Three experiments are described in which a partial identity priming procedure was used to investigate the hypothesis that orthographic onsets and rimes serve as units of visual word recognition. In Experiment 1, partial identity priming using word-final trigrams was observed only when the bigram corresponded to the orthographic rime unit. Nonrime primes were again ineffective primes. In Experiment 3, partial identity priming using word-initial bigrams was observed only when the bigram corresponded to the orthographic onset unit. Non-onset bigrams were ineffective primes. These differential priming outcomes cannot be explained by graphemic priming, prime frequency, or practice effects. They are consistent with the hypothesis that syllable onset and rime units serve as functional units of reading.