This study describes the development, implementation, and testing of a computer program intended to help second-language learners read more proficiently in English. After two decades of literacy software and a mountain of research literature, what should a new reading program look like and what should it be trying to do?
With a growing proportion of all written communication now taking the form of machine-readable text, computers have a certain inevitability as a principal medium for reading and writing instruction. Computers are the paper of the future, but with vastly greater capacity for storage, interactivity, and links with other message-bearing media. And yet, there is an impression in the educational literature that not much has been found to do, instructionally, with all this text.
In a recent review of first-language literacy software and its related literature, Reinking and Bridwell-Bowles (1991) argued that some problems at a very basic level remain to be solved. At the most basic level, it appears that very little of this software actually gets used by students very much, probably because it is rarely integrated with classroom work. This under-use problem is not, of course, confined to literacy but exists across the curriculum, according to a general report on computer use in Canadian schools by Collis, Kass, and Kieren (1989), and a similar one on British schools in The Sunday Times of May 26, 1996.
Another basic problem is that literacy software is rarely of very high quality. Most tutorials result from one-phase development processes with no formative evaluation. In spite of the amount of research reported, little of it feeds back to the development process, and most of it is speculative or descriptive. Few studies use control groups, refer to learning research, or test specific hypotheses about outcomes. Of course, some of this goes with the territory. Control groups are notoriously hard to organize in the usual free-access environment of a computer lab, especially if a longitudinal training study is intended; and, conversely, increasing control rules out a longitudinal study and introduces novelty effects. Research is hard to build into a development program, since keeping abreast of the latest research and the latest technology are both large undertakings with one or the other likely to get slighted, usually the research. And of course hypotheses are unlikely to drive the development process if both testing and research are in short supply.
Rather than basing their software designs on research, developers usually base them on instructional concepts originating off-line, in classroom materials or coursebooks. These materials may or may not be based on any theory of learning. For example, in the 1970s reading theorists advocated practice in "reading sub-skills" (guessing the meanings of unknown words, finding the main idea of a paragraph, tracking pronoun references), none of which were thought to require actually reading extensive stretches of text, and accordingly the reading coursebooks of this period were largely reading courses with nothing to read (for example Yorkey, 1970), and a number of software applications were developed following this approach. Then, in the 1980s, research emphasized the importance of reading extended text (Nagy and Anderson, 1985; Krashen, 1989), but the coursebooks persisted with the skills-in-short-texts approach, and the software followed the coursebooks rather than the research. Reinking and Bridwell-Bowles (1991) cite a study of 297 literacy-related software packages of which only 21 required students to read connected text, the remainder focusing on individual letters, words, sentences, or skills. (One reason for this could be that the smaller units are easier to realize as computer exercises; extended text presents some programming challenges.)
But even if good, research-based off-line instruction could be recoded as a computer program, this is no indication that it should be. Anything that can be done as well on paper can be done better on paper, given the relative costs of the two media. Computers are justified only if they can do something that a paper medium cannot, something unique. In 1983 Wilkinson argued that "priority should be given to applications that employ the unique characteristics of the computer for displaying text." However, in 1996 the majority of reading tutorials still do not display text in any unique way that profits from the computer's storage or data-handling powers. Reading and writing software remain, on the whole, little more than "electronic books" doing in the computer lab precisely what is done in the classroom, except that the teacher gets a break and the learner gets slightly increased freedom and feedback.
Turning to second-language software, one finds an almost identical list of complaints. Kleinmann (1987) and Kenning (1990) discuss second-language reading and writing software in terms almost identical to those of Reinking and Bridwell-Bowles. But there is at least one difference; as Kenning notes, the problem of nothing to read is not so prevalent in second-language software. The tradition of text manipulation routines has at least put extensive texts on screens for learners to process in various ways (Stevens and Millmore, 1990; Cobb, 1993a), which if nothing else involves some reading and interactions with text that may not happen with print.
Still, there is no indication that text manipulation software has ever been used very much, or integrated very deeply into curricula, or subjected to very much research, whether hypothesis-testing or descriptive (Cobb and Stevens, 1996). In one of the few studies of user behaviour, Stevens (1995) found that texts were not necessarily processed at any great depth as learners proceeded through a manipulation activity. One problem with text manipulation may be that it poses questions to students about a text (Which word is missing from this space? What is the correct order of these sentences?), and then insists on literally correct answers, having no way to evaluate alternative answers, or nearly good answers, or give useful hints.
Kenning suggests that the way forward in literacy software is to carry on with the extensive texts but to change the orientation, from having learners answer the computer's questions, to having learners use the computer to ask their own questions. To achieve this change, she proposes a "prosthetic" or cognitive tools approach to software development, in which the computer handles some of the language processing that learners cannot initially handle for themselves, gradually handing over more and more of the work as sub-skills are automatized and integrated into complex abilities.
In writing, for example, if learners cannot initially think of ideas, write sentences, and watch their spelling all at the same time, then the computer can take over some of this work, freeing learners to concentrate on some other part. Special writing tools can help as idea generators, spelling checkers, sentence completion models, etc. In reading, various tools can be built into or behind texts to help learners look up words, move between texts or parts of texts, highlight patterns in longer texts that might be invisible to them otherwise, etc. The tutor creates a "zone of proximal development" for the learner to grow into, or a "cognitive partner." Salomon, Globerson and Guterman's (1989) "Reading Partner" and "Writing Partner" are examples of this approach for first-language development in children.
While the cognitive tools approach to literacy seems reasonable and even inevitable, it still remains fairly undeveloped, particularly for adult second-language reading. As Kenning (1990) observed, "So far, word processing is the only realisation of the [cognitive tools] model to have received much attention, but there is a growing interest in two other types: database programs and interactive concordancing facilities." The exploitation of databases has been developed somewhat since 1990, capitalizing on the handy supply of databases on the World Wide Web. Interactive concordancing has been exploited less, yet it is arguably an even more promising technology for language learning within the tools paradigm. Concordancing is a text analysis tool, normally used by linguists and lexicographers, but which learners might be able to use for purposes of their own in reading. The intention in the present study is to explore the instructional possibilities of interactive concordancing as an aid to second-language reading.
In summary, developing a piece of second-language reading software in the mid-1990s is a tall order. To compensate for the shortcomings of the past and exploit the opportunities of the present, such a piece of software should do the following things: It should be extensively used by a large number of students over a lengthy period; it should be integrated into an ongoing curriculum; it should be based on theories deriving from basic research; it should be tested for learning effectiveness against a control group and this information fed back to the development process; it should involve the reading of extended texts; it should use the computer to do things with text that cannot be done or easily done on paper; it should invite students to ask rather than answer questions. The concordance concept and technology make it possible to group these desiderata within an extended program of software development. Chapter 1 discusses the idea and background of concordancing, reviews other attempts to adapt the idea to instruction, and proposes a specific, theoretically motivated, testable application of the idea.
Krashen (1989) offered a backhanded challenge to those who would develop literacy software. In a classic paper, he argued that skilled reading in a second language depends mainly on knowing a lot of words very well, and knowing a lot of words depends mainly on a lot of reading. By reading, of course, he meant reading books:
If you lack books, get them. My suspicion is that reading is not only a way to develop vocabulary, spelling, and other important aspects of competence, it is the only way. We have no choice. The problem is not always money; often it is a matter of priorities. Just think how many books can be bought for the price of one computer (p. 455).
This should be kept in mind by anyone proposing to divert part of learners'
reading time into a computer activity: in what way is this activity more
useful than sitting down with a book?
| contents |