Bell Tests and Biased Samples

["Edward Green" <spamspamspam3@xxxxxxxxxxx>]

I'm trying to understand the material at the web page of subject name:

http://www.mathpages.com/home/kmath420.htm

Some warm up questions ---

The author mentions a model where paired particles have "a definite
spin (or polarization) axis".  To fix our thinking, it seems to me we
must fix on one or the other -- we must fix on whether we are
considering a measurement returning detections along two orthogonal
axes (linear polarization), or an up/down value along one axis
(spin-1/2).

The author fixes upon spin-1/2 particles, consistent with a graph
showing "correlations" over a range 0,pi.  I am having difficulty
picturing an experimental set-up, however.  A Stern-Gerlach apparatus,
followed by a particle detector, measures "up" or "down". Doesn't this
mean that detectors relatively oriented at "0" and at "pi" are the same
up to a sign change in the results?  And shouldn't the correlation at
"0" be -1?

I don't understand the operational definition of the variables
measured.  Can somebody spell it out for me?