Duplicated sequences are one of the main sources of variation in eukaryotic genomes and are known to give rise to new genes and functions. Large (1-200 Kb) and highly identical (≥90%) duplications named segmental duplications (SDs) had a particularly important role in the evolution of African great apes (including humans). SDs compose around 5% of the human genome and are shared with the other African great apes more than what one would expect given their nucleotide divergence. In other words, the rate at which SDs appeared was specially high in the African great ape ancestor, after the split from orangutans. SDs that arose during that period are strong candidates to account for part of the phenotypic differences between these species that point-mutations cannot explain. Here, we use inferences of copy-number along great ape genomes to classify human SDs according to the period of time in which they appeared. We identify, first, human specific SDs, second, human SDs that appeared during the burst of duplications and, finally, older human SDs that are shared with all great apes, including orangutans. We explore the characteristics of these three groups of duplications trying to understand both the causes of the increase in duplication rate during the time of the African great ape ancestor and its phenotypic consequences. We also differentiate between tandem, non-tandem intrachromosomal and interchromosomal SDs. We find differences in length, gene content and Alu content between these groups. These differences point towards different duplication mechanisms of the SDs in these three types of duplications. Moreover, we use sequence similarity inside and outside shared exons in duplications to identify candidate signals of selection in human SDs.