Applied Psychology Of Readability Research Paper

During the middle portion of the twentieth century, the first readability formulas were developed. Consistent with the underlying philosophy of behaviorism, these formulas considered readability to be a property of the stimulus—that is, the text itself. Most of the commonly used readability formulas, such as the Gunning Fog index and the Flesch Readability scales, were developed during this time.

The so-called cognitive revolution of the 1960s renewed interest in natural language. In the last 25 years, especially, the psychological process of language comprehension has received considerable attention. A number of wide-reaching theories of text comprehension have been developed (Gernsbacher 1990, Just and Carpenter 1980, Kintsch 1988, Kintsch and van Dijk 1978). These models do not assume the text to be a static entity; instead, they emphasize the process of reading, or more specifically, the interaction between the text and the reader. Ironically, contemporary researchers have essentially returned to the same philosophical positions that drove researchers such as Huey and Dearborn a century before.

1. Early Readability Research

The challenge facing educational psychologists in the 1930s and 1940s was rather pragmatic, and is still one of the goals of contemporary researchers and educators: to evaluate and compare readability of texts used in a variety of school settings. In theory, this would allow teachers to select age (or ability-) appropriate materials for use in the classroom. In addition, educators were looking for a way to assess the readability of various books students might read on their own.

Understandably, these formulas tended to follow from behaviorally oriented research: identify the objective (although generally superficial), easily measured properties in the stimuli—the texts. Early work tended to focus primarily on such simple ideas as vocabulary difficulty, word frequency, and sentence length (e.g., Gray and Leary 1935). It is important to note that the question of why these variables were related to readability was considered largely irrelevant. An understanding that sentence length was important because it provided an estimate of the demands placed on working memory was decades away.

Within a few years, readability formulas were developed that could be applied virtually to any text (Dale and Chall 1948a, 1948b, Flesch 1948, Klare 1963, 1974/5, 1984, Vogel and Washburne 1928). Though each of these differed in some respect, in truth they were remarkably similar. Most of these scales used some combination of word length (presumably related to word frequency or word difficulty), number of sentences per paragraph (a crude measure of complexity), sentence length, and/or number of syllables per word. They soon became widely adopted in academic, educational, and other applied settings. Standardized testing procedures also made wide use of these formulas. Some governmental agencies even specified readability levels for things such as insurance policies; in order to be ‘approved,’ for example, an insurance policy must have a readability grade level of 9 or below.

Despite the sophistication of contemporary reading research, most of the readability formulas in wide use today continue to use the same key variables. For example, recent versions of Microsoft Word provide an assessment of readability using the Flesch Reading Ease scale, as well as a normed score using the Flesch-Kincaid Grade Level scale (which scales readability relative to US public school grade levels). Cassady and Greene’s program Grammarian also provides an assessment of readability using the Gunning Fog index. (The Gunning Fog index uses the number of ‘long words’: those of three or more syllables.) These formulas are somewhat successful, at least to a point, and provide a reference from which texts can be compared.

Perhaps such widespread use of these simple formulas is not warranted. After all, these formulas analyze only a few variables, and cases where the formulas fail are abundant. However, the limitations of the formulas (the fact that they analyzed only a few components) were also their most attractive feature—they were easy to use. Thus, these criticisms, even when recognized by the researchers, have not attenuated the popularity of readability formulas.

Do revisions intended to improve the readability of a text actually produce more easily understood texts? One of the most thorough investigations in rewriting and readability was performed by Britton and Gulgoz (1991). They used a text of roughly a thousand words long, detailing the success and failure of the air bombing ‘Operation Rolling Thunder’ offensive in the Vietnam war (Greene 1985). They compared the original text (which had a grade level equivalent of about 13.5 years, as measured by the Flesch-Kincaid Index) to one of three revisions: a readability revision, which used shorter sentences and more frequently used words, improving the readability level to 12 years; a principle revision, which modified the text according to the comprehension model of Kintsch and van Dijk (1978) and had a readability level the same as the original text (13.5 years); and a heuristic revision, performed by an expert in text revision (Britton himself ), designed to make the text as understandable as possible. The reading level of the latter version was lowered to grade 12. The only revision which actually improved comprehension was this final revision. Lowering the readability score itself produced no significant improvement—simply using shorter sentences and more frequent words was not sufficient to make a text more understandable.

Improving readability has one other benefit—texts written at a level consistent with a reader’s ability are monitored more accurately while reading. Simply put, comprehension monitoring refers to the accuracy with which individuals can assess their comprehension. After reading a passage, individuals are asked to predict later performance on a test. The correlation between predicted and actual performance provides a measure of comprehension monitoring accuracy.

Weaver and Bryant (1995) examined the effect of readability on comprehension monitoring. They had college students read texts that were either easy (reading level below grade 8), standard (reading level around grade 12), or difficult (reading level around grade 16). Comprehension performance was related to readability: recognition performance increased as reading material became easier (more readable). Comprehension monitoring, in contrast, showed an inverted U-shaped function. The greatest metamemory accuracy occurred not for the easiest materials, but for those materials best matching the readers’ ability (standard readability, approximately 12th grade).

