All conflicts have a pre-history. Few have as clear a beginning as the Nagorny Karabakh conflict. The basic positions – the Karabakh Armenians' determination to secede from Azerbaijan with the support of Armenia and Azerbaijanis' resolve to stop that happening – were adopted in February 1988 and that month saw turmoil erupt as if out of the blue in the form of demonstrations, strikes, political quarrels, flights of refugees and pogroms. That full-scale Armenian-Azerbaijani fighting only broke out at the end of 1991 is more a matter of weaponry than of intention.

The events of February 1988 were dramatic, sudden, and almost universally unanticipated in a Europe that had all but forgotten the power of nationalism as a political force. In that sense, by being the first serious nationalist quarrel of the late Communist era, the Karabakh conflict can be called both the most unexpected and the most predetermined of all these disputes. More than any others in Yugoslavia or the Soviet Union, the conflict was all but inevitable because its causes lay in the 'deep structure' of the relationship between its two parties in late Communist times. Four elements – divergent national narratives, a disputed territorial boundary, an unstable security arrangement and lack of dialogue between the two parties – had made fissures that would break Armenia and Azerbaijan apart, as soon as trouble began. Yet because the problem was both so new and so profound, no mechanism was found – or has yet been found – to repair the damage.

Contradictory national narratives pervaded both societies at all levels. Before fighting began, intellectuals had formulated detailed arguments that formed a national frame of reference for what happened on the battlefield. These positions were first staked out during the 1960s post-Stalinist 'thaw' initiated by Soviet leader Nikita Khrushchev, which created conditions for sanctioned, or 'orthodox', forms of nationalism. These proved difficult to control, however, as diametrically opposed versions of history were later propagated by writers such as Zia Bunyatov and Zori Balayan.

When the dispute broke to the surface in 1988, teams of pamphleteers and propagandists on both sides were ready to rush into the breach and they began producing works with titles such as 'Karabakh: the guilty party in the tragedy is well known'. Disappointingly little has changed in this regard. These ideological battles continue to this day on the internet among a narrow audience, yet opinion polls in both countries suggest an overwhelming majority of respondents find it impossible to countenance Karabakh being given to the 'other' in any peace agreement. It is customary on both sides to believe that to be without Nagorny Karabakh is to have an incomplete national identity, that Armenian or Azerbaijani nationhood is a stunted and wounded thing without it. This in turn feeds into a wider sensation of insecurity in the face of the threat posed by the 'other' and its allies, real or perceived.

We should not dismiss these fears as fantasies – after all the modern history of both Armenians and Azerbaijanis contains enough real instances of catastrophe and loss to provide grounds for genuine insecurity. In both countries, many more compatriots live outside the home state than inside it, as a result of war, expulsions and Great Power treaties. The Treaty of Turkmenchai of 1828 divided Azerbaijan into two – Russian and Iranian – parts and the Great Power politics of 1915-21 truncated Armenia. While the mass killing of Armenians in Anatolia in 1915 dwarfs in scale anything else committed in the neighbourhood in modern history, Azerbaijanis also suffered grievously from early twentieth century history – often at the hands of Armenians. It has been precisely those episodes where Azeris and Armenians suffered at each other's hands (rather than Russians, Georgians or others) that have been the focus of contemporary nationalist narratives.

The first major Armenian-Azerbaijani tensions can be traced to the end of the nineteenth century. Armenians and Azerbaijanis were the two major national groups in the borderland between the Russian, Ottoman and Persian empires and as such were intermingled over territory stretching hundreds of miles. Historically, the Armenians tended to be a more successful socio-economic group, with a heavier concentration in towns and cities as a prosperous merchant class. With the rise of nationalism and heightened Russian-Ottoman conflict at the end of the nineteenth century, the Armenians became both the most militant and the most vulnerable community in Ottoman Anatolia. And, while a few generations before mainly Shiite Azerbaijanis and mainly Sunni Turks might have found little in common, they increasingly found common cause – and were identified as one and the same by Armenians. This has fed through into the Armenian generalisation that Azerbaijanis are also 'Turks' – and therefore share complicity for the 1915 genocide.

The Armenian-Azerbaijani fighting of 1917-20 can be seen as a messy attempt to draw borders and build a viable state – a bloody process that was being played out across the rest of the Eastern Europe and the Balkans. It was also war by proxy, with Russia and Turkey continuing a long-running territorial conflict that had lasted for most of the previous century, each using the Armenians and Azerbaijanis respectively as their local allies.

