RADIOCARBON EVIDENCE OF THE MIDDLE TO UPPER PALAEOLITHIC TRANSITION IN SOUTHWESTERN EUROPE

In the present paper we systematically evaluate the radiometric database underlying the Middle to Upper Palaeolithc transition in southwestern Europe.The different models which attempt to explain the demographical processes underlying this transition rely to a large degree on radiocarbon chronology. We observe that: 1) with increasing age, dates on bone samples show large offsets against those on charcoal, often underestimating these for several thousand years BP and; 2) there is no proof for a persistence of Middle Palaeolithic industries into the time of the earliest Aurignacian in SW Europe. These data contradict the “EbroFrontier” model that distinguishes Late Middle Palaeolithic industries in the SW of the Iberian Peninsula from early Aurignacian ones in the NE. On the contrary, our data 3) imply a model of interregional shifts of populations contracting during severe cold and arid phases and expanding under warmer, interstadial conditions, raising ideas on a regional in situ development of the SW European Aurignacian out of Latest Middle Palaeolithic industries made by Neanderthals some 40.0 kyr cal BC.

Churchill and Smith evaluate the evidence for "the makers of the Early Aurignacian in Europe" and assert modern human presence in Europe "almost certainly" by ca.32.0 kyr BP, and -based on Rieks 1931 discovery of a modern human skull ("Stetten 1") from the base of the Aurignacian layer V of the southern German Vogelherd cave (Riek 1932;1934) -they see a "strong possibility" that modern humans "were there by ca.36 ka BP".Although the new radiocarbon dates from the same stratum vary by several thousand years BP, they confirm the antiquity of the Aurignacian deposits at Vogelherd cave (Conard and Bolus 2003).But stratigraphic attribution of hominid fossils can rarely be established with certainty since skeletal remains may have been reworked in into older layers, a problem especially relevant for old excavations, which are often less well-documented.A good example is the recent dating of a perforated shell (Littorina sp.), found as a grave-good in a Cro-Magnon burial at the famous eponymous rock-shelter of the Gra-vettian period, which had formerly been attributed to an Aurignacian horizon at this site (Henry-Gambier 2002;cf. Djindjan et al. 1999).Being aware of such difficulties Cabrera Valdés et al. (2000: 91) state that the "only well-known human evidence for the time range 40-35 ka is either of Neanderthal or undiagnostic type [...], while the presence of modern human types is not certainly recognized in Europe until about 30 ka".Nevertheless, the majority of models that today undertake efforts to explain the origin of anatomically modern humans is based on the "Out of Africa" hypothesis (Stringer and Andrews 1988;Stringer 2003), postulating colonization of Europe by early Homo sapiens with a simultaneous contraction of Neanderthal dispersal as expressed in the geographical spread of Middle Palaeolithic (MP) sites (Bocquet-Appel and Demars 2000).Recent studies in molecular genetics of fossil Neanderthal remains (Krings et al. 1997;1999;Ovchinnikov et al. 2000) and new finds of early Homo sapiens in Ethiopia (White et al. 2003;cf. Clark et al. 2003) have provided substantial support for the "Out of Africa" hypothesis, whereby anatomically modern humans immigrated from Africa through the Near East into Europe replacing indigenous Neanderthal populations (Stringer 2003).This process -roughly placed between 40.0 and 30.0 kyr ago -is generally assumed to be unidirectional, with modern humans spreading rapidly through Central Europe, finally arriving on the Iberian Peninsula (Zilhão and d'Errico 1999).

THE "EBRO-FRONTIER" AND RECENT CONTROVERSIES
According to the European colonization scenario, anatomically modern humans reached the Iberian Peninsula last.Consequently, most authors today believe that modern humans -following the arrival of the Aurignacian north of the Ebro basin (as represented by the antiquity of early Aurignacian radiocarbon dates from the North of the Iberian Peninsula) -and Neanderthals in the southwestern part of the peninsula coexisted for many thousands of years: i.e. between 40.0 and 30.0 kyr ago (Fig. 2; Zilhão 2000a;2000b;Zilhão and d'Errico 1999).In view of this "Ebro-Frontier"-model the Iberian Peninsula represents the major Neanderthal refugium (Vega Toscano 1993;vgl. Zilhão 1993) before their final replacement during the later Aurig- nacian, following ca.33,500 14 C BP (Vega Toscano 1990;Zilhão 1993).
This paradigm of a quite recent colonization of Europe by anatomically modern humans has been challenged during recent years by an alternative model of an in situ development of the Aurignacian out of the preceding regional Latest Middle Palaeolithic (LMP), based on the results of new excavations at El Castillo cave (Cabrera Valdés et al. 2001).
Due to the nature of population advances and the sparse typological evidence (i.e. the dispersal of Aurignacian I type split-based points in the northern part of the Iberian Peninsula; cf . Zilhão 2000a;2000b) the Ebro Frontier model relies to a great extent on methods of chronometric dating, each with its own implicit chronological notions.The alternative hypothesis of an Aurignacian in situ development in SW Europe is based on a series of chronometric data, most of which are also radiocarbon measurements.
In the present paper we have aimed to test the "Ebro-Frontier"-model and to evaluate the possibility of an Aurignacian in situ development, using the "Stage Three Project database" of last interpleniglacial (Oxygene Isotope Stage 3 = OIS 3) radiometric dates for European archaeological sites (1), completed for the Iberian Peninsula as a part of assessments of the demographic processes underlying the transition from the LMP to the Earliest Upper Palaeolithic (EUP) in SW Europe.

