Use of vocal tract in maternal-infant communication; non- vocal movements, gaze, facial expressions, etc.
These are largely emotional expressions and serving social and expressive functions. See Goodall (1995).
This event, 3+ mya., may have had some subtle triggering impact on restructuring of the vocal tract as well as in development of brain asymmetry.
After bipedalism, humans brains evolved into relatively larger more complex structures with left hemisphere language dominance, triple the neocortex of chimps, and also large relative gains in the cerebellum. This controls attention and movement so necessary for fine-grained observational learning (imitation) as well as skills like throwing and tool making-- and talking.
This includes the lowering of larynx, restructuring of tongue; finer controls of these features along with breathing control for prosodic aspects of language and singing.
These perhaps enhanced body consciousness, the ability to ascribe intent to others, observational learning ability, planning, tool making and use, etc..
All the large primates have somewhat differing social and sexual relationships; even the very closely related common and bonobo chimps. Humans are unique among these large apes in their high degree of male parental investment. One can imagine this is related to closer interpersonal male-female relationships which might lead to a greater degree of conventional communication. It may not be a coincidence that other species with close male-female bonds -- gibbons and parrots -- also share with humans a high degree of vocal coordination and interaction.
Accelerating accumulation of knowledge transmitted across generations, mother to offspring.
Known only in very limited cases in non-humans, this is a powerful function of human language.
Unknown in any other species, human language allows us to express our desires and beliefs --unlimited in number -- using the creative or projective aspect of human language. (See Limber, 1977)
Here is my best guess as to the steps in the evolution of modern human languages.
Building upon existing social communication, primarily of an unintentional, expressive sort; hominds became able to refer to objects and actions. These expressions --conventional but limited -- were learned just like other social customs of the familial group.
These expressions were somewhat productive in that new objects and actions might be substituted for the existing conventional ones.
This is analogous to human-based codes being used with dogs and apes, e.g. Kanzi -- with the very big difference it is the humans creating the conventions and their modifications.
"I think mommy wants me to pretend that I am asleep."
Modern human languages package multiple clauses into one continuous movement. These can be observed in young children during their third year as soon as they are able to "program" movements of four or more words at once. (See references in Limber, 1977.)
One speculative account of this is using the idea of James Mark Baldwin --the "Baldwin effect." This Darwinian process proposes that through the normal processes of learning, language skills were more or less gradually acquired by individuals as they matured in their familial hominid group. But because early acquisition of language was so advantageous in learning other skills, children who acquired language fastest had greater survival (fitness); hence humans became specialized in acquisition of language such that for us today it is NOT much related to other abilities (outside of vocabulary development.
(See Limber, 1982 for extended discussion.)
Clearly the above "steps", while, in what I am guessing as the chronological ordering, are NOT meant to suggest each was independent of the others. Certainly there is was extensive impact, say of language itself on the reorganization of the brain -- maybe even its size.