In the monkey, cortical auditory projections subdivide into a dorsal stream mostly involved in spatiotemporal processing, that projects mainly to dorsal frontal areas; and a ventral stream involved in stimulus identification, connected to the ventrolateral prefrontal cortex (VLPFC). We propose that in the human lineage, part of the dorsal auditory pathway has specialized in vocalization processing, enhancing vocal repetition and short-term memory capacities that are crucial for linguistic development. In the human, the vocalization-related dorsal auditory component tends to converge in the VLPFC with the ventral auditory stream and with projections involved in gestural control, and consists of a direct connection between the auditory cortex and the VLPFC via the arcuate fasciculus. and an indirect pathway via the supramarginal gyrus. Additionally, intraparietal and inferior parietal afferents to the VLPFC are associated with communicative hand gestures, with manipulation skills and with early tool-making. Although in general terms compatible with the mirror-neuron gestural hypothesis for language origins, this proposal underlines the participation of the dorsal auditory pathway in voice processing as a key event that marked the beginning of human phonology and the subsequent evolution of language. Instead, the mirror neuron system for gestures and the primitive vocalization network (ventral pathway) contributed to provide a communicative scaffolding that facilitated the emergence of human-like phonology. Furthermore, we emphasize the phylogenetic continuity (homology) between non-human and human vocalization and their neural substrates, something that is not usually stressed in the mirror neuron perspective.