Previous PageTable Of ContentsNext Page

Recent Reports 
 Support HRW 
About HRW
Site Map

Human Rights Watch - Home Page



Operation Allied Force began on March 24, 1999 after more than a year of effort by the international community led by NATO to find a negotiated solution in Kosovo. In June 1998, NATO Defense Ministers decided to charge NATO planners with the responsibility to produce a range of options, both ground and air, for military action should the diplomatic process fail to yield the desired results. By the fall, an estimated 250,000 Kosovo Albanians had been driven from their homes and some 50,000 were threatened by approaching winter weather.2 The United Nations Security Council adopted resolution 1199 (UNSCR 1199) on September 23, highlighting the impending human catastrophe and demanding a cease-fire and the start of real political dialogue. A Contact Group meeting in London on October 8 gave U.S. envoy Richard Holbrooke a mandate to secure agreement to the requirements of UNSCR 1199 in a mission to Belgrade. Activation orders for air strikes were agreed on October 13; that same day Holbrooke reported to NATO that Slobodan Milosevic, the president of the Federal Republic of Yugoslavia (FRY), had agreed to the deployment of an unarmed Organization for Security and Cooperation in Europe (OSCE) verification mission to Kosovo and to the establishment of a NATO aerial verification mission. Yugoslavia also agreed to reduce the numbers of security forces personnel in Kosovo to pre-crisis levels.

Despite initial stabilization, violence continued. Following a massacre in the village of Racak on January 15, 1999, NATO increased its state of readiness, issuing a "solemn warning" to Milosevic and the Kosovo Albanian leadership on January 28.3 This was followed by a second statement on January 30 that reaffirmed NATO's original demands, and delegated to Secretary General Javier Solana authority to commence air strikes against targets on FRY territory.

Parties to talks at Rambouillet in France, in February 1999, attempted to build agreement to protect the rights of all sides. After the first round of talks was suspended on February 23, a second round was convened on March 15. This second round was suspended on March 19 in the light of what NATO intelligence and OSCE observers saw as intensifying violence on the ground instigated by FRY security forces, and a build-up of FRY/Serbian forces in and around Kosovo. OSCE verifiers were withdrawn during the night of March 19-20, and Holbrooke flew to Belgrade on March 22 in a last-ditch effort to persuade Milosevic to back down and avoid a military confrontation. On March 23, following final consultations with allies, Javier Solana directed NATO's Supreme Allied Commander Europe (SACEUR), Gen. Wesley Clark, to initiate a "phased" air operation.4

Operation Allied Force Attacks

Operation Allied Force was initiated at 7 p.m. GMT (8 p.m. local time in Yugoslavia). Of thirteen (out of nineteen) NATO nations that made aircraft available for the operation (Belgium, Canada, Denmark, France, Germany, Italy, the Netherlands, Norway, Portugal, Spain, Turkey, the United Kingdom, and the United States), eight put their planes in action on the first night. Aircraft from the United States, the United Kingdom, France, Canada, and Spain conducted bombing, carrying out a succession of attack waves with almost exclusively precision-guided munitions (PGMs) against fixed and pre-selected targets. Long-range cruise missiles were fired by the United States and Britain. Though targets were hit throughout Yugoslavia across a mix of target types (for example, airfields, command and control sites, barracks, andheadquarters, particularly of the special police), the initial focus was almost exclusively an effort to neutralize the Yugoslav air defense system. In the first day, NATO hit fifty-three targets, largely air defenses and radar sites.5

The mission of Operation Allied Force, in General Clark's words, "was to halt or disrupt a systematic campaign of ethnic cleansing."6 Attacks would be along two lines, a "strategic attack line operating against Serb air defenses, command and control, VJ [Yugoslav Army] and MUP [Ministry of Interior] forces, their sustaining infrastructure, supply routes, and resources," and "a tactical line of operation against the Serb forces deployed in Kosovo and in southern Serbia."7 The initial attacks against air defenses and command and control elements were intended to "set the conditions for moving on up [the hierarchy of targets] to [include] the forces in the field."8

Following the attacks on air defenses and command and control centers, NATO chose targets to isolate Yugoslav forces and constrain their movement. According to Chairman of the Joint Chiefs of Staff Gen. Henry Shelton, this included their "ability to move both horizontally [and] laterally on the battlefield, the road and bridge network, which was key to that, and also...sustainment, particularly the POL [petroleum, oil, and lubricants], to start causing shortages, since this was a predominantly mechanized armored force."9

Although there were expectations on the part of some, including evidently many political leaders in NATO governments, that Allied Force would be a short campaign, the U.S. Department of Defense stated that it "made clear to our allied counterparts that Operation Allied Force could well take weeks or months to succeed."10 Regardless of this post-war claim, NATO operations began with just a limited number of cruise missile and air strikes. The carefully planned "phases" were quickly melded together and expanded to accommodate political and public sensitivities, as well as to escalate the intensity of operations to make progress towards forcing Yugoslav submission. According to U.S. Secretary of Defense William Cohen, "soon after the conflict began, entire classes of targets were delegated for approval by NATO's military commanders. And only certain sets of targets, such as those in downtown Belgrade, in Montenegro and those with a high likelihood of civilian casualties, were reviewed by the allied capitals and by higher political authorities."11

At the NATO summit in Washington on April 23, 1999, one month into the air war, alliance leaders decided to intensify the air campaign by expanding the target set to include military-industrial infrastructure, news media, and othertargets considered to be of a strategic nature.12 More aircraft and weapons were deployed in the theater of operations, and there was an intensification not only in the rate at which targets were hit, but also a shift from an initial eight-hour day to a twenty-four-hours a day campaign.13

With an increasing force and greater intensity of attacks, there were also increasing attacks on Yugoslav forces in and around Kosovo. However, by and large, the focus into the second month of bombing continued to be attacks on objects that would cut the supply lines and support infrastructure of the military forces. Not only was poor weather a prohibitive factor in mounting attacks on mobile forces, but NATO had to "learn" the Kosovo geography and the organization of Yugoslav forces. It was many weeks before it was able to track forces on the ground, identify key elements, predict their movements and activities, and attack them in urban settings. Nevertheless, NATO's air attacks, both against "strategic targets" and in the south, slowly had an accumulating impact on Yugoslav military operations. Air activity forced Yugoslav forces to remain largely hidden from view, traveling only under limited circumstances.14 Over time, attrition of heavy equipment accelerated, peaking at about the last week in May.

In the first month of Operation Allied Force, NATO reported that it averaged around 350 sorties per day, with nearly 130 attack sorties. By the fourth week, it was flying nearly two-and-a -half times the number of attack sorties per day than it flew during the first three weeks.15 NATO reported in early July that it had flown a total of 37,465 sorties, of which 14,006 were strike and suppression of air defense (SEAD) sorties and 10,808 were strike-attack sorties.16 By the end of the conflict, NATO had attacked over 900 targets.17

As more NATO forces were introduced and the attacks continued, the percentage of PGMs being used also declined. In the early days of Allied Force, "smart" weapons constituted more than 90 percent of the ordnance employed. By mid-May, this had declined to only 10 or 20 percent of the total, with guided weapons constituting about 35 percent of the 26,000 weapons employed throughout the course of the war.18

From the very beginning of Operation Allied Force, minimizing civilian casualties was a major declared NATO concern. According to NATO, consideration of civilian casualties was fully incorporated into the planning and targeting process. All targets were "looked at in terms of their military significance in relation to the collateral damage or the unintended consequences that might be there," General Shelton said on April 14. "Then every precaution is that collateral damage is avoided."19 According to Lt. Gen. Michael Short, "collateral damage drove us to an extraordinary degree. General Clark committed hours of his day dealing with the allies on issues of collateral damage."20

Though a couple of dozen incidents would dog NATO throughout the war in its press and propaganda battles with the Yugoslav government, from another perspective, the limitation of "collateral damage" was a political imperative to successful conclusion of an alliance war. In the words of Lt. Gen. Marvin R. Esmond, the senior Air Force operationsofficer, "NATO's success with precision engagement and minimal collateral damage was a key factor in holding the Alliance firmly together during the bombing."21

Documenting and Assessing the Civilian Toll

Because of keen public interest in the civilian toll from Operation Allied Force, Human Rights Watch assumed a major undertaking to document and evaluate the impact and effects of the NATO military operation. Human Rights Watch military consultant William M. Arkin and researcher Bogdan Ivanisevic conducted extensive research into the operation. During the war, they compiled a master chronological database from military sources, Yugoslav media and Internet reports, collating these with press and governmental reporting from the NATO countries. Tanjug (official Yugoslav news agency) and Yugoslav television and radio dispatches were monitored on the Internet and via the Foreign Broadcast Information Service (FBIS) and BBC Summary of World Broadcasts. The researchers corresponded with Yugoslav civil defense, military, and information ministry officials via E-mail, and scoured Yugoslav websites, particularly those maintained by official agencies. They also comprehensively monitored the Yugoslav press from March-June 1999, including: BLIC (Belgrade independent daily), Politika (Belgrade pro-government daily), Politika Ekspres (Belgrade pro-government), Vecernje Novosti (Belgrade pro-government daily), Glas Javnosti (Belgrade independent), Dan (Montenegrin pro-Belgrade/SPS daily), Pobjeda (Montenegrin pro-Podgorica government daily), and Vijesti (Montenegrin independent daily).22

Between August 2 and August 20, 1999, Arkin and Ivanisevic conducted a bomb damage assessment mission in Serbia and Montenegro. Human Rights Watch Executive Director Kenneth Roth accompanied the team on August 2-5. In twenty days, the team drove approximately 5,000 kilometers, visited ninety-one cities, towns, and villages, and inspected well over 250 sites (targets, reported targets, areas of civilian damage, stray craters, etc.). They met with officials from a dozen ministries in Belgrade, and in other locations met with regional, municipality, factory, and utility representatives. Taking eyewitness testimony and inspecting bomb damage, they were able to verify individual events and assess the veracity of wartime and post-war reporting.

