A Mach-Zehnder interferometer is set up so that a light wave is divided by a half-silvered mirror and travels on two different paths. After the photon-like wave function has been divided but before it is detected, the detector at the end of the interferometer is rapidly adjusted to one of two possible settings. If particulate photons exist, then the measured results from the two different settings depend on which path the conjectured particulate photon took. Under this assumption, the experimental results show that the conjectured photon made a choice of path after it had passed the point where the paths divide—a violation of our intuitive understanding of causality.

If, however, one assumes only the wave function exists (no particulate photons, as is justified in No Evidence for Particles), then quantum mechanics, as it is, gives the correct answer, with no after-the-fact choice involved. That is, the assumption of the existence of particles gets us in causality trouble, but no-particle quantum mechanics does not. So I would take the results of the Wheeler delayed-choice experiment as more evidence (in addition to that given in No Evidence for Particles) for no particles rather than as causality operating backwards in time (from the later time when the detector setting was changed, with the photon halfway to the second mirror, to the earlier time when the photon had not yet passed the first mirror).