Rule learning and complexity in simulated reading

Three approaches (Rule, Higher-order Unit, Whole Word) to decoding in a simulated reading task were studied. Rule complexity was varied in three experiments employing symbols ($, @, etc.) as stimuli and numbers as responses. CAI was the experimental vehicle. The 110 fifth and sixth grade boys and girls from an inner-city school were randomly assigned to one of the 11 groups in the three experiments. When the rule was simple (one-to-one correspondence) the rule approach was superior, but when the rule was complex (simulation of the final e rule in reading) the higher-order unit approach was superior. Higher-order unit subjects studied examples composed of the fewest stimuli possible to illustrate the rule.