Langugae Teaching Methodology
A Text for Teacher
Chapter Four (Published with Permission).
Reading: A Discourse Perspective
This chapter takes up some of the central themes which have been developing in the two preceding chapters and applies them to the development of reading in a second or foreign language. Some of the key theoretical and empirical perspectives on the nature of reading and learning to read are examined in the first section of the chapter. We then look at ways in which this theoretical and empirical work has found its way into the classroom through teaching materials and techniques. The key questions addressed in this chapter are:
Bottom-up and top-down views on reading
In Chapter 21 outlined two different views on the nature of listening: the bottom-up view, which suggests that successful listening is a matter of decoding the individual sounds we hear to derive the meaning of words and thence utterances; and the top-down view, which suggests that we use discoursal and real-world knowledge to construct and interpret aural messages. These two competing models of language processing have also had a central place in the debate on the nature of reading comprehension.
Until comparatively recently, the bottom-up approach dominated both first and second language research and theory. According to Cambourne (1979), it was the basis of the vast majority of reading schemes. Many people will recall with distaste the basic primers with their highly improbable stories which were used to develop early literacy skills. Although there is now a great deal of evidence which points to the inadequacy of the approach, it still has many adherents within the teaching profession. In addition, the ‘back-to-basics’ movement and introduction of ‘basic’ skills testing in schools indicates that the approach has a good deal of support beyond the profession itself.
The central notion behind the bottom-up approach is that reading is basically a matter of decoding a series of written symbols into their aural equivalents. Cambourne, who uses the term ‘outside-in’ rather than bottom-up, provides the following illustration of how the process is supposed to work:
According to this model, the reader processes each letter as it is encountered. These letters, or graphemes are matched with the phonemes of the language, which it is assumed the reader already knows. These phonemes, the minimal units of meaning in the sound system of the language, are blended together to form words (Phonemes are the individual units of sound in a language.) The derivation of meaning is thus the end process in which the language is translated from one form of symbolic representation to another.
One reason for the survival of this approach in the face or criticism is that it seems a reasonable and logical explanation of what happens when we read. Letters do represent sounds, and despite the fact that in English twenty-six written symbols have to represent over forty aural symbols, there is a degree of consistency. On the surface, then, it seems more logical to teach initial readers to utilize the systematic correspondences between written and spoken symbols than to teach them to recognize every letter and word they encounter by memorizing its unique configuration and shape (a practice favoured by proponents of the ‘whole-word’ approach.)
One of the assumptions underlying this phonics approach is that once a reader has blended the sounds together to form a word, that word will then be recognized. In other words, it is assumed that the reader possesses an oral vocabulary which is extensive enough to allow decoding to proceed. This assumption is not one that can be made with second language learners for whom any form of reading instruction ought to be linked with intensive aural vocabulary development. In fact, the assumption that phonic analysis skills are all that is needed to become a successful independent reader is questionable with first as well as second language readers. Most teachers of reading have encountered children who are able to decode print after a fashion and thereby ‘read’ without actually extracting meaning from the text.
A number of other major criticisms have been made of the phonics approach. Much of this criticism is based on research into human memory. In the first place, with only twenty-six letters to represent over forty sounds in English, spelling-to-sound correspondences are both complex and unpredictable. It was this realization which led to the development of primers, in which stories were composed of words which did have regular sound/symbol correspondences. Unfortunately, as many of the most common English words have irregular spellings and were therefore excluded, the stories in primers tended to be unnatural and tedious.
Research into human memory also provides counterfactual evidence. It has been shown that the serial processing of every letter in a text would slow reading up to the shown that the serial processing of every letter in a text would slow reading up to the point where it would be very difficult for meaning to be retained. An early study by Kolers and Katzmann (1966), for example, demonstrated that it takes from a quarter to a third of a second to recognize and assign the appropriate phonemic sound to a given grapheme. At this rate, given the average length of English words, readers would only be able to process around 60 words per minute. In fact, it has been demonstrated that the average reader can read and comprehend around 250-350 words per minute. Given the fact that we can only hold in working memory about seven items at a time, readers would, under the bottom-up model, very often forget the beginning of a sentence (and perhaps even a word) before they have reached the end.
