Climate and Climate Change
Natural climate change is characterised by infrequent but rapidly initiated, short-lived but global or hemispheric excursions (Mayewski et al. 2004) which exceed anything within normal human experience (deMenocal 2001; Mitchell 2008). Events in prehistory were centred on c. 6200, 4000-3800, 2200-2000 and 1000-800 BC. The beginning and end of the Neolithic period are bracketed by two of these.
Narratives are dangerous because of the absence of precise temporal correlations, so that cause and effect, cannot always be recognised, and because there may have been delayed responses in some systems. Nevertheless the following story, with time expressed as cal years BC, might be constructed. The period 5400 to 4000 BC was 1-2°C warmer than present in NW Europe (Davis et al 2003): temperature fluctuations within the Neolithic period are from a warm 'baseline'. NW European soils became wetter from around 5050 BC, more so after 4750 BC (Hughes et al. 2000; Nesje et al. 2001; Spurk et al. 2002; Langdon et al. 2003; Blaauw, van Geel and van der Plicht 2004; Magny 2004). Ice-rafting in the northern North Atlantic Ocean was earlier at 4700 BC (Moros et al 2004) than the same effect off western Ireland at c. 4350 BC (Bond et al 1997), coincident with accelerated sedimentation and lowered sea-surface temperatures in the southern Irish Sea and off Ireland (Keigwin and Boyle 2000; Scourse et al. 2002; Marret, Scourse and Austin 2004). Thermohaline circulation was possibly weakened (Broecker 2000; Oppo, McManus and Cullen 2003; Thornalley, Elderfield and McCave 2009).
The northern hemisphere cooled at 4500 BC for c. 400 years (Karlen and Larsson 2007), possibly principally felt in lower winter temperatures (Davis et al 2003). Increased frequency and intensity of westerly meridional winds by 4450-4400 BC led to greater storminess around the North Atlantic Ocean, seen in loess and sand-sheet/dune-building around the North Atlantic Ocean (Noren et al 2002; Wilson et al 2004; Jackson et al 2005) and erosion in coastal archaeological stratigraphies (Peeters 2009), most noticeably after 4000-3800 BC (Keatinge and Dickson 1979; Gilbertson et al 1999; Bjorck and Clemmensen 2004; de Jong et al. 2006; Melton 2008, 2009). Geomorphological evidence for increased storminess contrasts, however, with biological evidence from western Scotland for quiescence at this time (Birks and Williams 1983; Andrews, Gilbertson and Kent 1987; Andrews et al 1987; Russell, Bonsall and Sutherland 1995; Sugden 1999).
By 4350 BC soils were increasingly arid (Hughes et al. 2000; Nesje et al. 2001; Spurk et al. 2002; Langdon et al. 2003; Kalis, Merkt and Wunderlich 2003; Blaauw et al. 2004). A c. 2°C fall in mean July air temperatures at 4200 BC in northern Scandinavia (Grudd et al 2002; Helama et al 2002) might have lowered evapotranspiration rates, leading after 4100 BC to wetter ground conditions. Temperatures ameliorated after c. 4100-4000 BC (Cheddadi et al 1997; Grudd et al 2002; Karlen and Larsson 2007).
A lull of a few centuries in the rate or intensity of climatic change is suggested in some data-sets. Relative aridity intensified after 3800 BC (Hughes et al. 2000; Nesje et al. 2001; Spurk et al. 2002; Langdon et al. 2003; Blaauw et al. 2004) as NW European air temperatures became very warm (Karlen and Larsson 2007). A fall in air temperatures in northern Scandinavia at 3700 BC, with a much steeper fall at 3650 BC (Grudd et al 2002; Helama et al 2002) is associated with wetter soils after c. 3650 BC. Dune building recurred at c. 3950-3700 BC (Gilbertson et al 1999; Bjorck and Clemmensen 2004).
