Two reading tests have been constructed which require children to restore the missing words in a series of mutilated paragraphs. The missing words are virtually completely redundant to an efficient reader. For two independent groups of randomly selected children ac each grade level from 3 through 7 and two independent groups of University undergraduate students, results indicate that, if the word redundancy of passages of English written prose is estimated from restorations of mutilated versions of those passages by independent samples of the same population, then the estimated redundancies vary as simple powers of one another. This relationship appears to have potential significance for, among other things, readability assessment.