ABSTRACT
This chapter introduces a simple new procedure for obtaining a cumulative-type scale which should have properties of high reproducibility, high test-retest reliability, and high stability from sample to sample in rank order of cutting points. The H-technique produces a Guttman scale or a Lazars-feld latent distance scale with one important modification. The H-technique has been used on other sets of data with equal success, where the original item error, as indicated in the correlation tables with the provisional total score, does not exceed about 20 percent. Because of the high precision with which each cutting point on the cumulative scale is marked off by the H-technique, the rank order of the contrived items, if they are relatively widely spaced, should tend to remain invariant from sample to sample. The theoretical effect on reproducibility, test-retest reliability etc. appears to be almost the same as when no original item is used more than once.
