Case Study Research Purpose Meme
Here we firstly formally define the terminology we use in the rest of the paper. We then provide a collection of evidence sustaining our thesis that successful memes are the ones which tend to be dissimilar from all other memes. Finally, we develop a measure to evaluate the degree of uniqueness of a meme and we test the amount of variation it can explain in the success of memes, measured with the number of votes they can attract.
In this work, we use the following definition of a meme:
Definition 1 (Meme)
A meme is a cultural unit defined by an atomic concept. A meme is identified by a name and a template and it can be implemented in different forms.
Following this definition, one meme could be a class of jokes about anything related to social clumsiness: it can be used to describe a social situation where a person misbehaved or she did not know how to properly react. This is an actual meme and, among all possible names, people on the Internet decided to call this meme “Socially Awkward Penguin”. Our meme definition requires us to define what a template and what a meme implementation are.
Definition 2 (Meme Template)
A meme template is a piece of information that is used as signature of the meme itself and it identifies it in an unambiguous way.
In this definition, it does not matter what is the piece of information we use to identify the meme. In some works37 it can be a hashtag in Twitter. For this paper, we decide to focus on those memes whose template is a picture. The established “Socially Awkward Penguin” template is a picture of a left-facing penguin in a blue field. Finally, we can define what a meme implementation is:
Definition 3 (Meme Implementation)
A meme implementation is any human expression that puts together the meme template and some additional information, whose meaning is semantically related to the meme concept.
In our case, a meme implementation is a short text superimposed to the meme template.
A word of caution is needed when we have to solve the problem of setting two memes apart. Formally, two memes are different when the atomic concept they carry is different. However, isolating the atomic concept requires to reach an agreement between all users of the meme. The task is not easy: just as in the case of genes41, the boundaries between two memes are fuzzy. There are examples of memes which are being used interchangeably, with users often complaining that the two memes are the same42. Since we are focusing on a single data source, we can use an approximation to solve this issue: two memes are set apart when they use different names and templates.
The temporal information allows us to study how a meme's success evolves over time. Later in the paper, we focus on a specific event in a meme's lifetime. To do so, we need to introduce two concepts: offspring and parent. Our assumption is that cultural products are rarely completely original. Therefore, every meme, to some extent, is derived from at least one other pre-existing meme. The pre-existing meme is called a parent. The derived meme is the parent's offspring. Hereafter, we use the following formal definitions of offspring and parent.
Definition 4 (Offspring)
is the set of all memes. is the set of all memes that had an implementation at a timestep x ≤ t. Every meme (the set of the “original” memes present at timestep 0) is an offspring and the timestep t > 0 of its first implementation will be referred to as its appearance timestep.
For example, if a meme o has no implementation before February 2012 (), it means that o is an offspring with appearance time t = 4.
Definition 5 (Parent)
Consider an offspring o whose appearance timestep is t. The meme arg , i.e. the meme p for which the similarity function st (the similarity function that considers only information generated at all timesteps t′ < t) is maximum, is the parent of o.
In the definition, the meme similarity function s is the one defined in the Methods section. Examples and explanations about the concepts defined in this section are provided in the Supplementary Material.
We now define the shape of the meme similarity space. The meme similarity space is based on the meme similarity function s as defined in Methods. We calculate s for all meme pairs, resulting in the symmetric matrix S. Then, we create a network visualization of S. To increase readability, we impose the following constraints on the structure of the resulting graph: (i) it has to include all memes; (ii) it has to be composed by a single component; (iii) its edge density should not exceed the typical average degree of real world complex networks. The implemented procedure is a standard approach to visualize a matrix through a graph43. An analysis of the robustness of the network map and of the stability of the similarity measure used in the paper is provided in the Supplementary Material.
The result is depicted in Fig. 1 (left). Large successful memes tend to scatter in the outside of the network layout. There is a negative relationship between success and network centrality. We calculate such correlation and we report the results in Tab. I. Different network centrality measures and the number of instances and of votes of a meme per timestep present significant negative correlations.
