Daniel Cloud has written extensively on the domestication of words and language. Cloud credits Dennett with the idea that language could be domesticated - though he argues that Dennett didn't take the idea far enough. The earliest reference to domesticated memes from Dennett I can find is in his 1998 essays Memes: Myths, Misunderstandings and Misgivings AND Snowmobiles, horses, rats, and memes.
Dennett goes on to discuss the idea of domesticated memes some more in Breaking the Spell (2006), writing:
What I now want to suggest is that, alongside the domestication of animals and plants, there was a gradual process in which the wild (self-sustaining) memes of folk religion became thoroughly domesticated. They acquired stewards. Memes that are fortunate enough to have stewards, people who will work hard and use their intelligence to foster their propagation and protect them from their enemies, are relieved of much of the burden of keeping their own lineages going. In extreme cases, they no longer need to be particularly catchy, or appeal to our sensual instincts at all. The multiplication-table memes, for instance, to say nothing of the calculus memes, are hardly crowd-pleasers, and yet they are duly propagated by hardworking teachers — meme shepherds — whose responsibility it is to keep these lineages strong. The wild memes of language and folk religion, in other words, are like rats and squirrels, pigeons and cold viruses — magnificently adapted to living with us and exploiting us whether we like them or not. The domesticated memes, in contrast, depend on help from human guardians to keep going.
However, I notice that Adam Westoby seems to have written extensively on domesticated memes in 1994. He has the idea that memes domesticate humans as well as the idea that humans domesticate memes. Here's his 1994 manuscript. To quote from it:
The memes of theoretical natural science, as Wolpert (1992) points out, are highly "unnatural" memes, remote from "common sense". Like cattle or sheep, they have been bred for generations into the forms preferred by their domesticators (of whom some of the most important are other memes). Testability, generality, uniform vocabulary, unambiguous meaning, internal consistency, and so on - even taken singly such traits are rare memes, and to assemble them all requires long intentional selection. The domesticated memes of theoretical natural science, having embodied such significant adaptations to artificial circumstances, could no longer survive reintroduction to the wild. They can live and breed only with the aid of rather complex arrangements to sustain them. The cultivation of theoretical science (like keeping sheep) has come to rely on auxiliary breeds, such as scientists - rather like sheepdogs, who keep the flock together and bark at intruders. By comparison, much social science consists of more "common sense" memes, less "deformed" by domestic breeding. They more resemble semi-domesticated breeds which forage freely on the mountain slopes in summertime, but are herded in for the winter.Westoby is the earliest reference to the idea of domesticated memes I have found so far. Is this the true origin story for the idea that memes could be domesticated? Did anyone else come up with this idea earlier? Please let me know if there's an earlier reference that I'm currently missing.
The importance of domestication in cultural evolution is apparently an illustration of the superiority of memetics in this area - compared to other strains of cultural evolution. It looks as though meme enthusiasts got to this idea first - because they have a symbiosis-aware version of cultural evolution. Academics are now picking up the idea (for example, Joseph Henrich's latest book has culture domesticating humans in its subtitle) but they appear to be playing catch-up.