In a famous quip, Hilary Putnam once stated that "meanings just ain't in the head". It seems that the field of memetics and memetic theory is developing a couple of deep divisions: one between those who think of memes as genes and those who think of them as germs and the other between those who think memes are in the head, and those who think they ain't. In this essay I shall deal with these questions, but not attempt to further define what I think a meme is, which I have already done (Wilkins 1998a), based upon the cybernetic notion of an evolutionary gene of G. C. Williams (1966) and the Shannon-Weaver concept of a message:
A meme is the least unit of sociocultural information relative to a selection process that has favourable or unfavourable selection bias that exceeds its endogenous tendency to change.
Derek Gatherer (1998) has written a provocative and interesting paper attacking the "thought contagion" notion of memetics and claims that no population model of memes is possible. He follows Benzon's (1996) earlier approach of locating memes in the overt behaviour and artifacts of culture. He and Benzon are both anti-head memeticists, and I suspect that they prefer to see memes as Mendelian genes, observed by their effects. His target is Aaron Lynch's (1998) pro-germs and pro-head view (overlooking here some of Lynch's subtleties), derived from the later views of Dawkins (1982, 1993).
Gatherer notices a shift in the meme definitions provided by Richard Dawkins, and calls them Dawkins A, and Dawkins B:
Dawkins A: `..a unit of cultural transmission, or a unit of imitation.' (Dawkins 1976, p.206); `Examples of memes are tunes, ideas, catch-phrases, clothes fashions, ways of making pots or of building arches' (Dawkins 1976, p.206); `Popular songs and stiletto heels are examples. Others, such as the Jewish religious laws...' (Dawkins 1976, p.209); `Perhaps we could regard an organised church, with its architecture, rituals, laws, music, art and written tradition, as a co-adapted stable set of mutually-assisting memes.' (Dawkins 1976, p.212); `Memes for blind faith have their own ruthless ways of propagating themselves.' (Dawkins 1976, p.213).
Dawkins B: (referring to the original Dawkins A definition, above) `..I was insufficiently clear about the distinction between the meme itself, as replicator, on the one hand, and its `phenotypic effects' or `meme products' on the other. A meme should be regarded as a unit of information residing in a brain...... It has a definite structure, realized in whatever physical medium the brain uses for storing information....I would want to regard it as physically residing in the brain.' (Dawkins 1982, p.109); `The phenotypic effects of a meme may be in the form of words, music, visual images, styles of clothes, facial or hand gestures.....' (Dawkins 1982, p.109).
The different approaches I call the "genes" and the "germs" view of memes are based respectively on Dawkins B and Dawkins A. The genes view seeks to understand memes as analogous to the information units of selection and the causes of (in this case, sociocultural) evolution. Memes are units of cultural heredity, just as genes are units of biological heredity on this approach. The germs view seeks to explain the epidemiological spread of memes, like viruses through a population. It does not explicitly concern itself with selection or evolution.
Gatherer rejects Dawkins B in favour of the following definition of a meme, which he is revising from William Benzon's (1996) definition:
Meme: an observable cultural phenomenon, such as a behaviour, artefact or an objective piece of information, which is copied, imitated or learned, and thus may replicate within a cultural system. Objective information includes instructions, norms, rules, institutions and social practices provided they are observable.
While I agree with much of Gatherer's criticism and many of his conclusions, especially of the thought contagion approach to memetics, I am moved to argue against his major claim that memes aren't to be located in the heads of individuals and to assert the possibility and even immediate attainability of a populational memetics. I do this from a philosophical perspective, and philosophers are notoriously poor at prescribing the best routes for science to take, particularly in the area of methodology, so I won't do that. Instead I'll do the traditional philosophical job of armchair criticism, and I expect to be taken in that light. Mostly, I will reject the inference from a specific denial to a general one, the move I call "simply because". My view could be seen as a compatiblism between both the germs/genes division and between the in-head/out-of-head division. Such is the wishy-washyness of being a philosopher.