THE MIDDLE TO UPPER PALAEOLITHIC TRANSITION AND AURIGNACIAN ORIGINS
Throughout the different regions of Europe, including its southwestern part, i.e.France and the Iberian Peninsula, regional archaeological records best displayed in numerous stratigraphies of cave sites and rock shelters, show exclusively Aurignacian industries post-dating LMP ones attributed to Neanderthals (Fig. 1).At several sites the latest MP is found in sediments formed under interstadial conditions of moderate to temperate climate attributed to the Hengelo-period (Carbonell et al. 2000), the most significant warm interval in the second half of OIS 3 (2).
(3) Besides the rich evidence of Châtelperronien in France (Bosinski 1987;Demars 1996), in Northern Spain, level 10 from Cueva Morín, the small inventories from Labeko Koba IX and Ekain X, and those from El Pendo and A Valiña with their stratigraphical problems (Maíllo Fernández 2003) are the only inventories known from Cantabria.From Catalunya only few Châtelperronian points have been recorded embedded in Aurignacian inventories (cf.Canal i Roquet and Carbonell i Roura 1989: 337: l'Arbreda, Cova Pau, Reclau Viver).Besides the problems concrening El Pendo (Montes Barquín and Sanguino González 2001), stratigraphical disturbances may also account to explain the interstratifications of Châtelperronian with Aurignacian levels at Roc de Combe (Rigaud 2001) and Le Piage in France (d 'Errico et al. 1998;Rigaud 2001).ten sandwiched between LMP and Aurignacian layers (Carbonell et al. 2000;d'Errico et al. 1998).The few hominid fossil remains currently known may imply that the Châtelperronian should also to be linked to Neanderthals (Hublin et al. 1996;cf. Asmus 1964).Therefore, one may assign the Châtelperronian to the Middle Palaeolithic, as can be also argued from the evolution of lithic technology (Gouedo 1990;cf. Bodu 1990) and tool spectra (de Sonneville -Bordes 1972;Bosinski 1987;1990), both showing links to LMP industries.
In summary, the MP/UP transition of SW Europe comprises a well-established sequence of LMP, Châtelperronian-, and EUP Aurignacian-type techno-complexes (Fig. 1).

Radiocarbon evidence
During recent years numerous chronometric dates have been assembled from a large number of sites dating to the LMP and EUP on the European continent (Bocquet-Appel and Demars 2000; Davies 2001).The dates have been compared and interpreted in context with archaeological findings (d 'Errico and Sánchez Goñi 2003;Zilhão and d'Errico 1999) (4).Most of the dates, with radiocarbon-measurements forming the bulk of dates from the entire corpus of radiometric age-determinations, have been obtained from sites in SW Europe, alltogether providing more than 700 radiocarbon measurements >17,500 14 C BP, backed by more than 200 non-radiocarbon dates (TL/OSL; Useries; ESR).Due to its comprehensiveness the combined date list is a valuable instrument for studies on the demographic processes underlying the models in question.
However, due to a number of limitations of the different chronometric dating methods employed, in particular radiocarbon, caution in the meaning and interpretation of dates has repeatedly been expressed (e.g.Djindjian 1999;Pettitt 1999;Pettitt and Pike 2001).It is also most important to acknowledge that many of the 14 C-measurements on which the MP/UP transition is based range close to the technical limits of the dating equipment.With increasing age the dates become, as a rule, less reliable (one of the principles of the method underlying radiocarbon dating).This is reflected in increas-ing standard deviations as well as in the higher number of infinite ('greater than') age determinations -both parametres that largely depend on the technical equipment of the laboratory.
In the past developing radiocarbon methods did indeed produce a few dozen age determinations in the range of 10.0-12.5 half-lifes of radiocarbon.These were measured on large peat samples (ca. 100 g carbon) by a combination of thermal isotope enrichment with large 14 C-ß-counting systems (cf.Grootes 1977).Today, however, the technical limits of most modern laboratories for routine 14 C-(AMS)-measurements remain close to 9.5 half-lifes i.e. ca.55,000 14 C BP (Fig. 3).While some radiocarbon laboratories promise reliability of their measurements up to at least 40,000 14 C BP (Hedges and Pettitt 1999), others have much lower technical age limits around 5.0-7.5 half-lives of radiocarbon age (5), i.e. the time of transition from LMP to EUP.Such problems principally effect radiocarbon dating and, consequently, the interpretation of dating results.
The most comprehensive study so far on the beginning of the EUP and the earliest occurences of the Aurignacian in Europe has been undertaken by Zilhão and d'Errico (1999) who, using a selected data set, (1) confirm the validity of the "Ebro-Frontier"-model and (2) state that no Aurignacian older than 36,500 14 C BP (i.e.north of the Ebro basin) withstands their criteria of evaluation of sample taphonomy (6).This picture of earliest Aurignacian presence in Europe results from the omission of all radiocarbon dates older than 36,500 14 C BP on charcoal samples that appear systematically older relative to bone.
(5) Despite corresponding claims documented in the proposed analytical dating errors, we recognise that some radiocarbon laboratories, notably in earlier years, cannot achieve reproducible 14 C-measurements of such high age.A fair portion of the equipment earlier used in ß-decay counting quite apparently had rather large and at any rate often widely varying counter backgrounds, with statistical variations not always according to Poisson statistics, so that the reliability of the archaeological radiocarbon data available today is not in all cases beyond reasonable credence.Quality and reliability of dates thus relate to the age of the laboratory and the year a specific sample had been dated, rather than systematic offsets between conventional and mass spectrometric methods of radiocarbon dating as stated by d 'Errico and Sánchez Goñi (2003).
(6) It is noteworthy that any dates are worthless without interpretation of the circumstances that may have contributed to the date, but the taphonomy and context of samples for radiometric dating are difficult to evaluate years after excavation.Any such 'evaluation' finally remains a selection judged by the authors' personal criteria.Testing the available data for integrity of their statistical properties should, at least, be carried out.

