From the text... 'The negative and contradictory results of many of the earlier experiments in forestry are primarily due to the inadequacy of the experimental design. Treatments were seldom replicated over the experimental area and in the statistical analysis of the data, treatment effects could not be separated from random variation. Statistical methods wen~ used on a limited scale before World War II, for example in the construction of volume tables and sampling in forest inventory. After World \Var 11 they have been used extensively in experimental work in forestry and wood technology. But even today the necessity for a sound experimental design is not fully appreciated, particularly in Europe. It is often thought that statistical methods applied to biological problems are inadequate and even unreliable. The application of statistical methods of silviculture research is also impeded by the necessity for relatively large experimental plots. Sample plots of 0.4—1 acre are needed for many long-term experiments and these sample plots must be surrounded by a buffer zone of considerable width. The total area required for a field experiment with four treatments and four replicates will then be 12-16 acres. On this area the uncontrolled variation due to soil heterogenity will be considerable and this may obscure the true treatment effects offsetting the greater precision achieved by replication. On the other hand it should be realized that any experiment without replications is faulty as it fails to reveal the amount of random variation. Inferences drawn from these experiments are usually unreliable.'