Atmospheric circulation in the North Atlantic had weakened by 3400 BC (Bond et al 1997, 2001) and became stable for a few centuries. However, around 3200-3100 BC was a cluster of dune-building events on coasts facing the northern Atlantic (Caseldine et al 2005; de Jong et al. 2006; Holmes et al 2007). Summer temperatures in northern Scandinavia continued to fluctuate, falling at 3200 BC and recovering at 2900 BC (Grudd et al 2002), after which it became very warm (Karlen and Larsson 2007). Major climatic variations in the later Neolithic to c. 2200 BC appear fewer save for wetter soils after c. 2500 BC and an increase in dune-building after c. 2300 BC.
Some parts of the research community are making deterministic connections between climate and cultural change (Baillie 1995; Mayewski et al. 1996; Bogucki 1998; Sandweiss, Maasch and Anderson 1999; Berglund 2003; Brooks 2004; Bonsall et al. 2002; Tipping and Tisdall 2004; Gronenborm 2005, 2006; Turney et al 2005). It is essential to accept determinist arguments at thier face value and to resist the impulse to reject them 'in principle'. At present the connections between climate and culture in prehistory are largely only temporal associations. Testable hypotheses need to be developed that can identify the physical, ecological, economic and social mechanisms driving change. Bonsall et al. (2002) proposed an expansion in the late Mesolithic of opportunities for pastoral farmers as climate became dryer but this change would also have benefited hunter-gatherers, making it hard to see why lifeways should change. Tipping (in press) and Tipping and Tisdall (2004) suggested that climate change led to resource failure for late Mesolithic foragers, making the adoption of agriculture more tempting. There is little evidence so far for this (but there is little enough evidence for Mesolithic subsistence at all, let alone how it changed). The 'slighting of the sea' (Tauber 1981; Schulting 1998; Schulting and Richards 2002; Milner et al. 2004; Hedges 2004; Richards and Schulting 2006) has been interpreted as a response to climatic stress in the North Atlantic Ocean (Tipping in press; Tipping and Tisdall 2004) but its marked diachroneity across the Atlantic façade (e.g. Lubell et al. 1994) probably makes this unlikely. Changes in seasonality across Europe and through time need to be explored more (Davis et al 2003) because these will have affected the viability of crops in regions removed from south-west Asia as well as the availability of indigenous resources (Parks 2009). It cannot be assumed that present-day patterns existed in the Neolithic period.
Almost no palaeo-climatic data-set used above relates directly to Scotland, although each describes changes that will have impacted on Scotland. It cannot be expected that the archaeological community in Scotland should fund climatic reconstructions, but it can encourage the generation of high-resolution proxy records readily achievable in Scotland such as reconstructions of summer temperature from Pinus sylvestris tree ring data (Grudd et al 2002; Helama et al 2002).
Landscape and the Natural Environment
Relative sea-level rise was either very slow or had effectively ceased by the Neolithic period (Shennan and Horton 2002; Smith, Firth and Cullingford 2002), making this factor unlikely to have influenced Neolithic communities in most areas (Armit 2003; Behre 2005).
Armit (2003) drew attention to the losses of land available to Neolithic farming communities through blanket peat spread. This is probably overstated (Mills et al. (2003), although there are few localities in Scotland where blanket peat inception and spread have been systematically measured. Blanket peat inception seems to have occurred in the early-mid Holocene (Robinson 1987; Charman 1992; MacDonald et al 2006; Tipping 2008) and to have had climatic or pedogenic triggers. Once initiated, blanket peat continued to spread in and after the Neolithic period, but its spread does not seem to have accelerated within the Neolithic period. Blanket peat spread need not have confined human activities (Carter 1998; Tipping et al 2007). Behre (2005) argued that in north Germany early farming communities were confined by lowland raised mosses formed on marine mud as relative sea-level rise culminated but except for those in Aberdeenshire on older substrates (Tipping 2007) raised mosses cannot have significantly constrained human choice in Scotland.
The dominant vegetation cover of Scotland in the Neolithic period was woodland. Though present across much of the highlands by the Neolithic period, blanket peat seems not to have prevented tree growth in the way it can now, because it was still thin. Trees also in the main continued to live through the short-lived climatic fluctuations. There is currently no evidence that Scottish Calluna heaths developed via Mesolithic anthropogenic impacts in the ways suggested for English uplands (Simmons 1996). The distribution of major woodland types has been mapped at a broad scale by Bennett (1989), Tipping (1995) and Edwards and Whittington (2003).
