The experiment, devised to test McDougall's claim that the effect of training in rats is inherited, has been carried out for twenty generations. In addition to the trained line, a control line has been maintained parallel with it, from which a number of rats have been trained in each generation, but not used for breeding. For each generation of the trained line there is therefore a corresponding group of trained control rats for comparison, differing from the rats of the trained line only in that they have no trained ancestry. During the first fifteen or sixteen generations there was a progressive, though irregular decline in the number of errors made in each generation in both lines. In generation 18 both lines showed a marked increase in the number of errors made, with fluctuations in subsequent generations running closely parallel in the two lines. This parallelism of periodic fluctuations in rate of learning in the two lines makes it impossible to attribute a progressive change in the trained line, when it happens to be in the direction of decreasing number of errors, to the inherited effects of ancestral training. Our experiment is being continued, and therefore our conclusions must be regarded as tentative only. The results of the experiment up to the present, together with those of Crew's experiment, show however, that the progressive decrease in the number of errors in successive generations of McDougall's experiment, in which no control line was maintained, cannot be held to have established the operation of Lamarckian inheritance.

This content is only available via PDF.