So far the IEA enterprise has produced twelve volumes written by participants, a volume reporting the Harvard Conference that assembled to discuss the results, with some of its associated papers, and a data bank for future exploration. It has also attracted various critical articles, of which Dr. Freudenthal's is one. In any investigation, the subsequent critical discussion plays an essential part and is therefore to be welcomed by the participants, particularly when, as was said of the Harvard meeting "The conference was marked by commendable restraint and a readiness to acknowledge important difficulties and limitations that persist in spite of best endeavours". An undertaking can be criticized both for what it did and for what it failed to do. In both cases reasonable criticism will take account of what it was practicable, in the circumstances, to do. By this standard not all of Dr. Freudenthal's remarks are reasonable.
For example, he says that the IEA statistics were static and should have been dynamic, and while he is silent upon the meaning of 'dynamic' it may well be that what he has in mind is the testing of innovations by something analogous to randomised field trials. This point is discussed in the technical volume of the IEA series, and in the Harvard papers. In the lEA study the correlations, and therefore the regression coefficients that stem from them, were all derived from systems in being. If the systems had been quite different the correlations and coefficients might have been very different. If the systems had been slightly different it is unlikely that the coefficients would have been greatly changed, and this seems to offer some possibility of inferring the effects of small changes by using the regression equations, though such inferences would remain doubtful until they could be confirmed or confuted by experiment. But in education experiment is easier to commend than to carry out. The idea underlying the field trial type of experiment is that most of the variables are to be controlled by randomisation. But where, as in education, there is a large hierarchy of variables, randomisation would involve interfering with the system to an extent likely to be thought intolerable, as may be seen from the fact that it is hardly ever employed. Moreover even if the data did come from a controlled field trial the inferences from them would still be uncertain, because they would depend on the general background of ideas against which they were interpreted. This background is much wider, but also much vaguer, than the immediate evidence. For each investigator it is a distillation from his