The result could be an artefact of our network visualization procedure. The network is in fact a visualization with the purpose of illustrating the point, not proving it. To prove it, we integrate the network visualization with the heat maps in Fig. 1 (right). The heat maps highlight that there is a highly unexpected number of memes with low average similarity and high popularity (top left of the map). On the other hand, there are no memes with high average similarity and high popularity (top right of the map).
Our interpretation of this collection of facts is the suggestion that success eschews similarity. We now explore what are possible mechanisms explaining this anti-correlation.
In biological systems, successful genes are more likely to be passed to offspring and to mutate. In cultural systems, original ideas are likely to generate new ones. For these reasons, success and similarity are the two ingredients we focus on to study the parenthood odds of a meme.
For each meme m we know if it had offspring or not in the observation period, thus making parenthood a binary variable. We explain parenthood using a logistic regression. The two predictors are the average number of votes per timestep and the average similarity of the meme with all other memes. For parents, both measures have been calculated over the time span that precedes the parenthood event, because the appearances of offspring would increase their average similarity, thus invalidating the analysis. For non parents, we considered the entire lifespan of the meme.
In our estimated logit model, both the number of votes and the average similarity are significant, with p < .05 and p < .001 respectively. The coefficient of similarity is much stronger than the one of votes, being equal to −15.9 and 1.8 × 10−4 respectively. This means that each .1 increment in the average similarity score of a meme decreases by around five times its odds of being a parent, while to achieve a 20% increment in parenthood odds a meme must have 1, 000 more votes per timestep, which is unrealistic. We can conclude that the success of a meme does not contribute much to its odds of having an offspring. What matters most is that the meme has to have a low degree of similarity with the other memes.
The fact that memes are on average similar to each other and that the distribution of similarities does not change over time could reduce this discovery to circular reasoning. In other words, the results of the regression could be explained by the fact that the appearance of the most similar meme for a meme appears at a random time, thus parent memes are the ones for which this happens later, for random reasons. We disprove this objection in the Supplementary Material.
In Fig. 2 (Left) we show three boxplots, showing the distributions of average votes per timestep for different classes of memes. The first box, labeled as “NP” ( = “Non Parents”), is the distribution for non-parent memes. The second box, labeled as “BP” ( = “Before Parenthood”), is the distribution for parent memes in all timesteps before the appearance of the offspring. Consistently with the results of the logistic regression, parents before parenthood have roughly an amount of votes comparable with non parents. In the figure, the two distributions substantially overlap. The third box, labeled as “AP” ( = “After Parenthood”), is the distribution for parent memes in all timesteps after the appearance of the offspring. We can see that there is a noticeable decrease for all percentiles. The figure suggests that the appearance of an offspring has a negative effect on the success of its parent. This effect is quite strong (notice that the y axis is in logarithmic scale), and the loss is of one order of magnitude. A likely explanation is that the new offspring is very similar to its parent and therefore it “fishes in the same pond”: users rarely use both.
One could reject this explanation by arguing that memes are more popular at the beginning of their life cycle and that is when they are more likely to generate offspring too. This would mean that what is responsible for the decay in votes of a parent is not parenthood, but age. However, we observe that the peak in parenthood odds and in popularity appears at different ages, thus rejecting the hypothesis (see Supplementary Material).
Having an offspring seems to cause a loss in votes for the parent. We can quantify this affirmation by exploring the relationship between the average votes per timestep of a meme before parenthood and how much it lost after the parenthood. This relationship is depicted in Fig. 2 (Right). The figure suggests that the more popular a meme was before becoming a parent, the more popularity it lost after parenthood. Some very unpopular memes have actually gained popularity. However, all memes that had at least 20 votes per timestep lost popularity, with some of them reducing to 1% (in the plot 10−2) of their original popularity.
