minkwe wrote:local wrote:FrediFizzx wrote: 3 angles. How does that work?.

Simple. The source is the same, i.e., a singlet pair with the photons propagating in different directions to the detection stations. One station has its angle fixed, while the other station chooses between two angles. Concerns about different ensembles remain (weak versus strong objectivity).

Thank you for the interesting link, minkwe. Note that existing data sets for CHSH can be re-analysed for OB violation. It might be fun to try that.

I doubt that the lack of perfect anticorrelation is the key, because OB can be derived in other ways omitting that assumption.

You are welcome. Perhaps I should ask the question another way. Is any of you aware of any experiment which measured the particle pairs at the same setting on both sides and characterized perfect anti-correlation? This would seem like an excellent/obvious experiment to do as part of the myriad of experiments being done. I've been looking for such data and haven't found it after a long time of searching. His inequality leaves enough room - enough distance between LR and QM. As long as QM is close to giving the right predictions (in the sense that the singlet correlations are close to true), you can use his methodology to reject local realism.

So I reduced my expectations and started looking for raw Malus law data for entangled particles. No luck there either. The Malus experiment is done in almost every undergraduate physics class, yet nobody ever reports the comparison of the intensity without polarizer to the maximum intensity with polarizer.

I believe that such experiments have been done, but without a big spatial separation between the two measurement stations. Indeed, one can use the data from experiments in which the whole negative cosine curve is measured. The important thing is that it is observed to be close to full amplitude.

In some experiments this has been seen after using the “fair sampling assumption” to dispose of the detection loophole.

Already, at the end of his famous paper, Bell mentioned that one could never expect to observe *perfect* anti-correlation because of inevitable experimental errors. He showed how to modify his inequality to make it experimentally useful. You must show (up to some acceptable statistical uncertainty) that the negative correlation at equal settings is stronger than -1 + epsilon, where epsilon is a small positive number, you may choose it yourself; and you must show that his inequality holds (again, up to some acceptable statistical error), but sharpened by an amount delta; he shows how to compute delta from epsilon. The smaller is epsilon, the smaller delta will be. His original inequality is, in a good sense, robust to imperfect measurement.