[article | discuss (0) | ]
Graham Davies' "Using Computers in Language Learning" (with a separate section on ESL by John Higgins), is for the most part a useful guide through the maze of programming languages, authoring systems, exercise formats, etc., which face the teacher contemplating the use of computers for language learning. Unfortunately, however, much of the information in the text refers to British sources, limiting its usefulness to the American teacher. A more serious flaw is the underlying assumption that computers are best suited for mechanical drill, thus freeing the teacher for communicative activities. The reviewer suggests that the continued use of computers in this fashion may well lead to the demise of CALL before experimenters have had an opportunity to explore its full potential.
KEYWORDS: artificial intelligence, CALL, communicative activities, courseware design, drills
Ours is a curious science. The more our work becomes common and everyday, the less we seem to know what it is about. We are still not sure whether the computer is a supplementary aid to be taken into the classroom, or whether it is best used by learners on their own in a library or language lab. There is still no reliable statistical evidence to confirm that using computers does in fact increase students' proficiency in the language, although we have accumulated plenty of informal or anecdotal evidence to suggest that this is the case. Some of us are not even certain that such confirmation matters; perhaps the value of practicing with a computer is in fact some elusive gain of a sort which cannot be measured with an achievement test. What's more, with the current emphasis on teaching for proficiency, which is necessarily measured globally, it is no longer obvious that we should be looking for results in terms of
achievement on discrete-point tests at all, but instead through performance in communicative discourse—clearly much harder to do on a computer.
One senses then, a feeling of ambivalence: what we can do easily is probably not what we should be doing, yet we'll do it anyway, because the alternatives are too complicated, too difficult, or require a bigger machine. After attending sessions at CALICO on artificial intelligence and parsing strategies, we will go home and punch in yet another BASIC drill, because the Al people do what they do and we do what we do.
Using Computers reflects both this ambivalence toward computer-assisted language learning (CALL) and the apologetic attitude that usually accompanies it. The second edition of a handbook published in 1982, the text now carries the subtitle "A Teacher's Guide," which suggests just how common the use of computers has become. The aim of the guidebook is to provide the teacher with a survey of the field, from a brief history of attempts at CALL to a list of current software sources.
The present edition consists of six chapters ("sections"). The first, "New Technology for Linguists," discusses how computers are used to manipulate natural language, including such things as machine translation, literary analysis, speech synthesis, and—of course—CALL. After pronouncing a cautious, and by now probably unnecessary caveat, to the effect that computers "can never replace the 'live' teacher, especially in language teaching, where the emphasis is on communication" (Davies 1985:8), the author proceeds to sketch a brief history of early mainframe CALL efforts, mostly British. Since the late 1970s, he says, the scene has shifted to an emphasis on homespun materials for microcomputers, which he describes as ". . . hundreds of individuals . . . sitting in backrooms [sic] earnestly re-inventing the wheel" (Davis 1985:9). Nevertheless, the excitement of the early 1980s has now faded considerably, replaced by a more realistic and pragmatic view of the limitations of computer learning.
Section Two, "Approaches to CALL," contains most of the substantive material in the book. It should prove useful to the teacher who needs a concise discussion of a strategies for exercise design or answer-processing. Davies discusses the relative merits of the Socratic information-plus-question approach, the quiz format, fill-in exercises, multiple-choice, and the problems of free-format exercises of the type which allow for a variety of possible answers. He continues
with an overview of response analysis, the manner in which the program analyzes the student's input for errors. Next, he treats "partial matching," a way of detecting and pointing out incorrect letters; error diagnostic routines, which can not only point out the error but can comment on it in morphological terms ("La terminaison n'est pas correcte"). This is followed by a useful list of 20 "input validation routines," strategies for controlling and editing input to assure that it conforms to the format expected.
In the same section (perhaps better treated separately), Davies describes types of CALL programs: simulations, "text mazes" (short pieces of text with alternative solutions, in the manner of Write-Your-Own Stories), adventure games, cloze exercises, etc. Finally, still in the same section, Davies explores peripheral questions such as the use of audio, graphics, color, artificial intelligence, and interactive video.
Section 3, "Computers in English Language Teaching," is a separate contribution by John Higgins of the British Council, most of it echoing work Higgins has published elsewhere (e.g., Higgins and Johns 1984). Higgins here distinguishes between what he calls "serendipitous" software--programs which, although designed for some other purpose, turn out to be useful for language learning (for example, word-processing, or English adventure games for ESL practice)—and "dedicated" software, which was expressly designed for language instruction. Characteristically, Higgins has a broad view of what the latter could include. His ideas go far beyond the traditional drills described by Davies in the previous section. They include such things as exploratory programs which can be "taught" by the student, conversation programs of the ELIZA type, and his own micro-world programs, which he refers to collectively as GRAMMARLAND.