The appearance of an offspring has a negative effect on the popularity of the parent. We now address the question: what best explains the popularity of the offspring itself? A first answer may come by looking at the popularity of the parent. It is possible that the parent is passing to the offspring those characteristics that made it successful (or not). To test this hypothesis, we classified offspring memes in three equipopulated classes (i.e. each class contains the same number of offspring) according to the average number of votes per timestep of their parents before parenthood. So a parent meme can have either high, medium or low popularity. We then looked at the number of votes the offspring of these parents got in the timestep they appeared. This relationship is depicted in Fig. 3 (Left), where we have a boxplot indicating the offspring's votes on the logarithmic y axis. From the figure, we see that the popularity class of the parent is not able to explain a lot about the popularity of the offspring. There is a positive effect, but it does not appear significant. The median number of votes goes from 9 for the “low popular parent” offspring to 14 for the “high popular parent” offspring. The largest difference appears to be in the outliers from the 90th percentile on.
We test a second hypothesis. Given the shown anti-correlation between similarity and success, we expect to find significantly higher numbers of votes for the memes which are the most dissimilar from their parents. We explore this relationship with the same boxplot we examined previously. In Fig. 3 (Right), we put the offspring in three equipopulated bins according to how similar they are to their parent. As we can see, now we have significant differences between the three classes (please remember that the y axis has a logarithmic scale). As expected, the offspring that have low similarity with their parents have a median amount of votes equal to the 75th percentile of the “Medium” class. The “Low” similarity class offspring have also a median amount of votes ten times higher than the median amount of votes of the “High” similarity class. We conclude that while the popularity of the parent meme does not necessarily imply anything about the popularity of the offspring, their degree of similarity does, with higher number of votes connected to a lower degree of similarity.
Evaluating Meme Uniqueness
One could be tempted to predict a meme's future popularity by using average similarity and network topological measures shown in Tab. I, given their anti-correlation with the success of memes. However, both measures have some downsides. The average similarity does not control for groups of memes similar to each other but dissimilar from everything else. In this case, some high similarity values may increase the average similarity of memes that are indeed dissimilar to almost any other meme. Network topological measures, on the other hand, are highly dependent on how the network map has been built. If the criterion to select significant edges is not capturing the relevant information, the network map usefulness may be questionable.
We propose a method based on matrix factorization. We aim to evaluate what we define as “Meme Uniqueness” u. In our method, we make use of the entire similarity matrix S: we recursively correct the average similarity of a meme with all other memes' average similarities. In other words, if a meme is very similar only to highly dissimilar memes, then its uniqueness u is still high. First, we calculate the sums of the rows/columns of S. S being symmetric, the sum of row i is equal to the sum of column i: . To correct these sums recursively we need to calculate the average level of similarity of the memes by looking at the average similarity of the memes they are similar to, and then use it to update the average similarity of the original meme, and so forth. This can be expressed as follows: . We then insert kj,N−1 into ki,N obtaining: and rewrite this as: where: We note in the last formulation ki,N is satisfied when ki,N = ki,N−2 and this is equal to a certain constant a. This is the eigenvector associated with the largest eigenvalue, that is equal to one. Since this eigenvector is a vector composed by the same constant, that is the average similarity of the meme, it is not informative. We look, instead, for the eigenvector associated with the second largest eigenvalue. This is the eigenvector associated with the variance in the system, i.e. how fast the meme is converging to the average similarity. The faster a meme converges to the average similarity the less unique it is and thus we can formulate the meme uniqueness as: where is the eigenvector of associated to the second largest eigenvalue, µ is the function calculating its average and σ is the function calculating its standard deviation.
We now have to test if meme uniqueness is a good predictor of meme success. We calculate U for each timestep. We then calculate the Spearman correlation between Un (U calculated at the n-th timestep) with the popularity in number of votes of the memes at the timestep n + 1. We use the Spearman correlation because we are not interested in predicting the actual number of votes but only what meme will be ranked among the top memes. Correlation and p-values are reported in Tab. II. We focused on the timesteps after the ninth, because that is when we have a fixed number of memes. We can see that the correlations are much stronger than the one reported in Tab. I. The correlations are weak, but nevertheless significant, showing that uniqueness carries information about a meme's success and it could be used in a prediction task.
