Two tasks have been predominantly used over the past thirty years to measure false belief understanding: the Location Task and the Typical Box task. These tasks have produced robust findings that children fail false belief tasks at age 3 and pass false belief tasks at age 4. Recent theory, however, points out a shared confound in the 2-option Location and Typical Box tasks. This confound would allow children to perform successfully on the standard false belief tasks without understanding belief. Instead, children might be using perceptual access reasoning and reason that ignorant agents will "get it wrong" and be incorrect about reality. Modified 3-option tasks were used to address this confound by introducing a third, irrelevant option to the 2-option tasks. According to PAR, children who pass 2-option tasks should perform worse on the 3-option tasks because there are two "wrong" answers. We argue that subtle differences in salience between the false belief and irrelevant options in combination with one open-ended test question can draw children who use PAR toward one or the other in unpredictable ways. To demonstrate that other procedures will give more salience to the irrelevant options several studies are needed, each with minor variations in procedure that do not alter the basic false belief structure. Thus in five studies we varied superficial characteristics across tasks in order to test for a task effect across studies. We used the "continuously cumulating meta-analysis" (CCMA) approach, combining each replication study into a broader analysis (final N=113) for higher power. Our CCMA analyses provide strong support for the PAR hypothesis because 1) children performed worse on the 3-option tasks than the 2-option tasks, 2) children's proportion of false belief responses out of non-reality responses did not replicate across studies, 3) children's proportion of reality responses replicated across studies, and 4) the Location task was easier than the Box tasks across studies. These findings suggest that there is a lack of construct validity in traditional false belief tasks; thus, new methods of testing for false belief understanding are needed to determine at what point children acquire Theory of Mind.