Hard to miss , easy to blame ? Peacelines , interfaces and political deaths in Belfast during the Troubles

As Northern Ireland moves further from the period of conflict known as the ‘Troubles’, attention has increasingly focussed on the social and material vestiges of that conflict; Northern Ireland is still a deeply-divided society in terms of residential segregation between Catholic nationalists and Protestant unionists, and urban areas are still, indeed increasingly, characterised by large defensive walls, known as ‘peacelines’, which demark many of the dividing lines between the two communities. In recent years a body of literature has emerged which has highlighted the spatial association between patterns of conflict fatality and proximity to peacelines. This paper assesses that relationship, arguing that previous analyses have failed to fully take account of the ethnic complexity of inner-city Belfast in their calculations. When this is considered, patterns of fatality were more intense within the cores or ‘sanctuaries’ of highly segregated Catholic and Protestant communities rather than at the fracture zones or ‘interfaces’ between themwhere peacelines have always been constructed. Using census data at a high spatial resolution, this paper also provides the first attempt to provide a definition of the ‘interface’ in clear geographic terms, a spatial concept that has hitherto appeared amorphous in academic studies and media coverage of Belfast during and since the Troubles. In doing so it embodies both the material and demographic aspects of social division in Northern Ireland, and suggests an urgent need to reappraise the true role of these forms of social boundary in influencing patterns of violent conflict. 2014 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/3.0/).


Introduction
Over the past decade, an increasing amount of public and academic attention has become focussed on the 'peacelines' which divide highly segregated Catholic and Protestant communities across urban Northern Ireland, but most notably within Belfast. Despite the fact that Northern Ireland is now some fifteen years on from the Good Friday Agreement and a new consociational political dispensation, these peacelines have not only persisted but have actually increased rapidly in number since the substantive 'resolution' of the Troubles (McDowell, 2009;Nagle, 2009: p. 137;Shirlow & Murtagh, 2006: pp. 66e67). During the Troubles the existence of these defensive barriers was easier to justify in the face of the threat posed by paramilitary activities, but in a time of nominal 'peace', they often appear to outsiders as socially retrograde and anachronistic.
Another reason for the attention which peacelines garner is the fact that these are typically a feature of 'interface' areas, a wider term which has come to embody the spaces that lie between these highly segregated neighbourhoods. While the threat posed by paramilitary groups has largely receded in Belfast, these interfaces remain sites of contestation and antagonism, particularly in the summer months when tensions between the two communities rise during the marching season. Annual Orange parades to celebrate the victory in 1690 of the Protestant King William III over the Catholic James II are particularly problematic because they disrupt Belfast's delicate modus vivendi as Protestant marches are seen by Catholics to be both inherently supremacist at a general level (McGarry & O'Leary, 1995: p. 235), but also antagonistic a local level by transgressing the religio-political geographies which are at once both highly pronounced and severely contested (Bryan, 2000: pp. 1e10;Cairns & Smyth, 2002: p. 147). So in short, while the major constitutional conflict has been addressed, Northern Ireland is still far from being a 'normal' society because so little territory is 'neutral' in a religio-political sense (Graham & Nash, 2006: pp. 270e273), and it is in the contested margins or interfaces, where these battles are believed to have played out with increasing intensity over the last decade (Boycott, 2011;Brown, 2011).
While it is therefore clear that the boundaries between polarised Catholic and Protestant neighbourhoods remain sites of tension and hostility, this paper seeks to address specifically the relationship between peacelines, interfaces and conflict-related fatal violence during the Troubles (Gaffikin & Morrissey, 2011: p. 209;Mesev, Shirlow, & Downs, 2009: p. 901;Shirlow, 2003a: pp. 80e81). It analyses the association between peacelines and deaths on both spatial and temporal grounds, before addressing the role of interface zones in the Troubles more widely. It argues that while existing work has accurately identified deaths as occurring primarily in neighbourhoods bounded by peacelines, the scales of analysis employed have been unable to reflect the religio-political complexity of inner-city Belfast. By applying spatially-detailed data and sensitive methodologies we argue that death rates were actually lower in immediate proximity to peacelines than they were deeper within Catholic and Protestant enclaves. These findings extend to a fuller analysis of interface areas, giving putative form to what is currently a vague spatial concept. The overall result provides quantitative support for the qualitative research conducted by Allen Feldman for his landmark ethnographic study of west Belfast, Formations of Violence (1991), in which he identified that over the course of the Troubles it was the cores or 'sanctuaries' of these highly segregated neighbourhoods that became the primary arenas of conflict.
'An ounce of prevention is better than a ton of cure': historical and theoretical contexts Primary documents from the early phase of the Troubles in Belfast seem to make it abundantly clear why much of the city is now characterised by a complex network of defensive fortifications constructed between Catholic and Protestant areas. The first deaths in the summer of 1969 in the conflict that would become known as the 'Troubles' should be viewed not as the start of a process but as the result of a deepening political and social crisis in Northern Irish society since at least the mid-1960s. The causes of the Troubles are both complex and contested, with contributions by Whyte (1990) and McGarry and O'Leary (1995) remaining as scholarly landmarks in interpretations of the period. The unresolved argument between the Catholic minority, most of whom were nationalists and sought a united Ireland, and Protestants, the vast majority seeing themselves as pro-British unionists, re-emerged violently in the late-1960s. Over the next three decades, some 3500 people died as a direct result of the conflict, the bulk killed by loyalist and republican paramilitaries claiming to represent the interests of their wider Protestant and Catholic populations (Bardon, 2005: pp. 622e666;Bew, Gibbon, & Patterson, 2002: pp.137e187;Buckland,1981: pp.106e131;Foster,1989: pp. 582e592;Hanley & Millar, 2009: pp. 70e107;Whyte, 2003: pp. 309e316).