When trying to predict which of the newly born memes will be successful in the future, the meme uniqueness measure can be used to have an educated guess in the absence of any other external information. Without any information about the social network or social media through which the memes are shared, currently one can only do a random guess. In Tab. III we confront the number of correct guesses based on the meme uniqueness measure and on random trials. Again, we stop at timestep #8 because there are no more offspring after that time. While not perfect, the meme uniqueness measure still represents an objective alternative to random guess, yielding better results.
Journal of Memetics - Evolutionary Models of Information Transmission, 5.
Is Suicide Contagious?
A Case Study in Applied Memetics
- 1 - Introduction
- 2 - 2. Memetics and Social Contagion: Two Sides of the Same Coin
- 3 - Suicide Contagion - Infected by a Suicide Meme
- 4 - Priming your Mind with Suicide
- 5 - Testing the Model - Is Suicide Contagious?
- 5.1 - Research Objectives
- 5.2 - Materials and Method
- 5.3 - Participants and Recruitment
- 5.4 - Results
- 5.5 - Discussion
- 6 - But is this Memetics?
- 7 - Conclusion
The phenomenon of suicide contagion is demonstrated experimentally. An interpretation of the results is proposed using an understanding of memetics as contagion psychology informed by selectionist thinking. Using the term `meme' to denote an object of contagion and `contagion' to denote a process of spread by exposure, a selectionist explanation of why certain people might be susceptible to a contagion of suicide is provided. Specifically, it is suggested that people who have become socially isolated and culturally disenfranchised, i.e. those with reduced residual cultural fitness (compromised access to the means of cultural reproduction), might be at particular risk from suicide contagion. Finally, public health policy implications of this memetic understanding of suicide are briefly outlined.
1 IntroductionIn June 1962, the managers of a textile factory in Strongsville, USA, were obliged to close down their factory because of a 'mysterious sickness' that was affecting onsite workers.
On the evening of the closure, a news report described how at least ten women and one man had been admitted to hospital suffering from rashes and severe nausea. The news report also suggested that the cause of the sickness was a poisonous bite from an insect that had arrived in a shipment from England, and had taken up residence in the factory.
Several weeks later, with the plant still closed, a total of 62 workers had sought medical attention, having developed the symptoms of having been bitten by the bug. From initial rashes and nausea, the effects of the poisonous bite appeared to develop into chronic severe weakness punctuated with acute panic attacks.
The US Public Health Service `Communicable Disease Center' with a team of entomologists were called in to investigate this `June Bug', and took the expensive decision to fumigate and vacuum the huge textile plant.
After careful investigation, it was concluded that there had been no dangerous bug at all, English or otherwise. Rather, the episode had been an example of a process that social psychologists call `social contagion', a generic label used to describe the apparent infectious spread of opinions, emotions and behaviour by exposure to similar opinions, emotions or behaviours.
Social contagion usually occurs in contexts of uncertainty and stress, where people make use of information in the actions of people around them to make sense of situations, resolve ambiguity and to inform their own responses (Colligan, Pennebaker and Murphy 1982). Sure enough, when two social psychologists, Kerckhoff and Back (1968), interviewed victims of the mysterious `June Bug' illness they found that those who had been `bitten' had been suffering from undiagnosed stress and alienation.
2 Memetics and Social Contagion: Two Sides of the Same CoinI have argued in a previous paper that for applied memetics, an inclusive and pragmatic working definition of a meme is the object of contagion (Marsden 1998a). This is consistent with the popular understanding of memetics as the study of `infectious' elements of culture. By `infectious', what is meant is the quality of some acts, emotions and opinions, in certain contexts, to spread by exposure rather than by some deliberate attempt to influence (such as coercion or persuasion).
