Results

Examination of the 55 malting complex samples revealed a cereal-rich assemblage, dominated (unusually for Anglo-Saxon England) by rye grains (Secale cereale L.) (64%, n = 21,814), and secondarily bread wheat (Triticum aestivum L.) (28%, n = 9,518). Seven per cent (n = 2,363) of grains from the assemblage are hulled barley, likely 6-row (Hordeum vulgare ssp. vulgare) (Caroe 2022a). All three crops occur near ubiquitously in each sample. In total, 61,062 plant items were identified, belonging to 99 taxa/types. Abundances and ubiquity of key taxa are summarised in Table 3.

Plant item | Samples where present | Max. items per sample | Sum of items
no. | %
Cereal grains |  |  |  | 
Rye | 55 | 100 | 1,391 | 21,814
Free-threshing wheat | 55 | 100 | 997 | 9,518
Hulled barley | 53 | 96.4 | 267 | 2,363
Oat | 25 | 46 | 40 | 321
Chaff |  |  |  | 
Rye rachis | 22 | 40 | 125 | 699
Bread wheat rachis | 12 | 22 | 32 | 124
Hulled barley rachis | 15 | 27 | 48 | 204
Weedy/wild oat floret base | 2 | 3.6 | 8 | 12
Weedy/wild seeds |  |  |  | 
Weedy/wild taxa total | 55 | 100 | 544 | 9,446
Brome grass | 54 | 98 | 237 | 2,766
Corncockle | 42 | 76.4 | 120 | 1,183
Black bindweed | 36 | 65.5 | 480 | 1,791

The overall proportion of malting complex grains clearly showing signs of germination under standard light microscopy is 17% (n = 1,602 of 9,662 assessed); this increases to 46% (n = 4,450 of 9,662) following proportional reassignment of grains whose germination status cannot be determined (Caroe 2022a). Only one sample from the malting complex lacked germinated grains. A comparison by way of a ‘control’ with four similarly examined samples from the settlement part of the site, where malting is not believed to have been practiced, revealed that no grains from the settlement show clear evidence for germination (44% had indeterminate germination status).

Functional weed ecology

Intensity of cultivation

Figure 3a displays the weed ecology intensity model (Bogaard et al. 2016). Each smaller symbol represents a single modern arable plot (from either Haute Provence or Asturias), coded according to the type of agricultural regime it represents and plotted along the discriminant function extracted to distinguish between the two groups of high- and low-intensity farming systems (centroids – average scores for each group – are shown). Discriminant function values are calculated using the functional weed trait values per arable plot, entered as discriminating variables in the model, as shown in Fig. 3e. Figure 3b plots the scores of the 26 Sedgeford units, along the same discriminant function. All units cluster at the negative end of the axis and each is classed by discriminant analysis as low intensity. These observations clearly attest that the agricultural regime(s) from which Sedgeford’s assemblage originates was one of low fertility, implying low labour inputs per unit area.

Figure 3c shows no apparent trend when the datapoints are coded by feature within the malting complex. Similarly, there does not seem to be any relationship between the date of the episodes and the discriminant function when datapoints are coded by approximate date range; in other words, there is no evidence for change in level of labour input over time (Fig. 3d).

Soil disturbance levels

Figure 4a displays the disturbance model (Bogaard et al. 2022). Each smaller symbol represents a single modern arable plot (at either Laxton or Highgrove), coded according to the type of agricultural regime it represents and plotted along the discriminant function extracted to distinguish between high- and low-disturbance growing conditions (centroids are shown). Figure 4f displays the weed functional trait data used as discriminating variables and their relationship to the discriminant function. Figure 4b classifies the Sedgeford data based on the above model (Fig. 4a). Datapoints are clustered at the positive end of the axis, with discriminant analysis classifying all episodes as relatively high disturbance. Almost 90% of discriminant scores generated by our disturbance discriminant analysis exceed the ‘Lorsch baseline’, i.e. the lowest level of disturbance expected with use of a mouldboard plough (Fig. 4b).

According to Fig. 4d, there is little indication of a relationship between disturbance and area of the malting complex from which units derive. There is some suggestion (Fig. 4e) of a shift from a more to less soil-disturbed agricultural regime over time at Sedgeford, but the number of later-dated units is minimal.

Stable isotope analysis

The δ13CVPBD and δ15NAIR values for each sample are presented in ESM 2. Associated %C, %N values and C:N ratios are also there detailed.

