1. Introduction

Edmonds (2002) has set down three challenges to the memetics community: a) a conclusive case-study; b) a theory for when memetic models are appropriate; and c) simulation of the emergence of a memetic process. More recently, Duthie (2003) has called for a greater focus on practical examples, and a retreat from definitional debates. In doing so, Duthie echoes similar calls from Hull (2000), and the "Symposium on Memetics" panel (Hales 1998).

I should like to reply to these calls, essentially for two purposes: firstly to demonstrate that empirical case-studies can be carried out within the behaviour/artefact formulation of memetics, but also secondly to illustrate how difficult it is to steer clear of definitional debate. In the space of a letter, it is neither possible nor desirable to perform a thorough analysis, and what follows is, like the two case studies in Duthie, merely intended to serve as an illustration. It will be seen, however, that the issue of definition of a meme is crucial to the design and interpretation of the analysis.

In passing, it ought to be mentioned that Duthie's (2003) reference to the debate between "those who advocate the contagion-like or viral metaphor and those who prefer the gene metaphor" (quoting Aunger 2002), having resulted in "both groups appearing to claim that the other is retarding progress in memetics", is actually not strictly correct. The virus versus gene metaphor debate is in fact far older than the controversies of the late 90s, and can be seen most acutely in a comparison of the work of Cavalli-Sforza & Feldman (1981) against that of Cullen (2000). The "retarding progress" debate is much more recent and is really between fact-orientated/quantitative/behavioural approaches and ad-hoc-theory-driven/pseudo-psychological approaches within the "viral" school, broadly defined. It is about things like analysing data quantitatively as opposed to merely musing on it. In that respect, I am as firmly in favour of detailed case studies as Duthie (2003), as correctly suggested by Edmonds (2002).

2. The Case Study

2. 1 Background

Since the early 1960s, the Eurovision Song Contest has been an annual competition between member states of the European Broadcasting Union. The modern form of the contest originated in 1974, when the present judging system was adopted, and works as follows. Each participating country sends a representative to the competition, held in the territory of the current holder of the trophy. The songs are performed and a jury in each of the participant countries votes for its favourite song. The rules of the contest specify that the juries should judge the quality of the song and not of the performance. Juries are not permitted to vote for the song of their own country. Since the mid-90s, some countries have dispensed with juries and opted for a system of national telephone voting, effectively turning the entire interested population of the country into jury members. In the words of the well-known BBC disc jockey John Peel, "the reason the music industry is not interested in the Eurovision Song Contest, is that it is impossible to fix the result". That is certainly true. Nevertheless, there have been persistent suggestions that some countries appear to be regularly voting for each other in a kind of reciprocal altruism, and also that other political considerations have come into play - such as BBC presenter Terry Wogan's analysis of the 2003 contest, that the United Kingdom entry had suffered from "post-Iraq backlash".

The precise mechanics of voting are as follows. The jury awards its favourite song 12 points, the second favourite 10 points, the third 8 points, and then a further 7 songs in descending order from 7 to 1 points. During the period under consideration, the number of countries participating in the contest varied from 18 to 25. Therefore the probability of receiving no points at all from any particular other country in any particular year varies from 0.41 to 0.59. This system was used for the last time in 2003, and from this year will be replaced by a novel, more tournament-like, mechanism involving an eliminatory round. It is not clear what effect this will have on voting patterns, but it is probably partly intended to make collusive voting more difficult.

2. 2 Method and Results

Scores for all contests between 1974 and 2002 inclusive, were downloaded from http://www.kolumbus.fi/jarpen/ with permission. The average scores for each pairwise combination of countries were calculated over this period, in the form of three statistics: a) the average score awarded from country 1 to country 2; b) the average score awarded from country 2 back to country 1; and c) the average total shared between country 1 and country 2. A simulation of 100 contests was carried out, with completely random voting. A Perl script for the random contest is available on request. The statistics were calculated for each pairwise combination in the 100 random contests, and the resulting means and standard deviations used to derive z-scores for the statistics for the real contests. The country pairs deviating most from random voting are shown in Table 1.

