the majority containing only 1-5 fibres ≥3mm.
Altogether, it is clear that there is indeed a mutual relation-ship between the size, the amount or volume% of temper with the %AP, similar to the properties in the test tablets, even though these relations are far from straightforward. There is a far greater variation in %AP in relation to the amount of temper than observed in the test tablets. Clearly other factors than the quantities and size of temper played a role. These will be discussed in chapter 7.
6.6.2 RESULTS SCHAGEN
The %AP in 57 sherds out of the fabric sample of Schagen ranges from 26.1% to 50%, an extreme variation even com-pared to the pottery from Uitgeest. These differences cannot be due to the actual measurements as sherds from Schagen and Uitgeest were mixed during this process. The average %AP of most of the Schagen pottery is 38.7, much higher than that of Uitgeest14
. The average %AP is 36.5, 39.1 and 40 respectively for each class of the %AD (0-25, 25-50, >50). The influence of coarse temper also appears to be greater than for Uitgeest. The average %AP in sherds with no coarse material is 35%, against 39.5% and 39.4% for pottery with 1-5 and > 5 coarse fibres. Both averages are however based on a very wide range in values within each class. The results indicate that there is no relationship between the amount and size of temper and the %AP, which is also apparent in the distribution in fig. 6.9. This distribu-tion suggests that there are two groups of sherds, one cluster with a very high %AP and all other onesfalling within a lower %AP range. The %AP even seems to decrease in the cases with a high amount of temper.
Likewise, there is no correspondence between the porosity in the pottery with that in test tablets. For nearly half the sherds the %AP is higher than 39.5% (equalling 20 vol%) and no less than 72% has an %AP higher than 36.5 (10 vol%), while only 4 sherds contained more than 60 %AD15. Because of the low number of measurements, the %AP was also classified into two classes, adapted to the distribution in fig. 6.7b (table 6.9a,b). There is no correlation with the %AD or with the vol%. The %AP is in general much higher
than expected on the basis of set C and corresponds better to the values for set B 1, tempered with coarsely cut hay (see fig. 6.1). For this set a different clay— type 63, also from Schagen—was used with a higher porosity of its own. When this difference in the %AP of the clays is accounted for, the %AP values for the temper of set B1 and set C is minimal (see chapter 7).
Altogether, the results are not easy to interpret, but it must be concluded that other factors had a greater influence on the porosity than the amount or size of temper. Several possibilities will be explored in the next chapter.
6.7 Discussion
The results of the temper analyses suggest that the potters from both sites used at least one standard recipe for the amount of temper to be added to the clay. The definition of the standard amount of 30-60% AD for Uitgeest and 25-50 %AD for Schagen, is based on the frequency distribu-tions. In both sites, more than 50% of the pottery contained the standard amount. As expected, there are no clear or strict limits detectable in the amount of temper within each sample, independent of the level of precision of measurements. The frequencies show a slightly skewed normal distribution, higher amounts of temper occurring more frequently than lower amounts. This could indicate that there is a second recipe with higher amounts of temper and/or that a different, coarser type of temper was used for some pottery. The latter interpretation is based on the fact that there is a clear correla-tion between the amounts of temper and of coarse temper (≥3 mm). This is partly the result of the method itself. By using 3 ≈ 3 mm as the basic observation unit, fibres that were larger than 3 mm (in length) added more to the counts than smaller ones. Although this effect was not foreseen when the method was designed, it turned out to be a very useful correction of the method. In virtually all sherds, with a few exceptions in the sample from Schagen, the temper consisted mainly of fragments ≥3 mm, including very fine material (<1 mm). By counting the exact amount of fibres ≥3 mm in the standard surface area of 3 ≈ 3 cm, the mea-surements of the % areal density proved to be more represen-tative of the ‘real’ density of temper.
Vol.% > 20 10 - 20 -10 % AD 10 20 30 40 50 60 70 80 90 100 0 % AP 46 44 42 40 38 36 34 32 30 28 % AD 20 40 60 80 100 0 % AP 46 44 42 40 38 36 34 32 30 28 90 70 50 30 10 N fibres > 3mm 4 = >10 3 = 5 - 10 2 = 1 - 5 1 = 0
Fig. 6.7a,b Uitgeest-Gr.D. Distribution of the %AP and the %AD in 90 sherds, classified by (a) the vol% of temper and (b) the amount of coarse temper (fibres >3 mm)
Fig. 6.7 Uitgeest-Gr.D. Distribution of the %AP in relation to the %AD, the vol% and the amount of coarse temper (fibres >3 mm)
A
Fig. 6.7c The variation in %AP for each class of the vol% and the %AD of temper Fig. 6.7d As fig. 6.8c, but sherds with very fine or very coarse temper are excluded
N fibres >3 mm > 10 5 - 10 1 - 5 0 % AD
The number at each case refers to the amount of coarse temper (>3mm)
100 80 60 40 20 0 4 16 2 5 4 1 5 8 2 0 12 6 0 11 15 3 5 8 12 0 11 0 0 13 5 7 4 4 1 7 7 4 11 2 24 3 4 3 7 5 0 6 0 3 4 4 0 2 3 6 12 12 8 0 11 0 % AP 50 45 40 35 30 25 90 70 50 30 10 Vol.% > 15 7.5 - 15 - 7.5 %AD 90 80 70 60 50 40 30 20 10 0 % AP 50 45 40 35 30 25
Fig. 6.8a,b Schagen-M1. Distribution of the %AP and the %AD, classified by (a) the vol% of temper and (b) by the amount of coarse temper. Fig. 6.8 Schagen-M1. Distribution of the %AP in relation to the %AD, the vol% and the amount of coarse temper (fibres >3 mm).
A
Vol.% 11 8 1 22 5 1 2 7 N = >15 7.5-15 -7.5 % A P 50 45 40 35 30 25 %AD 0-25 25-50 >50 7
Fig. 6.8c Schagen-M1. The variation in %AP for each class of the vol% and the %AD of temper.
of the tablets of set C and to a lesser extent set B 3.1 were quite similar to those of the sherds, they could be used to estimate the volume % of temper in the latter. These esti-mates provided a ‘check’ on the counting method. The good match between the two methods makes it possible to use the test sets as a very quick way to estimate the temper vol% for any pottery complex with vegetable temper, assuming that the same tempers were used as in the present cases. Ideally, comparison should be based on the best matching test set (see fig. 6.10).
Most likely, the remains in the sherds consist of parts of graminae, but further specification of the specific type of material proved more difficult. The fibres could be from either the stems and leaves of wild grasses or cultivated plants. In both cases the material must have been prepared to obtain the specific size ranges observed in the pottery. The variable presence of coarse material may indicate that in some cases, especially in the Schagen sample, a different type of temper or a different method of preparation was used. The latter seems more likely at present. As the tablets of set B 1 and 2 show, there is quite a large variation in size even when the material is cut into one standard size under controlled conditions. It is also possible that dung from cattle or horses was used as the source material, although the test tablets suggest that this results in a different, finer type of remains. More test series need to be produced before the distinction between dung and plants can be defined properly.
Finally, it must be concluded that the results of the %AP measurements are rather disappointing, especially for Scha-gen. The data for Uitgeest did to some extent confirm the influence of the amount and/or size of temper on the %AP, but mainly for sherds with high amounts of temper, >50 %AD or more than 15 vol%. An increase in coarse material seems to have a much greater effect than the amount itself, although this effect is different, even contradictory, in both samples; in the sherds from Schagen, the coarser material tends to lower the porosity, while for Uitgeest the opposite trend is observed. This difference is not easy to explain; one possible cause is the amount of very fine organic ‘dust’ which may enhance the connections between pores and thus the %AP. Another factor may be the irregular distribution of the temper through a vessel wall and even within a sherd. For seven vessels from Uitgeest and Schagen the %AP was measured in two sherds from different parts of the vessel wall, for example a lower wall and a rim sherd. The differ-ences between the two sherds is minimally 0.6% and maxi-mally 2.6%, which equals 5-10 vol% of temper in the test tablets of set C. The vessels showing a high variation in %AP also showed a high variation in the amount of temper within and between sherds.
Fig. 6.9 Uitgeest-Gr.D. Examples of estimated vol% of temper. Scale 150%
Left and middle column: Sherds with increasing %AD and vol% of temper compared with the vol% in the test-tablets of set C.
composition intervene in the relationship between the amount of temper and the %AP, especially for the pottery of Schagen.