Objection 1 - Over generalisation. My first objection is that he has overgeneralised. Simply because we can now have no population memetics, or because it fails in Gatherer's aptly chosen example of the Windsor knot, it does not follow that we can not have it in any case. Gatherer wishes to extend the fact that we cannot easily, if at all, identify the knowing of "how to tie a Windsor knot" with the act of "tying a Windsor knot", and we can only say with certainty how many actual Windsor knots there are in a population. Again, just because in some cases we cannot say who "has" the Windsor knot meme, it is not thereby true that we can never know who has memes. Some memes are identifiable, and in quite specific terms. I have instanced the memes of scientific theories (Wilkins 1998a, 1998b) but I also expect that there will be a number of other examples of the attitudinal kind usually found in social surveys. These can be operationally defined by the process of iterative refinement exampled by Grounded Theory in health research (cf Wilkins 1998b).
Objection 2 - Over reliance on instrumentalism. Simply because we cannot assay memes except through behaviour, which is quite true, it does not follow that memes of a mental kind are simply hypothetically postulated entities (or HYPEs). To argue this is to confuse information with its expression, the cause with the effect. It is true that we reason backwards from phenomenon to mechanism in science, but HYPEs have this habit of becoming observable, measurable, entities in their own right. Electrons are a case in point (Hacking 1983). Once the very model of a HYPE, and treated as an instrumental theoretical notion, useful for calculating but not much else, electrons are now routinely manipulated; sprayed, bathed and no doubt washed over objects for our own purposes. Hacking suggested the Higgs Boson as an echt HYPE, but almost as soon as he did, it was observed. The process of natural selection itself has moved from being a hypothetical process to being an observed and repeatably manipulated phenomenon (Bell 1992). So the "simply because" here fails to support the blanket conclusion of the impossibility of mental memes.
However, the engram proposal is, I agree, quite wrong. The relation between neural net structures and dynamics on the one hand and cultural structures and dynamics on the other is many:many, and represents a Kolmogorov mapping between domains of over determination in both directions (figure 1). It is unhelpful to try to identify memes in terms of the neural structures that encodes them, because memes are multiply realised and neural nodes are involved in realising more than one meme, at least most of the time. While a single region of a brain may be involved in the storage and activation of a given concept or word, it does not follow that every brain does it in quite this way. Similar issues apply to genes and traits. While there happens to be a strong conservation of many structural and regulatory genes, such as Hox which is implicated in the expression of appendages, the variety of appendages so generated are very different. And not all appendages involve Hox or its orthologs. The "leg" genes are a heterogeneous set across lineages, what in taxonomy is called a polythetic set.
Figure 1. A Kolmogorov mapping of memes
Objection 3 - Memes are units of information. Although I agree with Gatherer that memes are not beliefs (in philosophy this is known as the distinction between propositions and doxastic attitudes to propositions), I disagree with him that memes are not sequences of information (and with Lynch 1998 that memes are single units; it is naive in the extreme to think that propositions are simple entities, in defiance of all logical and linguistic analysis since Frege). Memes are meaningful, semantic, information transmitted as a message from a source to a receiver. Note that semantic information is a notion that differs from the Shannon-Weaver concept of information, which is a signal-relative notion that may, or may not, have semantic content. The most "information-bearing" signal is one that is entirely random under Shannon-Weaver, because it takes a minimum message equal to the actual message to transmit it. Any semantic message is still a Shannon-Weaver message, but not all Shannon-Weaver messages are semantic messages. It follows that memes are messages, of the semantic variety.
Any message may, relative to an encoding protocol, be represented as a bit string, and so once we have such a conventional protocol (a formal "language" of the minimum kind) we can:
I shall say more on the structure of memetic information below. However, it must be noted that the observation of "objective" behaviours suffers exactly the same problems as semantic memes: we have no easy way to identify them, classify and relate them (ie, distinguish homologies from analogies) and measure them purely on the external, observable properties and structures alone. This is because we need to have semantic information about them before we can do any of these things. Again, I shall say more about this shortly.