The purpose of this paper is not to engage in an abstract discussion of the merits of this or that conception of a meme or memetics; indeed the purpose of using the meme neologism to describe objects of contagion is wholly pragmatic; to provide a rich, substantive research focus, and get on with doing memetics. Interestingly, contagion psychology shows that a meme, thus conceived, may have no essential properties, and is always a meme in context, thereby defying useful abstract discussion. For example, an act may spread by exposure in one context (and thus be a meme), but spread by coercion in another (and therefore not be a meme). Of course, if the term `meme' is used as a scientific-sounding basket label for all received ideas or socially learned behaviours, then the notion of a meme as an object of contagion, and memetics as a selectionist interpretation of contagion psychology, may seem unnecessarily restrictive. However, the advantage of the more modest conception of memetics is that it is useful within established social scientific models of human psychology, instead of being of interest only from without.
What follows is an example of how applied memetics might usefully proceed based on this understanding of a meme as an object of contagion, focusing on the peculiar phenomenon of suicide contagion.
3 Suicide Contagion - Infected by a Suicide MemeSuicide contagion is said to occur when exposure to suicidal acts appears to trigger copycat suicidal acts. This peculiar phenomenon challenges rational interpretation perhaps as much as the notion of suicide, that is, deliberate self-destruction, itself. Now, sense can be made of many memetic phenomena within an established social learning framework; that we use information vicariously available to us in other people's experiences to inform our own adaptive responses to ambiguous situations. However, suicide contagion does not seem to fit this model insofar as suicide is clearly not adaptive from the suicidal individual's point of view. So bizarre is the phenomenon, that despite the consistent finding of elevated suicide levels following suicide publicity (Marsden 1998b), some researchers find the whole idea of suicide contagion incredible. Nevertheless, strict official guidelines do exist on how suicide should be publicly discussed (e.g. CDC 1994) based on the assumption that individuals contemplating suicide who see suicide or suicidal individuals rewarded (in attention and/or positive evaluation) may infer that suicide is an appropriate response to their distress.
The suicide contagion hypothesis is that exposure to suicide is a suicide risk-factor for people experiencing an unresolved conflict as to whether suicide is an appropriate response to current unresolved distress. Whilst ethical considerations obviously preclude the direct experimental investigation of this hypothesis, one aspect of suicide contagion does lend itself to indirect experimental research. This is the impact of semantic priming.
4 Priming your Mind with SuicidePriming refers to the idea that the interpretation of situations can be involuntarily patterned by recent and frequently experienced events (Fiske and Taylor 1991). To take a trivial example, a habitual walk home after seeing a frightening movie may result in an increased level of `feeling spooked' because the themes evoked by the film remain salient in the mind, thereby producing contagion; the spookiness of the movie influences subsequent interpretations of events. More generally, priming suggests that the ways in which situations are interpreted can be influenced by ideas salient in memory that have been the focus of recent and/or frequent attention. Put simply, we tend use the meanings we have to hand in interpreting situations. Further, because some ideas cue other ideas, this aspect of what is known as social cognition may result in `spreading activation' whereby exposure to ideas cue not only those ideas but also related ideas.
"Thoughts of which one is consciously aware send out radiating activation along associative pathways, thereby activating other related thoughts. In this way, ideas about aggression that are not identical to those observed in the media may be elicited later. In addition, thoughts are linked, along the same sort of associative lines, not only to other thoughts but also to emotional reactions and behavioural tendencies." (Geen and Thomas 1986: 12)For example, in one experiment it was found that people became more aggressive following exposure to images of weapons, as aggressive ideas around weapons were primed in the mind of the exposed to them (Leyens and Parke 1975). The idea is that when activated by becoming the focus of attention, a concept and related concepts in the semantic network of an individual's memory are easier to retrieve, that is, the mind is primed with these concepts and will tend to use them to interpret situations (Higgins 1989, Fiske and Taylor 1991, Berkowitz 1984, Jo and Berkowitz 1994).