Table 1: Pairs of countries deviating the most from a random voting pattern. Pair 1 and Pair 2: the names of the countries (CY Cyprus, GR Greece, LA Latvia, EE Estonia, SE Sweden, DK Denmark, ML Malta, SK Slovakia, HR Croatia, GB Great Britain, MC Monaco, RO Romania, MK Macedonia, SL Slovenia, RU Russian Federation, CH Switzerland, YU Yugoslavia, IS Iceland, IE Ireland). 1 to 2: average votes from pair 1 to pair 2. 2 to 1: average votes from pair 2 to pair 1. Years: total number of years both countries have appeared together in the contest. Total: average sum of reciprocal votes for both countries. Recip z: z-score for total. 1-2 z: z-score for 1 to 2. 2-1 z: z-score for 2 to 1.

A z-score of 1.65 is significant at the 5% level, a score of 2.33 at the 1% level. All of the scores in Table 1 are therefore highly statistically significant. Some of these scoring patterns are distinctly asymmetrical. For instance, although Malta and Slovakia both have significantly elevated rates of reciprocal voting over the 3 years they have both appeared in the contest (1994, 1996 and 1998), Slovakia has tended to give double the number of votes to Malta as it receives in return. However, it should be noted that Malta performed quite well overall in those three years, (5th, 10th and 3rd respectively), so some of this may be attributed to genuine vote-catching qualities of the relevant Maltese songs.

The 16 significant pairwise voting relationships in Table 1 can be expressed graphically by means of a Venn diagram (Figure 1). This reveals the existence of two large power blocs and two smaller axes. The more compact of the two largest blocs is here termed the "Viking Empire", owing to its uncanny resemblance to the area under Norse control in the early Middle Ages, and consists of Scandinavian and Baltic States, along with Ireland. The second of the large blocs is more diffuse and is really an intersection of two axes, a "Maltese Cross" composed of Malta, Slovakia, Luxemburg and Croatia, intersecting with a "Warsaw Pact" composed of states of the former Communist east. The two peripheral axes are a Britain-Switzerland-Monaco triad (financial centres?), and a Greece-Cyprus-Yugoslav axis, now reduced to a Greece-Cyprus dyad following Yugoslavia’s last appearance in 1992 and the subsequent drift of Croatia into the connecting position between the "Warsaw Pact" and the "Maltese Cross".

Such classification is merely whimsical, but the basic fact remains that statistically significant reciprocal voting does tend to have a geographically non-random distribution, and in four of the five clusters, to be suggestive of older historical alliances (the "Maltese Cross" being the only exception). The question now is: can we apply memetics to this situation and if so, how?

3. Discussion

One thing is clear. There is no central co-ordination of voting in the Eurovision Song Contest. Reciprocal voting is something done by individuals, and since the introduction of telephone voting, it is something which is often done on a mass scale. It requires a large number of people to act independently of each other in a predictable manner in the expectation that other people, often hundreds of miles away and with whom they have never had any personal contact, will act in another predictable manner. At first glance, this appears to be potentially an imitative phenomenon. A person, having observed a contest, and having seen what voting patterns occur, may resolve to cast a vote in the next contest in the same way. However, if this is the case, we ought to see the gradual increase of collusive voting as the pattern spreads through the voting pools of the two countries concerned. The historical voting pattern between the leading pair in Table 1, Greece and Cyprus, is shown in Figure 2.

Figure 2: Voting pattern from Greece to Cyprus, and Cyprus to Greece within the period 1981 to 2002, encompassing all the years of the modern voting system era (post-1975) when both countries were represented (15 occasions). Mutual maximum marks have been awarded on 5 occasions, and unilateral maximum marks on a further 8 occasions.

Figure 2 demonstrates that there was initially no predictable pattern of voting. Indeed Greece gave no points to Cyprus in 1983. The pattern starts to establish itself after the mutual 8 points in 1985, leading to mutual maximum points in 1987. The pattern breaks down a little in the late 80s to early 90s, but by the late 90s mutual maximums seem virtually guaranteed. A slightly more complex pattern is seen when considering Sweden and Denmark in Figure 3. On the first three occasions, no points were awarded by either side, but from 1981 Denmark took the lead in awarding increasingly high points to Sweden. 1990 saw a sudden withdrawal of points by Denmark, but the relationship recovered immediately. Figure 4 shows the relationship between Iceland and Sweden. Initial high marks from Iceland produced little or no Swedish response, and the Icelandic vote gradually declined in the late 80s. However, the position then seems to have reversed with Sweden taking the lead in high voting for Iceland.

