This chapter sets out to analyse apologies in the framework of corpus pragmatics and evaluates the validity of a form-to-function approach in the study of apologies. Apologies have often been claimed to be formulaic in that they are most frequently realised by an Illocutionary Force Indicating Device (IFID). In form-to-function approaches, such IFIDs have been used as search tokens to retrieve apologies from corpora. Implicit apologies (that are defined by the absence of an IFID in their form) cannot be retrieved with such an approach. That is why less routinised speech acts are typically retrieved from corpora through manual searches in function-to-form approaches in order to be able to retrieve all realisation forms. As an alternative to manual searches, pragmatic annotation systems have been developed that can be used to retrieve speech acts from corpora. The present study compares the realisations of apologies across three different data collection procedures: (1) a form-to-function approach using IFIDs as search items, (2) a function-to-form approach relying on the pragmatic annotation system SPICE, and (3) a function-to-form approach employing manual searches of the corpus material. The results suggest that implicit apologies occur more frequently than previously thought. This poses a difficulty for form-to-function approaches, as this type of approach will not be able to retrieve implicit apologies.