One of the most comprehensive theories of music expectancy is the Implication-Realization [I-R] model by E. Narmour. It postulates that events implications are determined by bottom-up and top-down expectancy processes. The I-R original version postulated five bottomup processes, which cognitive reality has been identified. Several revisions of the model simplified it, reducing the analysis to only two processes: one in which, given a melodic interval, it is expected a following event close in register to the last tone (Pitch proximity) and other in which every interval implies change of direction, returning to the first interval tone register, being such implication more evident the higher the size of the given interval (Pitch reversal). In spite of the empirical support of the I-R reduced version, its application to the study of contemporary music composition has not been reported so far. The aim of the present research was to assess the validity of the I-R reduced version to describe bottom-up processes of melodic expectancy, present in the elaboration of the note-to-note level, during a contemporary music composition task. An experiment was run with 20 major students of music composition, who were required to compose a good continuation to 9 melodic fragments extracted from lieder by A. Webern. The first note of the composed continuation was analyzed to see if it satisfied the model’s implicative criteria. Results strongly supported the I-R revised version, except for small intervals. Further data analysis led to identify that small intervals had also clear implicative direction properties. As they were not reflected in the I-R reduced version, Pitch Reversal predictor was modified in order to capture them. Reformulation succeeded in describing data, except for one small interval. A concise study of this case led to hypothesize that answers could have been influenced by higher-level expectancy processes. If the three last notes of the fragment were considered instead of just two of them, the model could efficiently predict the found responses. Overall results support the I-R revised version to describe expectancy processes at the note-to-note level and suggest the model’s preliminary validity to describe expectancy processes that occur at higher levels of musical structure.