Do fluent, adult readers read whole words as ‘sight words’? Nooooooooooooo

A question that comes up repeatedly in regard to adult readers’ fluent reading is whether such fluent readers recognise whole words as ‘sight words’ or process through words so fast that it falls below the level of their conscious attention, rendering them unaware of what’s going on. In short, the answer is the latter! Just because something appears to be the case doesn’t make it so. We know from very many studies that what may seem to be happening on the surface is very different from what may be happening at a subconscious level.

Where did the idea that fluent readers read whole words by sight and without processing each letter or group of letters within a given word come from? In fact it all started over a hundred years ago with James Cattell, who reasoned that because adults could both read common words as fast as they could name letters and they could read words in context twice as fast as they could read words in isolation, this provided incontrovertible, scientific proof that adults were whole word readers. Oh that things were so simple.

The idea has also become known as ‘late stage sight word  theory’, which proposes that once a reader has read a word and become familiar with it, it ‘appears’ to be recognised instantly as a whole. As you can see from the way in which I have problematised the word ‘appears’, the reality is very different, and what the proponents of this theory fail to recognise is that the tests conducted by Cattell could just as easily have to do with the speed of motor processing or the rate of speech output NOT the speed of processing visual input.

Diane McGuinness is particularly scornful of such explanations about how people read and maintains that any attempt to ‘infer something about perception, cognition, brain processing and speech production from a single measure of response time is extremely naive’. As Keith Rayner put it in his brilliant study ‘Eye movements in reading and information processing: 20 years of research’, ‘Any single measure of processing time per word is a pale reflection of the reality of cognitive processing’.

What exactly is involved in the reading/speaking act? You have to look and focus, scan in a saccade or a number of saccades, transform the visual information into sounds and thus into the word, and finally execute the output or speak. Given that individuals response times vary from 450 to 800 milliseconds to read a word, what McGuinness asks is if there is any evidence why phonological decoding stops just because we read faster.

As brain studies show, most of what goes on in the mind happens under the level of our conscious attention. Imagine for a moment that you go on a training day. As the trainer appears before you, within milliseconds you have made all sorts of assumptions about what the person is like from what they are wearing, how old you think they are, their accent and other aspects of their speech, their body language, etc, etc. Very often, you won’t even be aware of having made these kinds of assessments being made unless something happens to surprise you and contradict first impressions. So, we form our impressions instantaneously and then spend subsequent time with the person in question confirming or questioning those first impressions. How many times have we had the conversation with someone who later becomes a friend and we find ourselves confessing that ‘When I first met you, I thought XXX, but then I discovered later that …

In terms of conscious processing, i.e. switching on our (deliberate) conscious awareness of a situation, our immediate reactions to situations are often very ponderous and require much-needed time to allow us to process relevant information. This also comes at the cost of then missing other events taking place around us. This often happens when we’re reading a text that contains words and information with which we are unfamiliar: the effort of processing the words (decoding) will often necessitate our having to re-read the text because we’ve lost sight of the big picture.

You will notice that the response to the hypothetical trainer will be multi-faceted and processing takes place in parallel, simultaneously: seeing, hearing, listening and smelling. As McGuinness states: ‘The brain can map grapheme-phoneme correspondence, analyze patterns of orthographic redundancy, register degrees of word familiarity, perceive context clues, and work out possible decodings of odd or unpredictable spellings, in parallel.’ She further adds that we do many of these things when we are listening to someone speak. In other words, the brain performs a vast array of analyses at a speed far in excess of anything we are consciously aware of never mind being able to control. The model proposed by late stage sight word theory is linear and sequential and fails utterly to account for the dazzling simultaneity of operations the brain is able to perform synchronously.

As McGuinness further points out, doffing her cap in the direction of Robert Glushko, whose research in this area was so seminal, ‘all the information about a word – visual, phonological, orthographic, semantic – is processed at the same time in parallel. This means that no matter how many elements contribute to successful decoding, processing is not carried out in separate (disconnected) pathways.

The subtlety of this kind of perspective on the reading process and the dual route (separate pathways) model is a bit like comparing a couple of telephone lines to the complexity of the internet.

None of this is to say that the decoding process is reduced in importance. Far from it! Decoding is a vital component of a complex mix and, as many in the research community assert, the more accurate and automatic the process of decoding, the more cognitive resources can be allocated to other aspects of textual interpretation.

We know from the work of Rayner (see also here and here) and his colleagues on the eye movements of experienced and fluent readers that they are sensitive to semantically related words, that their fixation durations decrease with the frequency of the same words in the text, that they ‘look longer at morphemes in long words that are more informative with respect to overall meaning of the word’, and that ‘fixation time in the region of a pronoun varies as to a function of how easy it is to make the link between the pronoun and its antecedent’. All of which is sophisticated stuff, but none of this can happen if the reader can’t decode the words on the page!

Glushko, R., ‘The Organization and Activation of Orthographic Knowledge in Reading Aloud’, 1979, Vol 5, No 4, 674-691.
McGuinness, D., (2004), Early Reading Instruction, MIT Press.
Rayner, K., ‘Eye Movements in Reading and Information Processing: 20 Years of Research, Psychological Bulletin, 1998, Vol. 124, No 3, 372-422