Despite the complexity that underlies the process of language production, speech is relatively error free. The errors that do occur, however, tend to be rule bound (e.g. phoneme exchanges occur within consonants or vowels, semantic exchanges occur within grammatical class, real words are more likely than non-words, and taboo words are less likely than neutral words). In addition, these errors are usually rapidly corrected, even under circumstances when overt speech perception is masked. These findings suggest some sort of filter or monitor, which allows for certain errors to occur but not others, and does so rapidly and without need for auditory processing. There are many models of such a system, with the perceptual loop monitor (Levelt, 1983) proposing that inner speech is monitored via the comprehension system, being the dominant one. The perceptual loop model is popular in part because it can account for a wide variety of patterns, including the “taboo words effect”- where people seem to be particularly good at avoiding errors that would result in taboo words. According to the perceptual loop theory, the comprehension of a taboo error is especially noticeable and so easier to detect and avoid (Motley et al., 1982). Other models that do not rely on comprehension of internal speech cannot obviously capture the taboo words effect, however they may be sensitive to factors correlated with taboo status such as arousal or valence and so may yet be viable alternatives to the perceptual loop model. The current study aims to a) replicate the original taboo word effect research, which- though widely cited- has yet to be replicated, and b) define what about the taboo words (e.g. emotional arousal/valence) might be driving the effect. By determining the relative role, if any, of arousal and valence in our ability to detect speech errors, these data will help distinguish between the models of error detection.