The notion that "replicators" play a uniquely distinguished role in biological evolution has been championed by Richard Dawkins (1976, 1989). Furthermore, Dawkins has argued that this idea can be generalised in a way which makes it applicable to any properly Darwinian evolutionary process, at least if that process gives rise to a growth in adaptive complexity (Dawkins 1983). It is evident, therefore, that if Dawkins' analysis is correct, it has profound implications for any attempt to realise a growth of adaptive complexity in artificial systems by Darwinian means.
This paper is concerned with trying to clarify just what, exactly, a Dawkinsian "replicator" might be. This has implications both for the specific field of Artificial Life, but also for the general debate in evolutionary biology about "units of selection" . I provide a reformulation which, I claim, captures the valid core of Dawkins' insight, while, at the same time, avoiding certain confusions and misconceptions which might otherwise be read into his views.