Policy Papers

Understanding terrorism and radicalisation: a network approach

Paul Ormerod , Andrew P. Roach |

  • RSS Feed Icon

Executive Summary

  • New research on networks has put individual well connected agents at the centre of the spread of many social phenomena, including religious ideas.
  • Much can be learnt from the spread of the medieval Cathar heresy by relatively few individuals, the perfecti.
  • The inquisitors sent to suppress them used various unsuccessful strategies, whose failures can be illuminated by modern network theory.
  • Eventually they achieved success by prioritizing the gaining of information about the movements of heretics and isolating their local contacts.
  • England in the 1550s provides an interesting opposite model of how a society can be radicalized through heavy-handed public persecution.
  • Using a network simulation suggests that the martyrdom of key Protestant leaders may have been a sufficient condition for the conversion of England to Protestantism: had Mary's government not burnt anybody, England might well have remained a Catholic country.
  • These cases suggest, first, that it is important for the rest of society that opinion formers within local communities and families are left alone to make their own judgements on extreme Islamist ideologies, unless they can be shown to have strong active links with terrorism.
  • And second, that it is likely to be more effective to isolate key hubs in a terrorist network by neutralising their close contacts, rather than run the risk of creating a cascade effect through direct persecution and martyrdom.
  • Finally, our work is an illustration of a recent important discovery about systems in which social influence is a determinant of behaviour and ideas: deep knowledge about them can be extracted from limited information.
  • Thus we do not need to know the exact details of a past network in order to be able to draw valuable implications for the present day.


The authors of this paper have recently been involved in an unusual collaboration, funded by an AHRC Knowledge Transfer Fellowship, to cast light on contemporary issues using historical examples and recent developments in the theory of social networks.

Particularly important has been the modelling of so-called 'scale free' networks. These are usually complex, consisting of a small number of very well connected hubs and a large number of far less well connected nodes. Examples include the world wide web and the pattern of human sexual contacts. Although a small number of classic articles in the social sciences can be identified going back almost 100 years which describe scale free networks, it is only in the past decade or so that they have come to be recognised as being applicable to many social and cultural situations. Detailed analysis of their properties only commenced in the late 1990s.

Our most recent work with this model has concentrated on the suppression of a network in the case of the Inquisition and the Cathar heresy in France in the 13th century; and on the spreading of a network in the case of the conversion to Protestantism of England in the mid-16th century. In each case there may be interesting lessons to be learned for the understanding of terrorism and radicalisation throughout the world in the 21st century.

Network theory: cascades and hubs

The religious beliefs of an individual are formed in many ways, but here we focus on the process of conversion. We assume that an individual can only be persuaded to change his or her religious beliefs by observing the beliefs of other individuals whose opinion the individual in question values. This dramatic simplification is in fact a good approximation to reality in many modern contexts, especially in cultural markets. On 'YouTube', for example, a video can suddenly and inexplicably become incredibly popular, receiving many millions of viewings because of recommendation by influential friends and family, whilst others which appear very similar in terms of both content and quality receive hardly any.

During the past decade, our understanding of such phenomena has been dramatically increased by advances in our mathematical knowledge of networks. In particular, in the current context, of how 'cascades' can either spread or be contained across such networks. The notion of cascades can be applied equally to ideas, consumer behaviour, or even epidemic disease. All readers will be familiar, for example, with the common cold. When a new form of it evolves, the chances of catching it are increased if a close family member has the virus. In general, however, the common cold is spread by purely random contact with other individuals. Someone on the tube you will never meet again sneezes, and you have caught the cold. By contrast, the assumption that individuals are connected at random turns out not to be a reasonable one to make in the case of the spread of sexual diseases.

There is a fundamental difference between the patterns of connections in a randomly connected network and that which is observed in the patterns of connections of sexual partners. In networks in which individuals have a random chance of interacting with each other, the number of people with which each of the individuals has contact will be very similar. But in terms of sexual contacts, a 'scale free' network, a completely different pattern of connection is observed. A few people have an enormous number of partners over any given period of time, while most people have only a very small number. The internet has a similar sort of mathematical structure, with a small number of sites receiving huge number of visits, and most sites having relatively few.

One notable difference between random and scale free networks is that in a random network, there is a threshold which has to be reached of 'infected' individuals before a 'virus' (an actual virus, an idea, the purchase of a particular type of product or brand) can spread generally throughout the population. But in a scale free network, in principle even if a single individual becomes infected, this can percolate across the entire network. Naturally, the chances of this happening if the person only has a few connections is very low, but it can still happen. Therefore in such a network viruses are very difficult to eradicate, always having the capacity to revive suddenly in a 'phoenix effect'. The reason for this extreme behaviour is the important role in such networks of the 'hubs', those individuals with large numbers of connections. Suppose we want to inoculate a population against the spread of a virus, a policy of inoculating people at random, even if a high percentage of the population is dealt with in this way, has only a small probability of success. If only one of the hubs escapes this policy, the virus will eventually reappear, even though it might be thought to have been eradicated.

Suppressing a network: inquisitors and Cathars

The Catholic church faced such a problem with hubs in attempting to deal with the Cathar heresy in the 13th century, the first organised challenge faced by the Roman Church for many centuries. Medieval observers were themselves quick to use the analogy of diseases in the context of the heresy. As the chronicler Peter of Vaux-de-Cernay observed 'one bunch of grapes can take on a sickly colour from the aspect of its neighbour, and in the fields the scab of one sheep or the mange of one pig destroys an entire herd'. Initially, a policy of attempting to inoculate the population of Southern France by terror was tried through the Albigensian crusade in 1209. The policy of suppression was essentially random, with individuals and groups singled out and punished. The problem was famously recognised in the statement attributed to the papal legate and monk, Arnaud Aimery at the storming of Beziers. Knowing from the confessions of these Catholics that they were mixed up with heretics, the crusaders said to the abbot: 'What shall we do, lord? We cannot tell the good from the bad'. The abbot is said to have replied: 'Kill them. For God knows who are his'.

However, the very structure of Catharism, based as it was on a small number of highly mobile and influential perfecti (holy men and women), militated against the success of such a strategy. Indeed, as early as the 1220s, when the above account was written, the policy of mass intimidation by terror was recognised to have failed. In general, the heresy lingered on even when it was believed to be suppressed. Catholic writers preparing reports for the 1274 Council of Lyon thought the threat was over, yet the last Cathar was only burnt in 1321 and a Cathar revival led by only around ten perfecti around 1300 caused a major panic among churchmen.

A very effective strategy of inoculation in scale free networks is to target the hubs themselves, though obviously this is far from easy. After 1231, specialist inquisitors started to be employed against heresy, mainly drawn from the learned order of Dominican or black friars. Following early failures they began to assemble a body of expertise which was summarised in a number of handbooks for inquisitors. The best known of these was by the Dominican friar, Bernard Gui (the name used by Umberto Eco for the caricature of an inquisitor in his novel The Name of the Rose), the Practica Inquisitionis, completed in 1323-24 from which most of the following is taken. The sickness metaphor was still present, but Gui saw himself as a physician applying 'different and specific medicines' for the varied diseases of heresy. He was interested in the connections of heretical sects; in the section dealing with Cathars he suggested a suspect be asked: 'Whether he had any familiar association with heretics; when; how; and who was responsible for it'. He was also keen to obtain information on the physical organisation of the network, interrogators should ask: 'Whether he received any heretical person or persons in his home; who they were; who brought them there;...who visited them there and escorted them thence.'

Gradually, the inquisitors evolved the strategy of what is now known as 'acquaintance immunisation' for preventing the spread of viruses across a scale free network. Catching the hubs - the perfecti - would be ideal but failing that, close contacts of the perfecti could be persuaded in a variety of ways not to disseminate news of their activities. Whilst the threat of torture or the stake remained in the background, it was rarely used. A particularly effective strategy was to target guides, individuals who made a living by travelling from village to village passing on information and leading perfecti. Gui himself realised that the beliefs of this group of people was largely irrelevant, provided that they did not spread news about the perfecti.

The various attempts made to contain and suppress the medieval Cathar heresy are almost a textbook for dealing with modern terrorist movements which approximate this type of structure, in which there is a small number of key individuals. General suppression by force and counter-terror has only a low probability of success. If there is a lesson for the 21st century it is the contrast with the American mistake in Iraq which was to treat the post-invasion situation from an ideological perspective, and to assume that most members of the Baathist Party were fervent supporters of Saddam Hussein. As a result, officials were dismissed en masse, and the basic functions of the state became difficult to carry out. As most of them were almost certainly rather innocuous careerists, just as happy to serve a different master, it would have made much more sense to have invested time in identifying and dealing with the real fanatics.

Spreading a network: Marian England

Targeting the key leaders of a threatening movement should therefore be an obvious and attractive policy, but the way in which they are dealt with is very important if the structure of influence approximates a scale free one. Heavy-handed policies can backfire spectacularly, as the Marian persecution in England in the 1550s illustrates.

