The so-called "forest before the trees" phenomenon refers to the perceptual processing of the global information before the fine-grained analysis in vision. This advantage can be reversed by the presence of local salience. The same advantage for processing global temporal information is found in audition, but not for musical experts. Here, we ask whether and how salience modifies the local/global temporal auditory processing. To do so, participants (13 non-musicians, 7 expert-musicians) had to search, within melodic sequences of 9 notes organized in 3 triplets, for local or global modification applied to one triplet. In some conditions, these sequences were further manipulated in timbre (brightness) to make one of the triplets salient. Results show that the advantage for processing global information was substantially decreased with salience, regardless of its congruence with the modification to detect. Furthermore, the performance to find a local modification was improved for non-musicians when salience was congruent with that modification. We conclude that salience draws attention to local information, and that once at the local level, attention is directed by salience to the locally modified information. This reorganization of local/global temporal auditory processing with salience appears to be more efficient for non-musicians than for expert musicians.