A material matter
In their analyses of data from the MP/UP transition Zilhão and d'Errico (1999) recognize that bone samples systematically turn out younger than charcoal ones, often diverging by several thousand 14 C BP.This phenomenon is independently and best illustrated by the example of the recently published 14 C data series from Kostienki 14 in Russia (Sinitsyn et al. 2002) (7).The main reason for Zilhão and d'Errico (1999) to trust the bone dates more than the ones produced from charcoal is that charcoal is often judged as 'mobile' material sensitive to stratigraphical disturbance, e.g. by processes of bioturbation.Based on the physical properties of the radiocarbon method of dating, there is no reason to rely more on the dating of bone rather than on charcoal samples, especially when the archaeological context is clearly identified, as is most likely to be the case with charcoal from hearths.For a variety of reasons, namely (1) the smaller initial content of datable carbon, (2) the well-known higher susceptibility of bone towards contamination with younger carbon, and (3) the disappearance of bone collagen due to decay (e.g.Schwarcz 2001), higher reliability should be given charcoal versus bone samples, especially when the samples are close to the detection limits of the radiocarbon method of dating.
Taken at face value, the 14 C age-distributions of the MP and UP of SW Europe more or less overlap in the entire period from around 38,000 to 19,000 14 C BP (Fig. 4, left), creating a "Coexistence Effect" (Conard and Bolus 2003).This results from the exponentially decreasing rate of 14 C-decay and the rapidly increasing susceptibility towards contaminaton in older samples, as shown by theoretical estimates of the surplus of modern carbon on dating (Grootes 1977).Due to the fact that the youngest LMP samples exclusively derive from samples of bone or burnt (cremated) bone, their young age-determinations are most likely due to effects of contamination and alterations of the physical properties of sample material during processes of combustion (cf.Schwarcz 2001; Gillespie 1997) and/or to difficulties in the techniques and methods of dating burnt bone (Lanting and Brindley 1999;cf. Gillespie 1997).It is often impossible to establish whether the material is contaminated or not, and apart from obvious anomalous samples, it is still difficult to judge until which age LMP dates may be considered reliable.
Having filtered for sample material type (i.e.bone versus charcoal), radiocarbon age-distributions for bone dates (Fig. 4, centre) appear similar to those that have already appeared in the entire data set, while a completely different picture of agedistributions is given with charcoal (Fig. 4, right).This is most clearly seen in MP charcoal dates that appear systematically older than those obtained on bone.Direct comparisons of both bone and charcoal samples, as undertaken exemplarily for the SW European MP on the one hand (Fig. 5, left) and the Aurignacian of the same region on the other (Fig. 5, centre; cf.Zilhão and d'Errico 1999), emphasize  In the construction method applied to the resulting 14 C-dispersion graphs, each individual radiocarbon date has been defined by its given median value and standard deviation.The corresponding individual Gaussian curves have been added, to give a curve of the summed 14 C dating probability (Geyh 1969).Because each Gaussian curve is normalised with equal area, using this method each date/sample is given equal weight, independent of dating precision.
the complex difficulties with which radiocarbonbone samples are generally endowed (Jöris et al. 2001;Schwarcz 2001).Demographic modelling of the processes underlying the MP/UP transition (Fig. 5, right) is thus strongly affected by the material dated.

Aurignacian origins
Against the background that different sample material may result in entirely different age-distributions, an 'alternative' radiocarbon chronology based solely on charcoal samples for the earliest occurences of the Aurignacian in Europe (1) pushes back earliest appearance of the Aurignacian to ca. 38,300 14 C BP and (2) produces two distinct geographical clusters of possible Aurignacian origin (Fig. 6), i.e. the southeastern Central European region to the East and the Pyrenean and Cantabrian area to the SW (Jöris 2003).(3) The results do not, however, reflect a geographical pattern that one would assume in a model of anatomically modern humans colonizing Europe from the East to the West.
Since the Aurignacian lacks any readily apparent 'cultural predecessors' outside of Europe, it cannot be excluded that this culture may have evolved locally within these two regions in Europe (cf.Cabrera Valdés et al. 2001).The geographical dispersal of the oldest sites dated using 14 C-charcoal samples could imply Aurignacian genesis out of LMP leaf-point industries in the East and out of the Châtelperronian in the SW.In these regions youngest reliable radiocarbon dates on charcoal for the LMP range between 38,800 and 35,900 14 C BP.In the construction method applied to the resulting 14 C-dispersion graphs, each individual radiocarbon date has been defined by its given median value and standard deviation.The corresponding individual Gaussian curves have been added, to give a curve of the summed 14 C dating probability (Geyh 1969).Because each Gaussian curve is normalised with equal area, by this method each date/sample is given equal weight, independent of dating precision.Fig. 6.Europe at the Middle to Upper Palaeolithic transition with Châtelperronian sites in the SW (Bosinski 1987;cf. Demars 1996), Uluzzian in Italy (Gioia 1990; Palma di Cesnola 1993), and leaf-point industries in south-central and southeastern Europe (Allsworth-Jones 1986; Bolus and Rück 2000) as well as 'last appearance data' (LAD's) for Middle Palaeolithic levels <40,000 and >35,000 14 C BP (white underlain) and East European leaf-point industries (grey underlain) in comparison with first appearance data (FAD's) for the Aurignacian in Europe based on radiocarbon dates >35,000 14 C BP (black underlain).All radiocarbon FAD's and LAD's derive from charcoal samples and are given in kyr 14 C BP, in case of repeated measurements given as weighted means (cf.Jöris 2003; cf.Tab.1-2).Palaeogeography of Europe corresponds to the glacial maximum of the last glaciation at around 22.5 kyr cal BC with ice margins dotted and coastline lowered for some 120 m.

