The author suggests that studies of infant speech sound data may be improved by statistical study. First, rank-order correlations were obtained from a sample of 1,047 records of speech sounds of infants from 1 to 30 mos. of age. 2 records were selected per month, 60 in all. Each record includes the sounds recorded during 30 expirations by a given infant. The results show the consistency of the sample randomly drawn. Second, the method of % of agreement offered a more detailed analysis of the reliability of the data. Third, adequacy of the size of the sample by range of value and number and percentage of value is > 0.7. Another sample was made by 64 records ranging from the 3d to the 30th mo. Satisfactory reliabilities were obtained by these various methods.