The researchers looked at how well a common statistical test works when the data comes from a finite, discrete population. They found that the test is too cautious, especially with small sample sizes or when the sample is a big part of the whole population. This means the test might not always give accurate results in these situations.