The Use of Covariance to Control Gradients in Experiments

In the spring of 1951 an experiment was devised to determine whether the exposure of tobacco seeds to different dosages of cathode rays would affect the growth of the resulting plants. The seeds were from a strain of tobacco which had been under controlled pollination since 1909, and, hence, the material used in the experiment was highly uniform with respect to its genetical background. The seven different treatments (the different dosages of cathode rays) were laid out in a randomized complete block experiment with eight replicates. The plot size was 2 rows by 10 plants with 3 feet between rows and 1.5 feet between the plants. The following measurements were made on each plant: