[EM] Range ballots vs range voting

[Ken Kuhlman]

In the statement that Rob LeGrand posted from the CAV/AAV board, I
find the following statement from the footnote particularly
interesting:

> Although Range Voting provides a generalization of Approval Voting,
> it is not obvious how many levels of approval voters should be
> allowed to indicate: 100, 10, or some other number?  Even 10 levels
> is asking a lot of voters, who--psychological experiments show--
> generally cannot make such fine discriminations.

Do any good references exist for these psychological experiments?  
The basic proposition made here is reasonable, but I wonder if a upper
bound on the practical range of a range ballot is known.

Also, has the idea of using a range ballot but compressing the ballots
to remove empty slots been discussed before on this list?   The net
effect would be to produce a "traditional" ranked ballot.  This would
also seem to nullify the argument being made in another thread about
the supposed tactical superiority of range voting due to current
ballot machines being able to handle the rating/range format.

-Ken