Section Four, "Ready-Made CALL Software, Authoring Packages and Author Languages," may prove less useful to the American language teacher, since many of the programs or packages described are of British origin. Seen as generic examples, however, these same descriptions give one a fairly good idea of the possible variety of CALL programs, their limitations and their strong points. For example, Davies' discussion of the preparation of cloze exercises is enlightening: Should we attempt to program the exercise so that it will accept whatever syntactically or semantically correct word could be typed in the blank? No, he concludes; the considerably greater amount of time needed to prepare
such a text would not be justified by a proportionate gain in usefulness to the student.
None of the three sources of software described in this section is, according to Davies, without problems. What he calls "off-the-peg" (i.e., off-the-shelf) software is often unsuitable to the teacher's needs, and can rarely be adapted. Authoring languages such as PILOT or MICROTUTOR offer the most programming freedom, but are also, for that reason, the most time-consuming. Davies prefers the authoring system, a program (such as, in this country, DASHER), which only requires that the teacher type in the cues and the correct responses in order to create an exercise. But again, he points out, there is an obvious drawback: the exercise can only do what the programmers designed it to do. Where will better designs come from? The answer, says Davies, is more teamwork: "The elegant solution to software programming is collaboration between linguists and programmers" (Davies 1985:75).
The book concludes with a section on BASIC, another on choosing hardware, a bibliography of mostly British references, and a Software Index--an alphabetical listing of more than 200 programs and packages, complete with a one-paragraph description of each and the address where it can be obtained. Unfortunately these, too, are mostly British.
As a guide for the American language teacher, Davies' handbook is clearly flawed.2 There is a more serious flaw, however, than the many references to unknown British products. Underlying a technically correct discussion of CALL are some questionable assumptions about the nature of the enterprise—assumptions once again tempered by the apologetic ambivalence we spoke of earlier:
1. The computer's role is that of surrogate drillmaster: "The computer can relieve teachers of much of the drudgery and release them to concentrate on communicative skills" (Davies 1985:15). This view of CALL is one which has been often criticized, even by Davies' co-author (Higgins 1983, Higgins and Johns 1984; Underwood 1984), on the same grounds that one could well criticize the use in the classroom of mechanical drills rather than communicative activities—because they cannot be shown to lead to language acquisition, and because they are boring. Davies seems willing to concede that the role of drillmaster may indeed be a boring one: the simulation programs described in another section are referred to as "as welcome change from drill and practice" (Davies 1985, 28). Nonetheless, he claims, dullness does not apparently affect
their usefulness: "Curiously, it does not seem to matter how 'imaginative' these programs are" (Davies 1985:16). As evidence, Davies cites the example of a "rather boring" French verb drill program used at his own university. Not only is it "popular" with the students, but the feeling among both students and staff is that those who have used it show improvement in their ability to handle irregular verb forms. (Such is the state of our science, as we noted above, that the author of a handbook must resort to anecdote in place of empirical evidence to support what would otherwise be an interesting claim.)
2. The correlate to assumption (1) is that CALL programs which would allow for a more open and less rigid (hence less boring) exercise format are somehow "too difficult and too complicated" to be worth attempting. And in order to design an "expert system" to help students learn language, we would need to know much more about how they learn:
The key factor which is inherent in the concept of the expert system is that the computer should be able to react 'intelligently' to every [my italics] response the student makes To a large extent this approach is a reaction against Skinnerian behaviorist teaching methodology, and in its extreme forms it manifests itself in a total rejection of anything which smacks of a drill. The author's personal view is that expert systems and artificial intelligence are worthy of exploration, but at present not enough is known about the way in which human beings learn a language for these ideas to be effectively applied in CALL. (Davies 1985:38).
Those of us who advocate exploring the use of artificial intelligence strategies for CALL would probably agree with at least part of Davies' statement: indirectly, such experimentation is a reaction against behaviorist methodology (with computers or in the classroom), although we would probably not go so far as to say there is no such thing as a good drill.3 On the other hand, we would take issue with the assumption that every student response would have to receive intelligent feedback. This is indeed difficult, if not hopeless. Yet there are useful stages of "semi-intelligence" between this extreme and the opposite, the completely unconscious program. For example, if we assume the program is operating within some limited domain, and that the student is abiding by certain rules outlining lexical or structural boundaries, then it is not hopeless to imagine a program that makes an intelligent response perhaps 90 or 95 percent of the
time—and apologizes with something like "Je ne comprends pas" the rest of the time.
There are such programs. Much of Higgins' work at the British Council has been along those lines, in addition to a certain amount of work here in the U.S. By way of example, I will describe briefly a project I have worked on: JUEGOS COMUNICATIVOS (Underwood and Bassein, 1985), a set of Spanish problem-solving games designed to accompany the college textbook, Puntos de Partido (Knorre, et al., 1985). The concept behind the games is that students will use Spanish sentences to solve the problem before them rather than merely supply grammatical forms on cue. And since they can use language rather freely, the program must be able to parse their input, and determine both its grammaticality and its appropriateness to the environment.