6.8 Firing methods
The influence of the firing techniques on the properties and characteristics of the fired product have been discussed at several instances in chapters 4, 5, and 6. In the following, the observations on the main variables involved, (a) the maximum temperature, (b) the rate of temperature increase and (c) the atmosphere around the pottery, the amount of available oxygen and (d) the duration of firing, are presented for the pottery of both sites. The primary distinction in methods is that between the use of a kiln or of a bonfire. In the latter, the control over the variables (a) to (d)d is much lower than in kilns. In general, firing in bonfires will result in more variation in the degree of oxidation across the vessel wall. As the test tablets with temper clearly demonstrate, the presence of organic material in the fabric is an important intervening factor in this respect (see paragraph 2). The evidence from the test clays, the chemical analyses, and the refiring of sherds together pointed to a restricted range of clay types being used for the pottery and to incomplete oxidation as a standard method. Although the organic mater-ial is burnt out or carbonized, the core of the pottery always still is reduced.
Evidence for bonfires
A strong indication for the use of open fires is the often large colour variation within the surfaces of one vessel, which is typical for the pottery of both sites. These differ-ences are caused by variations in the available oxygen as well as in temperature. In bonfires, the pottery is at least partially covered with or surrounded by fuel and the fire should be smouldering to avoid misfires. Because of these conditions, the atmosphere tends to be more or less neu-tral, non- to slightly oxidizing, but the amount of oxygen and the temperatures can fluctuate throughout the fire. In most sites from the Roman period in North Holland pottery slag is present, usually scattered among other occupation debris. Although oven remains are known from the Iron Age and Roman period sites in the coastal regions, there is as yet no conclusive evidence that these were used to fire pottery16.
The following observations were recorded.
– The colour(s) of the exterior and interior surfaces – The thickness of the ‘oxidized’ layer on the surfaces.
Signs of (secondary) burning and/or overheating were also noted.
– The degree to which the organic temper in the core was carbonized or combusted, for the sample of Uitgeest only.
– For a sample of 15 sherds from Uitgeest the original firing temperatures were estimated from a dilatometric graph.
The colour(s) was (were) used as an indication for the degree of oxidation that took place during firing. For two reasons, the descriptions are not based on Munsell codes, but on relative differences and similarities within the pottery. Firstly, the human eye can see more variation in colours than can be detected with the Munsell charts, but, more impor-tant, it would be impossible and useless to express the large colour variations which occurred within any one vessel by Munsell codes. Instead, the suggestions made by Shepard (1963, 213-224) are followed here.
6.8.1 RESULTS
Firing atmosphere
The results are discussed for both sites together as there is little difference between them. Five categories were distin-guished for the firing atmosphere and the resulting degree of oxidation (table 6.10):
1 Reduced: black
2 Neutral to incompletely oxidized: – neutral: mainly grey or grey-brown
– neutral to slightly oxidized: mainly (light)brown and buff, with some yellow and red patches
– incompletely oxidized: (large) parts of the vessels have a pale yellow to light orange colour
3 ‘Oxidized’: most of the vessels surface is yellow to light orange, varying to white or orange/red.
The term ‘oxidation’ is used here in a relative sense and only for the surfaces. No truly oxidized pottery was present in either of the settlements; the cores are always still reduced and the colours are never as bright as those of the refired sherds and the test tablets. Examples are shown in fig. 5.1. Furthermore, the classification is based almost exclusively on the evidence from the exterior surfaces. Only occasionally, the interior surfaces are other than a pale to dark grey colour, although the inside of the rims often did show some sign of oxidation. The colours of the base sur-faces varies.
Conform the evidence from other sites in the (western) Netherlands, the pottery from Schagen and Uitgeest was usually fired in neutral to slightly oxidizing conditions, whereas a small group of well- reduced ware was present as well.