Gatherer rightly makes much of the notion of variations in memes in response to Lynch's naive conception of "mnemons" (which, in the end, are just the "propositions" of late medieval Platonic philosophy). This point is well-made: anything that can be expressed as a sentence cannot be a unitary semantic entity. The sentence, "There is(are) <n> god(s)" is a complex structure with place holders for variables, as is the sentence, "Napoleon died in <year date>", and an infinite number of sentence tokens can be generated from the primitive relations (predicates) used to construct it. Using such primitive operators of the conventional natural language as "... died in ...", " some definite number of ...", along with the natural quantifiers "there exists" and cognates, and proper names and the natural numbers, etc, one can regenerate a literal infinity of statements from a very few, indeed quite small, number of stored and retrievable linguistic structures. If Chomsky's Transformational Grammar or something like it has any validity as a schematic for the neurology of language (cf Pinker 1994) then there is no real objection to memes being in the head in the sense that we can store a certain number of these constructions, and yet generate as many as needed on demand. Gatherer's objection would hold if we had to store all possible memes instead of only these variables and connectives. Instead, these semantic entities and their relations can be represented as a hypercube of however many dimensions there are particles. A given meme is therefore a well-formed coordinate in that resulting space. We can store some number of coordinates without having to store all of them, and yet be able to traverse this semantic space if it becomes important.
It's not up to philosophers to conjecture about mechanisms - that should be left to working researchers like Gatherer who use experiment and empirical data collection. However, I can make one suggestion: where memes might be located. If memes are messages, then they are entirely relative to considerations of encoding protocols, language (and logic) forms, and evaluative attitudes that determine recognition and responses. The meme therefore consists in the relationship between the neurological states and resulting behaviours of individuals and these contextual abstractions. This is illustrated in figure 2.
Figure 2. Culturally significant transmits as a subset of an individual's outputs
The output of the memes in an individual might be behaviours (Windsor knot tying), or artifacts (Windsor knots), or both. But a one-year old (let's call him Charlie) playing with Daddy's ties might randomly create the topology we call a Windsor knot, exhibiting neither the meme in thought, behaviour or in artifact. What counts here is the mental state - I hesitate to call it the intention, but at any rate the knowledge and understanding - its openness to inspection, the output, and the context... and above all the transmissibility of the result to others.
We have two entirely different issues here. What kind of entity, on the first hand,is a meme? On the other hand, what method and heuristic can we use to locate, measure, differentiate and model memes? Gatherer wishes to follow Benzon and locate them in physical structures, not of neurons and glial cells, but of artifacts and the actual behaviour patterns of cultures. Only these are measurable and objective. We can see a Windsor knot, and we can watch it being tied.
I have two objections to the Benzon-Gatherer "External Meme" thesis, but in so doing I do not wish to reassert the truth of Dawkins B - the thesis that memes are just units of information residing in brains. They are, I aver, informational structures, and they do reside in brains a lot of the time, but as I have argued before (Wilkins 1998a), they also often exist in the cultural relations between brains, and may not even exist in any brains at all at a time or ever.
Moreover, I expect that what goes on inside brains will sometimes be, not memes, but meme-like; call them "memoids" after Dawkins' term for design-like objects, "designoids" (Dawkins 1996). They will resemble memes in that they appear to be semantic structures encoded in neural networks, but their selection within the brain will often be hit-or-miss and inconsistent, despite Dennett's thesis of a strict internal selection (Dennett 1991, chapter 8). What counts isn't so much the selection process inside the head as the one outside, and that requires transmission.
The first objection to the exclusive External Meme thesis is again a "simply because" objection. Just because we can locate, measure, differentiate and model something is not reason to think that what we (now, or even in principle) cannot is nonexistent or nonsensical. This was the deep error of Skinnerian behaviourism, a holdover of old positivism, and it made of an epistemic limitation an ontological virtue. Although I am sure that neither Gatherer nor Benzon is a positivist, this is still an all-too-common error in science. While we may narrow down the search space of alternative HYPEs in this way, and Occam's Razor may seem to give weight to that approach, the history of science has shown (not least with the notion of a "gene") that it often leads to error and failure of research. We had better not kid ourselves that if we can't see it, it can't affect us, like kids hiding from monsters under the blanket at night.