THE EARLIEST AURIGNACIAN AND THE LATEST MIDDLE PALAEOLITHIC OF SOUTHWESTERN EUROPE
The question of whether or not the Aurignacian originated from the regional LMP or is the result of anatomically modern humans colonizing Europe can best be approached from the point of view of available radiometric data (Tab.1-2).In terms of this the "Ebro-Frontier"-model between ca.40,000 and 30,000 14 C BP is testable, since it requires two basic lines of evidence, which are (1) proof for Aurignacian presence NE of the Ebro basin at times for which (2) contemporaneous LMP is manifested in the region SW.Both lines can contribute to the colonization model in question, assuming that LMP lithic industries indeed represent Neanderthals, and that anatomically modern humans are the artisans of the Aurignacian.
While first appearance data (FAD's) of the Aurignacian in Europe -based on 14 C-charcoal samples -indeed demonstrate an early Aurignacian presence NE of the Ebro basin (Fig. 6), the second line of evidence is especially problematic to establish.Problems arise -as described in the preceding chapter-, in determining up to which age young LMP radiocarbon dates can still be regarded as reliable and, more importantly since only a few LMP sites in the southern part of the Iberian Peninsula are published in detail, with the exception of Gibraltar.
To derive 'realistic' latest-appearance estimates (LAD's) for LMP sites of that region, it is necessary to compare the radiocarbon dates availabe with nonradiocarbon age-determinations.But direct data pairs of different dating methods, radiocarbon versus non-radiocarbon, are unfortunately rare.Furthermore, it is necessary to acknowledge that the radiocarbon time-scale requires calibration to cor-  rect for variations in past atmospheric 14 C-levels (Fig. 7) (8) in order to achieve compatability with the results from other non-radiocarbon methods of dating.When calibrated, the 14 C-time interval covering the MP/UP transition proposed by different authors roughly corresponds to the period 40.6 -30.2 kyr cal BC (Tab.3), showing a remarkably clear distinction between LMP versus Aurignacian age-distributions on the calendric scale (Fig. 8).The oldest Aurignacian sites in SW Europe dated by radiocarbon are situated in the North of the Iberian Peninsula (Fig. 2; Fig. 9).These sites -in sum -form the base of the first line of arguments for the "Ebro-Frontier"-model, stating that the Aurignacian starts around 36,500 14 C BP in most of Europe (Zilhão and d'Errico 1999), whereas the Aurignacian in the southwestern part of the Iberian Peninsula does not date before 33,500 14 C BP (Vega Toscano 1990;Zilhão 1993).
Evidence for an Aurignacian presence in SW Europe dating prior to 36,500 14 C BP is sparse, and a single radiocarbon age on bone dated to 40,000 ± 1400 14 C BP (OxA-3727) obtained from the basal layer at Reclau Viver may be due to a reworked sample (Zilhão and d'Errico 1999).The upper part of the same level contains an Aurignacian I with split-based points (Canal i Roquet and Carbonell i Roura 1989).Similar problems seem to account for a radiocarbon measurement (GifA-97185: 37,200 ± 150 14 C BP; bone) from layer G at Caminade-Est, where "systematic refitting work carried out by J.-G.Bordes (1998) has demonstrated that around 30% of the archaeological material included in this layer comes from the underlying Mousterian deposits, to which the dated sample could conceivably be related" (Zilhão and d'Errico 1999, 17).The overlying layer F at this site has produced an Aurignacian I with split-based points, and a radiocarbon date of 35,400 ± 1100 14 C BP (GifA-97186) on bone (Rigaud 2001).
Whereas early Aurignacian sites are dispersed in SW France and Northern Spain (Fig. 2), oldest radiocarbon evidence for the Aurignacian (Tab.1; Fig. 9) is exclusively restricted to the provinces of Cantabria and Catalunya in Northern Spain.Here the sites of Abric Romaní, L'Arbreda and El Castillo are most controversially discussed, due to their history of research, the antiquity of radiometric measurements with radiocarbon dates going back to > 36,500 14 C BP, and typological arguments.
A series of radiocarbon measurements of charcoal samples from the new excavations of V. Cabrera Valdés immediately in front of the cave entrance at El Castillo gave results ranging between 41,100 ± 1700 (OxA-2477) and 37,100 ± 2200 (OxA-2473) 14 C BP (Tab.2) (Cabrera Valdés and Bischoff 1989;Cabrera Valdés et al. 1996, 2000;2001) for a layer inside the cave labeled "Aurignacian Delta" by H. Obermaier (Cabrera Valdés 1984;cf. Cabrera Valdés et al. 1996).Zilhão and d'Errico (1999) critically request (1) the stratigraphic correlation of the layer outside the cave with the inner deposits, (2) the cultural attribution of the recent finds, and (3) the association of the samples dated with the lithic industry, which they think has more affinities with Mousterian or Châtelperronian than the Aurignacian.According to Zilhão and d'Errico (1999), the Aurignacian of El Castillo is present only inside the cave in the upper portion of Obermaiers "Aurignacian Delta", while the lower part of this layer contains more MP types, and likely (8) We note, however, that all radiocarbon "age"-estimates are measured on the conventional 14 C-scale, which is very precisely defined as a dimensionless logarithmic ratio (Mook 1983), so that an independent calibration of all 14 C-ages is necessary if we wish to sensibly discuss the chronological implications of the 14 Cdata.Intensive progress in the construction of the calibration curves required for such age-transfer of (uncalibrated) radiocarbon dates into the calendric dimension has been made during the last few years (eg.Jöris and Weninger 1999a;1999b;2000; http:// www.calpal.de:CALPAL-2003; cf.van der Plicht 2002), with the period in question being best recorded in the data sets from Suigetsu in Japan (atmospheric; Kitagawa and van der Plicht 2000) and PS2644 in the North Atlantic (marine; Voelker et al. 2000).Between 46.0 and 28.0 kyr 14 C BP (45.0-30.0kyr cal BC) the combined 'synthetical' calibration data set (CALPAL-2003) allows for rough calendric estimates of the 14 C-time scale (Fig. 7), representing a pattern of highly fluctuating 14 C-levels, with periods of extremely high production of radiocarbon (steep parts in the calibration curve), i.e. between 42.0 and 35.0 kyr 14 C BP, and others with limited production rates, resulting in long 14 C-age-plateaux, i.e. between 39.0 and 35.0 kyr cal BC, normally accompanied with extreme age-distortions (cf.Beck et al. 2001).It is this complex 'pattern' of the calibration curve that allows for improvement of dating precison for periods of rising atmospheric radiocarbon contents as well as the age-distortions which make it difficult to interpret radiocarbon dates that fall into the period of extended plateaux.
