The Provence area in the 3rd millennium BCE can be characterized by a number of materials studies that highlight its archaeological complexity and a significant variability in the range of landscape zones settled. The work done by J. Cauliez (2010, 2011) provided a more structured chrono-cultural framework and a more diversified background to the ceramic components of the Late Neolithic in Provence. Therefore, this study utilizes the quadripartite chronological sub-division Cauliez defined (horizon 1: 3400-2900/2850 BCE; horizon 2: 2900/2850-2600/2550 BCE; horizon 3: 2600/2550-2400/2350 BCE; and horizon 4: 2400/2350-1950 BCE). One of the problems that this chronological division does not resolve is the crucial issue of the Bell Beaker event which overlaps with the various local traditions during horizons 3-4. Here, the sites with Bell Beaker pottery and associated materials will not be analysed separately. The site corpus includes, on one hand, 260 strictly Late Neolithic sites, and on the other, settlements belonging to the four Late Neolithic horizons (horizon 1: 8 sites; horizon 2: 19 sites; horizon 3: 29 sites; horizon 4: 7 sites). General assessment of these domestic sites, and the establishment of associated deposit typologies has been observed previously (Caraglio, 2015, 2016b), and will not be addressed here. However, this paper presents the final latest analysis of these sites, which involved an in-depth study of settlement patterns in the Provencal landscape to better comprehend the emerging social and cultural mechanisms operating at the dawn of the Bronze Age. New spatial data was generated from the georeferenced sites in the corpus through the creation of a relational database integrated with GIS (Geographic Information System). Through a combination of statistical exploratory and multivariate analysis of the spatial data, precise characterization of the topographic location of each site in Provence has been possible, as well as the identification of significant trends regarding topographical settlement preferences during the 3rd millennium BCE. Three Multiple Correspondence Analyses have been conducted on the archaeological corpus, with a fifth analysis being undertaken as a control, on sites randomly generated through GIS, using five location variables (altitude, slope degree, relief variation, topography and slope orientation). For the Multiple Correspondence Analysis, geocultural landscape units, geology and chrono-cultural phases have been included as comparative variables, but as they cannot be interpreted reliably, they have not been incorporated into the final analyses in order to reduce potential distortion of the results. Although we cannot propose narrowly defined location typologies, the results show the great variability in settlement location choice in Provence during the second half of the 3rd millennium BCE. Thus, in the Late Neolithic, populations seem to favour sites on plains or hillsides, around 150 m and up to 300 m in altitude, on flat or medium slopes, with average topographic variations and orientations to the east or south. Despite the relatively small number of sites known, the populations of horizons 1 and 4 (respectively 8 and 7 sites) seem to favour very diversified settlement areas, while those of horizons 2 and 3 (respectively 19 and 29 sites) would gather in very low areas, below 150 m in height. The deposits of horizon 2 were found mainly in the Vaucluse plains or the slightly sloping areas of Lower Provence quaternary. From horizon 2 onwards, exposure to the east tends to become more widespread. Locations during horizon 3 appear to show a dichotomy or complementarity between very conspicuous areas (steep slopes, significant topographical variations, minor prominences) and more geographically indistinct areas (weak or flat slopes, weak or non-existent topographical variations, plains), between Lower Provence and the Provencal coastal Plains, and on Marine Quaternary, Jurassic-Cretaceous or Miocene-Pliocene soils. Despite some limitations, both the spatial analyses presented here and a precedent case study in the Luberon area (Caraglio, 2016a) demonstrate both slight oscillations in settlement choices and permanence of land management at the end of the Neolithic. The global trends could be the result of specific voluntary and cultural choices, but the great variability of the Provencal landscapes should not be underestimated. We therefore believe that they are not due to chance, but indicate changes in settlement choices at the end of the Provencal Neolithic and that the spatial variables studied had a structuring role in the distribution of habitation. The results tend to illustrate the occupation of all ecosystems. This raises the question as to whether these different groups were forced to settle everywhere space was available due to the most ecologically favourable areas being densely settled already? Generally speaking, it is between the 29th and 24th centuries BCE, and most markedly during horizon 3 parallel to the development of the Bell Beaker, that the number of known sites seems to increase significantly. As such, it would be unsurprising if this illustrates the conquest of all the available "economic niches" (Higgs, 1975) by these Late Neolithic groups. These results would tend to justify the notion of a "full world" (Farina et al., 2003), perceptible through a denser population, greater stability, an evolution of agricultural land management and an interweaving of the socio-economic and symbolic spheres at the beginning of the Bronze Age.
