Neolithisation at the site Brandwijk-Kerkhof, the Netherlands: natural vegetation, human impact and plant food subsistence
- First Online:
- Cite this article as:
- Out, W.A. Veget Hist Archaeobot (2008) 17: 25. doi:10.1007/s00334-007-0108-8
- 488 Downloads
Brandwijk-Kerkhof (ca. 4600 to 3630 cal b.c.) is a Neolithic site, located on a river dune in the Dutch Rhine/Maas river area. The natural vegetation and human impact upon it have been investigated by analysis of pollen and macroremains from four cores that are located at increasing distances up to 20 m from the site. The relationship between the strength of human impact on the vegetation and the distance of the cores from the river dune has been investigated as well. The results show that the natural vegetation on top of the river dune consisted of deciduous woodland, while in the surrounding wetlands alder carr and eutrophic marsh vegetation dominated. Human impact of limited strength resulted in more open and disturbed vegetation. There is no correlation between the strength of the evidence of human impact in the pollen diagrams and the distance of the cores from the river dune. The evidence for presence of crop plants from the cores is compared with evidence from the excavation. The first presence of crop plants from ca. 4200 b.c. onwards corresponds with data from other Dutch wetland sites. Large-scale local crop cultivation cannot however be demonstrated.
KeywordsPollen analysisBotanical macroremainsHuman impactNeolithisationCrop cultivationSwifterbant culture
The cultural setting
The Neolithic started in the Netherlands with arrival of the fully Neolithic Linearbandkeramik culture in the southern loess region of the Netherlands at ca. 5300 cal b.c., resulting in a long-lasting open frontier between Mesolithic and Neolithic societies. From ca. 4750 b.c. onwards, the Neolithic Rössen culture was present in the loess region, followed by the Michelsberg culture at ca. 4300 b.c. In the remaining parts of the Netherlands, the Neolithisation process was gradual, characterised by introduction of pottery at ca. 5000 b.c., introduction of domestic animals at ca. 4700 b.c. and introduction of crop plants at least by ca. 4300–4000 b.c., although earlier introduction of crop plants (after 4500 b.c.) cannot be excluded (Brinkkemper et al. 1999; Louwe Kooijmans 2003; Raemaekers 1999). It is not, however, known how far this reconstruction of the gradual Neolithisation process, based on wetland sites where preservation of organic remains is excellent, is representative of southern and eastern sandy regions in the Netherlands (Louwe Kooijmans 1993).
What did the natural vegetation of the river dune and its direct surroundings look like?
How can we characterise human impact on the vegetation?
Is there a relationship between the strength of human impact on the vegetation and the distance from the river dune?
Which crop plants were present and when were crop plants first introduced at the site?
Was crop cultivation practised at the site?
The site at Brandwijk was partially excavated in 1991 by a team from Leiden University under the direction of Van Gijn and Verbruggen in a research project on Neolithic occupation on late-glacial river dunes. At Brandwijk, Swifterbant pottery was found in all layers. The flint assemblage indicated contacts with southern regions, and fragments of flint axes at the base and top of layer 50 indicate incorporation of the Michelsberg material culture (Raemaekers 1999). There were no structures present except for a series of pointed posts (dating to layer 50 or later). The bone assemblage was dominated by beaver, otter, deer and wild boar. Bones of domestic animals (dog, goat/sheep, cattle and pig) were present from initial occupation onwards but only formed a minority of the assemblage. Preliminary results indicate that the site was used as a temporary hunting/fishing camp. The similarity of the mammal and fish faunas through the layers suggests consistency of site function through time (Van Gijn and Verbruggen 1992; Raemaekers 1999).
Find layers at Brandwijk-Kerkhof
Age (cal b.c.)