Carbon stable isotope values

Figure 5 displays δ13C values for all samples, grouped by cereal taxon. The mean δ13C value for Sedgeford’s (six-row) barley of − 22.84 ± 1.11‰ is ~ 2‰ lower than that for both rye and wheat (Fig. 6; Table 4), a significant difference (F (2, 108) = 41.62, p = < 0.001) (ESM 1 Table S4). The expected − 2‰ offset in δ13C values for six-row barley if all taxa are cultivated in equal water conditions is thus observed. Adding 2‰ to the δ13C value for barley grains results in no significant difference between means (F (2, 108) = 0.20, p = 0.82) (ESM 1 Table S4).

Fig. 5: δ13C values for all samples, grouped by cereal taxon and coded by feature within the malting complex

Fig. 6: Mean δ13C values with associated SDs for all samples by cereal taxon. The mean for barley after compensating for expected 2‰ offset is shown

Feature | Taxon | Number of (single grain) samples | Mean (‰) | Standard deviation (‰)
All | Rye | 39 | − 21.03 | 1.00
Free threshing wheat | 39 | − 20.99 | 0.85
Barley | 33 | − 22.84 | 1.11
Kiln 1a | All | 23 | − 21.07 | 0.69
Kiln 2 | 30 | − 20.62 | 1.12
Kiln 3 | 29 | − 21.08 | 1.04
Steeping tank | 30 | − 21.06 | 0.94
Kiln 1 | Rye | 9 | − 20.92 | 0.82
Free threshing wheat | 9 | − 20.94 | 0.52
Barley | 5 | − 23.56 | 0.56
Kiln 2 | Rye | 10 | − 20.93 | 1.01
Free threshing wheat | 10 | − 20.56 | 0.84
Barley | 10 | − 22.38 | 1.47
Kiln 3 | Rye | 10 | − 20.97 | 1.23
Free threshing wheat | 10 | − 21.24 | 0.94
Barley | 9 | − 23.03 | 1.02
Steeping tank | Rye | 10 | − 21.19 | 1.03
Free threshing wheat | 10 | − 21.12 | 0.96
Barley | 9 | − 22.84 | 0.88

δ13C values for rye and wheat have similar ranges (3.97‰ and 3.47‰ respectively) (Fig. 5). Excluding one outlier (a grain from context 19061: value − 18.66‰), the range for barley is 3.02‰. Stroud’s (2023) research found that the δ13C value SD for single grains from one modern field was ± 0.33‰, with a maximum range of 1.24‰: figures indicating expected variability in water availability within a single agricultural regime. SDs and ranges for all taxa exceed these thresholds, i.e. data are consistent with each crop deriving from a range of water availabilities (Figs. 5 and 6; Table 4). However, heterogeneity in other environmental conditions (e.g. soil type, canopy cover, light or topography) may also contribute to this variability.

Statistical comparison of crop means within each feature reveal, once 2‰ is added to δ13C values for barleys to compensate for the physiological difference, no significant disparities between taxa (ESM 1 Table S4). All SDs for grains of each cereal within each feature exceed the ± 0.33‰ threshold, i.e. variability in values for grains within each feature are consistent with cultivation in different water availability conditions (Nitsch et al. 2015; Stroud 2023).

No clear trends are apparent in the distribution of datapoints by feature within the malting complex (Fig. 5). This is also evident in Table 4 and Fig. 7, which show minimal difference in mean δ13C values between the features (where all taxa are combined), and this is confirmed statistically (F (3, 107) = 0.63, p = 0.60) (ESM 1 Table S5). After compensating for the barley offset, SDs for each feature exceed the ± 0.33‰ threshold, suggesting that grains between the features were cultivated in varying water conditions.

Fig. 7: Mean δ13C values with associated SDs for all samples by feature within the malting complex. SDs are calculated after compensating 2‰ for the barley offset

Nitrogen stable isotope values

Figure 8 displays δ15N values for all samples, grouped by cereal taxon. The data exhibit considerable variability, with δ15N values between 0.06 and 12.66‰; SDs are ± 2.46‰ (rye), ± 1.79‰ (wheat), and ± 2.07‰ (barley).

Fig. 8: δ15N values for all samples, grouped by cereal taxon and coded by feature within the malting complex. The mean δ15N value for deer, minus 4‰ for the trophic shift (i.e. the wild herbivore baseline) is shown as a dashed line, with the expected grain/rachis offset shaded

