Subjects who got six hours of sleep a night for two weeks straight functioned as poorly as those who were forced to stay awake for two days straight.

this is literally not what the study finds. chatgpt generated article. For digit-symbol-substitution, by the end of the study (14 days) they find taht the 6 hour group, when corrected at the rate of learning of the 8 hour group, performed as bad as the ONE night sleep deprived group. Note that this is entangling sleep's effect on learning with its direct effect on cognitive performance. Also. for the serial-addition-substitution, after 14 days it was the 4 hour of sleep group that was as bad as the 1 night deprived group. maybe they only looked at the reflex-response task, where the 6 hour group barly scrapes into the 2-day-deprived group's margin of error.

Here is the article to see for yourself: https://pomf2.lain.la/f/a09sl5sp.pdf. Note that when they say "linear" they mean "not saturating" as opposed to a strictly linear effect.

ALSO ALSO with respect to the subjective sleepiness thing, using the power model and then saying that its due to nonlinearity of the observation variable, but then looking at the subjective sleepiness score which is visually VERY SIMILAR and saying "yep thats linear for the 0hr group and saturating for 4hr,6hr" doesn't pass muster.