The misfortune of Karabakh was that it was always caught in the middle. Geographically it was situated on the Azerbaijani side of the mountainous watershed that runs down between the two countries. Demographically it was mixed, as it evidently had been for centuries: the Armenians predominated in the hills, with more Azerbaijanis in the plains, as well as in the city of Shusha (or Shushi as it was known to its Armenian inhabitants). Culturally it had great significance for both sides. For Armenians, the meaning of Karabakh lay in the dozens of Armenian churches dotted around the territory, its tradition of local autonomy through the "melik" princes of the Middle Ages and the martial reputation of Karabakh Armenians. For Azerbaijanis, the associations were primarily with the khanate based around the great eighteenth century city of Shusha and with the great cultural flowering of composers and poets such as Vagif, Natevan and Uzeir Hajibekov. Karabakh was in short a culturally rich border-zone, like Alsace, Flanders or Kashmir and, like them, fated to be a battlefield.

In 1920-21 the only 'solution' of this dispute could come either by military victory – as basically happened in Anatolia, Zangezur and Nakhichevan – or by the imposition from above of a new structure by an imperial power. After the British failed to impose a settlement, the imperial arbiters turned out to be the Bolsheviks, whose 11th Army conquered Karabakh in May 1920. On 5 July 1921 the Bolsheviks' Caucasian committee, the Kavburo, under the chairmanship of Stalin ruled that the mountainous part of Karabakh would be part of Azerbaijan. In July 1923 the Nagorny (or Mountainous) Karabakh Autonomous Region (NKAO) was created within Azerbaijan, with borders that gave it an overwhelming Armenian population of 94 per cent of the total inhabitants.

This arrangement turned the NKAO into one of only two instances in the Soviet system of an autonomous province inside one union republic that had a strong affiliation to another union republic (the other instance, Russian-majority Crimea, though also unstable, has proved a less fissiparous case). The contradictions of this arrangement were never openly discussed, but the two national narratives were still far apart: many Armenians never accepted the 1921 decision and protests against it were made in 1945, 1965 and 1977. On 20 February 1988 the NKAO local Soviet voted to request the Soviet government to permit Karabakh to leave Soviet Azerbaijan and become part of Soviet Armenia. It was perhaps predictable that the vote took Azerbaijanis completely by surprise: to them that Nagorny Karabakh was part of their republic was a self-evident fact, reinforced by everyday news as well as by scholarly literature that stressed the territory's Azerbaijani heritage.

A key point must be made here, which is that these underlying structural tensions in the architecture of the region had little impact on the daily life of the residents living there. As most Armenians and Azerbaijanis will tell you, they traditionally had a better trading relationship with each other than either community did with Georgians; rates of intermarriage were also high. Soviet Karabakhis from both communities tended to be bilingual, on good terms with their neighbours and little concerned by the nationalist narratives being advanced by intellectuals in Yerevan or Baku.

It is an elementary mistake therefore to see the Karabakh conflict as a clash of 'ancient hatreds' or as a religious dispute. Links of culture, business and marriage still bind Armenians and Azerbaijanis together in Moscow, Georgia and Iran – anywhere in fact outside the zone of the Karabakh conflict.

What was it then that pushed these ordinary neighbours into conflict with one another? The precipitate breakdown in security and trust could be attributed to the rigidities of the Soviet state, which had failed to manage the political contradictions inherent in Nagorny Karabakh. The heavily centralised system had enforced security through fear but it had almost no mechanisms of resolving a dispute between two communities by consensus.

The Soviet Union created territorial units defined in terms of nationality, but it did not allow the free formation of horizontal civic bonds across society in the sense of participation in public life by organisations defined neither by nationality nor the Communist Party. By 1988 ideological socialism had vanished as a guiding principle, so, when crisis struck, it was all too logical that both Armenians and Azerbaijanis showed allegiance to their nationality and homeland, rather than to putative ideals of 'brotherhood' and 'workers' solidarity'. They had no interest in engaging in bilateral dialogue in pursuit of an acceptable compromise, even if they had wanted to. No Karabakh Armenians thought of pressing their case in Baku and no Baku leaders thought of inviting them: one Karabakh Armenian advocate of compromise, Valery Grigorian, part of a delegation from Stepanakert to Baku in 1991, was murdered on his return. Instead both sides instinctively sent delegations to Moscow and made telephone calls to the Kremlin to bend the ear of Communist Party officials. Both believed this was a dispute which Moscow would arbiter and in which there would be one winner.

