No abstract

Although group measures may be compared without high reliability coefficients, more reliable mazes than are at present available should be sought. The desirable criteria are high odd-vs.-even and high half-vs.-half correlation, since these insure not only similar ranking of individuals at various points, but also permanence in the function measured. The results of 14 experiments with various mazes are evaluated with reference to these criteria, and show the T-mazes superior to the right-and-left mazes; of the former, the 14-unit form is the best of those examined, giving a corrected coefficient of .90. Time, retracings, and perfect runs are found to be ambiguous or otherwise invalid measures of learning, leaving errors as the only desirable type of score.

No abstract

No abstract

No abstract