UDK 903(4-5)“633/634”:572.1/.4

Documenta Praehistorica XXX 

A new model for the spread 
of the first farmers in Europe 

Ron Pinhasi

Department of Biological Anthropology, University of Cambridge, UKron.pinhasi@chello.at 

ABSTRACT – The appearance and dispersion of the first farmers in Europe has been the subject of 
heated debate among anthropologists, archaeologists, and linguists for over a century. There is no 
consensus regarding two main aspects: (1) the extent to which the transition to farming was an indigenous process, and (2) the historical pattern in terms of the timing and tempo of the dispersion 
events. Morphological variability and affinities are assessed among Mesolithic and Early Neolithic populations of the Near East, Anatolia and Europe. Statistical results reveal regional and temporal differences in the dispersion process among these populations. Based on these results, a new model is 
presented for the spread of farming in Europe. 

IZVLE
EK – Pojav in 
irjenje prvih kmetovalcev v Evropi je e ve kot stoletje predmet vroih razprav med antropologi, arheologi in lingvisti. Zaenkrat 
e ni soglasja glede dveh glavnih vidikov: (1) 
do kak
ne mere je bil prehod v kmetovanje domoroden proces in (2) kdaj se je zaelo in kako hitro se je kmetovanje raz
irilo. V lanku ocenimo morfolo
ko raznolikost in sorodnost mezolitskih in 
zgodnje neolitskih populacij Blinjega vzhoda, Anatolije in Evrope. Statistini rezultati kaejo, da so 
med temi populacijami v procesu 
irjenja regionalne in asovne razlike. Na temelju teh razultatov 
predstavimo nov model raz
irjanja kmetovanja v Evropi. 

KEY WORDS – craniometric analysis; biological distances; morphological variability; Mesolithic; Early 
Neolithic; Europe; Levant; Anatolia 

INTRODUCTION 

There is a view prevailing among archaeologists, an-ceful and unique avenue to the study of our past. 
thropologists, geneticists, and linguists that skeletal Each discipline has its drawbacks and misconceptistudies of population affinities are a legacy of the ons. Some of these are the outcome of the type of 
past. Such scholars tend to point out that past stu-data under examination, others are the result of misdies of this sort were associated with the study of conceptions and prejudiced assumptions accumulahuman races and eugenics and are thus better off ted throughout the history of a particular scientific 
being put away. Another prevalent assumption is discipline. 
that as bone growth and development is strongly 
affected by various environmental factors such as The field of biological anthropology has a particuclimate, diet, and activity, it can tell us nothing about larly infamous history. One of the outcomes of decagenetically based biological variability among popu-des of racist research is a tendency among researlations. chers from other related disciplines to question the 

validity of biological anthropology as a scientific di-
These assumptions are just as outdated as the study scipline. The abolition of the concept of races and 
of human races. The application of skeletal techni-present emphasis on the uniformity of the human 
ques to the study of human populations is a resour-species is, in a sense, a positive consequence. How


Ron Pinhasi 

ever, such an approach disregards the fact that there 
exist both phenotypic and genotypic differences 
among human populations. Emphasising the fact 
that past and present human populations share biological similarities undermines the notion that each 
of these populations possess some unique biological 
variability. The fact that the concept of human races 
should be discarded does not imply that there are 
no significant differences in biological variability 
across space and time. 

There exist relatively large Neolithic skeletal collections from localities in Europe and the Near East. 
Most of these skeletons are from well-stratified, radiocarbon dated archaeological contexts. These contexts therefore provide the researcher with comparatively accurate spatial and temporal attributes. Archaeological data provides invaluable information 
about past humans’ behaviour and human-environment relationships. The archaeological record of Mesolithic and Neolithic populations is particularly rich 
and the analysis of these data in the context of human evolution is of great potential. 

Renfrew (1992) pointed out that perhaps the greatest advantage of cranial data is its time span. Thus, 
anthropometric analysis can be grounded on the 
one hand in the archaeological background which 
provides the relevant historical context, and modern 
genetics on the other, which provides a reference to 
both past biological variability and past population 
structure and affinities (e.g. maps of clinal distributions of genetic traits). 

At present there is a lack of consensus regarding the 
following aspects: 

 The extent to which the transition to farming was 

an indigenous process, involving some admixture 

between incoming farmers and local hunters, or 

a population replacement process. 
 The historical pattern in terms of the timing and 

tempo of the dispersion events. 

The first stage of the project involved the analysis of 
archaeological data from approx. 2200 sites/contexts. The data was analysed in regard to its spatial 
and temporal attributes (i.e., radiocarbon dates, and 
geographic location in longitude/latitude) as well as 
for archaeological culture (Pinhasi 2000). Emphasis 
was placed on observing the temporally based distribution of Mesolithic vs. Neolithic occupation across 
Europe by using Arcview GIS application. Maps of 
settlement patterns by chronological period are depicted in Figures 1–7. 

The analysis of archaeological data yielded the following results: 

 The density of Mesolithic occupation at the time 
of the appearance of farming varied greatly between regions. 
 The timing and density of Neolithic occupation 
also varied between regions. 
 The dynamic pattern of Mesolithic colonization 
will tend to obscure any clinal trends that are presently attributed to the Neolithic ‘wave of advance’. 
 Any arriving farmers would have met very competitive and interactive conditions with a divergent 
potential for gene flow. 
 The variability in population densities had a genetic effect on immigrant Neolithic farmers, and 
this should be modelled at a regional level. 
Five models for the spread of farming in Europe 
were developed, partially on the basis of the results obtained, and partially on the basis of other 
prevalent theories (Ammerman & Cavalli Sforza 
1984; Zvelebil & Rowley-Conwy 1984; Zvelebil 
1986; Zvelebil & Zvelebil 1988; Zvelebil & Dolukhanov 1991; Zilhao 1993; Dennell 1985). The first part 
of this paper will expose these models and specify 
the corresponding expectations of each model in regards to the ‘neolithisation’ process across Europe. 

Three main aspects of ‘Neolithisation’ will be examined in this paper. The first aspect begins with the 
assessment of the relationships among the Mesolithic groups in order to examine whether some of 
these groups are similar enough (morphologically) 
to be regarded as a single biological population. The 
section proceeds to examine affinities during the 
Early Neolithic period. The questions being addressed are as follows. What type of morphological affinities can one detect among the Early Neolithic populations? Are some groups similar enough to be regarded as belonging to a single population? This 
question is therefore identical to the one for the Mesolithic. However, morphological affinities must be 
interpreted in the specific context of the specified 
models. Particularly, the question of greatest relevance is whether a pattern of morphological affinities is compatible with a demic dispersal. 

The second aspect is the specification of a plausible 
area of origin from which the first farmers spread to 
Europe, and whether a linear correlation between 
distance from source area and morphological differentiation exists, as suggested by the findings of Menozzi & Cavalli-Sforza (1978). The two centres exa


A new model for the spread of the first farmers in Europe 


Fig. 1. Settlement pattern during the period 14 000–12 000 BP.



Fig. 2. Settlement pattern during the period 12 000–10 000 BP.



Ron Pinhasi 


Fig. 3. Settlement pattern during the period 10 000–9000 BP.



Fig. 4. Settlement pattern during the period 9000–8000 BP.



A new model for the spread of the first farmers in Europe 


Fig. 5. Settlement pattern during the period 8000–7000 BP.



Fig. 6. Settlement pattern during the period 7000–6000 BP.



Ron Pinhasi 


Fig. 7. Settlement pattern during the period 6000–5000 BP.



Fig. 8. The division of the studied groups according to geographic regions.



A new model for the spread of the first farmers in Europe 

mined are SE Anatolia, represented by the population from Çayönü, and Central Anatolia, represented 
by the population of Çatal Höyük. The possibility of 
other locations (Levant, Cyprus) is rejected on the 
basis of the evidently large distances between the 
PPNB populations from these regions and the Early 
Neolithic European populations. 

In the last part of this work a regional-based analysis of craniometric data is performed. The underlying hypothesis is that the process of neolithisation 
differed across space and time and thus needs to be 
assessed in greater depth (see Pinhasi 2000; Lahr et 
al. 2000). This is achieved by examining the degree 
of regional continuity between local hunters and incoming farmers in the various regions of Europe as 
a series of case studies. Temporal variation in spatial 
units is examined by modelling expected patterns 
of continuity (i.e., admixture) and discontinuity in 
various regions. The placement of a given specimen 
in a specific region was solely based on the geographic location of its associated site. The demarcation 
of the regions is depicted in Table 1 and in Figure 8. 

MODELS FOR THE SPREAD OF FARMING 
IN EUROPE 

Model 1 – Autochthonous transition across 
Europe 

This model proposes that the transition to agriculture occurred as a series of local events that took 
place in different locations and time across Europe. 
The model is based on the idea of ‘Neolithisation’ 
through a process of cultural diffusion but without 

any migration or expansions of farming populations. 
This model fits with the theory of in situ transition 
and cultural rather than demic diffusion of the Neolithic across Europe (see Whittle 1996; Dennell 1985 
and others). The model is represented in Figure 9. 

Under this model, regionally based changes in morphological variation and changes in sexual dimorphism are the outcome of an adaptive response to 
environmental changes such as changes in diet, activity pattern, climate and mobility, and/or changes 
in mating networks. Thus, changes are not expected to be external (i.e. gene flow from migrations), 
but rather internal. In such a case, changes should 
be more pronounced at the period between stage 1 
and stage 2, during which the Early Neolithic cultures appear across Europe. This is because this model places emphasis on a culturally-based transition. 
The archaeological record attests to the fact that in 
the case of most of the regions of Europe changes in 
lifestyle, mobility pattern and mating networks occurred during the initial transition to agriculture (i.e., 
during the Early Neolithic period). Less ‘functional’ 
morphological changes are expected among populations that were already ‘fully Neolithic’, and little 
would have changed during the subsequent period 
in terms of the above-mentioned factors. 


Fig. 9. A model of autochthonous transition across 
Europe. 

Region 
Description

1 Anatolia and the Levant 
Consisting of modern-day Turkey, Syria, Lebanon, and Israel. 

2 The Alpine region 
Consisting of present-day Austria and Switzerland, and the Alpine regions of Italy and France. 

3 Northern Europe 
Consisting of all specimens between longitude 6°E and 26°E, and north of latitude 51°N. Itincludes the area of present-day northern Belgium and Holland, Northern Germany, Poland, 
Scandinavia, and the Baltic states. 

4 Central Europe 
Consisting of Slovakia and the Czech Republic, south and central Germany up to latitude51°N, and mostly falling between 48°N and 51°N, and east of longitude 6°E. 

5 South East Europe 
Including mainland Greece and the Greek Islands, Bulgaria, Romania, Serbia, Macedonia,
Bosnia, Croatia and Hungary. 

6 Mediterranean Europe 
Including the Adriatic coastal region of Albania, Bosnia and Croatia, all of Italy (with the exception of the Alpine region), Corsica, Sardinia, and Mediterranean France. 

7 Atlantic Europe 
Including Great Britain, all parts of France and Belgium west of longitude 6°E and north oflatitude 44°N, Majorca, Spain and Portugal. 

8 Dniepr Region, 
Including the Mesolithic and Neolithic cultures of southern Ukraine. 

Tab. 1. Description of the regions analysed. 


Ron Pinhasi 

Model 2 – Incoming farmers with differential 
degrees of admixture 

Model 2a – Differential admixture across space 

This model is in congruence with the ‘demic diffusion’ model. The underlying assumption is that the 
Neolithic farmers spread across the continent and 
‘absorbed’ differential amounts of genes from indigenous Mesolithic populations (Fig. 10). Note that 
this model does not differentiate between scenarios 
which assume varying degrees of demic input from 


Fig. 10. A model that assumes a process of neolithisation with incoming farmers (EN) with differential degrees of admixture with local Mesolithic populations across space. 

the Mesolithic populations. Thus, the degree of admixture may have been weak or extensive. Nevertheless, the underlying assumption is that the spreading 
farmers absorbed Mesolithic populations along their 
path and not a scenario in which the greatest contribution to the Neolithic gene pool came from Mesolithic populations. 

Model 2b – Differential admixture across time 

– “Delayed Admixture” 
The second variant of model 2 is one which adds the 
element of delayed admixture. The underlying assumption is that the Early Neolithic farmers did not 
admix with the Mesolithic hunters, but rather that 
an admixture occurred later on during the Middle/ 
Late Neolithic period (Fig. 4). During the Early Neolithic period, the farmers would have undergone a 
process of ‘consolidation’, during which they expanded across the continent and intensified and diversi

fied their lifestyle. Subsequently, as the settlement 
pattern became denser and population density increased, mating networks and acculturation between 
Neolithic and Mesolithic populations became more 
prevalent. Consequently the genetic contribution of 
Mesolithic groups to the Middle/Late Neolithic populations would have been significant. 

Model 3 – Complete replacement without 
admixture 

This model assumes a demic diffusion process without any admixture. In figure 12 the Mesolithic populations are represented as moving northwards. Another possibility is their extinction without any genetic contribution. Consequently, all Early Neolithic populations are regarded as being directly affiliated 
with the Levantine and/or Anatolian Early Neolithic 

Fig. 12. A model of complete replacement, with 
some extinction and dispersion of Mesolithic popu-
lations northwards. Thatched lines represent ex-
tinction. 
populations and share no affinities with any of the 
European Mesolithic populations. Changes in the genetic composition of the dispersing populations due 
to stochastic effects (series of founder effects, isolation, etc.) are ruled out from this model as it assumes 
sufficient gene flow along the dispersion path. 

Model 4 – Admixture as a function of geographic region and distance from the source population 

This model is a variant of model 2a (Fig. 13). However, ‘differential admixture’ is assessed as a function 
of the geographic region under examination and its 

A B 
Fig. 11. A model that assumes a process of neolithisation with incoming farmers (EN) with differential 
degrees of admixture with local Mesolithic populations across time. A: 8000–6000 BP (Early Neolithic). 

B: 6000–4000 BP (Middle/Late Neolithic). 

A new model for the spread of the first farmers in Europe 


Fig. 13. Differential contribution of Mesolithic populations to incoming farmers across space and 
time. Contribution varies by region and with corresponding admixture levels varying in a scale between no contribution to intensive admixture. 

distance from the Near East/Anatolia. The model 
therefore places emphasis on examining the ‘Neolithisation’ process according to regions rather than 
as a pan-European phenomenon. Therefore, any theory which proposes a complex geographically based 
scenario with a differential admixture due to the impact of geographic/spatial factors is congruent with 
this model. 

MATERIALS 

The skeletal material consisted of the following: 

 Data collected during fieldwork. These data com

prised of: 

a. Cranial measurements. 
b. Mandible plus lower dentition 
measurements. 
 Data collected from the literature and from the 
ADAMS database, Department of Anthropology, 
University of Geneva, Switzerland. 

Description of the data is provided in Table 2 (a = 
Personal data, b = ADAMS, c = published data). 

METHODS 

All measurements were taken from the skull, mandible and teeth of each individual. In the case of the 
majority of the analyses performed, only a subset of 
the total set of variables was utilised. The subset of 
measurements utilised in the following analyses is 
described in Table 3. 

The sex of specimens studied by the author was evaluated by using three systems. 

 A general assessment of sex based on the criteria 

outlined in the book Standards for Data Collec

tion from Human Skeletal Remains, by J. E. Buik

stra and D. H. Ubelaker (1994). 
 An assessment of sex by means of a plasticine 

cast of the endocranial part of the external audi

tory meatus (Joachim Wahl, Landesdenkmalamt 

Baden-Württemberg, pers comm.). 
 An assessment of sex following the method deve

loped by Graw and co-authors (1999). 

Age estimation was based on the age assessment in 
the available registry books at the given institution, 
as well as an assessment of age based on the skull 
carried out by the author. The two main criteria for 
the assessment of age were the degree of cranial suture closure (specifically basi-occipital-sphenoid synchondrosis, as described in Acsádi and Nemeskéri 
1970), taken together with the presence of fully 
erupted second molars. This ‘system’ is in many respects inaccurate (see discussion in Schwartz 1995. 
206–211 regarding age assessment techniques using 
the skull). However, it is adequate to exclude the 
majority of sub-adult specimens. Thus, the overwhelming majority of skulls being measured were those 
of adult specimens. 

STATISTICAL METHODS 

The following is a brief description of the statistical 
methods utilised in the following sections. 

Univariate methods 

a. Kruskal-Wallis test 
The Kruskal-Wallis ANOVA by ranks test assumes 
that the variable under consideration is continuous 
and that it was measured on at least an ordinal (rank 
order) scale. It tests the null hypothesis that the different samples in the comparison were drawn from 
the same distribution or from distributions with the 
same median. Thus, the interpretation of the Kruskal-
Wallis test is basically identical to that of parametric one-way ANOVA, except that it is based on ranks 
rather than means. This procedure is more robust 
(albeit less sensitive) as it uses a rank order scale, 
and therefore can be applied to small sample sizes. 

b. Analysis of boxplots 
The analysis of boxplots is a straightforward procedure. The common boxplots (in SPSS) compare ranks 
rather than means and thus depict differences in 
group medians. It also allows the comparison of interquartile ranges per variable. 

Bivariate methods 

Following the univariate procedures, part of the data 
was scrutinised using bivariate analyses. In each of 


Ron Pinhasi 

Fig. 14. Minimum Frontal Breadth vs. Zygomatic 
Breadth for individual Mesolithic specimens, by 
group. 
the bivariate analyses, the individual data for a pair 
of variables (measurements) were plotted in a scatterplot. Using these scatterplots, the position of single finds and the variability of groups can be visually assessed. The simplicity of the scatterplot procedure is that it does not amplify between-group variability or distort the output. 

Multivariate methods 

a. Discriminant analysis 
This statistical procedure is in fact a set of closely 
related procedures under one broad term (Klecka 
1980). The procedure is applied when there exist 
two or more samples from potentially different populations and the researcher wishes to distinguish 
among them. Discriminant function analysis has two 
main applications: 

 Interpretation of the ways in which the groups 

differ from each other. Is one able to discriminate 
between the groups on the basis of certain characteristics? How well do the groups discriminate, 
and which characteristics are the most powerful 
discriminators (Klecka 1980). 

 Classification: to predict group membership from 
a set of predictors, the derived equations (canonical functions) combine the groups’ characteristics in a way that will allow one to identify the 
group to which each case most closely approximates (in terms of the relation of the individual 

case values for a set of traits and the average value of those traits for each of the groups). The 
case under examination may be of either a known 
or an unknown group (Tabachnick and Fidell 
1996), thus allowing both the comparison of the 
accuracy with which certain traits allow identifying the group of origin of known cases, and the 
potential classification to known groups of ungrouped cases. 

b. Principal Components Analysis 
Principal components analysis (PCA) is a powerful 
exploratory technique in which a large number of 
variables are reduced to a smaller number of factors 
(Tabachnick and Fidell 1996). The multivariate 
technique of principal components analysis is usually applied for the purpose of data reduction and 
de-correlation of the variables. However, in this 
work, principal components analysis is mainly applied as an exploratory tool in the search for underlying patterns/structures of relationships between 
discrimination and association of past populations 
and corresponding specific morphological features. 

c. Squared Mahalanobis Distances 
The generalised distance, D2, developed by Mahalanobis, provides an effective measure for estimating 
group differences between biological populations. 
The Mahalanobis distance statistic is often applied in 
the analysis of prehistoric populations (Van Vark 

Fig. 15. Nasion-Prosthion Height vs. Begma Height 
for individual Mesolithic specimens, by group. 

