Philip B. Gough

Decoding, Reading, and Reading Disability

To clarify the role of decoding in reading and reading disability, a simple model of reading is proposed, which holds that reading equals the product of decoding and comprehension. It follows that there must be three types of reading disability, resulting from an inability to decode, an inability to comprehend, or both. It is argued that the first is dyslexia, the second hyperlexia, and the third common, or garden variety, reading disability.

The simple view of reading

A simple view of reading was outlined that consisted of two components, decoding and linguistic comprehension, both held to be necessary for skilled reading. Three predictions drawn from the simple view were assessed in a longitudinal sample of English-Spanish bilingual children in first through fourth grade. The results supported each prediction: (a) The linear combination of decoding and listening comprehension made substantial contributions toward explaining variation in reading comprehension, but the estimates were significantly improved by inclusion of the product of the two components; (b) the correlations between decoding and listening comprehension tended to become negative as samples were successively restricted to less skilled readers; and (c) the pattern of linear relationships between listening and reading comprehension for increasing levels of decoding skill revealed constant intercept values of zero and positive slope values increasing in magnitude. These results support the view that skill in reading can be simply characterized as the product of skill in decoding and linguistic comprehension. The paper concludes with a discussion of the implications of the simple view for the practice of reading instruction, the definition of reading disability, and the notion of literacy.

Learning to read: An unnatural act.

The six-year-olds sight is as good as the adults (Amigo 1972), and his hearing is nearly so (Elliott and Katz 1980). The child has an excellent memory (Mandler, in press), and his learning ability is remarkable. Even a conservative estimate of the size of his vocabulary will show that he must have learned, on average, more than four new words every day since his first birthday (Carey 1978). He has already learned to speak and understand his native language with remarkable fluency. The average American six-year-old can already produce and recognize more than a dozen vowels and nearly 30 consonants of English. He can produce and understand literally thousands of different words, and he can comprehend virtually any sentence that one can form with those words. To be sure, his language acquisition is not complete. Over the next decade he may have to smoothe out some rough spots in his phonology (Templin 1953), his vocabulary will grow by many more thousands of words (Oldfield 1963), and he must capture a few syntactic niceties which still escape him (Chomsky 1969). But his mastery of English would be the envy of any college graduate learning English as a second language. Yet for all his cognitive and linguistic talents, the child has one peculiar linguistic shortcoming: he cannot read a word. Indeed, that is one of the primary reasons why we now send him to elementary school. His teacher

Grammatical transformations and speed of understanding

This study tested an hypothesis that the hearer of a complex sentence must transform that sentence into the underlying kernel sentence before understanding it, and hence that speed of understanding a sentence would vary with the number and nature of the transformations separating it from its kernel. Descriptive sentences of varying grammatical form were presented to Ss who were asked to verify them, and the speed of verification was taken as an index of speed of understanding. Active sentences were found to be verified faster than passive, affirmative faster than negative, and true faster than false. The true-false variable was found to interact with the affirmative-negative, indicating that the latter difference is not simply syntactical. The consistency of the results with the hypothesis was noteworthy, but transformational complexity was confounded with frequency and length.

The verification of sentences: The effects of delay of evidence and sentence length

Two experiments were conducted to explore further the relationship between the syntactic stracture of a sentence and its verification. In Exp. I, the evidence confirming or disconfirming a sentence was delayed for 3 sec after the sentence. Active sentences were verified faster than passive and affirmative faster than negative, contrary to an hypothesis that the hearer immediately decodes a complex sentence by transforming it into its underlying structure. In Exp. II, passive sentences were made shorter than active by deleting their agent phrases. Active sentences were still verified faster, indicating that the relation of verification time to sentence structure is not an artifact of sentence length.

How children learn to read and why they fail.

The present article considers the contrast between conceptions of reading as a natural and as an unnatural act, relying on the simple view of reading as a theoretical framework (Gough and Tunmer 1986). According to the simple view, reading comprehension is a product of both listening comprehension and decoding. Here it is argued that the comprehension aspect of reading depends on those same—natural—forces that govern acquisition of spoken language, whereas decoding depends on explicit tutelage, with little evidence that children will induce the cipher from simple exposure to written words and their pronunciations (sight-word instruction). Rejecting both sight-word and phonics instruction as inadequate in and of themselves, evidence is reviewed suggesting that successful readers require explicit awareness of the phonological structure of spoken words, which can and should be taught in kindergarten, prior to formal reading instruction. Beyond this point, reading success depends on a modicum of phonics instruction together with extensive practice with reading itself.

Two ideas about spelling: rules and word-specific memory

Abstract Two ideas have been prevalent in spelling research. One is that we spell by using rules which map phonemes onto graphemes (the rule idea); the other is that we rely upon word-specific memory of spellings (the memory idea). We tested the hypothesis that good spellers make significant use of rules in addition to word-specific memory. In Experiment 1, 23 college students were given a standardized oral dictation spelling test; a multiple regression analysis of item difficulty showed that variables associated with the rule idea explained significant variance in spelling accuracy when word-specific memory variables were statistically controlled. Similarly, when rule variables were statistically controlled, word-specific memory variables explained significant variance. Experiment 2 used an experimental design to test a second prediction of the rule idea: phonemes with high-ambiguity rules should be more difficult to spell than less ambiguous phonemes, even when word frequency, word length, serial position within the word, and the target letter are experimentally controlled. Twenty-nine subjects were given an oral dictation spelling test, and the results showed that there were more errors on high than on low ambiguity phonemes. In Experiment 3, word frequency and phoneme ambiguity were manipulated factorially, and results consistent with both the memory and rule ideas were found. Experiment 3 also allowed us to ask how word frequency and rule ambiguity interact, and this led to suggestions of how rules and memory might interact in a processing model of spelling.

The beginning of decoding

It is widely agreed that children recognize their first words in a different way than they later decode. One hypothesis is that sight words are recognized as wholes, another that they are recognized by parts. Two experiments were devised to compare these hypotheses. In one, children were taught a sight word accompanied by a salient extraneous cue and then tested for recognition of the word and the cue. In the other, children were taught sight words, then tested for recognition of each half of the word. The children were found to recognize the cue but not the word; they recognized one half of the word but not the other. The results support the idea that first words are recognized by selective association.

Forewarning, meaning, and semantic decision latency

Twelve Ss were asked to decide if pairs of nonsense syllables were identical or different. A second group of 12 Ss was asked to decide if pairs of words were synonymous or not. Prior to each pair all Ss were given either one member of the pair, an unrelated item of the same type, or the word “set.”Decisions about nonsense syllables were made faster than ones for words, and type of forewarning was also significant. In a second experiment 16 Ss were asked to judge the synonymity of 16 pairs of words where one was rare and one common. Ss had true and false forewarning. Forewarning was significant, but did not differentially effect judgments on rare and common words.

Comment on “The insufficiency of a finite state model for verbal reconstructive memory”

Braine (1965) has contended that, in the acquisition of a certain miniature language, what is learned cannot be represented by a finite state model because it requires the assumption that S invented some of its rules. It is argued that any grammatical description of what S learned requires this assumption, and that the finite state model is inadequate instead because it poorly represents what S does invent.