It is now generally accepted that words’ emotional content plays a role in lexical processing, but the literature offers incompatible findings concerning what this role may be. Here we use a large sample of lexical decision data (British Lexicon Project, Keuleers et al., 2012) and we carry out a series of analyses differing in the way emotional variables are treated. A variety of statistical approaches yielded common conclusions: when confounding variables are taken into account, emotional words, whether positive or negative, are processed faster than neutral words. This effect is categorical rather than graded; is not modulated by emotional arousal; and is not limited to words explicitly referring to emotions. We discuss this in terms of internally grounding words’ meanings in emotional experience, akin to the manner in which concepts may be grounded in perception and action.