A new model for the spread of the first farmers in Europe 

and Schaafsma 1992; Howells 1973; 1989; Keita 
1990; 1992). It is regarded by many anthropologists 
and biologists as being the standard and most effective measure of distance between two or more 
populations when the observed data are quantitative (Bedrick et al. 2000). It has been used by various researchers in the study of affinities within and 
between prehistoric populations (see for example, 
Bräuer & Rimbach 1990; Van Vark et al. 1992). 
The Mahalanobis distance is an effective measure 
when variables are correlated, and it includes both 
variances and correlations (Campbell 1984). According to Van Vark and Schaafsma (1992), Mahalanobis distances calculated from skull measurements 
can be used in order to trace historical events such 
as population influx, admixture, and drift. Thus, 
while morphometric similarities among groups/individuals are not the same as genetic relatedness, 
there exists a significant correlation between the 
two. It follows that Mahalanobis distances between 
samples of ancient populations are to a certain extent genetically based, and their interpretation in 
terms of between-population genetically based affinities is valid (Van Vark and Schaafsma 1992). 

PART 1 – ASSESSING POPULATION AFFINITIES 
BY PERIOD 

Affinities between the Mesolithic populations 
of Europe 

The bivariate graphs investigate the relationship 
among the European Mesolithic groups. The question in mind is whether one can detect any affinities/patterns between groups. 

The bivariate graph (Fig. 14) indicates that, with the 
exception of one specimen, the Danube Gorge material falls within the range of variation of the French 
and the Italian Mesolithic. The Iberian and French/ 
Danube/Italian samples have a relatively small overlap (the largest Iberian with the smallest French/Danube/Italian), while the German remains also show 
comparatively small values of min frontal breadth. 
The under-representation of individuals from southern France and Greece does not permit the drawing 
of inferences regarding the morphological variability 
and association of these groups. 

The pattern observed in the scatterplot on Figure 15 
is a much greater overlap among all groups, with 
the German variation encompassing almost all other 
variation. Individuals that belong to the Danube 

Gorge group are relatively clustered as is the case 
for Muge and Greece. In contrast, we see that the specimens of the German Mesolithic and Italian Mesolithic groups are more scattered. 

The scatterplots suggest some regionally-based discrimination between the European Mesolithic groups, 
with a considerable element of intra-sample variability. We therefore do not see a clear-cut morphological differentiation. Scatterplots do not apply distance algorithms that intend to maximise among-
group differences while minimising the effects of intra-sample variance (such as the squared Mahalanobis distance). One, therefore, may conclude at this 
stage that the Mesolithic groups of Europe are morphologically diverse. No visible patterns of affinities 
were observed. 

Affinities between the Early Neolithic populations of 
Anatolia, the Levant, south-east and central Europe. 

In this section we examine morphological relationships among the Early Neolithic populations of Anatolia, Levant and south-east and central Europe. The 
main issue that we opt to tackle may be phrased as 
follows. 

What is the relationship among the Early Neolithic 
groups from the Levant, Anatolia and the Early Neolithic groups from the various parts of Europe? 

We are now looking for more definite answers regarding the population relationships during the period of the appearance of the first farmers in Anatolia and the Levant, and subsequently in south-east 
and central Europe. We expect to detect: 

 Some similarities between many of the Early Neo

lithic European groups. 
 Some of the Early Neolithic European groups are 

morphologically similar to one or more of the Le

vantine, and/or Anatolian groups (in the case of 

Demic Diffusion). 

Univariate analysis 

Table 4 shows the groups (the Natufian sample and 
Early Neolithic) and corresponding codes used in 
the analyses. 

A Kruskal-Wallis Anova rank analysis was carried 
out in order to assess whether significant differences between the groups for the following set of variables exist: GOL, XPB, ZYB, NPH, NLH, NLB, and 
OBH. This non-parametric method is preferable over 


Ron Pinhasi 

Lat Long Location N Date Arch Period Zone Database Collection/Reference 

35.87 38.4 Abu Hureyra 5 10 790 PPNB 1 a 1

32.76 35.36 Ain-Mallaha (Eynan) 3 11 547 Epipaleolithic 1 a 1

30.23 35.53 Basta 3 PPNB 1 a 3

34.12 35.65 Bayblos 2 Chalcolithic 1 c Kurth 1973 

37.1 32.13 Çatal Höyük 49 7499 Neolithic 1 c Ferembach 1981 

38.23 39.65 Çayönü 8 9360 PPNB 1 a 1

31.42 35.06 Erq-El-Ahmar 2 11 000 Epipaleolithic 1 a 1

32.67 35 Fallah- Nahal Oren 9 10 046 Epipaleolithic 1 a 1

32.9 35.22 Hayonim 4 12 010 Epipaleolithic 1 a 1

32.00 35.00 Jericho 9 Chalcolithic 1 c Kurth 1973 

32.8 35.33 Pkein 29 Chalcolithic 1 a 1

48.47 20.52 Aggtelek 1 Neolithic 5 a 7

47.30 19.03 Agostonpuszta 1 Eneolithic 5 a 7

47.00 17.00 Ajaki Kokornaki 2 Eneolithic 5 a 7

47.18 19.12 Alsónémedi 5 4500 Badeni 5 a 7

41.39 21.58 Anza 3 EGEAN 5 b ADAMS

47.32 19 Aszód-Papiföldek 6 5669 Neolithic 5 a 7

37.99 23.73 Athens 1 Neolithic 5 c Boev 1973 

46.83 17.98 Balatonendried 1 Neolithic 5 a 7

46.7 21.26 Békés-Povád 4 Szakálhát 5 a 7

39.60 22.50 B'Koybeleiki 2 Neolithic 5 a 17

48.00 21.00 Bodrogkersztur 1 Eneolithic 5 a 7

47.30 19.03 Budakalász 21 Badeni 5 a 7

47.30 19.08 Budapest 1 Neolithic 5 a 7

47.30 19.06 Budapest-Andor 1 Badeni 5 a 7

44.12 26.46 Cascioarele, Locus 1 1 5598 Neolithic 5 a 5

44.18 28 Cernavoda 21 Hamangia 5 b ADAMS

46.3 24.12 Cipa 1 Cris 5 b ADAMS

47.25 18.95 Csepel Sziget 3 Neolithic 5 a 7

46.4 20.31 Csóka 1 Neolithic 5 a 6

46.22 20.25 Desyk-Olajkut 2 6570 Körös 5 a 6

46.12 20.15 Deszk 2 Cris 5 b ADAMS

43.23 24.95 Deveta_kata Pe_tera 1 Early Neolithic 5 b ADAMS

46.94 20.78 Endröd 1 6566 Neolithic 5 a 639 22.3 Franchthi Cave 3 Early Mesolithic 5 b ADAMS

44.35 23.91 Gîrlsti 2 Neolithic 5 a 5

44.08 26.63 Gumelnita 1 5557 Gumelnita 5 a 5

46.48 23.36 Gura Bacului 1 6650 Cris 5 b ADAMS

46.64 21.32 Gyula 1 Neolithic 5 a 7

37.42 22.18 Hageorgitika 1 Neolithic 5 b Boev 1973 

46.36 20.18 Hódmezövásárhely 1 6450 Körös 5 a 6

46.39 20.39 Hódmezövásárhely 3 6190 Late Neolithic 5 a 6

46.39 20.39 Hódmezövásárhely–B. 1 Körös 5 a 6

46.4 20.31 Hódmezövásárhely–G. 4 6050 Late Neolithic 5 a 6

43.00 27.00 Janka 8 5700 #NAME? 5 b ADAMS43 24 Jasa Tepe 1 6600 Karanovo 2 5 b ADAMS

42.36 25.54 Karanovo 2 6600 Karanovo 1 5 b ADAMS

42.36 25.24 Kasanlak 1 6500 Karanobvo 2 5 b ADAMS

34.54 33 Khirokitia 21 7350 PPN 5 b ADAMS

47.5 20.5 Kisköre-Gát 7 5942 Neolithic 5 a 7

47.00 19.50 Kkotac Tanya 2 Eneolithic 5 a 6

46.41 20.31 Kotacpart 7 6450 Körös 5 a 6

46.23 18.21 Lengyel 1 Neolithic 5 a 14

44.3 22.06 Lepenski Vir 16 7300 Mesolithic 5 b ADAMS

44.33 22.03 Lepenskivir 23 Starcevo 5 b ADAMS

45.00 18.00 Mogyorós 1 Neolithic 5 a 6

40.65 22.3 Nea Nikomedeia 13 8180 Early Neolithic 5 c Angel 1973 

47.73 18.37 Neszmély 1 Neolithic 5 a 7

46.38 18.57 Paradicsompuszta 1 Lengyel 5 a 7

48.14 20.67 Paszar 1 Eneolithic 5 a 7

47.00 19.30 Puszaistvánháza 3 Eneolithic 5 a 7

43.54 26.00 Rusé 3 5200 Mound Culture 5 b ADAMS

45.54 25.18 Sf.Gheorghe Bedehaza 1 Cris 5 b ADAMS

42.4 23.18 Sofia 2 Early Neolithic 5 b ADAMS

47.06 21.54 Solca 1 Cris 5 b ADAMS

44.49 20.42 Starcevo 1 Starcevo 5 b ADAMS

47.47 18.42 Sturovo 5 Zeliezovce 5 b ADAMS

46.29 20.26 Szegvár-Tüzköves 1 6000 Neolithic 5 a 7

47.67 19.08 Szentendre 1 Neolithic 5 a 7

38.52 23.98 Tharounia 1 E. Neolithic 5 a 17 


A new model for the spread of the first farmers in Europe 

Lat Long Location N Date Arch Period Zone Database Collection/Reference 

39.68 21.68 Theopetra 1 E. Neolithic 5 a 17
39.68 21.68 Theopetra 1 Mesolithic 5 a 17
47.17 26.33 Tirpesti 4 6240 Gumelnita 5 a 5
47.87 21.12 Tiszapolgar-Basatanya 8 5060 Eneolithic 5 a 7
47.00 18.00 Tököl 2 Bell Beaker 5 a 7
47.00 16.00 Tordekes Koponya 1 Lengyel 5 a 7
44.23 27 Varasti 43 5360 Gumelnita 5 a 5
46.94 20.23 Vészto-Mágori Halom 7 6200 Körös 5 a 6
46.52 18.26 Villánykövesd 7 Lengyel 5 b ADAMS
44.48 20.36 Vinca 9 Starcevo 5 b ADAMS
44.31 22.01 Vlasac 56 7700 Mesolithic 5 b ADAMS
39.30 22.80 Volos 3 Neolithic 5 a 17
43.48 7.30 Abri de Pendimoun 1.00 Cardial 6 c Riquet 1973 
40.06 –2.48 Alcazar del Rey 3.00 Neolithic 6 a 14
41.12 1.00 Arboli 1.00 Bell Beaker 6 c Alcobé 1973 
38.33 16.12 Arene Candide 1.00 Cardial 6 c Riquet 1973 
38.33 16.12 Arene Candide 3 Early Mesolithic 6 b ADAMS
42.37 13.37 Arma dell'Aquila 3.00 Cardial 6 c Riquet 1973 
43.06 0.24 Arudy 1.00 Late Mesolithic 6 b ADAMS
37.9 13.48 Buffa Cave 1 Neolithic 6 b ADAMS
43.48 7.3 Castellar 1 Cardial 6 b ADAMS
43.24 5.12 Chateauneuf 2.00 Cardial 6 b ADAMS
40.06 8.3 Condeixa 60 Cardial 6 b ADAMS
44.12 8.18 Finale Ligure 3 Cardial 6 b ADAMS
37.18 –2.06 Gerundia 1.00 Almeria 6 c Riquet 1973 
37.24 –2.54 Gorafe 2.00 Almeria 6 c Alcobé 1973 
43.18 2.54 Grotte de Bize 1.00 Middle Neolithic 6 c Riquet 1973 
43.24 5.12 Grotte Sicard 2.00 Cardial 6 c Riquet 1973 
37.18 –3.12 Gudaix 1.00 Neolithic 6 c Alcobé 1973 
43.12 2.18 Iziar 2.00 Mesolithic 6 b ADAMS42 13.3 La Punta 1 Early Mesolithic 6 b ADAMS
43.24 4.46 Llanes 1.00 Mesolithic 6 b ADAMS
43.03 13.04 Maddalena di Muccia 1 Cardial 6 b ADAMS
40.54 17.24 Monopoli 2 Cardial 6 b ADAMS
39.12 8.42 Muge 24 9000 Late Mesolithic 6 b ADAMS
39.06 8.42 Mugem arruda 10 Late Mesolithic 6 b ADAMS
38.37 8.58 Mugem moita 15 Late Mesolithic 6 b ADAMS47 7 Niederhergsheim 1 Campaniform 6 c Riquet 1973 
47.12 7 Oberentzheim 1 Campaniform 6 c Riquet 1973 42 13.3 Ortucchio 3 12 500 Mesolithic 6 b ADAMS
43.24 6.54 Peillon 1 Mesolithic 6 b ADAMS
41.38 2.18 Sabassona 1 Cardial 6 c Alcobé 1973 
41.09 1.12 Salamno 3.00 Bell Beaker 6 c Alcobé 1973 
42.54 2.54 Salces 4 Cardial 6 b ADAMS38 14.36 San Fratello 4 12 000 Mesolithic 6 b ADAMS42 13.3 San Teodoro 4 Mesolithic 6 b ADAMS
38.24 –9.06 Sesimbra 2.00 Almeria 6 c Alcobé 1973 
39.06 –0.12 Tabernes de Valdigno 2.00 Neolithic 6 c Alcobé 1973 
42.06 3.09 Toroella 1.00 Bell Beaker 6 c Alcobé 1973 
40.24 3.42 Urtiaga 2 11 500 Mesolithic 6 b ADAMS
41.18 1.09 Vilavert 1.00 Neolithic 6 c Alcobé 1973 
Key to location of collections:

1 Department of Anatomy and Anthropology, University of Tel Aviv, Israel.

2. Department of Anthropology. Hacetepe University at Beytepe Campus, Turkey.
3. Göttingen University, Department of Anatomy, Germany.
4. British Museum, London, England.
5. Francis Rainer Institute of Anthropology, Bucharest, Romania.
6. József Attila University, Department of Anthropology.
7. Natural History Museum, Department of Anthropology, Budapest.
8. Jena University, Germany.
11. State Archaeological Institute, Munich, Germany.
12. University of Vilnius, Fac. of Medicine, Dept. Anatomy, Histology & Anthropology.
13. Institute of History, Department of Archaeology, Tallinn, Estonia.
14. Panum Institute, Copenhagen, Denmark.
15. Natural history Museum, Prague., Czech Republic.
17. Department of Animal and Human Physiology, University Athens, Greece. 
Table 2. Summary table of data. 


Ron Pinhasi 

Measurement Measurement Source Presence in ADAM 
Acronym Description 
Cranial measurementsGOL Maximum cranial length 
BBH Basio-bregma height 
XFB Maximum frontal breadth 
XPB Maximum parietal breadth 
ZYB Bizygomatic breadth 
MFB Minimum frontal breadth (WFB) 
NPH Nasion-prosthion length 
NLH Nasion height 
NLB Nasion breadth 
OBH Orbital Height 
OBB Orbital breadth 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
Howells, 1973 
yes 
yes 
yesyes 
yes 
yesyes 
yes 
yesyes 
no 
Mandibular measurementsRAMH Projective height of mandibular ramus 
GONANG Gonial Angle 
MAXL Projective length of mandible 
RAMB Minimum width of ramus 
GONB Bigonial breadth 
CONDB Bocondylar breadth 
ANTH Symphysis height (id-gn) 
ANTTHIC Anterior thickness 
Martin, 1957 
Martin, 1957 
Martin, 1957 
Martin, 1957 
Martin, 1957 
Martin, 1957 
Martin, 1957 
Martin, 1957 
yes 
yes 
yesyes 
no 
no 
no 
no 
Dental measurementsLCMD Lower Canine mesiodistal dimension 
LCBL Lower Canine buccolingual dimension 
LP3MD Lower Third Premolar mesiodistal dimension 
LP3BL Lower Third Premolar buccolingual dimension 
LP4MD Lower Fourth Premolar mesiodistal dimension 
LP4BL Lower Fourth Premolar buccolingual dimension 
LM1MD Lower First Molar mesiodistal dimension 
LM1BL Lower First Molar buccolingual dimension 
LM2MD Lower Second Molar mesiodistal dimension 
LM2BL Lower Second Molar buccolingual dimension 
LM3MD Lower Third Molar mesiodistal dimension 
LM3BL Lower Third Molar buccolingual dimension 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
Hillson, 1996 
no 
no 
no 
no 
no 
no 
no 
no 
no 
no 
no 
no 

Table 3 – list of cranial and mandibular variables utilised.


Anova due to the small and uneven sample sizes of 
the groups. Results are described in Table 5. The 
analysis by rank indicates that the groups are significantly different at the p = 0.005 level. 

Figure 16 contains boxplot diagrams of the above 
groups by variable. The left column comprises box-
plots of group medians and inter-quartile ranges, 
while the right column comprises means and standard deviations and standard error (the boxed area). 
This arrangement allows one to compare the distribution and variance of scores per measurement, and 
group and to detect possible ‘distortions’ due to outliers. Group 2 (Çayönü) has the largest variance for 
GOL, XPB, and NPH, while group 3 (Abu Hureyra) 
has the largest variance for ZYB, NLH, NLB and OBH. 
The variance of other groups is much smaller. The 
Levantine Natufian group has a significantly greater 
mean for GOL and ZYB. The European Early Neolithic groups (5–8) and Çatal Höyük display similar 

means and variances for GOL, XPB, ZYB and NPH. 
In the case of nasal dimensions (NLH and NLB) and 
Orbital Height (OBH) some differences between these 
groups exist, as the Cardial (Impressed) Mediterranean Neolithic group has lower means. There is a significant degree of heterogeneity among the specimens of Southeast Anatolia (Çayönü) and Northeast 
Levant (Abu Hureyra), which is apparent from the 
examination of their rank and inter-quartile boxplots 
(left column), as well as their means and variances. 

In sum, the examination of these boxplots indicates 
that, all in all, the European groups plus Çatal Höyük (groups 4 to 8) show similarities in means and 
variance, as well as in medians and interquartile 
ranges for the variables examined. In contrast, the 
first three groups (Natufian, Çayönü and Abu Hureyra) differ from this group mainly in the dispersion of the 50% range (and variance in the case of 
the right column), as well as from each other. 


