A quantitative comparison of higher-order chromatin arrangements was performed in human cell types with three-dimensionally (3D) preserved, differently shaped nuclei. These cell types included flat-ellipsoid nuclei of diploid amniotic fluid cells and fibroblasts and spherical nuclei of B and T lymphocytes from peripheral human blood. Fluorescence in-situ hybridization (FISH) was performed with chromosome paint probes for large (#1–5) and small (#17–20) autosomes, and for the two sex chromosomes. Other probes delineated heterochromatin blocks of numerous larger and smaller human chromosomes. Shape differences correlated with distinct differences in higher order chromatin arrangements: in the spherically shaped lymphocyte nuclei we noted the preferential positioning of the small, gene dense #17, 19 and 20 chromosome territories (CTs) in the 3D nuclear interior – typically without any apparent connection to the nuclear envelope. In contrast, CTs of the gene-poor small chromosomes #18 and Y were apparently attached at the nuclear envelope. CTs of large chromosomes were also preferentially located towards the nuclear periphery. In the ellipsoid nuclei of amniotic fluid cells and fibroblasts, all tested CTs showed attachments to the upper and/or lower part of the nuclear envelope: CTs of small chromosomes, including #18 and Y, were located towards the centre of the nuclear projection (CNP), while the large chromosomes were positioned towards the 2D nuclear rim. In contrast to these highly reproducible radial arrangements, 2D distances measured between heterochromatin blocks of homologous and heterologous CTs were strikingly variable. These results as well as CT painting let us conclude that nuclear functions in the studied cell types may not require reproducible side-by-side arrangements of specific homologous or non-homologous CTs. 3D-modelling of statistical arrangements of 46 human CTs in spherical nuclei was performed under the assumption of a linear correlation between DNA content of each chromosome and its CT volume. In a set of modelled nuclei, we noted the preferential localization of smaller CTs towards the 3D periphery and of larger CTs towards the 3D centre. This distribution is in clear contrast to the experimentally observed distribution in lymphocyte nuclei. We conclude that presently unknown factors (other than topological constraints) may play a decisive role to enforce the different radial arrangements of large and small CTs observed in ellipsoid and spherical human cell nuclei.