The subject paper1concludes that number of faults per line of code is independent of whether Assembly language or a high order language (HOL) is used, contrary to previously published results by the author of this correspondence. However, the subject paper contains some technical errors and uses erroneous data. An explanation of the source of the erroneous data is given. Also, previously published information by the author of this correspondence on average number of operators plus operands per line of Assembly language code is corrected.

In this note, the number of faults or "bugs" per line of code is estimated based upon Halstead's software science relationships. This number is shown to be an increasing function of the number of lines of code in a program, a result in agreement with intuition and some current theories of complexity. The form of this function is investigated and an easy-to-use approximation is developed. An application to a moderately large software project is shown in which the predicted number of faults for program modules of various sizes agrees fairly well with the actual numbers of faults discovered.