A new model for the spread of the first farmers in Europe 

PART 2 – ASSESSING THE GEOGRAPHIC AND 
BIOLOGICAL DISTANCES OF EARLY NEOLITHIC 
EUROPEANS FROM POTENTIAL SOURCES 

Underlying the demic diffusion model is the assumption of a differential admixture between dispersing 
farmers and local Mesolithic populations, which 
would explain the observed SE–NW genetic cline. 
Thus, the hypothesis assumes that the more distant 
the original farmers from their centre of origin, the 
more they admixed with local hunters. Ammerman 
and Cavalli-Sforza (1984) regarded Jericho as the 
area of origin for the first farmers. This observation 
is inaccurate. While Jericho is one of the oldest agricultural sites in the region, there exist other contemporaneous populations from the PPNB period in the 
Levant and Anatolia. These populations display a 
high degree of heterogeneity, and thus cannot be accepted as a single ancestral population. Moreover, 
there is a time lag of approximately 2000 years between the first appearance of Neolithic cultures in 
this region and the appearance of the first Neolithic 
sites in south-east Europe. There is no reason to assume that the farmers that dispersed to south-east 
Europe came from Jericho and not from another location/culture. In order to assess the demic diffusion 
model it is, therefore, necessary to define more accurately a possible region from which the farmers 
dispersed to Europe. We saw in the previous sections that among the PPNB groups only Çayönü 
displays some similarities to the early Neolithic European populations. Moreover, in all analyses of 
Early Neolithic populations, the Çatal Höyük specimens are strikingly similar to the European Early 
Neolithic groups. 

This section examines the relationship of geographic 
distance from (a) Çatal Höyük, and (b) Çayönü to 
other European Early Neolithic sites. It does so by 
estimating the Mahalanobis D2 distances between 
corresponding group centroids in order to assess the 
following hypothesis: 

H1: Assuming that the site of Çatal Höyük represents the center of origins of the first farmers who 
dispersed to Europe, the further away the specimens/ site are from this centre, the further they 
are also in morphological distance. 

The methodology applied is the analysis of squared 
Mahalanobis distances between groups using the following set of variables: GOL, XPB, ZYB, MFB, NPH, 
NLH, NLB, and OBH. Cases with missing data were 
excluded (no substitution). Group means and codes 

Group Code N 
Levant Mes. – Natufian 1 9 
SE Anatolia – Çayönü 2 3 
NE Levant – Abu Hureyra 3 3 
Central Anatolia – Çatal Höyük 4 11 
Greece Neolithic 5 16 
Med. E. Neolithic 6 35 
SE Europe – E. Neolithic 7 19 
Central Europe E. Neolithic 8 27 
Total 123 

Tab. 4. Groups and corresponding codes. 

are described in Table 6. Geographic distances between groups were calculated using Spheric V. 1.05 
software for the calculation of geographic distance 
between two points on Earth based on their latitude 
and longitude co-ordinates. 

Table 7 outlines the groups and their Mahalanobis 
and geographic distances from Çatal Höyük. Clearly, 
the site with the smallest geographic distances from 
Çatal Höyük is Çayönü. Yet, this site has the largest 
Mahalanobis distance (7.42) from Çatal Höyük. The 
second largest squared Mahalanobis distance is of 
Lepenski Vir, which is 1169 km away. The smallest 
squared Mahalanobis distance is from Nea Nikomedeia (1.16), which is only 938 km from Çatal Höyük. However, the LBK site of Viesenhäuser Hof, 
which is 2253 km away from Çatal Höyük, has a 
squared Mahalanobis distance of 1.89. It therefore 
appears there is no correlation between geographic 
distance and morphometric distance in the case of 
the Early Neolithic groups. The most intriguing observation is, in fact, the LBK sites, which are more 
than 2000 km away from Çatal Höyük and have 
small squared Mahalanobis distances from this group. 

We see no linear relationship between Mahalanobis 
and geographic distances from Çatal Höyük (Figure 
17). The site of Çayönü is only 681 km from Çatal 

Code N Sum of Ranks Levant Mes. – Natufian 1 9 949 
SE Anatolia – Çayönü 2 3 136 
NE Levant – Abu Hureyra 3 5 108 
Central Anatolia – Çatal Höyük 4 18 1424 
Greece Neolithic 5 16 1061.5 
Med. E. Neolithic 6 35 2041 
SE Europe – E. Neolithic 7 19 1386 
Central Europe E. Neolithic 8 30 2074.5 

Kruskal-Wallis test: H (N = 135) = 20.29882 p = 0.005 

Tab. 5. Results of the Kruskal-Wallis analysis by 
rank (Natufian and Early Neolithic). 


Ron Pinhasi 


Fig. 16. Boxplots of medians and 25% quartiles (left), means and Standard Error (right) by variables. 

Höyük and is geographically the closest. Yet the Ma-H2: Assuming that the site of Çayönü represents the 
halanobis D2 distance between Çatal Höyük and Ça-centre of origins of the first farmers who dispersed 
yönü is 7.42, by far the largest figure. We can there-to Europe, the further away the specimens/site are 
fore reject the above hypothesis. We will then pro-from this centre, the further they are also in morphoceed and examine the same hypothesis regarding logical distance. 
the site of Çayönü. 


A new model for the spread of the first farmers in Europe 


Fig. 16. Cont. 

Table 8 presents the morphological (Mahalanobis 
D2 distances) and geographic distances of European 
Neolithic specimens to that of Çayönü. 

The first observation is that the squared Mahalanobis distances from Çayönü to the other sites are 
much larger. The largest Mahalanobis distance is to 
Condeixa (11.5), followed by Lepinski Vir (10.81). 
We observe no uniformity in scale of distances, as 
Viesenhäuser Hof, which is 2700 km away from 
Çayönü, has the smallest squared Mahalanobis distances from this group (3.76). There is clearly no linear correlation between geographic and morphometric distances (Fig. 18). 

These results indicate that morphological distances 
between the European Early Neolithic groups and 
either Çatal Höyük or Çayönü are not related to geo

graphic or to chronological distances per se. This finding is relevant, keeping in mind that we are looking 
only at Early Neolithic groups which could be regarded as representative samples of the first European 
farmers. We may speculate that the lack of association with geographic distance is due to the rapidity 
or particular route of the dispersal, thus explaining 
why a German LBK group may be morphologically 
more similar to an Early Neolithic Anatolian group 
than to a Danish Group. Moreover, the distances of 
the Early Neolithic European sites to Çatal Höyük 
are much smaller than the distances to Çayönü. 

PART 3 – ANALYSIS BY GEOGRAPHIC REGION 

The regionally-based approach was applied successfully in the case of the archaeological site analyses 


Ron Pinhasi 

GOL XPB ZYB MFB NPH NLH NLB OBH Valid N 
Çayönü 178.67 135.33 122.00 94.33 68.33 53.00 25.30 34.23 3 
Çatal Höyük 184.95 138.59 125.45 94.73 67.27 48.36 25.00 31.82 11 
Lepinski Vir 186.33 142.67 138.33 99.00 70.67 51.33 26.67 32.00 3 
Vészto-Mágori halom 187.00 134.17 123.00 96.17 65.33 48.58 24.78 32.15 6 
Nea Nikomedeia 180.82 138.82 123.36 94.45 66.45 46.82 25.18 32.00 11 
Schwetzinegn 179.80 138.10 123.40 95.66 64.41 46.98 24.74 31.80 10 
Sondershausen 187.40 139.20 126.60 98.40 67.52 50.46 23.90 32.82 5 
Viesenhäuser Hof 183.67 138.00 122.13 95.14 65.68 49.89 24.70 32.28 12 
Condeixa 179.00 137.12 122.69 95.31 65.15 46.12 23.15 30.69 26 
All Groups 181.99 137.80 123.89 95.47 66.02 47.95 24.39 31.74 87 

Table 6 Group means and codes utilised. 

(Pinhasi et al. 2000). On the basis of these analyses, 
it was possible to develop a model in which different 
degrees of admixture and interaction were hypothesised for the various regions of Europe (Lahr et al. 
2000). 

The model is examined using the available skeletal 
samples. The statistical procedures applied are PCA 
and discriminant function analyses. The first technique can be used in an exploratory manner, as no 
dependent variables need to be specified. The lack 
of grouping criteria makes it possible to perform 
analytical runs with relatively small numbers of specimens, and thus to narrow the temporal and geographical range. The second technique is mostly used 
for classification purposes. Nevertheless, its second 
value is in the interpretation of the findings. Due to 
word limits, results presented below include findings 
from the analyses of regions 1, 5 and 6. 

Region 1 – Turkey and the Levant 

In the Levant, it is common to distinguish between 
the Epipalaeolithic period – 13 000–10 500 BP, followed by two pre-pottery Neolithic periods – PPNA 
and PPNB. The PPNA is the period between ca. 

Group Name Mahalanobis Geographic 
ÇayönüLepinski Vir 
Vészto-Mágori halom 
Nea Nikomedeia 
SchwetzinegnSondershausen 
Viesenhäuser Hof 
Condeixa 
distance 7.424.29 
1.81 
1.161.75 
2.00 
1.892.45 
distance (km) 674.7481169.748 
1468.144 
938.592333.112 
2310.736 
2253.6152095.071 

Tab. 7. A comparison of Geographic and Mahalanobis distances from Çatal Höyük to European Neolithic specimens. 

10500–9300 BP. During this time, villages are found 
on a rather narrow territory, extending from the Damascus Basin in the north to the Trans-Jordan in the 
south (Yakar 1998). The lithic industry from this period shows discontinuity with the Natufian cultures. 
In the PPNB period (ca. 9300–7800/7500 BP) villages are on average larger than before, some of them 
reaching 10–12 hectares (Yakar 1998). 

The aims of the analysis are as follows: 

 To check the position of the specimens from the 

PPNB sites of Cayönü, Basta and Abu Hureyra in 

relation to the Natufians, and the Chalcolithic site 

of Pkein. 
 To assess the position of the Mesolithic and Early 

Neolithic specimens from Greece in relation to 

Khirokitia. 

All analyses on specimens from this region were performed on the data set after performing a Norm 
NORM V. 2.03 (2000) imputation procedure for missing data. The location of specimens utilised is depicted in Figure 19. 

Analysis of cranial dimensions 

The following set of cranial variables was selected: 
GOL, XPB, ZYB, MFB, NPH, NLH, NLB, and OBH. The 
total sample included a set of 106 specimens. Eigenvalues, means, standard deviations, and factor loadings are given in Table 9. 

Figure 20 is a scatterplot of the first two components. 
The facial height variables NPH and NLH, and cranial length variable (GOL) have high positive loadings on the first component. Thus, PC1 describes crania that are either long with tall faces, or short with 
short faces. The variables MFB, ZYB, and XPB load 
highly on the second component, while GOL and 
NPH have small negative loadings. Therefore, PC2 


A new model for the spread of the first farmers in Europe 

Fig. 17. A scatterplot of Square Mahalanobis dis-
tances vs. geographic distances from Çatal Höyük 
to European Neolithic specimens. 
describes crania that are either very broad (in both 
vault and zygomatic dimensions), and moderately 
short with short faces, or the opposite shape. Orbital Height (OBH) and MFB have high positive loadings on the third component (especially OBH), while 
XPB and ZYB load negatively on this component. 
Thus, PC3 describes crania that are either very broad 
with short orbits and narrow frontals, or very narrow with tall orbits and broad frontals. 

Mandibular variables 

The Eigenvalues, means, standard deviations, and 
factor loadings of the PCA of mandibular dimensions are given in Table 10. The analyses extracted 
three components, but only the first two have Eigenvalues above 1. 

Figure 21 is a scatterplot of the first and second components. The upper part of the scatterplot includes 
two outliers, one from Çayönü and the other from 
Abu Hureyra. The Jericho PPNB group have positive PC1 scores and are thus located in the right part 
of the graph. The largest range of variation is among 

Group Name Mahalanobis Geographic 
distance distance (km) 
Çatal Höyük 7.42 674.748 
Lepinski Vir 10.81 1619.438 
Vészto-Mágori halom 7.85 1856.873 
Nea Nikomedeia 7.26 1514.494 
Schwetzinegn 6.30 2763.415 
Sondershausen 6.66 2689.569 
Viesenhäuser Hof 3.76 2700.287 
Condeixa 11.50 2702.951 

Tab. 8 A comparison of Geographic and Mahalanobis distances from Çayönü. 

the Natufians. However, with the inclusion of the 
two outliers mentioned above, both Abu Hureyra 
and Çayönü have a large range of variation comparable to that observed for the Natufians. The scatter-
plot shows poor morphological differentiation between most groups, with the exception of Basta and 
Abu Hureyra, which are clearly separated. The factor 
loadings show that all variables load high on the 
first component, which may thus be interpreted as 
describing overall mandibular size. Ramus breadth 
and height have high positive loadings on the second 
component, while bigonial breadth and condylar 
breadth load negatively. Therefore, the second component describes two contrasting mandibular shapes – narrow mandibles, with broad and tall rami, 
and broad mandibles, with narrow and short rami. 

Dental dimensions 

The third morphological complex selected is one defined by the dimensions of the lower dental arcade 
(excluding incisors). A principal component analysis 
was performed on the same set, including 75 specimens. Eigenvalues, means, standard deviations and 
factor loadings are given in Table 11. 

Figure 22 is a scatterplot of the first and second 
components. The component loadings indicate that 
all variables load positive on the first component. 
The mesio-distal dimensions of the two premolars 
have the largest positive loadings on the second component. The highest negative loadings are of the bucco-lingual dimensions of the two premolars and the 
canine. However, these loadings are much smaller in 
scale. 

The specimens from Abu Hureyra are scattered in the 
upper left part of the scatterplot. The specimens from 

Fig. 18. A scatterplot of Square Mahalanobis dis-
tances vs. geographic distances from Çayönü to 
European Neolithic specimens. 

Ron Pinhasi 

Jericho form a cluster in the middle 
right part of the graph, next to the 
specimens from Basta (which, nevertheless, have comparatively smaller 
dental dimensions), and within the 
larger clusters of the Natufian and 
Çayönü groups. Most of the Natufian 
specimens show comparatively large 
tooth sizes. In terms of morphology, 
these results imply that the Abu Hureyra specimens have large mesiodistal premolar dimensions, but within 
an overall small dentition (negative 
PC1 scores). 

The results of the PCA using cranial 
dimensions indicate a morphological differentiation between the PPNB 
specimens from Basta, Abu Hureyra 
and Khirokitia on the one hand, and 
the Natufians, Çatal Höyük and Çayönü on the other hand. The second and third analyses displayed no separation between groups, with 
the exception of the relatively distinct dental dimensions of Abu Hureyra in relation to the other PPNB 
groups. In all three analyses, the intra- group range 
of morphological variability of the Natufians overlaps with that of Çayönü, Çatal Höyük, and Jericho 
PPNB. We may therefore conclude that a considerable amount of intra-group and inter-group morphological variability exists among the PPNB groups. 

Discriminant analysis 

Following the results obtained from the PCA, we now 
examine the same data using discriminant analysis. 
The analysis expands both the temporal and geographic scope, as it includes the Greek Early Neolithic 
groups and two Levantine Chalcolithic groups. The 
Levantine Chalcolithic samples represent the popu-

Fig. 19. Location of analysed skeletal samples from Region 1 (by 
site). 
lised in the first analysis are outlined in the Table 

12. 
The variables included are GOL, XPB, ZYB, MFB, 
NPH, NLH, NLB, OBB, and OBH. All groups were included in the discrimination process, with the exception of half of group 3 (Çatal Höyük) and group 4 

(E. Neolithic – Greece). Half of group 3 was excluded 
from the configuration of the canonical functions in 
order to evaluate the performance of the classification in the next step. Group 4 was excluded in order 
to determine its affinities in the classification phase. 
Results are described in Table 13. Two functions, 
explaining 92.4% of the variance, had Eigenvalues 
above 1. 

Results indicate that the variables mostly correlated 
with the first function are GOL and MFB, while OBB 
is highly correlated with the second function. Wilks’s 
lambda shows that 
residualdiscrimina-c. Factor loadings 

a. Means and 
standard deviations 
Mean S.D. 
GOL 181.29 10.51 
XPB 139.89 7.30 
ZYB 127.23 8.45 
MFB 99.02 12.20 
NPH 66.85 5.57 
NLH 49.85 4.50 
NLB 25.03 2.32 
OBH 31.96 2.00 

lations of the Levant that
succeeded the Neolithic in
this region. The groups uti-


b. Eigenvalues and variance PC 1 PC 2 PC 3 
GOL 0.58 –0.22 –0.15 
Eigenvalue Total Cumulative Cumulative XPB –0.07 0.60 –0.58 
% variance Eigenvalue % ZYB 0.43 0.57 –0.49 
1 2.18 27.28 2.18 27.28 MFB –0.07 0.74 0.41 
2 1.51 18.89 3.69 46.16 NPH 0.85 –0.21 0.06 
3 1.45 18.15 5.14 64.31 NLH 0.84 0.07 0.29 
NLB 0.45 0.11 –0.23 
Tab. 9a–c. Results of the principal component OBH 0.13 0.41 0.74 
analysis of cranial dimensions. 


A new model for the spread of the first farmers in Europe 


Fig. 20. Scatterplots of the 1st & 2nd and 1st & 3rd components. 

tion after the derivation of the first two canonical 
functions is small (Wilks’ Lambda = 0.755). We therefore only examine the scatterplot of the first two 
functions. Examination of the group centroids and 
scatterplot of individuals’ scores (Fig. 23) indicates 
that the Natufian specimens (and one Greek Neolithic outlier) are very large, and thus have larger than 
most Fcn1 scores. Some of the Çatal Höyük specimens, as well as some of the Nea Nikomedeia sample, are particularly small. Jericho-Chalcolithic specimens are the only group with all negative values on 
Fcn2 (narrow orbits), and thus different from the 
others (albeit some overlap with remains from Nea 
Nikomedeia), while those from Çayönü are the only 
group with only positive values on Fcn2 (homogeneously wide orbits). 

Classification 

In total, 80.5% of the selected cases were correctly 
classified (Tab. 14). 

Among the interesting aspects of this high classification result, we may observe that: 


 Most of incorrectly classified Natufian remains 
(18.2%) were classified as PPNB, while a remaining 9.1% as Levantine Chalcolithic. 

 All of the misclassified PPNB (Çayönü and Abu 
Hureyra) remains (25%) had higher probabilities 
of belonging to the Natufian group. Together with 
the observation above, this highlights that the Natufian and PPNB samples clearly have two distinct 
morphologies, which, nevertheless, overlap in a 
portion of their ranges. 

 This latter inference can further be extended to 
the Anatolian Neolithic remains of Çatal Höyük in 
relation to the PPNB material. The former had 
25% misclassified cases, all of which grouped with 
the PPNB sample. 

