The split-half method is a way to estimate the reliability of a test by randomly splitting it into two halves and calculating the correlation between the two halves. The closer the correlation between the two halves, the higher the reliability of the test.