Thursday, 7 February 2019

The Simple View of Reading: Still Conclusive After 33 Years

The Simple View of Reading is, as the name suggests, a simple (but not simplistic) representation of what the reading process entails, particularly when viewed with the novice in mind. It was proposed over a generation ago, but its timing could not have been worse, arriving as it did, during “Peak Whole Language” in the mid-1980s. It seems to have by-passed a generation of teachers in their pre-service education, but has thrived in cognitive psychology and speech-language pathology circles, where it has informed theory, research, and practice for the last three decades. Regular readers of this blog will know that one of my gripes is the ITE-auspiced erosion of teacher knowledge that has taken place in recent decades.

In this guest blogpost, retired Boston teacher Stephen Parker (@ParkerPhonics) takes a look at the Simple View of Reading, unpacks its key elements, and discusses its implications for reading instruction and for supporting struggling readers.

For many teachers, the Simple View of Reading was a gift that went missing in the mail. Here it is, re-packaged and ready to assist in the start of another Australian school year. 

Note –  
  1. You can download free copies of Stephen’s books on phonics instruction at his website.
  2. Stephen has now commenced his own blog, and this piece is published there as well. See this link.

*******

In 1986, Whole Language, a philosophy for teaching reading that rejected systematic phonics, was approaching peak popularity. The two founders of Whole Language, Ken Goodman and Frank Smith, had little patience for decoding, that is, for matching a sound with each letter (or letter group) in a word and then blending those sounds together, left to right, to sound out the word.
Goodman believed that “Matching letters with sounds is a flat-earth view of the world, since it rejects modern science about reading and writing and how they develop.” Smith added: “Reliance on phonics – or spelling-to-sound correspondence – is dysfunctional in fluent reading and interferes with learning to read”. Whole Language, in 1986, was nearly universally accepted by the educational establishment and by reading teachers everywhere.

All the more surprising, therefore, that in that same year, Philip Gough and William Tunmer proposed their Simple View of Reading. Surprising because their model placed “decoding” front and center, right along with “language comprehension,” as the two independent factors necessary for a child to read. The model succinctly states:

R  =  D  x  C

where R stands for “reading comprehension,” D for “decoding.” and C for “language (listening) comprehension.”

To understand this model today, 33 years after it was proposed, it’s important to know how Gough and Tunmer understood their three variables. It’s also important to recognize that the numerical values assigned to the variables D and C are multiplied, not added. 

Language comprehension (C) involves not only hearing words spoken by another, but also understanding what was just spoken. Reading comprehension (R), which comes later developmentally, is exactly what everyone means when they speak about reading something: not only producing the words (mentally or aloud), but also understanding what was just read. 

D stands for two distinct types of “decoding”: the decoding done by a beginner (the “sounding out” defined above) – as well as the decoding done by a skilled reader. Skilled readers don’t sound out words, rather, they simply recognize the words immediately, as “sight words.” These are sight words in the full sense, meaning they were created automatically and unconsciously precisely because the child was taught the letter-sound correspondences of the alphabetic code and then used those correspondences – at least initially – to sound out words. 

In other words, "skilled" decoding (automatic word recognition) directly depends upon the child going through the process of “primitive” decoding (matching letters with sounds and then blending those sounds to produce a pronunciation). Gough and Tunmer expressed it this way: “Word recognition skill is fundamentally dependent upon knowledge of letter-sound correspondences…We assume that decoding varies directly with knowledge of letter-sound correspondences.”

Since Gough and Tunmer referred to two distinct types of “comprehension” in their model, it will likely avoid some confusion if we take the liberty to rename two of the variables in the Simple View this way: 

RC  =  D  x  LC

Reading Comprehension is the product of Decoding and Language Comprehension. The variables D and LC in their model can each be assigned a numerical value that ranges from 0 (no skill) to 1 (perfection). Those numbers are then multiplied to determine RC (reading comprehension). 