Human Rights Watch confirmed ninety incidents in which civilians died as a result of NATO bombing (see Appendix A). The field mission visited forty-two of the ninety confirmed incident locations and collected primary source information on thirty other incidents. Sufficient corroborating information existed on twenty-two others to recognize their credibility (including five in which NATO has officially confirmed that it attacked nearby targets at the same time). Eight additional reported and claimed incidents have been eliminated altogether, three because they could not be verified or there was little corroborative reporting,23 and five because the reported deaths are actually presumed to be paramilitary policemen or soldiers (see Appendix C).24

Human Rights Watch has also assessed the veracity of information compiled by the Yugoslav government, including autopsy reports, death certificates, and photographic evidence of bomb damage and casualties. The government's two-volume White Book, NATO Crimes in Yugoslavia, provides information on seventy-five of ninety incidents. Other releases by the Ministry of Foreign Affairs and the Ministry of Health document, in a less comprehensive manner, other aspects of the civilian effects of the bombing. Human Rights Watch also met with representatives of the governmental Committee for Compiling Data on Crimes Against Humanity and International Law, which is compiling dossiers on eachinstance of civilian casualties, as well as the government Reconstruction Commission, which is responsible for repair of the public infrastructure.

The findings of Human Rights Watch's field operation facilitated a critical review of governmental and private reports from both sides of the conflict. Systematic on-site inspections facilitated the cross-checking of information compiled from press and Internet reports, as well as providing the basis for evaluating the detailed reporting on casualties by the Yugoslav government and private agencies. Our inspection of bomb damage and interviews with witnesses, survivors, and others enabled us to assess the accuracy of detailed reporting on deaths and bomb damage produced on the same incidents, for example, by the Yugoslav and other government sources. On the basis of spot-checking in the course of our own field research and correlation with other sources, some of these documentation sets, notably the White Book and the Ministry of Health photographic record, have been found to be largely credible.25

Civilian Deaths as a Result of Attacks

Human Rights Watch concludes that as few as 489 and as many as 528 Yugoslav civilians were killed in the ninety separate incidents in Operation Allied Force. In sixty-nine of the ninety incidents, the precise number and the names of the victims are known (see Appendix B). In another nine incidents, the number of victims is known and some of the names have been confirmed. In nine incidents, the number of victims is known but the names are unknown. In three incidents, the names and precise numbers of victims are unknown.26

Between 62 and 66 percent of the total registered civilian deaths occurred in just twelve incidents (see Table 1). These twelve incidents accounted for from 303 to 352 civilian deaths, based on the best available information. These were the only incidents among the ninety documented in which ten or more civilian deaths were confirmed.

Information drawn from the ninety incident reports allows a general picture to be drawn of the civilian deaths by the time, place, and circumstances in which they occurred. The deaths resulted from attacks on a range of targets, under different circumstances, and from a variety of munitions. Fifty-five of the incidents occurred in Serbia (including five in Vojvodina), three in Montenegro, and thirty-two in Kosovo. But between 279 and 318 of the dead-between 56 and 60 percent of the total number of deaths-were in Kosovo. In Serbia, 201 civilians were killed (five in Vojvodina) and eight died in Montenegro. A third of the incidents-thirty-three-occurred as a result of attacks on targets in densely populated urban areas.

Human Rights Watch was able to determine the intended target in sixty-two of the ninety incidents (68 percent). Of these, the greater number of incidents was caused as a result of attacks on military barracks, headquarters, and depots; thirteen were a result of attacks on bridges (and one tunnel); six resulted from attacks on telecommunications and air defense facilities; five each resulted from attacks on industrial facilities, oil installations, and airfields; and seven were as a result of attacks on convoys or on what were perceived to be military forces in the field. These latter incidents were the most deadly, while two of the ten worst incidents occurred as a result of attacks on bridges.

Almost half of the incidents (forty-three) resulted from attacks during daylight hours, when civilians could have been expected to be on the roads and bridges or in public buildings which may have been targeted. Overall, forty incidentsoccurred in April, forty-five occurred in May, four in June, and one in March. May 29 saw the most incidents (five), followed by April 14, May 30, and May 31 (four each).

Human Rights Watch was able to determine the weapon involved in the cause of the civilian deaths in only twenty-eight of the ninety incidents. Of these, twenty-one are incidents in which it can be confirmed that precision-guided munitions (PGMs) were used (though there could be others). This includes all of the attacks on bridges or targets in and around the Belgrade area. Cluster bomb use can be positively determined in seven incidents (another five are possible but unconfirmed).27 In almost all of the other instances, we have been unable to establish the weapon used.

Countervailing Claims

One disturbing aspect of the matter of civilian deaths is how starkly the number of incidents and deaths contrasts with official U.S. and Yugoslav statements. Speaking on September 9, 1999, Secretary of Defense William Cohen said: "Of the thousands of bombs that were dropped and the missiles that were fired, nearly all of them hit their intended target. Of all those thousands of weapons that were dropped and expended, approximately 20 had unintended consequences or were not on target."28 Gen. Wesley Clark, commander of NATO forces, in the war, stated on August 31 that there were just "20 incidents of collateral damage" in the entire war.29 Deputy Secretary of Defense John Hamre testified before Congress on July 21 that "Out of the 9,300 strikes [sic], we had 30 where we killed people...30 where we had damage we hadn't intended."30 In October, General Clark again repeated that there had been twenty incidents: "I just want to emphasize the incredible precision of the bombing; the fact that on 78 days, with over 23,000 weapons dropped or fired, there were only 20 incidents of collateral damage...that's an incident rate of less than 1/10th of 1 percent."31

However, the number of confirmed deaths is considerably smaller than both U.S. and Yugoslav public estimates. The post conflict casualty reports of the Yugoslav government vary, but coincide in estimating a civilian death toll of at least some 1,200 and as many as 5,700 civilians. On May 22, Margit Savovic, president of Yugoslav Committee for Cooperation with UNICEF said that "more than 1,200 civilians were killed and more than 5,000 [were] wounded."32 On July 14, Milovan Zivkovic, director of the Federal Office of Statistics, said at press conference that "estimations [of] about 1,200 killed have appeared, and some sources talk about more than 5,000 victims, some go even up to 18,000."33 According to the BETA independent news agency, Zivkovic also said that the 1,200 number publicized by the Yugoslav Committee for Cooperation with UNICEF pertained only to those killed during the two and a half months of the air campaign. "But the 5,000 and 5,700 numbers are exact as well, only they cover a longer period of time and various ways of losing life," he said.34 Ambassador Djorde Lopicic, chief of international law at the Ministry of Foreign Affairs (MFA), told Human Rights Watch on August 5, 1999, that 2,000 civilians had died and over 10,000 were injured from NATObombing. At the lower end, even the 1,200 figures is more than twice the civilian death toll of around 500 that Human Rights Watch has been able to verify for the ninety known incidents involving civilian deaths.

While there have been various pronouncements from the Yugoslav government regarding the number of civilian deaths, NATO has been far more silent. There has been only one informal U.S. government or NATO statement regarding the number of Yugoslav civilian deaths from the bombing. General Joseph W. Ralston, vice chairman of the Joint Chiefs of Staff, said in September that "Despite the weight of bombs dropped, Serbian civilian casualties were amazingly light, estimated at less than 1,500 dead."35 This estimate is three times the number calculated by Human Rights Watch.

