This pages discusses various ways physicists and others have tried to explain violations of Bell’s inequalities. For an introduction to Bell’s Theorem, see first bells-inequalities
Bell’s theorem has the structure of a mathematical theorem. It gives premises (assumptions) and from those premises it deduces specific inequalities that can not be violated. Since Bell Inequalities are violated in the lab, we know that one or more of Bell’s premises must be violated “by the universe”. John Bell himself set out with the goal to isolate locality and make arguments about locality — i.e. it would be his goal that any local theory should follow his inequalities and hence a violation of his inequalities would be a violation of locality.
Locality
My view on the implication of Bell’s Theorem is that to explain Bell inequality violations, we must give up locality as it exists in Bell’s sense, which I also think is as it exists in the relativistic sense (admittedly, this is contentious). Yet, this viewpoint might be the hardest one for many physicists to swallow. That is because, alongside Quantum Mechanics, General (and therefore Special) Relativity forms a core pillar in standard physics today. General relativity is not only extremely well experimentally tested, having never failed a lab experiment yet, it also provides our conception of the universe’s causal structure. In the theory of relativity, space-like separated events are entirely causally disconnected. For two events that are space-like separated, there is no invariant fact of the matter which one occurred before the other and thereby (barring retro-causality) no way one could have caused the other. In this view, asking something like “did Alice conduct her experiment before Bob or vice versa?” when the two experiments are conducted at space-like separation is a category error. One needs a preferred foliation of spacetime to even ask that question. A different way to state the above is that in different inertial reference frames, spacelike separated events may switch time ordering (this is the way that might be more commonly presented).
The Standard Model hinges (through incorporation via Quantum Field Theory) on our best attempt at reconciling Quantum Mechanics and Relativity. The reconciliation was partial and limited to microcausality.
Probably the interpretation (or theory) most known for violating locality is the Bohmian one. Ghirardi-Rimini-Weber (GRW) type objective collapse theories are also rather explicitly non-local.
Albert Einstein
“I cannot seriously believe in [quantum mechanics] because the theory cannot be reconciled with the idea that physics should represent a reality in time and space, free from spooky action at a distance”
Musing
In QFT (the basis for the Standard Model), we posit a commutation or anti-commutation structure on “operator-valued distributions” which vanishes at space-like separation when we are performing “second quantization” of some bosonic or fermionic field respectively. This “microcausality” requirement recovers the no-signaling theorem (no signals can travel faster than the speed of light), but it does not make the theory locally causal under Bell’s definition. So if locality is such a foundation of physics, why am I ready to give up on it? Simply speaking, Bell’s arguments are persuasive. His definition of locality fits my understanding of relativity’s definition of locality as not just signal locality but a fundamental causal structure of spacetime. It would have been intellectually nice if Einstein’s program could have succeeded and produced an ontic, local theory of the world. But it seems to me that violations of Bell inequalities in the lab rules that out.
Other Denials
Instrumentalism and Copenhagen
Note: I have also heard “operationalism” in place of instrumentalism. I think of them as almost exact synonyms. In colloquial speech, one might call the pure instrumentalist school of thought, the “shut up and calculate” school.
Info
Although oft attributed to Feynman, the quote “shut up and calculate” actually comes from N. David Mermin.
Many working physicists simply don’t engage with Bell’s Theorem. The theorem is not necessary to calculate scattering amplitudes, plasma temperatures, galactic rotation curves, or lattice spacings for example. Even in most of the physics which uses entangled states regularly, you can always just “do the calculations” as quantum mechanics tells you to do them without explicit worry about Bell’s theorem. We know quantum mechanical predictions are accurate. QFT, a subfield of QM, is possibly the most well tested physical theory of all time (I don’t want to get into an argument about whether GR is more well tested or not).
So it makes sense that, unless you are specifically working in the foundations of QM or adjacent fields, you might not need to engage much with Bell’s theorem on a day-to-day basis. So, for practical purposes, a lot of physicists are instrumentalists who apply the formalism of QM to get the results they need.
I think the Copenhagen “interpretation” (or spirit, or school, or what have you), though, goes a little bit further than pure instrumentalism although I wouldn’t begrudge anyone for using the terms interchangeably. In particular, there is a strict need to maintain both completeness of quantum mechanics and locality. EPR’s argument was rejected outright by Bohr (founder of the Copenhagen school) in his response. There is also a strong focus on complementarity—the somewhat vague idea that somehow physical reality is inextricable from the experimental set up, the discussions thereof, and the measurements made. It doesn’t really make sense to talk about the momentum of a particle or the position of a particle without also specifying the experimental procedure you use to measure (or manifest?) those quantities. Measurements must be described classically and in macroscopic terms, but the microscopic is described by the equations of quantum mechanics. This divide between macroscopic and microscopic is sometimes called “Heisenberg’s cut”, but exactly where that cut goes is never specified.
For an instrumentalist and/or a follower of Copenhagen, Bell’s argument against locality appears to be inapplicable. There can not be a posit of “what is there” independent of measurement apparatus so no well defined independent notion of “local causality”, beyond signal locality, should be expected. The no signaling theorem says Alice and Bob may not use entanglement to send faster-than-light signals to each other and from an instrumentalist perspective, there is no more story to tell.
Of course, many (myself included) feel physics should strive to explain more than instrumentalism or the Copenhagen doctrine. It would be nice if physics could actually tell us one level deeper of “what is there?“. But such further discussion might be more philosophical in nature.
