Reading and risk-taking: a role for the computer

John Higgins

This article discusses the skills of prediction and hypothesis-formation which are used by the expert reader, and the techniques which may be used to develop these skills among EFL learners. It goes on to describe a computer exercise which invites the learner to identify the source of a piece of text from a randomly chosen fragment. initially the fragment is only one word long, but the student can open a ‘window’ on it to take in more of the context, until he or she is ready to make a guess. The article goes an to describe the scoring system, the relationship between the language level of the learner and the selection of texts on which the computer draws, and possible developments and extensions to the program.

Reading, unlike listening, is not carried out in real time. While a listener must listen to and decode speech at roughly the same pace, and following the linguistic order in which the speaker produces it, no such limitations affect the reader, who is completely at liberty to glance ahead or back, or to make use of any graphic clues within the field of vision. The good reader does this quite unthinkingly, without being aware of any departure from the left-to-right order which is the (Roman alphabet) analogue of the one-way passage of time.

Readers may “swim” through a text in a linear way, but the technique of good readers is more akin to crossing a river by leaping from floating log to floating log, not always choosing to leap directly forward, but always conscious of the direction of the bank ahead. In the process, they can develop reading speeds well in excess of any normal speaking speed, just as the runner can outstrip the swimmer. They may on occasion “fall in” and get wet, i.e. lose the sense and have to re-read a sentence or two to regain it.

The technique which I have described as jumping from log to log is labeled by Kenneth Goodman sampling, predicting, testing, and confirming (Goodman 1967, 1971). Goodman points out that good readers accept a certain amount of disconfirmation; in other words, they realize that some of the guesses they make will be wrong, but are confident that they can backtrack in order to make a better guess if necessary. They will normally make it to the other side, i.e. extract meaning from the whole text, whereas the less accomplished reader who starts out with the linguistic equivalent of a dog-paddle is the one most in danger of drowning, of failing to make any sense of the text.

Barriers to effective reading

Developing better reading skills is, as Bright and McGregor (1970: 97) pointed out, often a matter of convincing a reader that more adventurous reading is also likely to be more effective reading. Carrying conviction may involve overcoming emotional as much as intellectual barriers: the commonsense appeal of a linear reading strategy—one that says, I must understand word one before I can understand word two—is powerful. Moreover, a background of learning experience in which accuracy and detail have been given high value, in which 90 per cent has been better than 70 per cent, will have trained learners not to settle for fuzzy understanding of an item instead of the precise wording given in the glossary.

Techniques for effective reading

For these reasons, persuading readers to take risks may involve rather stark procedures. To persuade learners to read faster, it may be necessary to flash texts up for impossibly short lengths of time, so as to convince them that some information still gets through, and that the information derived from that first rapid glance will be of great help if they return to the text in order to fill in detail. To persuade them to rely more on prediction (or guesswork), it may be necessary to invite them to speculate, in their first language if necessary, about tiny amounts of data. As an example of the kind of technique which can be used, a teacher can present the class with a list of apparently unconnected words and ask them which ones they would expect to see in a text on a given topic, say Scotch Whisky. The group will begin by rejecting all the words except those with an obvious lexical association with whisky (drink, alcohol, distil, for instance), but can be led to see possible associations with other words (stream, smoke, barley, nose), until finally they realize that the words in the list all occur in a coherent paragraph on the topic.

Another technique might be to give the learners a title and to ask them to predict as much of the content of the text as they can. Or one can give them a tiny fragment of text, three or four words, and ask them to infer the topic. It is these last two procedures which I have combined in the form of a computer exercise I call Close-Up (now renamed Pinpoint).

The name Close-Up derives from a style of television quiz game in which participants are shown an enlarged detail from a picture of an everyday object, possibly photographed from an unusual angle. The camera slowly zooms out, showing more of the picture, and the participants must call out, or press their buzzers, as soon as they are ready to guess what the object might be. The element of competition forces the participants to develop a strategy to balance wild guessing against over-caution, to keep in mind not only the question ‘Is this the right answer?’, but also another equally important question, ‘How sure am I that this is the right answer?’ (A related technique has been developed by Andrew Wright for the EFL classroom: the teacher projects a badly out-of-focus slide, and the class guess what the picture shows as the slide is very gradually brought back into focus.)