From the earliest stages of the conflict with the deployment of the British Army in 1969, a dominant strategic imperative has been to keep the two communities apart (JSC, 1969). Inter-communal blockades have been part of the city's social history since at least the early-nineteenth century (Hepburn, 1996: p. 123;2008: p. 220). However, these barricades were temporary affairs, appearing only in periods of increased tension. With the advent of the recent Troubles that changed as the number of permanent structures increased dramatically (BIP, 2012: pp. 12e13). The continual challenges faced by the security services in managing low-intensity interface hostilities between the Catholic and Protestant communities in large part explains the rationale for the building of permanent impediments between the two: It is true to say that there is a high incidence of hooliganism [at the New Lodge interface].At lunchtime and in the afternoons schoolchildren of different religions engage in throwing stones at each other. In the evenings, in connection with attendance at youth clubs and at nights after public houses close there are similar and frequently more serious incidents e only timely police intervention prevents many such incidents becoming riots (APS (MD), 1970).
Despite being designed to protect life and prevent violence, peacelines have themselves become a key source of conflict in debates on the Troubles (Jarman, 2001: p. 36). Recent research shows that for many residents of interface areas, physical barriers offer them protection (Byrne, Heenan, & Robinson, 2012;Calame & Charlesworth, 2009: p. 61) but academic critics have argued that peacelines have had a formalising effect on religio-political differences, rendering a 'culturally opposed immediate community as a menacing spatial formation' (Shirlow, 2001: p. 68). This creates fear which in turn engenders violence so peacelines and interfaces more generally are 'an outcome of the broader structuration of conflict but they also produce a dynamic of their own which, in turn, feeds back into the wider reproduction of both spatial and social distance' (Murtagh, 2004: p. 461). These are used 'to construct narratives of inclusion and exclusion that define communities and the ways in which they are rendered specific or differentiated' (Graham, 2002(Graham, : p. 1008. At a conceptual level, arguably implicit in all of the above criticisms is the sense of peacelines as articulations of state power. But if this is indeed the case it is in the qualified and partial exercise of that power, as the violence which has afflicted these areas can be seen as a manifestation of the sort of 'uncontrollable autonomy' identified by Lefebvre in The Production of Space (1991: p. 26). Furthermore, Till et al. (2013: p. 9) have recently argued that in many respects, such barriers represent not the authority of the state but rather the limits of that authority, or as Brown (2010) has put it, as being emblematic of 'waning sovereignty', declining civic deference and the inability of states to control their populations by any means other than the crudest of techniques. For Loyd, Mitchelson, and Burridge (2012: pp. 5e6) however, the proliferation of walls, prisons and other spaces of incarceration are symptomatic of a much wider crisis in late-capitalist societies in which these instruments of state violence act to protect the interests of wealth holders and accumulators. This may well be true, but in the Northern Ireland context walls have not acted to protect the interests of citizens from advancing 'would-be' citizens who both share common capitalist self-interests (Loyd et al., 2012: pp. 6e8) as is the case for example, on the border between Mexico and the U.S.A. or on the margins of the European Union. Rather, they have tended to define pre-existing boundaries between groups who cannot be reconciled in terms of religiopolitical identity, and therefore Reece Jones' (2011: p. 215) conceptualisation perhaps has greater salience when he has argued that borders more generally act 'as important sites for the performance of both security and citizenship. They are the line that symbolises the distinction between the 'homeland' and the outside world'.
The identification of these liminal spaces as performative zones gains even greater conceptual and empirical substance through a consideration of the work of the American cultural anthropologist Allen Feldman. In Formations of Violence (1991) he explores through a detailed and spatially-nuanced qualitative approach how the working-class Catholic and Protestant communities of west Belfast sought to organise space in ways that subordinated patterns of violence to the needs of everyday life: The sanctuary/interface complex was an ideological organization of the spatial dimension of human association and the spaces of violence. It attempted to preserve the subordination of violent enactment to the prerequisites of residence and kinship through spatial confinement of violence. The sanctum was constituted by a space that was reserved for residence and kinship and by a complementary space, the barricades-interface continuum, reserved for the ideological and material reproduction of community through violence (36).
The creation of the sanctuary is both contingent upon and instrumental in the development of 'topophilia', in which space is eulogised and set apart, necessitating defence (Feldman, 1991: p. 38). In the words of Bachelard (1994: pp. 217e218): Outside and inside are always intimate e they are always ready to be reversed, to exchange their hostility. If there exists such a border-line surface between inside and outside, this surface is painful on both sides. For Feldman, this surface has clear manifestation in the interface, designed to serve as the place of violence (1991: p. 37). Yet despite these socio-spatial formulations, the division between the interface and the sanctuary became compromised as the conflict ground on, worn away by both state actors and paramilitaries, the latter having justified their territorial activities as defensive: Yet paramilitary practice created cycles of violence that exceeded the spaces and performative prescriptions that had been predicated by the sanctuary/interface construct. This space was rapidly instrumentalized by paramilitary organizations as they converted the sanctuary into a base for operations. Chronic paramilitary violence eventually played a pivotal role in the disintegration of the community space as a sanctuary to the extent that it attracted retroactive violence aimed at the confessional space (Feldman, 1991: p. 39).
