Are online autism tests for adults accurate?

A self-observation questionnaire measures one thing only: the way you describe your own functioning at the moment you answer. That is useful, because it puts words on situations people often live through without naming them. It is neither an examination nor an objective measurement.

Published scales were built to help sort people, not to settle the case of one person. The Autism-Spectrum Quotient (AQ), described in 2001 by Baron-Cohen and colleagues, has fifty items. Its short version, the AQ-10, has ten. It was published in 2012 by Allison, Auyeung and Baron-Cohen.

A study by Booth and colleagues, published in 2013, compared the AQ-10 with the full version in 149 adults whose autism had been recognised by a professional and 134 adults without that recognition. The authors conclude that performance is close at the usual thresholds, and that the short version can serve as a first filter.

Because the make-up of the group changes everything. In the 2013 study, 149 of the 283 participants, or 52.7% (a calculation made here from the published group sizes), had a recognised autism profile. In the general population the share is far smaller.

The value of a result depends on that share. When half the group is concerned, a high score is a strong signal. When very few people are, a large part of the high scores belong to people who are not. The questionnaire has not changed; the situation has. Performance from a research study therefore does not carry over unchanged to someone answering from their sofa.

A high score means your answers describe traits that often turn up in this context. It does not say where they come from. Marked anxiety, an attention disorder or lasting exhaustion can produce similar answers on some items. Our ADHD or autism guide goes through these overlaps.

A low score is no more reassuring. The items on short scales mostly describe visible behaviour. A person who has learned, sometimes over years, to compensate in company may answer differently from what they live inside. Our guide on social camouflaging describes this. No short questionnaire can measure it on its own.

In both cases the practical conclusion is the same. The result establishes nothing and rules nothing out. It helps you decide whether it is worth talking to someone whose job it is.

Seven markers are enough, and every one can be checked before you answer the first question.

• The page says where the items come from, and names the scale and its publication when it reproduces one (AQ, AQ-10, RAADS-R or another).
• It states which version is used: the AQ-10 and the full AQ are not read the same way.
• The scoring is explained, not merely displayed.
• An update date appears on the page itself.
• The limits are written in the body of the page, not in small print.
• The result appears without creating an account and without leaving an email address.
• No result looks like a verdict: no autism percentage, no label, no sentence that concludes on your behalf.

The seventh marker is the most telling. A questionnaire that announces an “autism percentage” promises more than the published scales themselves, whose authors speak of a sorting tool or an aid to assessment. The RAADS-R, published in 2011, says in its title that it is meant to assist a professional's assessment.

1. Write down three to five concrete, dated situations where something weighed on you: a family meal, a meeting, an exchange where you were playing a part. One example is worth more than a score.
2. Work out how far back these situations go. How long the difficulties have existed is the first thing a professional will look for. If people knew you as a child, their memories count.
3. Talk to your family doctor. They are the ones who refer you, and waiting times are often long, so starting early saves months.
4. If you do not know where to start, our guide “I think I'm autistic: what now?” describes the pathway step by step.