Bottom-up models of reading are considered to be data or ‘text driven’ (Manzo and Manzo, 1995) with readers grasping the meaning from the text itself, how the text is organised and from each word in the sentence. Gough (1972) argued that the good reader focuses on every single letter in the text, and reads word-by-word in order to understand the text. For instance, if the reader wants to read the extract below from A Scots Quair, s/he should theoretically read each word in the sentence to comprehend the context, no matter their linguistic level but, as I will demonstrate, this method does not result in meaning. For Farrell (2009), further, reading is based on extracting propositions:
When we read, one thing we do is extract the propositions from the text. How?
By breaking sentences into their constituent parts and constructing the propositions from there. Comprehension then depends on the proposition we have extracted, which serve as the basis of what we understand and recall.
(Farrell, 2009, p.18)
From Farrell’s viewpoint, reading and understanding any text relies on the reader’s
knowledge of words and grammatical patterns in the sentence, then making inferences and the propositions.
Following the sort of reading suggested in Gough's model, I read each word in the extract and broke sentences into their constituent parts, but I still had difficulties in understanding the text because the grammatical pattern of the sentences was complicated, and there were a number of unfamiliar words. Further, because of my unfamiliarity with the topic,
vocabulary or dialect, I struggled to extract propositions about the text (Farrell, 2009). The basic grammatical structure in English is subject-verb-object, with the subject often
representing the noun for place or person, and the verb which identifies the action or being of the sentence (Schmitt, 2000). However, the grammatical structure in the extract is complex. In this clause, for example, 'with its great wings half-folded across the great belly of it'’, the naïve reader may have forgotten already the subject of the sentence, beast, or have not quite grasped that the author is describing what a shepherd would have once seen at dusk (gloaming) in Kinraddie’s mythical past. Additionally, the sentence contains 74 words, which is a long sentence even for a L1 user, has one subject beast, different types of verbs lay about, with a number of aspects in the past tense such as poked, and follows an oral, rather than traditional narrative form. Further, the syntactical and
grammatical basis of the text is to be found in Scots Gaelic and archaic English (childe).
I found it difficult to identify the meaning of ‘lay’, in 'In the Den of Kinraddie one such beast had its lair and by day it lay about the woods'. I was unsure if it was a transitive verb that requires a direct subject and object as in 'lay about the woods', or an adjective having the meaning of ‘secular’ (as in ‘lay person’, the meaning I know). The verb comes after a number of words that have different functions in the sentence, and that also confused me.
Had I followed the bottom-up model, I might have started with the graphics in order to comprehend the text. However, when I tried to break the sentences into their constituent parts I still could not derive meaning from the words alone or together. Moreover, I was unable to grasp the writer's intended meaning by combining words and linguistic structures because the grammatical and lexical patterns were above my linguistic level. So far, the model is unable to account for the difficulties foreign language will encounter reading texts such as these.
The following section discusses in more detail the bottom-up model from Gough’s (1972) original view as his 1972 model was considered as the most important bottom-up reading model.
2.3.1 Gough’s (1972) bottom-up model
The most influential bottom-up model was introduced by Gough (1972), and became known as the 'bottom-up' information-processing model focused on ‘one second of reading’. What made this model interesting when it appeared was its contrast to
Goodman’s top-down model discussed in Section 2.4.1. Instead of regarding a good reader as an ‘intelligent guesser’, Gough characterised the good reader as a passive decoder who makes little use of the text’s context. Following Gough's model, a reader would focus on analysing and breaking the words into segments, and ‘plod through the sentence, letter by letter, word by word’ (Gough, 1972, p.354). From Gough’s point of view, learning to read is learning to decode, namely, changing graphic characters into phonemes. As a result, the printed form can be changed into a spoken form while reading:
The reader converts characters into systematic phonemes… The reader knows the rules that relate one set of abstract entities to another… The Reader is a decoder... (Gough, 1972, p.310)
Gough’s (1972) linear model of reading states that the reader follows a number of stages when processing a text: 1) eye fixation, 2) letter identification, 3) phonological
representation, 4) understanding of words serially from left to right, and 5) absorption of visual stimulus. In other words, reading, according to Gough (1972), begins by capturing each letter to examine and identify an image from the text. After the image is identified, the decoding process starts. According to Williams (2006, p.355), the central feature of Gough’s model is that the processing moves in one direction, from the bottom, which is the perception of letters on the page, to the top, which is the cognitive process that constructs meaning. Now, I apply Gough’s model to my reading of the following extract:
When I began reading the text using Gough’s model to guide me, I started with the first word ‘Kinraddie’ which contains three syllables: Kin-rad-die. At first, I thought the syllable ‘Kin-' was related to the word ‘kill’ because the last syllable is ‘die’. Then I read the whole sentence, but found that the meaning did not make sense because of the verb
‘won’ which follows and bears no relation to ‘kill’. I then took each word in the sentence, decoded it, and to that extent I followed the bottom-up model. However, this failed to yield meaning. I did not know the meaning of the words, even when separating them into
syllables to try to understand each word individually. As I explained in the introduction, the word 'Kinraddie' has no linguistic correlation in my L1. I sought help from my
supervisors, who explained that ‘Kin’ is from Scottish Gaelic, ‘cean’ means ‘head’, while
‘raddie’, the place name, is fictional though the 'aidh' (anglicised as 'ie' here) is adjectival but in this instance ‘raddaidh’ has no place name meaning in Gaelic. I learned that the word ‘childe’ was an old spelling of ‘child’ and ‘Norman’ was a northern French national identity (see Chapter One). By then, I began to understand some of the context, and knew that I was missing knowledge of the language, culture, history, and place in which the novel was set. The importance of this kind of knowledge is missing from Gough’s model because he is interested in the relation of decoding to linguistics rather than to reading comprehension (Balota and Chumbley, 1990).
