Impersonal subjects of tensed sentences are expressed in different languages in different ways. In French, they are expressed by the lexical pronoun "on", and in languages such as Spanish and Italian, the reflexive pronoun is used to trigger "impersonal subject" interpretation. This so-called impersonal-reflexive construction is well-known and well-studied in the framework of generative grammar. Impersonal subjects of infinitives are "expressed" in the same way in French, Spanish and Italian, by the zero form called "arbitrary PRO" and not by lexical "on" nor by the reflexive pronoun.One puzzling fact is that there are no "impersonal objects" of the same type: lexical "on" in French cannot be used as an object, and there is no impersonal-reflexive object construction in Italian and Spanish. Another puzzling fact is that impersonal subjects have always human reference. Thus, a verb such as "to bark" selecting a dog as its subject cannot take an "impersonal subject" with "on" in French nor with the impersonal-reflexive in Italian and Spanish. Arbitrary PRO in infinitives is also limited to subject position, and it refers to human beings. Morphosyntactic studies on impersonal-reflexive constructions have uncovered facts which suggest that some properties derive from Discourse Representation properties.A detailed study of how the "impersonal subject" referent can be treated in a formal theory of discourse representation will be rewarding. It is interesting that "impersonality" has not been discussed in theories of Discourse Representation, although it must relate to some kind of discourse reference.

The main problems can be summarized as follows:

• Why are "impersonal subjects" not found in non-subject position?
• Why do "personal subjects" refer to persons, and not to animals or things?
• Why are "impersonal subjects" expressed in so vastly different ways?
• How do "impersonal subjects" relate to arbitrary PRO, the subject of argumental infinitives, and to the understood agent of agentless periphrastic passives?
• What other expressions of "impersonality" are used, and how are they interpreted?

The starting point of the investigation will be to review the procedure of identifying the discourse participants, by assuming them to be given at the beginning of the stretch of discourse. The "impersonal subject" can be interpreted as a "third participant", given at the beginning of the stretch of discourse and linked to specific morphosyntactic properties. The "first participant" is the speaker and the "second participant" is the addressee. Furthermore, morphosyntactic predicates can take a subject supplied by some specific discourse interpretation rules; in this case, the "impersonal subject" is supplied by the discourse interpretation rules. The project aims at combining the morphosyntactic approach to "impersonal subjects", which have been studied in many generative grammar studies of the Romance languages, and the approach implied by Discourse Representation Theory.