The distinction between the performative and the fatal is critical to an understanding of what the location of violence during the Troubles actually signifies. While both forms of violence are integral to the conflict, being linked at one level through a shared geographical theorisation of bodily practices and resistance (Thrift, 1997), at another level there is also an important difference. Interfaces have long acted as spaces for the symbolic performance of difference, but the data to be presented here suggests that it was actually within the cores of polarised communities that the fatal conflict played out.
Peacelines and interfaces are bad things at an emotional and psychological level: they engender fear and hatred of those on the other side of them (Shirlow, 2001: pp. 67e74;2003a: pp. 76e91;2003b: pp. 77e94). It is also impossible to dispute the notion that interfaces have historically, and continue to be the foci of intense inter-communal hostility and there is a body of literature which sees such communities as being economically, socially and spatially disadvantaged by the presence of the structures as natural commercial and infrastructural patterns are distorted and investment is deterred by the stigma of locating in an interface zone (Murtagh, 2002: pp. 45e52;Shirlow, 2006: p. 103;Shirlow & Murtagh, 2006: pp. 64e65). Interface violence has been an ongoing source of this negativity, and while it remains a serious and ongoing problem, the key fact is that such violence has tended to be more sporadic, more directed against property and less often resulting in loss of life (Bollens, 1999: p. 77;Cunningham, 2014: pp. 247e263;Jarman, 2003: p. 228; Jarman & O'Halloran, 2001: pp. 6e10). In this paper however, the focus is on the view that peacelines and interface zones may have acted to focus patterns of political killing during the Troubles. For some authors, this possibility has been merely implicit in their findings (Mesev et al., 2009: p. 901;Shirlow, 2003a: pp. 80e81), but others have more explicit, arguing that peacelines possibly acted as 'magnets' for fatal violence (Gaffikin & Morrissey, 2011: p. 209). This paper seeks to build upon this invaluable body of scholarship through a closer spatial and temporal examination of the evidence.

Data and methodologies
This paper uses geographical information systems (GIS) techniques to examine the relationship between patterns of fatality during the Troubles in Belfast and the boundaries between highly segregated Catholic and Protestant communities in the city. The analysis proceeds by addressing both the material manifestations of the social and political divide in Northern Ireland in the form of peacelines, before engaging with the more substantive demographic cleavage in the city's residential geography through an evaluation of the role of interface areas more generally. Geography has a particularly vital role to play in understanding the dynamics of the political conflict in Northern Ireland; as Stewart (1977: p. 56) has put it: Topography is the key to the Ulster conflict. Unless you know exactly who lives where, and why, much of it does not make sense.Ulster's troubles arise from the fact that people who live there know this information to the square inch, while strangers know nothing of it. This paper is based on two geo-referenced datasets that together allow us to explore in detail the micro-geographies of fatal violence during the Troubles. The first of these records the location of each conflict-related killing that took place in Belfast during the Troubles together with information on which group the victim and perpetrator came from. The second is data from the censuses of Northern Ireland for 1971, 1991 and 2001 which, uniquely in the UK, are available for the major urban areas at 100 m grid square using units that are consistent over time (Shuttleworth & Lloyd, 2009: pp. 213e216). The use of standardised areal units presents a solution to a major problem in time-series analyses meaning that it not necessary to rely on population estimates which would otherwise result from interpolating census datasets from one set of boundaries onto another (Gregory & Ell, 2007: p. 138). The benefit of using such a detailed census geography will become apparent very quickly in identifying where the boundaries between polarised Catholic and Protestant neighbourhoods lie, even if these boundaries are not marked by physical barriers in the form of peacelines. Northern Ireland's censuses have always included data on religion, as well as the more conventional census variables on topics such as household size and themes relating to deprivation. Unfortunately the 1981 census was taken against the backdrop of increased political tensions arising from the hunger strikes and suffers from some serious problems, specifically an underenumeration of Catholics (O'Grada, 1997: p. 209;Shuttleworth & Lloyd, 2009: p. 216). For this reason it is not used in this paper and the analysis presented here covers the period between the initial outbreak of violence in 1969 and 2001. The reason for ending in 2001 is so that the study period may be book-ended by that year's census, but at a more fundamental level it marks a point in Northern Ireland's history at which the political settlement embodied in the Good Friday Agreement could be seen to have taken hold, with power devolved to a new legislative assembly at Stormont, paramilitary decommissioning of weapons and most importantly, a dramatic decline in the level of political violence (Bew et al., 2002: pp. 244e245;English, 2003: pp. 331e336;Taylor, 2002: p. 390).
The database of killings was compiled by Malcolm Sutton and is hosted by the University of Ulster's 'Conflict Archive on the InterNet' (CAIN) (CAINa, 2013). This database was initially created to act as a memorial to those who had died during the conflict and formed the basis for a book, Bear in Mind These Dead.An Index of Deaths from the Conflict in Northern Ireland (Sutton, 1994). The process of updating the database has continued to the present to include new deaths as they occurred. While the rate of entries has declined substantially since the late 1990s and the signing of the Good Friday Agreement, the vital role of keeping the information up to date, and more specifically, correcting the information on individuals already within the record, continues. This highlights the uniquely interactive nature of the resource as one in which the people most directly affected by its contents, the bereaved relatives of those therein, have a direct input into the ongoing process of perfecting its accuracy and detail (Lynn, 2011). For the purposes of this study, deaths occurring within the Belfast Urban Area (BUA) between 1969 and 2001 form the basis for analysis.
