This paper deals with interpersonality in research article abstracts analysed in terms of interactional metadiscourse. The evolution in the distribution of three prominent interactional markers comprised in Hyland's (2005a) model, viz. hedges, boosters and attitude markers, is investigated in three decades of abstract writing in the field of applied linguistics in the broad sense. On the basis of a quantitative corpus survey of abstracts in "Journal of Pragmatics," two major points are made. One is that the distribution of hedges, boosters and attitude markers in abstracts, when compared with their distribution in research articles, supports the idea that abstracts are not just pale reflections of the full-length articles, but rather have a specific make-up, which can plausibly be linked to their function. The second point is that the use of interactional metadiscourse in abstracts has undergone interesting changes in the course of the past 30 years. On the whole, the degree of interpersonality realised by hedges, boosters and attitude markers diminishes over time, though notable differences exist with regard to the subcategories in the interactional domain. In the discussion section, we try to arrive at an explanation for the changes that have occurred, taking genre, discourse community, research practice and rhetoric strategy into account. (Contains 5 figures and 1 table.)