In the thirty instances acknowledged by the Defense Department, Deputy Defense Secretary John Hamre has provided the only analysis regarding the causes of civilian deaths. In Congressional testimony in July he said of the thirty incidents:

one third were instances where we damaged the target we wanted to destroy, but innocent civilians were killed at the same time. You will recall the time one of our electro- optically guided bombs homed in on a railroad bridge just when a passenger train raced to the aim point. We never wanted to destroy that train or kill its occupants. We did want to destroy the bridge and we regret this accident. As I said, 10 of the 30 instances of unintended damage fall in this category. For the remaining 20 instances, 3 were caused by human error that identified the wrong target, and two were caused by mechanical error by our hardware. In 14 instances we have not yet determined whether the unintended damage was caused by human error or mechanical failure. We will determine that to the best of our ability during our after action assessment. The one remaining ... [is the] bombing of the Chinese embassy. ... [It] was unique in that we had a legitimate target that we wanted to hit; the only problem is we had the target located in the wrong building. To my knowledge, this is the only example of this failing in all of our strike operations.36

The Standards Applied

The conduct of warfare is limited and restrained by the complementary standards of international humanitarian law, the laws of war. International humanitarian law (IHL) applies expressly and uniquely to armed conflict situations, with distinct provisions to regulate international and non-international (internal) armed conflicts. In evaluating NATO's use of military force in the Federal Republic of Yugoslavia, the laws of war provide the most relevant standards. The conflict in Kosovo reached the threshold of an internal armed conflict in the terms of humanitarian law in 1998, so that certain provisions of the laws of war then applied to both government forces and to the armed insurgency. With the initiation of the NATO bombing in March 1999, the conflict in Kosovo and all of the Federal Republic of Yugoslavia involving NATO and Yugoslav forces became an international armed conflict to which the full body of international humanitarian law applied.

The four Geneva Conventions of 1949 are the fundamental building blocks of international humanitarian law. Geneva Convention IV concerns the protection of civilians in time of war. The conventions are among the most widely ratified international treaties, and the norms they establish are largely considered customary international law, that is, norms that have obtained universal recognition and are accepted as binding upon all nations.

Further elaboration of the provisions of the four Geneva Conventions of 1949 is provided in the 1977 additional protocols to the conventions: Protocol I, relating to the protection of victims of international armed conflicts, and Protocol II, relating to the protection of victims of non-international armed conflicts. All of the 188 members of the United Nations are parties to the four Geneva Conventions of 1949; 155 states are parties to Protocol I and 148 to Protocol II. The Federal Republic of Yugoslavia acceded to the four Geneva Conventions on April 21, 1950, and to Protocols I and II onJune 11, 1979. Most NATO members are parties to Protocol I, which applies to the conflict in question: notable exceptions are France, Turkey, and the United States. Although the U.S. has not ratified Protocols I and II, it considers many of their provisions to be applicable as customary international law.37 In addition, the United States and NATO recognize as a matter of policy and have declared that the laws of war (LOW) apply to all cases of armed conflict, even if a state of war is not recognized.38

Protocol I Additional to the Geneva Conventions provides the basis for the evaluation of NATO's bombing. A basic principle of the laws of war is that the civilian population and individual civilians shall enjoy general protection against dangers arising from military operations. This turns in large part on the requirement that attackers must distinguish between civilians and combatants and between military objectives and civilian objects. They must avoid or minimize harm to civilians, and to this end may not attack civilians exclusively or combatants and civilians indiscriminately. Attacks may not be indiscriminate by intent, where the attackers deliberately set out to kill and maim civilians, or through negligence, where those carrying out an attack disregard their obligations to identify a specific military objective, and to take care not to cause disproportionate harm to civilians in attacking it. Damage to civilian objects and civilian casualties that are incidental to lawful attacks on military objectives are known in military terms as "collateral damage."

The most fundamental principle of the laws of war requires that combatants be distinguished from noncombatants, and that military objectives be distinguished from protected property or protected places. Parties to a conflict must direct their operations only against military objectives (including combatants).39 Under Protocol I, Art. 51, paragraph 4, indiscriminate attacks are prohibited. These include attacks that:
· are "not directed against a specific military objective";
· "employ a method or means of combat which cannot be directed at a specific military objective";
· "employ a method or means of combat the effects of which cannot be limited as required" by the Protocol; and
· "in each such case, are of a nature to strike military objectives and civilians or civilian objects without distinction."

Military objectives are defined as "those objects which by their nature, location, purpose or use make an effective contribution to military action."40

The term "means" of combat refers generally to the weapons used; the "method," to the way in which such weapons are used. Casualties that are a consequence of accidents, as in situations in which civilians are concealed within military installations, may be considered incidental to an attack on a military objective-"collateral damage"-but care must still have been shown to identify the presence of civilians. Protocol I, Art. 57 sets out the precautions required, among them to "do everything feasible to verify that the objectives to be attacked are neither civilians nor civilian objects..."; to "take all feasible precautions in the choice of means and methods of attack" to avoid or minimize incidental civilian casualties or damage to civilian objects; and to refrain from launching any attack "which may be expected to cause" such deaths, injuries or damage "which would be excessive in relation to the concrete and direct military advantage anticipated..."41 In its authoritative Commentary on the protocols, the International Committee of the Red Cross (ICRC) is clear on that what is meant by "feasible": "What is to take the necessary identification measures in good time in order to spare the population as far as possible."42

The principle of proportionality places a duty on combatants to choose means of attack that avoid or minimize damage to civilians. In particular, the attacker should refrain from launching an attack if the expected civilian casualties would outweigh the importance of the military objective.43 Protocol I, art. 57 ("Precautions in attack"), para. 2(b) requires those who plan and/or execute an attack to cancel the attack in such circumstances. The ICRC has noted that there isnever a justification for excessive civilian casualties, no matter how important the military target. Moreover, the argument of proportionality can never justify very high civilian casualties and damage whatever the military advantage envisioned: "Incidental losses and damages should never be extensive."44

The ICRC Commentary on Protocol I's art. 57 sets out a series of factors that must be taken into account in applying the principle of proportionality to the incidental effects attacks may have on civilian persons and objects:

The danger incurred by the civilian population and civilian objects depends on various factors: their location (possibly within or in the vicinity of a military objective), the terrain (landslides, floods etc.), accuracy of the weapons used (greater or lesser dispersion, depending on the trajectory, the range, the ammunition used etc.), weather conditions (visibility, wind, etc.), the specific nature of the military objectives concerned (ammunition depots, fuel reservoirs, main roads of military importance at or in the vicinity of inhabited areas etc.), technical skill of the combatants (random dropping of bombs when unable to hit the intended target).45

The Commentary provides a number of examples of the application of this principle:

All these factors together must be taken into consideration whenever an attack could hit incidentally civilian persons and objects. Some cases will be clear-cut and the decision easy to take. For example, the presence of a soldier on leave obviously cannot justify the destruction of a village.
Conversely, if the destruction of a bridge is of paramount importance for the occupation or non-occupation of a strategic zone, it is understood that some houses may be hit, but not that a whole urban area be levelled.46

In researching each of the incidents in which attacks led to civilian deaths we have sought to compile the facts from which to determine the nature of the real or perceived military objectives targeted; any facts relating to the care taken and procedures and criteria employed to confirm the military nature of the targets; analysis done by NATO to determine proportionality of the civilian deaths and the means of attack to the express military objectives; the correlation of civilian deaths to the location and nature of the targets selected; the timing of target selection as a factor in its appropriateness; the methods and conditions under which distinct weapons systems were employed; and the potentially indiscriminate nature of some weapons systems in general and under certain conditions.

In assessing specific attacks, with a view to general observations on the conduct of the air war, the primary issue is whether due care was taken for the protection of civilians. Was the prospect of civilian casualties sufficiently taken into account in the targeting, the weaponry employed, and the means and conditions under which weapons were employed? This involves a review of the selection of targets, and the procedures through which these are determined: Were the military objectives as defined and identified by NATO forces within the terms of humanitarian law? This report addresses the air war only through its cost in civilian lives as an indicator to be taken into account in assessing the larger picture of compliance with international humanitarian law.

Case Studies of Civilian Deaths

The ninety incidents involving some 500 civilian deaths provide a part of the picture from which to consider NATO's conduct of the war (two subsequent Human Rights Watch reports are planned to look in greater detail at targeting in Operation Allied Force and the use of cluster bombs).47 At issue is whether NATO effectively adhered to the humanitarianlaw imperative that the civilian population be protected against dangers arising from military operations. At the core is the principle of civilian immunity from attack and its complementary principle requiring the parties to a conflict to do everything feasible to distinguish civilians from combatants at all times. Several incidents, which accounted for a large proportion of civilian deaths, illustrate various problems faced in NATO actions, and are further presented below.

