Alan Sturgess' letter about "fiddling" test results ( TES , 4703) raises an 'old chestnut' about exam standards. If only it were as simple as he suggests. In some kinds of tests (pre-tested objective test questions, for example), pass marks can be fairly reliably determined in advance, but in others this is not sensible. With the best will - and expertise - in the world it is impossible to be absolutely confident about the demands of a set of examination questions, even when these may have been trialled in advance. Some tests and mark schemes turn out to be more demanding than others and this cannot be confirmed ahead of the test being taken and marked.
Therefore, decisions about a correct boundary mark can only be taken when all aspects of a test have been reviewed and compared with previous performance: the demands of different questions; the expectations of the setters; the actual performance of candidates; the statistical distribution of marks etc. At this point senior examiners and officers determine what they consider to be the most appropriate mark in the light of a multitude of evidence.
It is not and can never be an exact science and those of us involved in these processes usually do their best to be fair to both a hypothetical standard and to the candidates. There are undoubtedly potential weaknesses in the process, but currently it is the best that has been devised so far.