This paper outlines an argument for how development in child speech and gesture could shed light on language evolution: child acquisition can be thought of as two types of acquisition, one of which goes extinct (gesture-first, Acquisition 1) and is replaced by another (gesture–speech unity, Acquisition 2). For ontogenesis, this implies that children acquire two languages, one of which is extinct, and which again goes extinct in ontogenesis (it continues as “gestures of silence” rather than as gestures of speech). There is no way to get from Acquisition 1 to Acquisition 2. They are on different tracks. Even when they converge in the same sentence, as they sometimes do, they alternate and do not combine. This paper proposes that the 3~4 year timing of Acquisition 2 relates to the natural selection of a kind of gestural self–response labelled here “Mead’s Loop”, which took place in a certain psychological milieu at the origin of language. This milieu emerges now in ontogenesis at 3~4 years and with it Mead’s Loop. It is self-aware agency, on which a self-response depends. Other developments, such as theory of mind and shared intentionality, likewise depend on it and also emerge around the same time. The prefrontal cortex, anchoring a ring of language centers in the brain, matures at that point as well, another factor influencing the late timing. On the other hand, a third acquisition, speech evoking adult attachment, begins at (or even before) birth, as shown by a number of studies, and provides continuity through the two acquisitions and extinction.