Autism is something of a paradox. Since a seminal 1977 study led by Michael Rutter found that identical twins are much more likely both to be autistic than are fraternal sets, it has been understood that the disorder is heavily influenced by genetics. Three decades of research, however, have yet to reveal which genes are actually responsible.

A number of candidate genes have been identified. But with a few rare exceptions, science has not located autism mutations like those that cause cystic fibrosis or muscular dystrophy. It has not even found common genetic variants of the sort that affect people's risk of diabetes or prostate cancer.

One reason could be that there are no genes that have large effects on the condition: it could be influenced by dozens, each of which raises the risk by amounts too small to be detected. Another possibility is that very rare spontaneous mutations are responsible, which are specific to individuals and families.

A third explanation has been advanced by a team led by Francesca Happé, of the Institute of Psychiatry in London. It is that autism is probably not a single disorder at all.

Many people are now familiar with the idea that autism is a spectrum of disorders, ranging from high-functioning forms such as Asperger's syndrome to disabling conditions. Dr Happé's research, which she outlined last week in a keynote lecture to an international autism conference in London, suggests that even this is too simple.

For autism to be diagnosed, a child (or adult) must meet three criteria: they must show social impairment (such as failing to understanding other people's emotions), communication difficulties and non-social problems (such as repetitive behaviour and restricted interests - following a very rigid routine is a good example) . These traits, however, do not always go together. It is quite possible to have problems in one area alone: some parents and siblings of autistic people, indeed, show this.

Dr Happé, with her collaborators Angelica Ronald and Robert Plomin, has established from large samples of twins that the genes that underlie each element of the autistic triad do not really overlap. Autism may be a confluence of three disorders, each with different causes. It is no wonder few genes have been found: scientists may have been looking in the wrong place, at a condition that is not as homogeneous as it seemed.

This lack of a single explanation for autism, and the growing understanding that its component symptoms are not necessarily linked, has an interesting implication. It suggests that the behaviours that make up autism are not really abnormal: what is unusual is when genes place people at one end of a continuum for all three.

Perhaps everyone is a little autistic. Those we classify as such may simply have an extreme manifestation of traits that are present, to a greater or lesser extent, in us all.