A Framework for Readability Research: Moving beyond Herbert Spencer.
Albert J. Harris, Milton D. Jacobson · The Journal of Reading · 1979
Words. Spencer also anticipated another variable of readability equations in stating: economy of the recipient's mental energy... may equally be traced to the superiority of specific over generic words. That concrete terms produce more vivid expressions than abstract ones . The concept that abstract and concrete words relate toand concrete words relate to the difficulty of materials was incorcorporated into readability equations by Flesch (1950). Familiar Words and Short Words. Spencer sought reasons for the preference of Saxon to Latin words in English and attributed this to their greater familiarity through their greater use, and the fact that they are shorter. Both points have been incorporated into readability formulas; shorter words are determined by syllable or letter counts, and counts of familiar words are determined by counting words not contained in lists of familiar words. Using Spencer's Variables List of Familiar Words. Though Spencer identified unfamiliar words as a factor which contributed to the reading difficulty of a passage, there were no objective ways to distinguish familiar from unfamiliar words until Edward L. Thorndike's pioneering The Teacher's Word Book was published in 1921. This was a list of the 10,000 words occurring most frequently in a word count of 41 sources, including children's literature, elementary school textbooks, practical 392 Journal of Reading February 1979 This content downloaded from 157.55.39.49 on Mon, 29 Aug 2016 04:36:06 UTC All use subject to http://about.jstor.org/terms manuals, newspapers, the Bible, English classics, and adult correspondence-about 4,500,000 running words in all. Ten years later, Thorndike (1931) expanded his list to 20,000 words by making additional word counts from 200 sources and by taking some words from five other word lists. This list was further expanded to 30,000 by Thorndike and Lorge (1944). Thorndike's lists tell how common or familiar a word is in representative English material. In applications to readability, the Thorndike lists were used in formulas by Vogel and Washburne (1928), Washburne and Morphett (1938), Yoakam (1955) and Jacobson (1965). Two word lists by Edgar Dale (1931, 1948) were used in formulas by Gray and Leary (1935), Lorge (1944), Dale and Chall (1948) and Spache (1953). A list by Harris-Jacobson, Basic Elementary Reading Vocabularies (1972), was a major source for the revision of the Spache formula (1974) and for the Harris-Jacobson formulas (1973, 1974, 1975, 1976). It is evident that the development of these word lists clarified the determination of familiar and unfamiliar words and facilitated the inclusion of familiar words as an aspect of readability formulas. Sentence Length and Word Length. Relating sentence length to reading difficulty dates back to first efforts to quantify variables in formulas by Dale and Tyler (1934) and Gray and Leary (1935). Sentence length is measured by counting the number of letters, number of syllables or the number of words in each sentence in sample passages. As sentences begin and end in a characteristic way, sentence length can easily be determined in a mechanical fashion by manual counts or by computer. As manual counts are facilitated by counting words and computerized counts use either words or letters, most readability formulas use word counts to determine sentence length. In readability formulas, word length is determined by counting the number of letters or the number of syllables in each word in a sample passage. Counting the number of letters in each word is easy for a computer and easy, though tedious, for a person. Syllable counts are slightly more complex. Computer counts of syllables in a word must be based on estimates, while manual counts are relatively straightforward once certain ambiguous ways of breaking words into syllables have been resolved by the person doing the counting. Computers can estimate the number of syllables by counting the number of letters and dividing by 3.1127 (since most syllables are about 3 letters long) or by counting the number of vowels in each word and dividing by 1.1761 (since most syllables have one vowel). These estimating procedures give correlations of .98 and .96, respectively, with the number of syllables in a word (Felsenthal and others 1971). Thus word length seems to be adequately measured by counting letters, counting syllables, or estimating syllable