The Sutton database provides information on the name of each person killed, the date of the incident leading to their death, together with the person's age and gender. It also contains information on the background of the victim including their religion and whether they were civilians, members of either a republican or loyalist paramilitary group, or members of the security forces, a category which mainly consists of the British Army, the Ulster Defence Regiment (UDR), RUC and prison officers. For most of this paper the broad classes of 'security forces', 'republican paramilitary', 'loyalist paramilitary' and 'civilian' will be used with civilians being further sub-divided into Catholics and Protestants. In the vast majority of cases information is also given on the affiliation of the perpetrator of the attack using the same classes.
The database consists of a total of 3524 deaths in the period up to 2001, of which 1541 or 46 percent occurred within the BUA. The victim profile was overwhelmingly male at 94 percent of the total, and young, with a mean age of just 32. Loyalists were responsible for 67 percent of all random killings, in which the circumstances suggested that the victim was selected by chance or where the casualty was the unfortunate victim of an indiscriminate gun or bomb attack. Republicans accounted for approximately 29 percent of these killings while in contrast, they were responsible for 52 percent of 'targeted' killings, as opposed to 30 for loyalist paramilitary groups and seventeen percent for the various British security organisations. Over three-quarters of all victims in Belfast died in gun attacks, with eighteen percent being killed as a result of bomb blasts and just three percent due to blunt force trauma or stabbing. Fig. 1 shows that deaths were also heavily skewed to the early period of conflict, with almost one in five occurring within a single year of the Troubles, 1972. This is important because the temporal spread of fatalities effectively represents the inverse pattern to that of peaceline construction during the conflict, which intensified much later on from the 1990s, well after fatalities had peaked. Such was the intensity of violence in the city in 1972 that a U.S. State Department official wrote to the Belfast consul 'quite incredible e I believe you are now in the most dangerous posting in the world, not excluding Saigon' (Hillenbrand, 1972).
Within Belfast civilians made up a majority of victims at 64 percent, members of the security forces nineteen percent and the combined total for loyalist and republican paramilitaries was seventeen percent. Yet in terms of responsibility for these deaths, between them loyalists and republicans caused 84 percent of all fatalities, with thirteen percent being attributed to the security services. If we focus solely on the victims of paramilitary violence, of the 620 victims killed by loyalists, 71 percent were Catholic civilians, suggesting that these non-combatants presented a useful proxy for the aggressions of Protestant armed groups rather than members of organisations like the PIRA, who were much harder to target (O'Duffy & O'Leary, 1990: pp. 326e327). While republicans were responsible for fewer civilian deaths at 299 or 30 percent of the total, almost two-thirds of these victims were Protestant It is therefore difficult to find fault with O'Duffy and O'Leary's (1990: pp. 325e327) assessment that a PIRA non-civilian 'kill-ratio' of only just over three out of every five deaths constituted a 'success' rate which 'indicates a high degree of incompetence, error or malevolence with respect to civilians'.
The database also provided a textual description of where each killing took place. We used maps and other sources such as contemporary newspaper reports, to allocate a grid reference to each killing based upon where the incident occurred, or failing that, where the victim's body was found. We consider this method to be a more direct measure of conflict intensity than using the home addresses of victims. The primary reason for this has already been identified by Fay et al. (1997: pp. 50, 56); a substantial number of the victims were either not from Northern Ireland (in the case of the vast majority of members of the British Army, for example), or resided in suburban areas well away from the locations of greatest violence (as in the case of most members of the civilian police force, the RUC), thus providing a somewhat misleading picture. With this geographical reference it was then possible to convert the database into a layer for analysis within a GIS.
To summarise the pattern of deaths kernel density smoothing was used to estimate the intensity of fatal violence across Belfast. Fig. 2 shows quartiles of these densities. It indicates that the areas with the greatest number of killings were located within a compact part of inner-city Belfast covering the Catholic neighbourhoods of the Ardoyne, New Lodge, Falls and Protestant Shankill to the west and north of the city centre, but also including the Short Strand and Lower Ormeau, nationalist enclaves immediately to the east and south of the city centre respectively. It also shows the peacelines extant in 2001 and it is apparent that the greatest intensities of violence do indeed appear to be in neighbourhoods bounded by these structures. Fig. 3 develops the socio-economic picture of these areas in terms of both deprivation and segregation using a bivariate Local Indicator of Spatial Association (LISA) (Anselin, 1995: pp. 93e115). This method groups the census grid squares into statistically significant clusters in terms of the two variables, levels of deprivation and segregation, thus allowing us to define areas that are both highly deprived and segregated or less deprived but still segregated. The map shows that the highest levels of fatality were clearly concentrated in a relatively small band of inner-city neighbourhoods which were also characterised by high levels of deprivation and segregation using the bivariate LISA classification technique. However, it was not the case that all highly segregated and deprived parts of the city witnessed high levels of political deaths. Fig. 4 illustrates this, by tracking the density of deaths in these Catholic and Protestant areas of high deprivation and segregation in terms of their distance from Belfast City Hall in the centre of the city. This makes clear that there was a marked decline in levels of fatality as one moves outward from the inner-city areas (1 and 2 km bands) towards the outskirts even in areas characterised by high levels of deprivation and segregation. What this analysis means is that deaths were a feature only of particular highly segregated and deprived inner-city areas, while neighbourhoods of similar socio-economic profile in the suburbs saw far fewer fatalities.
