My take is that this is about parallax [1], although I haven't read
the paper and so that's an assumption.
Creatures with forward facing stereoscopic vision (like us) can judge
the speed and distance of an object if we can determine its size and
see something of the background. In the dark, some information is
missing when we rely on attached lights.
Most people know the approximate size of a car, and so with two lights
its easier to guess the distance (and crucially speed as the distance
between the lights seems to change).
I've noticed it's harder to cross a road when bikes are approaching
than cars, because if you're looking along the axis of travel then its
hard to gauge their speed.
Attaching more than one light to a bike allows us to judge the speed
better. It seems that it's irrelevant whether the spacing is
horizontal or vertical for our visual systems to make this judgement.
[1] https://en.wikipedia.org/wiki/Parallax