The program

In place of the photograph, Close-Up uses a text from which one word is randomly selected and displayed. The learner’s task is to identify the topic of the text by matching it to one of eight titles displayed on the screen. The initial screen looks like the illustration below.

One word is usually, though not always, insufficient, so the learner can press the M key (paying a 10-point penalty) to enlarge the window on the text to disclose the word before and the word after, and this can be repeated as often as desired, until the whole text is on display. After two such calls for help and one wrong guess (costing a 50-point penalty), the screen will look like this.

Two further calls for more words will produce significant new evidence from the punctuation, showing that this is an extract from dialogue.

At any point, the learner can make a guess by typing the number of the title chosen. If the choice is wrong, the word NO appears beside that title, and the process continues as before. If it is right, the word YES appears alongside the title, and the complete text is printed out below. The screen will now look like the illustration below. I have found that learners read the passage with close attention at this point, presumably to identify the fragment they have worked on and to see where it fits in to the whole.

From this point on, the program can proceed in one of two ways. In the more elementary version, another text from the same set of eight is chosen randomly and one word is selected from it. The same text will not occur twice consecutively, but may recur during the game. Thus a student who does badly at the beginning is almost bound to improve, simply because the texts become familiar and small extracts are recognized.

In the more advanced version, each text which has been correctly identified is replaced by another. The new title over-writes the current one in the list of titles and has the same chance of being selected as any of the other seven. In a session of play the student may encounter quite a number of different texts out of the hundred or so which would be available if the whole of the machine’s internal memory were used. (Naturally the hundred limit no longer applies.)

The main function of the scoring system is to give a structure to the session, to provide an attainable goal to be worked towards. The actual score values in this version of the program are more or less arbitrary and still experimental. Other values and procedures are being tried out.

Any scoring system, even in an examination, sets up a challenge to the student to work out the optimum strategy for winning. The reward for success will be the same whether achieved by luck or knowledge: that, after all, is what happens in life. It must be balanced by penalties both for failure (wild guessing) and for caution (asking for more help than is needed). In Close-Up the student begins with a stock of 200 points, and each extra bit of context asked for is bought at a cost of ten points. The penalty for a wrong guess is one third of the stock of points, but the stock is never allowed to drop below forty. The reward for a successful guess is double the stock of points, up to a maximum of 200. At the beginning of the game, the cost of help is relatively high and the penalty for mis-guessing relatively cheap. As the game proceeds, these relationships reverse. With 900 points in the bank, one can afford quite a number of ten-point calls for more words, but a single wrong guess now costs 300 points. In another version of the program, points won on one text are not put at risk by a wrong guess with the next. It remains to be seen which type of scoring leads to more effective use of the program.

The gambling principle

This system of scoring has some resemblance to those used in reading programs by Tim Johns (Higgins and Johns, 1983), although Johns often makes more overt use of a gambling format. Gambling is a realistic metaphor of many decisions and procedures in life (not necessarily having financial implications) including language use. Training somebody in a harmless environment to assess risk can be of value. In Close-Up the principle is limited to demonstrating that help must be paid for and that it also diminishes potential rewards. The student must work out the strategy which secures the highest reward, balancing the cost of the help against the risk of doing without it. But gambling is a topic which has strong cultural and emotional overtones, and I have encountered criticism for appearing to encourage it.

Target setting

Before the game begins, the student is asked to set a target, which must be in the range 500 to 3,000 points. The main consequence of the choice will be to fix the length of the session. A good student will need only three or four texts to reach a score of over 500, and will soon learn to set a more challenging target. Weaker students will find that 500 points occupy them for fifteen minutes or more and let them see eight or ten texts, which is probably about the right length of session at first. The principle of target setting is part of a more general principle, that of consulting the student on decisions which affect what he or she is doing. Such consultation is a general advantage that the computer has over mass teaching; it is a pity that it is not used more than it is.