 The material from Nea Nikomedeia is more variable, with some cases classified as PPNB, others 
as Anatolian Neolithic. 

 Finally, the Levantine Chalcolithic remains are 
very distinct from the rest; 100% of them were 
correctly classified. 

The high percentage of correct classification and the 
scatterplot of the first and second discriminant func

a. Means and 
standard deviations 
Mean S.D. 
MAXL 102.45 8.08 
RAMB 35.19 3.02 
RAMH 58.31 5.28 
GONB 95.19 7.52 
CONDB 115.87 8.21 
ANTH 32.97 4.25 

b. Eigenvalues and variance 
Eigenvalue % of Total Cumulative Cumulative 
variance eigenvalue % 
1 2.72 45.40 2.72 45.40 
2 1.08 17.93 3.80 63.33 
3 0.75 12.42 4.55 75.75 

Tab. 10a–c. Results of the principal component 
analysis of mandibular dimensions. 

c. Factor loadings 
Factor 1 Factor 2 
MAXL 0.69 0.25 
RAMB 0.48 0.66 
RAMH 0.62 0.37 
GONB 0.65 –0.57 
CONDB 0.80 –0.32 
ANTH 0.76 –0.12 


Ron Pinhasi 

tions indicate that discrimination between these groups was easily achieved with the given variable set. The 
most distinctive group is the Chalcolithic Levantine, which had all its 
specimens classified correctly and is 
comparatively homogenous. The second distinctive group is Nea Nikomedeia, with 83.3% of correct classification and a rather homogenous distribution of cases. The Natufians are 
variable. Although distinctive enough 
to have the majority of specimens 
correctly classified, some Natufian 
specimens approximate the morphology of the PPNB or the Chalcolithic. 

Summary 

The first two functions of the PCA of 
the cranial dimensions indicated the 
existence of three clusters: one of the 
Natufian, Çayönü, and Çatal Höyük; 
a second, of Khirokitia, and a third, 
of Abu Hureyra. The PCAs based on the teeth and 
mandibular dimensions showed little differentiation 
between the groups. 

The first discriminant analysis indicated a successful 
discrimination among most groups. In particular, 
the Natufians and the Jericho Chalcolithic group 
were separated from the rest with their high positive scores on the first function. The results show a 
lack of discrimination between Çatal Höyük, Körös, 
Nea Nikomedeia and Early Neolithic Greece. 

Fig. 21. A scatterplot of the first and second components (mandi-
bular dimensions). 
Neolithic specimens from Greece (see the first discriminant function analysis) than to Cayönü. We can, 
therefore, confidently conclude that Çatal Höyük is 
much closer to the European Early Neolithic groups 
than to any of the other Anatolian/Levantine PPNB 
groups. Once again, we observed extensive heterogeneity within and between Levantine PPN groups 
and the lack of affinities between these groups and 
the SE European groups. 

The analyses also demonstrate a lack of continuity 
between the Early Neolithic of the Levant and the 
Chalcolithic groups. The latter could easily be discri

minated from most 
of the PPN and south-c. Factor loadings 
east European Neolithic groups. The po-Factor 1 Factor 2 
sition of Khirokitia as LCMD 0.76 –0.15 
an outlier was once LCBL 0.73 –0.33 
again confirmed. If LP3MD 0.54 0.58 
LP3BL 0.71 –0.43 

a. Means and 
standard deviations 
Mean S.D. 
LCMD 6.88 .58 
LCBL 7.71 .76 
LP3MD 6.85 .51 
LP3BL 8.00 .57 
LP4MD 6.99 .52 
LP4BL 8.28 .55 
LM1MD 11.03 .64 
LM1BL 10.99 .59 
LM2MD 10.65 .66 
LM2BL 10.71 .65 
LM3MD 10.57 .85 
LM3BL 10.23 .80 

The PPNB specimens 
from Basta and Çayönü 
areinanintermediateposition between the Natufians and the Early Neolithic European groups. 
Çatal Höyük is clearly 
much closer to the Early 

b. Eigenvalues and variance LP4MD 0.46 0.75 
LP4BL 0.67 –0.25 
Eigenvalue % of Total Cumulative Cumulative LM1MD 0.73 0.29 
variance eigenvalue % LM1BL 0.83 –0.16 
1 6.22 51.85 6.22 51.85 LM2MD 0.68 0.13 
2 1.54 12.85 7.76 64.70 LM2BL 0.89 –0.15 
LM3MD 0.69 0.34 
Tab. 11a–c. Results of the principal component LM3BL 0.86 –0.12 
analysis on dental dimensions. 


A new model for the spread of the first farmers in Europe 

this culture belonged to the PPNB 
cultures of the Levant, then its odd 
position further strengthens the 
hypothesis of a large degree of heterogeneity among PPNB cultures. 

The PCAs cannot differentiate between the morphology of the Natufians and their PPNB successors, with 
the exception of the Abu Hureyra 
group, which is clearly an outlier. 
However, the first and second discriminant analyses show the discrimination between the Natufians and 
other populations. 

Region 5 – Greece and south-east 
Europe 

The affinities between Early Neolithic specimens from the Levant, Anatolia and south-east Europe were 
previously addressed. The aim of 
this section is to go further, and specifically examine 
the following questions: 

 Is there any evidence for continuity between lo

cal Mesolithic and Early Neolithic populations in 
south-east Europe? 

 What is the relation of the Early Neolithic groups 
from south-east Europe to the Anatolian Early 
Neolithic/PPNB populations? 

 What degree of morphological homogeneity can 
we detect among the Early Neolithic specimens 
from south-east Europe? 

The location of specimens studied is provided in Figure 24. 

PCA 

The PCA examines the relationship between Early 
Neolithic specimens from Greece and Çayönü, and 

Group Code 

Mesolithic Levant – Natufian 1 
PPNB – Abu Hureyra + Cayönü 2 
Neolithic Anatolia – Çatal Höyük 3 

E. Neolithic – Greece Various 4 
EN Greece – Nea Nikomedeia 5 
Chalcolithic Levant – Jericho 8 
Tab. 12. Groups utilised in the discriminant function analysis. 

the Mesolithic specimens from Italy and Greece (Mediterranean Mesolithic), and the Danube Gorge (Vlasac and Lepenski Vir Mesolithic). The PCA results are 
described in Table 15. The selected set of variables 
is similar to the one utilised for Zone 1. However, 
the nasal length and breadth dimensions were replaced by cranial height (BBH). This set was chosen 
in order to minimise the number of excluded cases 
due to missing data and yet retain the main vault 
and face variables utilised throughout this work. 

Figure 25 depicts the scatterplot of the individual 
factor scores on the first and second components. 
All factor loadings, with the exception of OBH (which 
has a small negative loading on PC1), are positive 
on the first component. High loadings on the second 
component are of OBH (0.93) and NPH (0.56). The 
first component, therefore, differentiates mainly according to the size of the vault. The figure indicates 
a differentiation between the two Mesolithic groups 
on the one hand, and the majority of the Neolithic 
specimens on the other. This is achieved by the first 
component, with Mesolithic specimens having positive loadings, while most Neolithic specimens have 
negative loadings. The main exceptions are the two 
Mesolithic specimens from Ortuccio and some specimens from Çatal Höyük, which have, respectively, 
small and large sizes. The Mediterranean Mesolithic 
specimens have low faces and low orbits and thus 
have negative PC2 scores. The Mesolithic specimens 
from Franchthi Cave in Greece are not closely asso-

Fig. 22. A scatterplot of the first and second components. 

Ron Pinhasi 

ciated with any of the Nea Nikomedeia specimens. 
The group of Greek Neolithic specimens from various locations shows much variability, with some 
specimens such as Athens-Agora, Hageorgitika, and 
Greek Neolithic remains from Volos positioned next 
to the Greek Mesolithic cluster. The specimens from 
the SE Europe Early Neolithic group vary in their factor scores, and thus do not form a distinct cluster. 

Discriminant analysis 

A discriminant analysis was performed on the same 
set. The variables selected were GOL, XPB, ZYB, NPH, 
NLH, NLB and OBH. The groups used and sample 
sizes are described in Table 16, and the results of 
the analysis in Table 15. 

The first part of the discriminant analysis examines 
the distribution of specimens from the seven groups 

a. Group Statistics 
GroupLevant-Natufian 
ÇayönüÇatal Höyük 
Greece-Neol. 
Nea Nikomedeia 
Jericho 
Arch. Period 
Mesolithic 
PPN 
E. Neolithic 
Neolithic 
E. Neolithic 
Chalcolithic 
Code 
1 
2 
3 
4 
5 
8 
N 
11 
4 
8 
none 
12 
6 
Total 41 

in relation to each other, and the location of group 
centroids in the discriminant space (Fig. 26). Function 1 has a strong positive correlation with ZYB, 
GOL and NLH, thus describing long skulls, with 
broad faces and tall noses. Function 2 describes contrasting shapes defined by long (GOL: +0.418) and 
narrow (XPB: –0.468) skulls, with moderately tall 
faces, and the opposite combination. 

The main observation is the disassociation of Khirokitia from all other groups analyzed. The Khirokitia 
centroid and associated specimens are clustered at 
the lower part of the graph, and are completely separated from all other groups by their high negative 
values on the second function (describing very short 
and broad skulls). The second observation is a cluster of Mediterranean Mesolithic specimens at the upper right section, and the proximity of their centroid 
to the centroids of the Danube Gorge Mesolithic and 
Danube Gorge Neolithic groups. The centroids of Çatal Höyük (code 4) and Körös (code 6) are very close 
to each other, while the centroid of Nea Nikomedeia 
is to the left. It is therefore possible to discern a cline 
of overall size (defined by facial breadth and height 
and cranial length) along Function 1, from the smaller groups (Nea Nikomedeia), to Çatal Höyük, to Körös, to the Mediterranean Mesolithic, with the Danube Gorge Mesolithic overlapping the range of variation of the latter and partially Körös. Although there 
is much overlap in each case, the trend in size is apparent. 

b. Eigenvalues and Wilk’s lambda 
Eigenvalues Wilk’s Lambda 

Function Eigenvalue % of Cumulative Canonical Test of Wilks’ Chi-square df Sig. 
Variance % Correlation Function(s) Lambda 

1 2.422 60.7 60.7 .841 1 through 4 .097 78.040 32 .000 
2 1.268 31.8 92.4 .748 2 through 4 .333 36.829 21 .018 
c. Structure matrix 
Function 
123 4 

GOL .540 –.178 .432 –.324
OBB .056 .770 .301 –.459
MFB .330 .069 .641 .153
NLH .267 .174 –.347 .017
NLB .012 .070 –.157 .071
NPH –.068 .069 –.119 –.068
OBH –.160 .052 .033 .603
XPB .040 .195 .537 .541


Tab. 13a–c. Results of the discriminant function 
analysis. analysis. 

CLASSIFICATION 

The classification included all groups plus an additional group of early Neolithic specimens for which 
posterior probabilities were recorded (group 8), but 
were not selected for the calculation of the discriminant functions. Altogether 61.4% the cases were correctly classified into one of the seven groups. Classification is 100% in the case of Khirokitia (group 1), 
followed by 90.9% in the case of the Mediterranean 
Mesolithic (group 3). In the case of the Danube 
Gorge Mesolithic group, the percent of correct classification is 58.3%, with 25% of the cases being misclassified into the Mediterranean Mesolithic and an 


A new model for the spread of the first farmers in Europe 

Fig. 23. A scatterplot of the first two canonical fun. 
additional 16.7% into Nea Nikome

dia. In the case of the Danube Gorge 

Neolithic group, classification is poor. 

Only 33% of the cases were cor

rectly classified, with single cases 

(i.e., 16.7% each) being assigned in

to groups 1, 2, 3 and 7. Among the 

Nea Nikomedeia specimens (group 

7), 72.2% are correctly classified, and 

misclassification occurs in groups 1, 

2, 3 and 6. In the case of Çatal Hö

yük, four specimens are misclassi

fied into the Nea Nikomedeia group, 

while single cases are misclassified 

into groups 2, 3, 5 and 6. Among the 

Körös specimens (group 6) correct 

classification is 50 %, with 37.5% of 

the cases misclassified into the Nea 

Nikomedeia group and a single case 

into the Mediterranean Mesolithic 

group. In the case of group 8 (Tab. 19), all cases were 
Is there any evidence for continuity between local 

assigned to one of the Early Neolithic groups. 
Mesolithic and Early Neolithic populations in southeast Europe? 

What we therefore see is a much higher degree of 

misclassification occurring among the Early Neoli- 
There is some evidence for Mesolithic/Early Neoli

thic groups. The Danube Gorge Mesolithic and Neoli- 
thic continuity only in the case of the Danube Gorge. 

thic groups have a relatively high number of cases 
Only about 50% of the cases of the Danube Gorge 

being misclassified. These results show that discrimi- 
Early Neolithic groups were classified into another 

nation between Khirokitia, the Mediterranean Meso- 
Early Neolithic European group or to Çatal Höyük. 

lithic and the Early Neolithic groups is clear. How- 
The other 50% were assigned to the two Mesolithic 

ever, discrimination between the Danube Gorge Me- 
groups and to Khirokitia. In the case of Çatal Höyük, 

solithic and other groups is not clearly achieved. 
only 20% of the cases were assigned to a Mesolithic 
group. With Körös, only 10% (a single case) were as-

Summary 
signed to a Mesolithic group. In the case of Nea Nikomedeia, about 30% of the cases were assigned to 

It is now possible to address the three questions po- 
Khirokitia, or to one of the two Mesolithic groups. 

sited in the introductory part of this section. 

What is the relation of the Early Neo-
Code 
Predicted Group 
Membership1 2 3 5 8 
Total 
lithic groups from south-east Europe 
to the Anatolian Early Neolithic/PPNB 
populations? 
Count 1 
% 
8 
2 
3 
5 
8 
1 
2 
3 
5 
8 
2 
1 
0 
0 
0 
72.7 
25.0 
.0 
.0 
.0 
0 
3 
2 
1 
0 
18.2 
75.0 
25.0 
8.3 
.0 
0 
0 
6 
1 
0 
.0 
.0 
75.0 
8.3 
.0 
1 
0 
0 
10 
0 
.0 
.0 
.0 
83.3 
.0 
11 
0 4 
0 8 
0 12 
6 6 
9.1 100.0 
.0 100.0 
.0 100.0 
.0 100.0 
100.0 100.0 
The Khirokitia group stands as a population distinct from other PPNB, 
Mesolithic and Early Neolithic populations. In the case of Çatal Höyük, 
about 50% of the specimens were 
classified into another Early Neolithic 
European group. However, none of 
the Early Neolithic European specimens was misclassified to the Çatal 
Höyük group. The position of the Çatal Höyük centroid is next to the Kö-
Tab. 14. Classification results. rös centroid, and in proximity to the 


Ron Pinhasi 


Fig. 24. Location of analysed skeletal samples from Region 5 (by site)
.


centroids of Nea Nikomedeia (on the left) and the 
Danube Gorge groups (on the right). We may therefore conclude that this group is much more similar 
to the Early Neolithic European populations than to 
any PPNB or Mesolithic groups. 

What degree of morphological homogeneity can we 
detect among the Early Neolithic specimens from 
south-east Europe? 

The above set of analyses was not set to particularly assess homogeneity, but rather population affinities and distances. However, the PCA gives us 
some idea about the range of intra-group variability. 
We see in Figure 22 that the Early Neolithic specimens are scattered and not clustered together as in 
the case of the Mediterranean Mesolithic. Nevertheless, given the existing variability of the Early Neolithic groups, most cases were still distant from the 
Mesolithic groups (with the exception of the Danube 
Gorge). This observation suggests that the apparent 
heterogeneity among the Early Neolithic groups is 
not due to their admixture with local Mesolithic populations, but rather due to other demographic, historical reasons. In addition, as this range of variability is not estimated, it is possible that it still complies with an expected range of variation in an average biological population. 

Region 6 – The Mediterranean regions of 
France and Italy 

The Mesolithic/Neolithic transition in the western 
Mediterranean region was a complex and diverse 
process (see Pluciennik 1997). Part of this complexity is due to the fact that this region extends over 
a very large area of ecologically diverse zones. The 
review also pointed out to a ‘delayed’ Neolithic occupation in most of the western Mediterranean region. 
The majority of secure dates are from the early part 
of the 7th millennium BP. This implies a gap of at 
least 1000 years between the Early Neolithic in the 
Balkans and northern Greece and the western Mediterranean. In addition, the review from the various 
localities indicated a hiatus of more than 500 radiocarbon years between the Mesolithic and Early Neolithic occupation layers at various sites. 

The questions to be addressed in this section are as 
follows: 

 Regarding the Mesolithic/Neolithic transition, is 

there evidence for Mesolithic-Neolithic morpholo

gical affinities? We know that the ‘Impressed Neo

lithic’ pottery culture extended all across this re

gion. 
 Do the ‘Impressed Neolithic’ specimens share mor

phological similarities with the Anatolian speci


A new model for the spread of the first farmers in Europe 

(group 7) contains various specimens from the following sites: 
Anza, Cipa, Deszk, Gura Bacalui, 
Devetakata Petera, Endröd, Hódmezövásárhely-Bodzáspart, Hódmezövásárhely-Kovács J. Tanya, 
Jasa Tepe, Karanovo, Kasalnak, 
Kotacpart, SF. Gheorge Bedehaza, 
Sofia, Solca, Sturovo, and Vészto-
Mágori. These are all Early Neolithic sites in the Balkans and 
Southern Hungary. The Mediterranean Mesolithic group includes 
specimens from Franchthi Cave, 
Arene Candide, Orttocio, La Punta, San Fratello and San Teodoro, 
and several other Mesolithic sites. 
The first analysis was performed 
on 72 specimens from the above 
set using the following variable 
mens from Çatal Höyük and some of the south-set: GOL, XPB, MFB, BBH, NPH, NLB, NLH, and OBH. 
east European specimens? The results of the analysis are described in Table 21. 
 Does the morphological analysis of western Me-Figure 28 displays the heterogeneity of the analysed 
diterranean Early Neolithic groups support the groups. No separation is achieved between the Mesotheory of a gradual logistic dispersal, or rather lithic and Neolithic groups. Positive PC2 scores are 
suggest a more rapid movement of farmers? associated with high orbits, low vault, and high and 
narrow noses. The two Mesolithic specimens from 
The location of specimens studied is depicted in Fi-Arene Candide (Fig. 18) belong to this type. Negative 
gure 27. PC2 scores are associated with low orbits, wide and 
low noses and high vaults. The two SE Early Neolithic specimens from Devetakata Petera and Leng-
PCA yel belong to this type. 
Analysis 1 Analysis 2 
Table 20 depicts the groups selected. The Impressed The second PCA was performed on averaged data. 
Neolithic is represented by two groups, one contai-The groups were selected so that each one reprening specimens from various sites (group 2) and the sents a single site (Tab. 22). 
second containing specimens from the site of Condeixa in Sardinia (group 4). The Nea Nikomedeia is Results are provided in Table 23. Twenty-two groups 
selected, as this site securely belongs to an early were included in the analysis. The variables select-
Neolithic occupation (group 5). The SE Europe group ed were GOL, XPB, MFB, ZYB, NPH, NLH, NLB, and 

Fig. 25. A Scatterplot of the first two components. 
a. Descriptive statistics 
Mean S.D. GOL 185.11 7.67 
BBH 136.59 6.39 
XPB 138.85 6.04 
NPH 67.04 4.45 
OBH 32.02 2.01 
ZYB 128.59 9.31 
MFB 95.67 4.35 

b. Eigenvalues and variance 
Eigenvalue % of Total Cumulative Cumulative 
variance eigenvalue % 
1 3.16 45.17 3.16 45.17 
2 1.26 18.00 4.42 63.17 
3 0.94 13.37 5.36 76.54 

c. Factor loadings 
PC1 PC2 GOL 0.78 0.02 
BBH 0.78 –0.05 
XPB 0.63 0.11 
NPH 0.66 0.56 
OBH –0.18 0.93 
ZYB 0.83 –0.25 
MFB 0.63 –0.07

Tab. 15a–c. Results of the principal component 
analysis. 


