( 2 )
Whereas previously the force varied over a continuous range with Q, it
takes now just two values, +F, the sign bqiqg determined by whether the
magnetic axis of the particle points more nearly in the direction of
the field or in the opposite direction. No attempt is made to explain
this change in the force law. It is just an ad hoc attempt to account
for the observations. And of course it accounts immediately for the ap
pearance of just two groups of particles, deflected either in the direc
tion of the magnetic field or in the opposite direction. To account
then for the
EinsteinPodolskyRosenBohm
correlations we have only to
assume that the two particles emitted by the source have oppositely di
rected magnetic axes. Then if the magnetic axis of one particle is more
nearly along (than against) one SternGerlach field) the magnetic axes
of the other particle will be more nearly against (than along) a paral
lel SternGerlach field. So when one particle is deflected up, the
other is deflected down, and vice versa. There is nothing whatever pro
blematic or mindboggling about these correlations, with parallel Stern
Gerlach analyzers, from the Einsteinian point of view.
So far so good. But now go a little further than before, and con
sider mn4arallel SternGerlach magnets. Let the first be rotated away
from some standard position, about the particle line of flight, by an
angle a. Let the second be rotated likewise by an angle b. Then if fhe
magnetic axis of either particle separately is randomly oriented, but
if the axes of the particles of a given pair are always oppositely
oriented, a short calculation gives for the probabilities of the various
possible results, in the ad hoc model,
where "up" and "down" are defined with respect to the magnetic fields
of the two magnets. However, a quantum mechanical calculation gives
1
P
(up,up)
=
P
(down,down)
=
3
(sin
e)
'
2
1
1
ab
P(up,down)
=
P(down,up)
=
3

7
(sin
T)
Thus the ad hoc model does what is required of it (i.e., reproduces
quantum mechanical results) only at (a

b)
=
0, (a

b)
=
s/2 and
(a

b)
=
n,
but not at intermediate angles.
Of course this trivial model was just the first one we thought of,
and it worked up to a point. Could we not be a little more clever, and
devise a model which reproduces the quantum formulae completely
?
No.
It cannot be done, so long as action at a distance is excluded. This
point was realized only subsequently. Neither EPR nor their contempora
ry opponents were aware of it. Indeed the discussion was for long enti
rely concentrated on the points la

bl
=
0, ~/2,
and
n.
3.
Difficulty with locality. To explain this denouement without mathe
matics
I
cannot do better than follow dVEspagnat /14, 15/. Let us re
turn to socks for a moment. One of the most important questions about
a sock is "will it wash
"?