Group 2: Neutral to partially ‘oxidized’
and from light to dark. The slightly more oxidized parts of the same vessels show brighter colours, from a soft yellow to areas with a pale orange to red colour or, alternatively, brighter yellow and yellowish white spots occur. None of the colours are really bright. The ‘oxidized’ layer is restricted in all cases to a few (2-3) mm of the outer surface, never exceeding 4 mm. The core is always still reduced and varies in colour from grey to black. Together these data suggest that the duration of firing was in general rather short. The oxidation on the interior surface is mainly determined by the way the vessels were stacked in the fire. The degree of oxidation of the interior in the vessels studied here varies. In most cases the rim is slightly oxidized but the rest of the wall is not. The colours vary from (dark) grey to a grey brown. Although data on the way of stacking have not been collected systematically, the colouring suggests that vessels were usually stacked and/or placed upside down in the fire. A small group of vessels in both sites has a slightly more oxidized inner surface and these may have been fired stand-ing up and/or not stacked.
Group 3: ‘Oxidized’ fabrics
A low number of vessels from both sites show a more intense oxidation of the exterior surface. In the sample of Schagen, the percentage of such fabrics are higher than for Uitgeest. The difference with group 2 is one of degree rather than kind. The surfaces still show variations in colours, but sometimes approach the colour-intensity of the test clays fired at 750-850°C17. The thickness of the oxidized layer is not significantly different from that of vessels in group 2, while there also is no clear relation with the degree of oxida-tion of the interior surfaces. In five cases, all vessels from the well 18-1 in Uitgeest, the oxidation is due to (secondary) burning. Three of these showed signs of melting. In the sample of Schagen, the more oxidized pottery consists mainly of the vessels that were used to cover the cremations (fig. 2.8) and several had small patches of secondary burn-ing. Of four more vessels, classified as oxidized, two stem from a hearth, two from a pit fill. It can therefore be con-cluded that complete oxidation never took place during the original firing process, except for possible accidents during (or after) firing. A second conclusion is that the duration of firing appeared to be more or less the same for all vessels in group 2 and 3. The possible relationship with the amount of temper is discussed in chapter seven.
Group 1: Pottery fired in a reducing atmosphere
A small but consistently present group of pottery from Uit-geest and Schagen has been fired under reducing conditions, 17 and 20% respectively. All of this ware has highly pol-ished exterior surfaces. The reduced ware is easily distin-guished from the rest of the pottery. The exterior surface is
black with a shiny gloss, the interior is a dull dark-grey to black, with a few exceptions from Schagen18. The intensity of the gloss depends partly on the state of preservation; abraded or weathered pottery in this class usually lost the shiny gloss.
It is ascertained beyond doubt that all of this pottery has first been fired in the ‘normal’ neutral to (slightly) oxidizing atmosphere just like the majority of the pottery. The reduc-tion took place only in the final stage of firing and both on the inside and outside surfaces. This ‘last minute’ reduction is indicated by remnants of the slightly oxidized layer just below the reduced surface, which were present in all cases. Obviously the reduction process was applied for only a short time and never long enough to completely replace the oxidated surface completely. The (original) oxidized layer has moreover the usual thickness of group 2 and 3. The reduction was most likely achieved by ‘smudging’. This way of producing a reduced and shiny ware is most often mentioned in the ethnographic descriptions (for example Hally 1983, Rye 1981, Saraswati & Behura 1966; Shepard 1963) and is still used in the modern ceramic industry (Keramiek 1973): towards the end of the firing process a reducing atmosphere is created by adding damp fuel (like dung) to the smouldering fire and/or cover the pottery with damp material to close off the oxygen supply. Damp fuel also results in smoke, that is in free carbon, which will adhere to the surfaces where these molecules will use the oxygen present in the vessel wall, thus adding to the reduc-tion. The carbon intensified the gloss, which was already present on the polished surfaces19
.
Clearly, the potters were first of all concerned with the ‘looks’ of this pottery. The short reduction time at the end of a ‘normal’ firing process can partly be explained by techno-logical conditions. As the organic matter in the fabric will use all available oxygen first in a reducing atmosphere, the reduced Fe in the clay can act as a violent flux through the formation of ferri-silicates in the surface layers. These can close off the pores and prevent the CO2formed by the tem-per to escape, which results in ‘bloating’; the vessel wall is literally being blown up. Such fluxing takes place at temper-atures from circa 900°C upwards (Keramiek 1973). A short reduction period after oxidation at lower temperatures will prevent this. Moreover the original firing temperature was probably lower than 900 °C.
measured temperature also can be expected to vary for different parts from one vessel. Only the testing of large samples can provide more or less reliable information about the common range of the T-max.