Rivers respond to landscape change. Analysis of data in the British Isles has shown limited evidence for accelerated river activity at c. 4180, 3780 and 3590 BC (Johnstone, Macklin and Lewin 2006): explanation of these is that they were climatic in origin. Disappointingly few catchments in Scotland have been investigated (www.aber.ac.uk/rivers/14c) but of these, few show extensive development of floodplains within the Neolithic period: an exception is the Carra Water in Kintyre (Tipping, Carter and Haggart 1994). But it is clear that some rivers had floodplains as wide as today by the Neolithic, such as the Dee (Tipping 2007) and the Kelvin (Tipping et al 2008), providing the context for the possible burial of Neolithic landscapes beneath later alluvium (Howard and Macklin 1999). Lacustrine sediments can also reflect landscape disturbance. The period 4000-3700 BC shows increased soil erosion (Edwards and Whittington 2001) but Edwards (2004) advises caution in interpretation of this. Loch Olabhat is exceptional in the Neolithic period in showing such a phase from c. 3430 BC (Edwards et al 2000; Armit 2003; Mills et al. 2003).
It is important to know the pattern of relative sea level change after its peak at c. 4500 BC, to explain coastal settlement distribution patterns. There are two models, a steady decline to present sea level (Shennan and Horton 2002) or a second period around c. 2000 BC of high relative sea level, the Blair Drummond Shoreline (Smith, Cullingford and Firth 2000). New data suggest that this second event was the more important in some parts of Scotland such as Orkney (Dawson and Wickham-Jones 2007) and Skye (Selby and Smith 2007), but it is not yet clear whether relative sea level actually fell between c. 4500 and c. 2000 BC. As in riverine environments there is the potential in some areas for the burial of Neolithic archaeological landscapes, particularly in the Outer Isles, but at most places on mainland Scotland this is unlikely: for example, the Neolithic platform in the upper Forth Valley at Arnprior (Ellis et al 2002) which lies on but is not overlain by marine sediments.
There is almost no objective evidence from which to evoke the experience of living in or moving through a wooded landscape (Edmonds 1998, 1999; Evans, Pollard and Knight 1999; Austin 2000; Brown 2000; Tipping 2003; Evans and Hodder 2006). It is challenging to use palaeoecological techniques in reconstructing species compositions on spatial scales comprehended by people, or the canopy cover, or the spacing of trees and the extent of scrubby and thorny tangle. There are no 'fossilised' landscapes and no secure present-day analogues. The interplay between different techniques which inform on different aspects and scales of the woodland environment (Kreuz 2008), applied to the same landscape (e.g. molluscan assemblages (Dimbleby and Evans 1974; Davies and Wolski 2001; coleopteran assemblages (Robinson 2000; Whitehouse and Smith 2004); on-site data on colluvial processes (Dreibrodt et al. 2009)) will prove invaluable in the future, as might new palynological modelling approaches (e.g. Caseldine and Fyfe 2006; Caseldine et al. 2007) which create the spatial vegetation patterns that most plausibly explain pollen records. A related problem is intervisibility in a wooded landscape, ignored by most (Tilley 1994; Gaffney, Stancic and Watson 1995; Gibson 2004) and circumvented by others (Cummings and Whittle 2003), but which is critical in much archaeological conjecture relating to the cosmological and territorial concerns of ancient people. Knowledge of pollen recruitment now allows reconstruction in broad terms of the vegetation cover local to particular archaeological sites, but it is currently not possible to to reconstruct tree density. A way forward might be to explore the spatial distribution of plant communities from soil pollen (e.g. Hannon et al. 2008) because soil surfaces have the smallest, most local pollen recruitment areas, though care in identifying sealed contexts uncontaminated by more recent pollen is necessary (Tipping, Carter and Johnston 1994).
