This paper assesses the limitations and the gains of using fiction films as specimens for the study of language. It argues that the conditions of possibility of film use are dependent on the focus and the framework of linguistic investigations. The dialogues’ lack of spontaneity is regularly cited as an argument against studying film speech, but since many non-spontaneous corpora are indeed used in linguistics, this issue requires a detailed discussion. The actual differences between film speech and naturally-occurring interactions are grounded here theoretically. Arguments drawn from social studies and art theory explain how the filters of representation and artistic creation can pose essential quantitative and qualitative problems to studies interested in spontaneous speech (some of these arguments also explain why foreign language teachers have enthusiastically embraced the use of films and TV-series in the classroom). From these limitations, the paper turns to the possible advantages of film corpora for studies tackling linguistic-pragmatic strategies in dialogue, as illustrated by qualitative and quantitative analyses. It concludes with recommendations based on the types of linguistic objects and approaches, indicating in which cases the use of films can pose essential problems, and in which cases it can prove a good crucible for qualitative studies, as well as an accessible workbench to elaborate quantitative hypotheses.