Applied to suicide contagion, this interpretation would predict that exposure to suicide should have a short-term impact on how a situation is interpreted as being potentially suicidal. To test the plausibility of this model, the following experiment was conducted as part of a D.Phil research project in 1999 (Marsden 2000).
5 Testing the Model - Is Suicide Contagious?
5.1 Research ObjectivesTo assess the empirical plausibility of the idea of suicide contagion by priming; specifically that mediate exposure over the Internet to the concept of suicide can influence how a situation is subsequently interpreted as being potentially suicidal.
Specifically, the research hypothesis was as follows:
Priming participants with the idea of suicide, by informing them that the topic of research was `young people and suicide', would result in an increased likelihood of interpreting a distressing and ambiguous situation described by a text as suicidal
5.2 Materials and MethodStandard experimental design of controlled exposure to the concept of suicide over the web was followed by data capture in a self-completion web questionnaire. Piloted in April 1999, the research took place in May 1999. Participants were randomly assigned to either an experimental group or a control group, all invited to first read a short text about a distressed student at university, and then indicate on a standard 5-point Likert scale, the likelihood that they thought that the student would commit suicide (1 = not at all likely, 5 = very likely). The only difference in materials used between the two groups was how the research study was introduced to them; in the experimental group the welcome screen introduced the study by saying that it was about young people and suicide, and in the control group about young people and stress. The texts and HTML questionnaires were published to the Internet at using Microsoft FrontPage and a custom CGI script that randomly directed participants from an initial index welcome page to one of the two texts. The responses were captured in a simple CSV (comma separated variable) database sent from the HTML form.
5.3 Participants and RecruitmentParticipants were UK Internet-users over the age of 18 recruited through chain email. Because of the potentially sensitive nature of the research, no incentives were offered, and instructions were given to those recommending individuals that only those over eighteen should be invited. Additionally, attention was called to a hyperlink provided on each page of the web questionnaire to an online support group, (Samaritans - although this was unnamed) to be used if participants found the exercise in any way distressing. Finally, participants were told that they could change their mind about participating, and withdraw from the exercise at any time.
5.4 Results67 UK Internet-users completed the task (36 male, 31 female, average age 22) over a two-week period in May 1999. The research hypotheses predicted that participants primed with the concept of suicide would be more likely to assess an ambiguous situation as more suicidal than participants not primed with the concept of suicide. An inspection of the mean results was consistent with this, with those primed with suicide interpreting the likelihood that a distressed individual would commit suicide at 2.42, and this compared to 1.77 in the control group. A one-way ANOVA (analysis of variance) revealed that this difference was statistically significant at the 0.001 level (F(1,65)= 11.79, p< .001) indicating a strong patterning of the data in a way consistent with expectations of the model.
Figure 1: Suicide is Contagious
5.5 DiscussionThe results of this study were consistent with the idea that exposure to the idea of suicide can prime the concept of suicide in peoples' minds and influence the interpretation of ambiguous situations by facilitating a perception of that situation as suicidal. Although the research hypothesis was confirmed by a strong patterning of the data, thus suggestive of such an influence for the group of participating adult UK Internet-users, the use of non-probability sampling would have precluded any possibility of generalising the results to a more general UK Internet-user population, even if this were considered a legitimate strategy. Further, although the results were continuous with expectations following from the model in terms of interpreting somebody else's situation as suicidal, this does not imply, other than extremely weakly, that participants might also be more likely to interpret their own situations as suicidal following exposure to suicide. Nevertheless, the results of this small study are consistent with such a view.
6 But is this Memetics?The above experiment is a simple example of applied memetics operationalised as the study of infectious acts, opinions and emotions. Now, the critic could argue that this understanding of memetics adds little to the already established field of social contagion research, apart from providing a useful label for the object of contagion. However, from the understanding of memetics proposed here, this is not a problem because memetic research is social contagion research.
Further, it may be possible to gain some leverage from the selectionist heritage that underpins Dawkins' neologism, and use it as a creative stimulus for generating hypotheses on why certain people appear to be susceptible to contagion. By way of example, I would like to conclude by proposing a selectionist understanding of why certain people, and not others, may be susceptible to suicide in general and suicide contagion in particular.
