The Easter headlines from the teachers' union conferences were stark:
"Union call for action on school violence," read one. "Armed pupils face airport-style checks," shouted another.
Every year, without fail, violence in schools is raised as a major issue by the unions at their annual gatherings. And so it should be, for even one attack on a teacher is one too many.
Studies carried out by the unions among their members often make alarming reading. In a recent survey of 2,500 teachers commissioned by the National Union of Teachers, almost a third of respondents said they suffered some form of physical assault at least once a year. A further third reported being threatened by pupils with the same frequency, and a quarter said they suffered threats from parents.
The National Association of Schoolmasters Union of Women Teachers found a similarly grim picture when it carried out a survey in the east of England last year. Extrapolating from its findings, it calculated a teacher was attacked once every seven minutes in the Eastern counties.
The union said the survey demonstrated "a very worrying picture" and that the incidents of verbal and physical abuse reflected national as well as local problems.
So the latest findings from the British Crime Survey, the most authoritative snapshot of crime rates nationwide, comes as something of a surprise. The rate of violence against teachers has dropped by more than 40 per cent in the past eight years, according to a report published last month.
Drawing on results from the crime survey, which reports each year on a representative sample of 40,000 people, Home Office researchers found that in 2002 and 2003, 1 per cent of teachers were physically attacked at work.
Between 1994 and 1998 the level of assaults on teachers was much higher, running at 1.8 per cent.
Verbal threats against teachers had also fallen, the survey revealed. Two per cent of teachers reported the problem each year between 1994 and 1998, but by 2002-3 that figure had dropped to 1.2 per cent.
Although teachers faced a higher-than-average risk of being attacked at work, the research showed, several other groups of workers were more likely to face assault. Security staff were 12 times more likely to be attacked, nurses three times and transport workers twice as likely. Managers and proprietors in agriculture and in service industries were also at higher risk than teachers.
When it came to verbal threats, the picture was similar. Three per cent of security workers had been threatened; 2.3 per cent of health workers and 1.5 per cent of leisure workers.
The threat of verbal or physical abuse had dropped significantly for all the high-risk groups since 1994, but the effect on teachers had been greater than on others. While the risk of assault for all workers had dropped by 25 per cent, the risk for teachers had dropped by 44 per cent.
While the risk of threats had dropped by a third overall, the risk for teachers had dropped by 40 per cent.
So how do these findings square with those blaring headlines from the teacher unions? Examine the unions' work more closely and some major drawbacks are revealed.
The first, and perhaps the most significant, of these is a perennial problem for any researcher with limited resources - the self-selecting sample. The NUT, for example, sent survey forms to 17,000 teachers but it received just 2,575 replies. A response of 15 per cent, such as this one, is by no means poor for such a survey.
Dr Sean Neill, a senior lecturer at the Warwick Institute of Education who analysed the NUT's data, said the union's surveys had been shown to be representative on other issues. For example, a survey on testing was later followed by a ballot of all members which showed a similar proportion opposed to the measures.
However, pressed as to why these findings should differ so markedly from the National Crime Survey data, he admitted the Government's research might be more solid.
"I would accept that the number of people who report serious incidents is likely to be biased towards those who experienced them," he conceded.
"These surveys are done in order to get a quick response to an immediate problem. The union may ring and say it needs a report in six weeks, and so it is more limited than the sort of research in which you can spend two or three years chasing up every single response."
The issue of self-selection is not the only problem with the teacher unions' research. Its questions do not always ask specific questions, as the Government's crime research does: "Have you been physically assaulted at work in the past year?"
In fact, it is hard to extrapolate from the NUT's research precisely what proportion of respondents had been attacked in the preceding year. They were asked not whether they had been assaulted, but whether they had experienced "pushing, touching or other unwanted physical contact". This, of course, would have included the teacher who was irritated by children rushing carelessly past along a corridor.
The category in the union's research which asked about "serious incidents" such as assaults, asked not if the respondent had been a victim, but only if he or she had witnessed an incident.
The issue of what precisely constitutes an assault is also tricky. While respondents to a crime survey might be expected to take a strictly legal approach, others' responses may be more subjective.
Tim Beech, the NASUWT eastern regional organiser who carried out its survey, has a simple formula. "I have no problem in defining assault," he said. "If someone believes because of the action of someone else they have been assaulted, then they have been assaulted. If they were standing in a doorway and someone pushed past them, I would count that as an assault."
Mr Beech's survey concluded that a teacher was attacked at work in the Eastern Counties once every seven minutes. The National Crime Survey data, when extrapolated in the same fashion, suggests that a teacher is attacked somewhere in England or Wales once every 20 minutes.
Even this lesser figure sounds worrying, of course. But put it another way: if 1 per cent of teachers experience assault at work each year, the average teacher can expect an attack once in every hundred years.
Professor Peter Smith, head of the unit for school and family studies at Goldsmiths college in London and an expert on violence in schools, says this improvement is both welcome and predictable. The increase in awareness of violence and bullying has led to a number of initiatives in recent years, he points out, and it would be sad if that had not had an effect.
"We shouldn't be complacent but a lot of good things are happening," he says. "We know that schools can do quite a bit, and although there is a lot more still to do, there has been some success."
A recent project on bullying in schools, carried out by Professor Smith with another Goldsmiths researcher, found an apparent decline in the problem over the past few years. And a book by Professor Smith which drew together perspectives on violence from across Europe found other countries had a similar "perception gap" between real and perceived levels of aggression in schools.
Researchers from Germany, for example, reported a growing perception there that violence in schools was getting worse. But there were no national data and recent research had failed to discover any such effect, they said.
This effect is not necessarily a bad thing, though, Professor Smith argues.
"It's probably good to have a bit of a perception gap," he says. "It would be a pity if it got so big that it led to an inappropriate response such as security guards around schools.
"But you've got to keep on at these things. It isn't just a matter of having a policy - you've got to keep working at it."
Violence at work: Findings from the 2002-3 British Crime Survey by Anna Upton. Home Office April 2004. Violence in Schools: The Response in Europe.
Edited by Peter K Smith. RoutledgeFalmer, 2003