Despite the spread of support for Protestantism amongst the elite, it appears that when the Catholic Queen Mary came to the throne in 1553 the bulk of the population still retained an affection for Catholicism. Yet, within a relatively short space of historical time, religious affiliation in England underwent a marked change. Again, there are arguments about the extent and timing of this change. It would certainly be reasonable to say that by the 1570s, a sizeable minority within England not only paid lip service to the new official doctrine, but embraced it enthusiastically and by the 1580s many adults had known nothing else. However, even as early as Elizabeth's accession in 1559, people in general were prepared to accept a settlement which has been described as 'a snapshot of King Edward VI's Church as it had been in 1552' with its doctrine of predestination and stripping out of most of the ceremonial elements of the 'old religion'.

Certainly, the increased threat to England from Spain during Elizabeth's reign increased both support for the Protestant settlement and hostility to Catholicism. But however we qualify the evidence, it does seem to be the case that within the space of a couple of decades, genuine support for Protestantism had grown markedly in England. How could such a rapid shift in opinion take place?

Modern network analysis points firmly to the impact of the policies adopted by Mary to restore Catholicism at least as far as it might have opened minds to the possibility of sympathy with the Protestant cause, and there is some evidence of this in contemporary accounts. In particular, the burnings of the Protestant leaders exercised a strong influence on public opinion. This had two aspects. First, the main one of a switch of allegiance from Catholicism to Protestantism. But, second, the effective elimination of the rival Protestant sect of the Freewillers, who were not persecuted on anything like the scale of their predestinarian contemporaries.

Church leadership under Edward VI formed a scale free network in terms of the number of people in the country who were aware of its members' existence. Many more people would have known of the Archbishop of Canterbury or York or the local bishop, than a village priest. But in terms of influencing people's religious choices, this network was latent rather than active. All this changed under Mary. The Catholic authorities had the intention of creating a climate of fear which they believed would suppress what they regarded as the Protestant heresy. They rapidly imprisoned many of the leaders of 'Reformed Protestantism' as the Edwardine doctrines have come to be known by historians and went on to articulate a policy of elimination of key figures through public burnings.

In contrast, the Protestant clerical elite were convinced from the outset that their death could cause the policy to rebound on the persecutors. They were well aware of the necessity of creating a good impression at the stake. To this end, then, they encouraged one another. On 8 February 1555, on the morning of his execution Laurence Saunders, a noted Protestant preacher in London and the Midlands, wrote to his wife and supporters: 'God's people shall prevayle: yea our blood shal be their perdition Who do most triumphantly spill it' (Foxe, 1576, p.428). A few weeks earlier he had written to Cranmer, Latimer and Ridley who were in prison also on the verge of execution. These three were probably the most prominent of all the Reformed Protestant leaders. Cranmer was the Archbishop of Canterbury, the religious head of the English Church, Latimer was Bishop of Worcester and a famous preacher, and Ridley, Bishop of Rochester, was the author of the 1552 Book of Common Prayer which defined the liturgy of the Reformed Protestant faction. Saunders' message was to make the most of what he termed God's 'unspeakable gift' and put on a good show. Latimer and Ridley famously rose to the occasion, with perhaps the most memorable of all the quotations of the English martyrs: 'Be of good comfort, master Ridley, and play the man. We shall this day light such a candle, by God's grace in England, as I trust shall never be put out.'

The net effect was to transform a scale free network of mere awareness into one in which individuals were actively influenced by the beliefs and actions of others. Under such conditions, an ideology (or a virus) can spread rapidly through a population. In network terms the martyrs were influential hubs who could in turn influence opinion formers in smaller local and familial networks.

The control group here is the Freewillers. Although most historians consider them a small fringe sect, they were based in the influential prosperous south east of England and had an attractive emphasis on free will as opposed to the predestination fashionable in mainstream Protestantism. Writing from prison in January 1555, for example, the future martyr, John Bradford commented that 'more hurt will come by them (the Freewillers), than ever came by the papists, inasmuch as their life commendeth them to the world more than papists.' Crucially, far fewer Freewillers were burned; in general they may even have decided to live and fight another day as a matter of policy. There was certainly talk of an accommodation with the Marian government. However, in network terms this was to forego the opportunity of becoming the kind of hubs the martyrs were. The Freewillers' disappearance by the early years of Elizabeth's reign was almost total.

To measure the possible impact of the Marian burnings on public opinion we created a model of 500 agents to represent Marian England and allocated to them religious allegiance in the proportion of 92% Catholic, 7% Protestant, 1% Freewillers and other Protestant sects. Each agent could convert if more than a certain percentage of the total weight of the individuals to whom he or she paid attention subscribed to the rival ideology. For the sake of argument we fixed the number of individuals at four and the percentage weight necessary to switch allegiance at 67%. In choosing this weight we were deliberately conservative, although the reasoning behind the numbers chosen is explained in detail in our Cultural Science paper and interested readers are referred to this.