Ron Pinhasi 

OBH. Eigenvalues and cumulative variance by factor are almost identical in magnitude to those obtained in the case of the above factor analysis. The factor loadings indicate that all variables, with the exception of NLB, load highly on the first factor. The 
highest positive loading on the second factor is of 
NLB (.940). The highest negative loading is of ZYB 
(–.40). The highest positive loadings on the third 
component are of MFB (.70) and XPB (.50), and the 
highest negative loadings are of OBH (–.426), and 
GOL (–.385). We can therefore deduce that the first 
component accounts for general size, while the second is positively correlated with nasal breadth and 
minimum frontal breadth, and negatively correlated 
with zygomatic breadth. The third component is for 
the most part a reflection of variations in frontal 
and parietal breadth. The Mediterranean Mesolithic 
specimens from San Teodoro, San Fratello, Franchthi 
and Arene Candide are located in the forth quadrant 
of the graph and thus have generally large dimensions with narrow noses and wide zygomatics. The 
specimens from Ortuccio, Condeixa and the two 
Muge sites have generally small dimensions, but 
share with the above groups the morphological fea

a. Site names and codes 
Site Code n 
Khirokitia 1 4 
Vlasac Mesolithic 2 12 
Franchthi Cave 3 2 
Ortuccio 3 2 
San Fratello 3 3 
San Teodoro 3 2 
Theopetra 3 1 
Çatal Höyük 4 11 
Lepinski Vir Neolithic 5 3 
Vlasac Neolithic 5 3 
Vészto-Mágori 6 6 
Desyk-Olajkut 6 2 
Nea Nikomedeia 7 13 
Athens-Agora 7 1 
B'Koybea 1 7 1 
Athens-Neolithic 7 3 
Tirpesti* 8 1 
Cascioarele* 8 1 
Gîrlsti* 8 2 
Kasanlak* 8 1 
Gura Bacului* 8 1 

* Cases not included in the DA run, but only in the classification. 
Tab. 16. Results of the principal component analysis. 

tures of narrow noses and wide zygomatics. The two 
Cardial Neolithic groups of A. Dell’Aquila and Finale 
Ligura and the specimens from Tirpesti, have average cranial dimensions, but particularly wide noses 
and narrow zygomatics. 

Figure 29 illustrates the positions of the groups in 
the two dimensional space of the first two components. Great variability is evident in the distribution 
of Impressed Ware Neolithic specimens. We see two 
of the sites at the top of the scatterplot, while Grotte Sicard is at the bottom. Variability is mostly along 
the second axis, and thus mainly reflects variation in 
nasal breadth among the Impressed Ware groups. 

In contrast, the Mediterranean Mesolithic groups are 
mostly scattered in the bottom right part of the plot, 
with the exception of the site of Ortuccio, which is 
positioned at the left part of the plot. The Early Neolithic sites of Vészto-Mágori and Nea Nikomedeia are 
clustered next to each other at the top centre part, 
very close to the sites of Çatal Höyük and Çayönü 
and in proximity to Tîrpesti. We see variability in 
the location of the Late Mesolithic sites, with Vlasac 
and Hoëdic at the top right (positive factor scores on 
both components), while the two Muge sites are towards the bottom left (negative factor scores on 
both components). Khirokitia is positioned in the 
middle of the scatterplot, and is closer to the Early 
Mesolithic groups and remote from the Early Neolithic Anatolian groups. The Vina and Lepenski Vir 
Neolithic groups are positioned near each other at 
the centre-right part of the graph, near the Vlasac 
Mesolithic group. Thus, among the analysed Early 
Neolithic groups, these two groups are clearly the 
closest to the Danube Mesolithic and to other Mesolithic groups. 

Summary 

The above results point to the large heterogeneity 
among the Impressed Ware Neolithic groups. In the 
case of the cluster and PCA analyses, Arma Dell’ 
Aquila and Finale Ligure are associated with Tîrpesti. Grotte Sicard is associated with Mediterranean 
Mesolithic groups, such as San Fratello and San Teodoro. The scatterplot of the PCA (Fig. 29) indicates 
that Condeixa is associated with the Muge groups 
and with Ortuccio, and is thus much closer to the 
Mesolithic groups than to the Early Neolithic groups. 
We can therefore deduce that in the case of the Mediterranean Zone, it is not possible to rule out admixture with local Late Mesolithic groups. The lack 
of a satisfactory sample size and geographic coverage 


A new model for the spread of the first farmers in Europe 

for the Early Neolithic Mediterranean does not allow an in-depth 
examination of the specific areas 
of admixture. 

The first Impressed Ware Neolithic sites appear along the Adriatic coast, and this culture only reaches the Iberian Peninsula about 
a millennium later. We may therefore assume a gradual spread of 
this culture along the Mediterranean coast, and generally along 
the east-west axis at the southern 
part of the continent. We cannot, 
however, based on the above results, detect whether this spread 
was gradual or rapid. 

The Danube Gorge groups display 
similarities to each other and suggest local continuity. It is important to distinguish between the Balkans zone, as a whole, and the specific Danube Gorge 
Groups. While in the Gorge evidence for continuity 
may be found, the rest of the Balkan populations 
show clear evidence of affinities to each other, to 
Nea Nikomedeia, and to Çatal Höyük. These results 
support the hypothesis of a dispersal of farmers 
from central Anatolia (represented by Çatal Höyük), 
to the Greek Mainland and the Balkans, as well as to 
the southern part of Hungary. 

ASSESSING THE RESULTS IN THE CONTEXT OF 
GENETIC STUDIES 

The above discussion suggests that a general spatial 
analysis of genetic data may suffice to reveal general clinal patterns, but that the association of these 
patterns with historical events of expansion and migration is ambiguous and problematic. In general, 
geneticists seem to overlook the complexity of historical processes, and the fact that such processes complicate our understandings of the observed genetic 
patterns across space. Interpolation, coalescence, 

Fig. 26. A scatterplot of the first and second discriminant functions. 
a. Eigenvalues and Wilk’s lambda 
Eigenvalues Wilk’s Lambda 
Function Eigenvalue % of Cumulative Canonical Test of Wilks’ Chi-square df Sig. 
1 
2 
3 
.907 
.785 
.239 
Variance 
43.1 
37.3 
11.4 
% 
43.1 
80.5 
91.8 
Correlation 
.690 
.663 
.439 
Function(s)
1 through 6 
2 through 6 
3 through 6 
Lambda 
.201 
.384 
.685 
99.427 
59.406 
23.497 
42 
30 
20 
.000 
.001 
.265 

b. Structure matrix 
Function 
1 2 
ZYB .733 –.320 
GOL .684 .418 
NPH .246 .267 
NLH .431 –.076 
OBH –.153 .188 
XPB .014 –.468 
NLB –.084 .264 

Tab. 17a–b. Results 
of the discriminant 
analysis. 

and other mathematical models and methods often 
applied by geneticists tend to smooth out differential densities of genetic markers in specific regions, 
and steep genetic boundaries in favour of clinal patterns. It has also been pointed out by Barbujani 
(1995), Barbujani and Bertorelle (2001), and Sokal 
(1991), that migratory events without admixture, 
and gradual dispersal with admixture, can result in 
similar geographic distributions of gene frequencies. 
As Barbujani and Bertorelle (2001.22) point out: “A 
cline or gradient, for example, may reflect adaptation to variables environments, or a population 


Ron Pinhasi 


Fig. 27. Location of analysed skeletal samples from Region 6 (by site). 

expansion at one moment in time, or continuous barriers, such as deserts, forests, and so on. Most imgene flow between groups that initially differed in portantly, it should include archaeological and bioallele frequencies”. Thus, the genetic analyses can-logical (i.e., skeletal) data, the only actual evidence 
not detect more particularistic, population-specific of past peoples in time and space. 
patterns. These specific patterns can only be revealed by the incorporation of non-genetic information. Support for ‘demic diffusion’ (model 2a) comes from 
This information should include geographic data the study by Semino and colleagues (1996) of frewhich take into account geographic barriers such as quencies of two distinct Y-chromosome markers, the 
seas, lakes, mountain ranges, and possibly ecological p12f2–8–kb and the 49a, f–Ht 15 alleles among 

3000 subjects, mainly from Europe. 
Predicted Group Membership Total The study revealed that the p12f2– 
Group 1 2 3 4567 8–kb allele is specific to western Eura-
Count 1 4 0 000004 sian populations. The frequencies of 
2 0 7 3 000212 this allele among Near Eastern and 
3 0 0 10 001011 European populations display a cline 
4 0 1 1 311411 decreasing from the Near East to 
5 1 1 102016 north-western Europe. In contrast, a 
6 0 0 100438 map of 49a, f–Ht 15 allele frequencies 
7 1 1 2 00113 18 displays a cline in the reverse direc% 1 100.0 .0 .0 .0 .0 .0 .0 100.0 tion, with its highest frequencies 

2 .0 58.3 25.0 .0 .0 .0 16.7 100.0 among north-western Europeans and 
3 .0 .0 90.9 .0 .0 9.1 .0 100.0 the Basque. On the basis of these geo
4 .0 9.1 9.1 27.3 9.1 9.1 36.4 100.0 graphic patterns, these results have 
5 16.7 16.7 16.7 .0 33.3 .0 16.7 100.0 been interpreted as suggesting that 
6 .0 .0 12.5 .0 .0 50.0 37.5 100.0 the 49a–f–Ht 15 allele can be consi
7 5.6 5.6 11.1 .0 .0 5.6 72.2 100.0 dered as a proto-European haplotype, 
while the p12f2–8–kb allele is a Near 
Tab. 18. A summary of the classification results. Eastern haplotype. The south-east to 


A new model for the spread of the first farmers in Europe 

Location Assigned Group 
Tirpesti 5 
Cascioarele 4 
Gîrlsti 6 
Gîrlsti 4 
Kasanlak* 6 
Gura Bacului* 7 
Tab. 19. Classification results for unselected cases 
(group 8). 

north-west distribution of the latter is therefore in 
agreement with the ‘demic diffusion’ model. However, an examination of the published table (Semino and colleagues 1996.Tab 1) indicates that the 
frequency distribution of this allele is not uniform, 
being high among Near Eastern, Greek, Albanian, 
and Italian populations (f > 20), and much lower (f 
< 8) among Spanish, French, Hungarian and English 
populations. It is therefore questionable whether 
such results actually support the ‘demic diffusion’ 
model. 

Lucotte and Loirat (1999) analysed the distribution 
of haplotype 15 at p49 (locus DYS1) of the Y-chromosome among present-day Europeans. Their sample included 2,418 individuals originating from 28 
different geographic locations in Western Europe. 
They found the highest frequencies of the p49 TaqI 
haplotype 15 among French Basques (72.2%), Spanish Basques (53.8%), and individuals from the 
Montpellier region in France (53.5%). Frequencies 
were considerably lower in south-western European 
regions and in central Europe, while a minor peak 
was detected in north-west Europe (Great Britain, 
Brittany, Northern France, Germany, Belgium). The 
spatial plotting of these results, using frequency contours, reveals that there exists a gradient from northwest Europe and the Basque area into south-eastern 
and peripheral countries. This pattern is opposite in 
direction to the cline observed for the HLA genes, 
but is in agreement with the clinal distribution observed in the fifth synthetic map produced by Cavalli-Sforza and his team (1994.294). This map has 
very wide bands, which may represent the pre-Neolithic relict populations. 

Similar results were obtained by Wilkinson-Herbots 
and colleagues (1996) (see also Richards et al. 1996) 
in their analysis of human mtDNA at site 73 of hypervariable region II. A reduced median network of 
mtDNA control region sequence data for European 
populations (Wilkinson-Herbots 1996.Fig 1) suggests the presence of 5 groups. The diagram indica

tes that the phylogenies of groups 1 and 4 are starlike, while those of groups 2, 3, and 5 are not. A possible explanation for the star-like structure of groups 
1 and 4 is that they reflect the population expansion 
of the Late Glacial Maximum (20 000 BP), perhaps 
from one or two relict populations (Wilkinson-Herbots et al. 1996). The estimated age of these two 
groups is around 25 000 BP, using an evolutionary 
rate of 11.81% originally obtained by Stoneking and 
colleagues (1992 as cited in Wilkinson-Herbots 
1996). Based on the same mutation rate, an estimate 
of 45 000 BP was obtained for group 5 and suggests 
that this population, together with parts of groups 2 
and 3, represents the pre-glacial populations of Europe. 

Chikhi and colleagues (1998a; 1998b) looked at 
nuclear DNA clinal variations across Europe. They 
found that out of 34 DNA alleles, 22 showed significant spatial structure. Approximately one third of 
the alleles were arranged in broad, statistically-significant gradients, while for some other alleles, long-
distance differentiation was evident. These results 
indicate that (1) patterns observed by the study of 
molecular markers (Menozzi et al. 1978; Cavalli-
Sforza 1994; Sokal et al. 1989) do not differ much 
from studies of non-molecular markers; and (2) there 
is evidence to suggest that isolation by distance has 
affected genetic variation at the molecular level. Chikhi and colleagues admit that more research is required in order to reveal whether these DNA markers are affected by selective forces. Nevertheless, 
the presence of a clinal pattern among these markers 
casts some doubt on the aforementioned argument 
of Fix (1996) for a selection-based clinal pattern. It 
thus remains to be investigated whether the European clinal pattern is the outcome of ‘demic diffusion’, which assumes a certain level of admixture 
with local Mesolithic populations, or a ‘steppingstone’ model of founder effects, which annuls any 
admixture. 

In a recent work, Chikhi and colleagues (2002) studied Y chromosome markers, using a genealogical 

Group Arch. Period Code 
Impressed Ware – various Impressed Ware 1 
Impressed Ware – Condeixa Impressed Ware 2 
Nea Nikomedeia Early Neolithic 3 
SE Europe – Various Early Neolithic 4 
Çatal Höyük Early Neolithic 5 
Med. Mesolithic Mesolithic 6 
Tab. 20. Groups selected and corresponding codes.



Ron Pinhasi 

likelihood-based approach and to 
examine their findings in the light 
of the Demic Diffusion model of 
Ammerman and Cavalli-Sforza 
(1984) and the Cultural Diffusion 
model (i.e. indigenous transition). 
Their main innovation was the 
evaluation of genetic data from 
European, Anatolian and Near 
Eastern populations, modelling 
for admixture between migrating 
Neolithic farmers and local Mesolithic populations, as well as for 
the effects of drift. The data set 
comprised 22 binary markers 
from the non-recombining region 
of the Y chromosome (NRY) in a 
large number of European populations (n = 1,007 chromosomes 
from 25 samples). These markers 
are considered to be the result of unique mutational 
events and are called unique-event polymorphisms 
and are thought to be rare enough to have occurred 
only once in the recent history of human populations. In order to estimate admixture, the genetic structure of the original Neolithic population was derived from three samples from Syria, Turkey and Lebanon. The Mesolithic (or Paleolithic) population was 
represented by two samples of Basques. 

The results of the analysis of Chikhi et al. indicate 
that there is a clear trend across Europe, with the 
proportion of Neolithic genes decreasing from modal values around 85–100% in Albania, Macedonia, 
or Greece, to around 15–30% in France, Germany, 
or Catalonia. The statistical significance of this trend 
was then assessed and quantified by combining information from the individual populations and their 
geographic distance from the Near East and by plotting the regressions. Estimated average p1 values 
across Europe were compared with the values given 
by Semino and colleagues (2000) for the same data 
set. Chikhi and colleagues found an average Neolithic contribution of 50% across all samples. Furthermore, Chikhi and colleagues (2002) point 
out that these figures 
are likely to be underestimates of the true genetic contribution of 
the Neolithic farmers. 
This is since the me

a. Eigenvalues and variance 
Eigenvalues % Total variance Cumulative % 
1 3.08 38.50 38.50 
2 1.30 16.20 54.70 
3 0.96 12.01 56.71 
Fig. 28. A scatterplot of the 1st and 2nd components. 
genes that can be traced back to ancestors in the 
Near East rather than the true proportions of Neolithic vs. Paleolithic genes during the initial formation of Neolithic settlements in Europe. Thus, according to Chikhi and colleagues, the contribution of 
Neolithic farmers to the current gene pool of the various European populations is in fact between 65 
and 100%. These results provide further support for 
the ‘Demic Diffusion’ model. Moreover, the study reveals the importance of the incorporation of demographic parameters such as admixture rates and drift 
in any study of the spread of farmers in Europe. 

A more geographically oriented approach was applied in a study of Simoni and colleagues (2000). 
They analysed more than 2600 sequences of the 
first hypervariable mitochondrial control region 
for geographic patterns in Europe. Spatial autocorrelations were used in order to examine the relationship between genetic variability in this region 
and geographic distances. However, only a limited 
geographic pattern was observed. An area of significant clinal variation 
wasidentifiedaround 

b. Factor loadings 
PC1 PC2 
GOL 0.79 –0.15 
BBH 0.76 –0.39 
XPB 0.42 0.16 
MFB 0.55 –0.27 
NPH 0.78 0.27 
NLH 0.77 0.38 
NLB 0.34 –0.56 
OBH 0.28 0.70
thod utilized estimates Tab. 21a–b. Results of the principal component 
only the proportion of analysis. 


A new model for the spread of the first farmers in Europe 

the Mediterranean Sea, but not in the more northern parts of Europe. According to Simoni and colleagues (2000.275): “A simple demographic expansion from the Levant is easy to reconcile with the 
gradients observed at many nuclear loci, but it is 
not easy to link with the fact that mitochondrial 
variation is clinal only in southern Europe.” These 
findings may suggest greater gene flow (female, at 
least) along the southern Mediterranean region than 
across the northern part of the continent (Simoni et 
al. 2000). These findings agree with the suggested 
model. As the Impressed Ware people expanded 
westwards, they would have absorbed differential 
amount of genes from local foraging groups. Signs 
of initial admixture were observed in section 9.8 in 
the case of some of the initial Impressed Ware sites. 