Smith (1978) has pointed out that the serial processing operations underlying the phonics approach are also contradicted by the fact that it is often impossible to make decisions about how upcoming letters and words ought to sound until the context provided by the rank above the one containing the item has been understood. Thus, in order to assign a phonemic value to a grapheme it is often necessary to know the meaning of the word containing the grapheme. If, for example, in reading a text one came across the graphemic sequence ‘ho--’, it would be impossible to assign a value to ‘o’ until one knew whether the whole word were ‘house’, ‘horse’, ‘hot’, ‘hoot’, etc. In the same way, it is impossible to assign a phonemic value to the vowel sequence in the word ‘read’ until it is known whether the sentence containing the word is referring to the present or the past. Additional evidence against the notion that reading proceeds through the serial processing of ever larger units of language has come from a line of research initiated by Goodman and Burke (1972). This research involved the analysis of errors made by the reader when reading aloud. Errors, termed ‘miscues’ by Goodman and Burke, provide evidence that something more than mechanical decoding is going on when readers process texts. The found that in many instances deviations from what was actually written on the page made sense semantically – for example, a child might read the sentence ‘My father speaks Spanish’ as ‘My Dad speaks Spanish’. ( If the child read ‘My feather speaks Spanish’, this would be evidence that he/she is not reading for meaning, but is decoding mechanically.)
Insights from sources such as the Reading Miscue Inventory led to the postulation of an alternative to the bottom-up, phonics approach. This has become known as the top-down or psycholinguistic approach to reading. As with the bottom-up models, there are a number of variations in this approach, but basically all agree that the reader rather than the text is at the heart of the reading process. Cambourne (1979) provides the following schematization of the approach.
From the diagram, it can be seen that this approach emphasizes the reconstruction of meaning rather than the decoding of form. The interaction of the reader and the text is central to the process, and readers bring to this interaction their knowledge of the subject at hand, knowledge of and expectations about how language works, motivation, interest and attitudes towards the content of the text. Rather than decoding each symbol, or even every word, the reader forms hypotheses about text elements and then ‘samples’ the text to determine whether or not the hypotheses are correct.
Oller (1979) also stresses the importance of taking into consideration psychological as well as linguistic factors in accounting how people read. He points out that the link between our knowledge of linguistic forms and our knowledge of the world is very close, and hat this has a number of implications for discourse processing. Firstly, it suggests that the more predictable a sequence of linguistic elements, the more readily a text will be processes. It has been suggested that texts for initial first language readers should come close to the oral language of the reader. (The fact that phonics primers written in phonically regularized language present relatively unpredictable language at the levels of clause and text probably helps explain the difficulty many readers have with them.) Even second language learners, despite their limited knowledge of linguistic forms, should be assisted by texts which consist of more natural sequences of elements at the levels of word, clause and text. A second way of exploiting the relationship between linguistic and extra linguistic worlds is to ensure not only that linguistic elements are more predictable but also that the experiential content is more familiar and therefore more predictable. (See also Alderson and Urquhart’s (1984) ‘text-reader-task’ structure for characterizing foreign language reading.)
The significance of content familiarity for compensating limited linguistic knowledge is explored more fully in the next section. Here we should note that just as linguistic sequencing above the level of the grapheme/phoneme was unfamiliar in phonics-based primers, so also was the content, the subject matter often being remote from the experiences and interests of the readers (For an extended discussion on linguistic and extra linguistic relationships, see Smith 1978, and Oller 1979.)
One of the shortcomings of the top-down model is that is sometimes fails to distinguish adequately between beginning readers and fluent readers. Smith, for example, advances the view that fluent readers operate by recognizing words on sight. In other words, fluent reading in non-ideographic languages such as English proceeds in the same way as fluent reading in ideographic languages such as Chinese, where readers must learn to identify characters by their shape. (in ideographic languages, the characters do not represent sounds. Rather they are derived from the objects and entities they are supposed to represent.) Of course, it does not necessarily follow that, as fluent readers proceed through sight recognition (assuming that this is how they read – something upon which there is by no means universal agreement), this is the way initial readers should be taught.