The most dramatic losses of civilian life from the NATO offensive in Kosovo came from attacks on fleeing or traveling refugees confused with military forces. These included repeated attacks on refugees over a twelve-mile stretch of the Djakovica-Decane road in Kosovo, in which seventy-three civilian refugees died (incident no. 19), attacks near Korisa in Kosovo (incident no. 57), in which as many as eighty-seven civilian displaced persons and refugees died, and two incidents involving attacks on civilian buses, at Luzane (incident no. 41) and Savine Vode (incident no. 46). Another dramatic loss of civilian life followed from an attack on Dubrava prison, which caused nineteen deaths (see below).

In these various incidents involving the deaths of Kosovar refugees, the principal issue is whether every feasible precaution was taken to accurately distinguish civilians from combatants. At the same time, there are questions regarding the decisions to attack on the basis of incomplete and/or seriously flawed information. The public statements by NATO spokespersons concerning particular attacks, and the changes in the way attacks were characterized, also bear some analysis, in particular insofar as they may seek to justify attacks in which civilian casualties were clearly excessive.

Moreover, there is a question as to whether NATO's extraordinary efforts to avoid casualties among its pilots precluded low-flying operations that might have helped to identify targets more accurately. This was and continues to be a major issue in the public debate about Operation Allied Force. For many weeks in the initial stages of the war, NATO airplanes were not flying below 15,000 feet. If the height at which the NATO pilots flew had little to do with identification and attack of the target, than the issue is irrelevant. But if precision would have been greater (and civilian casualties lessened) had NATO pilots flown lower, it could be argued that NATO was "obligated" to have its pilots fly lower.48 In the case of attacks such as those at Djakovica-Decane, in which flying at a higher altitude seems to have impeded a pilot from adequately identifying a target, the conclusion again is that inadequate precautions were taken to avoid civilian casualties.

The incident at Korisa (incident no. 57) also raises important questions of Yugoslav responsibility for some civilian deaths attributed to NATO bombing. In this case, NATO did not apply adequate precautions in executing its airstrikes. But Yugoslav military forces may share the blame for the eighty-seven civilian deaths at Korisa: there is some evidence that displaced Kosovar civilians were forcibly concentrated within a military camp there as a human shield.

Yugoslav responsibility of a more direct kind has been shown for killings at the Dubrava prison that Yugoslav authorities attributed to NATO bombing. Human Rights Watch researchers in Kosovo have found that some seventy-six prisoners there were victims of extrajudicial executions-cold-blooded murder-by Yugoslav forces in the days after NATO bombed the prison. The NATO attack on May 21 was, however, responsible for nineteen deaths at the facility prior to the massacre of prisoners; an earlier NATO attack killed four civilians at the prison (see incidents nos. 60 and 65).49

A third of all of the incidents in which civilians died-thirty-three-occurred as a result of attacks on targets in populated urban areas. Six incidents occurred in Belgrade, Nis, and Vranje (the latter two are towns in southern Serbia). Eight towns had two or three incidents each involving civilian deaths: Aleksinac, Cacak, Novi Sad, Surdulica, and Valjevo in Serbia and Vojvodina, and Djakovica, Pristina, and Prizren in Kosovo. The targets in almost all of these attacks were headquarters or military/police barracks and facilities, and/or factories. In these cases there was little doubt as to the apparent objective of the attack, or that these locations constituted lawful military objectives.

In one case, involving the use of cluster bombs in Nis (incident no. 48), the weapon employed was a decisive factor in the civilian deaths. Nis is one of seven confirmed and five likely incidents involving civilian deaths from cluster bomb use. Altogether, some ninety to 150 civilians died from cluster bomb use by the United States and Britain. In the case of the attack on the Nis airfield on May 7, the technical malfunction of the weapon points to the fact that cluster bombs should not be used in attacks in populated areas, let alone on urban targets, given the risks. After the Nis incident, there was a U.S. executive prohibition on further cluster bomb use.50 Nevertheless, British planes continued to drop cluster bombs, indicating the need for universal, not national, norms regarding cluster bomb use.

In three cases-the bombing of Serb Radio and Television headquarters in Belgrade (incident no. 30), the bombing of the "Marshal Tito" Petrovaradin (Varadinski) Bridge in Novi Sad (incident no. 2), and the bombing of the Belgrade Heating Plant (incident no. 7)-Human Rights Watch questions the legitimacy of the target. Regardless of NATO's legal determination that civilian radio and television were legitimate military objectives because of their role in internal and external propaganda,51 NATO did not take adequate precautions in warning civilians in the attack on the media headquarters, nor did the attack satisfy the legal requirement in terms of proportionality, given that the center was located in a densely populated urban neighborhood and was staffed twenty-four hours. After strikes on the Belgrade headquarters, moreover, Yugoslav state broadcasters were easily able to move operations to private and makeshift facilities.52 Similarly, in the case of the 04:35 a.m. attack on the New Belgrade Heating Plant on April 4, in which one civilian (the night watchman) was killed, NATO issued no warning and attacked a target located in an urban area.53 The risks involved to civilians in undertaking the two Belgrade urban attacks were grossly disproportionate to any perceived military benefit.

The attacks on the Novi Sad bridge and six other bridges in which civilian deaths occurred (Ostruznica, incident no. 37; Trstenik, incident no. 39; Nis, incident no. 51; Vladicin Han, incident no. 55; Pertate, incident no. 71; and Varvarin, incident no. 81) also were of questionable military effect. All are road bridges. Most are urban or town bridges that are not major routes of communications. Human Rights Watch questions individual target selection in the case of these bridges. U.S. military sources have told Human Rights Watch that bridges were often selected for attack for reasons other than their role in transportation (for example, they were conduits for communications cables, or because they were symbolic and psychologically lucrative, such as in the case of the bridge over the Danube in Novi Sad). The destruction of bridges that are not central to transportation arteries or have a purely psychological importance does not satisfy the criterion of making an "effective contribution to military action" or offering a "definite military advantage," the baseline tests for legitimate military targets codified in Protocol I, art. 52. Moreover, the risk in terms of civilian casualties in attacking urban bridges, or in attacking during daylight hours, is "excessive in relation to the concrete and direct military advantage anticipated," the standard of proportionality codified in Protocol I, art. 57.

In one final incident, a pilot targeted a large sanatorium complex in Surdulica in southeastern Serbia (incident no. 79) in what was suggested to be an error, the complex apparently being mistaken for a military installation located in the sametown. Other than the Chinese Embassy bombing in Belgrade (incident no. 49), which NATO claimed it had mistakenly identified as the Yugoslav Directorate for Supply and Procurement, this appears to be the only target attacked in error. U.S. officials have elliptically admitted to what happened at Surdulica, but have not mentioned the place name.54 In another incident of civilian deaths, at Tornik peak in the Zlatibor mountains (incident no. 12), Human Rights Watch has been unable to identify the intended target.

What follows is a discussion of the major legal and policy issues raised in selected incidents (others are discussed in Appendix A).

Refugees on the Djakovica-Decane Road, Kosovo

On April 14, during daylight hours, NATO aircraft repeatedly bombed refugee movements over a twelve-mile stretch of road between Djakovica and Decane in western Kosovo, killing seventy-three civilians and injuring thirty-six-deaths Human Rights Watch could document. The attack began at 1:30 p.m. and persisted for about two hours, causing civilian deaths in numerous locations on the convoy route near the villages of Bistrazin, Gradis, Madanaj, and Meja. NATO and U.S. spokespersons initially claimed the target was an exclusively military convoy and that Serb forces may have been responsible for the attacks on civilians. Pentagon spokesman Ken Bacon said that NATO commander Gen. Wesley Clark had received reports that "after the convoy was hit, military people got out and attacked civilians." "The pilots state they attacked only military vehicles," NATO said, adding that the "reported incident will be fully investigated once all mission details have been reviewed." There are also various NATO reports of Serbian deception in placing dead civilians at the site of the bombing. German Defense Minister Rudolf Scharping, in particular, put the blame for civilian casualties on Yugoslav forces.55

On April 15 NATO began to backtrack. It said one plane had "apparently" dropped a bomb on a civilian vehicle traveling with a military convoy. The reference to a strictly military convoy was modified: "Serbian police or army vehicles might have been in or near the convoy." NATO acknowledged that it had bombed civilian vehicles by mistake: "Following a preliminary investigation, NATO confirmed that apparently one of its planes dropped a bomb on a civilian vehicle traveling with a convoy yesterday," alliance spokespersons said.

