A Letter on:

Three Challenges for the Survival of Memetics

Such frameworks have considerable power over those that hold them for these people will see the world through these `theoretical spectacles' (Kuhn 1969) - to the converted the framework appears necessary. The converted are ambitious to demonstrate the universality of their way of seeing things; more mundane but demonstrable examples seem to them as simply obvious. However such frameworks will not continue to persuade new academics if it does not provide them with any substantial explanatory or predictive `leverage'. Memetics is no exception to this pattern.

For this reason I am challenging the memetic community of academics to achieve the following three tasks of different types:

• a conclusive case-study;
  
  
• a theory for when memetic models are appropriate;
  
  
• and a simulation of the emergence of a memetic process.

If these challenges are met, memetics will almost certainly survive [note 1], if not then it will not die immediately, just be increasingly ignored until it becomes merely a minor footnote in the history of science. As memeticists you have to decide! Will you stop the over-ambitious theoretical discussion and do some of the mundane foot-work that will actually advance knowledge of memetics processes? As David Hull said at the Cambridge meme conference [note 2]:

"Stop talking about Memetics and start doing it."

Challenge 1: A conclusive case study

• Exhibit a replicator mechanism - this needs to be something physical and not in the mind. The mechanism must provide a testable cause of the claimed evolutionary process. It must faithfully replicate with a low level of error or change (although there must be some variation). There must be no doubt that particular inheritable patterns have been accurately replicated many times over.
  
  
• The lineages of the replicator must be unbroken for long enough to allow a process of adaptation to exterior factors to occur. If a meme originates from a few central sources and is only replicated a few times away from these, then this is insufficient. Thus if lots of people copy an idea from a particular book and this does not then take on a replicative momentum of its own then this can not evolve. Even when there is a demonstrable ability to imitate and the population statistics suggest that there is an evolutionary process occurring it can still be the case that no sustained evolution is actually occurring (Edmonds 1989).
  
  
• Over a long time period the success of a replicated meme must be demonstrably correlated to identifiable comparable advantages of a meme in terms of the mechanism and context of replication. If reasons why one meme is more successful than another are only based on vague plausibility, then this is not enough.
  
  
• The dynamics need to be numerically consistent with the applicable theories of population genetics, e.g. Price's covariance and selection theorem (Price 1970, 1972).

Possible cases might include some of the following.

• Nursery rhymes. Here there is a demonstrable copying process of infants rote learning rhymes from their parents and school teachers. The rhyming mechanism and regular meter helps ensure the accurate replication across generations, and it might be possible to relate the success of rhymes to its features (e.g. how easy they are to remember). There is some evidence going back hundreds of years to the `chap books' in the first age of popular printing (Opie and Opie, 1997).
  
  
• Legal Phrases. Successful legal phrases (i.e. those that succeed in court cases) are repeatedly reused in legal documents such as contracts and articles. They are copied exactly so as not to open the opportunity for a new interpretation by a court. A study of their population dynamics and lineages could be made to show that a substantive evolutionary process occurred as a result.

Challenge 2: A theoretical model for when it is more appropriate to use a memetic model

It seems clear to me (a memetic agnostic) that some human beliefs are more sensibly considered to be of a non-memetic character. For example, I may gain the information that the number 192 bus leaving Stockport goes to central Manchester, and I may even tell someone else this fact. However, the chains of referral are likely to be very short - that is to say, it is likely that individuals will not rely on obtaining this type of information from long chains of communication due to the likelihood of errors being introduced. Rather they will tend to go back to the original source - the centrally originated timetable. The `fitness' of this information lies not in any intrinsic propensity for being communicated but rather due to its utility in utilising the bus system for personal transport, i.e. its truth.

For other information it may be more appropriate to model a pattern of information as if it had an evolutionary life of its own, separable from the advantage it confers on its `hosts'. For example it may be that the success of nursery rhymes is more strongly correlated with its memorability rather than any utility - that almost any monotonous rhythmic words might be as good as any other for the purposes of getting children to sleep or teaching them language, so that the reason why particular rhymes spread is due to their replicability. In such a case a memetic model might explain the variety and dynamics of rhyme spread in a way that is not possible with models based on individual advantage.