The patterns of fatal conflict around these central areas become all the more remarkable when we consider the major population shifts that occurred in Belfast over the period of the conflict. In the three decades between 1971 and 2001, the population of the BUA fell by approximately 25 percent as a massive slum clearance programme moved people out of the old segregated inner-city neighbourhoods to peripheral estates and new towns across and beyond Greater Belfast (Brett, 1986;Mathew, 1964;Melaugh, 1994). These momentous demographic shifts frequently resulted in the mere transplantation of the old urban patterns of sectarian settlement to new semi-rural spaces (Bardon, 2005: pp. 791e792;Darby, 1971;Darby & Morris, 1971;Osborne & Singleton, 1982: pp. 179e 180;Simpson, 1973). However the geography of conflict did not appear to follow suit. Fig. 5 measures Troubles deaths in five-year windows either side of 1971 and 1991 by distance from Belfast City Hall. This shows that relative to the changing population base, death rates actually steepened near the city centre as the population fell despite a dramatic drop in overall levels of killing over the  course of the twenty-year period. The fact that there were no increases in death rates at the edge of the city indicates that population trends seem to have had little impact upon the conflict's geography. It would seem then that demography, deprivation and segregation per se have limited explanatory power in the spatial distribution of Troubles deaths. Fig. 6 is a map of peacelines that were in situ in 2001, overlaid with various radials indicating a number of specified distances from each barrier. The most striking and well-known example in the city is that which separates the Catholic Falls from the Protestant Shankill district (Fig. 7), a gargantuan concrete and metal structure which is up to 18 m high in places, scything through the west Belfast cityscape. While walls such as that in Fig. 7 are strikingly obvious, the process of quantifying peacelines and interfaces is not as straightforward as at first it might appear. Jarman (2002: pp. 22e23) has noted that in some places peacelines make use of barriers in the built environment that are already in place, and which are residentially neutral, such as parks or public utility buildings (Jarman, 2002: pp. 22e23). Due to the variety of forms that peacelines take, and the fact that by no means all of the boundaries between segregated Catholic and Protestant areas are marked by such structures, the analysis in this paper will take two forms. The first will address those material structures or peacelines identified by the Belfast Interface Project (BIP) as existing in 2001 (BIP, 2012). However, it is recognised that there has existed a significant underenumeration of the actual number of barriers that have been built in the city in official statistics, a state of affairs which it has been argued by Melaugh (CAINb, 2012) may be down to the fact that responsibility for their construction, maintenance and quantification has never been vested in one single body. By definition, these extra physical barriers must have been constructed in areas where heavily segregated Catholic and Protestant districts were spatially propinquitous. Therefore, a secondary stage of analysis will examine the pattern of deaths in all interface zones, using a wider spatial formula to define these areas in order to attempt to understand how patterns of violence relate to demographic definitions of the cores and peripheries of highly segregated neighbourhoods.

Spatial problems
In a 2006 article in the Observer newspaper under the headline, 'Peacelines are where danger lurks', the claim was made that 85 percent of killings occurred within 1000 yards of the walls (McDonald, 2006), a feature which both helped to construct and reflect the place of peacelines in the urban terror imaginary. Whilst an accurate and stark observation, the use of such radials of proximity to draw associations between peacelines and political deaths raises deeper questions. If we refer back to Fig. 3, the contested religio-political geography of central Belfast is in reality so complex that a considerable portion of the inner city can be seen to lie within reasonably close proximity to a peaceline. Bollens (1999: p. 209) has accurately described this as a, 'hyper-segregated sectarian and peaceline geography'. Indeed, this has also led another leading expert on peacelines and interfaces in Northern Ireland to remark that this renders the use of such bandwidths of association to be 'potentially meaningless' (Jarman, personal communication, 2012). This in turn is a reflection on the capabilities and limitations of fundamentally interrogative technologies such as GIS; it is excellent at identifying and visualising seemingly obvious patterns, but these visualisations are often better at raising questions and challenging existing hypotheses than they are at providing answers (Gregory & Healey, 2007: pp. 638e653).
The best way to illustrate the potentially distorting power of scale is to map what these values mean on the ground. Fig. 6 shows that by applying a Euclidian bandwidth of 500 m to peacelines we effectively cover the entire urban area of inner-city Belfast, with the exception of a narrow strip land of no more than 200 m in width running down the middle of the Shankill district. Expanding the radius to 1000 m (which approximates closely to the 1000 yard distance used in the 2006 Observer article), means that we completely cover all of west and central Belfast, in addition to vast swathes of the north and east of the city, stretching beyond the western extents of the urban boundary. Due to the highly-complex religio-political geography of inner-city Belfast, the result of using such coarse bandwidths to analyse peaceline fatalities is to effectively imply that all deaths which accrued from the political conflict across the entire time period can be reduced to interface hostility.
If we explore in much closer spatial detail the relationship between Troubles deaths, background population and peaceline proximity (Fig. 8), we find that while death rates were indeed high in immediate proximity to peacelines, they actually increased as distance increased, peaking some 200e300 m away from the structures. While these distances may appear to be relatively small, we must remember that against the intricate religio-political geography of inner Belfast this constitutes the difference between the cores and peripheries of highly segregated communities such as the Ardoyne, Falls and Shankill.