As noted above, suicide is problematic because it is clearly not an adaptive act from the perspective of the suicidal individual. This has led some to suggest that suicide might be enabled by a heritable genetic variation that remains in the gene pool over time because it enables suicide when the residual inclusive fitness of an individual is particularly low (e.g. de Catanzaro 1980, 1981).
Whilst a genetic `enablement' of suicide may well be feasible, it is possible to apply an alternative cultural model using a similar logic. Specifically, if cultural variations, such as those that enable suicide, persist based on their likelihood of being adopted, then the inclusion of suicide in a culture as a meme is only viable over time if it has no systematically deleterious effect on its own reproduction. Now, one way that this could be possible is if those committing suicide were not significant contributors to the propagation of suicide themselves, so that their deaths would not negatively impact on the persistence of suicide. It is this point that provides a selectionist hypothesis for susceptibility to suicide contagion insofar as such susceptibility might be contingent on a reduced residual capacity to pass on culture. In such cases, suicide would not be maladaptive, because the suicidal individual would not be culturally `viable'.
This model leads to an empirical prediction pertaining to those most at risk from suicide contagion. The model would predict that those susceptible to suicide contagion should be those with a reduced residual capacity to spread culture, that is, those who become socially isolated and culturally disenfranchised. Indeed, over and above the possibility that suicide may be used as a strategy for increasing a waning cultural fitness, the fact that cultures are shared means that the suicide of those whose capacity to reproduce their culture has become compromised could actually increase the overall capacity of cultural relatives to pass on shared culture, including any suicide meme. In fact, if an individual actually represented a cost to the overall reproductive potential of the shared culture to which suicide is a part, then suicide could actually have a positive effect on the likelihood that that some `suicide culture' gets reproduced. In this model, differential ownership of the means of cultural reproduction, or simply put, marginality, would be a key variable in susceptibility to suicide contagion, and to suicide in general.
This memetic prediction is borne out by empirical research that has consistently demonstrated that low levels of social cohesion are one of the most significant risk factors in suicide (e.g. Durkheim 1970 , Halbwachs 1978 , Henry and Short 1954, Gibbs and Martin 1964, Maris 1981). In these cases, suicide contagion could be seen as a `fortuitous' cultural mutation that allows a culture to effectively rid itself of parasites that reduce its reproductive potential to maintain itself.
The tentative support for this memetic model might warrant formalisation for future research using, for instance, Hamilton's (1964) model of inclusive fitness in a cultural substrate. Specifically, as long as overall inclusive capacity of the set of culture that describes an individual is not reduced by suicide (or indeed self-sacrifice, as in the case of war), then any suggestion-by-exposure to suicide occurring within a constellation of pro-suicidal circumstances could be more likely to result in suicide. This is because the benefit of suicide to cultural relatives, in terms of enhancing their capacity to reproduce their shared culture multiplied by a degree of cultural relatedness to the suicidal individual, would be equal or superior to the direct cost to the suicidal individual of suicide in terms of any personal capacity to make such contributions.
Figure 2: Hamilton's Rule for the Biological Communication of the Individually Maladaptive Traits. Here: Cd = Cost in cultural reproductive potential of suicide to individual, r = coefficient of relatedness, and Br = Benefit in cultural reproductive potential of suicide to cultural relatives
With respect to practical public policy recommendations on reducing contagion, this interpretation provides an argument for curtailing any sensationalist media publicity around suicide, because such publicity might effectively increase the cultural fitness of the suicidal individual, that is, the capacity to pass on culture. Rather than merely refrain from the positive portrayal of suicide and suicide victims in the media as per current recommendations, this memetic model of suicide could be used to suggest that there could be a potential benefit (if reducing suicide levels is seen as beneficial) of not covering suicides stories at all, or at least reporting the deaths but not as suicides. Of course, there are important and controversial issues of censorship and free speech that are raised here, and it may be decided that suicide contagion is a lesser evil than what amounts to ideational eugenics.