There’s a lot to unpack here. First, when two numbers in the range 0 to 1 are multiplied rather than added, the result will usually be a number smaller than either of the multipliers (e.g. 1/2  x  1/4   =  1/8). I will present some examples below that demonstrate the significance of this fact. 

Second, D and LC are independent of each other. For instance, I studied Spanish for two years in high school nearly 50 years ago. Based on that experience, I can decode Spanish text quite well (D = 1) but I can’t understand spoken Spanish (LC = 0). So, in this case we have RC  =  D  x  LC  =  1  x  0  =  0. You can see then, that in using this model, it’s quite possible to decode text, yet fail (completely) to read that text. Decoding is a necessary but insufficient condition for reading.

Conversely, a child can have perfectly good (age-appropriate) language comprehension skills (LC = 1) yet have no decoding skill (D = 0). Many four and five-year-olds are living examples of this. But knowing a language does not make one literate. In this case, RC  =  D  x  LC  =  0  x  1  =  0. If D = 0, reading ability is 0 no matter how good language comprehension might be. Language Comprehension is also a necessary but insufficient condition for reading. The necessary AND sufficient condition for reading to take place is that BOTH decoding (D) and language comprehension (LC) have values greater than 0.

The Ideal
The ideal situation is one where a student attains, as quickly and efficiently as possible, a decoding score of 1. In such a case, RC  =  D  x  LC  =  1  x  LC  =  LC, or simply: RC  =  LC. In other words, once a child can fully and accurately decode the print on the page into sound, even if that sound only unfolds in her head, her reading comprehension (RC) will be every bit as good (or bad) as her current language comprehension (LC). For her, it will be as though the text on the page were being spoken by someone else. 

As D approaches 1, the finite task of “learning to read” evolves into the life-long task of “reading to learn.” From this point on, her reading comprehension (RC) and her language comprehension (LC) will increase in unison (RC = LC). Reading about a new subject will improve her language skill just as conversation about a new subject will improve her reading skill. A lifetime of learning via both reading and conversation lies ahead, as “sounding out words” gives way to “automatic word recognition.”

Less Than Ideal (Dyslexia)
There are many children in school – and countless adults in our communities – whose language comprehension skills (LC) are fully (or nearly) appropriate for their age (so 0.9 < LC < 1), but whose decoding skills are poor to non-existent (so 0 < D < 0.2). These individuals might know a couple hundred consciously-memorized sight words, but they don’t have the ability to accurately sound out unfamiliar words. Therefore, their labored reading does not create new sight words automatically, as previously discussed. Their ability to engage with text is severely constrained – as are their educational and vocational prospects.

For this large population of poor decoders, the Simple View yields something like this:  RC  =  D  x  LC  =  0.20  x  0.90  =  0.18. Their reading comprehension (18%) will be poor indeed. Gough and Tunmer called this condition “dyslexia.” Dyslexics can’t read because they can’t decode. 

“We take no position on whether there are one or more ultimate causes of dyslexia. But we suggest that there is a common denominator in every case of dyslexia… an inability to decode. This is not to say that we claim to have identified the ultimate cause of dyslexia; for this, one would have to push the question one step back and ask why they cannot decode.”
Gough and Tunmer conceded that the ultimate cause of dyslexia might well be genetic and neurological, but they left open the possibility that dyslexia could also result simply because the individual has never been taught, properly, how to decode. 

Less Than Ideal (Poor Language Comprehension)
Reading disability can occur the other way as well. A child can have relatively good decoding skills (D = 0.9) but poor language comprehension skills (LC = 0.3). This might be the case if the child comes from a disadvantaged background, or if the child has a developmental disorder that compromises LC, or if English is a second language. For him:  RC  =  D  x  LC  =  0.90  x  0.30  =  0.27. Again, we have a child who can barely read. 