What is needed is some (falsifiable) theory that (under some specified conditions) tells us when a memetic analysis is more helpful than a more traditional one. Such a theory would have to meet the following criteria.

• It would have to make some sort of prediction of when a memetic model was appropriate - i.e. it had explanatory or predictive value - and when not. In other words when it is helpful to model a pattern that has been copied as a self-interested meme.
  
  
• The theory would be workable on information that was sometimes possible to obtain, i.e. not based on unobtainable information (e.g. the composition of mental states).
  
  
• The theory would have to be understandable in terms of the credibility, appropriateness and clarity of its core mechanism. The assumptions under which the model works would need to be fairly clear and practically determinable.
  
  
• The theory would need to be validated against observable phenomena, not just established by the plausibility of its assumptions.

Challenge 3: A simulation model showing the true emergence of a memetic process

The criteria that such a simulation model should meet are the following.

• The micro behaviour of the individuals needs to be credible. That is they need to reflect patterns of behaviour that third parties [note 3] would accept as being really possible. Thus behaviour based on strong a priori assumptions (e.g. utility optimisation) or unmodified off-the-shelf algorithms (e.g. Genetic Algorithms) would not be suitable.
  
  
• The emergent behaviour must be demonstrably evolutionary in character by the criteria in Challenge 1. That is to say there must be substantial and repeated accurate replication of patterns. Patterns replicative success must be demonstrably due to their characteristics. There must occur long, unbroken lineages for the evolution to act on etc.
  
  
• The emergent memetic process must not be directly `designed into' the simulation. This can be a difficult criterion to judge but, at a minimum, there should be: no built-in and inevitable processes of replication or imitation; the emergent evolutionary process should be contingent upon certain conditions and settings; and the behaviour of the individuals not obviously distorted to encourage the evolutionary process to occur (i.e. they retain some descriptive credibility).

Such a simulation could be composed of a population of interacting and self-interested individuals that are evolving in a reasonably complex environment. It would need to be shown that a secondary process of, first, imitation and, later, evolution, arose out of their interactions, so that, eventually, the secondary evolutionary process would become substantially self-driven rather than in the direct interest of the individuals (in the sense of Challenge 2). The emergence of a memetic process goes beyond just comparing whether pre-determined genetic or cultural operators won out (or were more effective) - it is the equivalent of exhibiting a simulation of the emergence of life from the interaction of chemicals.

Notes

1. That is unless subsumed within a new theory that is more general and powerful.
   
   
2. Reported by Andrew Lord, and confirmed in a personal communication with David Hull. For a more prosaic version see Hull's contribution (Hull 2000) to the resulting book.
   
   
3. By "third parties", I mean academics outside the field who have no particular interest in promoting (or, indeed, denigrating) memetics, for example biologists.

References

Edmonds, B. (1998). On Modelling in Memetics. Journal of Memetics - Evolutionary Models of Information Transmission, 2. http://cfpm.org/jom-emit/1998/vol2/edmonds_b.html

Gatherer, D. (1998) Why the Thought Contagion Metaphor is Retarding the Progress of Memetics. Journal of Memetics - Evolutionary Models of Information Transmission, 2. http://cfpm.org/jom-emit/1998/vol2/gatherer_d.html

Hull, D. (2000) Taking Memetics Seriously: Memetics will be what we make it. In Aunger, R. (ed.), Darwinizing Culture: The Status of Memetics as a Science, Oxford: Oxford university Press, 43-67.

Kuhn, T. (1969) The Structure of Scientific Revolutions. Chicago: University of Chicago Press.

Price, G. R. 1970. Selection and covariance. Nature 227: 520-521.

Price, G. R. 1972. Extension of covariance selection mathematics. Annals of Human Genetics 35:485-489.

Rose, N., 1998; Controversies in Meme Theory.Journal of Memetics - Evolutionary Models of Information Transmission,2.http://cfpm.org/jom-emit/1998/vol2/rose_n.html

Opie, I. and Opie, P. (eds.) (1997) The Oxford Dictionary of Nursery Rhymes. Oxford; New York : Oxford University Press.

Back to Issue 2 Volume 6