A GUESSING MEASURE OF PROGRAM COMPREHENSION

An effective comprehension measure could be helpful in ranking program s on complexity. Measures involving filling in missing parts of a prog ram can be adapted from the prose domain. For example, doze tests have been applied to software. We evaluated a new measure based on a fill- in-the-blank exercise conducted by an automated tool. The subject is a sked to guess certain characters which are missing from a sample progr am which is displayed on a computer screen. The value of the measure i s derived on an information-theoretic basis from the number of incorre ct guesses. This measure has shown promise in experimental evaluations of its ability to measure program comprehension.