'It is an observable phenomenon in Northern Ireland, and elsewhere', the Irish Taoiseach (Prime Minister), Bertie Ahern, suggested in 1998, 'that tension and violence tend to rise when compromise is in the air'. Almost three times as many people were killed in South Africa while the agreement was being negotiated between 1990 and 1994 than in the previous four years. The use of political violence also rose dramatically during the Middle East peace process, and more than two-thirds of victims were killed during the Rabin-Peres era.

Nor does violence disappear when a cease-fire is declared. Instead, both the forms of violence, and violence-related issues, change in ways which threaten the evolving peace process. The threats come from four principal sources.

1. Violence by the state

Governments are often as divided about peace processes as their opponents. Governmental or quasi-governmental agencies may continue covert actions either to undermine the process itself or, as when the 'third force' in South Africa used violence to support the Inkatha Freedom Party, to influence the outcome of negotiations. Even if such pressure is not applied, the security apparatus built up during periods of violence is a potential danger when they end. Militants are released instead of imprisoned. The security forces augmented during the violence are abruptly reduced in number, endangering their jobs and personal security. Demands are made to reform the police force that regarded itself as the bastion against terrorism. Unless handled carefully, disaffection within the security forces has the potential to undermine the peace process itself.

2. The paramilitary fall-out

Disaffection within paramilitary organisations is a more obvious threat to progress. These are rarely the monoliths presented by their opponents. They are complex organisms performing different functions and providing umbrellas for different interests. During cease-fire periods these interests diffuse and fragment. At least four pose separate threats to peace processes:

2.a A return to political violence

Cease-fires are never unanimous, so the most obvious threat is that they will break down and political violence will return. The more disaffected members of the militants may desert to splinter groups or perform individual acts of violence. The less affected may go along with the majority view, but their agreement is conditional. Their continuing allegiance depends on measurable rewards from negotiation, including prisoner releases and the dismantling of the security apparatus. These rewards are rarely immediate. Consequently the pendulum may swing back towards the militants. It is worth remembering that the talks broke down in both Northern Ireland and South Africa, leading to periods of renewed violence in 1996 and 1992 respectively, before the peace process resumed.

2.b 'Tactical' Violence

A common fear among constitutional parties is that the pace of negotiation will be determined by the gunmen outside the negotiating rooms, and that their political surrogates may use the threat of violence to get their way; to paraphrase Clausewitz, that paramilitary negotiators were approaching politics as the continuation of war by other means.

Acts of violence during the process are taken as confirmation of this fear. The early stages of the Oslo process were hampered by Israeli suspicion of the PLO's motives, and later the government regarded Hamas' atrocities as evidence that Arafat either was unable to control his own people or was colluding with it for strategic advantage. During the early stages of the Northern Ireland peace process, unionists constantly warned that Sinn Féin would use IRA violence to remind other negotiators of their power. As the negotiations progressed, it became increasingly clear that violence threatened Sinn Féin's interests as much as those of the unionists. The complaints diminished, as had happened in South Africa.

2.c Spoilers: Zealots versus Dealers

The very involvement of paramilitary interests in negotiations implies that the purity of their cause has been compromised. It imposes strains on organisations that are essentially military, and it is difficult to find any instances when such a move was not accompanied by a split between two main groups - the Zealots and the Dealers. The Zealots often comprise radical groups, like the 'Real' IRA and the Continuity IRA in Northern Ireland, who picked up the torch (sometimes literally) they believed had been surrendered by the Dealers. In Sri Lanka the result was a succession of assassinations of Tamil rivals by the Tamil Tigers; in 1998 they murdered the Tamil mayor of Jaffna, having previously killed her husband, and then her successor as mayor. All major attempts to start negotiations in Israel-Palestine were accompanied by Palestinian attempts to bomb them away; the killing of 28 Muslims by the Israeli settler Baruch Goldstein in 1994, and the assassination of the Israeli Prime Minister Rabin in 1995, showed that zealots from both communities were eager to capsize the process. Both white and black dissidents threatened the South African process, although spoiler violence by blacks was relatively well controlled during the period of negotiation, perhaps a reflection of the dominance of Nelson Mandela and the ANC.

