[R-sig-ME] Modeling polar coordinates

[Titus von der Malsburg]

I'd like to fit mixed models for polar coordinates.  Each point
represents the outcome of a trial in an experiment.  Items and
subjects will be crossed random effects.  Fitting a model for the
radius component is business as usual but I don't know how to approach
the azimuth component.  The problem is that 2pi and 0 specify the same
point on a circle and that the difference between 2pi-0.1 and 0.1 is
0.2 rather than ~6.08.  Is there a way to deal with that in lme4?

R-seek didn't yield anything useful, maybe I used bad search terms.
Thanks in advance for any advice!

Best,
  Titus