minkwe wrote:Jochen,
I will wait for a direct response to every point in my previous post before I respond any further to you.
A tempting offer. Anyway, since I've already pointed out a major misconception that invalidates your argument, I suppose it can't hurt to look at the rest of it. First, however, your accusations towards me:
Re Adenier, it is interesting that your only defense is that "it seems" Adenier changed his mind. Lies, Adenier did not change his mind as you can see by reading an even later paper than the one you referred to:
http://arxiv.org/pdf/0705.1477.pdf
My assertion (by quote) was that he seems to now believe that only models that don't possess ideal detection are viable as a local realistic QM completion. The model in that paper does not possess ideal detection, so what I said was completely true. Also, his earlier attempt was not based on nonideal detection; hence, it's fair to infer that he has changed his mind on the issue.
Then your next excuse was that his paper was not published. Lies, Adenier's paper has been published, you didn't look hard enough.
I didn't say it wasn't published, I said it was submitted to J. Math. Phys. (as per the arXiv comment), but appears to have never been published there. The only place I could find it published is in a conference proceedings, which is not surprising since it was presented as a talk there. So again, your accusation of lying has no base in actual fact, and I'd like to see you retract it.
minkwe wrote:Now please address all the points in my argument and state precisely where you claim the argument fails, if you can.
I've shown where the argument fails: from the fact that detections on separate particles are not bound by the value 2, but rather by 4, it does not follow that
; I've even given you an easy to understand example of just why that reasoning fails. But fair enough, in order to further reduce your wiggle room, I'll also address your other points.
Yes I've claimed that, but you can scratch the 'according to you', since it's just according to basic statistics: given the same probability distribution
, then indeed the correlation between two different experimental runs will be the same, because it's just derived from that distribution.
This is just what the expectation value means:
for large number of measurements n, so in order to yield the same EV, there need to be approximately the same numbers
of +1 outcomes and
of -1 outcomes. However, equality of these numbers only holds for
, and for any two finite experiment runs i and j, both
and
, as well as
and
, will generally differ. Hence, in (almost) all real world experiments, such a function won't exist.
Again, only for the
case, which is never attained in experiments.
There's two problems here. The first one is the erroneous nature of the argument in point 1, which I've already noted. The second is that even if the reordering functions did exist (which they won't in general), then you can't perform such reorderings and expect everything to stay the same: the correlations are generally not preserved under such a reordering. Consider two fair coins: if they are uncorrelated, you will get two columns of (approximately) as many Hs and Ts in each case. Now, you can reorder the columns: putting all the Hs in the same row, as well as all the Ts, for instance: but then, it will look as if the coins are perfectly correlated. Alternatively, you could put all the Hs of one in the same row with all the Ts of the other (since they're fair coins, there will be equally as many of each), and vice versa, and hence, obtain perfect anticorrelation.
The only rearrangements you can validly perform on the measurement tables are row permutations; all other rearrangements will alter the correlations. This corresponds to re-labelling the experiment runs, and will not change anything about the expectation values. A Bell test experiment yields a meassurement table like the following:
- Code: Select all
Alice | Bob
A | C | B | D
+1| | | -1
-1| | -1|
| -1| | +1
.
.
.
| -1| -1|
+1| | +1|
+1| | | +1
Whenever you rearrange a single column, all other columns have to be rearranged in the same way; else, you change the correlations. So you can't, for example, just rearrange the rows for C and D, since D also has correlations with A, and C with B; thus, you can only permute the rows. But this doesn't change anything. Is this clearer now?