A good instrument does not settle a question by itself, and the literature Schmidt founded has followed a trajectory that is now familiar. Individual studies, his own included, reported small positive effects. Replications by other laboratories were fewer than the field claimed and more mixed than its summaries suggested.
When Bösch, Steinkamp and Boller pooled the random generator literature in a meta-analysis published in Psychological Bulletin in 2006, they found a small overall effect that was statistically significant and, decisively, an inverse relationship between effect size and sample size: the biggest deviations came from the smallest studies and the largest studies came closest to chance. That pattern is the standard fingerprint of publication bias, and the authors concluded the data were better explained by selective reporting than by psychokinesis.
Proponents, including researchers at the Institute of Noetic Sciences, disputed the modelling and argued the correction had been applied too aggressively. Schmidt himself continued to publish, later directing research at the Institute for Parapsychology in Durham and at the Mind Science Foundation in San Antonio, and he never accepted that his effect had been explained away. The dispute about the pooled literature remains open; the difficulty of getting a clean independent replication of a single Schmidt-style result does not.