Reporters from U.S. media went to the scene on April 15. They interviewed refugee survivors and observed shattered farm tractors, burned bodies identified as refugees, bomb craters, shrapnel, and bomb remnants with U.S. markings. The refugee column had apparently been divided in two main groups. Over the next few days, NATO wavered from insisting its forces attacked only military vehicles to an explanation that two convoys had been targeted, that the refugees had been at the rear of military columns, and that the civilian death toll was limited. On April 16, NATO spokesman Jamie Shea and Gen. Giuseppe Marini declared that "in one case and one only, we have proof of civilian loss of life. Otherwise, we are sure that we targeted military vehicles."

NATO finally admitted that the pilot of a U.S. F-16 mistakenly fired on what he believed to be military trucks, and expressed "deep regret." Later, on April 19, NATO modified its account of a single pilot's error, declaring that about a dozen planes had been involved in numerous attacks on the two convoys, dropping a total of nine bombs. Convoluted explanations continued for a number of days after the incident; NATO and the United States seemed incapable of reconstructing what had occurred. There were widespread press reports of the use of cluster bombs, which the United States denied.56

In addition to the press reporting of this incident and the endless damage control by NATO and U.S. spokespersons, Human Rights Watch obtained extensive forensic details of the incident from the Yugoslav government.57 No evidence whatsoever was ever produced to indicate Serb responsibility for any of the deaths, though Tanjug reported the deaths of three Serbian "policemen" in the bombings who it said "were securing the safe passage for the convoy."58 This tends to suggest that military or police were present in the refugee vehicles, but

Human Rights Watch found no basis to support the claim that the convoys themselves were composed of military vehicles.59

General Clark stated in September that NATO consistently observed Yugoslav military vehicles moving on roads "intermixed with civilian convoys." After the Djakovica-Decane incident, General Clark says, "we got to be very, very cautious about striking objects moving on the roads.60 Another NATO officer, Col. Ed Boyle, says: "Because we were so concerned with collateral damage, the CFAC [Combined Forces Air Component Commander] at the time, General [Michael] Short, put out the guidance that if military vehicles were intermingled with civilian vehicles, they were not to be attacked, due to the collateral damage."61 When this directive was actually issued, and why it may not have served to avoid the subsequent three incidents, remains an important question. Nevertheless, the change in NATO rules of engagement indicates that the alliance recognized that it had taken insufficient precautions in mounting this attack, in not identifying civilians present, and in assuming that the intended targets were legitimate military objectives rather than in positively identifying them.

Displaced Civilians in the Korisa Woods, Kosovo
On May 13, almost a month after the Djakovica-Decane incidents, as many as eighty-seven displaced Kosovar civilians were killed and sixty wounded when bombs were dropped during the night on a refugee camp in a wooded area on the Prizren-Suva Reka road, near the village of Korisa in Kosovo (incident no. 57). There have been various conflicting reports of the number of dead, from 48 to 87.62 The Yugoslav government claimed the attackers used cluster bombs, and the White Book published by the Ministry of Foreign Affairs includes photographs of the remains of tactical munitions dispensers (TMDs) it says are from the site. NATO spokespersons vociferously denied the use of cluster bombs,63 and Human Rights Watch has been unable to independently confirm that cluster bombs were indeed used in this attack.

In an official statement on May 15, NATO spokesman Maj. Gen. Walter Jertz acknowledged the attack, deeply regretting any "accidental civilian casualties." He insisted, nonetheless, that the attack was against Yugoslav army forces in the field:

This was a legitimate military target. The Serb claims of an attack involving cluster bombs against a non-military target are both false. NATO identified Korisa as a military camp and command post. Military equipment including an armored personnel carrier and more than ten pieces of artillery were observed at this location. The aircraftobserved dug-in military positions at the target before executing the attack. NATO cannot confirm the casualty figures given by the Serbian authorities, nor the reasons why civilians were at this location at the time of the attack.64

The NATO statement further stressed that military positions had been positively identified and that the bombs employed included laser-guided PGMs and non-guided gravity bombs:

Immediately prior to the attack at 23.30 - 11.30 pm - local time Thursday night an airborne forward air controller confirmed the target, so the identification and attack system of his aircraft, having positively identified the target as what looked like dug in military reveted positions, he dropped two laser guided bombs. Following his attack, he cleared his wingman to also attack the same target using two more laser guided bombs. Approximately 10 minutes later, the third aircraft engaged the target with...six gravity bombs. A total of 10 bombs were dropped on the target.65

The same day, Pentagon spokesman Kenneth Bacon said at a news briefing that the incident would be reviewed, but that major changes in operations should not be expected:

This accident at Korisa did not shake NATO's resolve in any way....NATO deeply regrets civilian casualties....We try very hard to avoid these casualties, but combat is inherently dangerous and accidents cannot be avoided...this mission, like every other, will be reviewed, and the airmen and their commanders will learn what they can from it and continue. But I don't anticipate that there will be a sweeping change. We can't cross legitimate military targets off the list, and we won't.66

On May 16, a Kosovar refugee who witnessed the NATO strike on Korisa reported to Deutsche Welle that FRY police forced some 600 displaced Kosovars to serve as human shields there before the attack. "We were told something bad would happen to us if we left the place," said the eyewitness, interviewed by the station's Albanian service. He said Serbian police hinted at what was about to happen. "Now you'll see what a NATO attack looks like," the refugee quoted one policeman as saying. The refugee said he finally went to sleep underneath a tractor only to be woken up by explosions and the cries of children and adults. He said he and others managed to scale a two-meter wall surrounding the plot and fled in the direction of the village as Serbian paramilitaries fired bullets around them.67

On the basis of available evidence it is not possible to determine positively that Yugoslav police or army troops deliberately forced civilians to group near them, nor to establish the motive for such action. It is not clear, for example, how potential attackers could be expected to have been aware of the refugee concentration in order to be deterred from attacking.

The laws of war expressly forbid shielding. Article 28 of the Geneva Convention IV stipulates that "The presence of a protected person may not be used to render certain points or areas immune from military operations." Geneva Protocol I, article 51(7), elaborates:

The presence or movements of the civilian population or individual civilians shall not be used to render certain points or areas immune from military operations, in particular in attempts to shield military objectives from attacks or to shield, favour or impede military operations. The Parties to the conflict shall not direct the movement of the civilian population or individual civilians in order to attempt to shield military objectives from attacks or to shield military operations.

The Protocol stresses, however, in art. 51(8), that such violations of the laws of war do not in any account release an adversary from obligations to respect civilian immunity. An authoritative new commentary on humanitarian law states: "If one party to a conflict breaks this rule, this does not exempt the other side from the regulations applicable in military attacks....The military commander must therefore take into account the column of refugees used by the adversary as a shield."68

For NATO, then, the question is whether its target designation was made with the knowledge that hundreds of displaced civilians were present in this wooded area-there is no evidence to this effect-and secondly, whether sufficient measures were taken to verify that the target had no such concentrations of civilians. On this score, the excessive civilian death toll in what NATO has itself described as a lamentable accident suggests that verification was inadequate.

Bombing of the Dubrava Penitentiary, Kosovo
Another case of Yugoslav deception involves civilian deaths and NATO bombing that damaged the large Dubrava penitentiary complex near Istok in Kosovo. According to NATO and former Dubrava prisoners interviewed by Human Rights Watch, Yugoslav Army and police forces were based adjacent to the penitentiary, which was fully operational well into the NATO air campaign, housing common and political criminals serving out their terms.

The Penitentiary Institute Istok, as it was officially called, was hit twice, causing civilian deaths among both prisoners and guards. In the first attack, at 1:15 p.m. on May 19 (incident no. 60), three prisoners and a guard were reported killed. The second attack occurred on May 21 (incident no. 65), in which at least nineteen prisoners were killed. According to a separate investigation undertaken by Human Rights Watch in Kosovo, based upon extensive eyewitness testimony, prisoners were hunted down by Serb police inside the penitentiary walls after the May 21 attack, and some eighty or so prisoners were killed.

The Yugoslav government initially reported nineteen people killed in the Dubrava Penitentiary as a result of the May 21 attack.69 However, four days later, the Yugoslav press reported from the official Tanjug agency that "in days long bombardment of the Penitentiary Institute Istok, some 100 prisoners died, and some 200 were wounded." On May 27, Tanjug quoted Vladan Bojic, judge in Pec's District Court, saying that ninety-six corpses had been pulled from the ruins. On May 29, the Yugoslav government stated that "The number of casualties in the Correctional Institution in Istok is increasing."70 On May 30, Tanjug reported a total of ninety-three killed.71 In July, the Yugoslav government claimed that NATO bombs killed ninety-five inmates and injured 196.72

While NATO readily acknowledged the air strikes at Istok and justified the attacks on the grounds that it had targeted military objectives "in the vicinity of a prison,"73 Human Rights Watch has determined that Yugoslav forces were likely responsible for the majority of deaths which occurred after the bombing. On May 22, according to eyewitnesses, prison officials ordered the approximately 1,000 prisoners to line up in the prison yard. After a few minutes, they were fired upon, and grenades were thrown at them from the prison walls and guard towers, killing at least seventy people. Over the next twenty-four hours, prison guards, special police, and possibly paramilitaries attacked prisoners who were hiding in the prison's undestroyed buildings, basements, and sewers, killing at least another twelve people.

