Artificial symbol-word correspondence in a simple paired associate learning task were used to determine whether disabled readers have a general problem dealing with complex and/or irregular rule systems. The performance of 36 normal readers and 36 disabled readers in grades 4 through 7 was compared. Disabled readers had IQ scores of 87 or above and reading achievement of at least 1.7 years below grade level. The 12 symbol-word correspondneces were differentiated on the dimensions of consistency and conditionality. Differences between normal and disabled reader groups were observed to be specific to the presence of a rule within the set to be learned. When no rule was present and the set was consistent, reader groups performed similarly well. When no rule was present and the set was inconsistent, reader groups performed similarly poorly. However, when a rule was present, which presumably could facilitate learning of the pairs, disabled readers performed significantly more poorly than normal readers. Disabled readers did not differ from normal readers in their ability to detect the rule within a set, as measured by a later transfer and verbalization task. Performance of normal and disabled readers on the paired associate task was not found to be significantly correlated with IQ score in either the rule or nonrule condition. Limited evidence also suggested that the presence of inconsistency in a task containing a rule more adversely affected the learning of disabled than that of normal readers. (Author/SEW)