This study compared accuracy of word identification in oral reading of materials in which polysyllabic words were spatially divided with performance on undivided materials of comparable difficulty. Retarded readers in junior high schools were tested with two forms of the Gilmore Oral Reading Test. The experimental form presented polysyllables divided into groupings containing positional and marker grapheme clues to pronunciation. The control version was left unsegmented. Subjects performed significantly better with experimental forms than with control versions. However, treatment effect was differentiated according to reading grade level. The 4.0-4.9 reading grade group showed a non-significant difference with segmented print, while those in the 5.0-5.9 and 6.0-6.9 groups registered experimental differences equivalent to six and nine months respectively. These results suggest that segmented print may well serve as an interim treatment for identifiable groups of retarded readers. (Author/LL)