This study looks at an important methodological issue regarding the design of corpora of pop culture sources for corpus-based media studies. More specifically, by way of a case study, it explores whether transcriptions are the only reliable means of representing telecinematic discourse. To this end, first, a qualitative analysis of scripts (the language planned by scriptwriters), transcripts (the language viewers actually hear), and subtitles (the language that generally reproduces what viewers actually hear) is carried out to highlight the similarities and differences among these sources. Second, a Multi-Dimensional Analysis (MDA) compares the dimension scores of transcriptions and subtitles of the same movies, applying Biber’s (1988) dimensions of register variation in English. Results of the MDA show small differences on particular dimensions with respect to the complementary co-occurrence of grammatical features.