And, of course, there are readers who are deficient in both decoding and language comprehension. Here the Simple View yields the expected result: RC  =  D  x  LC  =  0.20  x  0.30  =  0.06, a child who is nearly illiterate and who needs support with both decoding and language comprehension in order to make progress. 


The Simple View asserts only that both decoding (D) and language comprehension (LC) are essential to reading. Note that this theory is easily falsifiable. To counter the Simple View, one need only show:

1) There are students with good decoding skills and good language comprehension skills, but who nonetheless can’t read.

OR

2) There are students who can do one but not the other and yet can still read with skill and understanding. 

Whole Language instruction can be viewed as an attempt to falsify the Simple View in the second way. If Whole Language founder Ken Goodman had been correct in his assessment of reading as a psycholinguistic guessing game, with little-to-no need of letter-sound knowledge and decoding, Whole Language would have succeeded, at least with children having good language comprehension (LC) skills. Instead, Whole Language failed. 

In the 33 years since Gough and Tunmer proposed the Simple View, no one has been able to prove it false. In fact, experiments show that the Simple View accounts for nearly all the variation in reading comprehension found in the general population. 

I suspect the Simple View will never be falsified because it expresses, in a single eloquent equation, something that is fundamentally common sense: that in order to read alphabetic text, one must be able to transform that text into the sound it symbolizes and then understand the result. Such text, after all, is nothing more than sound that was previously encoded onto paper; decoding it, and understanding it, is reading.

Implications for Reading Instruction
Gough and Tunmer were circumspect about the instructional implications of their model of reading. In fact, they stated explicitly that they “do not wish to discuss the place of decoding in reading instruction.” They were, after all, trying to publish their model at the height of the Whole Language craze. Nonetheless, one can get a hint of what they thought from the following: 

“If decoding plays a central role in the reading process, then it seems sensible to give it a comparable place in instruction… If we were to learn that decoding plays no role at all in skilled reading, it does not follow that we should ignore decoding in reading instruction. It might well be that direct instruction in Synthetic Phonics is the fastest route to skilled reading.”
Personally, I think the implications of the Simple View for reading instruction are unavoidable. Consider these two scenarios.

Scenario 1: During the critical first two years of reading instruction, the bulk of time is spent on “invented spellings,” “pretend reading,” rote-memorization of sight words, and lots of guessing. (Children guess the meaning of unknown words based on pictures, context, or the word’s first letter.) When enough sight words have been consciously memorized, analytic phonics is used to have the children analyze those sight words and gradually “discover” the letter-sound correspondences of the code.  

Under the best of circumstances, this type of teaching can increase a given child’s decoding score (D) from 0 to near 1 over a period of about 6 years. (See, for example, Words Their Way by Donald Bear.) At the end of this period, RC will equal LC and the children who remain will be fully capable of “reading to learn.” I say “children who remain” because many children will have given up on reading in the first two years due to all the tedious sight word memorization and guessing. (Note: What I have just briefly sketched out here is today’s Balanced Literacy.)

Scenario 2: The teacher spends the first two critical years directly and systematically teaching the letter-sound correspondences of the alphabetic code to all students. Such instruction goes by the name phonics-first or Synthetic Phonics. From Day 1, students are taught not only the code, but also two critically important phonemic awareness skills: blending (for reading) and segmenting (for spelling). 

In a Synthetic Phonics program, students can reach the ideal (D = 1, RC = LC) and be on their way to a lifetime of learning, in two years rather than in six. In addition, less students will give up on reading because Synthetic Phonics logically explains spellings right from the start – and the hated drudgery of consciously learning sight words is bypassed entirely. 

To forestall the criticism that Synthetic Phonics spends two years on decoding (D) while ignoring the child’s language comprehension skills (LC), I’ll add the following. Every day, a Synthetic Phonics class should be split (roughly) in half – the first part devoted to explicit teaching of the code, and the second to the reading of classic children’s literature. The reading, however, is done by the teacher for the whole class, and it allows plenty of time for discussion about what was just read. In this manner, both D and LC increase daily, for every child, during the two years required to complete Synthetic Phonics.

