Numerosity and privileges of occurrence of various types of interjections (primary conventional, primary non-conventional, secondary, and onomatopoeic) were investigated in three different literary readings of Winnie-the-Pooh (Milne 1926), in one reading of Ulysses (Joyce 1960), and in an artistic performance by actors (the film The third man, Korda, Selznik, & Reed 1949). The spoken corpora, based on printed texts as source, consisted of 667 interjections. Ameka’s (1992 b, 1994) hypothesis that, parallel to their independence from ambient grammar, interjections would also be isolated temporally by preceding and following pauses, was not confirmed; for the entire corpus, only 39% of all interjections were thus isolated. However, an alternative hypothesis, that interjections serve an initializing function, was confirmed: Altogether, 77% of the interjections were found to be initializing, i.e., were preceded by a pause, introduced a speaking turn, introduced an utterance, and/or introduced a citation. Primary conventional interjections constituted the majority of interjections (overall 56%), but only two of these were common to all the corpora (oh and ah). By far the highest percentage (28 %) of primary non- conventional interjections occurred in the artistic performance of The third man. None of these occurred in either the novel or the screenplay of The third man, unlike the primary non-conventional interjections throughout the text of the literary readings. Functions of interjections are discussed in terms of Goffman’s (1981: 226) animators (literary readers, 26% of whose spoken interjections were added to those in the printed text) and principals (actors, 79% of whose spoken interjections were added to those in the printed text), in terms of literacy and orality, and in terms of the emotional stance and perspective of a speaker at the very moment of utterance.