The rate of word integration and the overprinting paradigm

In the first experiment, the letters of words were overprinted; that is, successive letters were presented on the same location of the screen. Word identification was difficult unless the rate of presentation was slow, about 3 letters/sec. The second experiment showed that the low level of word identification did not reflect difficulty with basic letter identification. Instead, it reflected the rate at which short-term memory can accept separate items. While such results refute models of reading that stress a letter-by-letter integration, they do not test those that stress construction of higher order units. Instead, the task forces letter-byletter integration by removing spatial information, a dimension required for construction of higher order units. As a result, it does not permit integration of the correct units. The third experiment illustrated the construction and integration of supraletter units. Letter groups (either syllables or corresponding nonsyllable groups! were overprinted, a technique that introduces the spatial dimension. Word identification increased dramatically, and subjects were able to exploit the familiarity inherent in syllabic presentations. Thus, integration is much faster when more appropriate units can be used, and construction of the units depends on spatial information. The fourth experiment showed that the construction of supraletter units is achieved prior to short-term memory: Even though memory was able to group the material, the subjects could not use syllabic structure when syllables were isolated temporally but not spatially. The results were discussed in terms of a model describing the construction of higher order units.