Due to the method of construction of the CALPAL-2003 record, i.e. the transferral of radiocarbon-dates from marine oxygene isotope records into the GISP2 Greenland ice core age model (Jöris and Weninger 1998;1999a;1999b;cf. Voelker et al. 1998;2002), 14 C-ages are reliably linked with palaeoclimate signatures (Fig. 7).This allows for high-precision age-transferral of 14 C-measurements into a calendrical age-model, combined with its positioning within the record of OIS 3 palaeoclimate change that is characterised by extremely rapid oscillations between cold and dry (glacial) and more temperate (interstadial) conditions (for the Iberian Peninsula cf.Carbonell et al. 2000;d'Errico and Sánchez Goñi 2003).correlates with layer 18 of V. Cabrera Valdés' excavation.Due to the fact that Obermaier did not observe a sterile horizon between the upper and lower portions of the "Aurignacian Delta"-sediment, one may conclude that the Aurignacian immediately followed 38,679 ± 744 14 C BP, which is the weighted mean (WM) of five radiocarbon measurements on charcoal samples from sub-unit b of layer 18.The underlying sub-unit c has produced five additional radiocarbon dates-again all on charcoal -resulting in a WM of 40,621 ± 750 14 C BP paired with three ESR dates on bone (Rink et al. 1996;WM: 38.4 ± 2.6 kyr BP).
However, a consistent group of radiocarbon agedeterminations (Tab.1), significantly older than 36,500 14 C BP, has been produced for the Aurigna-cian layers of L'Arbreda (Bischoff et al. 1989;Canal i Roquet and Carbonell i Roura 1989).Four charcoal samples taken in an artificial 5cm-horizon (5,50-5,55m below surface) in square E2 (BE 111) immediately sheltered by the travertine wall in the lower Aurignacian, have produced a WM of 38,307 ± 552 14 C BP (Bischoff et al. 1989).Furthermore a radiocarbon measurement of 35,480 ± 820 14 C BP (OxA-3730) on bone assigned to the same cultural unit is available.A few metres to one side (CE 103) a radiocarbon measurement, again on bone, ages the upper Aurignacian level to 37,340 ± 1000 14 C BP (OxA-3729), whereas one further measurement (Gif-6422) obtained earlier, is considered to be too young.Against the homogeneity of the available data we can hardly follow the vague assumptions of Fig. 9. Radiocarbon dates (first appearance data: FAD's) for the SW European Aurignacian, sorted for material dated and given in kyr 14 C BP, when required as weighted means.Zilhão and d'Errico (1999) who claim possible stratigraphical disturbances in the various parts of the area excavated, and cannot assign the spatial origin of samples to areas that may be stratigraphically problematic.On the contrary:The samples from square E2 come from the very base of the Aurignacian deposits, immediately next to the main profile published by Canal i Roquet and Carbonell i Roura (1989).
Although the Aurignacian of L'Arbreda displays some similarities with that of "layer 2" at Abric Romaní (e.g. in terms of the presence of Dufour bladelets; Carbonell et al. 2000, 18), G. Laplace and N. Soler have emphasized the likelihood that the small assemblage from this layer may have resulted out of palimpsests of different occupations, one during the late Gravettian, another during the Aurignacian, based on the similarities of some (6) backed tools with Gravettian points (Canal i Roquet and Carbonell i Roura 1989; Carbonell et al. 1994).This 'Aurignacian' layer from the initial excavations conducted by A. Romaní between 1909 and 1929 ("layer 2") is preserved in remnant deposits -labeled "level A" -along the back wall of the central part of the rock-shelter as well as in the Coveta Nord, where it is stratified between two travertine horizons (Bischoff et al. 1994)."Level A" -al-though only 2cm in thickness -produced "abundant faunal remains, dispersed charcoal and artifacts" (Bischoff et al. 1994: 544) and most likely represents a living floor sealed by the overlying travertine.Seven radiocarbon measurements on charcoals were obtained from the remnant "level A" deposits at three different places of the site all together.Five of these form a WM of 36,644 ± 373 14 C BP, with individual dates ranging from 37,290 ± 990 (AA-7395) to 36,390 ± 629 (WM of AA-8037A and AA-8037B) 14 C BP.These dates are in strong contrast to two measurements from the radiocarbon laboratory in Waikato, New Zealand, which -due to possible contamination with younger carbon -are significantly younger (NZA-1817; NZA-1818).The same laboratory has also produced a date of 36,590 ± 640 14 C BP (NZA-2311, contained in the WM given above).All these dates are stratigraphically consistent with a further radiocarbon measurement of a charcoal sample that was embedded in the travertine (USGS-2839: 36,600 ± 1300 14 C BP; Bischoff et al. 1994).Furthermore, the great antiquity of these deposits is confirmed by U-series dating of this travertine (Bischoff et al. 1994; with corrected U/Th-ages of between 39.1 and 42.9 kyr BP given in Carbonell et al. 1994) ( 9).Concerning the difficulties with the archaeological assignment of "layer 2" Zilhão and d'Errico (1999), plead for the stratigraphical reliability of this sequence.Whereas it is highly likely that the radiocarbon measurements from "level A" material do indeed date the Aurignacian, stratigraphical mixing on top of the Aurignacian layer may also have played a role, close to the wall of the rock-shelter, and only a few metres away from the sampled area (as observed by A. Romaní).
Contra the interpretation of Zilhão and d'Errico (1999) the evidence from Castillo, L'Arbreda, and Romaní strongly indicates earliest Aurignacian presence in the North of the Iberian Peninsula before 36,500 14 C BP (Fig. 10), confirmed by the constistency of both radiometric dating and stratigraphy.Insignificantly younger radiocarbon measurements -all on charcoal -derive from the lower part of layer XIII at La Viña (Ly-6390: 36,500 ± 750 14 C BP), from Isturitz (WM of two dates: 35,490 ± 413 14 C BP), and possibly at Tournal G (oldest date: Ly-1898: >35,800 14 C BP).
(9) Although less precise than radiocarbon dates, due to the high standard deviations, the results from U-series dating at Romaní approximately fit the ranges of the calibrated radiocarbon measurements (cf.CALPAL-2003: http://www.calpal.de).Towards the North (north of 44.0° N; Fig. 9) radiocarbon FAD's of the oldest regional Aurignacian turn out slightly younger (starting with 35,400 ± 1100 14 C BP  in layer F of Caminade-Est; Tab.1), and in the south of the Iberian Peninsula, the oldest Aurignacian is present in Beneito VIII (AA-1388: 33,900 ± 1100 14 C BP on charcoal).Further to the South and to the West radiocarbon dates for the Aurignacian are even younger than 33,500 14 C BP (Vega Toscano 1990;Zilhão 1993), with the Portugese site of Gato Preto, dated by TL to 38.1 ± 3.9 kyr BP, representing the oldest Aurignacian in the SW of the Iberian Peninsula.