Geology and landscape
Materials and methods
The size of samples used for pollen analysis was 1 cm3. The sample interval was generally 10 cm. The samples were prepared according to the standard methods (Fægri and Iversen 1989) and to each sample a Lycopodium tablet was added. The samples were analysed with a microscope at magnifications of 400× and 640× by W.J. Kuijper, L. van Beurden, A. Reinink and D. Paetzold. Identification was based on literature (Fægri and Iversen 1989; Moore et al. 1991) and the reference collection of the institute. Identification of Cerealia-type pollen was based on the diameter of the grain and the ratio of the diameters of the annulus and the pore. Cores A and C were partly analysed: only samples from that part of the core containing charcoal and/or bone remains were selected. Data were converted to percentage diagrams with the software programs Tilia v2.0.b.4 and TGView v2.0.2 (Grimm 1991–1993, 2004). The pollen sum (300 pollen grains if possible) includes dry land trees, shrubs, herbs and pterydophytes such as ferns and mosses, enabling representative analysis of changes in the dry land vegetation. Exaggeration is fivefold, if applied. The classification of plants into ecological groups is based on Schamineé et al. (1995–1999).
The four cores used for pollen and spore analysis were divided into samples of 5 cm thickness that were all used for macroremains analysis. The size of the macroremains samples was ca. 25 cm3. A fragment of wood prevented regular sampling of macroremains in core D at −6.27 m. W.J. Kuijper and L. van Beurden analysed the macroremains with a binocular microscope at a magnification of 6 to 40×, and identifications were based on the literature and the reference collection of the institute. The results were calculated and drawn up with Tilia v2.0.b.4 and TGView v2.0.2. The diagrams show absolute numbers of seeds. Straight lines indicate borders of biostratigraphical zones, while dotted lines indicate the part of the core that was subjected to pollen analysis. Names of plant species are given according to Van der Meijden (1996). The excavation included analysis of botanical macroremains as well, but only highly relevant results will be preliminarily presented here. The samples of macroremains from layer 50 of the excavation cannot be allocated to the base or top of layer 50. Galeopsis-type represents G. bifida/speciosa/tetrahit; Ranunculus repens-type represents R. acris/lingua/repens but it probably concerns R. repens here.
Radiocarbon dates of the cores of Brandwijk
Depth (m. - NAP)
Age cal b.c. (2σ)
Corresponds with layer
5120 ± 40
3990 (95.4%) 3790
5110 ± 50
4040 (1.4%) 4020
4000 (94.0%) 3780
5100 ± 90
4250 (95.4%) 3650
50 top/60 (50 base)
5110 ± 70
4050 (95.4%) 3710
5360 ± 100
4370 (95.4%) 3960
5470 ± 80
4470 (81.4 %) 4220
45 (50 base)
4210 (6.9%) 4150
4140 (7.1%) 4050
Zone D-II is characterised by a temporary decrease in the percentage of Quercus in the lower part of the zone, followed by an increase of shrubs such as Corylus and Viburnum opulus. The peak of Pinus probably reflects regional pollen deposited during a period characterised by very open vegetation. The concentration diagram shows a similar decrease of Quercus, peak of Pinus and increase of shrubs as well. The wetland vegetation in this zone is characterised by fluctuations of Alnus and an increase of the marsh/riparian vegetation. The values of Cyperaceae, monoletae psilatae spores, Apiaceae, Filipendula, Galium-type, Urtica dioica-type, Sparganium erectum-type, Valeriana and H. lupulus increase, which indicates open and eutrophic conditions. Altogether these changes indicate that both the dry land and the wetland vegetation became more open. At −5.51 m, the landscape furthermore gradually became wetter. This is indicated by the presence of Sagittaria sagittifolia-type, Nymphaeaalba-type, Potamogeton, Salvinia natans and Pediastrum. These indicate the presence of calm open water and this partly explains the open environment. S. natans is an indicator species of the Atlantic and has been found regularly in deposits from the Rhine delta (Zandstra 1966).
Zone D-III is characterised by a strong increase of dry land herbs, anthropogenic indicators and some wetland herbs, followed by an increase of dry land shrubs (discussed below). Quercus is locally present at −5.26 m. The low values and absence of macroremains of Tilia as well as the increased values of Fraxinus and Betula may be related to a rise of the water table in the extra-local area, since Tilia grows on relatively dry soils while both Betula and Fraxinus tolerate wetter soils. Alnus shows fluctuating but generally decreased values and water plants are still well represented.