Heterogeneity in manuring conditions can exist within a single field. However, modern crop studies using single grain samples suggest, for grains from a single arable experimental plot receiving ~ 25 tonnes of manure per hectare, a maximum expected range in δ15N of ~ 5.40‰, and maximum SD ~ ± 1.64‰ (Larsson et al. 2019). A comparable, unmanured plot had a maximum range of ~ 2.10‰, and maximum SD of ~ ± 0.56‰ (Larsson et al. 2019). These figures indicate the maximum expected variability within a single manuring regime. The implication is that each Sedgeford cereal taxon was cultivated under variable conditions in terms of manuring levels and/or in other environmental factors e.g. soil type, soil moisture level, or water table depth (cf. Hamerow et al. 2025, p 129). Most Sedgeford samples have low δ15N values compared to the calculated wild herbivore baseline (Fig. 8). 65% have values below the herbivore baseline. The low water content (21.07%) of Sedgeford’s free-draining soils (Sourced from the UK Soil Observatory CS Topsoil-Soil moisture map, https://mapapps2.bgs.ac.uk/ukso/home.html. Accessed 3 April 2025) may influence δ15N values; heavy, water-retaining soils are associated with increased δ15N values, likely due to the activity of denitrifying bacteria under anaerobic conditions (Hamerow et al. 2025, p 91). Figures 8 and 9 shows broad overlap in δ15N values for all taxa. Inter-crop differences are not statistically significant (F (2, 108) = 2.43, p = 0.09) (ESM 1 Table S6). Means for both rye and free-threshing wheat are both less than the wild herbivore baseline, with barley only slightly above (Fig. 9; Table 5). The averaged (‘bulked’) single grain values in Fig. 9 may viably be compared with Bogaard et al.’s (2013) manuring bands for expected δ15N values based on bulk samples of modern grains from plants experimentally cultivated in differing manuring regimes. Figure 9 shows that the wild herbivore baseline calculated for Sedgeford falls in the middle of the ‘medium’ manuring band as calculated by Bogaard et al.: clearly, these bands assume a different (lower) baseline value, and hence some caution is required in their interpretation with reference to Sedgeford data. Mean values for each Sedgeford crop taxon fall within the lower part of the ‘medium’ Bogaard et al. manuring band (2013, Fig. 1).

Fig. 9: Mean δ15N values with associated SD for all samples by cereal taxon Bogaard et al. (2013) manuring bands (dotted lines) and the wild herbivore baseline (dashed) with grain/rachis offset shaded, are shown

Feature | Taxon | Number of (single grain) samples | Mean δ15N (‰) | Standard deviation (‰)
All | Rye | 39 | 4.06 | 2.46
Free threshing wheat | 39 | 3.34 | 1.79
Barley | 33 | 4.41 | 2.07
Kiln 1 | All | 23 | 3.09 | 1.78
Kiln 2 | 30 | 4.16 | 2.08
Kiln 3 | 29 | 4.59 | 2.13
Steeping tank | 29 | 3.63 | 2.35
Kiln 1 | Rye | 9 | 2.96 | 1.53
Free threshing wheat | 9 | 2.80 | 1.72
Barley | 5 | 3.84 | 2.41
Kiln 2 | Rye | 10 | 3.84 | 1.23
Free threshing wheat | 10 | 4.08 | 1.76
Barley | 10 | 4.56 | 3.01
Kiln 3 | Rye | 10 | 5.01 | 2.94
Free threshing wheat | 10 | 3.58 | 1.45
Barley | 9 | 5.23 | 1.36
Steeping tank | Rye | 10 | 4.32 | 3.30
Free threshing wheat | 10 | 2.83 | 1.97
Barley | 9 | 3.75 | 0.82

Statistical comparison of mean δ15N values for crop taxa within each feature also suggests no significant differences (ESM 1 Table S6). Table 5 shows SDs for each taxon within each feature. Most sets of grain taxa within each feature (8/12) have a SD > ± 1.64‰ (Larsson et al.’s (2019) threshold for heavily manured fields), suggesting origins in more than one condition of manuring.

Significantly, SDs for each grain taxon within each feature (the mean for these being ± 1.96‰, n = 12) suggest almost as much variability in grain δ15N values as equivalent SDs for each crop across all features (mean for these SDs being ± 2.10‰, n = 3) (Table 5). The likely chronological separation of kiln 3 (later) from kilns 1 and 2 has been observed. Grains from different features were probably distinct not only in date but also in field-of-origin. Evidence for comparable variability within and between features implies significant heterogeneity in cultivation conditions within each field supplying the malting complex.

No trends are evident in the distribution of samples by feature type (Fig. 8). Differences in mean δ15N values between the features are not significant (F (3, 107) = 2.47, p = 0.07) (Fig. 10; Table 5; ESM 1 Table S7). SDs for samples (of all taxa) grouped by feature all exceed the ± 1.64‰ upper limit for grains from a single source (Table 5), implying grains recovered across features grew in more than one manuring condition.2021

One might imagine, being close to the coast, that seaweed (macroalgae) or fish remains were used to fertilise crops malted at Sedgeford. This is unlikely. Modern research (Gröcke et al 2021) suggests use of both seaweed and fish remains as fertiliser i.e. ‘marine biofertilisation’ cause elevated nitrogen stable isotopic values in soils and associated crop plants. The low stable nitrogen isotope values at Sedgeford suggest marine biofertilisation was not in use.