Albert Einstein
I like to think the moon is there even if I am not looking at it.
Counterfactual Definiteness and Realism
At various times, and even now, it is often stated that what Bell’s theorem rules out is something called “counterfactual definiteness” or “local realism”. Both of these terms, but especially “local realism”, are used somewhat vaguely so it becomes a bit hard to have a precise discussion of them. They both point in the general direction of “properties exist without measurement”.
Counterfactual definiteness is the notion that “if I had measured X, I would have obtained result y”. Such a statement is a counterfactual if in reality I didn’t measure X (maybe I measured Z), and some would say that Bell Theorem assumes that in that counterfactual claim, y is definite. In this sense, I see rejections of counterfactual definiteness as an attack on the outcome determinism that existed in Bell’s 1964 proof of his Theorem. As discussed extensively in bells-inequalities, outcome determinism was a consequence, not an assumption, in that proof. Furthermore, later Bell papers which relaxed the perfect anti-correlation constraint also removed the need to have outcome determinism show up anywhere in the argument. There may be other senses in which “counterfactual definiteness” could be used. I will not try to give a complete overview here.
Local realism seems to be particularly shifty but popular in usage. In its original usage from Clauser and Shimony, it appears to be a synonym for the factorizability condition1 (what we called Bell locality). But it seems Mermin used used these words in the sense of outcome determinism1, and we have already discussed that particular point to death. Other uses mostly seem to me to be “whatever Bell disproved, he disproved local realism”.
Musing
The theme of this section seems to be—if you become radically famous for one breakout paper, people will latch on to every aspect of that paper. Even aspects, like outcome determinism, which weren’t essential to the core argument but maybe you inherited from a previous analysis. I’m being cheeky here of course.
Statistical Independence
In derivation of Bell inequalities, we always had to use the condition of statistical independence: . Even Bell acknowledged that parameters must be free to get his theorems to go through. But one must acknowledge that this sense of the word free is very weak. All it requires is some way to randomize your parameter choices. For example, you could have agents with free will decide on the parameter settings. If one does not ascribe to free will, then you could have a random number generator pick parameter settings, or find the quadrillionth digit of pi in binary and use that as a code, or use random photon incidence rates from the CMBR from a randomly oriented telescope. The commitment is that no matter how one chooses parameter settings, the expectation values in consideration, which requires integration over the distribution , will always come out to reproduce the experimentally verified prediction of quantum mechanics. The correlations has to be just so reproduced and not, for example, reproducing some other hypothetical theory which gives a CHSH bound of instead of . In this way, it seems that giving up statistical independence puts extreme limitations on what experimental science we can do at all. At least it seems we make “free choices” in how we, for example, randomize drug trials and the like.
So, it is certainly logically possible that the world is local (in Bell’s sense) but we simply can’t apply the condition of statistical independence. But denying an assumption does not mean you’ve given a convincing argument or coherent theory that makes that inference plausible.
Info
Gerard ‘t Hooft, a Nobel winner in physics, has a theory of physics based on cellular automata that purports to be local but super deterministic. I have not studied his theory so I can not make any judgement call on it.
Single Measurement Outcomes
The interpretation (or theory) most famous for avoiding the usual single measurement outcome assumption we carry around in physics is the Many Worlds Interpretation, sometimes called the Everett interpretation. There, the radical idea is what if Schroedinger evolution was all there is? All measurements are Schroedinger evolution, particle entangling with apparatus, apparatus entangling with environment, and so forth, and the universe branches in order to create the perception, for records, memories, and/or agents that exist inside, that we only ever see one measured outcome.
We won’t go into the MWI details here, but, suffice to say that single measurement outcomes was an assumption in Bell’s theorem. So, strictly speaking, if you remove the single measurement outcome requirement you can avoid Bell’s exact conclusions. However, in MWI there is just the ontic universal wave function (of every universe) which is not fundamentally separable (ignoring “effectively separable”, e.g. via decoherence and branching). In my view, it doesn’t seem to make too much sense to talk about locality in an ontology that is monistic. Some MWI proponents, like Sean Carroll, freely admit the world is non-local. They don’t use MWI to avoid Bell’s theorem. There might be some MWI proponents however who insist that “branching happens locally”.
Musing
I have read Everett’s original (long version) dissertation and it was an interesting read. There is an interesting story there. Everett originally wanted to submit a long thesis laying out the details of the theory of the “Universal Wave Function”. However, John Wheeler, Everett’s advisor was worried that Niels Bohr (who was old but still around at the time) would not approve of such a dissertation. After talking to Bohr, Wheeler convinced Everett to shorten his thesis (from ~150 pages to ~25 pages) and the thesis was renamed to “On the Foundations of Quantum Mechanics” with the journal publication being named “‘Relative State’ Formulation of Quantum Mechanics”2. Niels Bohr reportedly still hated the idea and Everett quit physics after graduating.
Footnotes
-
Myrvold, W., Genovese, M., and Shimony, A. “Bell’s Theorem.” The Stanford Encyclopedia of Philosophy, edited by E. N. Zalta and U. Nodelman, 2024. https://plato.stanford.edu/entries/bell-theorem/ ↩︎ ↩︎2
-
Barrett, J. A., and Byrne, P., eds. The Everett Interpretation of Quantum Mechanics: Collected Works 1955–1980 with Commentary. Princeton University Press, 2012, p. 4, n. 4. ↩︎