Zone B-II is characterised by an increase of dry land herbs and shrubs and low values of Tilia, indicating absence of lime in the local vegetation. Fraxinus, Betula and Rhamnus cathartica show a slight increase and Prunus is present, all indicating presence of open patches in the vegetation. The increase of Pinus is probably related to an increased influx of regional pollen since the dry land herbs and the wetland taxa indicate relatively open vegetation. Anthropogenic indicators are present (discussed below). In the wetland vegetation, Alnus decreases, after which Poaceae, Cyperaceae, Asteraceae tubuliflorae and monoletae psilatae spores show high values. The peak of Asteraceae tubuliflorae indicates local presence of Asteraceae and corresponds with core C. The presence of Nuphar lutea-type, Myriophyllum verticillatum, Spirogyra and Zygnemataceae indicate a very wet phase (corresponding with core D).
Zone B-III is characterised by an increase of Tilia and stable though relatively high values of dry land herbs. Zone B-III probably represents a time period later than the upper part of core D and the analysed parts of cores A and C. Unfortunately, the pollen sums of the spectra are very low, limiting the possibility of making detailed conclusions, for example about the low values of Ulmus and Fraxinus. The concentration diagram indicates that the increase of Tilia is a true increase and not an artificial result of the decrease of Quercus. Poaceae (probably representing Phragmites), Alnus, Filipendula and Asteraceae tubuliflorae were dominant in the wetland vegetation although the macroremains diagram indicates a minor importance of Alnus.
The macroremains of the cores contain several edible plants, such as Quercus, C.avellana, Prunus spinosa, C. sanguinea and Rubus fruticosus, but remains of crop plants were however not found there. The wild food plant spectrum can additionally be extended by finds from the excavation of non-carbonised seeds and fruits of Malus sylvestris and Crataegus monogyna as well as carbonised seeds/fruits/nuts of C. avellana, C. sanguinea, Trapa natans and N. alba. The carbonised state of the latter taxa suggests that people collected and/or consumed the edible parts of these plants. A similar broad range of wild plant food sources is documented at the Hardinxveld sites and the Hazendonk (Bakels 1981; Bakels and Van Beurden 2001; Bakels et al. 2001; unpublished data, Faculty of Archaeology, Leiden University).
The results of the botanical analyses confirm the correlation of the cores based on stratigraphy and 14C dates, although the pollen diagrams from some cores correspond better with each other than the diagrams from others. Zone D-II corresponds with zone B-I since these zones reflect the vegetation before the increase of dry land herbs and shrubs. Zone D-III corresponds with zone B-II and the pollen diagrams of cores C and A. Limited correspondence of the pollen diagrams from cores A with cores C and D can be explained by the preservation state, the low pollen sums of core A and differences in the local vegetation.
Wood and charcoal identifications (Hänninen and Vermeeren 1998)
The wetland vegetation was dominated by eutrophic alder carr, consisting primarily of A. glutinosa; additional trees and shrubs were Salix, Rhamnus frangula and Myrica. The main pattern of the wetland vegetation can be characterised as the development from a dense alder carr (lower part, core D) via somewhat more open alder carr and marsh vegetation (middle part, cores A, B, D and the upper part of core C) to a combination of open alder carr, Phragmites vegetation, open marsh and open water (middle and upper parts of all cores). The macroremains diagrams of the cores clearly demonstrate that alder was always present in the near surroundings despite some fluctuations. Both the pollen and seeds comprise many wetland taxa that could have been part of the alder vegetation as well as of the marsh vegetation (for example, Apiaceae, Cyperaceae, ferns, Iris pseudacorus, L. salicaria, Solanum dulcamara and U. dioica). A clear distinction between these vegetation types is therefore not possible (Stortelder et al. 1998). The marsh vegetation and riparian plants indicate a water depth between 0 and 50 cm. Floating water plants and riparian plants can form floating vegetation mats in open water, thus resulting in peat formation. This concerns Phragmites australis, Typha angustifolia, Rumex hydrolapathum and N. lutea among others (Weeda et al. 1985–1994).