Temporal problems
The temporal dimension of the relationship between peacelines and Troubles deaths is even more unsettling than the spatial. Fig. 1 clearly identified that deaths during the conflict were heavily skewed towards the early phase of the conflict in the 1970s (Bew & Fig. 4. Mean death rates per km 2 by distance from Belfast City Hall for Catholic and Protestant bivariate LISA clusters of high deprivation and high segregation.  Gillespie, 1999;p. 57;McKittrick, Kelters, Feeney, Thornton, & McVea, 2007: p. 137;O'Leary & McGarry, 1993: p. 31). However the BIP (2012: p. 12) identify that the majority of peacelines were built in the 1990s, well after deaths had peaked. Indeed, more fortifications were built in the second half of the decade following the ceasefire declared by the PIRA in 1994 than were built in the years preceding it (BIP, 2012: pp. 12e13). While erection dates for most barriers remain elusive the respected CAIN service has conducted some research on these and has managed to identify specific years of construction for a number of locations around the city (CAINb, 2013). This provides the basis for a temporal analysis of patterns of fatality before and after the walls were put in place. The locations of peacelines for which the year of construction is known are shown in Fig. 9. Fig. 10 aggregates deaths at these locations at a range of spatial scales in the three years immediately preceding and following the building of the barriers. The graph clearly shows a marked decline in deaths in proximity to the peacelines in the years immediately following their construction. Indeed, deaths in proximity to peacelines peaked in the year before they were built and declined sharply and consistently thereafter. In part, this is a reflection of the broader temporal pattern already identified with deaths heavily skewed to the early part of the conflict (Fig. 1), but bearing in mind that most barriers date from the 1990s, these statistics, together with the spatial evidence, appear to question the notion that peacelines may have acted as 'magnets' for violence (Gaffikin & Morrissey, 2011: p. 209). Indeed, it would seem that they had the opposite effect and may have precipitated a decline in fatal levels of Fig. 6. Peacelines, distance radials and conflict deaths.   Mean (1971Mean ( , 1991Mean ( and 2001 population-based death rates and peaceline proximity. Fig. 9. Locations of peacelines for which year of construction is known. violence in the areas in which they were built, or at least in possibly displacing these killings to other areas.

Widening the focus: defining the interface
The shortcoming of all this analysis so far is that we have only addressed the materiality of the religio-political divide in Belfast. While peacelines have proliferated over the last twenty years it is true to say that not every boundary between highly segregated Catholic and Protestant areas is marked by a physical barrier. To address this problem, the window of analysis will be widened to incorporate interface areas more generally. Whilst it is illogical to have a peaceline without an interface, it is quite possible to have an interface without a peaceline. Indeed, many parts of the city are marked by such informal barriers, which might take the form of serendipitous physical features such as rivers or streams, or be the result of subtle urban planning interventions as with the construction of small industrial units to buffer the Protestant Tiger's Bay from the Catholic Lower Antrim Road (Bollens, 1999: p. 127;Jarman, 2002: p. 23). It may even be something as seemingly innocuous as 'a turn in the road, a local landmark or a row of shops' (Jarman, 2005: p. 6). There is no official criteria for what actually constitutes an interface area, although the CAIN website defines it as 'The boundary between Catholic (Nationalist) and Protestant (Unionist) areas, especially where two highly segregated areas are situated close to each other' (CAINc, 2013). According to Jarman and O'Halloran (2001: pp. 3e4) the Northern Ireland Housing Executive (NIHE) determine segregated areas as those with Catholic or Protestant populations of 90 percent or more. Using the 100 m grid squares it is possible to bring the demographic dimension into the analysis with remarkable spatial granularity.
To incorporate these areas into the analysis, all polarised Catholic and Protestant 100 m 2 grid cells in the city in either 1971, 1991 or 2001 (i.e. those with populations of one group or another equal to or in excess of 90 percent of the total), and with centroids (centre points) lying within 100 m of an equally polarised cell dominated by the other group, have been defined as 'interface cells'. This is in addition to all cells within 100 m proximity to a peaceline, regardless of whether those cells are populated or not. The definition of interface here covers approximately eighteen percent of the city's populated space. As Fig. 11 shows, this provides a wider spatial definition that embodies both the material and demographic boundaries which define the city. As is clear, much of the inner city forms a large, contiguous interface, running from the Ardoyne in the west and arcing through a wide swathe of inner-city north and west Belfast. Other substantial interface zones are to be found in and around the Short Strand in east Belfast and Andersonstown in the far south-west. Strikingly though, it is no longer simply the traditional working-class areas of the city which fall within this classification; more affluent parts of north Belfast now also come within this definition of the interface. This accords with qualitative research which has found that during the first half of the 2000s, the meaning of the interface seemed to be morphing to incorporate areas of high owner-occupation (Jarman, 2005: p. 6). Such largely lower-middle class areas were previously considered outside the reach of inter-communal hostility.
Using the same methodology as employed for peacelines, Fig. 12 shows death rates based on mean population across the period of the Troubles and distance from interface zones. Again, a very similar pattern emerges. Death rates increase outside the interface markedly, peaking sharply in the band 100e200 m outside. Once more, it is important to emphasise that while the distances involved appear small, they are Euclidian, and in terms of the elaborate patterns of religio-political segregation in inner Belfast, these areas frequently represent the difference between the cores and peripheries of polarised communities. 2 So in terms of the most extreme measure of political and social unrest, interface areas and those in immediate proximity to peacelines which equate to the peripheries of segregated communities were not as violent as the cores or 'sanctuaries' of respective Catholic and Protestant enclaves. Furthermore, this relationship holds across a variety of individual status groups during the Troubles (Fig. 13). All groups registered higher death rates beyond the interface zone. It is however interesting to note that deaths among Catholic civilians rose nearer to interface zones, a trend that is not found in any other group. This may reflect the greater tendency for loyalist paramilitaries to target Catholics in ethnically marginal spaces within Belfast at certain times during the conflict, particularly in the early-to mid-1970s (Rose, 1972).

