We present a thorough literature study of the most-massive star, mmax, in several young star clusters in order to assess whether or not star clusters are populated from the stellar initial mass function (IMF) by random sampling over the mass range 0.01 <= m <= 150Msolar without being constrained by the cluster mass, Mecl. The data reveal a partition of the sample into lowest mass objects (Mecl <= 102Msolar), moderate mass clusters (102Msolar < Mecl <= 103Msolar) and rich clusters above 103Msolar. Additionally, there is a plateau of a constant maximal star mass (mmax ~ 25Msolar) for clusters with masses between 103Msolar and 4 × 103Msolar. Statistical tests of this data set reveal that the hypothesis of random sampling from the IMF between 0.01 and 150Msolar is highly unlikely for star clusters more massive than 102Msolar with a probability of p ~ 2 × 10-7 for the objects with Mecl between 102 and 103Msolar and p ~ 3 × 10-9 for the more massive star clusters. Also, the spread of mmax values at a given Mecl is smaller than expected from random sampling. We suggest that the basic physical process able to explain this dependence of stellar inventory of a star cluster on its mass may be the interplay between stellar feedback and the binding energy of the cluster-forming molecular cloud core. Given these results, it would follow that an integrated galactic IMF (IGIMF) sampled from such clusters would automatically be steeper in comparison to the IMF within individual star clusters.