What are the characteristics of Labelese?

[Koen G]

hmmm...and what if you would take random Voynich words from the text, the same amount as the labels, and
of course with a size that has a similar low and high as the labels (I suspect this lies between 5...12 from my head),
and do the same calculations?

I was thinking about this too, but how would we interpret the results? Would we expect proper labels' statistics to be similar to, or rather different from that of randomly selected words? This would involve first comparing the statistics of the randomly selected sample to that of full text, to see to what extent the randomly selected words retain some of the full text tendencies. (I'm thinking that if you select 1000 words from the VM, some common VM words will be relatively common in the random sample as well).