... and evidence for the latest Middle Palaeolithic
For SW Europe the "Ebro-Frontier"-model assumes LMP industries at times when Aurignacian people had already populated the North and Northeast of the Iberian Peninsula (Zilhão 1993; cf.Fig. 2).For the second line of arguments underlying the "Ebro-Frontier"-model it is thus important to verify the contemporaneity of LMP industries in the SW of the Iberian Peninsula with the earliest Aurignacian in the North and Northeast, i.e. evidence persisting significantly later than 38,300 14 C BP.This is all the more difficult since Aurignacian layers in general overlay LMP ones and are subject to some reworking of deposits.Relying more on 14 C-bone data for the MP/UP transition, Zilhão and d'Errico (1999: 10) note the "apparent contradiction between stratigraphy and dating" because the radiometric dates do imply a significant chronological overlap.They try to explain this contradiction as "simply an artifact of serious errors of method and interpretation in the use of radiometric results".
Due to the problems of possible sample contamination and the fact that -as a trend -radiocarbon charcoal data are shown to be more reliable than bone data, we will base our study solely on radiocarbon dates on charcoal as well as on the chronometric data available from the different non-radiocarbon methods of dating (e.g.TL/OSL, U-series; ESR, Tab.2; Fig. 11), in order to trace reliable LMP LAD's in SW Europe Regarding the charcoal-dated LMP (incl.Châtelperronian) of SW Europe only two sites -Fuentes de San Cristobal (Mousterian) and Morín (layer 10: Châtelperronian) -have produced finite radiocarbon age-determinations younger than 38,300 14 C BP.At Morín the same sample (SI-951) has produced two contradictory radiocarbon measurements, one with an extremely high standard deviation of several thousand BP.These measurements along with a single sample (OxA-8591) for the LMP at Fuentes de San Cristobal with a high standard deviation (± 1900), cannot be interpreted as reliable proof for a long persistence of the LMP in SW-Europe.
One sample from the basal LMP layer X at Beneito in the SE of the Iberian Peninsula has been dated to 30,160 ± 680 14 C BP (unknown lab.-no.), whereas a second sample (AA-1387) on charcoal, obtained from the same layer, dates to 38,800 ± 1900 14 C BP and is thus similar in age to the WM of the radiocarbon measurements available for the LMP layer 18b of El Castillo (38,679 ± 744).Sample AA-1388 from the overlaying Aurignacian level VIII at Beneito confirms the antiquity of the AA-1387 sample.
Other sites that have produced radiocarbon agedeterminations younger than 38,300 14 C BP have either been dated several decades ago (e.g.Ermitons, Quinçay, Tournal, Brugas) or the dates have been obtained from "inadequate dating material" (Zilhão 2000a: e.g. Columbeira) that is likely to be contamined with modern carbon (Zilhão 1997: 35).Possible contamination is furthermore likely for some radiocarbon measurements from LMP layers that have been independently dated by non-radiocarbon methods.At Zafarraya (Hublin et al. 1995), for example, teeth of Capra ibex were dated by radiocarbon as well as by U-series methods, but due to a variety of reasons mostly related to the stratigraphical properties of this site, these samples are to be regarded "highly questionable and/or irrelevant" in contribution to the issue discussed (Pettitt and Pike 2001: 416).A 'transitional industry' from Gorham's Cave may be due to stratigraphic palimpsests, since both radiocarbon measurements obtained on charcoal diverge by some 10,000 14 C BP (OxA-7857 for context 24: 32,280 ± 420 14 C BP vs. OxA-7791 for context 18: 42,000 ± 1100 14 C BP). Stratigraphical inversion of layers most plausibly account for the El Pendo series (Montes Barquín 2000;Montes Barquín and Sanguino González 2001).
At Brugas and Columbeira non-radiocarbon age-determinations all dating older than 45.0 kyr ago have produced results significantly older than radiocarbon measurements.In contrast, two radiocarbon measurements on charcoal at Gorham's Cave (context 22/22D) are much older than the OSL-dated sediment.At Figueira Brava a radiocarbon measurement on a Patella sp.shell has been dated to 30,930 ± 700 14 C BP .Although Zilhão (2000) attributes the date to the LMP level 2, its precise stratigraphic proveniance remains unclear.Two U-series measurements of the same sample gave strongly divergent results with extremely high standard deviations.Whereas the younger date was given as 30.6 ± 11.8 kyr BP (SMU-232E1), the older one resulted in 44.8 ± 15.9 kyr BP (SMU-232E2).
For the period under discussion in general, nonradiocarbon methods of dating have to be regarded as less reliable due to the often extremely high standard deviations, when compared with 14 C (e.g.Schwarcz 2001).Even in cases where radiocarbon measurements appear in agreement with results obtained from the various non-radiocarbon dating techniques, this may be simply coincidental (10).
The LMP from Tournal C is dated to 33,600 ± 1300 14 C BP (Ly-1667) by radiocarbon and to 33.0 ± 3.6 kyr BP by TL on burnt silex (WM of two dates), whereas a sample from the overlaying Aurignacian of level G is older than 35,800 14 C BP (Ly-1898;Tavoso 1976;cf. Tab. 1).The Châtelperronian of Combe Sauniere has been dated by ESR (36.4 ± 2.5 kyr BP; Tab. 2) as well as by radiocarbon on two bone samples.Although the WM of the radiocarbon measurements (35,449 ± 572 14 C BP) is largely in agreement with the date produced by ESR, the individual measurements are strongly divergent.Moreover, OxA-6503 gave a date of 35,900 ± 1100 14 C BP, whereas a tripeptide-measurement of the same sample resulted in 38,100 ± 1000 14 C BP, close to the date that we propose for the earliest Aurignacian in the North of the Iberian Peninsula (L'Arbreda; cf.Fig. 10).
Non-radiocarbon dates younger than 30.0 kyr BP have been obtained from Bajondillo (ESR and U-series) with a WM of 26.7 ± 1.3 kyr BP and from Conceicão (TL), for a level overlaying the LMP (QTSL-CNC-11: 27.2 ± 2.5 kyr BP).TL-dates from Roche à Pierrot (St.Césaire) and Carigüela -although rather heterogeneous -indicate that LMP layers are older than 40.0 kyr BP.The same accounts for the TL and ESR-measurements obtained from the Le Moustier sequence.
A low 230 Th/ 232 Th-ratio explains the relatively young age for the MP finds from the EVS-cone at Almonda (Zilhão 2000b), and low U-contents for the samples SMU-248 and SMU-249 at Gruta do Escoural (Zilhão 2000b) point to the higher reliabi-lity of the SMU-250-sample, giving an age of 48.9 ± 11.0 kyr BP.
The only consistent series of U/Th-measurements comes from Foz do Enxarrique C with an WM of 33.8 ± 0.5 kyr BP.Unfortunately the relation between LMP lithic artefacts and the faunal material dated at this site remains unclear (cf.Zilhão 1997).

