Abstract Analysis of the shape of hierarchies of association to emotional and neutral words reported by John (1988) shows that care is needed in the choice of words for experimental studies of recall and the priming effects of affect-laden words. Emotional words tend to fall into a particular subset of words with associative characteristics and failure to recognise this may lead to inferences being drawn about the effect of affect which may more parsimoniously attributed to the associative characteristics of the words.