Perfectly right that standard deviations are not the same es errors! Compared to the average effect size, though, in this case they clearly show there's a very high variation in the observed results.
The 4h is really unusually high compared to real-life reading in bed. But compared to the other studies its still quite low! One of the other studies had a whopping 6.5h hours of "bedtime reading".
Right, but the point of the crossover design and mixed-effects model is that some of that variability is attributable to individual differences: you always fall asleep quickly; I toss and turn for a bit.
Imagine you fall asleep in 1 minute normally, but take 11 minutes after light exposure. I take 21 minutes, but 31 after light exposure, and a third person goes from 11 to 21. The standard deviation of each condition is the same size as the effect (10 minutes), but there’s still an effect.
The 4h is really unusually high compared to real-life reading in bed. But compared to the other studies its still quite low! One of the other studies had a whopping 6.5h hours of "bedtime reading".