Stanovich (1980), in his exhaustive review of reading models, criticizes the top-down notion that reading proceeds through the generation of hypotheses about up-coming text elements. He points out, for example, that the generation of hypotheses in the manner suggested by Smith would actually be more time consuming than decoding would be. In the light of the perceived deficiencies of both bottom-up and top-down models, he proposes a third which he calls an interactive-compensatory model. As the name indicates, this model suggests, that readers process texts by utilizing information provided simultaneously from several different sources, and that they can compensate for deficiencies at one level by drawing on knowledge at other (either higher or lower) levels. These sources include all those looked at separately in bottom-up and top-down processes, that is, phonological, lexical, syntactic, semantic and discoursal knowledge.
Stanovich claims that his alternative model is superior because it deals with the shortcomings inherent in other models. The major deficiency of the bottom-up model is that it assumes that initiation of higher level processes, such as use of background knowledge, must await lower level decoding processes. The top-down model, on the other hand, does not allow lower level processes to direct higher level ones. The interactive-compensatory model allows for deficiencies at one level to be compensated for at another. In particular, higher level processes can compensate fro deficiencies at lower levels, and this allows for the possibility that readers with poor reading skills at the levels, and this allows for the possibility that readers with poor reading skills at the levels of grapheme and word can compensate for these by using other sources of knowledge such as the syntactic class of a given word or semantic knowledge. Given deficiencies of lower level skills, poor readers may actually be more dependent on higher level processes than good readers.
Schema theory and reading
With the insight that there is more to comprehension that the words on the page has come an attempt to provide a theoretical model which will explain the way that our background knowledge guides comprehension processes. We have looked at aspects of this theory in Chapter 2 and 3, and here we take the ideas a little further. Much of the model-building has been carried out by researchers in the field of artificial intelligence whose ultimate aim has been to enable computers to process and produce discourse. Terms chosen by cognitive psychologists and computer specialist include ‘frames’, ‘scripts’, ‘scenarios’ and ‘schemata’. A useful introduction and overview of this work is provided by Brown and Yule (1983a).
One of the most widely reported theories in the cognitive psychology literature is Minsky’s ‘frame’ theory. Minsky suggests that human memory consists of sets of stereotypical situations (frames) which guide comprehension by providing a frame-work for making sense of new experiences. (Pearson and Johnson’s 1978 view of reading comprehension as a process of relating the new to the known is based on a similar notion.)
While this theory appeals to our commonsense notions about how comprehension might work, it does have a number of problems. For example, it provides no explanation of why one script rather than another might be selected to guide comprehension. Brown and Yule see this as a major shortcoming of the theory, and they detail their major objection by inviting the reader to decide which ‘frame’ should guide comprehension for a reader processing a newspaper article:
Consider the following new situation which presented itself at the beginning of a newspaper article.
The Cathedral congregation had watched on television monitors as Pope and Archbishop met in front of a British Caledonian helicopter, on the dewy grass of a Canterbury recreation ground. (The Sunday Time, 30 May 1982)
The problem should be immediately obvious. Is a ‘Cathedral’ frame selected? How about a ‘television-watching’ frame, a ‘meeting’ frame, a ‘helicopter’ frame, a ‘recreation-ground’ frame? These questions are not trivial. After all, it probably is necessary to activate something like a ‘recreation-ground’ frame in order to account for the definite description ‘the grass’ mentioned in the text. (Brown and Yule 1983a: 241)
The term which has gained greater currency in the literature than any other is ‘schema’, a term first used by Bartlett (1932). Like frame theory, schema theory suggests that the knowledge we carry around in our head is organized into interrelated patters. These are constructed from our previous experience of the experiential world and guide us as we make sense of new experiences. They also enable us to make predictions about what we might expect to experience in a give context. Given the fact that discourse comprehension is a process of utilizing linguistic cues and background knowledge to reconstruct meaning, these schemata are extremely important, particularly to second and foreign language learners.
