The 'No' majority in the independence referendum last year is being explained by an unusual alliance of Scotland's youngest voters, its average earners, Protestants and women.
A breakdown of voting patterns, drawn from a survey of 5,000 Scots conducted soon after the referendum day, runs counter to the widespread belief that there was a clear split between older and young voters, or that higher earners backed the United Kingdom.
It did, however, confirm other polling evidence that women were decisive in the result. While men were 53% for 'Yes', women were 57% for 'No'.
Voters earning more than £30,000 were found to be evenly split, while those earning less than £20,000 were 53% for independence.
It was those earning between £20,000 and £30,000 who can also be seen as deciding the outcome - voting 'No' by a margin of 56% to 44%.
It is not explained why that pattern emerged, but one possibility is that these voters felt they had something to lose if independence proved costly, while not having the higher earners' financial security with which to take a risk on constitutional change.
The research found that older voters were strongly in favour of Scotland remaining part of the UK, including 67% of those aged over 70.
But the academics behind the Scottish Referendum Study note that the youngest voters, aged between 16 and 24, appear to have voted 'No' as well.
Those aged 25 to 29 were in the age group most likely to vote 'Yes', with 62% for independence, yet teenagers and those in their early twenties were 54% for the Union.
The analysis does not offer any explanation for why younger age groups split that way.
The survey looked for correlation between voters' identities and the way they cast their referendum ballot.
It found a strong link on religious affiliation. Some 60% of Protestants voted 'No', while 58% of Catholics voted 'Yes'. Those adherents of the Church of England were 81% for the Union - the strongest correlation of all. However, the research does not indicate that religion was a cause of the way people voted.
It found a weaker correlation between owner occupiers and a 'No' vote.
Those born in other parts of the UK were much more likely to vote 'No', by 70% to 30%. There was a smaller 'No' majority for those born outside the UK, and those born in Scotland were evenly split.
The Scottish Referendum Study is linked to the British Election Study. It is run by academics at the universities of Edinburgh, Glasgow and Essex, and is funded by the Economic and Social Research Council.
Its findings are broadly in line with two polls carried out at the time of the referendum.
Though with smaller samples and less detail, these showed a similar gender gap, there was an indication that the youngest voters were pro-union, and there was a split between higher skilled, higher earners and those on lower pay with lower work skill levels.
In an extension of the research, a briefing paper from the Scottish Referendum Study shows the issues that had most impact on voters in the latter weeks of the campaign.
The arguments that worked best for the Unionist side were that new powers could be secured for Holyrood, short of independence, and that Scotland was already seen as having a fair share of resources or powers.
On the other side, the independence campaign was winning people over with the potential for tackling inequality and the possibility of avoiding budget cuts.