School of Education   Home   Contents   Previous Page   Next Page


This chapter includes discussions of the theoretical bases for the study. First, the psycholinguistic approach to reading comprehension is described. Cloze procedure as a measure of reading comprehension fits in well with the psycholinguistic approach, and the theoretical bases of cloze procedure are described next. Third, there is discussion of readability and how it relates to reading comprehension. The cloze and maze procedures as measures of reading comprehension are detailed next. The chapter concludes with a brief discussion of computer based measures of reading comprehension.

Psycholinguistic Approach to Reading Comprehension Measurement

The psycholinguistic approach to reading comprehension measurement provides a sound theoretical basis for this study. This approach can be explained in the terms of Osgoodís language communication model (Anderson, 1976, p. 15) which consists of three components: Source System, Message System, and Receiver system as depicted in Figure 2.1.

Figure 2.1  Osgoodís representational model (after Anderson, 1976)

The process of language communication commences with the writer encoding language into written form to produce printed words (or text). The reader views these printed words as sequences of visual stimuli. In order to decode these words into meaningful language, the reader makes use of the semantic and syntactic cues within the text. All the readerís past learning experiences, including language development, are used to perform this decoding. Reading comprehension then is defined as "the correspondence between the way in which a message is encoded by the writer and the way in which it is decoded by the reader" (Anderson, 1976, p. 16).

Conceptual Framework for Cloze Procedure

When Taylor introduced a new technique for measuring readability and reading comprehension in 1953, he named it the cloze procedure based on the gestalt psychology that people tend to complete a familiar but incomplete pattern by closing in the gaps. For example, a broken circle (Figure 2.2) is usually recognised as a circle because its shape is so well-known that people readily fill in the gaps mentally (Rye, 1982, p. 2).

Figure 2.2  A broken arrangement of curved lines is recognised as a circle

Taylor applied the same principle to language. Although this explanation had a number of short comings, Taylor proposed an alternative rationale based on Shannon and Weaverís generalised communication model. Anderson (1976) modified Shannon and Weaverís model for language communication and it can be used as a basis in the conceptualisation of the cloze procedure (see Figure 2.3).

Figure 2.3  A model for the language correspondence of a source system to a receiver system (Anderson, 1976, p. 15)

The language communication model describes the transmission and receiving of the message as coding procedures. The source (the producer of the message) in the cloze procedure is the author of the text. Noise (the deletion of words in the cloze procedure) distorts the message before it is received by the decoder (reader). In order to understand or decode the message, the reader needs to reconstruct the text using contextual cues and previous language experiences. This modelís usefulness extends to the generation of hypotheses. For example, "language style" studies focus on the source system, and investigations of decoder attitudes focus on the receiver system. The present study focuses on the message system. While the printed words are the same in each of a series of reading comprehension tests that are developed for the study, the mode or message system for presenting these will vary.


Readability and reading comprehension can be seen to be related concepts. While readability is characterised by the degree of comprehension with which a text is read, the ability of a reader to comprehend a text is influenced by the textís readability. In order to select a text for assessing reading comprehension, the readability of that text needs to be appropriate for the reading ability of the target population. Determining readability of a text can be quite complex. Rye (1982, p. 14) suggested that there are eleven major, often difficult to measure interacting factors (see Figure 2.4) which influence readability.

Figure 2.4  Factors in readability (Rye, 1982, p. 14)

Formulas, based on word difficulty and sentence complexity, have been developed for calculating readability. However, these formulas are complex and time consuming to administer. Propitiously, advances in technology have allowed the construction of computer programs that analyse texts quickly. Such a program is MacTexan (Anderson, 1991). It delivers information about text for analysis, including estimated measure of reading difficulty, year level for which the text is appropriate, sentence profiles, and word frequencies. This program has been used to measure the readability of the text chosen for the cloze and maze reading comprehension measures.

Cloze and Maze as Measures of Reading Comprehension

Andersonís analysis of thirteen cloze studies (1976, pp. 35-39) in relation to a variety of standardised reading comprehension tests produced reliability coefficients ranging from 0.25 to 0.82. He judged that the cloze tests (in those studies which reported reliabilities) were fair measures of generalised reading comprehension. Cloze tests which are compared to reading comprehension tests of the same passage measure specific reading comprehension. Anderson (1976, p. 39) concluded from his analysis of such studies "that cloze tests are valid measures of specific comprehension at primary, secondary, and adult levels."

