This post is my reply to a blog comment by billschnieder, since I do not think it is wise to start a discussion in the blog's comments directly, nor in private messages, because of both the reduced visibility of the discussion and the increased spam-like characteristics of such communication for the blog's poster or bill's pm inbox respectively. https://www.physicsforums.com/blog.php?bt=6779 [Broken] In summary, I do not think hidden polarizer variables have any bearing on the point that the blog post or Herbert's original explanation are trying to make. Depending on the effect they have, they can either be dismissed completely, be considered as something to go along with the polarizer settings and still keep the point of the blog post or Herbert's original explanation valid, or they can make the whole experiment pointless but in a way that will be obvious and detectable. More specifically, hidden polarizer variables could affect the QM predicted 100% coincidence rate with matching polarizer settings in two ways: result in non-detections at one or both of the polarizers, or result in opposite detections. The first way would contribute to "detector inefficiency" and the second to "experimental error" margins. The important thing here is that you can measure those effects, know their average magnitude, and know if it is enough to cause violations to the inequality as high as the QM prediction, making your experiment pointless. There is no case where these hidden variables could conspire to mask themselves as insignificant for matching polarizer settings and then cause a significant deviation for non-matching polarizer settings, because they and their effect should be local to each polarizer. If the effect is small enough, i.e. you get close to 100% coincidence rate with matching polarizer settings, you can just ignore these "polarizer hidden variables", or include their worst case in your calculations. For example, a 5% effect could increase the (0 deg;30 deg) and (-30deg; 0 deg) mismatch from 25% to 30% and could decrease the (-30 deg; 30 deg) mismatch from 75% to 70%... but that still is inconsistent with local determinism. A 10% margin of error on the other hand would make this experiment pointless. Since the idea of examples like this is that eventually, experimental setups will get good enough to get such error margins down to 0, I think their point is valid and unaffected by your objections. If you have a reason to believe that the error margin can not possibly be reduced sufficiently, that is another subject entirely. I guess it may indeed be the case that QM itself leads to some minimal percentage of non-detections, and angle-dependent non-detections are indeed one of the ways (the only way I personally know, actually) to build a local deterministic model matching QM predictions after non-detections get filtered out of the data.