For many years I’ve been a staunch defender of GCSEs. I can no longer defend them. Exams should test understanding and GCSEs, in my opinion, don’t. For years we’ve been told that the system was being “dumbed down”. I refused to accept it.

I argued, quite correctly I think, that the breadth of study today is far beyond what I learned at O-level and A-level. I also argued that some topics have moved down from degree level to ASA2 and from there to GCSE.

How could we consider this dumbing down?

Last summer, for a TV programme, I taught O-level for the first time in 20 years. Because of this I’ve changed my mind - not about the breadth or the coverage of difficult concepts in the specifications, but about the depth of coverage and, more importantly, how we actually test children’s understanding.

Thirty teenagers, predicted to get at least one grade A*-C in science, were selected to spend four weeks boarding in the fictional 1950s Charles Darwin grammar school. The first thing we did was make them sit a battery of tests. I had great pleasure marking Fifties-style, complete with lashings of red ink, crosses and bold lines through what I considered gibberish.

Best of all was the freedom to write sarcastic, scathing comments such as “I regret wasting good red ink on this paper”, or “Sadly, trees died in order for you to produce this pathetic attempt at an answer.”

One comment I recall receiving as a child in a grammar school came flooding back and I used it with relish. A boy had written a fairly good description of the digestive process - his was one of the better attempts - but the way it was written implied that everything happened in the mouth. “This appears to be the product of a confused mind,” I wrote at the top of his paper.

None of the children could write a good description of a simple process.

They couldn’t convey that they had really understood what actually happens in digestion. But slipping into modern-day teachingmarking mode, the boy’s description would gain almost full marks. All the “key words” I would look for were there, more or less in the correct context.

The pedagogy of the 1950s was revealed in that one exercise. Children had to write essays in science; a descriptive narrative to explain processes or concepts. Precision was a requirement. As well as memorising certain key formulae or step by step processes, children had to communicate ideas clearly and articulate their understanding in grammatically correct prose.

When I looked at the sample O-level papers and syllabuses we had for research purposes, they amounted to no more than a couple of A5 sides.

O-level questions were to the point, even brutal, and there was an element of “guess what the examiner wants” in the essay-style questions. But they really did test understanding.

Looking at GCSE exam papers for 2005 reveals page after page of questions broken down into multiple parts, nearly all requiring just a few words to complete or, at most, a few sentences. Specimen GCSE science papers, matching the new science orders, make for disheartening reading. There is an increase in multiple-choice tests, filling in the “gaps” with the missing word, and linking words and descriptions with a drawn line.

The promise of more computerised marking will limit the style of questioning to preclude essay-type answers almost entirely. How does all this test understanding? There are no questions requiring pupils to articulate an argument or provide any form of descriptive narrative about their understanding of an idea, process or concept that is longer than about two sentences.

Perhaps this is being tested in the coursework - yet recent guidelines on coursework confirm what we have long known, but not really addressed, that plagiarism, inappropriate and unintentional help from parents and a little too much guidance from some teachers lead to inflated grades.

I don’t wish to return to O-levels. They required a certain degree of intellectual capability just to understand the question, let alone provide an answer. But instead of just breadth, how about some depth in our specifications? Instead of merely giving marks for the correct key words, how about trusting examiners to mark how clearly pupils communicate real understanding with at least one essay-style question per discipline?

As I tell my trainee teachers, you only know if you understand something when you have to explain it to others. When will our pupils get their chance to show to us in examinations what they really know about science?

James Williams is the deputy headteacher and housemaster in the new series of That’ll Teach ‘em, starting on Channel 4 on Tuesday. He is a lecturer in science education at University of Sussex