Almost all deposition of alluvial sediments in Scottish rivers has happened since the Neolithic period. It therefore follows that in some environments there is a high probability of finding buried Neolithic archaeology and it must be a recommendation that this resource is explored in any archaeological strategy. It is, however, equally true that most erosion of pre-Neolithic fluvial terraces has occurred since the Neolithic and precautionary consideration of this in archaeological research strategies is also necessary. Localities where the present floodplain seems to have been the valley-floor in Neolithic times, such as the Rivers Dee (Tipping 2007) and Kelvin (Tipping et al 2008), need to be explored now. The critical interpretative problem in historical geomorphology remains the assignation of cause; whether autogenic, climatic or anthropogenic. The current paradigm, fed by metadata-sets, is that climate is the principal driver (Macklin, Johnstone and Lewin 2005) but case studies are needed that draw on archaeological data and proxy data for the history of land use. More case studies are also needed that reconstruct the appearance of valley floors in the Neolithic period. As these factors are not yet known, experimental or phenomenological descriptions cannot be made upon any reliable basis, nor it is possible to describe how people used or negotiated riverine landscapes (Brown 2000). One model is that low-gradient rivers would naturally have been anastomosing systems with many thin channels threading their way between dense riparian woods (Brown and Keough 1992): such reconstructions have very profound implications for seeing rivers as viable routes of communication.
Changes in Woodland Cover and Composition
More is known about what happened, and when, to the primary woodland than why. The possible impact of Mesolithic communities on these woods is unclear: if anthropogenic disturbance occurred at all (Tipping 2004) it was small in scale, localised, and was followed by complete tree regeneration, with perhaps a shift to hazel as the principal long-term effect (Turner, Simmons and Innes 1993). At or near the Mesolithic-Neolithic transition there is evidence in southern and western Scotland that vegetation was less frequently burnt, the so called 'charcoal fall': it is not known why.
New paradigms of climate change must affect any interpretations of vegetation change. In northern Scotland some populations of pine declined in the earliest Neolithic period, almost certainly through climatic impacts, though what these were is not known (Tipping et al 2008). The elm decline may be another example: its cause is not yet established and while disease remains for some workers the most plausible reason (Clark and Edwards 2004), climate deterioration has emerged as increasingly relevant (Parker et al. 2002; Edwards 2004). The elm decline no longer defines for most workers the beginning of the Neolithic period, because an anthropogenic cause for this is not currently widely supported. Nevertheless, statistical analyses of all 14C dates suggest that the primary decline in the British Isles began between 4393-4357 BC, very close to estimates for the adoption of agriculture. The decline ended, with elm trees probably being a rare component in woodland (Caseldine and Fyfe 2006), at 3470-3340 BC (Parker et al. 2002). The c. 1000 year range means that individual elm declines need to be independently dated. Reductions in populations of oak trees commonly accompany elm declines as 'classic' landnam indicators but these need not be anthropogenic because precisely synchronous 'dying-off' events occurred in oak populations across north west Europe at 4350 and 3970 BC, and later in the Neolithic period at 2820 and 2550 BC (Leuschner et al 2002). In parts of eastern and northern Scotland pine, elm and oak populations all died at the same time at the Mesolithic-Neolithic transition (Tipping and McCulloch 2003) and the effect this had on Mesolithic perceptions of nature (Larsson 2003) needs to be considered.
People may have drawn on woodland resources without depleting them, with Rackham (1977) and Taylor (1998) arguing for Neolithic coppicing. No evidence has yet come from Scottish sites for woodland management but the use of split oak planks at Warren Field, Crathes (Murray, Murray and Fraser 2009) demonstrates a very high level of skill in handling timber (see also Evans and Hodder 2006).
Following Tipping's (1995) brief review of evidence in Scotland of mid-Neolithic woodland regeneration there has been no discussion from Scottish analyses and few new data that can add to the picture or test ideas such as those of Dark and Gent (2001) concerning declining crop yields with time through disease. From Ireland, O'Connell and Molloy (2001) noted widespread woodland regeneration, broadly synchronous from c. 3600-3200 BC, persisting until c. 2500 BC. No clear reasons were given by O'Connell and Molloy (2001) save that regeneration may represent agricultural decline through declining soil fertility. There are few climatic changes at this time that might have deterred farming.