7 ConclusionThis paper presents one vision of memetics, as an integrated part of applied social science investigating substantive issues of human experience. Understanding memetics as contagion psychology, using selectionist thinking to inform interpretation, is certainly not the only way to conceptualise the nascent discipline, but it is hoped that it is one that will allow memetics, after a quarter of a century of discussion, to start providing useful insight into real-world issues and problems.
ReferencesBerkowitz, L., (1984). Some effects of thoughts on anti- and prosocial influence of media events: A cognitive neoassociationist analysis. Psychological Bulletin, 95, 410-427.
CDC (1994) Centers for Disease Control and Prevention. Suicide Contagion and the Reporting of Suicide: Recommendations form a National Workshop. Morbidity and Mortality Weekly Review, 43, (RR-6).
de Catanzaro, D., (1980). Human suicide: A biological perspective. Behavioral and Brain Sciences, 3(2), 265-290.
de Catanzaro, D., (1981). Suicide and Self-Damaging Behavior: A Sociobiological Perspective. San Diego: Academic Press.
Colligan, M.J., Pennebaker, J.W., and Murphy, L.R., (1982) Mass Psychogenic Illness: A Social Psychological Analysis. New Jersey: Lawrence Erlbaum Associates.
Durkheim, E., (1970). Suicide: A Study in Sociology. Translation J. A. Spaulding & G. Simpson, (1951) London: Routledge & Kegan Paul. Originally published in French 1897.
Fiske, S. T. & Taylor, S. E., (1991). Social Cognition. New York: McGraw-Hill.
Geen, R., and Thomas, S., (1986) The immediate effects of media violence on behavior. Journal of Social Issues, 42, 7-27.
Gibbs, J., and Martin, W.T. (1964). Status Integration and Suicide. Eugene, Oregon: University of Oregon Press.
Halbwachs, M., (1978) The Causes of Suicide (translation of 1930 Les Causes du Suicide) London: Routledge.
Hamilton, W., (1964). The genetical evolution of social behaviour. Journal of Theoretical Biology, 7, 1-52.
Henry, A., and Short, J. (1954). Suicide and Homicide. New York: McGraw-Hill.
Higgins, E. T., (1989). Knowledge accessibility and activation: Subjectivity and suffering from unconscious sources. In J. S. Uleman & J. A. Bargh, (eds.), Unintended Thought, New York Guildford Press, 75-123.
Jo, E. and Berkowitz, L., (1994). A priming effect analysis of media influences: An update. In J. Bryant, & D. Zillmann, (eds.). Media Effects: Advances in Theory and Research. Hillsdale, NJ: Erlbaum.
Kerckhoff, A. C. and Back, K. W., (1968). The June Bug: A Study in Hysterical Contagion. New York: Appleton-Century-Crofts.
Leyens, J. and Parke, R. D., (1975). Aggressive slides can induce a weapons effect. European Journal of Social Psychology, 5, 229-239.
Maris, R., (1981). Pathways to Suicide: A Survey of Self-Destructive Behaviors. Baltimore: Johns Hopkins Press.
Marsden, P., (1998a). Memetics and Social Contagion: Two Sides of the Same Coin? Journal of Memetics - Evolutionary Models of Information Transmission, 2. http://cfpm.org/jom-emit/1998/vol2/marsden_p.html
Marsden, P., (1998b). Operationalising Memetics, Suicide, the Werther Effect and the work of D. P. Phillips. Proceedings 15th International Congress on Cybernetics. Namur, Belgium. http://pespmc1.vub.ac.be/Conf/MemePap/Marsden.html
Marsden, P., (2000) The Werther Effect: Suicide Contagion in the Internet Age - A Critical Evaluation, Theoretical Reconceptualisation and Empirical Investigation. D.Phil Thesis (Sussex). http://www.viralculture.com/DPhilsummary.htm
Back to Issue 1 Volume 5