Figure 3: Voting pattern from Sweden to Denmark, and Denmark to Sweden within the period 1978 to 2002, encompassing all the years of the modern voting system era (post-1975) when both countries were represented (22 occasions). Mutual maximum marks have been awarded on 2 occasions, and unilateral maximum marks on a further 3 occasions.

Figure 4: Voting pattern from Sweden to Iceland, and Iceland to Sweden, within the period 1986 to 2001, encompassing all the years of the modern voting system era (post-1975) when both countries were represented (15 occasions).

These three examples demonstrate that collusive voting patterns are not steady, but wax and wane, often in visible response to the previous year’s vote.

4. Conclusion

In summary, it can be seen that:

1. the Eurovision Song Contest is a purely cultural phenomenon - there is no biological selective advantage associated with winning or losing.
2. Nevertheless, behaviour has evolved to share votes in mutually beneficial partnerships, often with one country taking a lead in courting another..
3. Many more of these have sprung up in recent years (see Table 1), as new countries join the contest, adopting almost immediately the process that has evolved gradually among the long-standing members.

We have here several of the component parts of a memetic evolutionary system: confidence that biological evolution is not at work, evidence of a contagious spread of a cultural trait, and finally adaptation of a different evolutionary system (i.e. the rules of the contest) to cope with changes in the system under consideration (the two-tier system now being introduced from 2004 is partly designed, one suspects, to eliminate some of the evolved behaviour discussed here). This is not exactly the simulation of the emergence of a memetic process, as requested by Edmonds, but it is the observation of the birth, life and (from next year) death of a memetic ecosystem. This may not be a conclusive case study - there are still far too many unanswered questions, for instance: why did collusion evolve where it did and not elsewhere?; was the initial stimulus intentional or a co-incidental consonance of votes that was in some way ‘interpreted’ as an overture to collusion?; what role did commentators’ complaints about collusion have in spreading/creating the phenomenon? However, I hope it satisfies about one and a half of Edmonds’ three criteria.

Regarding Duthie’s comments, I believe that the issue of definitional debates cannot simply be ignored. It is difficult to see how the above example could have been produced using the ‘thought contagionist’ framework. I would have been required to identify beliefs and then derive evolutionary epidemiological scenarios for their spread to other individuals, identify recruitment, retention and drop-out events, express them in ‘the calculus of mnemon instantiations’ and then somehow fit them into a complicated set of differential equations. Applying this process to the above scenario would seem to be impossibly contrived. In any case, there has been no thought contagionist analysis case that has actually gone through this entire process from beginning to end, nor in fact has there been one that has quantitatively handled any data. That paradigm has an as yet unbridged chasm between its symbolic theory (the calculus of mnemon instantiations and the differential equations), and its actual practice, which seems to be veering increasingly towards journalistic commentary on things like financial market and wars.

Duthie is right, the more practical case studies the better. But to achieve those case studies, we need to concentrate on memes as behaviours and artefacts. That way, the experimental material is right before us, forks, paperclips and votes, among many other things. In the parlance of Eurovision, other methods can expect "null points".

References

Aunger R (2002) The Electric Meme: A New Theory of How We Think. New York, NY: The Free Press.

Cavalli-Sforza LL & Feldman MW (1981) Cultural Transmission and Evolution. Princeton: Princeton University Press.

Cullen BS (2000) Contagious ideas : on evolution, culture, archaeology, and Cultural Virus Theory / Ben Sandford Cullen ; collected writings edited by James Steele, Richard Cullen and Christopher Chippindale. Oxford: Oxbow Books.

Duthie AB (2003). A Letter on: The Fork and the Paperclip: A Memetic Perspective.Journal of Memetics - Evolutionary Models of Information Transmission, 8. http://cfpm.org/jom-emit/2004/vol8/duthie_ab.html

Edmonds B (2002) Three Challenges for the Survival of Memetics. Journal of Memetics - Evolutionary Models of Information Transmission, 6. http://cfpm.org/jom-emit/2002/vol6/edmonds_b_letter.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

Hales D (1998) Report on the Panel Discussion that occurred at the Symposium on Memetics. Journal of Memetics - Evolutionary Models of Information Transmission 2. http://cfpm.org/jom-emit/1998/vol2/panel_discussion.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. http://cfpm.org/jom-emit/1998/vol2/gatherer_d.html

Back to Issue 2 Volume 8