What We Know About How Good Readers Read Words (page 2)

We know a great deal more about how word recognition occurs than can be explained in this article. The theory that explains the incredibly fast ability of the brain to recognize words and associate them with meaning is called parallel distributed processing. This theory is complex but its most important tenets are easily understood. Information about a word is gained from its spelling (orthography), its pronunciation (phonology), its meaning (semantics), and the context in which the word occurs. The brain processes these sources of information in parallel, or simultaneously. The brain functions in word recognition, as it does in all other areas, as a pattern detector. Discussion of parallel distributed processing and its implications for word identification can be found in McClelland and Rumelhart, 1986; Rumelhart and McClelland, 1986; and Seidenberg and McClelland, 1989. The theory is translated and explained simply and elegantly in Adams (1990). Beyond the fact that the brain responds to many sources of information in parallel and that it functions as a pattern detector, the following specific facts seem particularly pertinent to the question of what kind of phonics instruction we should have.

Readers Look at Virtually All of the Words and Almost All the Letters in Those Words (McConkie, Kerr, Reddix, & Zola, 1987; Rayner & Pollatsek, 1989).  For many years, it was generally believed that sophisticated readers sampled text. Based on predictions about what words and letters they would see, readers were thought to look at the words and letters just enough to see if their predictions were confirmed. Eye-movement research carried out with computerized tracking has proven that, in reality, readers look at every word and almost every letter of each word. The amount of time spent processing each letter is incredibly small, only a few hundredths of a second. The astonishingly fast letter recognition for letters within familiar words and patterns is explained by the fact that our brains expect certain letters to occur in sequence with other letters.

To decode the unfamiliar word knob, for example, the child who knew many words that began with kn would immediately assign to the kn the n sound. The initial kn would be stored in the brain as a spelling pattern. If the child knew only a few other words with kn and hadn't read these words very often, that child would probably not have kn as a known spelling pattern and, thus, would have to do a quick search for known words that began with kn. If the child found the words know and knew and then tried this same sound on the unknown word knob, that child would have used the analogy strategy. Likewise, the child might know the pronunciation for ob because of having correctly read so many words containing the ob spelling pattern or might have had to access some words with ob to use them to come up with the pronunciation. The child who had no stored spelling patterns for kn or ob and no known words to access and compare to would be unlikely to successfully pronounce the unknown word knob.

Readers Divide Big Words as They See Them Based on Interletter Frequencies (Mewhort & Campbell, 1981; Seidenberg, 1987).  The research on syllabication rules show that it is quite possible to know the rules and still be unable to quickly and accurately pronounce novel polysyllabic words and equally possible to be able to pronounce them and not know the rules. Good readers "chunk" or divide words into manageable units. They do this based on the brain's incredible knowledge of which letters usually go together in words. If you did not recognize the word midnight in print, you would divide it between the d and the n. For the word Madrid, however, you would divide after the a, leaving the dr together. Interletter frequency theory explains this neatly by pointing out that the letters dr often occur together in syllables in words you know (drop, dry, Dracula). Words with the letters dn in the same syllable are almost nonexistent. This also explains why beginners might pronounce f-a-t-h-e-r as "fat her" but children who have some words from which the brain can generate interletter frequencies will leave the th together and pronounce "fath-er."

Although summarizing what the brain does to identify words runs the risk of oversimplification, it seems necessary before considering what we know about instruction. As we read, we look very quickly at almost all letters of each word. For most words, this visual information is recognized as a familiar pattern with which a spoken word is identified and pronounced. Words we have read before are instantly recognized as we see them. Words we have not read before are almost instantly pronounced based on spelling patterns the brain has seen in other words. If the word is a big word, the brain uses its interletter frequency knowledge (based on all the words it knows) to chunk the word into parts whose letter patterns can then be compared. Meaning is accessed through visual word recognition, but the sound of the word supports the visual information and helps to hold the word in memory.