Journalists who visited the Dubrava prison on May 21, just after the morning bombing, reported seeing deaths on the order of ten or twenty.74 Serb authorities again opened the prison for journalists on May 24. Reporting for the BBC, Jacky Rowland said it was unclear how the victims in the prison had died:

Walking around the prison we counted forty-four bodies; about half of these appeared to be the victims of the first bombing raid on Friday [May 19], still lying under blankets on the grass. Then we were taken to a room in a damaged cell block where there were twenty-five corpses. The men appeared to be ethnic Albanians, some of them had shaved heads, others had longer hair. A couple of the corpses had their trousers pulled down around their knees. We were told they had died between Friday and Sunday although it was not clear how all of them had met their deaths, nor why they were all in one relatively undamaged room.75

The Washington Post, wrote:

This time, the official version-that bombs again were to blame-did not match what reporters saw at the scene, where twenty-five more ethnic Albanian corpses were on display. The corpses were piled in the foyer of a clinic. Except for a ruined dining hall, however, no new bomb damage was visible inside the prison, and none of the newly dead had been crushed, or touched by the concrete dust that covered the dining hall floor.76

Post-war visits to the prison by journalists confirmed that prisoners had been killed after the bombing.77

In the two attacks on the Dubrava prison, NATO did not apply adequate precautions in executing its airstrikes on nearby military objectives, and therefore must be held accountable for the civilian deaths that occurred as a direct result of those attacks. But Yugoslav forces must be held fully responsible for seventy-six of the claimed ninety-five deaths at Dubrava, as these were prisoners who were executed extrajudicially well after the NATO strikes.

Serb Radio and Television Headquarters
One of the worst incidents of civilian deaths, and certainly the worst in Belgrade, was the bombing of state Serb Radio and Television (RTS) headquarters in Belgrade on April 23 (incident No. 30). According to military sources, there was considerable disagreement between the United States and French governments regarding the legality and legitimacy of the target, and there was a lively public debate regarding the selection of Yugoslav civilian radio and television as a target group.

The NATO attack was originally scheduled for April 12, but due to French disapproval of the target, it was postponed. According to military, media, and Yugoslav sources, Western news organizations, who were using the facility to forward material from Yugoslavia, were alerted by NATO government authorities that the headquarters would be attacked. Attacks also had to be rescheduled because of rumors that foreign journalists ignored warnings to leave the buildings.78 When the initial warnings were given to Western media, the Yugoslav government also found out about the intended attack. When the target was finally hit in the middle of the night on April 23, according to RTS and Yugoslav government officials, authorities were no longer taking the threats seriously, given the time that had transpired since the initial warnings. As a consequence, sixteen RTS civilian technicians and workers were killed and sixteen were wounded.

Paragraph 7 of the 1956 ICRC guidelines describing lists of targets that are legitimate military objectives includes "installations of broadcasting and television stations; [and] telephone and telegraph exchanges of fundamental military importance."79 In a May 13 letter to NATO Secretary-General Javier Solana, Human Rights Watch questioned the legitimacy of the target group in the Yugoslav war. The reasoning was that the system was not "... being used to incite violence (akin to Radio Milles Collines during the Rwandan genocide), which might have justified their destruction. At worst, as far as we know, the Yugoslav government was using them to issue propaganda supportive of its war effort. And, in fact, NATO has stated that it bombed the television facilities because they were being used as a propaganda tool of the Milosevic government." As a consequence, Human Rights Watch believes that "While stopping such propaganda may serve to demoralize the Yugoslav population and undermine the government's political support, neither purpose offers the `concrete and direct' military advantage necessary to make them a legitimate military target."80

Even if one could justify legal attacks on civilian radio and television, there does not appear to be any justification for attacking urban studios, as opposed to transmitters. After strikes on the Belgrade and Novi Sad headquarters, Yugoslav state broadcasters were able to easily move operations to other facilities. In this case, target selection was done more for psychological harassment of the civilian population than for direct military effect. The risks involved to the civilian population in undertaking the urban attack thus grossly outweighed any perceived military benefit. What is more, NATO failed to provide clear advance warning of the attacks "whenever possible," as required by Protocol I, art. 57(2).

Cluster Bombs and Civilian Deaths
One of the issues of most intense public interest that has emerged from Operation Allied Force is NATO's use of cluster bombs. As noted, there are seven confirmed and five likely incidents involving civilian deaths from cluster bomb use by the United States and Britain. Altogether, some ninety to 150 civilians died from cluster bomb use. The first confirmed incident was on April 10 (incident no. 14) and the last was on May 13 (incident no. 57).

The most serious incident involving civilian deaths and the use of cluster bombs occurred on May 7 in Nis (incident no. 48). The mid-day attack on Nis airfield, which is located inside the urban zone, killed fourteen civilians and injured twenty-eight. Cluster bomb submunitions fell in three widely separated areas: near the Pathology building of the Nis Medical Center in southeast Nis; in the town center near the Nis University Rector's Office, including the area of the central city market place, the bus station near the Nis Fortress, and the "12 February" Health Center; and near a car dealership and the "Nis Express" parking lot across the river from the fortress.

NATO confirmed the attack on Nis airfield,81 and on May 8, NATO Secretary General Solana confirmed NATO responsibility for the attack, stating that "NATO has confirmed that the damage to the market and clinic was caused by a NATO weapon which missed its target."82 According to U.S. Air Force sources, the CBU-87 cluster bomb container failed to open over the airfield but opened right after release from the attacking airplane, projecting submunitions at a great distance into the city.83

After the incident in Nis, the White House quietly issued a directive to the Pentagon to restrict cluster bomb use (at least by U.S. forces).84 Human Rights Watch considers this to have been the right move, but is concerned, given these risks, that cluster bombs were being used in attacks on urban targets in the first place. The mid-May prohibition against the further use of cluster bombs clearly had an impact on the level of civilian deaths as the war continued, particularly as bombing with unguided weapons (which would otherwise include cluster bombs) significantly intensified towards theend of the month. Nevertheless, the British air force continued to drop cluster bombs (official chronologies show use at least on May 17, May 31, June 3, and June 4),85 indicating the need for universal, not national, norms regarding cluster bomb use.

2 Documentation of abuses against ethnic Albanians in 1998 and 1999 can be seen in two Human Rights Watch reports: "Humanitarian Law Violations in Kosovo," October 1998, and "A Week of Terror in Drenica," February 1999.

3 On Racak, see Human Rights Watch, "Report on the Massacre in Racak," January 1999.

4 U.S. Department of Defense, Joint Chiefs of Staff (DOD/JCS), Joint Statement on the Kosovo After Action Review, October 14, 1999. Operation Allied Force was originally planned to be prosecuted in five phases: Phase 0 was the deployment of air assets into the European theater. Phase 1 would establish air superiority over Kosovo and degrade command and control over the whole of the FRY. Phase 2 would attack military targets in Kosovo and those FRY forces south of 44 degrees north latitude (in other words, targets beyond Kosovo but not yet including Belgrade), which were providing reinforcement to Serbian forces into Kosovo. This was to allow targeting of forces not only in Kosovo, but also in the FRY south of Belgrade. Phase 3 would expand air operations against a wide range of military and security force targets throughout the FRY. Phase 4 would redeploy forces as required. U.S. Department of Defense communications with Human Rights Watch, October and November 1999.

5 Dana Priest, "Tensions Grew with Divide over Strategy," Washington Post, September 21, 1999, p. A1.

6 Gen. Wesley Clark, Remarks to the American Enterprise Institute regarding military action in Yugoslavia, August 31, 1999. See also Testimony before the U.S. Senate Armed Services Committee Hearing on Lessons Learned from Military Operations and Relief Efforts in Kosovo, October 21, 1999. NATO's objectives for the conflict in Kosovo were set out in the Statement issued at the Extraordinary Meeting of the North Atlantic Council held at NATO on April 12, 1999 and were reaffirmed by Heads of State and Government in Washington on April 23, 1999. They included:

- A verifiable stop to all military action and the immediate ending of violence and repression;
- The withdrawal from Kosovo of the military, police, and paramilitary forces;
- The stationing in Kosovo of an international military presence;
- The unconditional and safe return of all refugees and displaced persons and unhindered

access to them by humanitarian aid organizations; and

- The establishment of a political framework agreement for Kosovo on the basis of the

Rambouillet Accords, in conformity with international law and the Charter of the United Nations.

