MEN accused of rape in England and Wales should count themselves lucky: only one in 19 will be convicted of the crime. Their odds are still better in some places. In Gloucestershire just one of 116 reports of rape led to conviction last year. Liz Kelly, an academic with the government's ear, dubs the system a ?culture of impunity?. The Home Office has asked for opinions on why so few reported rapes lead to a conviction, with a view to changing the law. But, so far, the consultation has revealed uncertainty and rancour over what is wrong at the moment.

The police now record more rapes than ever. That is probably a good sign because it suggests they are more trusted (the numbers have also been bumped up slightly by changes in how cases are counted). But it would be a better sign if convictions had kept pace with allegations. They have not. Between 1991 and 2004 the number of rapes recorded by the police rose by 247% (see chart). In the same period, prosecutions went up by 54% and convictions rose by just 34%. The conviction rate in England and Wales used to be about average. It is now one of the lowest in Europe.

One reason for the widening gap, police officers say, is that they now investigate trickier cases. They are more likely to pursue allegations about assaults in the home, directed against current or former partners. It is hard to convince prosecutors that such cases have any chance of success; and, even if the lawyers can be persuaded, juries often cannot. Jurymen, like most people, tend to think of rape as something done by men who lurk behind trees.

Campaigners disagree. They point out that many coppers lack proper training: a report by the Association of Chief Police Officers last year found that just ten out of 42 police forces in England and Wales had special sexual-assault investigation-teams. And researchers at London Metropolitan University say that police officers, even those trained to investigate sexual offences, almost certainly overestimate the proportion of claims that are false.

Betsy Stanko, an academic who has examined 700 case files in London, has found another reason why the police drop cases: kindness. Officers know that some kinds of victim stand little chance of being believed in court. They include people who had been drinking before the attack took place (one in three cases) and the mentally ill (one in five). In such cases, coppers often take what Ms Stanko terms ?sympathetic withdrawal statements? and do not press the matter further.

If the police believe courts do not look kindly on many rape victims, they are right. Scant evidence, the difficulty of weighing contrasting accounts and the severity of the prison sentence (not to mention the dismal experience in jail for a convicted rapist) make convictions unlikely. Even after police ?filtering?, fully 72% of rape prosecutions failed in 2004, although many defendants were found guilty of related crimes.

Ministers are now considering allowing rape victims to give recorded evidence to court?although they would still face cross-examination. Expert witnesses may be allowed to explain why traumatised victims do not necessarily go immediately to the police. Most controversially, ministers may review the definition of ?capacity to consent?. At present, a drunk person may be regarded as capable of consenting to sex, even if he or she does not remember doing so afterwards.

The Criminal Bar Association, which represents both prosecution and defence barristers, is sceptical. The lawyers believe jurors are quite capable of distinguishing the kind of drunkenness that enables consent (though it may be regretted afterwards) and a state of deeper inebriation, which makes consent meaningless. They point out, too, that if expert witnesses are allowed into courtrooms, defence lawyers will simply summon their own.

Nor is Ms Kelly convinced that legal reforms will make much difference. Since so few cases get as far as the courtroom, she believes, effort must be concentrated on changing the attitudes of the police and the public towards rape.

Ms Stanko is working on a new approach. Her analysis of London cases suggests there are four broad types of offender?the stupid, the psychotic, the opportunist and the predatory. Once offenders have been profiled, she believes it will become much clearer if they are targeting a particular group. Juries may feel more confident in convicting rapists if they can see a pattern in the kind of victim who gives evidence?women with mental health problems, say, or people who had been drinking in a club shortly before an attack. London's police will shortly consider whether to adopt such a method.

Rape is a distressingly common experience. Six years ago roughly one woman in 20 and one man in 100 told the British Crime Survey that they had been raped at some point since the age of 16. And, at the moment, victims' prospects are poor. Frequently disbelieved, disregarded and humiliated in court, they often feel they have been assaulted twice.