In order to deal with the metonymic grounding of illocutionary meaning, Panther and Thornburg (1998) have proposed the existence of illocutionary scenarios. These are complex structures generally consisting of three main parts: a before, a core, and an after, which specify traditional Searlean felicity conditions in a cognitive-model theory format. These scenarios are accessed metonymically. Perez and Ruiz de Mendoza (2002), on the basis of previous work by Ruiz de Mendoza (1999) and Perez (2001), have refined Panther and Thornburg's account to make it include additional pragmatic variables like power and solidarity relations (necessary to differentiate requests from orders, for example), and a cognitive version of Leech's (1983) well-known cost-benefit scale (useful to assess the degree of politeness of directives and commissives). In this paper, it will be our purpose to explore how illocutionary scenarios are constructed and exploited linguistically to convey illocutionary meaning. We will first argue that illocutionary scenarios (e.g. requesting, offering, apologizing) are high-level situational models constructed through the application of the high-level metonymy SPECIFIC FOR GENERIC to multiple low-level situational models. Once created, an illocutionary scenario may be accessed metonymically. Such scenarios are then applied to specific situations through the converse metonymy, GENERIC FOR SPECIFIC. On the basis of a corpus of examples derived from different sources (novels, film scripts, forum emails), we have identified central and peripheral elements of illocutionary scenarios belonging to most directive, commissive and expressive speech act categories. We have identified elements common to all of them and have constructed a broader higher-level description that we call the Cost-Benefit Cognitive Model. The corresponding notion in Leech's pragmatic theory was formulated to apply to directive and commissive speech acts. However, we have found that the scale also applies to expressive speech acts to the extent that they are regulatory of speaker-hearer interaction. Our cost-benefit model is thus a cluster of submodels that show family resemblance connections. Finally, we study a number of conventional and non-conventional linguistic realizations of the various parts of the Cost-Benefit model and explore the way in which such realizations are used to produce illocutionary meaning. We argue in this connection that the non-semantic part of a construction has a realizational potential that may be captured by means of sets of semantic conditionalsbased on the Cost-Benifit Cognitive Model (Francisco Jose Ruiz de Mendoza and Annalisa Baicchi)