Ordinary people thus soon lost whatever bonds they had had with their neighbours and friends. Once the political dispute had begun, residents in Shusha and Stepanakert could only be identified as either Armenians or Azerbaijanis – an act of identification that when it came to mixed marriages had to be deliberately made. They had no third option to choose, apart from voting with their feet and leaving their hometown altogether. It is indicative that the Armenian-Azerbaijani community with the strongest civic consciousness, the one in Baku, survived the political strains the longest and fell apart only in January 1990.

Some have seen this breakdown in relations as being economically motivated – as being a quarrel over scarce resources – yet this analysis also does not stand up to scrutiny. The tensions within Karabakh pre-dated by decades the depressed Soviet economy of the 1980s and the province was not appreciably poorer than many other parts of the Soviet Union, having average economic indicators for Azerbaijan and being slightly poorer than Armenia. A promise of a "small renaissance" by Gorbachev and a big influx of investment from Moscow spearheaded by Arkady Volsky in 1988-89 were not enough to persuade the Karabakh Armenians it was within their interests to stay within Soviet Azerbaijan.

The breakneck speed with which events occurred in early 1988 suggests how fragile the situation was. First a large group of Azerbaijanis fled southern Armenia, complaining of brutal harassment (a smaller group had left some months before). Then the Nagorny Karabakh Soviet made its unprecedented vote. This led to rallies in Stepanakert and Baku, and then to the largest demonstrations ever seen in the post-war Soviet Union in Yerevan. At the same time the post-war Soviet Union experienced its first major strike by workers in Nagorny Karabakh.

What tipped the balance from confrontational politics into outright conflict were the pogroms in Sumgait on 28-29 February 1988. These came about through a combustible mix of ingredients: a depressed and polluted town with a mixed ethnic community; a sudden influx into Sumgait of Azerbaijanis fleeing Armenia; a town leadership that was either confused or absent; silence from the authorities in Moscow; reports that two Azerbaijanis had been killed in Karabakh; and a series of angry demonstrations. All of this flared into a ferocious pogrom in the Armenian quarter of the town, leading to a belated armed intervention by the Soviet army. The official death toll (almost certainly reliable because it tallies with lists of named victims) was 26 Armenians and 6 Azerbaijanis. Sumgait effectively made the Karabakh dispute into the Karabakh conflict, with only a lack of access to weaponry preventing an immediate escalation into war.

Although there have been many conspiracy theories about Sumgait, the weight of evidence suggests that it was a grassroots upsurge of violence, mishandled by the Soviet authorities, but not provoked and still less instigated by them. Moscow played a less than glorious role in the outbreak of the conflict, but it is wrong, as many in the Caucasus still do, to blame the start of the dispute in 1988 on manipulation by the centre, with most ordinary Armenians and Azerbaijanis relegated to the role of bystanders. All the Politburo documents available from the time show that the central decision-making authority of the Soviet Union was immediately out of its depth, unable to cope with an unwelcome and unfamiliar challenge. Gorbachev was immediately wrong-footed because the republican Communist Parties of Armenia and Azerbaijan refused to comply with his wishes. In an outburst of frustration typical of this period, Gorbachev lashed out at the Party leaders in both Baku and Yerevan, complaining to a Politburo meeting convened to discuss the Sumgait events on 29 February 1988 that, "We need information and it's hard to obtain it – both sides are hiding it. Everyone is involved."

In retrospect Gorbachev could probably have used only two tactics in 1988 to dampen down this growing conflagration. It is to his credit that he did not try the first, which was mass repression. Mass arrests of all political activists in Armenia and Azerbaijan might have cowed the populations of both Soviet republics in the short term, but this would have carried a big political cost for Gorbachev and would surely have only postponed trouble to a later date. The second option was still more fantastic even for the most liberal leader in Soviet history. This was to begin a democratic debate designed to find consensus – by flying to the region, holding talks, commissioning an independent enquiry into the problems associated with the dispute and negotiating a compromise solution. In the Soviet Caucasus of 1988, this kind of initiative was simply unimaginable. Significantly, when both Armenians and Azerbaijanis continue to criticise Gorbachev for his role in the Karabakh dispute, it is not for his failure to be a fair mediator, but for his failure to use the authority of the Kremlin to award Nagorny Karabakh to its just owner – them.

The evidence shows that Moscow in fact lost full control of both Armenia and Azerbaijan as early as 1988. Gorbachev replaced both Communist Party leaders in Baku and Yerevan in May 1988, only to find that their successors took an even more robust line on the Nagorny Karabakh issue. Both Soviet republics embarked on a process of nation-building, adopting new nationalist symbols, creating new institutions, and deporting en masse their respective Armenian or Azerbaijani minorities.