There is an apparent discrepancy between findings 
from mitochondrial DNA studies (Richards et al. 
1996; Richards et al. 1998; Wilkinson-Herbots et 
al. 1996), which suggest a Palaeolithic ancestry to 
modern western European populations, and the findings from nuclear DNA (Chikhi et al. 1998a; 1998b) 
and classical markers (Menozzi et al. 1978; Cavalli-
Sforza et al. 1994), which suggest a Neolithic demic 

diffusion from the Levant as the main contributor 
to the European gene pool. This incompatibility can 
be explained if one is to accept that the gene pool 
of all modern European populations is to some extent admixed, with differential percentages of Palaeolithic and Neolithic ancestral contributions. This supposition is supported by evidence from mtDNA studies (Richards et al. 1996), which points to the 
mixed indigenous Palaeolithic and Neolithic (i.e., 
demic diffusion) ancestry of European populations. 
In the case of Y chromosome markers, results obtained by Semino et al. (1996; 2000) suggest that the 
contribution of incoming Neolithic farmers to the 
gene pool of the current European populations was 
as low as 22%. These results contrasts with results of 
the analysis of Y chromosome markers by Chikhi 
and colleagues (2002) suggesting that the genetic 
contribution of the incoming farmers may have been 
around 70%, and thus more in agreement with the 
demic diffusion model than with the concept of cultural diffusion without dispersal of farmers. 

The analyses of variability and morphological affinities suggest that the Epipalaeolithic populations from 
the Levant (Natufian) differed to a fair extent from 

Location N* Date (bp) Lat. (N)** Long. (E)** Arch Period Group Code 
Arma Dell'Aquila II 3 42.37 13.37 Impressed Ware 1 
Condeixa 60 40.06 8.30 Impressed Ware 1 
Finale Ligure 3 44.12 8.18 Impressed Ware 1 
Grotte Sicard 3 43.24 5.12 Impressed Ware 1 
Arene Candide 3 38.33 16.12 Early Mesolithic 2 
Franchthi Cave 3 39.00 22.30 Early Mesolithic 2 
Ortuccio 3 12500 41.54 13.42 Early Mesolithic 2 
San Fratello 4 12003 38.00 14.36 Early Mesolithic 2 
San Teodoro 4 42.00 13.30 Early Mesolithic 2 
Nea Nikomedeia 11 8180 40.65 22.30 Early Neolithic 3 
Vészto-Mágori 7 6200 46.94 20.23 Körös 3 
Tîrpesti 4 6240 47.17 26.33 Pre-Cucuteni 3 
Vlasac 56 7755 44.31 22.01 Late Mesolithic 4 
Hoëdic 18 47.21 2.52 Late Mesolithic 4 
Muge-Arruda 10 39.06 8.42 Late Mesolithic 4 
Muge –Moita 14 38.37 8.58 Late Mesolithic 4 
Teviec 31 9025 47.00 3.00 Late Mesolithic 4 
Çatal Höyük 50 7499 37.10 32.13 E. Neolithic 5 
Çayönü 9 9360 38.23 39.65 PPN 5 
Khirokitia 21 7368 34.54 33.00 PPN 5 
Lepenski Vir – E. Neol. 23 44.33 22.03 Starcevo 6 
Vinca-Neol 9 44.48 20.36 Starcevo 6 
* Sample sizes are given prior to the casewise exclusion due to missing values** Latitude and longitude are in the North East quadrant and in decimal notation 

Table 22. Groups analysed in PCA mean data analysis. 


Ron Pinhasi 

the Mesolithic population of the 
Danube Gorge, western Mediterranean and central Europe. As discussed before, no close similarities were observed between Early 
Neolithic and Mesolithic European 
groups in any of the studied regions, with the possible exception 
of the Danube Gorge, Mediterranean Europe, and the unstudied 
region of Atlantic Europe. However, there were also no clear affinities observed between the Epipalaeolithic Near Eastern groups 
and any other Mesolithic or Neolithic groups. These results imply a 
third scenario: that the original 
Epipalaeolithic population from 
which the first Anatolian farmers 
descended is yet to be “discovered”, as there are at present no 
skeletons, and meagre evidence for Epipalaeolithic 
occupation in Anatolia (Özdogan 1999). According 
to Özdogan (1999), the late phase of the Upper Palaeolithic period is either absent or poorly represented in both central and eastern Anatolia. The absence of sites from this period seems to reflect the 
scant occupation of these regions, rather than any 
excavation bias. 

The observed variability between Levantine and European Epipalaeolithic/Mesolithic groups should be 
studied in relation to the Upper Palaeolithic populations of Europe and the Near East. Thus, genetic studies that point to an Upper Palaeolithic ancestry of 
modern European populations should take into consideration population bottlenecks and segregation 
during the Late Glacial period, which can perhaps 
account for the above noted morphological variability during the Mesolithic. 

Fig. 29. A scatterplot of the 1st and 2nd components. 
the first synthetic map in the study of Cavalli-Sforza 
and colleagues (1994). The claim that the observed 
genetic cline is in fact the outcome of a ‘demic diffusion’ across Europe as proposed by Ammerman 
and Cavalli-Sforza (1984) is only partially in agreement with the current results (genetic, archaeological and palaeobiological). An examination of the 
model outlined in figure 10.4 indicates that the SE– 
NW genetic cline is, in fact, a synthesis of a southern 
(Mediterranean) cline and the northern/western 
cline generated by the expansion northwards and 
westwards of the Early Neolithic cultures of SE and 
central Europe. In other words, the new model 
agrees with the demic diffusion model in respect to 
western (i.e., Atlantic and western Mediterranean 
regions) and northern Europe, but not in the case 
of south-east and central Europe. One possibility is 
that the genetic cline is partially the outcome of the 
subsequent expansion of the Early Farmers, and partially the outcome of gene flow during later historical periods. 

c. Factor loadings 
a. Means and 
standard deviations 
Mean S.D. 
GOL 185.65 6.39 
XPB 138.53 3.75 
MFB 95.55 3.21 
NPH 66.32 2.98 
NLH 49.56 2.13 
NLB 24.66 1.20 
OBH 31.43 1.32 
ZYB 130.60 7.90 

The current findings ques
tion the interpretation of


b. Eigenvalues and variance GOL 
PC 1 
0.62 
PC 2 
0.30 
PC 3 
–0.39 
Component 
1 
2 
3 
Eigenvalue 
3.17 
1.37 
1.21 
% Total 
variance 
39.59 
17.15 
15.08 
Cumulative 
% 
39.59 
56.74 
71.82 
XPB 
MFB 
NPH 
NLH 
NLB 
OBH 
0.67 
0.52 
0.74 
0.85 
0.05 
0.48 
0.02 
0.36 
–0.17 
–0.15 
0.94 
0.24 
0.51 
0.70 
–0.21 
–0.24 
–0.07 
–0.43 
Tab. 23a–c. Results of the PCA analysis on group ZYB 0.75 –0.40 0.14 
means (n = 22).sis of cranial dimensions. 


A new model for the spread of the first farmers in Europe 

ASSESSING THE MODELS FOR THE SPREAD OF 
FARMING IN EUROPE 

Model 1 – Autochthonous transition across 
Europe 

The ‘indigenous’ model views the transition to farming as the result of cultural diffusion, without associated migrations or dispersals of human populations. Under such a model, local European hunter-
gatherer populations adopted farming either as a 
form of independent ‘discoveries’ or consequential 
to the arrival of knowledge and technology from the 
Neolithic cultures of the Near East and Anatolia. Underlying this approach is the assumption that the 
transition to farming occurred at different times and 
in different manners in the various regions of Europe. Archaeologists that support this model provide evidence which suggests continuity in material 
culture, and emphasise the economic, technological 
and cultural complexity of many of the Mesolithic 
cultures, which are viewed as a sort of ‘pre-adaptive’ stage followed by the choice of an agricultural 
lifestyle. 

The central tenet of this model is that in each region, 
the Early Neolithic farmers are the descendants of 
local Mesolithic populations. These descendants may 
differ in some morphological aspects from their ancestral Mesolithic predecessors due to admixture with 
other Neolithic groups and changes in morphology 
associated with their new lifestyle. Nevertheless, 
their morphology must display some affinities to 
their Mesolithic ancestors or similarities to contemporaneous neighbouring Mesolithic groups. Moreover, one should also see regional differences in 
morphology, as local continuity with minimal gene 
flow is assumed for each of the regions addressed. 
If this was the case, one should expect to see morphological distances across space being greater than 
morphological distances by period. However, the results of the analysis of the squared Mahalanobis distances between Mesolithic and Early Neolithic groups 
(Pinhasi 2003) indicate that this is not the case. 
The only possible case of regional continuity observed was in the case of the Danube Gorge region, 
where distances between these Mesolithic groups to 
the Early Neolithic context from Lepenski Vir are 
smaller than distances between the former and other Early Neolithic European groups. It is possible 
that, had the data set contained specimens from Atlantic Europe and the Iberian Peninsula, one would 
have found similar indications for local continuity 
in these regions. 

No firm evidence for local continuity was obtained 
in the case of the Levant region (Pinhasi 2003). The 
analysis of Levantine groups demonstrated a possible continuity between the Natufians and the 
Northern Levant and Jericho PPN in terms of their 
mandibular and dental dimensions. However, other 
PPN groups showed lack of affinities to the Natufians. One of the most intriguing results was the large 
degree of variability among the PPNB ‘Initial Neolithic’ groups from the Levant and Anatolia. The position of Khirokitia and Abu Hureyra as outlier populations was particularly noticed in numerous analyses. Moreover, the Natufians as a group display a 
large degree of heterogeneity, which suggests that 
they either represent several biological populations, 
a period of morphological transition with differential 
expression through time and space (the Natufian 
samples derive from several sites spanning a few 
thousand years), or still a period of comparatively relaxed selective pressures (in relation to earlier Upper 
Palaeolithic groups) that led to increased diversity. 

The regional analysis indicated lack of continuity between the Early Neolithic populations of the Levant 
and Chalcolithic groups. The latter could be easily 
discriminated from most of the PPN and south-east 
European Neolithic groups. The observed differences reflect populations that were isolated enough 
from each other to allow the development of regionally-specific morphological patterns. It is then possible that, after the initial phase of adoption of agriculture in the Levant and Anatolia, the PPN populations became isolated from each other and did not 
share a large mating network, as in the case of various hunting populations. The existence of biologically diverse PPNB groups in this region is likely, 
considering the time span of 2000 years or more during which agriculture communities existed prior to 
the first spread of farming into Europe. In fact, we 
see a similar process of diversification following the 
first period of the Neolithic in Europe. This diversification is evident culturally in the appearance of diverse stylistic groups in east and central Europe 
(Sherratt 1983; Thorpe 1996). What is important in 
this scenario of a period of differentiation of PPN 
populations in the Levant prior to expansion of agriculturists outside the region is that the group that 
contributed to the gene pool of European farmers 
(or originated it) need not be representative of the 
entire PPN Levantine population, but rather of one 
isolated and already somewhat differentiated group. 

The other possibility is that the heterogeneity among 
the PPNB populations reflects differences in mor


Ron Pinhasi 

phology among their ancestral Epipalaeolithic populations. As there are hardly any human remains from 
the Epipalaeolithic period of Anatolia, we presently 
know very little about the late hunters from this region and cannot therefore evaluate this hypothesis. 

Beyond the zone of Anatolia and the Near East we 
find a striking degree of morphological similarities 
among Early Neolithic populations. Thus, the first 
farmers from Nea Nikomedeia and other Greek Neolithic sites are similar to the first farmers from Çatal 
Höyük, as well as to specimens of the Körös and 
Starevo cultures and to the first LBK groups of Central Europe. The analyses showed no similarities between Greek Neolithic specimens and the Mesolithic 
specimens from Franchthi Cave, and thus no evidence for local continuity in the region of mainland 
Greece. 

Equally, there appear to be no affinities between any 
of the LBK groups and the German Mesolithic specimens from Ofnet, Hohlestein, Kaufertsberg, Stetten 
or Bottendorf (analysis of region 4 in Pinhasi 2003). 
These findings agree with Thorpe (1996.29), who asserts the following in regard to the appearance of the 
LBK culture of Central Europe “…there have been 
a few archaeologists arguing that the Mesolithic population of Central Europe as far north as the fringes of the north European plain and as far west as 
the Paris Basin played much part in the agricultural transition of the region, apart for some evidence for continuity in stone tool production.” 

The relationship between the Mesolithic and Neolithic populations of Northern Europe was examined 
in the analysis of region 3 (Pinhasi 2003). The discriminant function analysis performed included a 
group of four Mesolithic specimens (group 1), one of 
which was the Mesolithic specimen from Spiginas, 
Lithuania. The other three were specimens from the 
Danish sites of Koelbjerg, Vedbaek and Korsor. Discrimination between this group and the rest of the 
(Neolithic) Danish groups was achieved along the 
first canonical function. Because of the small sample 
size of the Danish Mesolithic sample, the possibility 
that some affinities existed between this group and 
Neolithic populations of the area cannot be ruled out. 
Nevertheless, the lack of association (in the form of 
easily achieved discrimination) between the two periods is a theme which has been noticed throughout. 

In sum, the review indicates that local continuity 
may have occurred in the Danube Gorge region and 
southern Scandinavia, and possibly also in the Medi

terranean zone and Atlantic Europe, although the 
evidence for these latter areas is still tentative and 
dependent on more in-depth analyses. Therefore, in 
the case of the latter zones it is necessary to examine this claim with larger skeletal samples. From 
the point of view of the debate on the origins of 
agriculture in Europe, the model of an autochthonous development does not fit the current findings 
regarding the regions of the Levant, Anatolia, SE 
Europe and central Europe. 

Model 2a – Demic diffusion 

The ‘wave of advance’ model argues for a large-scale 
advancement of farmers from the Near East or Anatolia into Europe in a more or less constant expansion rate of 1.1 km/year (Cavalli-Sforza et al. 1994). 
The hypothesis was largely based on a model for 
the ‘wave of advance of advantageous genes’ originally proposed by Fisher (1937). This mathematical 
model assumes that a biological population that 
grows at a constant rate reaches local saturation and 
spreads at a constant rate of migration, randomly in 
all directions, tending to grow and move away from 
its centre of origin at a radial constant rate of advance. Ammerman and Cavalli-Sforza adopted this 
theoretical model and applied it, with some modifications, to their study of Neolithic demic diffusion. 
Ammerman and Cavalli-Sforza (1984) used Jericho 
as the centre of origin from which the farmers dispersed towards Europe. The main supporting evidence underlining their model was as follows: 

 The expansion of farmers from the Middle East 

to Europe was very slow, gradual, and regular, 
thus more compatible with the expansion of people than of a technique. 

 Knowledge from contemporary ethnographic observations regarding demographic growth and 
expansion allows one to predict that the diffusion 
of agriculture is compatible with demic expansion. 

 Ethnographic observations based on African Pygmies suggest that hunter-gatherers show little tendency to acculturate when in contact with farmers. 

 A study of the modern geographic distribution of 
genes in Europe strongly suggests diffusion from 
a centre of origin in the Middle East, as well as 
other less important migrations (Menozzi et al. 
1978). 

The results obtained do not support the wave of advance model. Firstly, the analysis of population variability during the Early Neolithic revealed a contrasting pattern of high heterogeneity within the 
Early Levantine and Anatolian populations as op


A new model for the spread of the first farmers in Europe 

posed to an apparent homogeneity among the first 
farming populations of Europe. We also saw that the 
Mahalanobis D2 distances of the Early Neolithic 
groups from (1) Çatal Höyük and (2) Çayönü did not 
reveal any correlation between the geographic distances from each of these sites and Mahalanobis distances measured from remains from other Early Neolithic sites. Moreover, both Çatal Höyük and Çayönü are much closer (morphologically), to the early 
Neolithic European populations than to any of the 
PPNB Levantine groups (i.e., Basta, Abu Hureyra, Jericho, etc.). 

Secondly, the ‘wave of advance’ model explains the 
observed genetic SE–NW cline as being due to the 
gradual absorption of the Mesolithic populations 
(Ammerman and Cavalli-Sforza 1984). If this was 
the case, then one should clearly see some morphological similarities between Mesolithic and Early Neolithic populations due to the impact of absorbing 
some of the Mesolithic gene pool. Furthermore, one 
should observe a clinal increase in similarity between Neolithic and Mesolithic groups, as the advancing wave would progressively be composed of a 
greater and greater proportion of Mesolithic genes. 
However, the craniometric analyses showed no evidence for such a pattern. The analysis revealed very 
limited and not overwhelming evidence of continuity 
of a Mesolithic morphology into Neolithic populations. This evidence was of possible local continuity 
in the case of the Danube Gorge Neolithic, and in 
the western Mediterranean. The analysis of northern 
Europe (region 3) also indicated possible continuity, 
but this could not be further assessed due to the lack 
of specimens from the Early Neolithic and Mesolithic 
periods in this area. The Atlantic European zone was 
not examined due to the lack of appropriate skeletal samples. Yet in the regions in which agriculture 
originated and first appeared in Europe (i.e., regions 1, 4 and 5), we found no evidence for affinities between Early Neolithic populations and the 
preceding Mesolithic inhabitants. 

Thirdly, Barbujani (1995) pointed out that a demic 
diffusion without any absorption of Mesolithic populations can also result in the observed SE–NW cline. 
This has been outlined in various models of stochastic change (i.e., non-selective) due to drift and a series of founder effects (see Wright 1921; 1933; 1951; 
1969; Koningsberg 1990; Relethford 1991; 1996). 
Similarly, Sokal et al. (1991) pointed out that persistent long term demic diffusion originating from a 
single source population, and repeated migrations 
by different ethnic groups along established corri

dors, will leave similar indistinguishable marks on 
gene frequency surfaces. The ‘demic diffusion’ model, 
therefore, does not agree with the observed findings, as no clinal pattern was observed, but rather 
clear morphological affinities between geographically distant Early Neolithic populations. 

Model 2b 

Model 2b proposes a scenario in which the admixture between Mesolithic and Neolithic farmers was 
delayed due to a period of consolidation, and therefore occurred in the Middle/Late Neolithic period rather than during the Early Neolithic period. The corresponding biological pattern should show some affinities between Mesolithic and Middle/Late Neolithic populations in a given region. Such a scenario 
must also presume similarities between Early Neolithic and Middle/Late Neolithic populations in a given 
region, as it would have been the former who underwent a period of population growth and consolidation which continued into the Middle/Late Neolithic. 