The quit option

It is good practice in computer-assisted learning to provide an accessible means of changing or abandoning an activity which is no longer being enjoyed, and this has been done here by allowing the student to press Q at any time. This gives a choice of terminating the program, or else of making a fresh start with a new group of texts.

Student level

So far I have not commented on the level of student that the program is aimed at. The level depends largely on the selection of text, and with simplified texts the program could be used at any level at which one would contemplate introducing a simplified reader for extensive reading. The texts themselves are limited by the screen format to a maximum length of about sixty words, which is of course much shorter than is required for most types of reading practice but is long enough to show some features of coherence. The form of the program which I have been using contains unsimplified texts (largely pastiche in order to avoid copyright problems) of enormous diversity. This, I find, suits the early and mid-intermediate level learner. The activity is a challenge, but an attainable one. The students at this level (judging only from a sample of three who have tried it extensively) seem very willing to spend between forty-five and seventy-five minutes per session with the program, have returned for further sessions, and are sometimes reluctant to give up their place to another student. One encouraging observation I have made is the way in which the program encourages the students to use dictionaries and other reference books, not as props during the reading but as ways of following up what has intrigued them, after the exercise is over.

Higher levels

For upper-intermediate or advanced students, the program in its present form is ridiculously easy, just as it is for native speakers. It can be turned into a far more challenging activity by selecting texts which are not on a variety of topics but which are, for instance, successive extracts from a single long text, divided into short chunks, each given a headline. Now the lexis on its own provides much less help, and the student’s task is to identify features which will place each fragment in the structure of a larger piece of discourse. To persuade academic colleagues of this, I prepared aversion in which the eight texts were extracted from a two-page summary of Chomsky’s work, and they agreed that this was tough going.

Flexibility

The power of a program of this type derives partly from its potential flexibility. Since any text which fits the length limitation can be entered, one could create versions suitable for ESP students in a variety of disciplines, as well as at different levels. The first extension of the program that I plan is to create a text loader, i.e. a simple program which will allow teachers to create new files of texts chosen to suit their own students.

Unpredictability

In this type of program the computer is carrying out “unintelligent analysis”, simply identifying word boundaries and matching text to title. It has no understanding of the semantic content. Its selection of word or text is quite random, which means that it will generate easy and difficult tasks in an unpredictable way, picking a revealing content word on one occasion e.g. nuclear, or a word in the middle of a long sequence of structural words on another:

I have been criticized for this unpredictability, and asked to fix the program so that it either always or never produces items which are very easy or very difficult. This I decline to do. It seems right to have some unpredictability in an activity of this kind.

Why use a computer?

A paper-and-pencil equivalent of this activity could easily be used in a class. The teacher could have a number of texts on the desk and could write the titles or headlines on the board or overhead projector. He or she could now write or say one word, invite the class to speculate on which text it could be from, and then, as Close-Up does, could widen the window by reading out more of the text. Indeed, in an institution which can afford an overhead transparency maker and the rather expensive materials it requires, this could be done with direct reproduction of authentic text on the OHP, so that the window on the text is more literally a window, and all the relevant graphic information is present.

This technique, however, would miss the element of commitment which arises when an individual or small group works with a computer. It would be difficult to reproduce the need to evolve a winning strategy which comes from making guesses and seeing at once how they are scored. It would be difficult, too, to spare the time from other classroom commitments to fit in enough practice. Early experiences suggest that this is an activity which students enjoy so much that they will make their own time for it if a machine is available to them in free periods or in the evening.

References

Bright, j. and McGregor. G. 1970. Teaching English as a Second Language. London: Longman.

Goodman, Kenneth. 1967. “Reading: a psycholinguistic guessing game”. Journal of the Reading Specialist, Vol. 6.

Goodman, Kenneth. 1971. “Psycholinguistic universals in the reading process” in Pimsleur and Quinn (eds) The Psychology of Second Language Learning.Cambridge: Cambridge University Press.

Higgins, J. and T. F. Johns. 1984. Computers in Language Learning. London: Collins