Washington-based psychologist John Gottman is able to watch a video clip of a husband and wife talking and predict with 95 per cent accuracy whether they will still be married 15 years later.

It is known as “thin-slicing”, the ability of our unconscious mind to make judgments based on very narrow samples of experience. According to self-proclaimed “intellectual adventurer” Malcolm Gladwell, who reports the Gottman story in his book Blink, we shouldn’t be surprised. Instinct matters, Gladwell says, and nowhere more than in the classroom.

So when American psychologist Nalini Ambady gets students to judge teachers based on 10-second clips of them teaching, there is one problem: 10 seconds is too long. Ambady’s research reveals that students can predict teaching quality with astonishing accuracy based on a clip without sound lasting, incredibly, two seconds.

Veterans of the classroom know this. Phil Beadle, author of How to Teach, says you can spot a poor lesson as soon as you walk in if students are wearing coats, bags are on desks, and students are holding pens while the teacher is talking.

This is topical because in her farewell speech after five years at the helm of Ofsted, chief inspector Christine Gilbert said something rather startling.

She publicly revealed education’s open secret: that being graded “outstanding” rarely equates with outstanding teaching. Indeed, only 30 per cent of outstanding schools exhibited teaching of the highest quality. The solution, it seems, is that Ofsted will observe more lessons in future.

There are contradictions here. First, remember that schools which have already been judged outstanding aren’t going to be re-inspected unless required by the annual desk-based data exercise - a process akin to an iPhone app that allows you to point your phone at a night sky to confirm you are looking at stars.

So, a large number of schools will be ruled out of the inspection process by default. More significantly, I’m not sure that what Ofsted calls outstanding teaching is what we and our students might recognise as such. Ofsted’s reliance on a rather mechanical view of students’ progress may actually militate against great teaching. Here’s why.

In his study of ability, The Talent Code, Daniel Coyle cites a 1995 study that compares the classrooms of Japanese and American schoolchildren. In Japan, 13 to 14-year-olds spend 44 per cent of class time “inventing, thinking and actively struggling with underlying concepts”. In the US, Coyle reports that American students spend less than 1 per cent doing the same. Put bluntly, “the Japanese want their kids to struggle”.

Good teachers understand that sometimes a lesson will be better if the students don’t know what the learning objective is. Sometimes students will regress as well as progress. To improve we sometimes have to slip frustratingly backwards.

I’m in that white-knuckle phase of life watching my son learning to drive. If anything demonstrates the myth of continuous upward progress, it is this. One day his gear changes are smooth; the next we are lurching without dignity through a sequence of juddering false starts. He is learning and it is a messy, occasionally stomach-churning, business.

Although it is unlikely to tick an Ofsted inpection box, in the many classrooms I visit I like to see students doing more and the teacher doing less, and questions that initially seem to demonstrate lack of comprehension, but illuminate a deeper level of engagement with the topic.

That is why I agree with Michael Gove that modular exams have reduced too much learning to a conveyor-belt of test preparation. But we delude ourselves if we think that merely by observing more lessons, Ofsted teams will uncover better teaching.

After all, it is a brave teacher who takes serious pedagogic risks when an inspector calls. The Ofsted obsession with spotting students making demonstrable progress in a 20-minute observation window can force even our most talented teachers to pick up and deliver an off-the-shelf version of what good teaching is supposed to look like. Teaching thus becomes tame, pre-packaged, synthetic and less personal.

And because the stakes are so high, few teachers and leaders have the nerve to not play the Ofsted game and replicate some Meccano-style teaching model.

So, I’m not sure I agree with Ms Gilbert’s theory that inspectors should watch more lessons. It may be that the real teaching and learning - the genuinely outstanding stuff - happens when inspectors’ backs are turned and they are off analysing their data.

Geoff Barton is headteacher of King Edward VI School, Bury St Edmunds.