Universal explanations are sought for what might have been highly contingent events, and it might be advisable to return to well-understood site-by-site interpretations. The elm decline, for example, was far more complex than the term implies, and multiple causes, each compounding the effect of others may be more realistic interpretations (Rackham 1980). The charcoal fall may have resulted from different causes depending on context. Edwards (1988, 1989, 1998) saw this as possibly representing the cessation of Mesolithic manipulation of upland woods by fire as farming was adopted; Tipping and Milburn (2000) found that the charcoal fall occurred in many environments and argued that fires were natural and ceased with the change to a wetter climate. New ways of understanding the detailed form of woodlands and of their management are needed: the Berglund model of contrasting changes in tree and non-tree pollen implies an adversarial relation between people and woods which is wrong.
It is striking how little recent attention has been given to mid-Neolithic woodland regeneration in Scotland: its potential significance, with its hint of agricultural failure, raises many questions.
Proxy Data on the Earliest Agricultural Impacts
Purported anthropogenic impacts on woodland may need to be re-interpreted because some landscapes may have been naturally open (Fenton 2008), clearings natural rather than anthropogenic (Brown 1997, 2000) and woods impacted by deteriorating climate. The critical observation must be in the recognition of positive indicators of land use and not in evidence for woodland loss, but all pastoral pollen indicators grow in natural grassland and their expansion in a pollen record may reflect only reductions in tree pollen. Microscopic charcoal alone is not an unambiguous indicator of human activity.
The least ambiguous indicator of agricultural activity is Avena/Triticum (oat/wheat) pollen, although Avena fatua is a wild grass. Hordeum type (barley type) and 'cereal-type' are ambiguous in what they indicate. Because elm declines can be of mid-Neolithic age, cereal-type pollen occurring before the elm decline need not be significant markers of either an 'invisible' Neolithic or precocious Mesolithic farming: all finds must in future be AMS 14C dated. Cereal-type pollen continues to be reported, though outwith Scotland, from contexts immediately prior to c. 4400 BC (O'Connell and Molloy 2001; Innes, Blackford and Davey 2003).
Excluding pre-elm decline examples, the earliest cereal type pollen record in Scotland currently known is of Avena Triticum pollen at Achany Glen 2, Lairg (Smith 1998), from c. 4200 BC, but this is from a shallow and slow-growing peat in which chronological precision is low. In general the earliest evidences come from the islands, on Orkney at c. 4100 BC and c. 3950 BC (Bunting 1994, 1996) and c. 3950 BC near Skara Brae (de la Vega-Leinert et al 2007), on the Western Isles at c. 3950 BC (Mills et al. 2003) and c. 3800 BC (Bohncke 1988). South of the Great Glen on the mainland, some oat/wheat pollen grains amongst many barley type grains at Warren Field, Crathes have a Bayesian-defined maximal age-range 3820-3700 BC (Tipping et al. 2009), inseparable in age from the cache of carbonised grains at nearby Balbridie (Fairweather and Ralston 1993) and within the age ranges defined by Brown (2007) from AMS 14C dating of carbonised remains. These finds need not be the earliest even at the sites where they are recorded because cereal pollen has very limited dispersal, particularly in woodland, and there is a very high likelihood that single pollen grains will be missed in analyses where the temporal resolution is poor and the pollen sum is low.
It must be borne in mind that pollen grains reported as of cereal type are from cultivated grasses. There have been strong criticisms of pre-elm decline cereal pollen from the potential for sample contamination, the imprecision of identification or incomplete reporting (e.g. Tipping 1995; O'Connell 1987; Bonsall et al 2002; Behre 2007; Brown 2007), which may have resulted in new finds being played down (Macklin et al. 2000; Tweddle, Edwards and Fieller 2005), as well as strong defences (Tinner, Nielsen and Lotter 2007). Analyses must be more rigorous and the presentation of data much more comprehensive. Sediments must be analysed in thin, contiguous samples and to very high sums. 'Optimising' approaches wherein samples are scanned well beyond the total for all other grains (Edwards and McIntosh 1988) is of uncertain value because this is a "seek and ye shall find" approach. Anomalously large grass pollen grains will eventually be encountered but does success really inform the debate? Multivariate statistical analyses of grass pollen grains might be useful (Tweddle et al 2005) although most finds will be of single grains, impossible to classify by such techniques. For individual finds, the medium in which pollen is embedded must be stated, the key/s used indicated, all size measurements reported, their preservation stated because size measurements are distorted by crumpling, photographs of grains published, SEM analyses of sculpturing perhaps standard, sediments directly AMS 14C dated, and sufficient 14C assays obtained to (a) identify anomalies and (b) apply wiggle-matched and/or Bayesian approaches to refining chronologies.