Some ethnographic evidence on the T-max and duration of firing was presented in chapter 2.4. In general, a temperature of 700-800 °C is considered high enough to produce durable pottery21
, but higher temperatures can easily be achieved in open fires and are also mentioned for pottery from ethno-graphic and archaeological contexts (e.g., Nicholson 1989). The test tablets (fig. 4.1) show that at temperatures up to 850 °C the degree to which iron and other minerals such as Ca, Na, Mg, and Mn are recrystallized into brightly coloured minerals, is quite low. Compared to these the colours of the sherds are even less bright. The colours of a small sample of sherds from Uitgeest, refired at 850 °C, also differed from the original vessels. Dilatometric measurements were carried out for a sample of 15 sherds from Uitgeest. The reconstructed T-max varies from 800-950 °C, but was around 850 °C for most sherds. Unfortunately these results cannot be trusted and are probably too high22. Taken together, the evidence indi-cates that the maximum firing temperatures were probably 800-850 °C or lower. This conclusion is indirectly supported by the changes that took place in the refired sherds; at 950 °C the Ca-rich inclusions had changed in size and quantities, indicating that chemical changes had taken place, changes that have not been seen in the pottery as excavated.
The duration of the firing can only be estimated in conjunc-tion with the data on temper; the combusconjunc-tion of organic material will use the available oxygen. In virtually all pottery from Uitgeest and Schagen the temper is burnt out of the surfaces. The degree of carbonization and combustion of the temper in the core of the sherds varies23. As a control on this relationship, the degree of combustion of the organic material in the core of the sherds in sample 1, Uitgeest, was examined as well. In the majority of sherds, carbonized remains were still present. Vessels that showed mainly burnt out fibres usually showed a higher degree of oxidation as well. This correlation is independent from the amount of temper and suggests that these vessels were fired in more oxidizing cir-cumstances or for a longer time. As the thickness of the oxidized layer on the surface is not different from the other vessels, the latter is unlikely to be the main factor.
6.8.3 CONCLUSION
The pottery from Schagen and Uitgeest was fired in open fires. The similarity between the data from both sites suggest that the tradition of a specific firing method was continued from the late Iron Age up to the third century AD. The dura-tion of the firing process must have been relatively short and the supply of oxygen relatively low, as in most cases only the outer surface is oxidized to some, but always incomplete
degree. The atmosphere in the fire can best be described as neutral to slightly oxidizing with variations in the amount of oxygen within the fire. The degree of ‘oxidation’ of the exte-rior surface varies for each vessel. A small group of vessels from each site was possibly fired in a more oxidizing atmos-phere. This pottery consists of two types of fabrics, one with an orange to red colour and one with a yellowish to white colour. This could indicate a difference in clay types, with different Ca : Fe ratios. However, as discussed above (chapter 5.4), it cannot be excluded at present that these differences are also related to postdepositional processes of leaching and infiltration of both elements. Further research is needed to sort out the many possible explanations.
No firm data are available for the maximum firing tempera-tures, but all circumstantial evidence suggests that this was not higher than circa 850-900 °C. The pale colours point to the incomplete recrystallization of iron and calcium com-pounds, as do the differences between the original and refired sherds. The incomplete oxidation is related also to the presence of organic temper. As the organic material was burnt out of the outer layers of the vessel and, to varying degrees, also from the cores, the firing was probably stopped shortly after some oxidation of the exterior surfaces, but before any oxidation in the core had taken place. If this interpretation is correct, the oxidation of the colouring agents, Fe and the Ca : Fe ratio, would have been incom-plete, whatever the maximum firing temperature was. In a small number of firings, the potters chose that very moment to start the reduction process. The reduction was limited to a short period, just long enough to obtain the reduction, the shine, of the outer layers of the surfaces. It is quite clear that reduction was applied purposefully and for a special group of pottery. The combination of highly polished surfaces and reduction was meant to express the specific meaning of this group of pottery within the total assemblage and is also characterized by specific forms. These forms however underwent a change during the Roman period. In Uitgeest the pedestalled bowls seem to have been a part of the ‘standard’ household inventory, but nonetheless a special vessel. Although, mostly broken, pedestalled bowls still occurred in the 3d C. AD in Schagen, larger and different forms of reduced ware were found more frequently in con-nection with ritual deposition of pottery in pits (for more detail, see chapter 8.13).