CONCLUSIONS AND DISCUSSION
Our analyses of the radiometric evidence for the MP/UP transition in SW Europe has augmented a range of criteria for quality control in data mining as well as in terms of advanced numeric data processing.The results emphasize sample material, problems of possible adhering contamination, and on measuring precision.Comparisions of bone and charcoal radiocarbon dates points to extreme discrepancies between these sample categories, with bone dates being systematically too young (Jöris et al. 2001;cf. Fig. 4-5).These findings do not refute the reliability of radiocarbon measurements of bone in general, but call for circumspection in evaluation of quality and reliability of such dates.Against this background, bone dates may contribute less than initially hoped to the understanding of the chronological issues surrounding the transition from the LMP to the EUP in Europe, and in particular we must remain cautious when interpreting 14 C-bone data close to the detection limits of the radiocarbon method (Fig. 3).

On the age of the MP/UP transition in SW Europe
Evaluation of LMP radiometric dates from SW Europe <40,000 14 C BP and <43.5 kyr non-14 C BP (Tab.2; Fig. 11) has shown that minimal reliability -if any -is given for dates younger than 38,300 14 C BP. Based on radiocarbon dates obtained on charcoal, SW European LMP LAD's range between 35,900 14 C BP (Morín, level 10: SI-951-A) and 38,800 14 C BP (Beneito: AA-1387), with a WM of Beneito-, Castillo-(level 18b), Barbas-, Ermitons-, Fuentes de San Cristobal-and Morín-LAD's of 38,391 ± 381 14 C BP (Fig. 11,left).This SW European LMP LAD is statistically identical with the FAD of the archaic Aurignacian at L'Arbreda (WM: 38,307 ± 552 14 C BP) and is in overall agree-(10) Strikingly, at some sites radiocarbon measurements and non-radiocarbon methods have produced identical ages -at least within the errors contained in the standard deviations (e.g.Combe Sauniere, Zafarraya).This is the more surprising, since the calibration records that are available today, imply that radiocarbon age-determinations underestimate the calendric scale for several thousand BP within the time-period under discussion.ment with LMP LAD's available for other parts of Europe (Fig. 6), implying a sudden and simultaneous ending of the MP all over Europe.
To summarize, we note • that the LMP discontinues over the whole of SW Europe at ca. 38,300 14 C BP, • at a time for which similar LMP LAD's and Aurignacian FAD's can be fixed in various parts of Europe (Fig. 6).In terms of calibrated radiocarbon-'years' this transition falls into the shift from the interstadial conditions of GI 11 (GI = Greenland interstadial) to the stadial ones of GS 11 (GS = Greenland stadial) at around 39.9 kyr cal BC (Fig. 8; cf.Tab. 3) (11).
We can therefore confirm the high antiquity of the oldest Aurignacian in the North of the Iberian Pensinsula, i.e.Cantabria and Catalunya, but -in contrast to the radiocarbon record -evidence from the non-radiocarbon methods of dating for the LMP (Tab.2; Fig. 11, bottom) does not withstand criticism and the data available are too weak to prove LMP persistence in the southern half of the Iberian Peninsula until 33,500 14 C BP or later, at least not until the Aurignacian of Gato Preto.
Although the available radiometric dates represent regionally different patterns of hominid presence, we cannot confirm the existance of an "Ebro Frontier", that geographically distinguishes between Aurignacian industries to the NE and LMP ones in the SW against the background of these data.Rather our data implies that the southern part of the Iberian Peninsula was already void of MP hominids long before the appearance of the first Aurignacians.