In this model, Gough (1972) states that the reader is not a guesser since he observes:
A guess may be a good thing… But rather than being a sign of normal reading, it indicates the child did not decode the word in question rapidly enough to read normally. The good reader need not guess; the bad should not. (Gough, 1972, p.317)
As an, arguably, 'poor' reader, I found myself trying to guess the meaning of the text, since bottom-up type strategies were failing me. According to Gough’s (1972) analysis of reading, Goodman’s (1967) guessing, as I discuss in Section 2.4.1, may be used when reading has broken down, and decoding is difficult or impossible in application. Guessing, however, did not help me here. Guessing may work for fluent or skilled L1 readers with a threshold level of understanding, but even a L1 speaker might struggle with words such as
‘Kinraddie’, ‘meikle’, or ‘gryphon’ unless they come from, or know people from, that region of Scotland, or have encountered these words elsewhere. It seems clear that applying a strict decoding process as suggested by Gough’s model cannot help either. I could identify letters, break words into graphemes and phonemes to pronounce whole
words like ‘Kinraddie', but I still did not understand the text. While I would decode the words, I could not comprehend them because decoding, while necessary, is not sufficient for reading comprehension, as Gough and Tunmer (1986) later acknowledge. The five linear processes that Gough (1972) suggests are involved in processing text failed once more to yield meaning. Based on this discussion, the following section discusses the problems of Gough’s bottom-up model.
2.3.2 Some problems with Gough’s (1972) model
According to Birch (2007), bottom-up models, such as Gough’s might help the reader
‘read faster and with better comprehension, because more efficient bottom-up reading leaves more attention for higher level processing’ (p. 114). However, Hedgcock and Ferris (2009) state that Gough’s model suffers from several weaknesses:
... including its equation of reading with speech, its narrow focus on
“sentences” (rather than propositions or texts), and its reliance on ill-defined (and untestable) processing mechanisms. (Hedgcock and Ferris (2009, p.19) These processing mechanisms, according to Hedgcock and Ferris (2009), include graphemic, phonological and syntactical processes. Rumelhart (1977) criticized the bottom-up model for its neglect of the role of the reader’s background knowledge, and I referred to this in my attempts to read A Scots Quair text above. Because it is a linear model in which comprehension goes in one direction from the bottom-up, Rumelhart (1977) argues that it cannot account for higher levels of reading interaction (see below).
Moreover, the bottom-up model of reading was not used to understand REFL processes until 1973, when Eskey stated that the bottom-up model undervalued the contribution of the reader because it failed to identify how the reader might use his/her background knowledge to comprehend the text. In other words, reader schemata or background knowledge, which I discuss in Chapter Three, were not recognised as having any role in comprehending the text.
As an EFL reader, I felt I lacked any skill in reading this text. I had no adequate
background knowledge about the topic and I was only using the information on the page.
Even when I succeeded in analysing the grammatical context of A Scots Quair, which took a long time, I was unable to understand the text because the semantic process that includes understanding the meaning of vocabulary was absent. I had the grammar but I did not have
the meaning or understanding of the orthography. Decoding and analysing the grammatical units in context is only part of the reading process. Even in the L1, in order to be a good reader, it is not enough to understand the structure and syntactical units. For reasons such as these, Williams (2006) criticises bottom-up models, stating that ‘they cannot account for context effects’ (p.365). For instance, readers reading in their L1 often miscue, for
example, correcting the reading mistakes immediately without waiting until the end of the sentence.
Later, Gough and Tunmer (1986), in a work known as ‘a simple view of reading’, stated that the reader could decode anything, but not necessarily understand everything s/he decoded. Gough and Tunmer (1986) stated that decoding is not sufficient for
comprehension but it is necessary for reading. They described reading, ‘R’ (reading comprehension), as an equal product of decoding, ‘D’ (decoding), and ‘C’
(comprehension). If ‘D’, that is, if the reader’s decoding skills, are zero then the
comprehension skill is zero: R = D x C. In other words, the ability to decode is at the core of reading ability, so that learning to decode is tantamount to learning to read (Gough and Tunmer, 1986). The reading process according to Gough’s and Tunmer’s (1986) viewpoint is a product of listening and decoding. For example, according to a simple view of reading, if I cannot read any of the words from the above Scots Quair extract that means my overall reading ability is zero (see Chapter Seven).
Despite these criticisms, I cannot ignore the bottom-up model in REFL because aspects of it are often used by learner-readers who depend on local strategies such as decoding to understand the text, and I used it in reading A Scots Quair. Gough’s bottom-up model appears to fit the audio-lingual method of teaching a foreign language that I discuss in Section 4.2.3, which views reading, according to Swaffar (1988, p.129), as a habitual matter, a stimulus-response reaction to the written symbol, as ‘verbal mechanics’. The audio-lingual method considers the decoding of the sound representation relationship as the main characteristic of learning the target language. Moreover, the bottom-up model involves processes that work in the opposite direction to the top-down model. While the reader in the bottom-up model is working from text to meaning (from the particular to the global), the top-down reader is working from meaning to the text (from the global to the particular). The following section discusses the top-down model processes in more detail.