The action-sentence compatibility effect (ACE; Glenberg & Kaschak, 2002), a hallmark finding in Embodied Cognition, implicates the motor system in language comprehension. In the ACE, people process sentences implying movement toward or away from themselves, responding with actions toward or away from their bodies. These processes interact, implying a linkage between linguistic and motor systems. From a theoretical perspective, the ACE has been extremely influential, being widely cited evidence in favor of embodied cognition. The present study began as an attempt to extend the ACE in a new direction, but eventually became a series of attempts to simply replicate the effect. Across 8 experiments, I tested whether the ACE extends to a novel mouse-tracking method and/or is susceptible to higher-order cognitive influences. In 3 experiments, attempts were made to “disembody” the ACE by presenting participants’ names on the computer screen (as in Markman & Brendl, 2005). In each experiment, the ACE could not be disembodied, because the ACE did not occur. In further experiments, the ACE was not observed in reading times, regardless of response mode (mouse movements vs. button-presses) or stimuli, including those from the original research. Similarly, no ACE was observed in physical movement times. Bayes Factor analyses of the current experiments, and the previous ACE literature, suggest that the evidence for the ACE is generally weak: Many studies considered as positive evidence actually support the null hypothesis, and very few published results offer strong evidence for the ACE. Implications for the embodiment hypothesis are discussed.