2. Summary

The theoretical case against readability formulas has been argued many times (e.g., from the present perspective by Kintsch and Vipond 1979). Their merits are clear, and center on their ease of use. But the drawbacks are equally clear. At best these formulas provide an oversimplification, too often a serious distortion. Nevertheless, they continue to be used, and will be as long as comprehension research does not provide viable alternatives.

Great progress into the reading process has been made in the past 30 years. But why have these advances not led to the development of viable alternatives to readability formulas?

The answer to this question has three parts. First of all, most of the research discussed above has dealt with the construction of internal representations, often called text-bases. A great deal is now known about how to construct comprehensible, recallable, and summarizable texts for given populations. However, while this knowledge is quite adequate for laboratory research on discourse processing, simple, easy-to-use procedures for practical applications have not yet been developed.

The second part of the answer involves a fundamental change in the nature of readability. Readability has typically been considered as a property of the text alone. Clearly, true readability is more complicated than that and is probably best considered as an interaction between the text itself and the readers’ ability and knowledge. Unfortunately, a simple readability formula incorporating readers’ abilities has yet to be developed.

The final part of the answer is purely pragmatic. When using a text in an experiment—often a relatively brief text—it does not matter whether the procedures for text analysis are time-consuming, require expertise, and so on. For the applied settings in which most readability formulas are used, though, ease of use matters a great deal. What is needed is a large-scale effort directed toward the development of practical methods of text analysis that allow non-experts to do with relatively little effort what only experts can do today with considerable effort. While such an enterprise would be nontrivial, it is certainly feasible. For the moment, however, readability measures have little connection to the cognitive science research on text comprehension.

In light of this fact, Bruce et al. (1981; see also Weaver and Kintsch 1991) have proposed that readability formulas be used only when the following criteria are met:

(a) The material will be freely read, and reading time will not be determined by ‘external factors’ such as experimental control or (in their example) captions that are presented along with television programs (for the hearing impaired).

(b) The text is ‘honestly written’: that is, not directly written in such a way to satisfy the criteria of readability formulas.

(c) The higher-level features of a text, such as organization and intention, correlate with the sentence and word features. As current formulas make little use of the top-level information, such material is otherwise ignored.

(d) The purpose of the reading is similar to that of the readers in the validation studies. If the two groups are reading with different strategies, then the formula may not be valid across the situations.

(e) The formulas are applied across a large number of texts, and individual texts are not singled out. Bruce et al. (1981) emphasize that readability scores are composite averages and may not apply to specific texts.

(f ) The population of readers is the same in the ‘real-world’ setting as it was in the validation studies. If a text is to be used as a remedial text, it should not be validated on a group of average readers.

Bibliography:

1. Britton B K, Gulgoz S 1991 Using Kintsch’s computational model to improve instructional text: Effects of repairing inference calls on recall and cognitive structures. Journal of Educational Psychology 83: 329–45
2. Bruce B, Rubin A, Starr K 1981 Why Readability Formulas Fail. Center for the Study of Reading, Champaign, IL
3. Dale E, Chall J S 1948a A formula for predicting readability. Educational Research Bulletin 27(January): 11–12
4. Dale E, Chall J S 1948b A formula for predicting readability. Educational Research Bulletin 27(February): 37–54
5. Flesch R F 1948 A new readability yardstick. Journal of Applied Psychology 32: 221–33
6. Gernsbacher M A 1990 Language Comprehension as Structure Building. Erlbaum, Hillsdale, NJ
7. Gray W S, Leary B E 1935 What Makes a Book Readable? University of Chicago Press, Chicago
8. Greene M P 1985 US Air Power: Key to Deterrence. Maxwell Air Force Base
9. Just M A, Carpenter P A 1980 A theory of reading: From eye fixations to comprehension. Psychological Review 87: 329–54
10. Kintsch W 1988 The role of knowledge in discourse comprehension: A constructive-integration model. Psychological Review 95: 163–82
11. Kintsch W, van Dijk T A 1978 Toward a model of text comprehension and production. Psychological Review 85: 363–94
12. Kintsch W, Vipond D 1979 Reading comprehension and readability in educational practice and psychological theory. In: Nilsson L G (ed.) Perspectives on Memory Research. L. Erlbaum Associates, Hillsdale, NJ, pp. 329–65
13. Klare G R 1963 The Measurement of Readability. Iowa State University Press, Ames, IA
14. Klare G R 1974 5 Assessing readability. Reading Research Quarterly 10: 62–102
15. Klare G R 1976 A second look at the validity of readability formulas. Journal of Reading Behavior 8: 129–52
16. Klare G R 1984 Readability. In: Pearson P D, Barr R, Kamil M L, Mosenthal P (eds.) Handbook of Reading Research. Vol. I. Longman, New York, pp. 681–744
17. Vogel M, Washburne C 1928 An objective method of determining grade placement of children’s reading material. Elementary School Journal 28
18. Weaver C A III, Bryant D S 1995 Monitoring of comprehension: The role of text difficulty in meta-memory for narrative and expository text. Memory & Cognition 23: 12–22
19. Weaver C A, III, Kintsch W 1991 Expository text. In: Barr R, Kamil M, Mosenthal P, Pearson P D (eds.) The Handbook of Reading Research. Longman, White Plains, NY, Vol. 2, pp. 230–45