Climate controlled population dynamics?
Based on radiocarbon evidence, the oldest Aurignacian sites of SW Europe cluster in the North of the Iberian Peninsula (Fig. 12: A; cf.Fig. 9), while the southern French Aurignacian does not date to before 35,400 14 C BP and Aurignacian sites from the southern part of the Iberian Peninsular do not pre-date 33,500 14 C BP (Vega Toscano 1990;Zilhão 1993).This pattern may best be explained by a northward spread (Fig. 12: B) of Aurignacian populations in a severe cold phase (GS 9 = Heinrichevent 4: H4; Fig. 8) shortly after GS 11.In a recent study F. d 'Errico and M. F. Sánchez Goñi (2003) have characterized the highly arid desert-steppelike H4-environments found over large areas of the Iberian Peninsula as inhospitable: conditions that (11) Whereas our FAD's for the Aurignacian of SW Europe preceded those proposed by Zilhão and d'Errico (1999) by some 1800 14 C BP, in calibrated terms FAD-differences would be in the range of only 1.3 kyr cal, still placing the transition into GS 11. may have triggered a northward shift of populations (Fig. 8, 12: B), followed by a further expansion into the NE as well as into the SW at the onset of GI 8 (Fig. 12: C).

Makers of the Aurignacian
Based on the chronometric dates that cover the European MP/UP transition there is neither radio-metric nor stratigraphic indication for incoming populations that may have colonized Europe from East to West (12), nor is there reliable proof for a late persistence of LMP younger than 38,300 14 C BP. Against these data, the geographical pattern of both LMP LAD's and Aurignacian FAD's (i.e. the MP/ UP transition) can be best explained by a contraction of regional population dispersal into glacial refugia during GS 11.Such an interpretation would imply local in situ developments (13) of Aurignacian industries simultaneously in two different areas of Europe, separating the SW European Aurignacian without leaf-points from that of south-central and southeastern European sites which are characterized by the addition of a few of these artefact types.Consequently, this would imply that Neanderthals did indeed produce at least the earliest Aurignacian industries (cf.Churchill and Smith 2000).
Since the hominid fossil evidence is sparse and not entirely unambiguous during the Aurignacian, it would appear that unambiguous evidence of anatomically modern humans is not known prior to the European Middle UP.

Fig. 1 .
Fig. 1.Cultural sequence of Late Middle Palaeolithic to Middle Upper Palaeolithic technocomplexes in SW Europe compared with the hominid record.

Fig. 3 .
Fig. 3. Decay of radiocarbon with time relative to its initial activity in relation to the dating limits of radiocarbon laboratories.The Middle to Upper Palaeolithic transition (MP/UP) is close to the detection limits of most radiocarbon laboratories.

Fig. 4 .
Fig. 4. Age-distributions [kyr 14 C BP] of Middle Palaeolithic (to the left of each diagram) radiocarbon-dates of SW Europe compared with those available for the Upper Palaeolithic (to the right of each diagram) of the same region comprising all 14 C-data (left), those sorted for bone (center) and samples on charcoal (right).In the construction method applied to the resulting 14 C-dispersion graphs, each individual radiocarbon date has been defined by its given median value and standard deviation.The corresponding individual Gaussian curves have been added, to give a curve of the summed 14 C dating probability(Geyh 1969).Because each Gaussian curve is normalised with equal area, using this method each date/sample is given equal weight, independent of dating precision.

Fig. 5 .
Fig. 5. Age-distributions [kyr 14 C BP] of radiocarbon samples on bone (to the left of the left and centre diagrams) compared with those derived from charcoal samples (to the right of the left and centre diagrams) for Middle Palaeolithic and Aurignacian technocomplexes in comparison with age-distributions based on charcoal samples (right) from the Middle Palaeolithic (MP) and Aurignacian (Au.).In the construction method applied to the resulting 14 C-dispersion graphs, each individual radiocarbon date has been defined by its given median value and standard deviation.The corresponding individual Gaussian curves have been added, to give a curve of the summed 14 C dating probability(Geyh 1969).Because each Gaussian curve is normalised with equal area, by this method each date/sample is given equal weight, independent of dating precision.

Fig. 8 .
Fig. 8. Calibrated radiocarbon dates on charcoal of Middle Palaeolithic and Aurignacian of SW Europe (data identical with those in Fig. 5, right) against the GISP2 palaeoclimate record from Fig. 7, showing the abrupt ending of the Latest Middle Palaeolithic (LMP) at the onset of the Aurignacian (Au.) in SW Europe.For regional differences the Aurignacian is distinguished into A, B, C1, C2, C3.

Fig. 11 .
Fig. 11.Radiometric dates ('last appearance data': LAD's) for the SW European Middle Palaeolithic, with radiocarbon measurements (left) sorted for material dated and given in kyr 14 C BP, when required as weighted means, and non-radiometric measurements (right) sorted for method of dating and given in kyr BP, when required as weighted means.
Tab 3. Calibration of seven fictive radiocarbon measurements [BP] spanning around the Middle to Upper Palaeolithic (MP/ UP) transition with typical errors in the range of 2% of total radiocarbon BP.Calibration was established using the CAL-PAL-2003 data set (see annotation 8).FAD -First appearance data.