The Reading Eye: Word Perception

The problem Snell faced in designing his 19th century eye chart is symmetrically reversed from the problem faced by researchers in reading regarding word perception. Snell needed a visual object to serve as an “optotype” which had the property of shrinking or enlarging itself in close intervals. It also had to be generally recognized by the public as a stable object called by the same name, and it must not represent anything meaningful which a patient might not recognize. Letters of the alphabet met the criteria.

The problem reading researchers faced involved strings of letters, individually nameable serially as a, b, c, etc., but collectively representing a part of speech, a spelling, a map to pronunciation, a semantic load, and a syntactic potential. To find out how much time elapses between exposure to a word and consciousness of its syntactic potential—do readers need to recognize all of the possibilities before they can be said to have perceived a word?

*

The printed word is packed with potentials, a small, delicate box bridging gaps between human consciousness made by time or space. In his book Science in Action (1987), Latour theorized “block boxes” as sealed, scientifically-certified containers filled with and protecting actants, a vetted set of living and non-living agents functioning together in sequence or in parallel in the same predictable and trusted way to do something, say, get rid of germs. All of the aspects of a successful single word in bold above are separate agents.

People purchase milk from grocery stores with full faith that drinking it will be nutritious, safe, and sanitary. Why? A black box. Pasteurization. It is what it is and always will be. We rely on it. People use all of their words with similarly benign expectations. Day in and day out we trust words to do what we need them to do whether talking together or reading. We slip them quickly and fluidly into slots, notice meaningful word particles, access meaning, etc. with alacrity.

Word perception seems pretty straightforward on the surface. You hear “bottoms up,” you drink your beer. You read “total amount due,” you pay your bill. That’s because you can recognize these words instantly, everything you need to know to change a tire in a flash, these miraculous black boxes non-problematical and essential to perform or respond effectively.

Word perception as a technical term has distinct conceptual edges, not always in exact alignment in competing models even among highly knowledgeable experts, but mainstream reading scientists concur: Useful knowledge about word perception for understanding reading as a cognitive process involves accessing the meaning of the word in isolation and in context. The goal of word perception is lexical access.

Common phrases like word recognition, word identification, word analysis, word attack, and word perception each get at nuances in processing morphemes during reading which the mature reader does automatically, effortlessly, and unconsciously. For purposes of this essay, I’m defining ‘word perception’ in alignment with the mainstream as all of the sensory stimulation, neurological highways, and cognitive loops and nodes a printed word must pass through to transition from a string of letters to activate lexical access, that is, finding the entry in one’s mental dictionary and bringing stored meaning to bear in the task of reconstructing a message.

***

The central element in the transition from beginning to intermediate phases of reading development is arguably the intensity of the labor required to perceive a word. Readers at very early stages often approach the task with extremely strong demands on their attention, even working through initial identification of a word one letter at a time. Some children exert so much effort to learn simply to identify a string of letters that they need equally intensive instruction. Most readers, particularly those coming from quality preschool programs, after just weeks to months of beginning direct instruction in phonics begin to automate some identification tasks and are soon ready to leave the nest and develop their own identity as a reader.

Intermediate readers have a high degree of automation in word identification processes in texts using familiar words that appear with high frequency. They also function with more complex syntactic structures involving coordination, subordination, and modification. Increased sophistication in using traffic signals like adverbial connectives, relative pronouns, and transistion words and phrases strengthen coherence and structural organization.

Their ongoing development occurs generally in the areas of fluency, vocabulary, metacognition, text structures, and world knowledge, especially when they have easy access to a variety of interesting texts in a community of voluntary readers. Motive driven by interest, need, and curiosity fuels expeditions into more challenging texts. Comprehension begins to drive textwork. How do the eyes contribute?

***

How long does it take to identify a word¹? To answer this question, first it helps to understand methods researchers have used historically, including but not limited to response time, brief presentation, and estimates from reading texts. Each method has produced well-established scientific conclusions.

Response time is straightforward: How much time passes between presentation of a word and the production of a response? The accuracy of this method depends on the kind of response expected. Measuring the time between presentation of the word and pronunciation of the word isn’t a logical way if the goal of word perception requires lexical access (accessing meaning).

For one thing, after the word is identified, in order to say the word, the reader must access a motor program, send a command through a neural pathway to the speech articulation anatomy, and carry out the command—identification gets muddled by irrelevant response-execution time. It takes experienced readers after practicing the task as much as 400 msec to say a word—and that doesn’t count the most important thing: lexical access. What does the word mean? This approach doesn’t collect that data. Some have argued that using nonsense words as a steady diet to teach children short-circuits the full perceptual loop because perceptual completion ends at articulation, not at lexical access.

Brief presentation methods have also been used and criticized. The researcher flashes a word on a screen for 60 msec, and the subject responds by saying the word and giving a synonym. Theorists have argued against this method; the visual presentation might last 60 msec, but other processes occurred such as speech articulation and lexical access which go unmeasured. In other words, the retinal image, which we know lingers after a word goes dark, still must access a motor program, etc., but the time this mental activity takes is ignored.

Estimates from time spent reading have some practical appeal, but this method has its problems as well. Time spent reading a passage involves many processes happening holistically beyond word identification. Researchers have no idea how many words a reader skipped. If rough patches appear in the text, readers can slow down to half of their reading rate. Further, it’s virtually impossible to predict in naturally occurring reading which patches are going to be rough for which readers.

After summarizing the research up to 1989, Rayner and Pollatsek concluded the following:

“We are certain that 700 msec (categorical reaction time) is too slow for an estimate of time to [lexical access]… We are pretty sure, but not certain, that 400 msec (naming time) is also too slow…. We are also pretty sure that 60 msec (rapid exposure) is too fast…. …[S]omething like 150 msec is about right, although we wouldn’t be too surprised if word identification took place in as little as 100 msec or as much as 200 msec after the words is first sensed by the eye” (p.68).

Idiosyncratic conditions have also been tested. Is it possible to identify a word, that is, know the meaning of a word, without having fully seen the word? Word masking techniques have demonstrated that it is possible. A word is exposed for 20 msec (remember the lowest threshold for rapid exposure reponse is 60 msec), then masked. Subjects are not able to say what the exact word was, but they know its meaning.

Unconscious priming is another effect supported by evidence. Priming involves presenting a word (the prime), measuring response time, presenting another word (the target), measuring the response time, and then comparing the pattern to another priming set. For example, when the word DOG is presented as a prime and CAT is the target, the word CAT is identified 30-50 msec faster than when the pair is FAN (prime) and CAT.

The “Stroop Effect” (Stroop, 1935) is easily replicated. Preparation involves writing a color word in a different color ink. Take a word like RED and print it in green ink. Now print the word RED in red ink. Test the first green lettered RED and record response time. Then test the red lettered RED. Stroop demonstrated that subjects are 200 msec slower to identify “green” red than “red” red. Another experiment discovered that readers are slower to identify the word CAT when it is printed over a line drawing of a dog. Generally, these experiments suggest that priming can slow lexical access down as well as increase it.

The following language expresses the overarching conclusion Rayner and Pollatsek reached regarding the average speed of silent word recognition including lexical access:

1. Note: Much of the remaining discussion is drawn from Rayner and Pollatsek’s (1989) graduate school textbook supplemented with original sources.