Another watershed came in January 1990, when the Soviet leadership lost Azerbaijan, first yielding the streets of Baku to the nationalist opposition and then making things many times worse by sending in the army to the city to crush the Azerbaijani Popular Front and killing dozens of civilians. The first victims of this crisis were Baku's remaining Armenians, more than 90 of whom died in pogroms before they were evacuated from a city that finally lost its reputation as a haven of multiethnic tolerance.

The end of the Soviet Union in 1991 transformed the Karabakh conflict from a civil war to an interstate one. As the 15 former Soviet republics became independent states within their Soviet boundaries, Nagorny Karabakh was formally recognised internationally as being part of Azerbaijan – a central issue that still stands at the heart of the dispute. At the same time the transfer of Soviet weaponry to each side increased the destructive capacity of both combatants.

In purely physical terms, the geographical and demographic realities on the ground set an Azerbaijani side that surrounded Nagorny Karabakh with Azerbaijani towns and villages against an Armenian side that was separated from nearby Armenia but had far greater control of Karabakh itself. The military outcome of the conflict thus hinged on the ability of the Azerbaijanis to blockade Nagorny Karabakh and force their terms on it, set against the capacity of the Armenians to break out of the military cordon and link up with nearby Armenia.

The blackest episodes of the war from the Armenian perspective are the remorseless Azerbaijani shelling of Stepanakert in 1991-92 and the sweep through Karabakh by Azerbaijani forces in 1992. The Azerbaijanis recall with horror the worst massacre of the war when an estimated 485 people, the vast majority civilians, were killed in and around Khojaly by the Armenian forces in February 1992. They also suffered by far the greatest humanitarian crisis with the occupation by Armenian forces of seven regions around Karabakh in 1992-94 and the expulsion of half a million people from those regions. The aftermath of this expansion of the war to a vast swathe of territory outside Nagorny Karabakh (if Karabakh itself is included, then Armenians currently control 13.6 per cent of the internationally recognised territory of Azerbaijan) remains an open wound in the unsolved conflict.

These were brutal acts of war. But the blame for them should be shared with the outside world for failing to commit more resources to securing the warring parties' commitment to a settlement. These brutal acts arguably stemmed from a military logic in a situation of kill-or-be-killed (although it is harder to see a strategic logic in Khojaly, where warped motives of revenge may have played a big role). To prevail, the Azerbaijanis needed to tighten the economic and military noose around Karabakh, an aim that they bloodily failed to achieve. In turn the Armenians needed to transform Karabakh from a vulnerable enclave into a defensible fortress – a task they bloodily achieved.

A recurrent theme of the Karabakh dispute since 1988 has been the lack of an effective and authoritative outside arbiter. The conflict can be said to have originated as a clash of political ambitions for a traditionally disputed territory at a moment of imperial decline. This quickly turned into a matter of elementary security, when the imperial centre was unable to protect vulnerable communities and both sides fell back on their own resources.

The dynamics of conflict of course soon transformed what had already begun. After February 1988, economic factors also began to fuel the conflict and do so to this day. Profiteering arms-dealers, greedy looters, middlemen and mafiosi all came to have a stake in the growth of the conflict and in its non-resolution. Also, the awfulness of war and the hateful propaganda it has spawned and the desire for revenge caused by a conflict that has taken 20,000 lives have progressively undermined the common culture once shared by Armenians and Azerbaijanis and created an antipathy that is now projected back into the past.

What are the implications of this brief analysis of the roots and evolution of the conflict for attempts to resolve it? First, that socio-economic conditions were not to blame and that Armenian-Azerbaijani social and economic links, although damaged over the years, are entirely reparable. Second, that previous political arrangements for Karabakh have contained lethal structural flaws that perpetuated mutual insecurities, and that only the latter-day Great Powers of Russia and the West have the capacity to devise a new overarching security structure that makes all sides feel safe from potential aggression. Finally, that the heart of the conflict lies within the narrow political narratives that Armenia and Azerbaijan have come to employ both in imagining themselves and the other. To break out of the prison of the conflict, they need to begin the titanic effort of a genuine dialogue about their common future. Since 1988 Nagorny Karabakh has come to represent an ever-widening chasm, but both Armenians and Azerbaijanis need to re-imagine it not only as a territory but as a bridge, on which they have a firm footing but are freely joined to the other side.