A number of studies show that, in the average, as a sentence increases in length it increases in difficulty (e.g., Coleman, 1962, Bormuth 1966). Average sentence length has long been one of the clearest predictors of text difficulty.

New readability formulas

Critics of the formulas and formula developers questioned the reliability of the criterion passages, criterion scores, and the reading tests on which the formulas had been developed and validated. The arrival of cloze testing stimulated the development of new criterion passages, new formulas, manual aids, computerized versions, and the continued testing of text variables.

The Coleman formulas Edmund B. Coleman (1965), in a research project sponsored by the National Science Foundation, published four readability formulas for general use. The formulas are notable for their predicting mean close scores (percentage of correct cloze completions).

Coleman was also the first to use cloze procedures as a criterion rather than the conventional multiple-choice reading tests or rankings by judges.

The four formulas use different variables as shown here:

C% = 1.29w - 38.45

C% = 1.16w + 1.48s - 37.95

C% = 1.07s + 1.18s + .76p - 34.02

C% = 1.04w + 1.06s + .56p - .36prep - 26.01

Where:

C% = percentage of correct cloze completions;

w = number of one-syllable words per 100 words

s = number of sentences per 100 words

p = number of pronouns per 100 words

prep = number of prepositions per 100 words

Coleman found multiple correlations of .86, .89, .90, and .91, respectively, for his formulas with cloze criterion scores. The use of cloze scores as criterion consistently provides higher validation coefficients than does use of the multiple-choice scores. This may be a partial reason for the high correlations shown here.

The Bormuth studies Recognizing the problems of having more reliable criterion passages, John Bormuth conducted several extensive studies, which gave a new empirical foundation for the formulas. His first study (1966) provided evidence of just how much changes in a number of readability variables beside just vocabulary and sentence length can affect comprehension. Cloze testing made it possible to measure the effects of those variables not just on the difficulty of whole passages but also on individual words, phrases, and clauses.

His subjects included the entire enrollment of students (675) in grades 4 through 8 of Wasco Union Elementary School district in California. Their reading levels went from the 2^nd through the 12^th grade. He used 20 passages of 275 to 300 words each, rated on the Dale-Chall formula from the 4^th to the 8^th-grade levels of difficulty. He used five cloze tests for each passage, with the fifth-word deletions starting at different words.

Reading researchers recognized that beginning readers relate differently to word variables than do better readers. For this reason, special formulas have been developed for the earliest primary grades such as the Spache formula (1953) and the Harris-Jacobson primary readability formula (1973).