Although the sediment of the investigated cores consisted of peat alone, the presence of pollen, seeds and spores of N. alba, N. lutea, Potamogeton, S. natans and M. verticillatum demonstrate the presence of stagnant or slowly running, open water with a depth of 0.5–2 m. These indicators of open water are regularly present in limited numbers but reach their maximum values and diversity during the formation of layers 50 and/or 60, and correspond with deposition of clay on the northwestern side of the river dune then. Open water increased accessibility of the river dunes and might have influenced the selection of areas for human activity on the river dune.
Human impact on the vegetation
Signs of human activity can only carefully be linked with known occupation periods since the chronology of the cores is not known in detail and since there was no section available that linked the stratigraphy of the cores with the excavation. The lowest spectra of core D (−6.60 to −6.46 m) and C (−6.35 to −6.15 m) may reflect human activity that is probably related to layer 30. In the relevant spectra of the pollen diagram of core D, the signs of human impact are very weak, characterised by the relative high diversity of dry land shrubs and presence of taxa that indicate open patches (U. dioica-type, Apiaceae, Mentha-type). In the macroremains diagram from core D, the evidence of the pollen diagram is supported by the presence of sand. In the macroremains diagram and lithostratigraphy of core C, human impact is reflected by the presence of sand, charcoal and the presence of U. dioica. It is the combination of lithostratigraphy, pollen and macroremains of two cores that allows distinction of more open vegetation due to very small-scale human impact. Absence of this possible anthropogenic horizon in cores A and B can be explained by their shallower depth (Fig. 3).
The lower part of zone D-II corresponds with layer 45 and the base of find layer 50. The open landscape in this zone as discussed above may be the result of more or less continuous low-level human influence, resulting in an increase of dry land shrubs and open wetland vegetation. The increasing water level nevertheless played a role as well, as indicated by the presence of indicators of open water and the changes in the wetland vegetation. Zone B-I corresponds with D-II but does not show changes that can be linked to human activity with certainty, despite being closer to the river dune.
The changes in zone D-III correspond with the formation of layers 50 and/or 60. This occupation period has resulted in the strongest anthropogenic signal compared with earlier phases. The pollen diagram of core D shows an increased influx of the regional Pinus pollen, indicating a more open landscape, a slight decrease of Tilia, presence of the anthropogenic indicators Cerealia, Artemisia, Plantago lanceolata, Polygonum persicaria-type and Polygonum convolvulus-type (Behre 1981) and high values of Poaceae, Cyperaceae, S. erectum-type and Apiaceae. Furthermore, Alnus shows low values and the pollen curve of Filipendula shows high values, which may indicate clearance of Alnus carr (Weeda et al. 1985).
Zone B-II corresponds with zone D-III, representing the top of layer 50 and/or 60. The percentage of Tilia and Alnus in zone B-II is relatively low again. The diversity of dry land shrubs is high, especially in the beginning of the zone, although the number of dry land herbs that indicate human impact is low (Chenopodium album, Artemisia and P. lanceolata). In contrast with cores C and D, pollen of Cerealia is absent, which might be related to the presence of different activity zones on the river dune, colluviation processes and/or differential preservation.
The pollen diagram of core C, corresponding with D-III and B-II, shows human impact in the middle two spectra. Anthropogenic indicators are, for example, Cerealia pollen and spores of Pteridium. The macroremains and lithostratigraphy clearly show the occupation horizon, indicated by presence of sand, charcoal, fish remains, a decrease of Alnus, and presence of Caltha, B. erecta, Filipendula, U. dioica, Eupatorium cannabinum and Mentha aquatica/arvensis. The pollen diagram from core A does not seem to contain strong indicators of human activity despite correspondence with those parts of other cores that show clear human impact; the low pollen sum however restricts the accuracy of the diagram. The macroremains and lithostratigraphy of this core again show human impact: sand, charcoal, bone remains (including fish remains) and relatively high values of U. dioica and E. cannabinum.
