Demic diffusion into Greece: three key observations
In this paper we argue that agriculture arrived in southeastern Europe by demic diffusion and present a major modification of Ammerman & Cavalli Sforza's wave-of-advance model for this process. Our arguments rest on three observations:
1. the concentration of Neolithic settlement in areas of Greece and the southeastern Balkans that were sparsely occupied by an indigenous Mesolithic population;
2. the patchiness of the settlement patterns, and
3. the obvious preference of the immigrants for the floodplains of rivers and lakes.
None of these observations is entirely new, but none has received the attention they collectively deserve.
Greece had a small, scattered Mesolithic population found so far only in the coastal zone of the south and west, but the Neolithic settlements are located mainly outside areas of prior occupation. Perlès, evaluating mainly from lithic technology the evidence for the origins of the Greek Neolithic, also concluded that the thessalian Neolithic was the result of people moving into a hitherto uninhabited area, although at Franchthi Cave a Mesolithic element persisted into an intrusive Neolithic.
The same applies to the southeastern Balkans if we exclude the Lepinski Vir sites of the Iron Gates of the Danube. Except for some coastal sites, the Mesolithic is unknown in Bulgaria and in southern Yugoslavia, and Chapman stresses the lack of indigenous population in nearly all the southeasternmost Europe. We leave aside here recent suggestions of an extensive Mesolithic population on the now-submerged shelves of the Black Sea, because we regard their complete wipe-out as implausible.
The patchiness of the earliest Neolithic settlements in Greece and the Balkans is striking. The dense site pattern in Thessaly contrasts sharply with their absence in Macedonia and the sparse scatter in Boeotia, Attika, Epiros and the Peloponnese. This variability has been attributed to uneven exploration and low discovery rate wherever dwelling mounds were not part of the Neolithic landscape, but that argument cannot be sustained, now that surveys of the southern Argolid, Nemea, Berbati, Boeotia, Messenia, Langadas in Macedonia and Nikopolis in southern Epiros have confirmed [it]. Moreover, this dispersed pattern persisted for several millenia, suggesting that the mere presence of open space was not enough cause for expansion.
Barker has summarized the prevailing view on early Neolithic land preferences: "Today the most profitable arable soils tend to be those of the great plains, such as those in Macedonia in Greece, or the Tavoliere, Maremma an Po plains in Italy. However, [...] the fertility of the great plains and valleys could only be tapped economically with modern technology [...] Naturally waterlogged, many of the major plains and valleys have been far more important regions of animal grazing in the past than of arable soil."
This runs counter to Sherratt's view who had realized much earlier that early neolithic farmers might have chosen floodplain soils for the water they stored after the floods of spring. He was ignored because of a pervasive belief that the traditional farming methods of Greece and the southeastern Balkans illustrate prehistoric agricultural practice. As Barker stated: "Mediterranean ards and ploughs are designed to scuffle the topsoil, prepare the seed bed and trap moisture in the soil. The lightness of the implement is the reason why the fertile but waterlogged plains and river valleys of the Mediterranean region have so often been left for animal pasture until the modern era."
This extrapolation is dangerous, because no venerable old ways can be traced beyond or even into the Middle Ages. The removal of whole landscapes from consideration only because they did not suit assumed ancient farming methods is quite unreasonable.
A geo-archaeological investigation of the late Quaternary history of the Thessalian plain, carried out from 1982 to 1992, has enable us to specify in detail the landscapes occupied by Neolithic farmers and to infer the reason for their choices. To do so, we must first consider the nature of river floodplains.
River landscapes
Almost everywhere in Europe the rivers of late glacial and early pst-glacial times carried large loads of coaese sediment supplied by icecaps or, in Greece, by mountain glaciers and the sparsely vegetated landscape of a dry glacial climate. Overloaded relative to their discharge, the rivers formed networks of shallow channels that endlessy silted up here and eroded there as the large, coarse sediment load choked each in turn. Today, the sand-and-gravel floodplains of braided rivers are the common kind in dry, mountainous Greece.