The breakdown of the interface and the 'sanctuary': comparative and policy implications
The patterns uncovered in the above analyses challenge current orthodoxies on the relationship between peacelines, interfaces and geographies of extreme violence during the Troubles. They also challenge us to explain them. Why did the patterns emerge and more importantly, why do they matter?
The patterns of fatality identified in this paper do support the work of Feldman (1991) in which he identified that the 'sanctuary/ interface' was a social construct designed by residents to protect their respective Catholic and Protestant enclaves, but it was one that collapsed as both sanctums became the primary locations of fatal violence and the interfaces were left to fester as the 'dead zones' of dereliction and low-intensity conflict. As time wore on both loyalist and republican paramilitaries found it increasingly difficult, or rather impossible to offer spatial protection to 'their' communities and were thus confined to retaliatory strikes against each other's sanctuaries (Northern Ireland Cabinet, 1971). The inability of either paramilitary grouping to offer comprehensive defence to their respective sanctuaries led to an iterative and intensifying cycle of violence. In October 1993 a PIRA bomb exploded prematurely in a busy fish shop on the Shankill Road, above which senior loyalist paramilitaries had been scheduled to meet (Wood, 2006: p. 170). The bomb killed ten people including the PIRA bomber himself, eight other victims being innocent Protestant shoppers and just one being a member of the loyalist Ulster Defence Association (Bew & Gillespie, 1999: p. 277). Fig. 10. Deaths within three spatial ranges in the three years immediately prior to, and following the construction of the peacelines identified in Fig. 9.
However, the backdrop to this assault on the Protestant sanctuary of the Shankill Road was a series of loyalist attacks on Catholic enclaves: The loyalist onslaught directly challenged the IRA's raison d'être, and the pressures on the organization in Belfast to respond were enormous. After all, the Provisionals had come into existence to defend the city's vulnerable Catholic communities, and any failure to hit back at those behind the attacks could provoke uncontrollable freelance operations by IRA units (Moloney, 2002: p. 414).
It is vital to understand the true role of the interface during the conflict because it has profound implications for how we conceptualise the Troubles as a conflict. Interpretations of conflict are not neutral but constitutive and the implicit assumption that the focus on interface hostility potentially sets up is that deaths during the conflict primarily occurred as a result of the sort of ongoing territorial antagonisms which have come through in the archival materials and secondary literature that this paper has dealt with. Certainly interface zones were, and indeed remain places of considerable tension between the Catholic and Protestant populations (Alleyne, 2012;McDonald, 2012aMcDonald, , 2012bMoriarty, 2012). However, those tensions have historically tended to manifest themselves in terms of frequent low-intensity conflicts, which albeit had the potential to escalate into more serious confrontations. Rapidly after the initial outbreak of violence in 1969 the sanctuaries rather than the interfaces increasingly became the sites of fatal violence between the protagonists. Deaths arising from lowintensity stand-offs were comparatively rare during the Troubles and the vast majority of victims died, either deliberately or sometimes accidentally, as the result of planned actions and spatial transgressions by political or nominally-political actors such as paramilitaries and the security forces.
If the focus on material boundaries in relation to conflict deaths has been understandable but perhaps misplaced, this has clear implications for the frequency with which Belfast is compared to other urban sites of ethnic conflict around the world, most notably Jerusalem. The ongoing Israeli West Bank barrier, which runs through and around Jerusalem bears little resemblance to Belfast's peacelines in either its conception, reception or purpose. Bollens (2013: pp. 186e206) argues that in Jerusalem, as in Sarejevo and Beirut, the fortification was designed to insulate the entire urban space from exogenous threats of ethnic violence, and by constricting access to the city, to neutralise the dangers emerging from suspect minorities within. In Belfast, this was not the case as the peacelines are piecemeal and partial, still reflective of the highlycomplex religio-political geographies that exist within the city, while Jerusalem has since been nominally 'reunified' (Kliot & Mansfield, 1999: pp. 217e218) even if it remains highly segregated in reality. Even more significant is the historical fact that the construction of Belfast's peacelines has been largely consensual, an acknowledgement of pre-existing and long-standing ethnic divides within the city. In Jerusalem on the other hand, the building of the current West Bank barrier has clearly acted to marginalise the city's minority Arab population (Asser, 2010;Bollens, 1999: pp. 95e97;Tait, 2013), while in Belfast it could be said that working-class Catholic and Protestant neighbourhoods have been equally stigmatised by the construction of these barriers. It is symptomatic that it is their potential removal that is leading to tensions rather than their erection because for better or worse, these peacelines have contained the demographic pressures within the city that it will be argued are a more serious threat to inter-group stability (Boal & Douglas, 1982: pp. 339e340;Bollens, 1999: p. 108). This research therefore clearly has important policy ramifications in the light of contemporary political moves towards the removal of peacelines (Blevins, 2013;Buckler, 2013).
If peacelines and interfaces were not therefore the epicentres of political fatality during the Troubles, they were, and continue to be the focus for an ongoing, lower-intensity form of conflict, and it is through an understanding of interface population dynamics over the course of the Troubles that we can start to understand what the stakes really are in this regard. It is generally agreed in the academic and public spheres that violence has precipitated population change in Northern Ireland (Boal, 1987: p. 118;Darby, 1976: p. 25;Kennedy & Birrell, 1978: p. 104). However, it is likely to be in the mundane and quotidian manifestations of inter-communal hostility that the best explanations lie for the patterns of demographic decline in interface zones that come through in Fig. 14. The graph shows a remarkably clear association between rates of population loss and proximity to the interface, and while most of the inner city suffered dramatic population decline during the Troubles, that decline was most acute in the interface and in areas closest to it. Yet while such loss might appear extraordinary at first glance, in the wider comparative context of post-industrial, north British cities it is much less so. Over the same thirty year period, the core population of Liverpool and Manchester also fell by a quarter, Glasgow by considerably more (CDU, 2013). So while there is a tendency to think of the experience of Belfast or Northern Ireland more generally as exceptional (Aughey, 2005: p. 63), in terms of transnational processes of urban decline and suburbanisation its experience is far from unique.