Previous analyses of intra-population variability by 
period (Pinhasi 2003) show that the only similarities between Mesolithic and Middle/Late Neolithic 
groups was that in both periods sexual dimorphism 
and variability in general is more pronounced than 
in the case of the Early Neolithic period. However, 
when scrutinising specific indices and other morphological variables by period, it becomes apparent that 
many differences between the populations of these 
two periods exist. 

The analysis of inter-population distances by period 
(Pinhasi 2003) pointed to regional differentiation 
during the Middle/Late Neolithic, with some Late 
Neolithic groups from central Europe and Hungary 
having large distances from each other. We know 
from the settlement pattern analysis (Pinhasi et al. 
2000) that the evidence for Mesolithic occupation in 
south-east Europe is meagre. The apparent regional 
diversification during the Middle and Late Neolithic 
period cannot be associated with a delayed Mesolithic-Neolithic admixture in the case of this region. 
We must, therefore, conclude that a delayed admixture process is not supported by the data from the 
zones analysed in this work. 

Model 3 

The third model proposes a complete population replacement without admixture. To a fair extent, the 
obtained results are compatible with this scenario. 


Ron Pinhasi 

The small distances between Çatal 
Höyük and Early Neolithic European 
populations, and among the latter, 
do not support differentiation by distance. The small distances between 
Visenhäuser Hof, Schwetzingen, Kotacpart and Çatal Höyük suggest that, 
at least in southeast and central Europe, Early Neolithic populations are 
strikingly similar to each other. However, as mentioned before, we found 
some evidence for local continuity in 
the Danube Gorge, as well as possible continuity in Denmark and along 
the Western Mediterranean coast. 
The possibility of admixture between 
the local foragers and incoming farmers in these areas should be taken 
into account. Therefore, this model must be rejected, 
as it does not agree with the complete picture. 

Model 4 

The fourth model proposes that admixture was a 
function of the geographic region under consideration and its relevant distance from the centre of origin of the source population. Various geographic barriers exist in Europe, which more than likely affected the rate and direction of demic diffusion. These 
include the Mediterranean Sea, the Alps, the Pyrenees, and the Carpathians, to name but the conspicuous few. If one adds to this list an obvious preference among the dispersing farmers for fertile soils, 
river valleys and water sources, then the path by 
which the Early farmers dispersed looks much more 
specific and less like a radial dispersal. Moreover, 
the ecological preferences and geographic boundaries cannot be simply taken into account by slightly 
modifying existing formulas for dispersal rates, but 
rather requires an in-depth anthropological approach 
to assess past dispersal patterns. 

Van Andel and Runnels (1995) brought forth a modified model for the dispersal of the first farmers 
into Europe. They based their model on the earliest 
occupation radiocarbon dates used by Ammerman 
and Cavalli-Sforza, plus additional dates from various sources. 

Van Andel and Runnels point out that the wave of 
advance model assumes the following: 

 An initially logistic population growth curve, 

which yields a continuous advance across a broad 
front. 

 Local migratory activity that is, to a first approxi

mation, continuous and random in direction (Fig. 

10.2a). 

This model assumes that population increase and 
migratory activity occurred only at the wave front, 
and that the rate of advance was roughly constant. 
Well behind the wave front, population growth slows 
down due to lack of room for expansion. 

Van Andel and Runnels developed this model taking into account geographic barriers to dispersal. 
They contend that the Aegean Sea might be regarded as a geographic barrier. While sea-travel was apparently possible, it created a bottleneck that limited the number of migrants. Van Andel and Runnels 
(1995) suggest a two-phase colonisation model. At 
the initial stage, colonists from the Levant arrived 
early and almost simultaneously on Crete, at Franchthi Cave and in Thessaly, but probably only in 
small numbers (Fig. 10.2b). In the second step, migrating farmers, possibly from central Anatolia (T2 
in Figure 10.2b) dispersed, reaching the northern 
part of Greece as well as Macedonia and Thrace. Van 
Andel and Runnels (1995) assert that the lengths of 
each step and the intervals between them were dictated by geography and by population growth in 
each of a slowly rising number of parent areas. However, taking a less environmentally-determined approach, one could also argue that the second wave 
could have been triggered by other culturally induced factors. 

Van Andel and Runnels (1995) argue that the original dispersal westwards was by sea routes. They 
propose a dispersal from central Anatolia (Hacilar, 

Fig. 30. The Wave of advance model. (from Ammerman and Caval-
li-Sforza 1984). 

A new model for the spread of the first farmers in Europe 

Can Hassan, and Çatal Höyük) to Crete (Knossos) 
and the southern Peloponnese (Franchthi Cave), 
reaching the Larissa Basin, Thessaly and Macedonia. 
This model is based on the chronological discrepancies between “older” radiocarbon dates from Knossos, Franchthi, Nea Nikomedeia and Sidari, and 
“younger” dates from Macedonia, south Bulgaria 
and the Sava Valley. As has been previously mentioned, Perles (2001) proposed to differentiate between 
“Initial Neolithic” dates in Greece, which cluster 
around 8800 cal BP, and the “Early Neolithic” occupation, which clusters around 8100 cal BP. The initial phase belongs to “pre-pottery” sites, which share 
many similarities with the ‘pre-pottery” sites in Anatolia and the Levant. 

The first dispersal event by sea-route would have 
thus been part of the dispersal of a “pre-pottery” culture which prevailed in Anatolia and the Levant and 
reached Crete, Cyprus, and the southern Peloponnese around 9000 cal BP. The second dispersal event 
would have originated from central Anatolia approximately a millennium later, and Anatolian migrants 
would have appeared in Thessaly and rapidly spread 
across south-east and central Europe. This model is 
illustrated in Figure 32. 

Although the study of Van Andel and Runnels provides a more complex and realistic model of the initial spread of farmers into Europe, involving more 
than one historical event, as a model it is based on 
the concept of demic diffusion. The results obtained 
in this work do not disagree with the demic diffusion model in regards to the logistic, temporal, advance of farmers across Europe. This work did not 
bring forth any new evidence that indisputably supports the assertion of a rapid dispersal. The analysis 
of longitude and latitude of Early Neolithic sites and 
corresponding radiocarbon dates for first occupation 
showed a positive correlation between date and latitude, and a negative correlation between date and 
longitude, which supports a SE–NW linear advance 
pattern of Neolithic settlements (Pinhasi et al. 2000). 
One should then expect to detect a corresponding 
clinal pattern from the craniometric data, similar to 
the one observed for the settlement pattern analysis 
and Cavalli-Sforza’s gene frequencies. However, the 
analysis of skeletal data showed no indication for 
such a cline. The pattern observed is of a logistic dispersal of farmers from the south-east to the northwest, without any morphological cline. 

In sum, while Van Andel and Runnels’ model agrees 
in many aspects with the obtained results, it never

theless works with the underlying assumption of 
‘demic diffusion’, and thus that as farmers dispersed 
they absorbed the local Mesolithic populations. We 
must, therefore, reject this model because we can 
not accept the assumption of a gradually increasing 
input of Mesolithic genes into the Neolithic gene pool 
as the wave of advance progressed north-westward. 

Model 5 

Zilhao (1993; 2001) proposed a model that views 
the spread of a Neolithic lifestyle across Europe as a 
punctuated process with two main pulses. The first 
pulse would have begun around 6800–6400 BP, 
characterised by the spread of farming along (1) the 
Danubian route, and (2) the Mediterranean route. 
According to Zilhao, while the spread of farming 
along the Danubian route was rapid and involved 
the absorption of local Mesolithic groups, the spread 
of farming along the Mediterranean coast was slower, due to the predominance of hunter-gatherer 
groups in these regions. Consequently, hunter-gatherer bands and a more mobile settlement system continued to exist along the western Mediterranean 
shores for some time. 

Fig. 31. The original wave-of-advance model (a), 
modified (b) by the addition of a barrier to gra-
dual movement (sea, desert or mountain range), 
combined with strong preference for a specific but 
rare environment (large floodplains) from Van 
Andel and Runnels 1995). 

Ron Pinhasi 

A second pulse would have occurred after 6000– 5500 BP, in which 
agricultural populations would 
have reached northern Iberia, western France, the Low Countries, the 
British Isles, and Scandinavia. Zilhao contends that in contrast with 
the first pulse, the spread of farming 
in these regions is mainly the result 
of the adoption of these practices by 
local hunter-gatherer groups, rather 
than being due to an incoming wave 
of farmers. 

The results from the regional analysis agree to a fair extent with Zilhao’s model. However, there are a 
few points that must be stressed. Firstly, this model 
examines the later stages of the spread of farming 
in Europe, and thus does not apply to the first stages of the process. Dates for the first pulse post-date 
the arrival of the first farmers in south-east Europe. 
Secondly, the results obtained did not indicate any 
absorption of Mesolithic populations along the Danubian route. The Danube Gorge sites of Lepenski 
Vir and Vlasac, in which local continuity probably 
occurred, are situated in the south-eastern part of 
the river and thus in a zone marginal to the westward direction of dispersal. Thus, the issue is the extent of genetic absorption of hunter-gatherer bands 
proposed in Zilhao’s model. If the demographic process underlying the progression along the Danubian 
route involved but a negligible amount of gene flow 
from Mesolithic bands, then one should not expect 
to detect such a pattern in the craniometric analyses. 

In sum, the results obtained do not contradict Zilhao’s model. However, this model does not examine 
what happened during the preliminary stages of the 
process of spread of farming, and therefore does not 
offer a sufficiently extensive reconstruction to account for the spread of farming as a comprehensive 
historical event. 

A NEW MODEL FOR THE SPREAD OF FARMING 
IN EUROPE 

A new model is proposed. Its main tenets are as follows. The PPNA and PPNB were the periods in which 
we have truly ‘transitional’ agricultural communities. During these periods the Neolithic mode of life 
gradually developed, first with the development of 
domestic wheat, and later with the domestication of 

sheep, goats, cattle and, possibly, pigs. The PPNA period lasted for approximately 800 years, and was 
succeeded by the PPNB period, which lasted approximately 2000 years (see Yakar 1998). The time-
span of these periods implies that the transition from 
a hunter-gatherer semi-sedentary lifestyle, which is 
best exemplified by the Natufians, to the fully sedentary agricultural Neolithic village, which is best exemplified by Çatal Höyük, was not an abrupt change, 
but rather a gradual development. The PPNB cultural zone consists of the Levant, Anatolia and Cyprus, 
and possibly extended further west to other parts of 
the Aegean. The boxplot analysis (Fig. 16) illustrated 
that PPNB populations have an extensive degree of 
morphological heterogeneity. The observed heterogeneity must be associated with one or more of the 
following aspects: 

 Heterogeneity due to geographic/cultural isola

tion between some of these early agricultural populations, which occupied a very large geographical area. 

 Heterogeneity due to morphological differences 
inherited from different ancestral Epipalaeolithic 
populations in the Levant and various Anatolian 
regions as the first farming practices spread throughout a “culture zone”. 

 Heterogeneity due to differential adaptations to 
changes accompanying the transition to agriculture. 

Özdogan (1995) proposes a scheme in which the 
Early Neolithic cultures of Anatolia may be considered as two distinct entities: (1) the Neolithic of southeast Anatolia, which is related to the Mesopotamian-
Levantine tradition, and (2) the indigenous Neolithic 
of the Anatolian plateau. These entities are also separated temporally as the south-east Anatolian cul-

Fig. 32. Van Andel and Runnels’ multi-phase colonisation model 
(Van Andel & Runnels 1995). 

A new model for the spread of the first farmers in Europe 

ture begins about 2000 years before the Central 

Anatolian culture (Özdogan 1999). The former phase 

is characterised by the Neolithic culture of Çayönü, 

while the site of Çatal Höyük belongs to the latter. 

Following this scheme one should expect to see 

more similarities between specimens from Çayönü 

and the Natufians than between the latter and Ça

tal Höyük. However, the results of the analysis per

formed are somewhat contradictory in this respect: 


 The results of the analysis of squared Mahalanobis distances between groups (Fig. 10) indicated 
that the morphological distance between Çayönü and Nahal Oren is larger than the distance between the latter and Çatal Höyük (Pinhasi 2003). 

 The PCA analyses of region 1 (Figs. 20, 21, 22) 
point to a lack of differentiation between the Natufians, Çatal Höyük and Çayönü groups. 

 The discriminant analysis of region 1 indicates 
that the centroid of Çayönü is close to the centroid of the Natufian group, and that they are 
both distant from the Çatal Höyük centroid (Fig. 
23). 

We may therefore conclude that the analyses discussed above do not unequivocally support local continuity between Çayönü and the Natufian populations. 

Nevertheless, various analyses showed that the Çatal 
Höyük group is similar to Early Neolithic European 
groups of south-east and central Europe, and that 
this similarity contrasts with the lack of 
association between the latter and Abu 
Hureyra, Khirokitia and Basta. Based 
on these results it appears that the first 
farmers that colonised Europe did not 
originate from the Near East, but rather 
from central Anatolia. Their best represented type population is the one from 
Çatal Höyük, which represents the successful culmination of the 2000 years 
of agricultural development in Anatolia. This suggests that these farmers 
first arrived in south-east Europe through western Anatolia, and not by sea 
travel through the Greek Islands. The 
remarkable homogeneity among the 
first farmers, taken together with the 
differentiation between them and Mesolithic populations from these regions, implies lack of admixture between 
farmers and hunter-gatherers, and supports an initial ‘logistic dispersal without admixture’. 

Figure 33 illustrates the proposed model. The dark 
area represents the zone of the first farmers. The 
farmers would have arrived and dispersed across 
this zone without any significant admixture with the 
local foragers. The thatched circle in the Balkans represents the Danube Gorge. In this micro-region, the 
Mesolithic population possibly underwent an autochthonous transition to a Neolithic lifestyle, with or 
without some limited admixture with the incoming 
farmers. The spotted grey zone is the zone of the 
Early Impressed Ware culture. 

Around 6000 BP, the dispersal of farmers would 
have continued in two main directions. One group 
of farmers advanced westward along the Mediterranean region, eventually reaching the Iberian Peninsula. This advance would have been of the Early 
Impressed population. The other group represents 
the dispersal to northern and Atlantic Europe. This 
dispersal would have occurred between 6000–4000 
cal BP. The dispersed populations would have been 
the descendants of the first farmers (dark zone) who 
expanded westwards and northwards. This zone 
(northern and Atlantic Europe) was more densely 
occupied by foraging populations, and therefore it 
is postulated that the transition to agriculture in this 
region took different demographic paths. In some 
cases, the foraging populations became Neolithic 
through autochthonous development; in other cases 
the incoming farmers absorbed some local foraging 
tribes. 

Fig. 33. A multi-stage model for the spread of farming across 
Europe. 

Ron Pinhasi 

SUMMARY AND CONCLUSIONS 

This work elaborated upon the complex set of events 
that gave rise to the spread of farming across Europe. In light of the findings from the various craniometric analyses, any simplistic model that explains 
the prevalent pattern of population diversity across 
the continent as an outcome of a single evolutionary 
or historical process was rejected. 

The review of the genetic findings suggests that the 
gene pool of modern European populations displays 
mixed contributions from ‘indigenous’ European Palaeolithic ancestors and from the demic diffusion 
of the first farmers from Anatolia. However, this statement only offers a broad generalisation at a continental level. The main point of argument, therefore, 
is what proportion each group contributed to each 
modern European population, and more importantly, the historical/demographic process that resulted 
in the observed genetic structure of modern European populations. 

The model proposed as the outcome of this study is 
in broad agreement with the genetic findings, in the 
sense that it posits a more complex demographic 
process than was previously postulated by the majority of genetic studies. We saw that an original dispersal of farmers occurred during the 8th millennium 
BP. The source population was most probably located in central Anatolia. The dispersal of the first farmers is, therefore, in agreement with the ‘Demic Diffusion’ model in the temporal sense of a ‘cline’ in radiocarbon dates as one progress from the south-east 
to the north-west of the European continent. However, the demographic aspects of this dispersal cannot be explained by a logistic pattern of absorption 

of local foraging populations. The proposed model 
differentiates between a first expansion without admixture (in the case of most of south-east and central Europe), and a second subsequent expansion 
with some admixture (in the Mediterranean zone of 
Italy and south-east France). This two-phase model 
does not necessarily imply any temporal hiatus or 
change in the rate of dispersal across the continent. 
It does, however, speculate that the first expansion 
was more rapid than that of the subsequent phases. 

Further research is required in order to illuminate 
the nature of the spread of farming in the Western 
European regions. Based on the regional approach, 
one expects to detect temporally based regional variations in the degree of admixture between local 
hunters and migrating farmers. With additional craniometric data from these regions, and the application of corresponding craniometric analyses based 
on the above methodology, it would become possible to bring forth a more detailed model in the near 
future. 

ACKNOWLEDGEMENTS 

I would like to thank Dr. Mihael Budja for inviting 
me to take part in the stimulating and well organised Neolithic Seminar in Ljubljana, in November 
2002. I would also like to thank Dr. Robert Foley and 
Dr. Marta Lahr from the Department of Biological 
Anthropology, University of Cambridge, who supervised and guided my PhD dissertation, and for their 
constructive comments and corrections. Finally, I 
would like to thank all the people at the various research institutions mentioned above who gave me 
access to the skeletal material and hosted me during 
my field work at their institutions. 


A new model for the spread of the first farmers in Europe 

REFERENCES


ACSÁDI G. and NEMESKÉRI J. 1970. History of Human Life Span and Mortality. Budapest: Akadémiai. 

AMMERMAN A. J. and CAVALLI-SFORZA L. L. 1971. 
Measuring the rate of spread of early farming in Europe. Man N.S. 6: 674–688. 

1973. A population model for the diffusion of 
early farming in Europe. In C. Renfrew (ed.), The 
Explanation of Culture Change: 343–357. 

1984. The Neolithic Transition and the Genetics of Populations in Europe. Princeton, NJ: Princeton University Press. 

ANGEL J. L. 1973. Early Neolithic people of Nea Nikomedeia. In I. Schwidetzky (ed.), Die Anfäng des 
Neolithikums vom Orient bis Nordeuropa, Teil 
VIIIa: 103–112. 

BARBUJANI G. 2000. Geographic patterns: how to 
identify them and why. Hum. Biol. 72: 133–153. 

BARBUJANI G. and BARTORELLE G. 2001. Genetics 
and the population history of Europe. Proc. Natl. 
Acad. Sci. 98: 22–25. 

BARBUJANI G. SOKAL R. R. and ODEN N. L. 1995. 
Indo-European origins: a computer-simulation test of 
five hypotheses. Am. J. Phys. Anthrop. 96: 109– 132. 

BARBUJANI G., BERTORELLE G. and CHIKHI L. 1998. 
Evidence for Paleolithic and Neolithic gene flow in 
Europe. Am. J. Hum. Genet. 62: 488–491. 

BEDRICK E. J., LAPIDUS J. and POWELL J. F. 2000. 
Estimating the Mahalanobis distance from mixed continuos and discrete data. Biometrics 56: 394– 401. 

