In noisy backgrounds, listeners parse the concurrent auditory streams (“stream segregation”), and selectively focus on one stream as it unfolds over time (“selective attention”), performing the auditory scene analysis. Stream segregation is thought to remain immature in 12 year-olds with normal hearing. Here, we sought to investigate developmental changes in the neural signature of auditory stream segregation from childhood to adulthood. Children (n = 17), adolescents (n = 12) and young adults (n = 20) were presented with sequences of sounds consisting in coherent auditory figures imposed on stochastic backgrounds (“tone clouds”). These figure-ground sequences have been shown to elicit distinct EEG responses, including the object-related negativity (ORN) and P400, which reflect the processing of concurrent auditory objects. Participants were also presented with a consonant identification task in three conditions: in quiet, in the presence of one interfering talker, and in the presence of speech-shaped-noise. Results indicate a progressive improvement in the behavioural figure-ground segregation. Amplitude of both ORN and P400 decreased with age; and was not significantly correlated with the behavioural performance. There was no clear relationship between ORN/P400 and speech perception in noise. Yet, our results suggest a protracted development of stream segregation from childhood to adulthood. This will be discussed in relation with the literature on auditory scene analysis and the development of the central auditory pathways.