The JUEGOS program consists of two somewhat separate components: a language-independent parsing program, and a set of language-specific data files in which word classes, morphological rules, syntactic patterns, and semantic attributes are specified for the benefit of the parser. (An example of the sort of teamwork Davies refers to, the parser was written by a computer scientist, Richard Bassein, while the data files were developed by the author, a linguist.) The program's ability to parse intelligently depends on an environment limited to that of the mini-world displayed on the screen, plus, as we suggested above, certain rules regarding the range of permissible structures.
An example of one of the activities from JUEGOS is the game "Las cinco diferencias." Two pictures are displayed together on the screen, identical except for five unambiguous differences (door on the house/no door on the house, big car/little car, dog/no dog, etc.). The student is instructed to describe the five things which are different in the picture on the right using the verbs ser, tener, or hay. With each input sentence, the program parses left-to-right to see if it can recognize a pattern. If it can, the parser then checks the elements in that pattern for problems of gender/number agreement, missing words, etc. If problems are detected, the questioned words are highlighted and some stored message is displayed to give the student a hint as to the nature of the problem (¿Masculino o femenino?, etc.). If no syntactic problems are found, the program checks the lexical content of the input sentence to see if it agrees with the semantic attributes for this particular picture. If semantic anomalies are found, messages are again displayed. The following is a sample dialogue about one of the pictures
(numbers are for reference; computer responses in caps):
1. Las maletas son mas grande
2. ¿SINGULAR O PLURAL? [highlights grande]
3. Las maletas son mas grandes
4. TIENES RAZON
5. Hay una ventana
6. ¿COMO . . . ?
7. Hay una ventana en casa
8. FALTA EL ARTICULO [highlights casal
9. Hay una ventana en la Casa
10. SI, PERO NO ES UNA DIFERENCIA
11. Hay dos perros
12. YO NO VEO MAS QUE UNO
13. Hay un perro
14. TIENES RAZON ...
As clever as it may be, the program is not prepared to give an intelligent response to every possible student sentence, even in this limited micro-world. When faced with totally unexpected, yet parsable, nonsense such as "Hay una casa en el hombre" ("There is a house in the man"), the best the program can offer is '¿Como?" which the user takes to mean roughly "That's good Spanish, but I don't get it." Also, the structure in (7) is technically correct (in another, irrelevant, sense), but the program does not know that, either. Yet there is much that the program does know, as should be clear from most of the responses above. My point is that given a limited domain such as this, and with certain structural restrictions, it is possible to design a program which demonstrates a certain intelligence most of the time. And given that the aim of such a program is to provide a means for students to explore how sentences are built in Spanish, rather than submit them to a rigorous test of their grammatical knowledge, then "most of the time" is probably sufficient.
Finally, let us consider Davies' statement that we do not know enough about language learning to pretend to design effective intelligent CALL. In one sense, this is undoubtedly true. While much has been published on student learning variables—attitude, motivation, personality, cognitive skills, learning strategies, and so forth—as yet we know very little about how these variables affect, or are affected by, the use of CALL exercises.4 One could imagine that an intelligent learning system would need to possess at least two kinds of knowledge: (1) knowledge about the language itself and how it is to be used
within the domain of the activity at hand, and (2) acquired or "learned" knowledge about the types of errors the student makes during the activity, so that the program can adapt itself accordingly.
Vernick and Levin (1985) describe work in progress in the design of an intelligent grammar tutoring system for ESL which goes one step farther. They argue that an intelligent CALL system would need three different kinds of knowledge. While (1) and (2) above would enable the system to create an internal model of the learner's past and present performance, the system would also need to know the most appropriate pedagogical strategies for handling the various types of errors that arise. Thus, they argue, the system would be able to decide what kinds of feedback the student needs (suggestions for further practice? review?), how much to provide, and when to provide it. Despite the skepticism of Davies (and others), Vernick and Levin's work is proceeding on the assumption that we do know enough to design an intelligent learning system—albeit complicated indeed.
Although there is, then, much that we do not know about the use of computers for language teaching, there is clearly much that we do, if we will simply reflect upon the accepted principles of our profession. What is the place of the computer as a supplementary aid? In the classroom, certainly (let the whole class try to "beat the system"); in the lab, perhaps—but not isolated in little booths, since one of the advantages of computers is the way they tend to encourage group action and "off-screen dialogue." Do we need statistical evidence in favor of CALL? Probably, but that shouldn't stop us from trying things in the meantime. In fact, precisely because we do not know what works and what does not, we should be wary of clinging to the familiar and the obvious.
