04.12.2012 Views

The SRA Symposium - College of Medicine

The SRA Symposium - College of Medicine

The SRA Symposium - College of Medicine

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Papers<br />

<strong>of</strong> vowels per word, some on the number <strong>of</strong> consonants per word. Some count a sentence whenever<br />

a period, question mark or exclamation point is found, some whenever a colon or semi-colon<br />

is found.<br />

Hochhauser (1997) analyzed one consent form using six s<strong>of</strong>tware programs, not including Micros<strong>of</strong>t<br />

Word. As shown in Table #2, they did not all agree.<br />

Table #2: Comparison <strong>of</strong> six readability s<strong>of</strong>tware programs<br />

Six readability<br />

programs<br />

Number <strong>of</strong><br />

sentences<br />

Number <strong>of</strong><br />

syllables<br />

Words per<br />

sentence<br />

Flesch<br />

Reading Ease<br />

Flesch-<br />

Kincaid<br />

Range 30-35 1,201-1,313 22-25 37-45 12.1-15.5<br />

Average 31.9 1,279 22.9 40.5 14.0<br />

Using the same readability formula but in different programs found the Flesch-Kincaid varying<br />

from 12.1 to 15.5—a difference <strong>of</strong> 3.4 grades! Such s<strong>of</strong>tware differences show that readability grade<br />

levels will vary considerably for the same formula depending on which s<strong>of</strong>tware program is chosen<br />

to calculate reading grade level. That’s a major problem for consent form readability research.<br />

Writing computer programs to count sentences, syllables and words is more difficult that it appears.<br />

In an early attempt to computerize readability formulas, Fang (1968) developed a computer<br />

program for the Flesch Reading Ease Score, but noted that “<strong>The</strong> basis <strong>of</strong> the syllable counter is<br />

that one vowel equals one syllable. However, there are many exceptions, and many exceptions to<br />

the exceptions. Two years later, Coke and Rothkopf (1970) stated that “A word may be defined<br />

as any set <strong>of</strong> alphanumeric characters delimited by blanks and punctuation marks, while a sentence<br />

can consist <strong>of</strong> all words occurring between two periods.” But consent forms usually include<br />

headings or subheadings that would become part <strong>of</strong> the next sentence, as well as many periods<br />

(such as M.D., R.N., e.g., etc.), colons, semicolons and hyphenated technical terms that make exact<br />

counts difficult. Counting syllables is even harder. Coke and Rothkopf tested three methods for<br />

estimating syllable counts in readability formulas: 1) total number <strong>of</strong> vowels, including “y”, 2) total<br />

number <strong>of</strong> consonants, and 3) total number <strong>of</strong> letters, and concluded that the number <strong>of</strong> vowels<br />

per word had the best correlation to the researcher’s syllable count by hand.<br />

Since readability s<strong>of</strong>tware programs do not disclose the methods used to count sentences, words<br />

or syllables, it’s impossible to know if computerized formulas are consistent with the scores that<br />

would be obtained if calculated by hand. <strong>The</strong>se computational choices explain the puzzling findings<br />

by Mailloux, Johnson, Fisher, et al, (1995) who found inexplicable differences in their comparison<br />

<strong>of</strong> four readability s<strong>of</strong>tware programs: “It is interesting to note that, despite the fact that<br />

the Flesch-Kincaid formula and Gunning Fog Index formulas were reported to be identical across<br />

s<strong>of</strong>tware programs the Flesch Reading Ease Formulas were nearly identical, the s<strong>of</strong>tware programs<br />

provided different grade results. This finding is difficult to explain because if the formula were<br />

truly identical, no discrepancy should be found” (p. 224). Unfortunately, they did not compare the<br />

computerized readability grade levels with a level calculated by hand, so they could not state which<br />

s<strong>of</strong>tware readability was most or least accurate; that comparison study has yet to be done.<br />

But such discrepancies should be expected if different programmers choose different ways to<br />

count sentences, words, and syllables. <strong>The</strong> discrepancies are not with the formulas, but with the<br />

programmers who convert the formulas into s<strong>of</strong>tware code. <strong>The</strong> discrepancies found by Mailloux,<br />

et al, might also be due to their methodology <strong>of</strong> scanning the documents into a computer and analyzing<br />

them via readability formulas. Plus, they did not state if they “cleaned” the file (to remove<br />

extra periods) before running the readability formulas.<br />

118 2005 <strong>Symposium</strong> Proceedings Book

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!