In the first instance we made our allocations at random and proceeded to run the model 500 times, each sequence consisting of 500 steps. Not surprisingly in almost every case, Marian England turned completely Catholic. However, if only two of the most connected agents (0.4%), the hubs, were allocated to mainstream Protestantism then Catholicism ended up as a minority in a significant number of instances (7.4%) and in nearly every case the number of Catholics declined from the initial allocation. In almost every scenario the Freewillers quickly disappeared unless they too were deliberately allocated one of the two most connected hubs.

Our key finding was that the burnings, and the resulting creation of an active scale free network on which the actions of the martyrs influenced opinions, could have been (and we stress the conditionality) a sufficient condition for the conversion of England to Protestantism. The experiment also suggests that had Mary's government not burnt anybody, England might very well have remained a Catholic country. This is not completely surprising to historians, given the longstanding emphasis the historical literature has placed on the role of the martyrs and the conservatism of the population revealed by modern research. There were plenty of reasons why her English subjects disliked Mary, such as her marriage to the King of Spain, her close relationship with the Pope (tension between the monarchy and papacy long pre-dated the Reformation), and her loss of Calais, England's last colonial territory in France. Her death and Elizabeth's accession were welcomed by many. But even so, without the martyrs, the reformers under Elizabeth might have found their task of restoring Protestantism much harder.

Our results suggest that the persecuted Reformed Protestants were right and that the burnings offered an unprecedented opportunity to influence opinion. The spread of Protestantism throughout England and the complete marginalisation of Catholicism took several decades to complete, but the seeds were sown in the 1550s. Those who were persecuted by the Marian government shrewdly chose martyrdom as an opinion-changing strategy as well as for its own spiritual benefits. Modern understanding of network theory shows that they were right to do so.

There may well be lessons in this exercise which would illuminate the problem of dealing with modern terrorist groups. Martyrdom is potentially much more powerful if it is undergone by prominent adherents of the ideology. At present militant Islamic terrorists tend to be drawn much more from the rank and file. It would be more effective for their cause if Osama bin Laden and some of his top associates, the hubs within their network, were to become suicide bombers themselves or to be martyred publically by the authorities. As long as this does not happen, the opinion forming potential of martyrdom will be self-limiting.


Bringing the two case studies together suggests that it is important for the rest of society that opinion formers within local communities and families are left alone to make their own judgements on extreme Islamist ideologies, unless they can be shown to have strong active links with terrorism. Even then it is likely to be more effective to isolate the key hubs by neutralising their close contacts, rather than run the risk of creating a cascade effect through direct persecution and martyrdom.

Our work is also an illustration of a recent important discovery about systems in which social influence is a determinant of behaviour and ideas. Somewhat paradoxically, deep knowledge about such systems can be extracted from very limited information. We do not need to know the exact details of a network in order to be able to draw valuable implications; very limited information is sufficient. Our analysis of both the Cathar heresy and the Reformation in England shows that examination of a relatively small amount of contemporary evidence can provide an analytical framework which can give useful insights into effective counter-terrorism strategies.

Further Reading

A.L. Barabasi, Linked: the new science of networks (2002)

P. Ormerod,'Extracting deep knowledge from limited information on evolved social networks', Physica A, 378 (2007)

R La Violette, R Colbaugh and K Glass, 'Deep information from limited observation of robust yet fragile systems', Physica A, (2009)

P. Ormerod,'Hayek, "The Intellectuals and Socialism", and Weighted Scale Free Networks', Economic Affairs, 26 (2006)

Andrew P. Roach, The Devil's World: Heresy and Society 1100-1300 (Harlow, 2005)

P. Ormerod and Andrew P. Roach,'The medieval inquisition: scale-free networks and the suppression of heresy', Physica A, 339,(2004)

P. Ormerod and Andrew P. Roach,'Emergent Scale-free Social Networks in History: Burning and the Rise of English Protestantism', Cultural Science, Vol.1, No.1, (2008)


Papers By Author

Papers by Theme


Sign up to receive announcements on events, the latest research and more!

To complete the subscription process, please click the link in the email we just sent you.

We will never send spam and you can unsubscribe any time.

About Us

H&P is based at the Institute of Historical Research, Senate House, University of London.

We are the only project in the UK providing access to an international network of more than 500 historians with a broad range of expertise. H&P offers a range of resources for historians, policy makers and journalists.

Read More