When icecaps and glaciers had receded and vegetation again protected the land, many rivers changed to sinous, meandering channels and wide floodplains. In spate, when the flood exceeds bank height, its velocity drops abruptly and fine sand, silt and some clay are deposited on the banks, while coarse sand and gravel remain in the channel. Natural levees develop that consisted of silty clay and fine sand with diagnostic mm-scale laminations, while the heaviest clay settles only in the quiet waters of the backswamps. Eventually the shaoling channel and rising levees compel the river to seek a new course lower down in the floodplain. For a few centuries two active channels may co-exist, but in the end one is abandoned and, together with its levees, remains as broad ridge. This process, called avulsion, repeats itself every 500-1000 year, and a floodplain ultimately consists of many raised abandoned and a few active levee-and-channel complexes.
The floodplain is thus a dynamic landscape on a human time scale, but the abandoned levee/channel systems are dry most of the time and never deeply submerged. They make good dwelling sites and the rising mounds themselves add further security. [...]
Late Quaternary landscape evolution in Thessaly
Hundreds of neolithic mounds (magoules, s. magoula) dot the Thessalian plain. Halstead has suggested that thier occupants farmed ligth soils watered by rain in small woodland clearings on elevated river terraces, using hill country and floodplains for hunting, grazing and wood-cutting. Thus Anatolian farmers migrating to southeastern Europe needed only to apply the presumed dryland farming practices of their country of origin. We shall show here that this is not correct in Greece nor probably in Anatolia.
Neolithic pollen records, although difficult to interpret in terms of human exploitation do not reflect major woodland clearing in Greece until some 4000 years ago, and even then degradation by grazing may have been a main cause. There is little botanical evidence for major exploitation of the terraces and foothills before the Bronze Age. Any clearing of the willow and alder cover of floodplains, on the other hand, would be poorly reflected in the pollen record.
The thessalian plain consists of two subsiding basins separated by a ridge of Pliocene marl (the Revenia) and surrounded by mountains. The Peneios river rises far to the west on the Pindos range, traverses the Trikala basin and, joined by several tributaries, crosses the median ridge in the Kalamakiou gorge. Continuing first eastward and then north across the Larisa basin, the river reaches the Aegean through two more narrows, the Rodia Gorge and the Vale of Tempe. During the last to glacial-interglacial cycles the river deposited a complex upper floodplain in the Larisa basin, the High Terrace, then incised its channel. The High Terrace, although important for the early Palaeolithic need not be discussed here.
Another floodplain, now also incised above Larisa, formed late in the last glacial. This Low Terrace jas a complex past that conflicts with the view of Jarman that "stratigraphic evidence of the relationship between Neolithic settlements [in Thessaly] and the modern land surface does not suggest that here has been much change in the relevant period, and the modern situation can probably be taken as a rough indication of pedological condition at the time."
The Low Terrace comprises the deposits of three phases of extended deposition when the sediment supply exceeded for a time the rate of subsidence of the area. They are separated by long intervals when deposition was restricted to the vecinity of the river. A widespread lower unit of sand and gravel 10-15 m thick, the Agia Sophia alluvium, was deposited 40,000 to 27,000 years ago on a wide, braided floodplain. The river then withdrew to the centre of its valley and a mature palaeo-sol capped the landscape, until between 14,000 and 10,000 years ago the river once more spread its deposits across the plain. Less extensive than its predecessor but also a braided floodplain of sand and gravel, this Mikrolithos alluvium is only a few metres thick. Some 10,000 years ago the landscape stabilized and the Non-calcareous Brown soil formed on the Mikrolithos surface.
More than 8500 years ago the river began to lay down sediments across the entire Mikrolithos and much of the Agia Sophia surface. Unlike the older units this Girtoni alluvium consists of the silty clays and fine sands of meandering floodplain and displays the typical laminations of levee deposits. It is capped today by the Girtoni soil.
The Girtoni phase ended late in the Neolithic. An early Late Neolithic cemetery (Tsangli-Larisa phase) dug into its deposits defines its end.
