Four priming experiments investigating the functional use of onsets and rimes as identification units in normally and poorly reading children, matched on reading age level, are reported. Experiments 1 (onsets) and 2 (rimes) used monosyllabic words. High- and low-frequency bigram letter clusters were primed. Primes turned out to be more effective when they coincided with the rimes of target words then when they did not. For onsets this was not the case. The effect of priming was stronger in low-frequency letter clusters. For rime units there was a significant prime by rime coincidence interaction, consistent with data presented by Bowey (1990). A differential effect of rime priming was obtained for the ability groups in high-frequency letter clusters. Only in the normal reader control group were response latencies negatively affected by noncoinciding primes in high-frequency rime units. This finding suggests that these subjects probably had better access to letter information in the final part of words. In Experiments 3 (onsets) and 4 (rimes) bisyllabic compound nouns were used. Elements of the second syllable were used as primes. Main group effects were found in both experiments, but the effects obtained with monosyllabic words were not replicated. It is concluded that onset/rime mechanisms primarily operate within the boundaries of monosyllabic words and/or in stressed syllables.