7 Special Department of Defense Press Briefing with General Wesley Clark, Supreme Allied Commander, Europe, Topic: Kosovo Strike Assessment Also Participating: Airmen and Analysts from Operation Allied Force and Post-strike Assessment Work, Brussels, Belgium, September 16, 1999.

8 Testimony of Gen. Henry Shelton, to the U.S. House Armed Services Committee, April 14, 1999.

9 Hearing of the U.S. Senate Armed Services Committee, Lessons Learned from Military Operations and Relief Efforts in Kosovo, October 14, 1999.

10 DOD/JCS, Joint Statement on the Kosovo After Action Review, October 14, 1999.

11 Hearing of the U.S. Senate Armed Services Committee, Lessons Learned from Military Operations and Relief Efforts in Kosovo, October 14, 1999.

12 DOD/JCS, Joint Statement on the Kosovo After Action Review, October 14, 1999.

13 U.S. Secretary of Defense William S. Cohen, Remarks as Delivered to the International Institute for Strategic Studies, Hotel del Coronado, San Diego, California, Thursday, September 9, 1999.

14 DOD/JCS, Joint Statement on the Kosovo After Action Review, October 14, 1999.

15 Anthony H. Cordesman, "The Lessons and Non-Lessons of the Air and Missile Campaign in Kosovo," CSIS, revised September 29, 1999, p. 17.

16 Jane's Defense Weekly, July 7, 1999, p. 21.

17 Statement of the Honorable John J. Hamre, U.S. Deputy Secretary of Defense, before the U.S. House Permanent Select Committee on Intelligence, July 22, 1999.

18 Gen. Wesley Clark, Remarks to the American Enterprise Institute regarding military action in Yugoslavia, August 31, 1999.

19 Testimony of Gen. Henry Shelton, to the U.S. House Armed Services Committee, April 14, 1999.

20 Testimony of Lt. Gen. Michael Short before the U.S. Senate Armed Services Committee Hearing on Lessons Learned from Military Operations and Relief Efforts in Kosovo, October 21, 1999.

21 Statement of Lt. Gen. Marvin R. Esmond, U.S. Deputy Chief of Staff, Air and Space Operations, United States Air Force, October 19, 1999.

22 These are collectively referred to as "Yugoslav press reports" in footnotes throughout this report and Appendices. When significant individual articles are referenced, they are listed individually.

23 These are Turkovac near Leskovac (April 11), Kastrat east of Kursumlija (April 26), and Smederevo (May 21).

24 These are Kursumilija and Prizren (March 25), Nis and Pristina "refugee" camps (March 29), and Stavaljska breza village near Sjenica (April 6). Like army soldiers, paramilitary troops are considered combatants in the context of the Yugoslav war. and as such are excluded from this assessment of civilian (i.e., noncombatant) deaths.

25 One major exception to the largely credible nature of the government's White Book is the case of Dubrava prison, discussed in this report. It should also be noted that the general accuracy of the documentation contained in the White Book and the Ministry of Health's photographic record concerning civilian casualties and damage to civilian objects contrasts sharply with the sweeping statements made by the Yugoslav government of 1,200 to 5,000 civilian deaths during the war.

26 There is some uncertainty as to the precise number of civilians killed in about three incidents. These include the April 12 attack on the Djakovica-Klina road (incident no.17), the May 1 attack in Luzane which destroyed the Nis Express bus (incident no. 41), and the May 13 Korisa attacks (incident no. 57). In Djakovica-Klina, where the best information only indicates that "several" civilians were killed, Human Rights Watch uses five civilian deaths as its estimate. In the case of Korisa, reporting of civilian deaths varies from forty-eight to eighty-seven.

27 In three other cases, Yugoslav authorities claimed that civilian casualties were a result of cluster bomb use, but Human Rights Watch could find no evidence to corroborate the claims or found evidence to refute them.

28 Remarks as Delivered to the International Institute for Strategic Studies by Secretary of Defense William S. Cohen, Hotel del Coronado, San Diego, California, Thursday, September 9, 1999.

29 Gen. Wesley Clark, Remarks to the American Enterprise Institute regarding military action in Yugoslavia, August 31, 1999.

30 U.S. Congress, House Permanent Select Committee on Intelligence, Hearing on the Bombing of the Chinese Embassy, July 21, 1999.

31 Testimony before the U.S. Senate Armed Services Committee Hearing on Lessons Learned from Military Operations and Relief Efforts in Kosovo, October 21, 1999.

32 "Savovic: Deca cine 30 odsto od ukupno ubijenih civila," [Savovic: Children Make 30 Percent of the Total Number of Killed Civilians], Politika, May 23, 1999, p. 12.

33 "Temeljno utvrditi posledice agresije," ["To Establish a Full Account of the Consequences of the Aggression"], Politika, July 15, 1999, p. 17. In a report about the same press conference, the independent Danas newspaper quoted Zivkovic as saying that the G-17 group of independent economists had estimated the number of civilians who lost their lives in NATO attacks to be 5,700. "NATO ubio 5.700 civila," [NATO Killed 5,700 Civilians], Danas, July 15, 1999, p. 5.

34 "Zivkovic: Steta od NATO bombardovanja neprocenjiva," [Zivkovic: Damage Caused by NATO Bombardments Unmeasurable], BETA, July 14, 1999.

35 Gen. Joseph W. Ralston, Vice Chairman, Joint Chiefs of Staff, AFA Policy Forum: "Aerospace Power and the Use of Force," September 14, 1999.

36 Statement of the Honorable John J. Hamre, U.S. Deputy Secretary of Defense, before the U.S. House Permanent Select Committee on Intelligence, July 22, 1999.

37 U.S. Army, Operational Law Handbook 2000, Chapter 5, p. 2. The U.S. views the following Protocol I articles as either legally binding as customary international law or acceptable practice though not legally binding: art. 5 (appointment of protecting powers); art. 10 (equal protection of wounded, sick, and shipwrecked); art. 11 (guidelines for medical procedures); arts. 12-34 (medical units, aircraft, ships, missing and dead persons); art. 35(1)(2) (limiting methods and means of warfare); art. 37 (perfidy prohibitions); art. 38 (prohibition against improper use of protected emblems); art. 45 (prisoner of war presumption for those who participate in the hostilities); art. 51 (protection of the civilian population, except para. 6, reprisals); art. 52 (general protection of civilian objects); art. 54 (protection of objects indispensable to the survival of the civilian population); arts. 57-60 (precautions in attack, undefended localities, and demilitarized zones); art. 62 (civil defense protection); art. 63 (civil defense in occupied territories); art. 70 (relief actions); arts. 73-89 (treatment of persons in the power of a party to the conflict; women and children; and duties regarding implementation of GP I). The U.S. specifically objects to art. 1(4) (on the applicability of Protocol I to certain types of armed conflicts-wars of national liberation from "colonial domination," "alien occupation", and "racist regimes"); art. 35(3) (environmental limitations on means and methods of warfare); art. 39(2) (limits on the use of enemy flags and insignia); art. 44 (expansion of definition of combatants, relaxing of requirement to wear fixed distinctive insignia recognizable at a distance; reducing threshold of lawful combatant status to requirement to carry arms openly during military engagement or in military deployment preceding an attack; when visible to an adversary); art. 47 (non-protection of mercenaries); art. 55 (protection of the natural environment); and art. 56 (protection of works and installations containing dangerous forces). See Comments by Michael J. Matheson at "The Sixth Annual American Red Cross - Washington College of Law Conference on International Humanitarian Law: A Workshop on Customary International Law in the 1977 Protocols Additional to the 1949 Geneva Conventions," Session 1: "The United States Position on the Relation of Customary International Law to the 1977 Protocols Additional to the 1949 Geneva Conventions," American University Journal of International Law and Policy, vol. 2, no. 2 (1987), pp. 419-20.

38 It is the position of the U.S., U.N., and NATO that any military forces engaged in operations will apply the "principles and spirit" of the laws of war in their operations. It is DOD policy to comply with the laws of war "in the conduct of military operations and related activities in armed conflict, however such conflicts are characterized." (DOD Directive 5100.77 of July 10, 1979, para. E(1)(b))

39 Protocol I, Art. 48.

40 Protocol I, Art. 52(2).

41 Art. 57, Precautions in attack:

1. In the conduct of military operations, constant care shall be taken to spare the civilian population, civilians and civilian objects.
2. With respect to attacks, the following precautions shall be taken:
(a) those who plan or decide upon an attack shall:
(I) do everything feasible to verify that the objectives to be attacked are neither civilians nor civilian objects and are not subject to special protection but are military objectives within the meaning of paragraph 2 of Art. 52.and that it is not prohibited by the provisions of this Protocol to attack them;
(ii) take all feasible precautions in the choice of means and methods of attack with a view to avoiding, and in any event to minimizing, incidental loss of civilian life, injury to civilians and damage to civilian objects;
(iii) refrain from deciding to launch any attack which may be expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and direct military advantage anticipated;
(b) an attack shall be canceled or suspended if it becomes apparent that the objective is not a military one or is subject to special protection or that the attack may be expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and direct military advantage anticipated;
© effective advance warning shall be given of attacks which may affect the civilian population, unless circumstances do not permit. (...)