Floodplain deposition then became restricted again until the Deleria alluvium formed in Hellenistic or Roman times, while the present floodplain (Peneios alluvium) is still being built by the Peneios river. Soil erosion due to human exploitation of the land was responsible for both historical floodplains and probably for part of the Girtoni as well.
Above Larisa the Peneios is incised below the Kalamakiou gorge. Because of this, the Girtoni alluvium forms a terrace some 5 m below the Agia Sophia surface and raised about as much above the historical floodplain. The famous magoula of Argissa and several other mounds were built on the Agia Sophia surface at the edge of the then active floodplain and are now elevated from 5 to 15 m above the river. Below Larisa the slope of the land is gentle and the Girtoni covers much of the older surface. [...]
Early Neolithic land selection and farming practice
Clearly the Thessalian floodplains, as floodplains have done elsewhere, offered Neolithic farmers dry dwelling places and much arable land on abandoned levee/channel systems and many availed themselves of the opportunity. What specifically were the advantages that made this a sensible decision?
Floodplain soils
Active or abandoned levees and overwash fans (where floods break through the levee) have good, light, sandy and silty loam, easy to work and annually refreshed with water and new silt. The loam is suited to cultivation with primitive methods. The heavy, water-logged clay of the low wet backswamps, on the other hand, was useful only for grazing and the gathering of reeds and whithies; this environment has only recently yielded to modern technology. Pleistocene terrace and fan soils, usually gravelly or sandy with a poor water-holding capacity, were less attractive than the loam of the levees.
Frequency of flooding
In the Larisa basin rain, averaging 500-600 mm/yr, falls almost entirely in the winter. In the Pindos, source of the Peneios, the precipitation of up to 2000mm/yr also falls mostly in the winter months, much of it as snow. The main discharge of the river begins in November, reaching a peak in March and annually 3000 x 106 cu. M of water pass through the vale of Tempe. The annual variation, measured over 35 years, amounts to about one-fourth of the average and before present controls were installed extensive flooding was a yearly experience.
Values from the middle of this century do not necessarily apply to the time from about 9000 to 4000 BP that concerns us here. Still, precipitation in the early Holocene, although perhaps no greater than today, may have varied less from year to year then because it is subjected to a strong orographic effect and was buffered by the dense woodland later removed by human exploitation.
Spring floods and farming
In the Mediterranean climate of the plains the annual variation of winter and spring rains greatly exceeds that of high mountain run-off. Therefore, light soils of good water-holding capacity combined with reliable annual floods allowed cultivation of the lowlands with a higher expectation of a good harvest than rainfed farming on fans and terraces could provide. With the low yield and seed-to-harvest ratios of early grains ranging from 2.5 to 4, this may well have been the key to steady population growth. The difficulty of clearing the oak woodland on the terraces compared to the open vegetation of the floodplain may have offered an additional incentive.
We have no indication of the season of sowing and it would be difficult to say what the evidence for this might be, but Leake refers to spring-sown grain on the drying margins of Lake Ioannina. We also note that the often-heard objection that erosion during flooding would destroy crops sown previously does not hold. Except for a few wash-over fans, useless anyhow because of their coarse sediments, flooding is very gentle as is clear from the undisturbed, thinly laminated loam deposits of active and abandoned levee systems.
In the sparsely settled land of Neolithic Greece, a predictable harvest was critical because no help could be secured from other regions no stricken by drought. Only after better technology and crops more tolerant of drought opened a much wider range of soils was it possible to use the surplus of one region to insure survival in another. Thus Halstead's model of the control of surpluses as a force in the evolution of complex societies does not apply until the later Middle Neolithic at the earliest.
Spring floodwater farming has also been discounted because of supposedly poor yields on wet ground. This is not necessarily so; early cereals did flourish in many spring-wet soils. Also Davies & Hillman have identified populations of emmer (Triticum dicoccum), a main staple of the Early Neolithic in Greece, that tolerate wet soils. A well-adapted crop was therefore available that enabled Neolithic farmers to exploit the advantages of floodplains.