Stratigraphic approaches not using pollen might be pursued. A novel approach to the introduction of millet (Panicum miliaceum) in central Europe has used a species-specific lipid biomarker in lake sediments (Jacob et al 2009) but it may be telling that such lipid analyses suggest the same date of introduction as conventional plant macrofossil and pollen analyses. Ultimately, of course, some workers in the discipline may be attempting the 'fool's errand' of establishing an absence of evidence.
The Scale of Neolithic Agriculture
Arable agriculture originated in savannah and its translation across Europe (Colledge, Conolly and Shennan 2005) probably required open ground (Bogaard 2004; cf. Edwards 1993). It is critical to define the spatial scale and duration of early Neolithic clearances (Buckland and Edwards 1984). Vera (2000) argued that pollen data have seriously under-estimated the amount of open ground in the early-mid Holocene created naturally by large herds of wild animal grazers, but this argument was successfully refuted by Mitchell (2005). O'Connell and Molloy (2001) described extensive early Neolithic woodland clearance at some Irish sites, with the extent of woodland possibly halved. Caseldine and Fyfe (2006), also in Ireland, suggested from new modelling techniques (below) that 'landnam' created a landscape in which overall at least 12 percent was open, but the best-fit suggested that individual openings were small and not necessarily intentional anthropogenic creations.
Pollen diagrams elsewhere are dominated by tree pollen for at least three reasons: (a) trees produce and disperse pollen much more abundantly than herbs and open spaces are undoubtedly significantly under-estimated by pollen analyses; (b) woodland clearance can lead to greater pollen production in remaining trees and their greater ease of transport to pollen sites; (c) many pollen sites are from large-diameter basins which receive pollen from enormous distances, making detection of open spaces nearly impossible (Edwards 1979). Tipping et al. (2009) modelled pollen analyses at the early Neolithic timber hall at Warren Field, suggesting that the hall stood in a clearing some 2km across containing scattered trees: this cleared area would not be detected in analyses from large diameter peat basins. There is no evidence for early Neolithic woodland clearance in Scotland to have been substantial or extensive. Kalis et al. (2003) suggested for central Europe that this might only reflect the absence of extensive grazed grassland. In the Northern and Western Isles, woodland loss within the Neolithic may have been near-total but the cause of that woodland decline, climatic or anthropogenic, has not been established, and woodland gave way to heath (Edwards, Whittington and Hirons 1995), which is not a direct product of agricultural activity.
The on-site evidence for agriculture is considered by Rowan McLaughlin (see also Rowley-Conwy 2004; Thomas 2004; Bogaard and Jones 2007; Jones and Rowley-Conwy 2007) but wider, landscape-scale implications involve, for example, the idea of shifting cultivation or slash-and-burn farming to explain the apparently evanescent, suggested, transient Neolithic settlement pattern in England (Edmonds 1999; Thomas 1999). Slash-and-burn techniques are usually practiced on infertile soils and the need for people to have resorted to such practices on the generally fertile soils of north-west Europe has been questioned (Rowley-Conwy 1981, 2003). Bogaard (2002, 2004) has also recently argued that fields were cultivated for long periods, and that farmers were sedentary. Comparable plant analyses need to be undertaken in Scottish contexts. It should perhaps be expected in addition that farmers would know how to amend or improve the nutrient status of cultivated soils. Neolithic examples of 'man-made' plaggen soils are known (Bakels 1997; Guttmann 2005) with Guttmann arguing for in situ cultivation of Mesolithic midden heaps at coastal sites, but how widespread such practices were remains unknown. The preservation of Neolithic domestic landscapes on Shetland (e.g. Whittle 1986) need to be revisited with the application of new scientific approaches.