The development of the vegetation after occupation is represented in zone B-III (pollen and macroremains), and in the macroremains diagrams in the upper part of cores A and C. Decrease of human impact in zone B-III is characterised by partial recovery of Tilia, decrease of the percentage and frequency of dry land shrubs other than Corylus, disappearance of P. lanceolata, decrease of charcoal and absence of bone remains. The upper parts of the macroremains diagrams from cores B, C and A all show a decreased number of taxa after the period of occupation. Mainly macroremains of dry land trees, Alnus (dominant) and Carex remain present, representing the natural climax vegetation. These developments indicate that the vegetation recovered and became denser than during the period of human impact. Woodland was still present after occupation on the higher parts of the river dune, indicated by the presence of remains of Tilia, Quercus, Fraxinus and Corylus, despite the small surface of the river dune, the high water table and the effects of occupation.
Summarizing, the diagrams show some evidence of human impact, correlated with layer 30, layer 45 and/or the base of layer 50, and with the top of layer 50 and/or 60. The first two signs are of minor strength, indicated by an increase of the percentage and diversity of dry land shrubs and a slight increase of wetland herbs that indicate increased presence of light and nutrients, in the pollen diagrams and macroremains diagrams (Filipendula ulmaria, U. dioica, B. erecta and E. cannabinum). These changes indicate that the vegetation became somewhat more open. Human impact on the dry land vegetation was however limited, as indicated by the absence of strong changes in the curves of dry land tree taxa in the pollen diagrams (corresponding with absence of evidence for local crop cultivation, discussed below). Generally speaking, it is not possible to reconstruct precisely which parts of the diagrams reflect occupation periods and which parts reflect a lack of occupation, despite the small distance of the sample locations from the river dune (especially concerning core D). It is furthermore not possible to exclude influence of non-human disturbing factors such as the gradual rise of the water table and the influence of wild animals such as beaver and wild boar. The limited strength of human impact at Brandwijk corresponds with the pollen diagrams of the Hardinxveld-Giessendam sites, where it is even more subtle.
Only at ca. 4000–3800 cal b.c. (5100 b.p.) is strong evidence of human impact visible in the pollen diagrams, corresponding with the formation of layer 50 and/or 60. While the relationship with changes in the vegetation and human activities remains unclear for the previous phases/signs of human impact, it is highly likely that these signs are related to human activities, since they are characterised by the presence of charcoal, pollen of crop plants and classical anthropogenic indicators (Behre 1981), followed by partial recovery of the vegetation afterwards. Trees that show recovery after human impact are Tilia and Alnus, which indicates that these were used in some way by people. The analysis of wood and charcoal remains indeed supports the view that Alnus was used (Hänninen and Vermeeren 1998). The use of Tilia is not demonstrated by the wood and charcoal identifications though this might be related to the fact that Tilia wood is rarely preserved.
There are several possible reasons for the strength of the evidence of human impact, described above, that corresponds with formation of layer 50 and/or 60. In the first place, the intensity of occupation (duration of occupation, number of people, etc.) and the fact that the evidence might represent two occupation periods (the top of layer 50 as well as layer 60) may play a role. Secondly, distance to places of human activities may also play a role. Archaeological investigation by coring demonstrated the presence of only layer 50 on the northwestern side of the river dune, while remains of other layers were concentrated on its southern side (although the analysis of the cores presented here demonstrates the presence of charcoal and bone remains in layers other than 50 north of the river dune as well). This can explain why the lower and middle parts of core D do not show distinct evidence of human activity. Thirdly, the introduction of crop plants at Brandwijk contemporary with the formation of layer 50 may play a role. It is not possible to differentiate between these hypotheses according to the data currently available.
Influence of increasing distance from the river dune on evidence of human impact
The results only allow analysis of the signs of human impact related to the top of layer 50 and/or 60 to test the hypothesis on the relationship between the evidence of human impact and the distance of the cores from the river dune where this activity took place. Comparison of the anthropogenic evidence in the four cores indicates that there is no linear relationship. It is possible to recognize this evidence of human impact in all cores, and the percentages of dry land shrubs and of dry land herbs characteristic of human impact does not strongly differ between the cores. Only in core A is the signal relatively weak; this is assumed to be related to the poor preservation of the material.