2.d Family Feuding: Internal paramilitary violence

The determination of militants to exercise control over their own communities does not diminish when they enter a negotiation process. They are unlikely to hand over this negotiating card, even at the risk of destabilising the peace process itself. Violence between factions within the Palestinian community continued after the first transfer of land to Palestinian control in 1995. Punishment beatings persisted in Northern Ireland after the Good Friday Agreement as both loyalists and republicans exercised what they regarded as their policing role.

3. Violence in the Community

The fragmentation of paramilitary organisations is not the only violent accompaniment to peace processes. Two other forms of violence may also seriously undermine them - the revival of direct confrontations between ethnic rivals, and a rise in the conventional crime rate.

3.a Return to the Streets

When a cease-fire is declared the discipline of the military campaign diminishes, but the underlying sectarian hatred remains, taking the form of riots and undisciplined confrontations with ethnic rivals or the police. They can become a serious threat to a peace process. The paramilitary representatives who enter negotiations are not divorced from the instincts and antagonisms of their communities, and may feel the need to support them. During the early months of negotiation in South Africa sectional violence was partly orchestrated between the ANC and the IFP, but spilled over into tit-for-tat killings and more general violence. The emergence of the 'street guerrilla', mostly teenagers, in the Basque Country arose partly from ETA's weakening authority, but was also partly a reaction to the popularity of the peace movement. In Northern Ireland the declaration of the cease-fires, and the ending of direct violence between organised paramilitaries and the security forces, was marked by a return to more direct violence between Catholics and Protestants, especially during the marching seasons from 1996 to 1998.

3.b Ordinary Decent Crime

A rise in conventional crime may appear to present a less obvious danger. The crime rate has risen to such a degree in South Africa, however, as to seriously undermine post-settlement peace-building. It is a threat to inward investment, tourism and general confidence. By the mid-1990s the high level of conventional crime had far outpaced political violence as a destabilising factor. By 1998 the daily homicide rate was 52, and still rising. More ominously, the barrier between ordinary crime and South Africa's underlying racial tensions, never sharp, became increasingly blurred. Fifteen hundred white farmers were attacked between 1994 and 1998, resulting in more than 200 murders, and threatening to create a loop back from post-settlement civil violence to the violence familiar from the earlier struggle.

4. New security-related issues in negotiation

During peace negotiations new substantive issues emerge, most of them security related - notably early prisoner releases, decommissioning of illegal weapons and policing. The South African peace process actually began with the release of prisoners, when Nelson Mandela and other ANC leaders were set free in February 1990. The rate of prisoner releases has been a constant source of bitter dispute in Israel-Palestine, but was implemented relatively smoothly in Northern Ireland. Decommissioning of weapons is a more complicated matter. The 1991 South African National Peace Accord did not ask the ANC to disband paramilitary units nor hand over their arms caches; it required only that firearms should not be displayed at public meetings. The stubbornness of Ulster Unionists in demanding decommissioning and of Sinn Féin in rejecting it in Northern Ireland, however irritating, arose from the need of each side to keep its primary constituency on board, and on the symbolic association between decommissioning and surrender. Police reform is an equally emotional issue in negotiations. It is axiomatic that divided societies require a police force that reflects the divisions. Section 195 of the 1996 South African constitution, for example, insisted that the police and defence forces 'be broadly representative of the South African people' and by 1996 16 000 former guerrillas had been absorbed into the army. There are plenty of other examples of former combatants entering the security forces, although not all attempts at integration succeeded.

All of the forms of violence detailed above are separate threads in a single weave. The pattern that unites them is the central role of violence both before and after the declaration of a cease-fire. Each distinct form, however, demands different policy approaches, for governments, international and regional bodies, negotiators and NGOs attempting to move towards a fair and lasting settlement. The identification of these approaches is the next research priority.

--------------------------------------------------------------------------

Acknowledgements

The research upon which this article is based was started in 1998, when I was a Senior Jennings Randolph Fellow at the United States Institute of Peace, and completed in 1999 when I was visiting professor in the Kroc Institute at the University of Notre Dame. I want to record my thanks to both institutions, as well as to INCORE at the University of Ulster, at which I am Senior Fellow.