A more difficult question concerns the wide gap between our present limited concept of CALL and the growing interest in teaching language for proficiency—the ability to use the language in practical situations. One admittedly limited response is the problem-solving program of the JUEGOS type described above, or the ELIZA-type conversational program (Underwood 1982, Kossuth 1984), an open-ended, heuristic approach to language practice. Such programs may fall short of satisfying Davies' requirement for intelligent feedback for every response the student might make, yet perhaps we need not be so demanding. Our intuitions as language teachers suggest that working with a machine that can simulate language use in an unpredictable way (i.e., leaving
some sort of "information gap") offers a fair approximation of communication. The machine will not always be right, but if it succeeds often enough so as not to be frustrating, something will be taking place that is helpful to the students. Granted, we must remain humble about such "communication"—as I have pointed out elsewhere, "There can be no better 'communicative' learning environment than the warm and responsive presence of other human beings" (Underwood 1984:80).
One nagging doubt I am left with about CALL, one which books like Using Computers never seem to assuage, is that we may run out of ideas, and stall, without ever quite developing an adequate theory to help us decide what we should (or should not) be doing. There we will be, still "sitting in backrooms re-inventing the wheel," long after our students have lost interest and moved on to something else, as they inevitably will. (We need only recall the Great Language Lab Episode of the 1960s and 70s, or—more recently—Atari video games.) Our profession being what it is, we are used to warning of bandwagons. Yet what if we end up a long ways out of town on the CALL bandwagon before we realize it is the wrong bandwagon? Thus, warns Putnam (1982), "when the music stops [we] may find [ourselves] far from home with nothing but broken instruments and badly frayed sheet music."
1 Davies, Graham. (1985). Using Computers in Language Learning: A Teacher's Guide (2nd edition), with a section on the use of the computer in English Language Teaching by John Higgins. London: Centre for Information on Language Teaching and Research, Information Guide No. 22, pp. iv-159.
2 A different kind of flaw is the print quality of the text itself. Produced on a Commodore 4032 with SUPERSCRIPT and printed on a Commodore 8026, the result is a spacey, typewriter-style print that the author found unpleasant and difficult to read.
3 recall Paulston's useful three-way distinction regarding drills: mechanical, meaningful, and communicative. The mechanical drill can be performed without knowing what the words mean, while the meaningful drill cannot. What makes a drill communicative, on the other hand, is the "information gap," the unknown information which the student must supply. In this sense, any uncontextualized CALL drill will be mechanical, while a contextualized exercise requiring a lexical choice or the answer to a question about the context will be meaningful. It is my contention that we should strive for, at the very least, the latter type of exercise; at the same time we should explore ways of making them truly communicative. (See Christina Bratt Paulston, "Structural pattern drills: a classification," Foreign Language Annals 4.2:187-193, 1980).
4 For a preliminary discussion of learning styles and CALL, see Martin K. Phillips, "Approaches to the design of language teaching software," CALICO Journal 3.4:36-48 (June 1986).
Higgins, John. 1983. "Can computers teach?" CALICO Journal 1.2:4-6.
Higgins, John and Tim Johns. 1984 Computers in Language Learning. New York: Addison-Wesley.
Knorre, Marty, et al. 1985. Puntos de Partida (2nd ed.). New York: Random House.
Kossuth, Karen. 1984. Suggestions for comprehension-based computer-assisted instruction in German. Unterrichtspraaxis 17 (Spring):109-115.
Putnam, Constance E. (1983). "Foreign language instructional technology: the state of the art," CALICO Journal 1.1:35-41.
Underwood, John. 1984. Linguistics, Computers and the Language Teacher: A Communicative Approach. Rowley, Massachusetts: Newbury House.
Underwood, John, and Richard Bassein. 1985. JUEGOS COMMUNICATIVOS: Spanish Games for Communicative Practice for the Apple IIe, IIc, II+, based on Knorre, et al., Puntos de partida (2nd ed.). New York: Random House.
Vernick, Judy and Lori Levin. 1986. "Intelligent grammar tutoring in ESL." Paper presented at the 1986 TESOL convention, Anaheim, California.
John Underwood (Ph.D., UCLA, 1981), currently Assistant Professor of Hispanic Studies and Linguistics at Mills College, Oakland (California), is well-known as both an innovator and a critic in the field of computer-assisted language learning. His CALL projects include JUEGOS COMMUNICATIVOS, problem-solving games; CORREO, a Spanish electronic mail system for communication and composition; and Alexis, a multi-activity system for French, Spanish, and German at Mills. Among his publications, the best-known is Linguistics, Computers, and the Language Teacher: A Communicative Approach (Newbury House, 1984), winner of the 1985 Mildenberger Medal of the Modern Language Association.
Department of Foreign Languages
Oakland, CA 94613