There are several explanations for the limited relationship between distance from the river dune and evidence for human impact. A first explanation is that the relief of the subsoil is more complex than known at the moment. This could explain human impact on vegetation on higher ground if such terrain was present close to cores C and D, and in similar evidence of human impact in all cores. Geological sources do not, however, confirm this hypothesis. A second explanation is that the distance is too short for a decrease in the evidence for human impact with increasing distance from the river dune. Although it is not uncommon for evidence of human impact to be reflected in pollen diagrams over a distance of 20 m from the site of activity, comparison of pollen diagrams of the nearby Hazendonk shows that such a distance does not exclude the possibility of a decreasing anthropogenic signal. These pollen diagrams from the Hazendonk, reflecting the vegetation on the southwestern side of the river dune near the location of human activities, show that even a distance of 3 m results in a decrease of the evidence of human impact (Out, unpublished data). A third explanation is that the distance from the cores to the place of human activities is too large since these activities were mainly concentrated on the southern side of the dune during occupation. This hypothesis does not however correspond with finds of archaeological remains on the northern side of the dune. A fourth explanation is that the signs of human impact at Brandwijk-Kerkhof are overruled by the pollen signal from the much larger river dune Brandwijk-Donk, located nearby to the north. It is, however, unlikely that the pollen rain from the larger dune resulted in the presence of herb pollen at Brandwijk-Kerkhof or that the occupation at both dunes took place at exactly the same time, as implied by this argument. A fifth explanation is that the pollen source area of all cores was relatively large compared with the Hazendonk due to open vegetation at Brandwijk, resulting in a highly similar pollen rain at all the sampled locations. This explanation is supported by the presence of open water to the north of the dune during formation of the top of layer 50 and/or 60. Moreover, it is not possible to exclude secondary deposition of pollen due to erosion or people dumping waste, resulting in a similar pollen spectrum over 20 m.
Introduction of crop plants: local crop cultivation?
Local crop cultivation at Dutch Neolithic wetland sites has been subject of debate (see Bakels 2000). For Brandwijk it is not clear whether cereals were cultivated locally or whether they were cultivated in the nearby southern sandy regions outside the marsh area and brought in from there. In the pollen cores, cereal pollen was found in cores C and D. The local conditions at these core sites, however, would have been unsuitable for crop cultivation because of the high ground water level and risk of flooding as indicated by the nearby presence of open water. The cereal pollen was therefore probably dispersed from the river dune itself. There are several arguments that support local crop cultivation on the dune: finds of pollen grains and macroremains of cereals including a small number of chaff remains of free-threshing naked barley, and finds of pollen and macroremains of taxa that may represent arable weeds (not discussed here). There are however also several arguments against local crop cultivation: the excavation did not produce flint artefacts with sickle gloss (Van Gijn, personal communication) and only a single fragment of a quern (a mano; Verbaas, personal communication), the surface of the river dune is rather small for crop cultivation on a subsistence scale, the pollen diagrams do not show large-scale woodland clearance, the presence of cereal pollen can be explained by crop processing rather than cultivation and the presence of pollen grains could be correlated with the presence of waste (see Bakels 1986). Similarly to the Hazendonk, there is not enough evidence to show that cereals were cultivated at Brandwijk, and cereal cultivation at Brandwijk on a considerable scale seems very improbable (Bakels 2000). However, it is possible that there were small plots of arable land on the river dune (horticulture). Considering the few domestic animals in the bone assemblages, it seems that crop cultivation was probably of minor importance for subsistence at Brandwijk. The absence of strong indications of local cereal cultivation furthermore corresponds with the absence of strong signs of human impact on the vegetation as discussed above. The indications for cultivation of poppy at Brandwijk are even less strong than for cultivation of cereals, since the macroremains contained only non-carbonised seeds and concentrations were not found. It is furthermore not clear whether poppy grew as a crop plant or as an arable weed (Bakels 1982). Poppy has been found at only one other pre-Roman Dutch wetland site, Schokland-P14, but the age of this find is unclear (Neolithic; Gehasse 1995).