42 ICRC, Commentary on the Additional Protocols, pp. 681-682, para 2198. The ICRC notes that some delegations to the diplomatic conference that adopted the protocols interpreted the terms "everything feasible" to mean:

everything that was practicable or practically possible, taking into account all the circumstances at the time of the attack, including those relevant to the success of military operations. The last-mentioned criterion seems to be too broad, having regard to the requirements of this article.

The interpretation considered too broad by the ICRC, however, appears verbatim in the other authoritative commentary on the protocols, M. Bothe, K. Parsch, and W.Solf, New Rules for Victims of Armed Conflicts: Commentary on the Two 1977 Protocols Additional to the Geneva Conventions of 1949 (The Hague: Martinus Nijhoff, 1982), p. 362.

43 The principle of proportionality is codified in Protocol I, Art. 51 (Protection of the civilian population), section 5:

Among others, the following types of attacks are to be considered as indiscriminate:...
(b) an attack which may be expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and direct military advantage anticipated.

44 ICRC, Commentary on the Additional Protocols, p. 626, para. 1980: "The idea has also been put forward that even if they are very high, civilian losses and damages may be justified if the military advantage at stake is of great importance. This idea is contrary to the fundamental rules of the Protocol....The Protocol does not provide any justification for attacks which cause extensive civilian losses or damage. Incidental losses and damage should never be extensive."

45 ICRC, Commentary on the Additional Protocols, p. 684, para. 2212.

46 Ibid., p. 684, paras. 2213-2214.

47 For an early account of the use of cluster bombs by NATO in the Federal Republic of Yugoslavia and Human Rights Watch's position, see Human Rights Watch, "Ticking Time Bombs: NATO's Use of Cluster Munitions in Yugoslavia," A HumanRights Watch Short Report, vol. 11, no. 6(D), May 1999; and Human Rights Watch, "Cluster Bombs: Memorandum for CCW Delegates," December 16, 1999.

48 The question as to what extent the military is obligated to expose its own forces to danger in order to limit civilian casualties or damage to civilian objects is examined in William J. Fenrick, "Attacking The Enemy Civilian As A Punishable Offense," Duke Journal of Comparative and International Law, 1997, p. 546, located at

49 The eighty-seven deaths in Korisa are counted in the Human Rights Watch total of 500; the seventy-six at Dobrava prison are not.

50 Human Rights Watch discussions with U.S. Air Force and Joint Chiefs of Staff officers, October 1999.

51 This issue will be discussed in greater detail in Human Rights Watch's upcoming report on targeting in Operation Allied Force.

52 The second largest broadcast center in Yugoslavia, in Novi Sad, was hit the next day but there were no civilian casualties. Officials told Human Rights Watch that after the attack on Belgrade, RTS evacuated the facility.

53 Seven oil storage tanks were damaged in the attack, as were the pump house and pouring station on the Sava river.

54 U.S. Congress, House Permanent Select Committee on Intelligence, Hearing on the Bombing of the Chinese Embassy, July 21, 1999.

55 NATO, SHAPE News Morning Update, April 15, 1999; Reuters, 150059 GMT April 15, 1999.

56 Joie Chen and Jamie McIntyre, "As Serb Force Grows, Limits of Air Attacks Become Apparent," CNN The World Today Broadcast, April 19, 1999; Sarah Chayes, "General Daniel Leaf Explains the Refugee Bombings," National Public Radio, All Things Considered Broadcast, April 19, 1999.

57 FRY Ministry of Foreign Affairs, NATO Crimes in Yugoslavia, vol. 1, pp. 1, 21-26, 32-37; FRY Ministry of Housing, "Photo Documentation of Civilians Who Were Killed By NATO Attacks, from 24.03 until 20.05.1999."

58 Tanjug, Pristina, April 15, 1999.

59 At least two eyewitnesses told Human Rights Watch that the convoy was interspersed with military vehicles. Interviews with Kole Hasanaj, Meja, July 25, 1999, and with Safet Shalaj, Djakovica, July 25, 1999.

60 Special U.S. Department of Defense Press Briefing with Gen. Wesley Clark, Supreme Allied Commander, Europe, Topic: Kosovo Strike Assessment; Also Participating: Airmen and Analysts from Operation Allied Force and Post-Strike Assessment Work, Brussels, Belgium, September 16, 1999.

61 Ibid.

62 FRY Ministry of Foreign Affairs, NATO Crimes in Yugoslavia, vol. II, pp. 1-17. Though the White Book states that there were "only" forty-eight victims in Korisa, Yugoslav and Western press, as well as the U.S. State Department and the U.N. report figures of eighty to eighty-seven victims. Based upon Human Rights Watch investigations and discussions with Western journalists who attempted to reconstruct the incident, it appears that more that forty-eight people definitely died in the Korisa attack. The range of deaths is thus used.

63 Transcript of Backgrounder given by Peter Daniel and Major General Walter Jertz, in Brussels, May 15, 1999.

64 NATO, Subject: Press Release (99) 079, Statement by the NATO Spokesman on the Korisa Incident, May 15, 1999.

65 Transcript of Backgrounder, May 15, 1999.

66 Transcript, U.S. Department of Defense News Briefing, May 15, 1999.

67 Reuters 152249 GMT, May 15, 1999; Kosovo Chronology, Timeline of events 1989-1999 relating to the crisis in Kosovo, released by the Department of State, Washington, DC, June 18, 1999.

68 Hans-Peter Gasser, "Protection of the Civilian Population," in Dieter Fleck (ed.), The Handbook of Humanitarian Law in Armed Conflicts (Oxford: Oxford University Press, 1995), p. 505, para. 506. Hans-Peter Gasser is a Senior Legal Adviser of the ICRC.

69 Information provided by Yugoslav civil defense authorities; FRY MFA, NATO raids on manufacturing and civilian facilities on May 21st and in the night between May 21st and 22nd 1999.

70 FRY Ministry of Foreign Affairs, NATO Raids on Manufacturing and Civilian Facilities on May 29th and in the Night Between May 29th and 30th 1999.

71 Yugoslav press reports; "Identifikovano 86 mrtvih," DAN, May 27, 1999, p. 2; "Jos sedam leseva," DAN, May 30, 1999.

72 FRY Ministry of Foreign Affairs, NATO Crimes in Yugoslavia, vol. II, p. 319.

73 NATO, Operation Allied Force Update, May 22, 1999, 0930 CET. See also Transcript of Press Conference given by Mr. Jamie Shea and Col. Konrad Freytag in Brussels on Saturday, May 22, 1999.

74 Jacky Rowland, "Bombs, Blood and Dark Despair," Scotland on Sunday, May 23, 1999; Paul Watson, "NATO Bombs Ignite Prison Chaos-KLA Officers Reported to be Among Inmates," Toronto Star, May 22, 1999; Associated Press, "NATO Hits Kosovo Jail Again Friday Night," May 21, 1999.

75 Jacky Rowland, "Istok Prison's Unanswered Questions," BBC World News, May 25, 1999.

76 Daniel Williams, "Kosovo Revisited: At War's End, Old Places Seen in New Light," Washington Post, June 26, 1999.

77 Carlotta Gall, "Stench of Horror Lingers in a Prison in Kosovo," New York Times, November 9, 1999.

78 Human Rights Watch interviews with Air Force and Joint Staff planners. See also Dana Priest, "Bombing by Committee: France Balked at NATO Targets," Washington Post, September 20, 1999, p. A1.

79 ICRC, Commentary on the Additional Protocols, p. 632, para. 2002, note 3.

80 Human Rights Watch letter to Javier Solana, May 13, 1999.

81 NATO (SHAPE), ACE News Release - Press Release 99-05-02, May 8, 1999.

82 Transcript of Press Conference given by the NATO Secretary General, Mr. Javier Solana, in Brussels, on Saturday, May 8, 1999 (including Maj. Gen. Jertz).

83 Human Rights Watch correspondence with a U.S. Air Force officer, November 1999.

84 Human Rights Watch discussions with U.S. Air Force and Joint Chiefs of Staff officers, October 1999.

85 U.K. Ministry of Defense, Royal Air Force, Operation Allied Force News and Downloadable Images (

Previous PageTable Of ContentsNext Page