Reading about early development of agriculture, there's a lot of uncertainty about domestication being deliberate and conscious or not. When do we have some certainty?
Some of the first archaeological evidence of deliberate selective breeding comes from ancient China. Starting approximately in the 8th century BCE and continuing through the Qin Dynasty (about 200 BCE) farmers began deliberately breeding pest-resistant rice as part of a larger expansion and refinement of collective agricultural knowledge [1,2]. By the 1st century CE, farmers were breeding purpose-specific rice strains and experimenting with breeding methods [2].
Plants were, of course, not the only organisms subject to selective breeding. Xenophon, writing in the 4th century BCE, advised his readers not to breed hunting hounds which are “hook-nosed, grey-eyed, blinking, ungainly, stiff, weak, thin-coated, lanky, ill-proportioned, cowardly, dull-scented, [or] unsound in the feet” before going on to explain a number of behavioral issues, some of which he describes as “natural defects” in temperament, and some the product of “unintelligent training”; he tells his readers to “mate them [bitches] with good dogs”, with “good” dogs being defined as having “pluck, keen noses, sound feet and good coats”[3].
It’s particularly noteworthy that Xenophon warns his audience that “The colour of the hounds should not be entirely tawny, black or white; for this is not a sign of good breeding: on the contrary, unbroken colour indicates a wild strain...tawny and the black hounds should show a patch of white about the face and the white hounds a tawny patch.” [3] We now know that pigmentation changes-- particularly white or brown patches-- are part of what has been termed domestication syndrome, and are correlated with docile and friendly behavior. Arrian, writing of the “Celtic hounds” (greyhounds) in the 2nd century CE, states that the “temper” of greyhounds is also a result of “good and bad blood”, and that the reader should breed hounds with “the most attachment to man, and to whom no human countenance is strange”, whilst avoiding animals that “are afraid of people, and are astounded at noise…[or] pay no attention to you.” [6] This passage shows that the high heritability of behavior and temperament in dogs was not only known, but used as a criteria for selective breeding.
In the 3rd century BCE, Mango of Carthage was writing his 28-volume treatise on agriculture. It incorporated both Punic and Amaziɣ (Berber) traditional knowledge on both plant cultivation and animal husbandry. The book was so well-regarded that it was taken by the Romans during the destruction of Carthage and brought back to Rome for translation into Latin [4] . One section of the text gives advice on the selection of bulls for breeding. It also provided source material for Roman agriculture writers such as Varro, Pliny, and Columella, who then added to his ideas [5,7] . Verro writes in the 1st century BCE of “weeding out” inferior sheep, and of purposely breeding ewes which consistently produce twins [7].
Columella, in the 1st century CE, writes that “experience has taught the way to produce other variations of color in this kind of animal,” and goes on to describe his uncle creating a new sheep color by crossing African rams with Greek sheep. He also notes that “the marks of the father generally persist in the offspring” and that “a dark lamb is often the offspring of a white ram, but a white lamb is never bred from a red or brown sire.” [5]. Sheep coloration is determined by a single gene, with the white variant being dominant; this observation is possibly the earliest explicit reference to dominance and recessiveness of genetic traits.
Even though our records are incomplete-- for example, there is a dearth of documentation on early plant and animal domestication in the Americas-- it is clear that by the 1st century CE, humans were aware of the basic principles of heredity and applied them to domestic plants and animals.