The first option is loosely based on the trichromatic theory of three light receptors in our retina, which are sensitive to the spectrum as a whole, and whose peak sensitivities lie where we recognize, but not see these colours. They do not belong to light or nature, but to our post processing of these signals, which means that fewer light receptors will pass less discerning information.
The second option was an attempt at figuring out which paints reflected wavelengths that we recognized as the whole spectrum by means of intermixture. Yellow and blue do not give away green because the latter belongs to them, specially when their optical combination is achromatic (see dot paintings), but it's due to a slight deviation in reflectance from overlapping layers.
The third option is based on the ideal, but not yet accomplished, three colours that give away the additive ones (cyan and magenta give blue, magenta yellow gives red, etc) according to their model, with the addition of a key element to tone them down. Since what printers use belongs to a colour space, cyan, magenta and yellow are processed to measure up to their idealized selves.
The psychological colours are those whose names we have learned to identify and make combinations with. Magenta and cyan fall from this group, because they're either blueish red or greenish blue, respectively. They were inspired by the second stage of retina processing by grouping the "red", "green" and "blue" light receptors' stimulus in opponent cells, pitting red against green and their positive signal (i.e. yellow) against blue.
The last option is for when you consider rose and aqua, for example, between red and magenta, and between green and cyan, respectively, to make a more appropriate opponent axis to cross with the yellow and blue one. They're perpendicular in some colour spaces which account for human perception such as CIELAB and the Natural Colour System. Please check these two out for an example of how RGB and CMYK relate to them.