If you ask 160 meter operators with many years of experience, a modest speed of about 15 WPM is best for operator copy. Much faster is difficult to copy in noise, much slower is more difficult also.
Going much slower than 15 WPM is worse also, because the rhythm is very difficult to sort.
This is for experienced humans, not machines.
As a matter of fact I constructed with a program written in a general purpose programmimg language white noise (random numbers) and a sign wave 600 Hz, for different speeds starting at 5 wpm and signal to noise ratios, for me it turned out that with plain Dutch language random chosen words 6 characters long, the optimum for head copy was flat between 5 and 10 wpm. Just the correct copied words were maximum in number for the lowest S/(S+N) ratio in that speed range. 15 wpm was less and higher, deteriorates fast (needs less noise to copy reliable)
Some results (wav files) can still be found on my website http://pa0wv.home.xs4all.nl/zelfbouw.html
Of course it requires without noise more concentration to copy 5 wpm then 15 wpm without, but still that was my result.
I remember DJ1YFK did also this kind of tests, as I read some time ago on his website.