In a fine but neglected palaeo-environmental study of the early Neolithic Körös culture in Hungary, Kosse has pointed out that the floodplain landscape is also rich in natural resources, wild deer and boar, fish and shell-fish, water birds and a wide range of plants that coud be exploited to increase the security of subsistence.
Floodwater farming in other landscapes
Seasonal flooding is not limited to rivers. The reliable supply of floodwater at the right time of year which drew Neolithic farmers to the Larisa basin was available from other sources elsewhere in Greece and in the Near East, although usually to a more limited degree.
Of the alternatives, lakes in basins of internal drainage (poljes) which are common in many karst landscapes are the most important. Fed by winter and spring run-off or by seasonally rising springs, the lakes flood in the winter, then fall as summer comes, exposing a widening shore of fresh moist silt suitable for the growing of spring crops.
This is not mere speculation: it has long been known that in the past the margins of seasonally fluctuationg Greek lakes were farmed when the water was low. Colonel William Leake, indefatigable 19th-century traveller, described the profitable grain harvest from the seasonally dry bed of Lake Karla and noted that in Epiros grazing and the hay harvest were very productive on the emergent shores of Lake Ioannina.
We have suggested elsewhere that some Neolithic sites in the Peloponnese (e.g. Franchthi Cave, Alepotrypa and Lerna) used spring-watered coastal meadows. In the Levant also springs provided a reliable water supply sought by early Neolithic farmers.
Still, Neolithic settlement in Greece was far denser in Thessaly than at springs and on lake shores elsewhere. If those settings, of which Greece has many, suited early farmers just as well, why are their sites so sparse and small there? We attribute this to another quality of large floodplains: they permit a dispersed two-dimensional network of settlements and a far more efficient use of space than any alternatives and allow a higher population growth. In this respect the thessalian plain is almost unique in Greece.
Greek floodplains are usually covered with the sand and the gravel of torrential rivers, but the Peneios leaves most of its coarse sediment behind at the foot of the Pindos range in the western Trikala basin where its gradient drops sharply. In addition, the Kalamakiou and Rodia gorges, dam-like at flood stage, back up the waters and so further enhance the deposition of fertile loam.
Were floodplains attractive to early farmers elsewhere? This seems indeed to be so. Kosse, who understood floodplains, also noted a preference for the crests of natural levees by Neolithic settlers in the Great Hungarian Basin stating, "all Körös sites were located immediately by rivers and lakes". She calls the culture a "floodplain adaptation in a fairly dry climatic region". Sherratt agreed, saying that the Körös (c. 7500 BP) setlers had "a particular preference for the levees immediately adjacent to the Palaeo-channels".
Barker, in discussing the Neolithic landscape of southern Yugoslavia, noted that many sites occur in large poljes such as Bitola (Macedonia), perhaps in order to exploit the margins of seasonal lakes. Others are found in valleys between what he called, Higgsian site-catchment terms, 'light and heavy arable' (translation: levee and backswamp) deposits. Dennell & Webley's fine study of Neolithic land use in southern Bulgaria also stresses riverine soils. In eastern Italy, the Tavoliere is also similar in many ways to the Thessalian basins. The region contains some thousand Neolithic sites, largely placed on the composite floodplain of many small rivers.
We believe therefore that the emphasis on floodplain and wet bottomland environments and the evidence discussed above justify a new working hypothesis concerning land use and early Neolithic migration.
Discussion
Who were the farmers who so desired a reliable natural irrigation that they chose to live in active floodplains?
For more than a century, early European agriculture has been regarded as derived from the Near East. In the 1920s and 1930s Gordon Childe developed a model for the diffusion of agriculture from Near Eastern oases to Europe by direct cultural contact. A considerable similarity in Neolithic material cultures from the Near East and Southeastern Europe supports this view. More recently Ammerman & Cavalli-Sforza showed that the earliest Neolithic sites are in southeastern Europe and systematically decrease in age westward. They attributed this to the spread of a farming population (demic diffusion) described with a wave-of-advance model developed originally for biology.
