It is worth taking a deeper look at Fig 4 of the paper by Malewar et al. mentioned above. Don't worry about the two graphs (one of Requip XL and one of a different formulation of ropinirole CR), what follows applies to both.

On an initial dose the graph shows:
Time (hr)
0-5, plasma levels rising linearly from 0
5-15, plateau
15-32, plasma levels falling linearly to 0

Roughly 1/6th of the area under the curve is in the second day. The consequence of this is that, assuming doses are taken every 24hr, levels will build slightly over a few days. Roughly speaking, we can see this effect by cutting out the area beyond 24 hours and adding it to the initial section of the graph. So, at the start of subsequent days, when we take a pill at time 0 (which contributes nothing at time 0) we get all our benefit from the previous day's pill, giving a total approximately half that of the plateau. (It's a pity that for those taking levodopa as well, this coincides with the time levodopa levels are at their lowest.) So, if one pill is taken per day, on subsequent days there is a swing of about 50% from highest to lowest concentrations.

We can reduce variability by halving the dose, but taking it twice as often. (This should not be done by cutting a pill in half.)

John