Error in glasshouse experiments

Summary The proportions of plot variances that are assigned to error in glasshouse experiments at Bayfordbury are examined to determine the number of replications required per treatment.The relationship between average percentage standard error and weight of tomatoes picked by the end of a week is examined to discover which is the most suitable week on which to base an estimation of early crop. This was found to be the cumulative weight of fruit picked by the end of the fourth week after picking commenced.The gradients apparent on glasshouse benches and in crop houses were studied so that the experimental designs used could take these into account. Randomised blocks are suitable for experiments with small pots and germinating seeds on the benches. Where practicable a Latin Square design would be most suitable for cropping experiments with tomatoes grown in the beds, especially in the unequal span house orientated east and west, at Bayfordbury.