In the 1970s another view gained favour which, chiefly on theorical grounds, accepted an independent indigenous development. "Early farming in Europe 'always' occurred in areas where there were already hunter-forager communities. These cannot be regarded as irrelevant to the pattern of agricultural expansion". This statement sums up the views of the 'indigenists' as Ammerman has called them, and it is one with which we firmly disagree.
We have shown at the outset that the Neolithic advance in Greece and the southern Balkans proceeded mainly in areas not occupied by an indigenous Mesolithic population, while farther west and north such a population did exist. The difference, rarely stated explicity, has bedevilled the dialogue between demic diffusionists and indigenists. The debate on demic diffusion by Ammerman and Zvelebil is revealing in this regard, but to avoid entanglement with events in central and western Europe, we limit ourselves here to the early Neolithic of Greece and adjacent southeastern Europe.
Ammerman pointed to three propositions essential to the indigenist model:
1 there was a settled Mesolithic population ready to accept farming as a way of life;
2 late Mesolithic and early Neolithic population densities were similar in any given region, and
3 continuity existed in settlements across the region.
As we have noted, these conditions are not met in Greece where the evidence indicates that agriculturists settled almost solely in regions without a history of Mesolithic occupation.
Therefore, while accepting the essence of the demic diffusion model, we focus on the preference for floodplains by arriving farmers and their apparent success there. What does this preference mean in terms of their origin and how does it affect the demic diffusion hypothesis and the wave-of-advance model of Ammerman & Cavalli-Sforza?
Traditionally, the origins of Neolithic farming have been sought in the Levant or southern Anatolia. The exact location is not important in this discussion, but Anatolia being the closest, it is there that we seek the people who selected with such care environments that guaranteed an adequate annual harvest, either on the margin of seasonally rising and falling lakes or in the floodplains of rivers.
Not a great deal is known about settlement patterns and land preferences of the Anatolian Neolithic. In the west and north many medium-sized rivers have floodplains, but while future exploration may turn up Neolithic settlements there, so far they have not be seen. The only area known to be well populated in the early Neolithic are the Beysehir-Sugla and Konya basins in the south-central part of the country. At least 9500 years ago, the fine-grained alluvium of fan deltas and the margins of seasonal lakes in those basins provided a reliable harvest for a sizeable population, and a network of villages such as Can Hasan, Süberde and Çatal Hüyük spread across them.
The similarity of this setting to the lands preferred by immigrants into Greece and elsewhere in southeastern Europe is obvious and for the next step in our argument we take south-central Anatolia as our starting point. We shall compare the pattern of the earliest settlements (Figure 1) with the wave-of-advance model developped by Ammerman & Cavalli-Sforza.
In its simplest form this model (Figure 2A) rests on two assumptions:
1 an initially logistic population growth curve which yields a continuous advance across a broad front, and
2 a local migratory activity that is, to a first approximation, continuous and random in direction.
Population increase and migratory activity occur only at the wave front and the rate of advance is roughly constant. Well behind the wave front the population growth slows due to lack of room for expansion. Ammerman & Cavalli-Sforza, while recognizing several variants on this theme, regarded its basic form as adequate for a series of tests based on genetics and chronology. Those tests did not conflict with th model.
To increase the realism of the model other factors must be taken into account besides demography (Figure 2B). More complex models were indeed analysed by Ammerman & Cavalli-Sforza which included discontinuities in space and time as well as barriers and areas unsuited for exploitation. Those models are our point of departure here.
The Aegean Sea might be regarded as a barrier (as are mountains and deserts), but travel by sea was apparently not by itself a problem. It is swift and accelerates migration and shapes the wave front, but it is also a bottleneck that limits the number of migrants. Colonists arrived early and almost simultaneously on Crete, at Franchthi and in Thessaly (Figure 1), but probably only in small numbers.