Parker and Hasbroukís (1992) study of maze tests reported validity coefficients of 0.85 and 0.82 between Guthrieís maze tests and Gates-MacGinitie Vocabulary and Comprehension Tests. The mean validity coefficient over five similar studies was 0.58. These authors also examined studies with comparisons between maze, cloze and standard reading comprehension tests. Two separate studies obtained correlations of 0.52 and 0.55 between maze and standard tests, with similar results for the cloze. These statistics are indications that cloze and maze tests are valid measures of reading comprehension.

Computer Based Measures of Reading Comprehension

There is a difference in the "message system" when a reading comprehension test is presented on a computer screen as opposed to on paper in a traditional style comprehension test. The language communication model suggests that the readerís decoding of the message may be affected by the message system. Several studies on reading comprehension have compared paper and pencil tests to computer based tests.

In 1983, Gerell and Mason found that students achieved higher comprehension scores on computer rather than paper when the reading content was presented in "chunks". Conversely Farstrup (1985) in a study with 85 adults found that participants fared better on paper. Blanchard (1987) found no significant differences between reading comprehension tests on paper versus computer for 30 primary school students.

Andersonís study (1987) with primary school students also reported no significant differences between versions of the GAP test, even though there were major differences in presentation. Students completing the computer based tests had to select each passage in turn, complete each passage in order of presentation, and could not re-examine their completed work. The paper version did not have these restrictions. Remedial students, in spite of the presentation differences, performed measurably better on the computer version. Anderson attributes the higher scores on the computer to increased student motivation.

In other studies of computer based testing, it was found that paragraph comprehension is slowed by low resolution screens (Bunderson, Inouye and Olsen, 1989). In a study of testing on a university computers, Whittington, Saslaw and Carrigan (1995), reported that scores for the paper and pencil version of reading comprehension exceeded to a significant degree the scores for the computer version. When questioned, students told of problems with the computer screen such as difficulty seeing bold print, blurred vision, frustration with having to page back and forth between text and questions, and difficulty keeping their main thoughts due to having to backtrack from screen to screen. Students also preferred to complete the paper versions of the tests with other students in a familiar environment as opposed to the computer version on the university mainframe.

The differences in test performance in these studies can be attributed to elements summarised by Rye (1982) as factors influencing readability: the ability and desire to read, type of print, line spacing, organisation of material, and physical environment. Other factors, like syntax, sentence length, word length, word frequency and subject matter are assumed to have been the same in the paper and computer versions of the tests completed in the various studies.

Many of the limitations of computer based testing noted in the above studies do not apply to the present study. First, high resolution screens greatly reduce difficulties with screen clarity, and second, the hypertext format allows easy movement from one part of the test to another, passages may be completed in any order, and it is easy for answers to be changed.

Definition of Terms

Browsers are software applications used to view hypertext documents, interpret the links embedded in the documents and access the linked documents on demand.

Cloze Procedure is a method of systematically deleting words from a text selection and then evaluating the success a reader has in accurately supplying the words deleted.

Computer based testing "uses a computer to give exactly the same test as one in a paper and pencil format. That is, it has the same test questions and presents them in exactly the same order as the paper and pencil version of the test." (Bugbee, 1996, p.283)

Download refers to the act of copying files from a remote computer to the userís computer.

HTML refers to Hypertext Markup Language, the programming language used to create hypertext documents.

Hypertext is a software based tool that allows users to move from one document or part of a document to another via linked keywords.

Internet is "a loose connection of thousands of smaller networks in different countries around the globe. It links hundreds of thousands of academic, government, military and public computer systems together, enabling (literally) millions of people from diverse cultures to share information and other resources" (Neely, 1995, p.10).

Maze is a "multiple choice extension of cloze", in which "the blank space characteristic of cloze is replaced by three choices: a target, a semantically incorrect foil, and a syntactically incorrect foil." (Gillingham & Garner, 1992)

Online refers to the successful connection with another remote computer.

Readability is an estimate of the difficulty of a text. Formulas, which include factors such as sentence length, word length and word frequency have been developed to calculate readability.

Reading Comprehension is defined from a psycholinguistic viewpoint as "the correspondence between the way in which a message is encoded by the writer and the way in which it is decoded by the reader" (Anderson, 1976, p. 16).

Upload refers to sending files or information from the userís computer to another computer.

World Wide Web (WWW) is a hypertext based system linking information and files on different computers around the Internet.