My second objection is related also to the faults of Skinnerian behaviourism - the denial of what we are pretty sure is real and significant. Skinner denied the reality of psychological states - the External Meme thesis seems to be predicated on the denial of concepts. Questions raised by Gatherer of what in philosophy is called the K-K principle (if we know, we must know that we know, ad infinitum) aside, surely if I know how to tie a Windsor knot, some causally relevant structure exists inside my brain? And it must be one that satisfies the relevant semantic relation to the culture that defines Windsor knots and which therefore makes it "about" Windsor knots.
Now I am not here trying to revivify the extreme reductionism of engrams, the one meme-one engram theory, if ever it was seriously proposed. Memes are, in my perspective, multiply realised and multiply instantiated by neural net structures (as neural nets are wont to do). The view proposed by Delius (1991) that a meme will have a single locatable and observable structure ("a constellation of activated neuronal synapses") is not viable. In philosophical jargon, memes are not supervenient properties, for a property that supervenes on a physical system must be identical if the physical system is. The same meme might be realised in a number of physical systems, consistently with supervenience, but identical or very closely similar neural nets might encode quite distinct memes. This is because memes are relations in semantic space, not neurological space.
Let me explain by returning to Charlie's Windsor knot accident. If his knot matches exactly that of the very metre of Windsor knots, Prince Phillip's butler, when none of the "normal" intentions or concepts are possible, then the "Windsor knot" that results is only an accident. We would not call it a cultural artifact if the wind "tied" a bullrush in that knot, and we wouldn't here, except in an honorary sense. There is a continuum of intentionality and "aboutness" from the bullrush to Prince Phillip's butler. There is no absolute point at which mere topology makes that meme, and the resemblance the bullrush has for an observer is merely subjective - it looks like one so the observer, who has learned about Windsor knots, "reads" that topology that way.
But what makes that knot a Windsor knot is the fact that it results from the transmission from one agent to another (with, I suppose, some Urbutler in the House of Windsor as the Last Common Ancestor) of the "how-to" of Windsor knot tying. The more-or-less formal algorithm is what gets transmitted. This constitutes a message, an abstract array of symbols that are interpreted in the context of a consensual, or conventional, language and set of behaviours and standards. So, we cannot really say that it is the physical knot, nor even the tying behaviour, that is the meme, at least not without the interpretive context and semantic structures of a culture.
So, can we make memetics an operational science at all? I believe we can, and the reason is that in many cases we already have just the semantic apparatus needed to identify memes, and that resolving memes is a matter of resolving them enough to identify them. Goodenough and Dawkins (1994) had no trouble identifying instances of the St Jude Chain letter, even across cultural divides, and with a suitable knowledge, say, of Spanish, one could identify instances of it across languages. For most situations being investigated, seeing a Windsor knot being tied is evidence enough that there is a Windsor knot meme about. Eccentric examples of multiple personalities and other exotic exemplars are not enough to shake our confidence that there are memes, that usually people have them, and that we can count them well enough to model them in a populational, statistical, manner.
A general observation: Gatherer rightly worries about the academic respectability of memetics, just as serious researchers have worried about, say, the Gaia hypothesis - which is only now coming into serious contention, even by such respectable academics as William Hamilton (Hamilton and Lenton 1998) (of hamiltonian inclusive fitness) - or complexity theory. The debasement of memetics by quick and easy metaphors and popularised science to serve metaphysical agendas and political ideologies, with which we are all too familiar, is just the latest instance of serious evolutionary theory being perverted in that way, beginning with Spencer and the Edinburgh radicals of the 1840s, through social "Darwinism", "cultural evolution" theories, eugenics positive and negative, and so forth.