More critical was the preference of the migrants for environments that were far from ubiquitous. In a landscape where only scattered and usually small patches were inviting, the wave filled up only the desirable areas, then leapt to the next ones. This converted the wave front from a smooth bulge into the tongues of the incoming tide as it first advances across the sand. In terms of overall advance the basic model is a useful approximation, but the dynamic of the modified version is different.
The original model implied a continuous movement driven by steady population growth immediately behind the front. In its modified form (Figure 2B), the small number of shipborne arrivals at large suitable sites such as the Larisa basin had much time before another move became necessary or even possible and so migration occurred discrete steps. The lengths of each step and the intervals between them were dictated by geography and by the population growth in each of a lowly rising number of parent areas.
Of course, in Greece other environments that floodplains permitted settlement and survival in many but widely separte places. It is our contention, however, that only the few large, fertile floodplains, such as the Larisa basin, supported populations ultimately large enough to start the next migratory move. In later stages of the migration the Morava-Vardar area in the Balkans and the tavoliere in east-central Italy may have been similar growth and jumping-off points (Figure 1).
Table 1 provides some insight here. During the 1000-year span of the Early Neolithic just over a hundred sites were established in the Larisa basin, probably gradually. Of those one-third did not survive into the Middle Neolithic, but as many new ones were founded before the end of this phase. The Early and Middle Neolithic intervals, ending about 7500 BP, thus seem to have been a time of steady but not very rapid growth, in accord with a mere 25% increase of the average site area.
This changed drastically during the Late Neolithic (7500-6500 BP). The 30 sites that failed to survive the Middle Neolithic were replaced by many more new ones, and the total in the Larisa basin rose by half to 150 before the Bronze Age brought a reduction. There was also a modest but real increase in the number of large sites. Many new sites were established outside the floodplain on fans and on the flanks of the Ravenia ridge between the Trikala and Larisa basins. Perhaps we may see here the impact of new crops and technology better adapted to dry-land farming.
The still unsatisfactory state of the Neolithic chronology in Greece frustrates a quantitative analysis of the growth of the initial settlement, but even the Larisa basin, region of major growth, required some 1500 years, from about 9000 to 7500 BP, to reach saturation. It may be no coincidence that the first settlements in Yugoslavian Macedonia and southern Bulgaria date to c. 7800 BP (Figure 1), implying that a new spill-over had begun. This time the origin was Thessaly and a little later, for Bulgarian sites, the Morava/Valdar area.
Local population increase has often been regarded as the main force that drove the Neolithic migration, an assumption explicit in the wave-of-advance model. Figure 1 shows that only six centuries after they settled south-central Anatolia the first farmers arrived in Greece, almost a thousand kilometres across the sea. The rapidity of this new phase of movement cannot be comfortably attributed to population growth, even if we accept a high rate of growth at the front of the wave of advance. There is, moreover, no evidence for a dense population in Anatolia 9000 years ago and many less distant areas were open to settlement to relieve any population pressure. Neither does the settlement pattern of the migration into southeastern Europe (Figure 1) suggest the crowding that would force another step.
Our modified model partly resolves this paradox, because the limited availability of preferred lands at a considerable distance from each other does not imply a large population at the point of origin. It so explains the lack of evidence for a wide-spread dense population exhausting available resources, but it raises new questions. Was it an overwhelming preference for specific environments that motivated people to go far across the seas into the unknown rather than to adjust to lands that were not yet scarce at home? How did these migrants locate the thessalian plains which high mountains shield from view by land and sea? Or were those lands perhaps not so unknown, because earlier wandering seafarers had already found them as Broodbank & Strasser have suggested and as the use by Neolithic farmers of Melian obsidian known since the Mesolithic implies?
These questions suggest that, as we move beyond the basic, heuristically fertile wave-of-advance model, understanding the dynamic of the introduction of farming into Europe will require a firm chronology of the Neolithic and the identification of detailed patterns of Neolithic settlement and its environments. Such information will permit the quantitative estimates of population growth and resource exploitation that are essential to explain fully the spread of agriculture across Europe.
