Working backward from clues in ancient skeletons, Dutch researcher Bart de Boer has built plastic models of an early hominin's vocal tract and recreated the sounds our ancestors may have made millions of years ago.
     Non-human primates have an organ called an air sac - a large cavity that connects to the vocal tract - linked to an extension on the hyoid bone known as the hyoid bulla. Modern humans have neither an air sac nor an extension on the hyoid bone, but Australopithecus afarensis - a hominin species that roamed Africa approximately 3.9 million to 2.9 million years ago - had a hyoid bulla, and likely an air sac as well. De Boer built models of the human vocal tract both without an air sac, like modern humans, and with one, as Australopithecus afarensis would have had. By pushing air through the models, he could hear what various vowels sounded like.
     The air sacs acted like bass drums, resonating at low frequencies, and causing vowel sounds to merge; according to Charles Harvey at New Scientist, [an Australopithecus] would have had a greatly reduced vocabulary. What, then, might our ancestors' first words have been? With air sacs, vowels tend to sound like the 'u' in 'ugg', but studies suggest it is easier to produce a consonant plus a vowel, and 'd' is easier to form with 'u'. "I think it is likely cavemen and cavewomen said 'duh' before they said 'ugg'," says de Boer.

Edited from NewScientist (23 November 2011), Discover Magazine (28 November 2011)