There is but one way to teach reading that fully aligns with the Simple View. That way is Synthetic Phonics.



© Stephen Parker & Pamela Snow (2019)

3 comments:

  1. This is the best account of the SVR that I have come across. Nevertheless, I do have some questions. One is this. You say “in order to read alphabetic text, one must be able to transform that text into the sound it symbolizes and then understand the result.” Are you proposing here that the way skilled readers comprehend text is by first translating it to sound, and then comprehending the resulting phonological representation?

    ReplyDelete
    Replies
    1. Thank you for that compliment Max. Coming as it is from such a well-known reading research scientist as yourself, I’m pleased.

      I’m a teacher, not a research scientist, but I’ll attempt to answer the question you pose. For beginners, that is to say, unskilled readers, the path to reading comprehension, apart from some rote-memorized sight words, is through conversion of text into sound. I’ll assume most folks would agree with that.

      Your question, however, asks about skilled readers. Those are readers who have, over time, built a fairly large collection of sight words that have been created automatically – something possible only because they at one time engaged in the “primitive” decoding (sounding out) that distinguishes the beginner. For these skilled readers, both types of sight words, those rote-memorized and those created automatically, are stored in a personal mental dictionary.

      Am I proposing that skilled readers must first translate text into sound, either by reading aloud or by sub-vocalizing mentally, in order to comprehend the text? My answer, at least for single words, is no. I think there must be 2 distinct routes from text to meaning, one direct, via finding the word in one’s mental dictionary, and the other indirect, via using learned grapheme/phoneme correspondences (GPCs) to produce sound.

      For “words” (symbolic representations of sound) like COLONEL, CHOIR, 35, and #, there is no possibility of sounding them out first and then accessing meaning. Meaning, provided by the mental dictionary, comes first, followed by sound: KERNUL, KWIRE, THIRTY-FIVE, HASH TAG.

      On the other hand, real words never before encountered (e.g. CHURLISH, ALACRITY), and nonsense words (GRINCH, FLAM, PLID) are not in the mental dictionary, yet a skilled reader has no trouble converting all of them into sound. In this case, meaning follows sound (if the sound is in the individual’s speaking vocabulary).

      I suspect the 2 routes operate in a BOTH/AND fashion rather than in EITHER/OR mode. In her book “Beginning to Read,” Marilyn Adams claims that sound activation is “an automatic and immediate consequence” of seeing text, and that for skilled readers, both meaning and sound are activated “nearly instantaneously.”

      Adams sees two advantages in such automatic sound activation. First, the sound route provides “alphabetic back-up” of words that are in the mental dictionary. This accounts, I believe, for the fact that skilled, educated readers can have mental dictionaries with as many as 70,000 entries – a feat that would be impossible without the back-up sound system. Second, she views sound as aiding in the comprehension of complex sentences: sentences with multiple clauses and 25+ words. In this case, the sound track aids the reader’s running short-term memory so necessary in order to accurately process long sentences.

      Max, since you’re an expert on “dual-route theory,” I’d certainly welcome some additional input from you.

      Stephen

      Delete
    2. That all seems correct to me, Stephen. To summarise: the rule-based route has a crucial role to play in learning to read, but it contributes very little to skilled reading, where the dictionary lookup route has the dominant role. This view is supported by a large amount of literature.

      As you have pointed out, the “D” in the SVR equation has two different meanings.

      The SVR equation RC = D x LC when “D” stands for “the decoding done by a beginner (the “sounding out” defined above)” applies to learning to read (not skilled reading).

      The SVR equation RC = D x LC when “D” stands for “the decoding done by a skilled reader” (i.e. the dictionary lookup route) applies to skilled reading (not learning to read).

      Nothing wrong with that as long as people when using the term “decoding” make it clear which of these two meanings they have in mind.

      Delete