Fig. 10: Mean δ15N values with associated SD for all samples by feature within the malting complex. Manuring bands (dotted lines) and the wild herbivore baseline (dashed line) are shown, with the grain/rachis offset shaded

Comparing carbon and nitrogen stable isotope values

Figure 11 displays δ15N values plotted against δ13C values for all grains, coded by taxon and area of the malting complex. Datapoints of each type show considerable overlap and there is no significant relationship between δ15N and δ13C values (Pearson’s product-moment correlation co-efficient = 0.15, p-value = 0.12).

Fig. 11: δ15N values plotted against δ13C values for all samples, coded by cereal taxon and feature within the malting complex. The wild herbivore baseline is shown as a dashed line and expected grain/rachis offset shaded

Five clear outlier datapoints (representing different taxa and malting complex features) have elevated values for both δ13C (>–21‰) and δ15N (> 7‰), seemingly attributable neither to measurement or machine error, nor to post-depositional contamination. These may relate to salinity (Sedgeford approximating the coastal zone); research suggests that salinity may elevate both plant δ15N and δ13C values (Heaton 1987; van Groenigen and van Kessel 2002; Yousfi et al. 2010; Hussain and Al-Dakheel 2018). Elevated δ13C values in saline environments likely relate to partial stomatal closure in plants under salt stress, whilst elevated δ15N may be due to raised pH acting to increase discriminating volatisation of lighter N14 isotopes in soil (van Groenigen and van Kessel 2002).

It has been argued (e.g. Caroe 2022b, p 340; Faulkner 2022) that Sedgeford is an example of an early medieval ecclesiastical or lordly estate centre. Supportive evidence for Sedgeford’s representing an estate centre under ecclesiastical control includes the recovery of two metal writing styli from the cemetery part of the site (Jolleys et al. 2019, p 76). Styli at other early medieval sites, including Flixborough, have been associated with elite ecclesiastical oversight, although this is somewhat contentious (Loveluck 2001). Were this the case, it is to be expected that the site would receive imported crops cultivated on an agricultural ‘hinterland’, extending some distance away, for processing in the malting complex: an alternative possible explanation for outlier δ13C and δ15N values.

Seasonality

‘Seasonality’ scatterplots – the output of correspondence analysis – are displayed in Figs. 12 and 13. In each case, weed species are coded by sowing time association (Table 2). Figure 13 shows the same data following square root transformation. Seven weed species (out of 11 eligible) from the assemblage are clearly associated with a sowing season (i.e. are either autumn or spring-germinating).

Fig. 12: Correspondence analysis plot showing cereal taxa and weed species distributed according to associations in samples from the assemblage. Weeds are coded according to seasonality class

Fig. 13: Correspondence analysis as in Fig. 12, with square root transformation applied to data

All three autumn-germinating weed taxa strongly cluster together towards the negative (left) end of the x-axis together with rye. In Fig. 13, this cluster also includes barley. Wheat occurs at the positive (right) end of the x-axis, the only nearby seasonality-associated weed being (spring-germinating) Fallopia convolvulus. Oat also falls towards the positive end of the x-axis. Of the most abundant weed species in the assemblage (Table 3), two are clearly associated with a particular crop taxon: spring-germinating Fallopia convolvulus with wheat and autumn-germinating Agrostemma githago with rye.

Occurrence in both plots of (spring-germinating) Plantago lanceolata, Anthemis cotula and Chenopodium album in a cluster with autumn-sown rye and all three autumn-associated species at the negative end of the x-axis likely reflects the recognised capacity for spring-germinating species to persist among autumn-sown crops (Bogaard et al. 2001). Some mixing is expected among autumn-sown crops since autumn ploughing advantages autumn-germinating weeds but does not exclude spring-germinating taxa (by contrast, spring sowing destroys autumn-germinating weeds). This means that autumn-germinating weeds are a stronger indicator of autumn sowing than spring-germinating are of spring sowing.

To summarise, the correspondence analysis scatterplots are compatible with consistent and complementary sowing seasons for different cereals and by implication, crop rotation. The implication is that rye was an autumn-sown crop, with barley perhaps also autumn-sown. Wheat and oat were probably spring-sown. Given that rye dominates the assemblage, it is not surprising that, of the four cereals, rye shows the clearest seasonality pattern.