On the word-superiority effect

SummaryFour experiments tested Johnston's (1981) hierarchical-activation account and McClelland and Rumelhart's (1981) interactive-activation account of the word-superiority effect. The hierarchical-activation account claims that the effect depends on masking. Using both a luminance manipulation and direct feature removal, however, we showed that the word-superiority effect can be obtained without a mask. Moreover, when words were made hard to read by displaying them in a column, the context advantage with both the luminance and the feature-removal techniques was reduced to zero. With masking, in contrast, a context effect remained because of lower accuracy in the no-context (control) condition. Thus, masking confounds enhancement on context trials with inhibition on no-context trials and is neither a necessary nor a desirable procedure with which to demonstrate a context advantage.The advantage of word context is reduced when words are printed with extra space between the letters. The amount of space needed depends on the size of the letters — the reduction requires a manipulation about the size of a blank character. Although both define visual features spatially, neither model includes an internal representation of space. Thus, neither model can represent the configural manipulations, and neither can explain why column orientation and blank characters reduce the word-superiority effect.