Introducing some more challenging questions and requiring a mark of 90 per cent for an A* grade may seem like a way of ensuring it is really hard to get. But is it really? If all candidates take the same exam, most of it still has to be accessible to those who will achieve at the lower end.

One definition of a challenging question would be where the candidate has about a 5050 chance of getting it right. Simple arithmetic shows that you can’t have too many questions like that and expect anyone to get 90 per cent, even if all the other questions are so trivially easy for them that they are virtually certain to get them right.

From a measurement point of view, given the statistical uncertainty that surrounds a particular score, a candidate who gets 90 per cent correct is pretty much at the ceiling of what a test can measure. It seems more intuitively obvious that someone who gets 100 per cent cannot really have been stretched, and the difference between 90 per cent and 100 per cent is usually not much more than the error of measurement.

There is an important difference between a candidate who plays it safe and avoids making careless errors on a paper that is really too easy for them (hence gets over 90 per cent) and one who copes creditably with difficult questions that are challenging and unpredictable (hence gets a much lower percentage correct).

Elite universities who require the A* may prefer the latter but they will get the former.

Despite these ‘top end’ problems, which affect a small but vocal minority, A-level remains well suited to the needs of the vast majority of students who take it.

We should certainly not return to the so-called “golden age” when almost a third of candidates failed and about another third achieved pass grades but were left feeling they did not really understand any of it.

Robert Coe is director of secondary projects at Durham University’s curriculum, evaluation and management centre.