The first expansions occurred to the west  and the east  during the Miocene-Pliocene

A third assemblage corresponds to cultivated Cannabis, which is usually found together with cultivated cereals such as Avena , Hordeum , Secale  and Triticum , as well weeds such as Centaurea  or Scleranthus  species. Using these phytosociological affinities, McPartland et al.  attributed the Cannabis-like pollen types recorded in the literature to wild Cannabis if this pollen occurred together with steppe assemblages; to cultivated Cannabis when it was part of crop assemblages; or to Humulus if the dominant pollen assemblage corresponded to temperate deciduous forests. Based on these premises, these authors developed a more complicated identification algorithm that also considers the relationship between arboreal and non-arboreal  pollen . Using these criteria, the oldest known pollen compatible with Cannabis was found in 19.6 Ma-old rocks from the NE Tibetan Plateau , vertical grow system which was proposed as the center of origin of Cannabis .

Interestingly, this proposal roughly coincides with the former hypothesis based on indirect biogeographic evidence. The use of DNA molecular phylogenies calibrated with fossils of related genera such as Humulus, Celtis, Morus and Ficus allowed to estimate the age of divergence of Cannabis and Humulus to 27.8 Ma . Using the same DNA phylogeny and the associated molecular clock, the divergence between C. indica  and C. sativa  would have occurred in the Middle Pleistocene approximately 1 Ma . These authors noted that there is a gap of ca. 8 million years between the age of origin estimated by the molecular clock and the first fossil pollen encountered. In spite of this, they favored the mid-Oligocene age for the origin of Cannabis on the NE Tibetan Plateau , assuming that, as demonstrated by the presence of Artemisia and other steppe elements, the region was covered by this type of vegetation, which would have been particularly well suited for the development of Cannabis. A previous study using Bayesian calibration estimated the divergence between Cannabis and Humulus to have occurred 21 Ma , which is closer to the age  of the first fossil pollen evidence mentioned above . According to pollen and seed fossil records, mobile grow systems Cannabis would have experienced some expansion from its center of origin to Europe and East Asia well before the evolutionary appearance of the genus Homo .

The suggested dispersal agents are water  and animals. Most parts of the Asian continent were colonized by Cannabis during the Pleistocene , before the onset of the Neolithic, when humans domesticated the first plants. During the Pleistocene, glacial-interglacial recurrence could have contributed to Cannabis diversification without human intervention. It has been suggested that Cannabis underwent recurrent range contractions  and expansions  that facilitated allopatric processes, possibly leading to the differentiation between the European  and Asian  subspecies, which would have diverged nearly 1 Ma . The first is considered the putative hemp ancestor , and the second is the putative drug ancestor. In addition to fossils and DNA phylogenies, archeological evidence is of paramount importance to reconstruct domestication and anthropogenic diffusion trends within Cannabis. The main types of evidence of Cannabis from archeological sites are pollen, seeds, fibers, fiber/seed impressions, carbonized remains, phytoliths and chemical remains. During the historical period, written and graphic documents are also of fundamental help . Pollen identification has some additional clues related to its abundance in sediments, as Cannabis produces much more pollen than Humulus, which is usually underrepresented .

This is especially useful in cases of very high percentages of this pollen type, which are difficult to explain unless the sediments come from a former hemp-retting site. Indeed, when flowering male hemp plants are soaked in a retting pond to separate the fibers from the stalk, large quantities of pollen settle into pond sediments. In these situations, the percentages of Cannabis pollen in sediments may reach 80–90% of the total, but percentages over 15% or 25% have been considered sufficient to infer hemp retting . The anemophylous pollen of Cannabis may be transported long distances; therefore, the finding of a few grains or their scattered occurrence throughout a stratigraphic section is not necessarily evidence for the local presence of the parent plant. Recent developments in molecular DNA analytical methods have increased the probability of identifying Cannabis sativa in lake sediments and comparing these results with the abundance of Cannabis/Humulus pollen. For example, in lake sediments from the French Alps, Giguet-Covex et al.  found a good agreement between DNA and pollen records for the period 1500–1000 yr BP but significant inconsistencies for the periods 2000–1500 yr BP and 500 yr BP-present.