This concept represents a method of assessing the consistency of a measurement instrument, such as a survey or test. It involves dividing the instrument into two equivalent halves and correlating the scores on those halves. A high correlation suggests that the instrument is producing consistent results across its components. For example, a researcher might administer a 20-question personality inventory and then compare the scores on the odd-numbered questions with the scores on the even-numbered questions. If individuals who score high on one set of questions also score high on the other set, the instrument demonstrates a degree of consistency.
This technique is valuable in psychological research because it provides a relatively straightforward way to estimate the reliability of a test without requiring multiple administrations. This saves time and resources, and also avoids potential issues related to test-retest reliability, such as practice effects or changes in the examinee over time. Historically, it provided a practical alternative in situations where repeated testing was not feasible. However, the results are dependent on how the test is split, and different splits can lead to different estimates of reliability.