Widdowson (1983) has reinterpreted schema theory from an applied linguistic perspective. He postulates two levels of language: a systemic level and a schematic level. The systemic level includes the phonological, morphological and syntactic elements of the language, while the schematic level relates to our background knowledge. In Widdowson’s scheme of things, this background knowledge exercises an executive function over the systemic level of language. In comprehending a given piece of language, we use what sociologist call interpretive procedures for achieving a match between our schematic knowledge and the language which is encoded systemically. Cicourel, who was one of the first to utilize the notion of interpretation in the behavioural sciences, suggested that:
Through the use of interpretative procedures the participants supply meaning and impute underlying patterns even though the surface content will not reveal the meanings to an observer unless his model is directed to such elaborations. (Cicourel 1973:40)
Widdowson has related this notion to the work on schema theory. He argues that there are two types of schema operating through language. The first of these is concerned with propositional meaning, while the second relates to the functional level of language, either of which can provide textual connectivity. Thus, in Chapter 2, we saw that the connectivity of text cannot always be explained exclusively in terms of the language in the text. It also depends on our interpretive ability to make connections which do not exists in the text, but which are provided by use from our schematic knowledge of the subject in hand, or the functional purposed which the different text elements are fulfilling.
An important collection of articles on schema theory and reading is Carrell et al. (1988). The collection contains conceptual, position papers which provide a theoretical framework, and also a number of empirical studies showing how schema are implicated in reading comprehension.
Research into reading in a second language
A number of studies have been conducted into the influence of schematic knowledge on the comprehension processes of second language readers. Aslanian (1985) set out to discover what interpretive processes went on in her learners’ heads as they completed a multiple choice/gap test of a reading passage. After the subjects had completed the gap test, Aslanian took them individually through the test and asked them to tell her why they had chosen one item rather than another. Transcripts of these interviews indicate that correct responses were often provided for the wrong reasons, and that incorrect responses were sometimes provided by readers who, in fact, had an adequate grasp of the passage as a whole. Aslanian’s study shows that schematic knowledge structures can either facilitates or inhibit comprehension according to whether they are over- or underutilized.
If readers rely too heavily on their knowledge and ignore the limitation imposed by the text, or vice versa, then they will not be able to comprehend the intended meaning of the writer. Whether one has understood the text or not depends very much on text variables such as sentence structure and length, vocabulary intensity, number of new concepts introduced, the difficulty and novelty of the subject matter, etc… To understand the reader and the nature of the act of reading more clearly and comprehensively, one needs also to find out and describe the reader’s strategies and reactions with regard to the reading tasks, and to see how the reader copes with the reading tasks and solves the problems. (Aslanian 1985: 20)
In a rather different study Nunan (1985) set out to test whether the perception of textual relationships is affected by readers’ background knowledge. Subjects for the study were 100 second language, high school students who were divided into two groups. Group A consisted of longer-term learners who were orally fluent but had reading problems. Group B consisted of recently arrived immigrants. Test materials consisted of passages on familiar and unfamiliar topics from high school texts. Readability analyses showed the unfamiliar passage to be easier than the familiar passage. Both passages were analyzed for the cohesively marked logical, referential and lexical relationships they contained (Halliday and Hasan 1976). Ninety-six relation-ships were identified which were matched across both passages, and these were used as the basis for constructing the test items. To construct the test, a modified cloze procedure was used in which the second end of the cohesive tie was deleted. The task for the reader was to identify the relationship and reinstate the deleted item.
Examples of the cohesive relationships and the way they were used to construct test items are set out below:
Test item: Usually there would be no difficulty in deciding whether a living thing is a plant or an animal and it can be classified immediately. There are __________ some very tiny creatures which scientists know to be living, but cannot be sure whether they are plants or animals. (Deleted item: ‘however’. Cohesive type: adversative conjunction).
There is no difficulty in deciding that a bird is living and a stone is non-living, but not all things are as easy to distinguish _______. (Deleted item: ‘these’. Cohesive type: referential demonstrative).
Test ITEM: Green plants grow towards the light. This is because plants need ______ for energy. (Deleted item: ‘light’. Cohesive type: lexical reiteration).
For both groups of subjects, the perception of textual relationships was significantly easier in the familiar, though syntactically more difficult, passage. Group A was also superior to group B, indicating that length of formal and informal exposure to the target language is also a significant factor in success.