However, what does set the city apart is the fact that those patterns of decline have not affected the Catholic and Protestant populations of Belfast equally. Fig. 15 shows that population loss in the interface and surrounding zones has dramatically and Fig. 12. Mean (1971 population-based death rates and interface proximity. Fig. 13. Population-based percentage difference in observed versus expected death rates by categories of victim against distance from the interface.  disproportionately affected the non-Catholic population of the city. 3 While non-Catholic population loss across the city as a whole was drastic at around 45 percent, in five of the seven interface proximity bands it outstripped even that figure. This in turn has had an interactive effect on the dynamics of interface hostility during the Troubles and beyond, with Bollens (1999: p. 108) arguing that Protestant population decline in interface areas has been a far greater cause of antagonism than any motivations accruing from the existence of peacelines and/or interfaces in and of themselves. So the interface and surrounding areas delineates a space which is not just a political battleground but a demographic one; it has become a wry truism in Northern Irish political circles that republicans were too shrewd to admit they had lost the war, while loyalists were too foolish to realise they had won (Campbell & Hagerty, 2011: p. 66;O'Doherty, 2008;Patterson, 2012: p. 249). However, if Protestants more generally had won the political struggle, they had unquestionably lost the demographic one on the interface. An example of just such an area comes in the Torrens estate, which is surrounded by much larger Catholic districts such as the Ardoyne and Oldpark. Here the small remaining Protestant population was, depending on your viewpoint, either finally purged or chose to vacate the area en masse during the summer of 2004 as the sectarian boundaries continue to harden (BBC, 2004;McKittrick, 2004): Ms Ferguson's 91-year-old grandmother, Betsy McClenaghan e who was also moving e told one reporter: "I have come through the blitzes and the 1932 riots but I've never known an upheaval like it." The intimate intensity of the cruelty in such places can be viscerally shocking. Earlier this year, in a Belfast newspaper, a 15-year-old Torrens resident called Sarah Kell e whose father, Trevor, a taxi-driver, was murdered by a republican gunman e described how certain local Catholic schoolchildren regularly shouted "Where's your da?" at her. Graffiti on their street echoed the taunts: "Sarah Kell, where's your da, ha ha ha" (McCartney, 2004).

Conclusions
This paper has presented new empirical research which challenges current thinking on the spatial relationship between peacelines, interfaces and conflict deaths during the Troubles in Belfast. It has argued that existing interpretations of that relationship, while ostensibly accurate, have failed to factor the spatial and religio-political complexity of Belfast into their calculations, and in the process overlooked the fact that death rates were higher in the heartlands or 'sanctuaries' of respective Catholic and Protestant enclaves. As Shirlow (2003b: pp. 81e82) has already rightly identified, peacelines were built 'around those arenas within which the most disproportionate patterns of violence transpired' [our emphasis]. It appears that within those arenas more complex dynamics were at play. Through a novel methodology it has sought to give quantitative substance to the nebulous spatial concept of the 'interface' and the resulting analyses have confirmed the patterns of fatality observed in relation to peacelines, with death rates once again higher in zones outside the interface. This is not to argue that peacelines or interfaces have no role to play in explaining patterns of fatality during the Troubles; it may well be the case that these social and physical structures can indeed be associated with distributions of deaths, acting to channel patterns of aggression and thereby rendering certain localities more vulnerable than others (McKittrick et al., 2007(McKittrick et al., : pp. 1084e1085, 1090e1091, 1166(McKittrick et al., , 1250, but this is an area that clearly requires further research. In a longerterm historical context, interfaces have acted in the Northern Ireland conflict as second-order or performative centres of low-intensity violence rather than fatality, and they continue to operate as the battlegrounds for ongoing, indeed intensifying demographic disputes. These findings matter for a number of reasons. A wealth of research cited here has established the economic, social and environmental harm caused by peacelines. However, the evidence presented in this paper suggests that it is little wonder that is those who live nearest to the walls that are most reticent to see them removed (Byrne et al., 2012: p. 13); over the course of the Troubles these areas witnessed lower levels of fatality than in neighbouring 'sanctuary' zones and it is therefore far from irrational for these residents to associate their sense of security with proximity to the peacelines. In the current political context and for the sake of communities blighted by peacelines, decisions regarding their future removal need to be based on a closer reading of the spatial evidence, and a clearer understanding of the historical contexts and legacies of conflict from which they emerged. Taking the long view, it appears that certain parts of Belfast (Cunningham, 2013: pp. 58e 59), rather than particular types of neighbourhood have acted as centres of violent conflict despite radical changes in the urban fabric and population base over time. Above all else, the role of peacelines and interfaces matters because 'meta-conflicts' matter (Brubaker, 2002: pp. 173e174;Horowitz, 1991: p. 2) and the focus on interfaces and peacelines as the primary sites of fatal violence can provide (albeit unintentionally) a distorting and constitutive framing which implies that deaths during the Troubles arose primarily out of localised territorial disputes rather than much deeper political antagonisms.