While some cultural items may indeed be propagated by imitation and other forms of copying, it is clear that a large number are not. In particular, many are also (re-)constructed. For example, a student taking notes in a lecture does not simply copy any spelling error that the lecturer happens to write down, but will in fact, in her own notes, correct the error and in doing so re-construct the correct spelling. As such, cultural propagation is partly preservative, but also partly (re-)constructive, to different degrees in each particular case. As such, it is not only a matter of inheritance, as is gener- ally the case for biology, but also of reconstruction. Whichever of these is more important in any given case is an empirical question, but either way, the direct analogy with biological evolution is considerably weakened by this fact.Correcting spelling mistakes is rather a lot like the error correction involved in DNA copying mechanisms. So, this seems like a bad example of an alleged disanalogy - it's a clear case where there are similar error correction mechanisms in both domains.
The authors go on to say:
The mechanisms involved in cultural transmission (with rare exceptions such as rote learning) have, in various degrees and forms, both preservative and constructive functions. This is quite unlike the biological case, in which the proper function of the copying mechanism (replication) is preservative alone.
It is true that memes are "constructively" modified inside human brains.
However, genes are "constructively" modified inside human bodies too. For example viral genes may undergo adaptation to their human host, which may facilitate transfer to their human relatives. Or bacterial genes may develop resistance to anti-bacterial herbs that their host consumes.
In both cases, adaptive evolution takes place inside the host before being transmitted to a new host.
To argue that cultural transmission of memes between hosts is "constructive" - while organic transmission of viral genes between hosts is not - one has to consider selective processes in the cultural case, while screening them out of the biological case. The result is then not a fair comparison.
Sperber is just ignoring selective processes that take place inside brains during learning in odder to make his point. If you don't consider the evolution of memes within brains, then meme evolution seems to involve amazing intelligent-design processes not present in the organic case. However, once you understand that selection and evolution take place within brains during learning, meme evolution and gene evolution start to look much more similar to one another again.
The paper gets at least one thing right. It says:
How deep is the analogy between biological and cultural evolution? Memetics assumes that it is deep indeed; that the main relevant details of the biological case have direct equivalents in the cultural case, such that there is, for example, a cultural phenotype, which achieves a certain level of (inclusive) fitness, which will in turn determine the phenotype’s relative success in the population.
I endorse this description. Inclusive fitness and kin selection apply to memes - in cultural kin selection. Also, the phenotype/genotype split is useful in the case of culture - probably most useful if "genotype" refers to heritable information and "phenotype" refers to its products.