I'll be here on Friday 12-2pm EDT: https://bluejeans.com/767230868 The anterior post-N400 ERP positivity (PNP) has received recent attention as a possible index of the cost of recovery from violated predictions. Previous studies have found that the PNP is larger at unexpected words in strongly vs. weakly constraining contexts, but crucially, only when the unexpected word was still plausible. Contrary to these findings, in a previous experiment, we found evidence for the same constraint-based PNP effect in separable verb-particle constructions even though the target particle was completely implausible. We attempted to replicate our finding in a pre-registered study of 100 participants, but were unable to replicate the effect. Our results therefore support previous research suggesting the PNP is not sensitive to constraint at implausible words. This unfortunately also means that the PNP was not sensitive to our constraint manipulation and therefore that we could not tell whether readers predicted the verb particle or not.