The second stage of analysis used the description of apologies outlined in Blum Kulka et. al. (1989) to code a smaller subset of the material, derived from the posts which contained the most frequently used lexemes associated with the speech act of apologising (sorry and apologise).