DNA evidence indicates that all livestock, including cattle but with the exception, perhaps, of pig, were introduced into Britain (Bailey et al. 1996; Bollongino et al. 2005). In southern Scandinavia cattle were introduced from c. 4000 cal. BC (Price and Noe-Nygaard 2009). There is no evidence yet in Scotland of wild cattle in Mesolithic contexts. It is possible that wild mammals were also introduced by people to, in particular, the outer isles, and Searle has argued for a surprising number of species in Ireland (Searle 2008). Stable isotope data indicate that cattle grazed on grass immediately upon their introduction whereas aurochsen and red deer grazed from the woodland floor (Noe-Nygaard, Price and Hede 2005), although the influence of climate change on nitrogen cycling may affect these data. The scale of livestock-keeping is hard to estimate from archaeological finds from specific 'feasting' sites. Nor is it simple to compare quantities of surviving remains between robust species (cattle) and gracile species (sheep and pig) especially if dogs have been present on the site. Grazed grassland is recorded in most pollen diagrams, though seemingly insufficient to indicate extensive areas of pasture (cf. Kalis et al. 2003) except in the areas where climate has induced less wooded circumstances in northern mainland, and the Northern and Western Isles. Noe-Nygaard et al. (2005; Noe-Nygaard and Hede 2006) argued for the creation of new grassland at the onset of the Neolithic through climate change, the collapse of woodland and the slowing of sea-level rise, all of which would be important in the climatically more vulnerable circumstances of Scotland (above). In some settings, as at Warren Field, cattle and crops appear to have been segregated (Tipping et al. 2009), as they were in traditional shieling systems and perhaps for similar reasons (Kalis and Zimmerman 1988).
- Tipping (1995) was critical of the lack of application of 14C dating to Scottish vegetation histories. Almost all more recent analyses have dating controls, though often not enough, and no Scottish sequence has yet been 14C dated with wiggle-matched or Bayesian precision, which will be needed if archaeological and palaeoecological data are to be compared.
- Simulation modelling (Bunting and Middleton 2005), the approach used by Caseldine and Fyfe (2006), Caseldine et al. (2007) and Tipping et al. (2009) (see also Caseldine, Fyfe and Hjelle 2008), tries to estimate probabilistically the most likely distance from a pollen site that trees of different species grew, contrasting pollen production and dispersal characteristics to create some scenarios of tree distribution that are more likely than others. The future application of this approach has to be encouraged.
- " Palaeoeconomic analyses must be directed far more explicitly to excavated and well-dated archaeological sites. Off-site analyses must be related to on-site data. Too many data are from analyses where human activities are secondary to ecological questions and unsuited to describe human activities. A focused approach, choosing pollen sites that describe landscapes at human spatial and temporal scales, linked as networks by independent dating, could test, for example, the balance between foraging and farming, differences in land uses between halls and houses, the spatial separation of ritual and routine, sedentism and mobility, agricultural success and failure.
- More direct dating is needed of excavated crop remains and animal bones (Brown 2007; Price and Noe-Nygaard 2009). Much more work is needed to be done towards understanding the balance of arable and pastoral activities. The abundance of cattle bone from some archaeological sites has meant an interpretative emphasis on their significance in Neolithic society (Edmonds 1999; Ray and Thomas 2003), but the absence in the palaeoecological record for large or even modest expanses of grazed grassland in the Scottish mainland suggests limited exploitation of herded animals: this paradox needs to be explored. New work on using fungal spore assemblages as indicators of grazing pressure (Blackford and Innes 2006) might provide insights, but currently such approaches cannot distinguish domestic from wild animals. Other approaches to defining the scale or intensity of agricultural activities might come from multi-disciplinary investigations of landscape change, such as at Loch Olabhat (Mills et al. 2003), with the caveat that such changes are not unambiguous indicators of human intervention.