The introduction of cultivated plants at Brandwijk at the time of deposition of layer 50 is an essential step in the Neolithisation process in the central Netherlands. Based on the results from Brandwijk and the Hardinxveld-Giessendam sites, it must be concluded that crop plants were first introduced between 4220 and 3940 cal b.c. (layer 50), while there are strong indications that they were not present before 4370 cal b.c. (end layer 45). As explained in the introduction, the Brandwijk data fills the gap between that from the Hardinxveld-Giessendam sites and the Hazendonk. The timing of the introduction of crop plants at Brandwijk furthermore strongly corresponds with data from other Dutch wetland sites (Swifterbant S3: 4300–4000 b.c. and Hazendonk phase 1: 4000 b.c.). Emmer and naked barley have also been found as dominant crops at other Dutch wetland sites that are part of the Swifterbant culture and the related Hazendonk group. The sites concerned are Noordoostpolder-P14 (Gehasse 1995), Swifterbant S3 (Van Zeist and Palfenier-Vegter 1981), Hazendonk phase 1 and 2 (Bakels 1981), and Rijswijk-Ypenburg (Van Haaster 2001), Wateringen 4 (Raemaekers et al. 1997) Schipluiden (Kubiak-Martens 2006) and Hazendonk phase 3 (Bakels 1981), respectively. Considering the influence of the Michelsberg culture on the material culture at the time of the first presence of crop plants at Brandwijk, contacts with this culture might have resulted in the introduction of crop plants as far north as the Rhine/Maas river district. Earlier introduction of crop cultivation under the influence of the Rössen culture remains an alternative hypothesis (Brinkkemper et al. 1999, p. 82), especially for the more southern sandy dry regions and the IJsselmeer basin from which there is no data so far from the relevant period.
The natural vegetation of the river dune at Brandwijk consisted of deciduous woodland on its higher parts combined with alder carr and eutrophic freshwater marsh vegetation on the lower slopes and in the surrounding wetland area. The pollen diagrams indicate that human impact was moderate. Only the changes in the pollen diagram contemporary with the top of layer 50 and/or 60 can accurately be attributed to human impact. The pollen and macroremains diagrams show recovery of the vegetation after this period of human impact. Signs of human impact on the river dune are characterised by a slight decrease of Tilia, an increase of shrubs (V. opulus, R. cathartica, C. sanguinea, L. vulgare, Sambucus and Prunus), dry land herbs characteristic of human impact (Artemisia, P. lanceolata, P. persicaria-type and Polygonum convolvulus-type), Poaceae and Cyperaceae, wetland herbs that indicate disturbance and open patches (F. ulmaria, U. dioica, B. erecta and E. cannabinum), and a decrease in macroremains of Alnus. There is no relationship between increased distance from the dune and strength of the evidence of human impact. The open character of the vegetation may play a role here, although this remains speculative and needs further investigation in the context of comparable sites.
The botanical data from Brandwijk reveals new information on the Neolithisation process in the central Dutch wetlands: there is evidence for introduction of T. dicoccum, H. vulgare var. nudum and P.somniferum ssp. setigerum between 4220 and 3940 cal b.c. There are, however, no strong indications of local crop cultivation at the site itself and the details of the process of introduction of crop cultivation remain unknown. Considering the function of the site, the variety of wild food plants and the indications against local crop cultivation, the importance of crop plants in subsistence seems to have been limited.
This article is dedicated to Prof. Corrie Bakels on the occasion of her 65th birthday. I would like to thank Corrie for providing the data from Brandwijk. She collected the material herself in the field together with Marten and Wim. I am also grateful to her for stimulating years of education. I would like to thank W.J. Kuijper, L. van Beurden, A. Reinink and D. Paetzold for analysis of the data, and C.C. Bakels, W.J. Kuijper, A.L. van Gijn, A. Verbaas, M. Verbruggen, C.E. Vermeeren, M. van Waijjen, L.P. Louwe Kooijmans, O. Brinkkemper, L.I. Kooistra, referee J. Meurers-Balke and J. Greig for their kind support and useful comments during the realisation of this paper. This study was supported by NWO, Netherlands Organization for Scientific Research.