I'm of the view that the term memetics is now a liability, at least until it can be rehabilitated with some actual research instead of conjecture and anecdote. I tend to think of this whole field as cultural Darwinism, stressing the essentially amoral and nonevaluative dynamics of culture, and drawing no conclusions about what should be transmitted, instead seeking to know why things are transmitted, and adopting the generally pluralistic nature of Darwinism.
In Darwinism proper not everything must be explained in terms of its adaptation - Sewall Wright is centrally installed in modern synthetic Darwinism. Memetics is analogous to genetics not selection, and it is a subdiscipline of cultural Darwinism dealing with the evidence for causal factors that create the behaviours and artifacts we see and their rates of transmission.
As soon as we use a metaphor like "thought contagion" we are insert all kinds of socially-weighted evaluations associated with the terms "pathogen", "parasite", "disease" and the like, none of which play a useful role in a Darwinian model of any kind of change. The arbitrary difference between a predator and a parasite is the rate at which it kills its host, but we tend to think of parasites as, well, parasitic and therefore nasty, and predators as having a more noble ecological role. But in the end there is only selection and drift, and I refer readers to Paul Ewald's excellent treatment of this topic (Ewald 1994) for a fuller discussion.
Bill Benzon and Derek Gatherer have engaged me in electronic conversation on this topic over a period of time. Benzon was good enough to send me a selection of his papers, and Prof Juan Delius a copy of his paper, which makes my present disagreement with them an act of ingratitude. Despite my conversations with them and many others, including Mark Mills, for which I am continually grateful, nobody is responsible for the errors contained herein but me.
Bell, G. (1997). Selection: The Mechanism of Evolution. London: Chapman and Hall.
Benzon, W. (1996). Culture as an evolutionary arena. Journal of Social and Evolutionary Systems, 19:321-362. http://www.newsavanna.com/wlb/CE/Arena/Arena00.shtml
Dawkins, R. (1976). The Selfish Gene. Paperback edition 1978. Granada Paladin, London
Dawkins, R. (1982). The Extended Phenotype. Paperback edition 1983. Oxford University Press, Oxford UK.
Dawkins, R. (1993). Viruses of the mind. In: Dennett and his Critics. Dahlbom, B. (Ed.) Blackwell, 13-27.
Dawkins, R. (1996). Climbing Mount Improbable. NY: Viking.
Delius, J.D. (1991). The nature of culture. In: The Tinbergen Legacy, MS Dawkins, TR Halliday and R Dawkins (eds.). London: Chapman and Hall.
Dennett, D.C. (1991). Consciousness Explained. NY: Little Brown.
Ewald, P. (1994). Evolution of Infectious Disease. Cambridge: Cambridge University Press.
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
Goodenough, O.R. and R. Dawkins (1994). The `St Jude' mind virus. Nature, 371:23-24.
Hacking, I. (1983). Representing and Intervening. Cambridge: Cambridge University Press.
Hamilton, W.D. and Lenton T.M. (1998). Spora and Gaia: How microbes fly with their clouds. Ethology Ecology and Evolution, 10:1-16.
Lynch, A. (1998) Units, Events and dynamics in memetic evolution. Journal of Memetics - Evolutionary Models of Information Transmission, 2. http://cfpm.org/jom-emit/1998/vol2/lynch_a.html
Pinker, S. (1994). The Language Instinct. London: Penguin.
Wilkins, J. S. (1998a). What's in a meme? Reflections from the perspective of the history and philosophy of evolutionary biology. Journal of Memetics - Evolutionary Models of Information Transmission, 2:2-33. http://cfpm.org/jom-emit/1998/vol2/wilkins_js.html
Wilkins, J.S. (1998b). The evolutionary structure of scientific theories. Biology and Philosophy, 13: 479-504. http://www.wehi.edu.au/~wilkins/theories/evoltheory.html
Williams, G.C. (1966) Adaptation and Natural Selection: A Critique of Some Current Evolutionary Thought. Princeton NJ: Princeton University Press.
| Back to Issue 1 Volume 3 | Back to Commentaries on Gatherer|