As already indicated, this study was designed to test the effect of schematic knowledge on the comprehension of various types of textual relationship. Schema theory suggests that reading involves more than utilizing linguistic and decoding skills; that interest, motivation and background knowledge will determine, at least in part, the success that a reader will have with a given text. This study showed that background knowledge was a more significant factor than grammatical complexity in determining the subjects’ comprehension of the textual relationships in question.
This study has a number of pedagogical implications. The first of these stems from the fact that reading skills are not invariant: that is, they do not depend solely on a knowledge of the linguistic elements that make up a text. Reading is a dynamic process in which the text elements interact with other factors outside the text; in this case most particularly with the reader’s knowledge of the experiential content of the text. This suggests that there is a need to relate the language being taught to the context which carries it. In instructional systems where the target language is the medium of instruction, the teaching of language ought not to be divorced from other school subjects. ‘Language across the curriculum’ is as important, if not more important, for second language learners as for first language readers (for suggestions on teaching language and content see Mohan 1986; Brinton et al. 1989). Widdowson goes so far as to suggest that foreign as well as second languages might be taught through the other subjects on the curriculum:
A foreign language can be associated with those areas of use which are represented by the other subjects on the school curriculum and … this not only helps to ensure the link with reality and the pupils’ own experience but also provides us with the most certain means we have of teaching language as communication, as use rather than usage. (Widdowson 1978: 16)
Another study into the perception of textual relationships in a cross-cultural context is reported by Steffensen (1981). Steffensen identified relationships signaled by conjunctions in culturally significant sentences in two texts. One of these described an American wedding and the other an Indian wedding. A sentence was considered to have cultural significance if the relationship it contained could not be predicted from everyday knowledge, but which required familiarity with the culture from which it was drawn. Sample sentences from Steffensen’s test passages are as follows:
(i) Actually it was surprising that the men were in such good shape because they had a stag party on Thursday and didn’t get in until 3 am
(ii) The ushers seated some of the bride’s friends on his side of the church so things wouldn’t look off balance.
(iii) They did not create any problem in the wedding, even though Preema’s husband is their only son.
(iv) Her husband and in-laws picked ‘Uma’ for her new name since her husband’s family calls him ‘Shiva’.
Steffensen had American and Indian subjects read the passages and then recall as much of the content as they could. The recall protocols were then analyzed to determine whether the relationships being investigated were recalled by the subjects. This analysis revealed that Americans did better on the text containing American cultural content and Indians did better on the text containing Indian cultural content.
Steffensen concluded from her study that when readers are exposed to texts which describe aspects of a culture foreign to the reader, there will be a breakdown in the perception of textual relationships. A breakdown in relationships at the linguistic level reflects a breakdown in comprehension at the experiential level, that is, at the level of content. Her findings therefore support the contention that the process of reconstructing meaning is one of mapping the linguistic content onto extra linguistic context (see also Oller 1979). In setting out the pedagogical implications of her study, Steffensen suggests that what, at first sight, is a linguistic problem, may in fact be a problem of background knowledge. In such a case, teaching learners the facts about the customs in question would probably be more effective than drilling them in aspects of the language.
Considerable research has also been conducted into the strategies employed by good readers. This research has been selectively used to justify various proposal for pedagogical action . Walter (1982), in her book on learning to read in a second or foreign language, says that good readers utilize the following strategy when encountering a difficult text. First of all, they read the text slowly, pausing to consider what they have read. They then reread the text, looking from one part of the text to other parts in order to make connections between these different parts, and to make a mental summary of what they have read. She claims that most of the people who read in this way remember both the general points and the details of what they have read better than those who use other strategies. Her book provides numerous exercises deigned for learners to search through a text and mentally organize the information it contains.
Reading and social context
Reading is usually conceived of as a solitary activity in which the reader interacts with the text in isolation. While not wishing to pre-empt the discussion in the section which follows, it is worth pointing out that the two language lessons we look at (classroom extracts 4.1 and 4.2) show that reading lessons are generally anything but solitary.