BIAGI P. 1991. The prehistory of the early Atlantic 
period along the Ligurian and Adriatic coasts of 
northern Italy in a Mediterranean perspective. Rivisita di Archeologia 15: 46–54. 

BONSALL C., LENNON R., MCSWEENEY K. and STEWART C. 1997. Mesolithic and Early Neolithic in the 
Iron Gates: a Palaodietary Perspective. J. Eur. Archaeol. 5: 50–92. 

BORONÉANT V. 1996. The Mesolithic complexes in 
the Balkans and Danube Basin. Series Coloquia 7: 
59–76. 

BOYADZIEV Y. D. 1995. Chronology of prehistoric 
cultures in Bulgaria. In D. W. Bailey and I. Panayotov (eds.), Prehistoric Bulgaria: 149–187. 

BRÄUER G. 1988. Osteometrie. In R. Knussmann 
(ed.), Anthropologie: Handbuch der vergleichenden Biologie des Menschen, Vol. I/a: 160–231. 

BREUNIG P. 1987. C14 Chronologie Des Vorderasiatischen, südost-und Mitteleuropäischen Neolithikums. Wien: Böhlau Verlag. 

BRÖSTE K. and JORGENSEN J. B. 1956. Prehistoric 
Man in Denmark. A Study in Physical Anthropology. Vol. 1. Copenhagen: Einar munksgaard. 

BUIKSTRA J. E. and UBELAKER D. H. 1994. Standards for Data Collection from Human Skeletal 
Remains. Arkansas Archaeological Survey Research 
Series No. 44. Fayetteville, Arkansas. 

CAVALLI-SFORZA L. L. 1997. Genes, peoples and languages. Proc. Natl. Acad. Sci. 94: 7719– 7724. 

1996. The spread of agriculture and nomadic pastoralism: insights from genetics, linguistics, and 
archaeology. In D. Harris (ed.), The Origins and 
Spread of Agriculture and Pastoralism in Eurasia: 51–69. 

CAVALLI-SFORZA L. L., MENOZZI P. and PIAZZA A. 
1993. Demic expansions and human evolution. Science 259: 639–646. 

1994. The History and Geography of Human Genes. Princeton, N.J.: Princeton University Press. 

CHERRY J. F. 1981. Pattern and process in the earliest colonization of the Mediterranean islands. Proc. 
Prehist. Soc. 47: 41–68. 

1990. The first colonization of the Mediterranean 
islands: a review of recent research. J. Med. Arch. 

3: 145–221. 
CHIKHI L., DESTRO-BISOL G., PASCALL V., BARAVELLI V., DOBOSZ and BARBUJANI G. 1998a. Clinal variation in the nuclear DNA of Europeans. Hum. Biol. 

70: 643–657. 
1998b. Clines of nuclear DNA markers suggest a 
largely Neolithic ancestry of the European gene 
pool. Proc. Natl. Acad. Sci. 95: 9053– 9058. 


Ron Pinhasi 

CHIKHI L., NICHOLS R. A., BARBUJANI G. and BEAUMONT M. A. 2002. Y genetic data support the Neolithic demic diffusion model. PNAS 99: 11008– 11013. 

COLEMAN J. E. 1992. Greece, the Aegean, and Cyprus. In H. Ehrich (ed.), Chronologies in Old World 
Archaeology. Vol. I.: 247–288. 

DEMOULE J-P and PERLES C. 1993. The Greek Neolithic: A new review. J. World Prehist. 4: 355– 416. 

DEMOULE J-P. 1993. Anatolie et Balkans: la logique 
évolutive du Néolithique Égeen. Anatolica 19: 1– 18. 

DENNELL R. W. 1985. European Economic Prehistory: A New Approach. London: Academic press. 

DUMITRESCU V. 1982. The Prehistory of Romania: 
From earliest times to 1000 BC. In J. Boardman, I. E. 

S. Edwards, N. G. L. Hammond and E. Sollberger 
(eds.), The Cambridge Ancient History. 2nd Ed., 
Vol. III, Part I: 1–74. 
EFSTRATIOU N. 1995. The neolithisation of Europe: 
the Greek experience. Poroilo o raziskovanju paleolitika, neolitika in eneolitika v Sloveniji 22: 63– 

75. 
EHRICH R. W. and BANKOFF H. A. 1992. Geographical and chronological patterns in East-Central and 
Southeastern Europe. In H. Ehrich (ed.), Chronologies in Old World Archaeology. Vol. I: 375–396; 
Vol. II: 341–363. 

FEREMBACH D. 1969. Aprrecu sur le peuplement de 
l’ Anatolie et du moyen-orient au néolithique. Türk 
Tarih Kurumu Belleten 33: 137–143. 

1970. Les hommes du gisment néolithique de Çatal Höyük. Bulletin de la Societé d’Anthropologie du Sud-Ouest 6: 7–14. 

1982. Mesures et indices des squelettes humains 
neolithiques de Çatal Höyük Turquie. Paris: 
CNRS. 

FISCHER R. A. 1937. The wave of advance of advantageous genes. Ann. Eugen. 7: 355–369. 

GARAANIN M. 1982. The Stone Age in the Central 
Balkan Area. In J. Boardman, I. E. S. Edwards, N. G. 

L. Hammond and E. Sollberger (eds.), The Cambridge Ancient History. 2nd Ed., Vol. III, Part I: 75– 
135. 
GIMBUTAS M. 1991. The Civilization of the Goddess. San Francisco: Harper Collins. 

1992. Chronologies of Eastern Europe: Neolithic 
through Early Bronze Age. In Ehrich R. W. (ed.), 

Chronologies in Old World Archaeology. Vol. I: 
295–408; Vol. II: 364–384. 

GRAW M., CZARNETZKI A. and HAFFNER H. T. 1999. 
The form of the supraorbital margin as a criterion in 
identification of sex from the skull: investigations 
based on modern human skulls. Am. J. Phys. Anthrop. 108: 91–96. 

HARDING R. M., RÖSING F. W. and SOKAL R. R. 
1989. Cranial measurements do not support Neolithisation of Europe by demic expansion. Homo 40: 
45–58. 

HENRY D. O. 1989. From Foraging to Agriculture: 
the Levant at the End of the Ice Age. Philadelphia: 
University of Pennsylvania Press. 

HOURS F., AURENCHE O., CAUVIN J., CAUVIN M.-C., 
COPELAND L. and SANLAVILLE P. 1994. Atlas des 
sites du proche-orient 14 000–5700 BP. Travaux de 
la Maison de l’Orient Méditerranéen No. 24. 

HOWELLS W. W. 1973. Cranial Variation in Man: a 
Study by Multivariate Analysis of Patterns of Difference Among Recent Human Populations. Papers of 
the Peabody Museum, Archaeology and Ethnology, 
Vol. 67, Cambridge Mass.: Harvard University Press. 

1989. Skull Shapes and the Map. Papers of the 
Peabody Museum, Archaeology and Ethnology, 
Vol. 79, Cambridge Mass.: Harvard University 
Press. 

1995. Who’s Who in Skulls. Ethnic Identification 
of Crania from Measurements. Papers of the Peabody Museum, Archaeology and Ethnology. Cambridge Mass.: Harvard University Press. 

KAISER T. and VOYTEK B. 1983. Sedentism and economic change in the Balkan Neolithic. J. Anthrop. 
Archaeol. 2: 323–353. 

KEITA S. O. Y. 1990. Studies of ancient crania from 
Northern Africa. Am. J. Phys. Anthrop. 83: 35–48. 

1992. Further studies of crania from ancient Northern Africa: an analysis of crania from First 
Dynasty Egyptian tombs using multiple discrimi


A new model for the spread of the first farmers in Europe 

nant functions. Am. J. Phys. Anthrop. 87: 245– 

254. 
KLECKA W. R. 1980. Discriminant Analysis. New-
bury Park: Sage. 

KONIGSBERG L. W. 1990. Temporal aspects of biological distance: serial correlation and trend in a prehistoric skeletal lineage. Am. J. Phys. Anthrop. 82: 
45–52. 

LAHR M. M. 1996. The Evolution of Modern Human Diversity. A Study of Cranial Variation. Cambridge: Cambridge university Press. 

LAHR M. M., FOLEY R. and PINHASI R. 2000. Expected regional patterns of Mesolithic-Neolithic human 
population admixture in Europe based on archaeological evidence. In C. Renfrew and K. Boyle (eds.), 

Archaeogenetics: DNA and the population prehistory of Europe: 81–88. 

LUCOTTE G. and LOIRAT F. 1999. Y-chromosome 
DNA haplotype 15 in Europe. Hum. Biol. 71: 431– 

437. 
MAKKAY J. 1984. Chronological links between Neolithic cultures of Thessaly and the Middle Danube region. Acta Archaeologica Academiae Scientiarum 
Hungaricae 36: 21–28. 

1988. Cultural groups of SE Europe in the Neolithic: the PIE Homeland problem and the origins 
of the proto-Greeks. AI1N 10: 117–137. 

1992. Neolithic model of Indo-European prehistory. Journal of Indo-European Society 20: 193– 

198. 
MANTEL N. 1967. The detection of disease clustering 
and a generelised regression approach. Cancer Res. 

27: 209–220. 
MARTIN R. 1957. Lehrbuch der Anthropologie in systematischer Darstellung, mit besonderer Beruücksichtigung der anthropologischen Methoden. 3. 
Stuttgart: Fischer. 

MELAART J. and MELLINK M. J. 1992. Anatolian chronology. In R. W. Ehrich (ed.), Chronologies in Old 
World Archaeology. Vol. I.: 207–223; Vol. II: 171– 

184. 
MENOZZI P., PIAZZA A. and CAVALLI-SFORZA L. L. 
1978. Synthetic maps of human gene frequencies in 
Europeans. Science 201: 786–792. 

ÖZDOGAN M. 1993. Vina and Anatolia: A new look 
at the very old problem or redefining Vina Culture 
from the perspective of the Near eastern tradition. 
Anatolica 19: 173–193. 

1995. Neolithic in Turkey. The status of research. 
Readings in prehistory. Studies Presented to Ha-
let Çambel: 41–59. 

1996. Neolithisation of Europe: a view from Anatolia. Part 1: the problem and the evidence of East 
Anatolia. Poroilo o raziskovanju paleolitika, 
neolitika in eneolitika v Sloveniji 20: 25– 61. 

1997. The Beginning of Neolithic economies in 
Southeastern Europe: an Anatolian perspective. 
Journal of European Archaeology 5: 1–33. 

1999. Concluding remarks. In M. Özdogan (ed.), 
Neolithic in Turkey: 225–236. 

PERLES C. 2001. The Early Neolithic in Greece. 
Cambridge: Cambridge University Press. 

PINHASI R. 2003. The Appearance and Dispersion 
of the First Farmers in Europe. A Novel Approach 
to an Old Debate. Unpublished doctorate dissertation, University of Cambridge. 

PINHASI R., FOLEY R. and LAHR M. M. 2000. Spatial 
and temporal patterns in the Mesolithic-Neolithic 
Archaeological record of Europe. In C. Renfrew and 

K. Boyle (eds.), Archaeogenetics: DNA and the population prehistory of Europe: 45–56. 
PLUCIENNIK M. 1997. Radiocarbon determinations 
and the mesolithic-neolithic transition in southern 
Italy. Journal of Mediterranean Archaeology 10: 
115–150. 

PRINZ B. 1987. Mesolithic Adaptations on the Lower 
Danube. Vlasac and the Iron Gates Gorge. British 
Archaeological Series BAR 330. 

RADOVANOVI I. 1996. The Iron Age Mesolithic. 
Michigan: International Monograph in Prehistory, 
Series 11. 

RELETHFORD J. H. 1991. Genetic drift and anthropometric variation in Ireland. Hum. Biol. 63: 155–165. 


Ron Pinhasi 

1996. Genetic drift can obscure population history: problem and solution. Hum. Biol. 68: 29– 

44. 
RENFREW C. 1987. Archaeology and Language. The 
Puzzle of the Indo-European Origins. London: Penguin Books. 

1992. Archaeology, genetics and linguistic diversity. Man N. S. 27: 445–478. 

RICHARDS M., CÔRTE-LEAL H., FORSTER P., MACAULAY V., WILKINSON-HERBOTS H., DEMAINE A., et al. 
1996. Paleolithic and Neolithic lineages in the European mitochondrial gene pool. Am. J. Hum. Genet. 59: 185–203. 

RICHARDS M. B., MACAULAY V. A., BANDELT H. J. 
and SYKES B. C. 1998. Phylogeography of mitochondrial DNA in Western Europe. Ann. Hum. Genet. 

62: 241–260. 
SCHAFER J. L. 1999. NORM: Multiple imputation of 
incomplete multivariate data under a normal model, 
version 2. Software for Windows 95/98/NT, available from http://www.stat.psu.edu/~jls/misoftwa.html. 

SCHWARTZ J. H. 1995. Skeleton Keys. The Introduction to Human Skeletal Morphology: Development, 
and Analysis. New York: Oxford University Press. 

SEMINO O., PASSARINO G., OEFNER P. J., LIN A. A., 
ARBUZOVA S., BECKMAN L. E., DE BENEDICTIS G., 
FRANCALACCI P., KOUVATSI A., LIMBORSKA S., et al. 
2000. The Genetic Legacy of Paleolithic Homo sapiens sapiens in Extant Europeans: A Y-Chromosome 
Perspective. Science 290: 1155–1159. 

SEMINO O., PASSARINO G., BREGA A., FELLOUS M. 
and SANTACHIARA-BENERECETTI A. S. 1996. A view 
of the Neolithic demic diffusion in Europe through 
two Y chromosome-specific markers. Am. J. Hum. 
Genet. 59: 964–968. 

SHERRATT A. 1983. The development of Neolithic 
and Copper Age settlement in the Great Hungarian 
Plain. Part I: the regional setting. Oxford Archaeol. 

2: 13–41. 
SIMONI L., GUERESI P., PETTENER D. and BARBUJANI G. 1999. Patterns of gene flow inferred from 
genetic distances in the Mediterranean region. Hum. 
Biol. 71: 399–415. 

SIMONI L., CALAFEL F., PETTENER D., BERTRANPETIT J. and BARBUJANI G. 2000. Geographic patterns 
of mtDNA diversity in Europe. Am. J. Hum. Genet. 

66: 262–278. 
SOKAL R. R., ODEN N. L. and WILSON C. 1991. Genetic evidence for the spread of agriculture in Europe 
by demic diffusion. Nature 351: 143–144. 

STATISTICA V. 4.3. 1993. User manual. Statsoft. 

TABACHNICK B. G. and FIDELL L. S. 1996. Using 
Multivariate Statistics. New York: Harper Collins. 

THEOCHARIS D. R. 1973. Neolithic Greece. National Bank of Greece. 

THORPE I. J. 1996. The Origins of Agriculture in 
Europe. London: Routhledge. 

TODD I. A. 1980. The Prehistory of Central Anatolia, 1: The Neolithic Period. Studies in Mediterranean Archaeology Vol. 60 Götborg: Paul Aströms 
Förlag. 

TREUIL R. 1989. La Néolithisation dans les Balkans: 
quelques réflexions prématurées. In O. Aurenche 
(ed.), Néolithisations: 129–144. 

TRINGHAM R. 1971. Hunters, Fishers and Farmers 
of Eastern Europe 6000–3000 BC. London: Hutchinson. 

TRUMP D. H. 1980. The Prehistory of the Mediterranean. New Haven: Yale University Press. 

VAN ANDEL T. H. and RUNNELS C. N. 1995. The earliest farmers in Europe. Antiquity 69: 481–500. 

VAN VARK G. N. and SCHAAFSMA W. 1992. Advances in the quantitative analysis of skeletal morphology. In S. Saunders and M. Katzenberg (eds.), 

The Skeletal Biology of Past peoples: Research Methods: 225–257. 

WELLS P. S. 1992. The Neolithic chronology of Central Europe: Southern Germany, Switzerland, and 
Austria. In Ehrich R. W. (ed.), Chronologies in Old 
World Archaeology. Vol. I: 356–374; Vol. II: 326– 

333. 
WHITTLE A. 1996. Europe in the Neolithic. The Creation of New Worlds. Cambridge: Cambridge University Press. 


A new model for the spread of the first farmers in Europe 

1998. Fish, faces and fingers: presence and symbolic identitiees in Mesolithic-Neolithic transition 
in the Carpathian basin. Poroilo o raziskovanju 
paleolitika, neolitika in eneolitika v Sloveniji 

25: 133–151. 
WILKINSON-HERBOTS H. M., RICHARDS M. B., FOSTER P. and SYKES B. C. 1996. Site 73 hypervariable 
region II of the human mitochondrial genome and 
the origin of European populations. Ann. Hum. Genet. 60: 499–508. 

WRIGHT S. 1921 Systems of mating. Genetics 6: 
111– 178. 

1933. Inbreeding and recombination. Proceedings of the National Academy of Science, 
USA 19: 420–433. 

1951. The genetical structure of populations. 
Ann. Eugenics 15: 323–354. 

1969. Evolution and the Genetics of Populations, Vol. 2, The Theory of Gene Frequencies. 
Chicago: University of Chicago Press. 

YAKAR J. 1991. Prehistoric Anatolia. The Neolithic 
Transformation and the Early Chalcolithic Period. 
Monograph Series of the Institute of Archaeology Tel 
Aviv University, No. 9. 

1994. Prehistoric Anatolia. The Neolithic Transformation and the Early Chalcolithic Period. 
Supplement No. 1. Monograph Series of the Institute of Archaeology Tel Aviv University. Monograph No. 9a. 

1998 The socio-economic structure of prehistoric 
communities in the southern Levant, ca. 13 000– 
8000 BP. Poroilo o raziskovanju paleolitika, 
neolitika in eneolitika v Sloveniji 25: 53–63. 

ZILAHO J. 1993. The spread of agro-pastoral economies across Mediterranean Europe: a view from the 
far west. J. Med. Archaeol. 6: 5–63. 

2001. Radiocarbon evidence for maritime pioneer 
colonization at the origins of farming in west Mediterranean Europe. Proc. Nat. Acad. Sci. USA 

98: 14180–14185. 
ZVELEBIL M. 1986 Mesolithic prelude and Neolithic 
revolution. In M. Zvelbil (ed.), Hunters in Transition: Mesolithic Societies of Temperate Eurasia 
and Their Transition to Farming: 5–16. 

1989. On the transition to farming in Europe, or 
what was spreading with the Neolithic: a reply to 
Ammerman. Antiquity 63: 379–383. 

ZVELEBIL M. and ROWLEY-CONWY P. 1984. Transition to farming in Northern Europe: a hunter-gatherer perspective. Norwegian Archaeological Review 

17: 104–128. 
ZVELEBIL M. and ZVELEBIL K. V. M. 1988. Agricultural transition and Indo-European dispersals. Antiquity 62: 574–583. 

ZVELEBIL M. and DOLUKHANOV P. 1991. The Transition to Farming in Eastern and Northern Europe. 

J. World Prehist. 5: 233–278.