The social context of the second language reader is taken up by Wallace (1988), who explores in depth the circumstances in which such readers acquire and maintain literacy. She points out that learning to read is different from learning to speak, in that there is often a much stronger motivation to communicate orally than there is to communicate through reading and writing. If the use of literacy skills is a normal and accepted part of the behaviour of those with whom learners come into contact, then there is a much greater likelihood that the learners are going to want to read. In other words, learners are socialized into reading, and the motivation for learning to read is not only (or even primarily) for enjoyment or information, but because the aspiring reader wants to gain access to a ‘community’ or readers.
In cultural terms, reading and learning to read will mean different things to different learners. A young Polish doctor who is literate in her first language will have different expectations and views on the nature of literacy from an elderly Hmong woman with no formal schooling. This observation has obvious implications for the classroom; we must take these differences into account, not only in the texts we select, but also in the ways in which we go about teaching literacy.
The cultural, social and political implications of literacy are highlighted by Wallace in the following manner:
Achieving literacy is seen not only as concomitant with certain social and occupational roles but in some countries goes together with political rights, such as the right to vote. Literacy is valued as part of adulthood, of full citizenship. It is clearly partly for this reason that we do not talk of children – or at least younger children – as being illiterate. Illiteracy is a stigma reserved for adulthood. Nonetheless, as teachers of reading we need from the beginning to see functional literacy as the goal; we need, that is, to show our learners that being literate, for children as well as adults, is part of day-to-day life in a personal and social sense. Children need from the beginning to see that reading is purposeful, that it helps us to achieve things. (Wallace 1988: 3)
Types of reading text
In Chapter 3 we looked at genre theory and its implications for language use. In this section we apply some of the insights from both genre theory and schema theory to reading. You will recall that one of the claims of genre theory is that language exists to fulfill certain functions and that these functions will largely determine the structure of the text and the language it contains. Schema theory suggests that we need to utilize information not explicitly contained in the text (i.e. ‘inside the head’ knowledge) to comprehend more texts adequately. As you read the following texts, consider the interpretive procedures required to understand them. What additional information does the reader have to bring to the texts in order to understand them?
System 816 feature incredible storage capacity, the flexibility to handle a large variety of applications, and the speed to get the job done fast. And you can choose from more the 3,000 CP/M 8 or 16 bit programs as your needs grow. With the ability to expand from single to multi-user, network up to 255 systems, and upgrade as the technology advances, this equipment stands the test of time.
Lemon Cream Sponge
500 ml cream 1 sachet gelatine
500 ml milk 3 tbs sugar
juice of two lemons 1 punnet strawberries
Melt gelatine in small quantity of hot water. Blend milk, cream and lemon juice together. Add gelatine mixture, stirring constantly. Chill several hours until set. Decorate with strawberries.
If you want to do the job right, the saying goes, make sure you use the right tool. Don’t try to stop a tank with a Roach Motel. So more than in other fields, scientists are debating this questions of appropriate technology as microcomputers begin to move into research facilities and R and D departments around the country.
New apple-fresh Cuddly Fabric Softener. A fresh new experience! New Cuddly leaves your clothes so soft and easy to iron; and now new Cuddly adds the crisp fragrance of freshly picked apples to your wash. Only new Cuddly Fabric Softener has two softening agents working together to soften your whole wash; and new Cuddly also helps eliminate static cling. Use apple-fresh Cuddly in your final rinse. Do not pour undiluted onto fabrics.
For anyone with a passing knowledge of the field, text 1 is fairly obviously about computers. However, in order to get from the text all that the author intended, one would need to utilize a certain amount of background knowledge. It is unlikely that someone with only a passing knowledge of computers will be able to say with any certainty whether the text is about micro, mini or maxi computers.
Text 2 is quite evidently a recipe, and it displays the generic structure and language of recipes. In structural terms, it consists of a title, specifications, ingredients and procedure. Linguistically, it contains language which is typical of the ‘recipe genre’. Predicable function words such as articles and prepositions tends to be omitted, and the procedure usually consists of a set of imperative statements listed in chronological order.
Our knowledge of text genres can, in fact, sometimes mislead us. Text 3 is a case in point. This text is extracted from an article on the use of appropriate technology which was published in a popular journal. However, some people have said that it is about advertising, having been misled by the second sentence, which is couched in the language of advertisements but which is, in fact, a metaphor. (A ‘Roach Motel’, by the way, is an American insect trap. The insects check in, but never check out!)